The CUNY 2019 program will run from Friday 8:40AM until Sunday 4:50pm (March 29-31).  Please plan your travel accordingly.


The PDF version of the program is available here. Note: this document is 707 pages long and should probably not be printed

Note: schedule subject to change

Thursday March 28, 5-7 pm, Registration/Check-In, Embassy Suites Lobby 


Friday March 29

7:30 - 11:00 AM

Registration/Check-in in the Lobby outside of The Mall ballroom

8:40 - 9:00 AM Introductory Session 
9:00 - 10:35 AM

10:35 - 10:50 AM

Coffee Break

10:50 - 12:30 PM

Talk Session 2

12:30 - 1:50 PM

Lunch Break

1:50 - 2:40 PM

Talk Session 3

2:40 - 2:55 PM

Coffee Break

2:55 - 4:30 PM

Talk Session 4

4:30 - 6:30 PM

Poster Session A & Reception          


Saturday March 30

8:15 - 10:00 AM

Poster Session B + Coffee                                                        

10:00 - 12:00 PM

Talk Session 5

12:00 - 1:20 PM

Lunch Break

1:20 - 2:55 PM

Talk Session 6

2:55 - 4:40 PM

Poster Session C + Coffee & Snacks

4:40 - 6:20 PM

Talk Session 7

6:30 - 8:30 PM

Dairy Center Party


Sunday March 31

9:00 - 10:35 AM

Talk Session 8

10:35 - 10:50 AM

Coffee Break                                                                         

10:50 AM - 12:05 PM

Talk Session 9

12:05 - 2:00 PM

Poster Session D +

Boxed Lunches served

2:00 - 2:10 PM


2:10 - 3:20 PM

Talk Session 10

3:20 - 3:35 PM

Coffee Break

3:35 - 4:50 PM

Talk Session 11

5pm  End of Conference meetup at Shine, across the street


Detailed Schedule

Friday (Day 1)

8:40-9:00 AM Introductory Remarks

Talk Session 1, 9:00 - 10:35 AM, Friday (Day 1)

Chair:  Edith Kaan

9:00 - 9:45am: The neurobiology of reading in deaf and hearing adults Karen Emmorey (invited speaker)

9:45 - 10:10am: Do late first-language learners of ASL rely on word order to comprehend simple sentencesQi Cheng and Rachel Mayberry

The present study used a sentence-picture matching experiment with 4 types of target sentence conditions, contrasting in event probability and noun animacy, to explicitly investigate how deaf individuals with early language deprivation interpret simple, subject-verb-object (SVO) sentences in American Sign Language (ASL). The results suggest that late L1 learners do not fully rely on basic word order when comprehending simple ASL sentences in real time, but mostly adopt the probable event strategy, similar to three-year-olds as reported in literature. These findings suggest that even the simplest form of sentence processing is being affected by early language deprivation.

10:10 - 10:35am: Revisiting a critical period for syntax in a second language: New analyses provide the same results Tianhu Chen and Joshua K. Hartshorne

Hartshorne, Tenenbaum, and Pinker (2018) analyzed data from 669,498 subjects to provide the first-ever estimate of how the ability to learn syntax changes with age, finding a sharp, discontinuous drop in learning rate at 17.4 years. To address concerns that this finding was an artifact of how the data were analyzed, we re-analyzed these plus additional subjects (total N=1,136,105) using a new, more flexible analytic model and scoring ability using Item Response Theory. These innovations had a negligible impact on the results. We discuss the theoretical and practical implications of these findings for the study of human sentence processing.
10:35-10:50 AM Coffee Break

Talk Session 2, 10:50 AM - 12:30 PM, Friday (Day 1)

Chair:  Elsi Kaiser

10:50am - 11:15am: Structure before content: long-distance dependency in sentence production Shota Momma, Masaya Yoshida and Victor Ferreira

In long-distance dependency processing, it is widely known that comprehenders predict the gap before encountering gap-hosting verbs, without knowing the words intervening between the filler and the gap (Stowe, 1986; Omaki et al., 2016, a.o). Here we show that, in sentence production, speakers plan the syntactic structure of the verb phrase containing the gap before producing the filler and before planning the words between the filler and the gap. We argue that speakers establish the syntactic dependency between the filler and the gap before planning the intervening words, just like comprehenders do.

11:15 - 11:40am: Number attraction effects in production: errors and speech rate profiles narrow a production-comprehension contrast Margaret Kandel, Cassidy Wyatt, Lalitha Balachandran and Colin Phillips

In four production experiments, we compared number attraction effects on subject-verb and reflexive-antecedent agreement in both a traditional preamble-completion elicitation paradigm and a novel full-sentence elicitation paradigm. Previous studies using the preamble paradigm show attraction for both verbs and anaphora, unlike in comprehension, where only verbs show attraction. While our results from the preamble paradigm align with previous findings, the more naturalistic paradigm shows the same contrast seen in comprehension, with robust verb but minimal anaphor attraction. Using a forced-aligner, we looked for attraction effects in the production time-course of correct sentences; timing delays showed similar distributions to attraction errors.

11:40 - 12:05pm: Remember 'him', forget 'her': Gender bias in the comprehension of pronominal referentsVeronica Boyce, Titus von der Malsburg, Till Poppels and Roger Levy

What gender information is conveyed through pronouns such as 'he', 'she', and 'they'? We investigate using a diverse set of role nouns and both comprehension and production experiments. Comprehenders treat 'he' pronouns as a stronger signal of the referent being male than 'she' pronouns are of the referent being female, but there is no corresponding asymmetry in pronoun production preferences. From this mismatch, we conclude that comprehension and production are miscalibrated with regard to conveying referent gender, in a way that could lead listeners to infer more maleness of referents than the speaker intended to convey.

12:05 - 12:30pm: Sentence-planning strategies in adults who stutter: An eye-tracking studyKerianna Frederick and Yi Ting Huang

The hallmarks of stuttering are highly visible, but it remains unclear what production processes give rise to these patterns. Past studies investigate words and sentences separately, leaving open how these processes interact. Using an eye-tracking while speaking task, the current study manipulated the ease of lexical retrieval and word position in sentences. When initial pictures were easy, typically fluent adults began planning the next element. In contrast, for adults who stutter, fixations during initial pictures were unaffected by lexical properties or word position. These findings suggest that adults who stutter may adopt less flexible strategies for planning words in sentences.
12:30-1:50 PM Lunch Break

Talk Session 3, 1:50 - 2:40 PM, Friday (Day 1)

Chair:  Adele Goldberg

1:50 - 2:15pm: Testing processing explanations of word order universalsMichael Hahn and Richard Futrell

A prominent class of language universals are the Greenbergian harmonic word order correlations (e.g., whether a language puts verbs before or after objects strongly determines whether it has prepositions or postpositions). A prominent explanation is in terms of maximizing processing efficiency. We computationally evaluate such processing-based explanations. We create counterfactual word order grammars that, when applied to syntactic corpora of 51 languages, optimize three metrics of processing efficiency (dependency locality, surprisal, parsability). We find that most of the Greenbergian correlations emerge when optimizing word order for these metrics, supporting processing-based explanations of word order universals.

2:15 - 2:40pm: Crosslinguistic word orders enable an efficient tradeoff of memory and surprisalMichael Hahn, Judith Degen and Richard Futrell

A prominent class of language universals are the Greenbergian harmonic word order correlations (e.g., whether a language puts verbs before or after objects strongly determines whether it has prepositions or postpositions). A prominent explanation is in terms of maximizing processing efficiency. We computationally evaluate such processing-based explanations. We create counterfactual word order grammars that, when applied to syntactic corpora of 51 languages, optimize three metrics of processing efficiency (dependency locality, surprisal, parsability). We find that most of the Greenbergian correlations emerge when optimizing word order for these metrics, supporting processing-based explanations of word order universals.
2:40-2:55 PM  Coffee Break

Talk Session 4, 2:55 - 4:30 PM, Friday (Day 1)

Chair:  Peter Gordon

2:55 - 3:40pm: Language Processing is Both Incremental and Segmental -- and the Balance May Shift with Aging - Elizabeth Stine-Morrow (Invited Speaker)

3:40 - 4:05pm: Reference and perspective taking across the adult lifespanRaheleh Saryazdi and Craig Chambers

The present study explored whether younger and older adults differ in using common ground information during online referential interpretation. Participants followed instructions from a director to point to objects in a visual display. The target object was accompanied by either a competitor or a control object. We also manipulated whether competitor/control objects were presented in privileged or common ground. Older adults were less likely than younger adults to use common ground information online. However, older adults' comparable performance in the control condition suggest the observed pattern arises from differences in cognitive control or metalizing abilities rather than slower information processing.

4:05 - 4:30pm: When is declarative memory necessary for audience design? Evidence from amnesia Si On Yoon, Melissa Duff and Sarah Brown-Schmidt

In conversation, speakers tailor language based on the partner’s knowledge, a process termed audience design. One view suggests access to episodic memory of shared experience is critical to audience design. We ask if audience design emerges in individuals with amnesia, who have declarative memory impairment, in situations where the patient shares some knowledge with one partner, and different knowledge with the other partner. Individuals with amnesia adjust expressions depending on the current addressee’s knowledge, like healthy adults; Participants produced longer expressions for partner-novel vs. partner-familiar images. This finding suggests partner-specific audience design does not require access to the episodic record.
4:30 - 6:30 PM  Poster Session A & Reception


Saturday (Day 2)

8:15 - 10:00 AM Poster Session B & Coffee

Talk Session 5, 10:00 AM - 12:00 PM, Saturday (Day 2)

Chair:  Adrian  Staub
10:00 - 10:45am: Atypical Grammatical Profiles in Children Can Arise from an Interaction between Weak Language Aptitude and Typical Language Input Laurence Leonard (invited speaker)

10:45 - 11:10am: Certain ungrammaticality or uncertain grammaticality: Deciding between frequent errors and infrequent grammatical structuresMaayan Keshev and Aya Meltzer-Asscher

Sentence processing studies typically assume error-free input, but production/perception errors are common in language use. Recent models suggest that comprehenders engage in rational noisy-channel processing, pulling the interpretation towards more probable “near-neighbor” analyses. We investigate processing under uncertainty using subject/object relative clause ambiguity in Hebrew. Four self-paced reading experiments and a sentence completion experiment show that readers apply elaborate knowledge regarding the distribution of alternative structures in the language, and that they are willing to compromise subject-verb agreement to refrain from (grammatical but) highly improbable structures. This shows that the prior probability of alternative analyses modulates the interpretation of agreement.

11:10 - 11:35am: Online parsing strategies are influenced by verb-specific and language-general biases -  Zoe Ovans, Kathleen Oppenheimer, and Yi Ting Huang

During comprehension, listeners parse ambiguous sentences like “Jab/Choose the elephant with the carrot” based on knowledge of verbs’ syntactic distributions. For example, “jab” often occurs in VP-attachment contexts (carrot as instrument) and “choose” often occurs in NP-attachment contexts (carrot modifies the elephant). What happens when lexical biases are uncertain (e.g. for infrequent verbs)? We tested whether listeners avoid making parsing commitments or rely on verb-general tendencies. Using a visual-world eye-tracking task, we found that for infrequent verbs, listeners relied on the verb-general VP-attachment bias in English, suggesting that listeners track language-general syntactic distributions and use this information during parsing.

11:35 - 12:00pm: Selective effects of cognitive control on the P600: Evidence from a large-scale individual differences studyTrevor Brothers, Eddie Wlotko, Lena Warnke, Edward Alexander, Lin Wang, Maria Luiza and Gina Kuperberg

The processing of semantic anomalies during comprehension has long been linked to the elicitation of the P600 component - a neural signature that has been theorized to reflect the detection of conflict between the comprehender’s prior mental model and incoming linguistic input. In the present study, we employed a large-scale individual differences approach to examine the role of Language Experience, Working memory, and Cognitive control in the detection and resolution of semantic incongruities during discourse processing (N =77). Both behavioral detection accuracy and the amplitude of the P600 were selectively associated with individual differences in cognitive control.
12:00 - 1:20 PM Lunch Break

Talk Session 6, 1:20 - 2:55 PM, Saturday (Day 2)

Chair: Matt Lowder
1:20 - 2:05pm: The dynamics of recognizing words in people with language and hearing impairments: Revealing mechanisms of disordered language; pushing the boundaries of basic theoryBob McMurray (invited speaker)

2:05 - 2:30pm: That was a question?: Accommodating variability in intonation interpretationsAndrés Buxó-Lugo and Chigusa Kurumada

In two experiments, we investigate whether listeners can rapidly change their interpretations of question/statement prosody in response to new input patterns. Participants went through a pre-exposure phase, an exposure phase, and a post-exposure phase. Through each phase, participants listen to sentences and answer whether they heard a question or a statement. During exposure, participants were assigned to either a question-biasing condition, a statement-biasing condition, or a control condition, and they got feedback about their responses. Listeners quickly accommodated their interpretations of intonational contours between pre- and post-exposure, in accordance to the condition they were assigned to.

2:30 - 2:55pm: Verbatim memory for surface features: Evidence from stress shift - Jennifer Bellik and Tom Roberts

Recall of “surface” features of linguistic expressions is significantly worse than recall of “gist” (i.e., broad semantic content). The Regeneration Hypothesis (RH) suggests this is because surface features are not encoded verbatim in memory, but rather "reconstructed" from gist. However, the RH is supported primarily by evidence from recognition memory tasks on alternations with subtle meaning correlates (e.g. active/passive), and are thus not purely 'surface' manipulations. We present an auditory recognition memory study manipulating application of the Rhythm Rule, a surface-level phonological alternation. Results suggest listeners are sensitive to this change, indicating they do store verbatim surface features in memory.
2:55 - 4:40 PM  Poster Session C & Coffee

Talk Session 7, 4:40 - 6:20 PM, Saturday (Day 2)

Chair: Jared Novick

4:40 - 5:05pm: Lexical access in comprehension vs. production: Spatiotemporal localization of semantic facilitation and interference Julien Dirani and Liina Pylkkanen

Humans understand words faster when they are preceded by semantically related primes. Interestingly, the effect is reversed when naming objects. The locus of this interference effect is highly debated and could occur at the lexical level, or at post-lexical stages of processing. The present work took advantage of the high temporal resolution of magnetoencephalography to address the spatio-temporal localization of the semantic interference effect in production and the facilitation effect in comprehension. Results show an early facilitation effect in comprehension in the left middle STG and, a late effect of interference in the insular cortex pointing to a post-lexical locus.

5:05 - 5:30pm: Lexical Selection by Competition in Word Production: Evidence from New ParadigmsMark Koranda and Maryellen MacDonald

Whether lexical selection in production is directly affected by alternative candidate words (competition) is debated. Yet the critical data, changes in latency of the most common (dominant) name for a picture, are interpreted differently by competing accounts. We made some extensions that go beyond studying dominant predictors of dominant naming latency: we established secondary name accessibility, name and image agreement, and also tested likelihood of dominant naming. Overall fit between models with secondary versus dominant predictors were compared. Secondary predictors were the better fit for dominant naming likelihood, but not latency, suggesting selection may be competitive, independent of latency factors.

5:30 - 5:55pm: Ambiguity avoidance independent of non-linguistic similarityKumiko Fukumura, Sandra Villata, Francesca Foppolo, Celine Pozniak, Maria Nella Carminati and F.-Xavier Alario

Research suggests that speakers use fewer pronouns when the referential candidates have the same (rather than different) biological gender and are more similar, even in a non-gendered language. Here we examined how gender congruence might affect ambiguity avoidance by investigating the use of French gendered pronouns and Italian null pronouns. Gender congruence between the referents resulted in fewer French pronouns, regardless of whether pronouns agree with biological or grammatical gender. Neither gender congruence affected Italian null pronouns. Hence, speakers avoid gender ambiguous pronouns independent of non-linguistic similarity, indicating that ambiguity avoidance and similarity-based interference affect pronoun use via different constraints.

5:55 - 6:20pm: Effects of semantic similarity on referring expression production in Mandarin Chinese -Yangzi Zhou, Holly Branigan and Martin Pickering

Our study investigated how semantic similarity between entities can cause interference (similarity-based interference effect (SIE)) in speakers’ referential choices in Mandarin Chinese and its cause. Using a picture description paradigm, we found that Mandarin speakers tend to be more explicit (rather than less explicit) when the discourse includes another semantically similar entity. More importantly, concerning the cause of SIE, we found that besides semantic competition between two semantically similar entities, SIE is dependent on speakers’ communicative effort to avoid using referentially ambiguous expressions. This suggests the influence of both speaker-internal and addressee-based production constraints on speakers’ referential choices.

Sunday (Day 3)

Talk Session 8, 9:00 - 10:35 AM, Sunday (Day 3)

Chair: Jennifer Arnold
9:00 - 9:45am: Is language spared in Williams syndrome? Developmental timing as a lens - Barbara Landau (invited speaker)

9:45 - 10:10am: Discourse Effects on the Source-Goal Asymmetry Monica Do, Anna Papafragou and John Trueswell

A bias for goals (…to a lamppost) over sources (…from a chair) exists in language and memory for motion events. In a see-and-describe task, we show that the goal-bias in language is partially attributable to the status of sources as discourse-old while goals are discourse-new, and thus, relevant to communicate to an addressee. Using the change detection paradigm, we show that mentioning the sources improves speakers’ memory for sources, but the goal-bias persists. We conclude that discourse factors play a role in event cognition and can influence what speakers choose to include in their messages.

10:10 - 10:35am: Prediction supports infants' language processing and language developmentTracy Reuter, Casey Lew-Williams and Lauren Emberson

Recent theories claim that prediction – the pre-activation of representations during language processing – supports language development. However, prediction’s role in development is uncertain. Here, we present 3 experiments that contribute to this ongoing debate. Findings suggest that prediction and comprehension emerge concurrently in infancy (Exp.1), that infants’ prediction abilities correlate positively with their vocabulary size (Exp.1), and that prediction and comprehension are distinct processing mechanisms (Exp.2). Furthermore, in an ongoing longitudinal study (Exp.3), we are evaluating whether infants’ prediction abilities forecast their developmental outcomes. In sum, these novel findings suggest that prediction supports both language processing and language development.
10:35 - 10:50 AM  Coffee Break

Talk Session 9, 10:50 AM - 12:05 PM, Sunday (Day 3)

Chair: Les Sikos

10:50 - 11:15am: A probabilistic account of VPE interpretation in contextJeffrey Geiger and Ming Xiang

Identity- and discourse-driven accounts of VPE interpretation have both been proposed, but their proponents have tended to gloss over the possible role of information from the broader context or the linguistic antecedent, respectively. In two experiments, we show that neither account sufficiently captures the facts of VPE interpretation in context, with the distribution of interpretations intermediate between what is predicted by an identity and a discourse mechanism. We formalize these findings by comparing three probabilistic models of VPE interpretation - under identity, discourse availability, or a combined strategy - with the combined-strategy model most likely to generate our experimental data.

11:15 - 11:40am: The mental representation of universal quantifiers: evidence from verificationTyler Knowlton, Paul Pietroski, Justin Halberda and Jeffrey Lidz

The meaning of sentences like "every circle is blue" could be represented in speakers’ minds in terms of individuals and their properties (e.g., for each thing that’s a circle, it’s blue) or in terms of relations between groups (e.g., the blue things include the circles). We offer evidence that this formal distinction is psychologically realized and has detectible symptoms. Participants were found to have better memory for cardinality (a fundamentally group-property) following most-statements compared to existential-statements and following every-/all-statements compared to each-statements. This supports the idea that meanings are specified in the mind at a finer grain-size than truth conditions.

11:40am - 12:05pm: Delayed attachment commitments for parenthetical relative clauses: An eye-tracking study Marju Kaps, Alexandra Lawn and Jesse Harris

We investigate how not-at-issue content is integrated into a sentence by comparing the online attachment of parenthetical relative clauses to that of better-studied restrictive relative clauses. In an eye-tracking experiment, we report an early penalty for high attachment in restrictive relative clauses (as predicted by processing principles such as late closure or recency) but not for parenthetical relative clauses. These findings suggest that the processor delays structural integration of not-at-issue material, even when attachment is disambiguated by verb agreement. A preliminary or underspecified parse may initially be computed for parenthetical clauses during comprehension
12:05 - 2:00 PM  Poster Session D + Boxed lunches served
2:00 - 2:10 PM  Announcements

Talk Session 10, 2:10 - 3:20 PM, Sunday (Day 3)

Chair: Lee Osterhout

2:10 - 2:55pm: Contributions of experience to the mind and brain: insights from studies of language in blindness - Marina Bedny (Invited Speaker)

2:55 - 3:20pm: Semantic projection: Recovering human knowledge of multiple, distinct object properties from natural word collocationsGabriel Grand, Idan Blank, Francisco Pereira and Evelina Fedorenko

We demonstrate that rich, flexible world knowledge could be built bottom-up into the lexicon simply by tracking word co-occurrences in natural corpora. Specifically, we introduce a domain-general method for extracting context-dependent similarities among concrete nouns from Distributed Semantic Models: (i) constructing a “scale” representing some property (e.g., size / ferociousness) by linearly connecting the word-vectors of antonyms (e.g., “small”→“large” / “safe”→”dangerous”); and (ii) projecting word-vectors of nouns from some category (e.g., animals) onto this scale. Across 35 category-property pairs, such “Semantic Projection” recovers human similarity ratings such that, e.g., “dolphin” and “tiger” are similar in size but differ in ferociousness.
3:20 - 3:35 PM Coffee Break

Talk Session 11, 3:35 - 4:50 PM, Sunday (Day 3)

Chair: Wing-Yee Chow

3:35 - 4:00pm: Harry Potter and the language-knowledge interface: ERPs reveal retrieval of domain and specific knowledge during readingMelissa Troyer and Marta Kutas

We recorded ERPs while young adults read sentences about Harry Potter (HP)/general topics. Across three studies, individuals’ HP knowledge was moderately-to-strongly correlated with N400 amplitudes to contextually-supported endings of HP sentences, but not to unsupported endings nor with N400 effects (supported — unsupported endings) for general-topic sentences. Single-trial analyses confirmed this was not solely due to the proportion of items an individual knew. HP knowledge also modulated N400 amplitudes to “incorrect” endings categorically- or event-related to HP sentences. We propose domain knowledge influences the quick availability of relevant/related information used to make sense of sentential words in real time.

4:00 - 4:25pm: Capturing continuous effects of context during naturalistic story readingShannon McKnight and Albert Kim

4:25 - 4:50pm: Predictive pre-updating: Converging evidence from electrophysiology and eye-blink rateTal Ness and Aya Meltzer-Asscher

"Prediction during sentence processing was suggested to involve two qualitatively distinct mechanisms: “pre-activation” of representations in long-term memory, and “pre-updating” of working memory (WM) representation to include the predicted content. Pre-updating was recently demonstrated as an increased P600 amplitude prior to a highly predictable word, with a greater effect for participants with higher WM capacity. We present converging evidence for pre-updating and the influence of WM capacity, from two additional experiments, using event-related potentials (ERP) and event-based eye-blink rate (EBR) experiment. We also show that pre-updating occurs in highly constraining contexts regardless of the specific word predicted by the comprehender."

Poster Sessions


Poster Session A Friday (Day 1) 4:30 - 6:30 PM


  • East & West End Rooms: posters #A1-A41
  • The Ballroom: posters #A42-A64
  • Pre-Function Area: posters #A65-A80

When does repair occur?

How often do we detect incoherence in conversation? When we do, do we correct it or ask for clarification? Recent literature suggests that the answer to both of these questions is "far less often than we would think." The present pilot study investigates the hypothesis that this observation reflects a relationship between conversational repair and tangible consequences. When presented with instructions for a simple task that contained one incoherent instruction, participants were slightly more likely to initiate repair when the incoherent instruction was essential to the successful completion of the task. Potential sociolinguistic motivations for this phenomenon are also explored.
Nathaniel Paley, Eliyahu Spivack, and Bruno Galantucci

When is irony influenced by communicative constraints? ERP evidence supporting interactive models

The present ERP study tested the impact of contextual biases and speaker’s features on the time course of irony analysis in order to tease apart modular and interactive models' predictions
Sendy Caffarra, Arman M. Haeri, Elissa Michell and Clara D. Martin

The semantics-pragmatics of typographic emphasis in discourse

Typography is typically deemed a low-level visual feature of the text that does not interact with language processing. Although a few emphasis studies provide off-line evidence that type adds a contrastive layer of meaning to the interpretation of referential expressions, it is unclear whether this extends to real-time processing and whether type can bring a referent into discourse focus. This study provides on-line evidence for the visual-emphatic, contrastive, and discourse focus effects of typographic emphasis in English by means of two eye-tracking experiments manipulating capitals or italics. We argue that typographic emphasis serves semantic-pragmatic functions in sentence and discourse processing.
Jefferson Maia and Robin Morris

Distributional regularity of suffixes facilitates acquisition of gender: Eye-tracking evidence from two closely related languages

We examine two hypotheses about the role of transparency of noun suffixes in gender acquisition: 1) form-to-meaning mapping, and 2) distributional regularity of suffixes. We contrasted Russian and Bulgarian, closely related languages with similar gender systems but different distribution of opaque noun suffixes across genders. Two groups of 4-to-6-year-old children participated in production and eye-tracking experiments. We found that suffix transparency facilitated gender acquisition in both languages, but opaque FEM nouns were disproportionally affected in Bulgarian. This is evidence in favor of the regular distribution of opaque suffixes across genders, as predicted by the Distributional Regularity hypothesis.
Tanya Ivanova-Sullivan, Irina Sekerina, Glenn Stark and Lidiya Tornyova

The Role of Tagalog Voice Information on Sentential Argument Order in Production and Gaze Patterns in Comprehension

Is voice morphology in Tagalog, a verb-initial, flexible word order language, used to order and anticipate upcoming nominal elements? Tagalog verbs contain a voice affix that identifies a syntactically prominent argument and cross-references its thematic role. In a sentence continuation task, native adults showed strong effects of voice morphology on word order. However, voice morphology did not influence looks to upcoming arguments in a visual world comprehension experiment. Nor were looks to the agent preferred; instead, there was a general preference for looks to animate entities. These findings suggest non-syntactic information dominated syntactic information in Tagalog online processing.
Ivan Paul Bondoc and Amy Schafer

A large-scale deconvolutional study of predictability and frequency effects in naturalistic reading

Using three large naturalistic reading datasets, this study attempts to replicate previous findings from constructed experiments of additive effects of word frequency and predictability on measures of human comprehension effort. While results show strong effects of word predictability, they show no generalized effect of word frequency over and above predictability. This result is inconsistent with influential theories of sentence processing that posit distinct context-independent effects of lexical frequency and suggests that frequency effects are more attenuated in everyday reading would be suggested by constructed experiments, possibly as a result of task artifacts.
Cory Shain

Prosodic effects on attachment in Brazilian Portuguese

This research explores how prosody influences the attachment of final adverbial phrases in ambiguous Brazilian Portuguese (BP) sentences (Marcela ouviu que João tinha ligado # na segunda-feira: ‘Mary heard that John had called on Monday’). In English, Clifton et al. (2002) found that a prosodic boundary before the adverbial increased high attachments to the first verb (heard), while Carlson & Tyler (2018) showed that contrastive accents on the first or second verb drew attachment to the accented verb. We investigated the effects of both prosodic cues on corresponding sentences in BP, finding that prosodic boundaries affected interpretation but not accents.
Aline Fonseca, Katy Carlson and Andressa Silva

The dynamic generalized linear mixed effect model: Modeling intensive binary time-series data from the visual-world eye-tracking paradigm with GLMM with crossed random effects

The aim of this research is to present a novel model specification that takes into account change processes (autocorrelation [AR] and trend) in binary time series eye-tracking data, and variability across trials, persons, and items. By ‘dynamic’ we mean that the model considers change processes in the generalized linear mixed effect model (GLMM). The “dynamic GLMM” is a GLMM with crossed random effects (random person and random item effects). This is an extension of the model specification presented in Cho et al. (2018) in that a trend effect (the change in the mean level per unit time) is now considered.
Sarah Brown-Schmidt and Sun Joo Cho

24-month-olds (and adults) exploit negative sentences to constrain their interpretation of novel word meanings

This study shows that 24-month-olds can use their understanding of both affirmative and negative sentences to constrain their interpretation of novel word meanings. Children who witnessed a person label a monster using an affirmative sentence “It’s a bamoule!” associated the novel word with the monster. However, children who witnessed the person label the monster using a negative sentence “It’s not a bamoule!” did not make the same association. This ability to understand negative sentences so early might support language acquisition, providing infants with a tool to constrain the space of possible referents for word meanings.
Alex de Carvalho, Cécile Crimon, Anne Christophe and John Trueswell

Implicit causality: A comparison between English and Vietnamese verbs

We conducted a large-scale study (n=98) of 149 verbs in Vietnamese using ‘because’ sequences to investigate the crosslinguistic generalizability of implicit causality (IC) effects in relation to verb class (Agent-Patient, Agent-Evocator, Stimulus-Experiencer, Experiencer-Stimulus), specifically Vietnamese vs. English (Ferstl et al.'11) IC verbs’ behaviors. We found a correlation in the overall subject bias between Vietnamese and English ICs (p<0.001). The subject biases in three verb classes are correlated (p’s<0.05), except in Stim-Exp verbs. Our results suggest that semantic verb class relates to IC bias though there are crosslinguistic variations, highlighting the importance of language-specific IC norms—rather than assuming crosslinguistic generalizability.
Binh Ngo and Elsi Kaiser

Syntactic adaptation following short-term experience: neural correlates and relationship to cognitive control

Processing of infrequent garden-path sentences improves following exposure. We hypothesized that if this syntactic adaptation is driven by detection of a mismatch between the predicted and actual structure, then cognitive control mechanisms for conflict resolution may be involved. If so, (1) adaptation could be sub-served by frontal regions linked to both comprehension and cognitive control, and (2) the extent of this adaptation might correlate with individual differences in cognitive control (as indexed by Stroop). Activation within the left pars opercularis showed these markers, demonstrating that comprehension may adapt to short term experience using cognitive control subserved by the frontal cortex.
Kelly Sharer and Malathi Thothathiri

Resting State Power Predicts Cognitive Control and Language Abilities

Both cognitive and linguistic task performance have been related to stable brain functioning, namely resting state power in certain frequency bands as measured by EEG. The present study assessed individual differences in pivotal linguistic skills, language learning abilities, and cognitive skills. Lower resting state beta power was related to an increase in verbal fluency and vocabulary size. For working memory, executive control, and processing speed better performance was related to higher power (in different frequency ranges). The opposite relationship between power and skill for language and cognition suggests these are supported by different oscillatory networks.
Suzanne Jongman

Factive and manner-of-speaking islands are an artifact of nonlinearity in the acceptability judgment task

‘Island’ phenomena have been a central topic in theoretical syntax for many years. A prominent example of islands is that extraction out of a sentential complement introduced by factive and manner-of-speaking verbs is less acceptable than extraction from a clause intruded by “bridge” verbs. We aimed to replicate Ambridge and Goldberg (2008) who argued that extraction from a sentential complement is unacceptable in proportion to its discourse salience. We failed to replicate their results and found that there is no true island effect for such structures: instead there are separate penalties based on verb-frame frequency and the presence of extraction.
Yingtong Liu, Rachel Ryskin, Richard Futrell and Edward Gibson

Referential efficiency across adulthood: communicative strategies and cognitive control

Recent work characterizes efficient speakers as those whose referential choices facilitate listeners’ visual search. In this way, redundancy can be a means of achieving efficient communication if it’s helpful to listeners. This requires the ability to track contextual information and change strategy, likely relying on attention and WM. Adults aged 19-82 participated in picture-naming and story continuation tasks where visual scenes varied: polychrome vs monochrome displays –where color modification is efficient in polychrome only– and scenes with 1-character vs 2-characters-of-different-gender –where pronouns are more efficient/appropriate in 1-character scenes. Results revealed participants with better cognitive skills were redundant in efficient ways.
Madeleine R. Long, Hannah Rohde, Michelle Oraa Ali and Paula Rubio-Fernandez

Referential predictability and topicality diverge in implicit causality

The relationship between predictability and topicality is highly debated in current theories on pronoun comprehension and pronoun production. Some studies suggest that discourse topics are predictable, even in part defined by being predictable. Other research assumes that they are distinct and have separate influences on pronoun interpretation and production. For stories involving implicit causality, we find that meta-linguistic judgments on topicality and predictability do not pattern together, which suggests that they are distinct properties. As for individual differences, participants with higher print exposure showed a stronger subject bias in judgments on predictability, but not topicality.
Shuang Guan and Jennifer Arnold

Can garden path effects be reduced to predictability?

Marten van Schijndel and Tal Linzen

Ambiguous word recognition in sentence contexts by L2 speakers

This paper presents the results of an experiment investigating the effect of sentence context on the recognition times (RTs) of ambiguous words in second language (L2) learners with native speaker (L1) participants as a control group. Twenty-four native English speakers and 24 advanced L2 English speakers (L1: Arabic) performed a self-paced reading task on English sentences that included ambiguous words (e.g., solution). English L1 participants were able to use the preceding subordinate-biasing context to activate the subordinate meaning as much as the dominant meaning. However, L2 participants’ result indicates supremacy of the dominant meaning when processing L2 ambiguous words.
Essa Batel

Lexical prediction and the processing of argument structure in English psych verbs

In a self-paced reading study, we show that psych verbs are harder to process than non-psych verbs. However, a fronted experiencer argument facilitates the processing of the psych verb.
Russell Simonsen and Dustin Chacón

One star, two star, red star, blue star: Subsequent memory reveals both lexical and discourse influences on prosodic reduction

We tested why repeated words are reduced in prosodic prominence. Participants described a route through a series of pictures, then tried to recall the referents. Some referents appeared twice; other lexical items were used twice to describe two different referents. Prosodic reduction across mentions predicted recall, supporting the claim that reduction stems in part from cognitive activation. Specifically, duration predicted recall regardless of referent status and appears driven by lexical repetition, but intensity predicted recall only when the referent was also repeated. This supports a multiple-source view in which reduction from multiple sources, including both lexical and referential repetition.
Rachel Peters and Scott Fraundorf

Prediction or processing burden? --The online comprehension of the classifier-noun pair in Mandarin Chinese

Research on Mandarin has shown that the classifier’s meaning allows for prediction of the following noun. Our experiments manipulated and standardized the classifier-noun pair with a purpose to pinpoint the processing pattern of the entire NP. We found that apart from prediction, the bottom-up incremental incorporation is also a competing mechanism that might impose some burden through the entire classifier-noun comprehension process. We propose a generalization that even if earlier linguistic information boosts the reading efficiency by facilitating prediction, the information incorporation and reconstruction is definitely a competitive mechanism that enhances the tension between top-down prediction and bottom-up incrementality.
Yuhan Zhang and Peng Zhou

The Effect of Context on Local Syntactic Coherence Processing

We present an eye-tracking reading study investigating whether or not contextual information can guide parsing towards ungrammatical processing of local syntactic coherences. The results suggest that the former is the case. The findings will be discussed with respect to their compatibility with various accounts, such as self-organized parsing (Tabor et al., 2004), simple recurrent networks (Konieczny 2005, Konieczny et al., 2009), bottom-up/top down balancing (Gibson 2006), good enough processing (Ferreira & Patson 2007), rationality (Hale 2011), and noisy channel processing (Levy 2008).
Hanno M. Müller and Lars Konieczny

Verb phrase ellipsis avoids troughs in the ID profile: An information-theoretic account to VPE based on evidence from rating and reading time data

While verb phrase ellipsis (VPE) is well-studied theoretically and psycholinguistically, it remains unclear why speakers use ellipsis at all. We present an acceptability rating and a self-paced reading study that show that the uniform identity hypothesis (Jaeger & Levy 2007) is adequate to explain encoding preferences: By using VPE speakers avoid troughs in the information density profile. They are the more likely to use a VPE instead of the corresponding redundant full form the longer this full form would be. This is reflected in degraded ratings and faster readings times for the longer full forms.
Lisa Schäfer, Robin Lemke, Heiner Drenhaus and Ingo Reich

If you don’t have anything nice (or interesting) to say, don’t say anything at all

Studies often highlight comprehenders' use of event plausibility during sentence processing, emphasizing facilitation for plausible and predictable real-world situations. Ideally such an approach would be integrated with another facet of communication, namely comprehenders' expectations that speakers say things that are (at least somewhat) interesting and newsworthy. The current study contrasts these two dimensions—knowledge about situation plausibility (priors over eventualities) and knowledge about utterance probability (likelihoods of a speaker choosing to formulate a message about a particular eventuality). Participants are shown to endorse higher numeric values following "announce" in items like "Bill [thinks/announced] that Judith has ___ Facebook friends".
Hannah Rohde and Michael Franke

Distant relatives: Resumptive pronouns can inherit agreement features of implied antecedents

Agreement attraction effects are a window to linguistic representations during production and comprehension. We investigate how the discourse representation of the distractor modulates agreement attraction. We examine grammatical gender “attraction” between a resumptive pronoun (RP) and a noun outside of the relative clause complex, in Hebrew sentences like 'a spoon.F is a utensil.M that you can eat food with it.F'). Results of one SPR experiment and two acceptability rating tests suggest that formation of agreement is sensitive to the distractor’s discourse referent, and that this is true even when features are not semantically meaningful (i.e. grammatical gender).
Maayan Keshev and Aya Meltzer-Asscher

Processing referential expressions in German Sign Language: The effect of overt localization

In this eye-tracking study, we tested whether overt manual localization with the INDEX sign in German Sign Language increases the prominence and hence the accessibility of a discourse referent and how this interacts with a referent’s grammatical role. The data suggest a conjoined effect of the factors overt localization and grammatical function i.e., subject preference. Localization using the INDEX sign seems to increase the prominence of a referent similar to a prosodic focus marker in spoken languages and can therefore be analyzed as a manual focus marker.
Anne Wienholz, Derya Nuhbalaoglu, Markus Steinbach, Annika Herrmann and Nivedita Mani

Gender agreement processing: inflectional endings and stereotypes

"Most nouns denoting professions are grammatically M in Russian. What about female professionals? Firstly, an F noun can be formed, but many M nouns have no established F counterparts. Secondly, an originally M noun can be used with M and F agreeing forms. We examined the second option in a self-paced reading study. The main result is that F agreement is expected with all nouns denoting professions, even stereotypically male ones. With stereotypically female professions, there is virtually no reaction to the surface mismatch between the originally M noun (that also ‘looks M’ due to its inflection) and the predicate."
Natalia Slioussar and Anastasia Generalova

Co-speech movement behavior at floor exchanges and interruptions

We investigate co-speech bodily movements, specifically eyebrow and head movements, of two pairs of interacting speakers as they participate in a structured turn-taking experiment wherein participants help each other complete common nursery rhymes. We hypothesize that kinematic characteristics of these movements will vary across: conversational floor exchanges, interruptions, and non-turn-adjacent regions of speech. We find that gestural density for head and brow movements differs dependent on turn type, with the greatest co-speech gestural density occurring at felicitous floor exchanges and the lowest gestural density observed in non-turn-adjacent speech. This indicates that co-speech movements may be reflective of successful conversational interaction.
Samantha Gordon Danner, Jelena Krivokapic and Dani Byrd

Priming at a distance: phonology as a passenger in retrieval

Filler-gap dependencies are fertile ground for testing theories of working memory. Previous work demonstrates that comprehenders are sensitive to syntactic information across dependency lengths but lose sensitivity to plausibility in longer dependencies. In a self-paced reading experiment, we manipulated Filler (Control/Implausible/Phonological Overlap) and Length (Long/Short). We replicate previous plausibility results, and we find that phonological onset overlap between filler and verb eases processing for long but not short dependencies. We propose that phonological features are not maintained in working memory but may be reactivated along with syntactic/semantic features of the filler in a later retrieval event.
Austin Kraft and Dustin Chacón
A29 Variability in language input: Effects on the robustness of novel word representations Michelle K. Tulloch and Marieke van Heugten

The structural signalling effect of filled pauses during reading

In work by Bailey and Ferreira (2003), listeners who heard filled pause disfluencies (i.e., “uh uh”) at structural non-boundary locations judged the sentence stimuli as ungrammatical more often than when the disfluency was at a boundary location. They interpreted this as evidence of a structural signalling effect of filled pauses. The present work seeks to examine whether this effect extends to the reading of written stimuli. In a self-paced reading paradigm, participants’ reading times were slower in the first three regions (words) after a printed disfluency in a nonboundary location. This suggests the effect is robust across perceptual paradigms.
Ralph Rose

Anaphora resolution by Japanese learners of English: Constrained by syntax and semantics

Previous L2 processing studies have found that Japanese learners of English to not adhere to Principle A during anaphora resolution but are instead constrained by semantic features. However, these studies only investigated whether processing was constrained by Principle A and not by other syntactic factors. As such, the current study investigated the L2 English processing by Japanese learners of English to determine if both syntactic and semantic cues can cause an interference or intrusion effect during anaphora resolution using relative clauses. The results indicate that both syntactic and semantic features caused a facilitatory intrusion effect at the reflexive anaphor.
Michael Mansbridge and Katsuo Tamaoka

Testing error-based learning models of structural priming: Context and tense bias effects

"Three experiments showed no evidence that structural priming is affected by the predictability of the prime structure, contra claims of error-based learning models, which assume that priming occurs when experimental participants predict an incorrect prime structure. Experiments 1 and 2 manipulated the context preceding PO/DO ditransitive structures so they were more or less predictable, but this did not affect structural priming. Experiment 3 manipulated predictability using different aspect (progressive/non-progressive) in transitive/intransitive structures. Although aspect had a strong effect on the transivity bias, it did not affect priming. The results favour residual activation models of structural priming over error-based learning models."
Roger van Gompel

Resolution of ambiguous relative clauses of Turkish L2 speakers of Japanese

We investigate the relative clause attachment preferences of L1 and L2 speakers using a set of cross-translated sentences containing ambiguous phrases such as the servant of the actress who was on the balcony. We find native Japanese speakers have a high attachment preference (the servant was on the balcony), while native Turkish speakers have a low attachment preference (the actress was on the balcony). When Turkish L2 learners of Japanese are tested on Japanese examples, we find that their attachment preferences in Japanese reflect the processing biases from their native languages.
CHUNHUA BAI and Yuki Hirose

Not necessarily first language “attrition”: on-line reflexive processing among Mandarin-English late bilingual speakers

Using a word-by-word speeded comprehension task with two-alternative forced-choice, we investigated reflexive processing among Mandarin-English late bilinguals. We showed that during on-line reflexive processing, bilingual speakers have higher accuracy and shorter reaction time than their non-bilingual peers. In addition to the linguistic task, we conducted a battery of cognitive tasks to measure working memory capacities and cognitive control abilities such as attention maintenance, response inhibition and mental shifting. Significant correlations between cognitive tasks and the linguistic task suggest that general cognitive functions play a significant role during on-line processing, especially when the comprehension demands are high.
Wenjia Cai

ERP evidence for long-distance lexical predictions in German particle verb constructions

Using ERPs we investigated whether readers lexically predict particles in sentences containing verb-particle constructions. Sentences were compared which licensed either one or a small set of highly plausible particles. At violations of the expected particle/s, we found evidence for a larger positivity from 600-900 ms in sentences with only 1 plausible particle. We interpret the larger late positivity as more costly reanalysis of a parse which was more specified due to having predicted the particle’s identity. We conclude that long-distance lexical predictions are made when constraint is high and there is little risk involved in making the prediction.
Kate Stone, Shravan Vasishth and Titus von der Malsburg

Gestures evoke large, but not small, plural set sizes

The current study focuses on whether gestures can evoke the paucal (i.e., a few) versus non-paucal (i.e., a lot) distinction of the plural similarly to how referent size can be conveyed via gesture. 40 participants viewed videos of a speaker uttering sentences like “These cupcakes are for the party” while performing either a large or small gesture on the plural. These results suggest that gestures performed during the processing of plurals can evoke non-paucal sets, but not paucal sets. These data are consistent with work on number processing showing that small and large numbers are processed via different systems.
Madeline M. Nicol and Nikole Patson

Grammaticized resumption in sentence processing: Disrupting rather than facilitating

Many studies share the intuition that resumptive pronouns (RPs) aid in retrieving an inaccessible filler. We examine this notion using both simple and highly taxing sentences in Hebrew. Two SPR experiments suggest that RPs introduce anaphoric ambiguity and may disrupt the retrieval of the filler. Another experiment, testing complexity ratings of center-embeddings, also suggests that RP are unhelpful in recovering a constituent unavailable in WM, and may even increase the perception of complexity. Overall, our results suggest that both in simple sentences and in cases of high WM load, RPs do not support comprehension or parsing of long-distance dependencies.
Maayan Keshev, Mor Ovadia, Hila Davidovich and Aya Meltzer-Asscher

Agreement attraction effects in the comprehension of grammatical sentences

We demonstrate reliable agreement attraction effects in reading grammatical sentences (ungrammaticality illusion). Attraction effects were found across three experiments, both in prepositional phrases and in object relative clauses. These results suggest that attraction effects do cause ungrammaticality illusion in normal sentence processing and therefore can not be reduced only to repair mechanisms.
Anna Laurinavichyute and Titus von der Malsburg

Toward a Process-Memory Account of Non-native Sentence Processing

Laurent Dekydtspotter, A. Kate Miller, Mike Iverson, Yanyu Xiong, Kyle Swanson and Charlène Gilbert

Individual differences in second language learning via syntactic priming: examining proficiency, attention and motivation

We examined across- and within-group variation in syntactic priming by 1) comparing English L2 French learners’ and native speakers’ primed production of two syntactic alternations; 2) analysing the effect of proficiency, attention and motivation. In line with implicit learning accounts of syntactic priming, learners, like native speakers, showed long-term effects of syntactic priming; however, there was no stronger priming in learners than in native speakers. Our results show that learners with greater proficiency were more likely to show priming and that the effect of attention on syntactic priming may be modulated by syntactic complexity. Motivation did not affect syntactic priming.
Marion Coumel, Katherine Messenger and Ema Ushioda

Quantifiers, Restrictors, and Illusory NPI Licensing

We investigated the mechanisms driving illusory NPI (ever) licensing by universal quantifiers (every), which can license NPIs in their restrictor but not their scope, in four speeded acceptability studies. Study 1 found illusory licensing of NPIs outside of a universal quantifier’s restrictor. Illusory licensing was not found for existential quantifiers (some, Study 2), nor when the universal quantifier had an overt restrictor pre-nominally (Study 3) or post-nominally (Study 4). Against a faulty memory retrieval account, these results suggest that uncertainty in the semantic processes that identify a universal quantifier’s restrictor may temporarily license structurally unlicensed NPIs.
Luis Hildebrandt-Belmont and E. Matthew Husband

Online processing and interpretation of verb phrase ellipsis

This study investigates the relative availability of strict and sloppy readings during the interpretation of verb phrase ellipsis, and furthermore examines whether a mismatch between the gender of the subject and the retrieved pronoun in the sloppy reading requires an additional (and costly) step of processing. Results from eye tracking minimally indicate that both readings are equally accessible, although a gender mismatch effect in the sloppy conditions supports the hypothesis that changing the gender value of the retrieved material is costly.
Kathleen Hall and Masaya Yoshida

The approximate number sense bootstraps scalar implicature of comparatively modified numerals in impoverished contexts

Numerals modified by the comparative quantifiers 'more/fewer than' pose an interesting puzzle: How do interlocutors limit their expectations about potential values in the face of uncertainty? Experimental participants are shown a priming context such as 'We need more signatures on the petition. How many did we get?,' where expectations about the range and likelihood of potential values should be weak, and are asked to indicate range limits and the most likely true value. Participants' responses show two Weber fractions. We argue that it is the approximate number sense which bootstraps scalar implicature of comparatively modified numerals in impoverished contexts.
Christoph Hesse and Anton Benz

Evidence for shared conceptual structure for psychological and physical events

In an implicit categorization task, participants grouped causal physical verbs (e.g., Sally broke the vase) with frighten verbs (e.g., Max frightened Sally) and non-causal physical verbs (e.g., The vase broke) with fear verbs (e.g., Sally feared Max). This suggests a common representation of causation underlying frighten verbs and causal physical events, and supports the analysis that frighten and fear verbs have different meanings that account for their different syntactic realizations. These results provide strong support for the existence of systematic mappings from semantics to syntax. We discuss further implications for linguistic representation and language acquisition.
Jayden Ziegler, Annie Chai and Jesse Snedeker

When circumstances change, update your pronouns.

"Language is frequently ambiguous, with the same sentence having several possible interpretations. One particularly prevalent example is third-person pronouns. We model pronoun interpretation as an inference over a generative model of the speaker. A key advantage of the generative intuitive theory approach is that it incorporates a definition of world knowledge which is highly flexible rather than just facts and heuristics. We test our model by manipulating the way the world works and comparing model predictions to judgments from participants recruited via AMT, and we show that pronoun interpretation is indeed heavily influenced by dynamically updated beliefs about the world."
Joshua Hartshorne, Mariela Jennings, Tobias Gerstenberg and Joshua Tenenbaum

Input Modality Effect on L2 Lexical Recall and Recognition

The present study examined the modality effect on explicit memory of L2 English words by employing recall and recognition paradigms. Fifty Korean-speaking learners of English participated in two retention experiments, one with high frequency words and the other with relatively low frequency words. The results show that the production benefit is enduring and generalized to the retention of linguistic information in an L2. The results of the two retention experiments showed that the production effect stands out itself and also bears a memory decrement for L2 input of silent mode in L2 learners.
Jeonghwa Shin

Reevaluating Pragmatic Reasoning in Online Language Games

Results from recent work testing formalizations of Gricean maxims using one-shot web-based reference-games is mixed. Some studies indicate Bayesian (RSA) models closely predict human (pragmatic) behavior; others suggest participants rarely go beyond literal meanings in such studies. We investigated whether listeners in such tasks reason as pragmatically as presumed. Results indicate that participants interpret utterances based on a combination of literal meaning and salience of particular referents. To the extent that one-shot web-based experiments accurately elicit the depth of pragmatic reasoning seen in typical human interactions, these findings indicate that a simpler model than RSA can better explain human behavior.
Les Sikos, Noortje Venhuizen, Heiner Drenhaus and Matthew Crocker

Korean Verb Prediction in sentence comprehension

Research on prediction has investigated whether it is based on semantic/heuristic cues, syntactic cues, or both. Many studies have shown that syntactic cues are overridden by heuristics-based semantic cues/processors. However, Momma et al. (2016) found that changing Japanese case markers from N+Acc (grammatical) to N+Gen (ungrammatical) affects the predictability of an upcoming verb. This study expands previous findings by testing Korean, another verb-final language using eye-tracking measures. We failed to find an effect of grammaticality, which suggests that Koreans rely primarily on semantic/heuristic cues when making a prediction about a verb. It is not aligned with Momma et al. (2016).
Sea Hee Choi, Nayoung Kim, James Yoon and Kiel Christianson

Not all wug-tests are created equal: Cognitive load impairs access to the phonological grammar

Speakers 'probability match' on novel words, producing a distribution of output forms matching the distribution of form types found in the lexicon, even following complex and subtle statistical trends (Ernestus and Baayen, 2003; Hayes et al., 2009). This phenomenon has led to the adoption of models of the Phonological Grammar which are inherently probabilistic (Goldwater and Johnson, 2003; Hayes and Wilson, 2008). We argue that increased cognitive load impairs speakers' access to the probabilistic grammar. Exp. 1 shows consistent 'probability matching' behavior across participants, while Exp. 2 increases participants' cognitive load, and shows less probability matching and inconsistency across participants.
Claire Moore-Cantwell and Dave Kush

Words Take Time: Auditory Stimuli and Strategic Processing in Semantic Priming

This study examines participants’ task-related strategy use in auditory semantic priming (SP) experiments. Traditionally used to study access to semantic representations of visually presented words, SP is now used with auditory stimuli, despite inherent differences in the processing time-course. Visual SP is highly susceptible to strategy use when inter-stimulus interval (ISI) and relatedness proportion are high. Whether auditory SP is similarly susceptible is unclear. No strategic processing effects were found in auditory SP while manipulating ISI and relatedness proportion across multiple experiments. This differs from visual SP, and supports the use of auditory SP for studying access to semantic representations.
Yosiane White

Relative clause processing in a flexible word order language: Evidence from Hungarian

Hungarian subject-extracted (SRC) and object-extracted (ORC) relative clauses both have flexible word order. SRCs can occur either in a VO (local) or OV (non-local) configuration, while ORCs can occur in VS (local) or SV (non-local). In a self-paced reading experiment, we found a reading time advantage at the RC verb for the more frequent (=non-local) SRC and ORC types, despite there being a longer distance between the filler and gap. Our findings support an expectation-based (i.a. Hale, 2001; Levy, 2008), rather than a memory-based (i.a. Gibson, 1998; Lewis & Vasishth, 2005) account of relative clause processing.
Eszter Ronai and Ming Xiang

Pragmatically (ir)rational: loss aversion bias in L2 speakers of English

Recent research suggests that people are less prone to irrational decision-making in their L2. We explore this in a study investigating loss aversion bias in native Chinese and Italian speakers and do not replicate the finding that L2 users are “more rational” in their decision making than L1 speakers. We use Tversky and Kahneman’s classic decision problem (1981) and find a significant framing effect for both the Chinese and the Italian group, consistent with the behavior of L1 speakers is previous studies. This suggests that highly proficient L2 speakers behave similar to L1 speakers in respect to loss aversion.
Zoe Schlueter, Chris Cummins and Antonella Sorace

Covert Structure and Zero Morphology in Sentence Processing

A word such as "visit" can be used as a verb (I expect to visit) or as a noun (I expect a visit) due to the presence of zero morphology. Past studies have uncovered some grammatical properties of these categorically ambiguous words, but no research has been conducted on how people process these words in reading. Through an eye-tracking experiment, this study shows that the zero-derived form of categorically ambiguous words results in a slower reading time, which we argue is due to the structural complexity of those words.
Alexandra Krauska and Masaya Yoshida

Processing causatives in naturalistic data

Here we propose a computational processing approach that simulates human processing of verb meaning in naturalistic utterances. We exemplify this method in simulating the processing of causatives in large corpora of three different genres: child-directed speech, adult-to-adult dialogues, and written language. The goal of this paper is to test whether there is genre-specific information that helps processing. Our main finding is that in child-directed speech verb meaning can be successfully processed by solely relying on the semantics of neighboring words, whereas in the other two genres structural information of neighboring words is of more importance.
Guanghao You, Moritz Daum and Sabine Stoll

How much does verb semantics determine verb syntax?

Verbs vary in terms of which syntactic frames they can appear in. The Semantic Consistency Hypothesis posits that there is a systematic relationship between a verb’s semantics and its syntactic behavior. We present a large-scale investigation of 83 English verb classes covering 2,167 verbs and 10,976 verb/frame combinations. We test six commonly invoked semantic features and one control feature. This analysis covers 5,159 verb/frame combinations per semantic feature. Semantic consistency, averaged across verb classes, is near ceiling for our six critical semantic features. Semantic Consistency is much lower for the control feature, lending support to the Semantic Consistency Hypothesis.
Mariela Jennings, Martha Palmer and Joshua Hartshorne

Comprehension of agreement mismatch errors in language transfers to non-congruent musical preferences

"The effects of language on musical processing have received very little attention. Existing results do not investigate the question of which cognitive mechanisms are shared between the two domains. In this study, we investigate whether participants are less sensitive to “errors” in chord progression after they are exposed to a sentence in which the subject and verb mismatch in number. Our results show that after reading a grammatical subject-verb agreement sentence and answering a comprehension question, participants show a preference for the musically congruent chord progression. However, when they read a mismatched (ungrammatical) sentence, their preference shifts significantly."
Mythili Menon and Drew Colcher

An ERP study on semantic effects on the processing of adjunct island violations

In EEG experiments, we examined adjunct islands with gaps of variable acceptability as determined by adjunct verb type. We showed that the effect of the a ±Extractable verb and a ±Plausible filler dissociate and interact. We found an N400 as a main effect of plausibility, and a probable P600 in the -Extractable, +Plausible condition that may show a selective construal of an adjunct clause as an island.
Annika Kohrt and Dustin Chacón

Individual Differences in Word Knowledge and Working Memory Capacity Modulate Verb Bias Effects

Individuals use verb bias to anticipate the resolution of locally ambiguous sentences, but it remains unclear whether cognitive abilities, such as word knowledge (WK) and working memory capacity (WMC), influence its use. We used direct object / sentential complement (DO/SC) ambiguous sentences and manipulated the consistency between verb bias and syntactic structure. Participants completed a self-paced reading task, a vocabulary test to assess WK, and complex span tasks to assess WMC. Mixed effects hierarchical modeling revealed that WMC modulated the interaction between bias and resolution at the critical and post-critical words, whereas offline acceptability ratings were modulated by WK.
Kathryn Bousquet, Tamara Swaab and Debra Long

Can identity conditions on ellipsis be explained by processing principles?

Chung (2006) argues that preposition omission under sprouting (John danced, but I don’t know *(with) who(m)) is degraded as compared to sluicing due to an identity condition requiring all words that are omitted in the sluice to be given in the antecedent. We present a rating and a self-paced reading experiment investigating whether this pattern can be explained by processing mechanisms: The wh phrase is more likely and therefore processed faster (Hale 2001) under sluicing. Our data show that not only mismatches violating Chung’s condition are degraded and read more slowly.
Robin Lemke, Lisa Schäfer and Ingo Reich

Agreement attraction in a mixed agreement system: Evidence from Hindi

We examined agreement attraction in Hindi, where subject-verb agreement and object-verb agreement occur in complementary structural contexts. Subject agreement structures and Object agreement structures were tested using a speeded binary-choice production task. Clear attraction effects were observed for object agreement, and modest effects for subject agreement. We found some evidence in support of a retrieval hypothesis for subject agreement: subject attractors led to more attraction than object attractors, but not for object agreement: subject attractors and object attractors led to a similar attraction rate. Overall, the attractor’s grammatical role did not appear to consistently modulate the rate of agreement attraction.
Sakshi Bhatia and Brian Dillon

Matrix verb transitivity acts as a cue for parasitic gapping in Swedish

In the current study, we used a filled-gap, eyetracking while reading paradigm to investigate whether matrix verb transitivity affects the likelihood of the parser engaging in parasitic gapping in Swedish complex subject relative clauses. Differences were found in total durations for sentences having intransitive vs. ditransitive matrix verbs, with the latter showing longer fixation durations when a filled gap (but not a gap) was present at a final gap site. We take this as an indication that ditransitive matrix verbs promote parasitic gapping during incremental parsing by providing a form of look-ahead for an upcoming second gap.
Damon Tutunjian and Anna-Lena Wiklund

Linear (not logarithmic) effects of lexical predictability: A challenge to surprisal theory

According to surprisal theory the primary determinant of online processing difficulty is the negative log-probability of a word given the preceding context (i.e. surprisal). In the present study, we tested the empirical predictions of suprisal theory in two high-powered experiments (self-paced reading (N = 216), cross-modal naming (N=36) - in which contextual probability was parametrically manipulated (1% vs. 20% vs. 90% cloze). In these experiments - and a separate experimental meta-analysis of prior eye-tracking studies - we observed significantly better model fits for linear compared to log-transformed predictability, contrary to previous correlational corpus studies.
Trevor Brothers and Gina Kuperberg

Linguistic focus as predictive attention allocation

We report a display-change eye-tracking study testing the hypothesis that readers use linguistic focus to predictively allocate attention to information in parafoveal vision. We asked whether readers show a larger parafoveal preview benefit for focused material. We found that readers spent longer reading: (1) target words with invalid parafoveal preview; (2) focused words. We found no interaction of focus and preview, suggesting that focus does not facilitate information uptake from parafoveal vision. However, readers were less likely to skip focused material, indicating that they are aware of which positions in a sentence are most informative prior to fixation.
Shayne Sloggett, Amanda Rysling and Adrian Staub

Individual strategies for resolving lexical and prosodic cues to certainty

Verbs like "think" have two different interpretations: they can indicate genuine uncertainty ("I think it's this way?"), or they can be used to hedge an assertion ("I think we should drop this"). The two interpretations are prototypically associated with different prosodic tunes, which elicit reliably different certainty judgments on a continuous rating scale. Here, we used these two tunes to probe explicit judgments of whether a speaker is certain or not. In this metalinguistic task, we found individual differences between participants in how they resolve conflicting lexical and prosodic cues.
Marie-Catherine de Marneffe, Miranda Dickerman, Micha Elsner and Shari Speer

A cross-sectional study of non-native Spanish sentence processing

In this cross-sectional self-paced reading study, we tracked the development of second-language Spanish learners’ sensitivity to changes in word order and the lexical class of verbs in sentences. A group of native Spanish speakers also completed the task as a control. The results revealed that in all groups, subject-verb-object (SVO) word order facilitated processing when compared to object-verb-subject (OVS) word order. However, only native Spanish speakers and the most advanced Spanish learners were able to use word order cues to make predictions about the class of verb that would appear in a sentence.
Russell Simonsen and Dustin Chacón

Contrastive effects with color, material and scalar adjectives in English, Hindi and Hungarian

We investigated the derivation of contrastive inferences with color, material and scalar adjectives using Sedivy’s classic paradigm (2003, 2004) in three languages with prenominal modification: English, Hindi and Hungarian. Unlike Sedivy, we predicted that contrastive effects should be observed not only with scalar and material adjectives, but also with color adjectives since color contrast is normally visually salient. The eye-tracking results supported our prediction, with the strongest contrastive effects (Target boost) being observed with scalar adjectives, whereas color and material adjectives elicited a mixture of Target boost and Competitor reduction. An unpredicted result, English speakers revealed the strongest contrastive effects.
Paula Rubio-Fernández and Julian Jara-Ettinger

Multiple grammars for singular they

Singular "they" has long been used refer to non-specific antecedents of unknown gender. Recently, it appears to have expanded to include specific referents with known gender. The present study used cluster-analyses to explore grammatical and pragmatic licensing conditions. Three clusters were revealed: participants who rejected "they" with definites, participants who accepted "they" for definites with unmarked gender but not marked gender, and participants who accepted "they" with any animate referent. Additionally, social distance between the speaker and referent affects acceptability of "they". Finally, younger participants and those with more familiarity of non-binary individuals were more accepting of singular "they".
Sadie Camilliere, Amanda Izes, Olivia Leventhal and Daniel Grodner

Large neural network language models learn representations of incremental parse states

We ask whether large neural network language models learn incremental representations of syntactic parses. We treat the models as black boxes and perform behavioral experiments on them to elicit information about their incremental syntactic representations as if they were human subjects in psycholinguistics studies. We show that RNNs maintain representations of parse states as evidenced by MV/RR and NP/Z garden path effects reflected in incremental surprisal values. However, RNNs differ from human behavior in that they show decreased garden path effects with increasing time from the onset of the local ambiguity to the disambiguator, in a reversal of digging-in effects.
Richard Futrell, Ethan Wilcox, Takashi Morita, Peng Qian, Miguel Ballesteros and Roger Levy

A case study of long-distance dependencies in Korean with wh-scope

This study investigates the processing of long-distance dependencies in Korean, a head-final language, with a focus on wh-scope. The results of this study support the full constraint driven principle that the processing of filler-gap dependencies is conditioned by the need to satisfy grammatical requirements (theta-role assignment and wh-scope licensing) as well as to create a gap as soon as possible.
So Young Lee

Developmental parsing across SES: Trade-offs between cue reliability and input quantity

By age five, children exploit word order to assign subjects as agents, verb semantics to anticipate objects, and verb syntax to infer roles. Trade-offs between input quantity and cue reliability raise questions of what comprehension is like when input varies with SES. We examined comprehension of actives and passives in children from varying SES. For definite NP1s, children were more accurate for actives compared to passives, but those with larger vocabulary revised passives to a greater extent. For pronoun NP1s, accuracy across constructions improved with vocabulary size. These findings suggest that knowledge of verb-specific biases enables effective revision agent-first biases.
Zoe Ovans, Kathleen Oppenheimer and Yi Ting Huang

L1 acquisition of polarity sensitivity: The case of "either" and "too"

In an experimental investigation of how children acquire the polarity sensitivity of the negative polarity item (NPI) "either" and its positive polarity item (PPI) counterpart "too", we observe that a) children aged 3-5 exhibit a weak but stable inkling of the contrasting licensing requirements for both items and b) at age 6 they exhibit a categorical understanding of the licensing conditions for "either" but not for "too". This adds to a growing body of results suggesting that NPIs are easier to acquire than PPIs but also raises the question of what 6-year-olds have learned about "either" that 3-5-year-olds are missing.
Naomi Francis, Leo Rosenstein and Martin Hackl

How Interaction affects (Un)certainty about the Partners’ Perspectives

We consider perspective-taking in comprehension, demonstrating that listeners are sensitive to the speaker’s behavior beyond just the knowledge that can logically be attributed to her: when the speaker’s behavior indicates ignorance in the situation, listeners ignore objects she cannot see when processing her instructions, but when she instead shows that she is knowledgeable in the situation (especially through interaction), listeners’ consider that the speaker is also knowledgeable about hidden objects. We explain this pattern using the multiple-perspectives framework, which probabilistically combines the influence of the speaker’s and the listener’s perspective and explicitly encodes uncertainty about partners’ perspectives.
Xiaobei Zheng, Daphna Heller, Suzanne Stevenson and Richard Breheny

A rational model of word skipping in reading: ideal integration of visual and linguistic information

In a rational model of reading, word skipping decisions should be complex functions of the particular word, linguistic context, and visual information available. In contrast, simple heuristic models of reading only predict additive effects of word and context features. Here we test these predictions by implementing a rational model with Bayesian inference, and predicting human skipping with the entropy of this model’s posterior distribution. Results showed a significant effect of this entropy in predicting skipping, above a strong baseline model including word and context features, suggesting complex interactions between visual input and linguistic knowledge predicted by rational models of reading.
Yunyan Duan and Klinton Bicknell

Language and Aging: neurocognitive correlates of accent processing

Many older adults have increased difficulty processing spoken language, including accented speech. Decreased hearing acuity is often cited as the cause, but this is not unequivocally supported by research. In this study we examined older adults’ processing of foreign- and native-accented speech, and explored how this is affected by variation in cognitive and linguistic abilities, using EEG/ERPs. Participants were highly accurate in comprehension of both native and non-native accented speech, but EEG results show that older adults have difficulty processing foreign-accented speech (in terms of sensitivity to both semantic and pronoun errors), more so than native-accented speech.
Fatemeh Abdollahi and Janet van Hell

That project was a rollercoaster: An ERP test of deliberate metaphor

Conceptual metaphor theory (CMT) holds that all metaphors are processed via cross-domain mappings between two concepts. Deliberate metaphor theory (DMT) proposes that this process of conceptual mapping is only recruited when processing deliberate metaphors, i.e. metaphors that are intended to be explicit comparisons and are processed as such; according to DMT, non-deliberate metaphors involve comparisons that are not explicit, and so do not invite cross-domain mappings. In ¬this ERP study, we compared the N400 effect for deliberate and non-deliberate metaphors. We found that the N400 amplitude does not differ for deliberate and non-deliberate metaphors, supporting CMT over DMT.
Sophie Greene, Daniel Altshuler and Joanna Morris

Trade-offs between iconicity and structure in the evolution of combinatorial phonology

One design feature of language is its combinatorial phonology, allowing the formation of an unbounded set of utterances from a finite set of building blocks. Since these building blocks must be small and meaningless, however, combinatoriality appears to be in conflict with iconicity. We investigate the emergence of iconicity and combinatoriality in an iterated learning experiment where participants produce signals using a slide whistle instrument. We find that, as languages become increasingly more combinatorial, iconicity emerges rapidly but subsequently disappears. We discuss implications of our findings for ideas about the interaction of biases for iconicity and structure in language evolution.
Matthias Hofer and Roger Levy

ERP correlates for the meaning and structure of ambiguous phrases in context

We investigated the differential impact of nearby local and long-distance global contexts on ambiguous 3-word phrases. The target phrases (wrecked commuter ship, furry tiger paining, tall wheat tractor, automatic copying machine) were preceded by biasing global contexts. ERPs at the 2nd word indicated that global context mattered more. Left-Branching Biasing Context ‘ship for wrecked passengers’ facilitates interpretation compared to Right-Branching Biasing Context ‘passenger ship that is wrecked’. ERPs at the 3rd word indicated that the local context mattered more. Long-distance cues can suppress sensitivity to immediately preceding contextual cues. This supports sentence processing models positing immediate integration of long-distance context.
Rachel Brown, Vicky Tzuyin Lai and Thomas Bever

Listeners’ predictions of sentence lengths are categorical, not gradient

We report two identical pre-registered studies designed to test whether listeners can use F0 declination to predict the length of a sentence, as Grosjean (1983) first argued. We systematically fail to replicate Grosjean's finding that listeners can differentiate minimally different sentences that continue for six more words versus three more words after a syntactically well-formed potential last word. In light of both our results and earlier failures to replicate this difference, we argue that sentence continuation predictions are best understood as categorical — whether a sentence is ending or continuing at all — not gradient predictions of continuation length.
Amanda Rysling, Charles Clifton, Nick Van Handel, Ria Geguera, Hanna Choi and Anne Cutler

Illusory licensing of wh-phrases in Japanese: A preliminary study using speeded acceptability judgment task

"This study shows grammatical illusions in Japanese wh-licensing using speeded acceptability judgment experiment. Ungrammatical sentences which have a licensor in syntactically ineligible position got higher acceptability than those without licensors. A well-known example of illusory licensing is NPI-licensing in English/German, which is retrospective/possibly pragmatic licensing. On the other hand, Japanese wh-licensing is prospective, purely syntactic, in other words, licensees precede licensors and they can be licensed by only morpho-syntactically appropriate elements. In this respect, this study demonstrates the generality of cue-based memory retrieval in licensing processes. Further studies are planned to reveal incremental processes of illusory Japanese wh-licensing."
Itsuki Minemi

The non-quantum nature of parsing the late-closure ambiguity

This study investigated if sentence processing is analogous to the both-ness feature of quantum physics. Participants read sentences such as “While Anna dressed the baby spit up on the bed” and answered two comprehension questions that targeted comprehension of the main clause, the mis-parse and the correct parse. There were six conditions: Corr-then-Mis, Mis-Corr, Corr-Main, Main-Corr, Mis-Main, and Main-Mis. Results showed participants either correctly or incorrectly parsed the sentences, only rarely asserting that “Anna dressed the baby” and that “Anna dressed herself”, which is compatible with the earlier version but not the later version of the good-enough sentence processing account.
Zhiying Qian and Gary Dell

Poster Session B Saturday (Day 2) 8:15 - 10:00 AM


  • East & West End Rooms: posters #B1-B41
  • The Ballroom: posters #B42-B64
  • Pre-Function Area: posters #B65-B80

Syntactic interlanguage speech benefit: an ERP study

The present ERP study is aimed at testing whether an interlanguage speech benefit can be observed in the syntactic domain.
Sendy Caffarra, Ana Gonzalez and Clara D. Martin

Active antecedent search in cataphora processing: Insights from neural oscillations

We investigated the interplay between structural and lexical information during cataphora resolution. We used EEG to measure beta oscillations while participants read sentences containing a cataphor that triggered a search for a later antecedent. We varied whether a later name (i) matched/mismatched the cataphor's gender and (ii) was a structurally acceptable antecedent for the cataphor. We found a gender-mismatch effect at the name when it was an acceptable antecedent for the cataphor (decreased beta power to mismatch), but no such effect was observed when the name was not an acceptable antecedent.
Dave Kush, Ashley Lewis, Andrew Jahn, Luca Campanelli, Clinton L. Johns and Julie Van Dyke

Sound-meaning systematicity in early word learning

Early word learning relates both to a word's semantic content, its phonological form, and how these features systematically vary across words. Research to date has treated systematicity as a word-level value that does not change based on an individual child’s vocabulary. We sought to address this potential problem by investigating the degree of sound-meaning systematicity in individual children’s vocabularies. We found that in several different languages, children's vocabularies tended to demonstrate more sound-meaning systematicity than would be expected based on other word characteristics.
Justin Kueser and Laurence Leonard

Lingering Misinterpretation in Garden-path Sentences in Native and Non-Native Speakers

Non-natives often misinterpret garden-path sentences and persist with the misinterpretation (Jacob & Felser, 2016). Yet, why misinterpretation lingers is debated. To explore this issue, we conducted two experiments. In Experiment 1, natives/non-natives read temporarily ambiguous sentences like "After Mary woke up her husband drank coffee." and chose a picture corresponding to the content of the sentences. In Experiment 2, they read texts referring to the interpretation of the temporarily ambiguous sentences. The results provide further evidence that natives/non-natives persist with misinterpretation and show that persistence of misinterpretation results from a failure to erase the memory trace of the initially-assigned interpretation.
Hiroki Fujita and Ian Cunnings

Computer modeling suggests patterns of perceptual availability of phonological structure during infant language acquisition

Recent work on child language acquisition has stressed the importance of top-down lexical and phonotactic constraints for acquiring phoneme categories and phonological feature distinctions. However, to prevent the acquisition process from being circular, the speech signal itself must also provide evidence for these abstractions. We use unsupervised neural autoencoder models to simulate the learning of phoneme categories and phonological features from acoustics alone. Results show partial discovery of phonological categories, asymmetries in feature acquisition that replicate findings from infant phone discrimination studies, and fine gradations in perceptual availability between features which can be subjected to experimental testing.
Cory Shain and Micha Elsner

Evaluating a Minimalist Grammar Parser on Italian Relative Clause Asymmetries

"A top-down parser for Minimalist grammars (Stabler, 2013) can be combined with complexity metrics to relate parsing behavior to memory usage, and successfully used to model sentence processing across a variety of phenomena (Kobele et al., 2013 a.o.). At the core of this approach is a theory which explicitly connects grammatical structure and memory resources. Extending the range of phenomena correctly modeled by the parser is crucial to confirm the empirical feasibility of the approach. We test the parser on Italian relative clauses asymmetries: an interesting testing ground for the MG parser due to the pro-drop nature of the language."
Aniello De Santo

Generalized Quantifiers and Working Memory: Disentangling Encoding from Verification

"A large amount of literature has shown that the type of quantifier in a sentence significantly affects the verification procedure used to arrive at a truth-judgment (Pietroski et al., 2009). Few studies have explored effects of different quantifiers on cognitive load during early comprehension, in order to distinguish between quantifier characterization and verification procedures. This pilot study employed pupillometry measures in an auditory/visual verification task, to ask whether there are effects of quantifier type on working memory specifically during encoding, before subjects are allowed to engage in verification; and how these results reflect on the default representation of quantifiers."
Aniello De Santo and John Drury


Recent research suggests that speakers use fewer pronouns (more nouns) when referential candidates are more similar. We examined the level of representation at which such an effect might arise in English, Italian, and French. In all languages, the referents’ non-linguistic similarity lowered pronoun rates, with fewer pronominal expressions when target and competitor could be the referent than when only the target could be the referent. Neither the antecedents’ semantic similarity nor their phonological similarity reliably lowered the pronoun rate. Hence, the choice of using a pronoun or noun is affected by the referent’s non-linguistic similarity rather than their linguistic similarity.
Kumiko Fukumura, Coralie HERVE, Sandra Villata, Francesca Foppolo and F.-Xavier Alario

Not-so strategic prediction: N400 effect of predictability regardless of overall predictive validity

In order to better understand the effects of experimental context on predictive processing, the present study used event-related potentials (ERPs) in two experiments to ask whether changing the overall predictive validity of the experimental stimuli will strengthen (or weaken) predictive processing and, as such, modulate the effect of cloze probability on the N400, a negative-going ERP component peaking at around 400ms after stimulus onset. We found that predictive validity did not modulate the effect of cloze probability on the N400 response at all and propose that comprehenders engage in predictive processing even when the experimental context does not encourage prediction.
Wenjia Zhang, Wing-Yee Chow, Bo Liang, Hsuan-Chih Chen and Suiping Wang

Does implicit training lead to generalization? Evidence from an understudied construction

This paper explores whether implicit training on a novel dialectal variant leads to generalization regarding its syntactic and semantic properties. Previous research shows that speakers process novel dialectal variants more easily following repeated exposure, suggesting that exposure leads to generalized knowledge. This paper probes this conclusion using the Negative Auxiliary Inversion (NAI) construction (e.g., "didn’t everybody eat", on the reading ‘not everybody ate’), which is semantically unambiguous for native speakers. Experiment results suggest that input that pushes participants to develop more target-like form-meaning connections may be necessary for fostering a deeper understanding of the linguistic constraints governing unfamiliar syntactic structures.
Frances Blanchette, Erin Flannery and Carrie Jackson

Disappearing causative overgeneralization errors across five languages: The role of semantics.

The current study investigated how learners of five languages (English, Hebrew, Hindi, Japanese, K’iche’ Mayan) avoid causative overgeneralization errors while maintaining productive use of grammatical constructions. With 715/720 participants currently tested, preliminary analyses suggest that – with the exception of K’iche’ – within each language, children at least from age 5-6 are able to use the inherent semantics of a given action to determine which type of causative marking applies to individual verbs, and hence to avoid overgeneralization errors. Furthermore, the results suggest a degree of universality in the extent to which particular actions/verbs prefer more- or less-transparent causative marking.
Ramya Maitreyee, Ben Ambridge, Bhuvana Narasimhan, Rukmini Nair, Dipti Sharma and Soumitra Samanta
B12 Individual variation in the interpretation of complex sentences Margreet Vogelzang and Esther Ruigendijk

NPI Illusions are a problem of quantification

Much recent work has investigated the extent to which negative polarity items (NPIs) are susceptible to illusions of grammaticality. This grammaticality illusion has led some researchers to suggest that NPIs engage a cue-based retrieval mechanism to identify their licensors. Recent work has shown that NPI illusions are surprisingly fragile: Not all NPIs are illusorily licensed and some licensors may not always spuriously license NPIs. Against this background, the present study shows that only quantificational negation (e.g., [no fisherman]NP) produces NPI illusions through six speeded acceptability judgement studies. We propose that difficulty in representing scope in early processing underlies these illusions.
Wesley Orth, Masaya Yoshida and Shayne Sloggett

Incremental Generation Drives "Efficient" Language Production

While the Uniform Information Density hypothesis has been influential throughout the computational psycholinguistics literature, I argue that a general framework of Incremental Generation serves as a causal mechanism to these empirical output trends. Owing to the modular nature of the language production system, any correlates with lexical access time will also connect to output linear order. I explore such issues using statistical modeling over an extremely large sample of English verb-particle data and argue that, to whatever degree the output of the language production system is ‘efficient’ in information ordering, this is an emergent property of Incremental Generation.
Spencer Caplan

A causal influence of domain-general cognitive control on sentence comprehension

Language input arrives quickly, yet listeners rapidly coordinate cues to guide real-time comprehension. We investigated domain-general mechanisms that underlie listeners’ ability to correct language-processing errors as input unfolds. We tested whether engaging cognitive-control processes – which detect and resolve conflict – can facilitate revision during real-time language comprehension, by recording eye-movements as listeners followed temporarily ambiguous instructions. We observed earlier revision when cognitive control, but not attention, was elicited on a preceding trial. These results extend previously observed correlations between cognitive control and language processing by revealing a causal link that connects domain-general cognitive-control procedures to subsequent language processing.
Nina Hsu, Stefanie Kuchinsky and Jared Novick

What dependencies can we [ [interpret the semantics] and [construct __ ]] ?

In processing filler-gap dependencies (FGDs), conjuncts are typically considered 'islands' that do not permit extraction. Our study investigated whether gaps are permissible in the final conjunct when the events described by the conjuncts form a natural series of events. In a judgment study, we found that participants preferred extraction from conjuncts when the conjuncts formed a natural series of events. In a self-paced reading experiment, we found that the semantic relation of the two conjuncts had no effect on the processing of a gap in the final conjunct.
Peter O'Neill and Dustin Chacón

The psychological reality of Construction Grammar: A Visual World eye-tracking study investigating the semantic processing of argument structure

Theories of Construction Grammar suppose that argument structure constructions contribute to sentence meaning. However, it remains unclear if semantic cues from argument structure can be accessed in on-line sentence comprehension. In a Visual-World eye-tracking study, we could show that participants – when hearing a sentence – used argument structure knowledge in order to identify scenes resembling the abstract meaning of the sentence’s argument structure construction, even if that scene did not depict any lexical items comprised within the sentence. The results suggest that abstract construction knowledge is applied in sentence processing at an early stage of comprehension.
Simon Kirsch and Lars Konieczny

Reward-based decision-making during reading for speed versus for accuracy

Daniel J. Schad, Sophia Czapka and Shravan Vasishth

Syntactic prediction without lexical activation: Evidence from both (of the)…and

The arrival of the upcoming disjunction ‘or’ has been shown to be facilitated by the presence of its associated word ‘either’. It is unclear whether this pattern is entirely due to the parser building a prediction based on syntactic structure, as some constructions using ‘either’ require the presence of ‘or’, or if it is, at least in part, due to a mere association between the two words. Here, we use the two elements, ‘both…and’, which have a weaker association than ‘either…or’, to determine whether the facilitative pattern previously demonstrated varies based upon the strength of the association between the elements.
Kelly-Ann Blake, Frederick Gietz and Margaret Grant

Predicting form during sentence processing: Investigating lexically-based processes

This research explored the role of basic lexical processes in pre-activating form during sentence processing (phonology/orthography). In Experiment 1, participants heard high cloze probability sentences like “The tourists expected rain when the sun went behind the cloud” while viewing visual arrays with a cloud (target), clown (phonological competitor), bear and globe (unrelated distractors) (see Ito et al., 2018). In Experiment 2, participants heard the most target-related word from the context (“rain”) with identical arrays. In both, clown was fixated significantly more than globe; E2 reveals the extent to which activation spreads within the lexicon, providing a potential explanation for pre-activation.
Anuenue Kukona and Aaliyah Summerfield

ERP evidence of object agreement attraction in comprehension

We investigated agreement attraction in German SOV structures using ERPs and grammaticality judgments. While most attraction research in comprehension has focused on PP-modifier constructions, the question whether attraction effects occur in obligatory sentence constituents, e.g., direct objects (DO), is less clear.  We found a reduced P600 at the ungrammatical plural verb in response to a plural attractor. These effects were persistent, leading to increased misjudgments in plural DO sentences. These results add to the body of findings showing that attraction errors can arise in a wide range of syntactic configurations and languages.
Robin Schäfer, Sol Lago and Titus von der Malsburg

Language influences event memory after sleep

Yaqi Wang, Silvia Gennari and Gareth Gaskell

Acceptability of homophonous sequences in adjective-derived adverbs

The combination of -ly adverbial and adjectival suffixes can be observed in naturalistic speech when a speaker is “backed into a corner” syntactically, and are also infrequently attested in corpora like the Corpus of Contemporary American English (Davies, 2008-) in examples like this self-aware New York Times quote (Brantley, 2008): “…is being seriously silly. Or is it sillily serious? Sillily -- that can't be right.” The present study uses a lexical decision task as a method for exploring gradient differences in acceptability, and the combinations of phonological and morphological constraints that lead to gradient acceptability.
Lauren Ackerman and Shiloh Drake

The partial and complete island repair of stripping

The results of three acceptability judgment experiments reveal that non-contrastive stripping completely ameliorates the effects of a definite relative clause island violation, and that corrective stripping partially ameliorates these effects, whether the correlate was utterance final in the antecedent or not, replicating and extending Potter (2017). These results run counter to prior claims in the literature (Griffiths and Lipták 2014). The central theoretical contribution is that definite relative clause island constraints, at least, cannot be exclusively reduced to processing factors (Hofmeister and Sag 2010); there are some irreducible grammatical aspects to these locality constraints (Fox and Pesetsky 2005).
David Potter and Katy Carlson

Speaker choice in language production: Optional passive marker BEI in Mandarin

We tested whether optional word mention is explained by two separate accounts. One is the interference account, which states semantically similar words interfere with lexical retrieval and therefore affect sentence production (Hsiao & MacDonald, 2016; H&M). The other is the noisy-channel account, which states there are various of noise in the communication procedure, such as humans’ inherent bias (Gibson et al., 2013; Levy, 2008). Speakers choose the linguistic form that facilitates robust information transmission. Our findings support the noisy-channel account over the interference account: The usage of optional functional words like BEI is driven by informative and efficient communication.
Yingzhao Zhou, Yingtong Liu and Edward Gibson

Flexible prediction: Global Context and Speaker Reliability Affect Sentence Processing

Trevor Brothers, Liv Hoversten, Shruti Dave, Tamara Swaab and Matt Traxler

Structure-sensitive pronoun processing even in the absence of Principle B effects

Principle B of Binding Theory prohibits coreference between a pronoun and an antecedent in its local clause. Previous studies have reported conflicting results regarding whether English speakers use Principle B to filter the interpretation of pronouns in real-time comprehension. We ask whether structural constraints guide early pronoun resolution processes in Vietnamese, where a pronoun can corefer with a local referential antecedent. The results from our self-paced reading experiment suggest that despite the lack of clear Principle B effects in the language, Vietnamese speakers still prefer non-local antecedents, showing a similar structural bias in processing to that observed in English.
Thuy Bui and Brian Dillon

Toddlers rapidly infer multiple polysemous meanings in familiar and unfamiliar languages

Up to 80% of words have multiple, related meanings (polysemy), yet work on word learning has almost uniformly assumed one-to-one mappings to meaning. Using looking-while-listening, we present the first evidence that toddlers recognize multiple meanings. In an English-meaning condition, toddlers were tested on their ability to recognize meanings for polysemous words (e.g., a baseball cap, bottle cap). Another condition prompted toddlers with the same English words, but targets instead corresponded to the word’s polysemous extension in an unfamiliar language, (e.g., lid for Spanish’s “cap”, tapa). Toddlers demonstrate both recognition of familiar word-meaning pairs and an ability to infer new meanings.
Sammy Floyd, Mariesa Cay, Adele Goldberg and Casey Lew-Williams

Similarity-based interference and morphological retrieval in Portuguese sluiced sentences

Cue-based retrieval models predict reduced retrieval accuracy when the retrieval cue overlaps with more than one item in memory (similarity-based interference, SBI). This study addresses the extent to which morphological gender and number interact with structural biases (Locality bias) during antecedent retrieval in Brazilian Portuguese sluiced sentences. The results of two self-paced reading studies indicate that both number and gender cues can generate SBI effects when the Locality bias is violated, corroborating previous findings (Harris, 2015, 2019). However, gender cues elicited earlier SBI effects than number cues did, suggesting that not all morphological cues have equal status in retrieval.
Alexandra Lawn and Jesse Harris

Is explicit teaching of second languages always helpful? An artificial language study

The role of explicit teaching in L2 acquisition was investigated in an artificial language learning task with either explicit instruction or not. We also manipulated the similarity of the artificial language to learners L1. Comprehension and production tests showed that L1 dissimilar languages were learned equally well with just exposure to the language or with additional explicit instructions. L1 similar languages were also learned well from just exposure. But surprisingly, we found explicit instruction negatively impacted the L1 similar languages.
Hiroko Yamashita and Franklin Chang

L2 learners can predict during complex sentence processing

This study aimed to further specify the conditions under which L2 listeners can generate predictions. Using a visual world eye-tracking paradigm, the current study found evidence for L2 prediction during complex sentence processing. Similar to native speakers of English, advanced Mandarin L2 learners of English showed significantly more anticipatory looks to the targets while listening to complex sentences (e.g., relative clause sentences with a complex noun phrase) containing a semantically biasing verb than a neutral one. These results provide more evidence supporting the claim that L2 listeners use the same mechanisms for prediction as L1 listeners.
Eunjin Chun, Joshua Daniel, Chenyue Zhao and Edith Kaan

Reduplicated words are automatically decomposed, even when they are fake: MEG evidence from Tagalog pseudoreduplicants

Morphologically complex words are decomposed in the visual system based on their orthographic properties such that forms which imitate a morphological form (e.g. brother) are decomposed automatically even in the absence of any morphological relationship between their parts. This study of Tagalog aims to inform visual word recognition for words with non-linear morphological processes. Our results suggest two important implications for neural correlates of morphological decomposition. First, reduplication is comparable to affixation in that it is automatically parsed by the visual system during word recognition. Second, phono-orthographic cues to morpheme boundaries aid in this automatic decomposition.
Samantha Wray, Alec Marantz and Linnaea Stockall

A role for theory of mind in the N400 differentiating metaphors of mental and physical content

The processing of two types of metaphors were compared with ERPs: those where the intended meaning is based on physical properties of the lexical concept (Phys), e.g., dancers are butterflies, vs. those based on mental properties (Ment), e.g., teachers are books. The larger N400 for Phys metaphors indicated that the processing differences with Ment concern the search in memory for the metaphor-relevant properties, rather than the final derivation of the intended meaning. ToM skills modulated these differences, suggesting that such skills make the search for the metaphor-relevant properties less demanding, especially when metaphors involve mental contents.
Paolo Canal, Luca Bischetti, Chiara Bertini, Irene Ricci, Serena Lecce and Valentina Bambini

Sensitivity to Stroop interference facilitates anticipation-driven comprehension

The study aims to investigate whether the goodness of anticipation-driven comprehension might differ as a function of readers’ sensitivity to semantic access rather than a function of readers’ WM storage capacity. By conducting picture description and self-paced reading comprehension together with cognitive tasks (Stroop and Reading span), we demonstrated that the differences of readers’ sensitivity to Stroop significantly predicted RT variations on the anticipatory use of lexical and structural information. We speculated that automatic semantic access, indicated by Stroop performance, could help readers to utilize available information in the processing of upcoming information.
Hongoak Yun and Eunkyung Yi

Topicalization modulates the processing of a new topic in Chinese concession: An eye-tracking study

In this study, participants read Chinese concessive sentences suiran…dan… ‘although…but…’ where the first clause contains a special double-subject structure [NP1 NP2 PREDICATE]. NP1 was made salient or unmarked by being put before or after the connective suiran (saliency effect), meanwhile the subject in the second clause was manipulated as a continuous or a completely new topic (shift effect). We found significant interactions between topic saliency and topic shift (Region 1: total fixation; Region 3: total reading and total fixation), suggesting that the topic established syntactically in terms of its position with the connective modulates the building of larger discourse topic.
Siqi Lyu, Jung-Yueh Tu and Charles Lin

Beat gesture increases cognitive load during online contrastive reference resolution

In this study, we examined how beat gesture and contrastive pitch accenting affect comprehenders' cognitive load via pupillometry during reference resolution in spoken discourse. In a visual world paradigm, participants heard pairs of referring expressions that either contrasted in color or differed in both color and shape. The results indicated that listeners' pupil size was larger in trials with beat gesture than without it; however, pupil size did not differ by pitch accent or contrast type, and all interactions failed to reach significance. These findings suggest that, unlike auditory cues, visual cues may increase comprehenders’ cognitive load via multimodality.
Laura Morett, Jennifer Roche, Scott Fraundorf and James McPartland

When are gaps understood as symptoms or signals?

Recently, turn-taking gaps have been viewed as a symptom of predictive planning mechanisms. Other works has shown that gaps can take signaling functions due to politeness and response bias . We tested whether gaps take on signaling functions independently of these factors by testing how gaps affect the processing of no responses to negated antecedents and scalar implicatures. Overall, these experiments show that gaps lengths can have signaling functions beyond politeness and response bias.
John Michael Tomlinson, Jr. and Ina Baier

Real-time spoken word recognition in sentences by adult and child cochlear implant users

When recognizing isolated words, cochlear implant (CI) users are slower than normal hearing (NH) listeners to commit to the target, and they show heightened activation for competitors. We examined how these atypical word recognition dynamics manifest in sentences. Adults and children with CIs and NH completed a visual world paradigm task to examine the dynamics of lexical competition of words in semantically neutral sentences. Results suggest that even a sentence lacking semantic information leads listeners to wait even longer, and show reduced competitor activation as a result. Implications for clinical speech perception measures are discussed.
Kelsey Klein, Ashley Farris-Trimble and Bob McMurray

ERPs Reveal Predictive Activation of Word Form Features in Sensory Cortex

Phillip Gilley, Leif Oines and Albert Kim

Effects of frequency and simplicity in L2 English causative motion production

How do L2 speakers of English choose a linguistic form when describing a scene involving causative motion? Our results showed that L2 speakers omitted Path expressions more often than L1 speakers. However, L1 and L2 English speakers did not differ in their choice of structure for expressing Cause. Both groups used the canonical transitive construction most often. Among L2 speakers, level of proficiency showed no significant effect. Our findings support the Complex Adaptive System Principles model (Hawkins & Filipović 2013) which predicts that L2 speakers prefer forms that are frequently-used, structurally simple and resemble those in their L1.
Chun Zheng, Josh Weirick and Elaine Francis

Production training benefits both comprehension and production of grammatical gender agreement in L2 German

In an artificial language learning study, Hopman and MacDonald (2018) showed that production-based training led to more accurate and faster performance on post-training comprehension tasks targeting agreement features than comprehension-based training. We tested whether these findings extend to comprehension and production skills in classroom-based learners by adapting the training paradigm to investigate the learning of grammatical gender assignment and agreement among beginning L2 German learners. The production group was more accurate than the comprehension group on all tests targeting comprehension and production of grammatical gender, underscoring the advantage of production over comprehension training for developing both comprehension and production skills.
Valérie Keppenne, Elise W. M. Hopman and Carrie N. Jackson

Thinking ahead has its limits: Structural prediction with correlative and quantificational "both"

Expressions such as "either" can be used in a correlative structure ("either...or") or as a quantifier ("either student"). Previous research found that sentence-initial correlative "either" facilitated coordination processing by generating a prediction for clausal coordination. This prediction appeared even for quantificational uses. In the present study, a self-paced experiment found that correlative "both" triggered a structural prediction sensitive to its syntactic position. In an eye-tracking experiment, however, there was no evidence of structural prediction when "both" was unambiguously quantificational. We conclude that while correlatives generate predictions about upcoming structure, predictions across correlatives are not equally strong.
Stephanie Rich and Jesse Harris

Experimental insights on the relationship between pragmatics, lying, and misleading

This study presents results from a response time experiment that presented participants with a story containing a straightforward lie, straightforward truth, false explicature, or false implicature and gathered yes/no judgments to the questions “is this statement a lie?” and “is this statement misleading?” A significant gradation in yes/no responses supports theories of linguistic meaning with a tripartite distinction (bare linguistic meaning, explicature, implicature). This gradation, along with response time analyses, indicates that theories of lying should adopt such a theory of meaning as a foundation. The results also help to clarify the relationship between the notions of lying and misleading.
Benjamin Weissman

Facilitatory interference reflects direct-access retrieval

This project investigates two prominent claims about the source of facilitatory interference effects in sentence processing. Facilitatory interference occurs when a distractor eases the processing of an ungrammatical dependency. Previously, it has been argued that such results might reflect differences in retrieval speed, or that they could even be reduced to interference at the stage of encoding. We report two experiments (speeded acceptability judgments + drift diffusion modeling, and self-paced reading) that provide evidence against these claims. Results are discussed within a direct-access memory framework.
Dan Parker and Adam An

Effects of animacy and unaccusative verb type on “transitivity bias”

We report results from two self-paced reading studies testing whether subject noun animacy has an effect on reading early-closure garden-path sentences. Since noun animacy is a cue for agentivity, unaccusative verbs which assign a non-agent role to the subject were used. Subject animacy was manipulated for alternating and non-alternating unaccusatives. As expected, alternating verbs showed difficulty at the disambiguating region with animate subjects. Non-alternating verbs showed earlier sensitivity to subject animacy, immediately post-verbally, suggesting that readers make predictions about thematic and argument structure based on animacy. Specifically, inanimate subjects seem to be assigned a non-agentive thematic role early on.
Peter Nelson and Amit Almor

The State of the Onion: Grammatical aspect modulates object representation in event comprehension

Tracking of object states is a fundamental mechanism in event comprehension. We asked whether viewpoint aspect can modulate object representation during the processing of change-of-state events. In an EEG sentence-picture-verification task, participants read sentences, with verbs presented either in the simple past or the past progressive (John] [chopped/was chopping] [the onion.]). Thereafter, they saw a picture of the object in its changed-state (chopped onion), its original-state (unchopped onion) or an unrelated object (cactus). Whereas the simple past (chopped) facilitated attention to object state-change, the progressive (was chopping) defocused it as reflected in P300 modulation: Grammar thus modulates online event representation.
Julia Misersky, Ksenija Slivac, Peter Hagoort and Monique Flecken

Noisy-channel sentence comprehension in aphasia: the role of noise in the context

Recent psycholinguistic theories propose that comprehenders integrate prior world and language knowledge with a model of potential noise to infer the intended meaning of a sentence. This framework has been proposed to account for failures to use syntactic cues in persons with aphasia (PWA). We find that PWA make more inferences overall than controls suggesting they expect a higher rate of noise and they do not appear to modulate the rates of inferences based on the amount of noise in the input.
Rachel Ryskin, Edward Gibson and Swathi Kiran

The Effect of Foreign Bias on Processing L2 English Speech

This experiment investigated whether negative associations towards foreigners affect the learning of incorrect labels by an L2 speaker (e.g., calling a crayon ‘marker’). In critical trials, the target object (e.g., marble) overlapped phonologically with the non-target (crayon) when it was mislabeled marker but not when it was labeled correctly. Participants were more attracted to the non-target when it was mislabeled compared to when it was correctly labeled, suggesting that participants learned the incorrect labels. Crucially, participants with less foreign bias (measured by an Implicit Association Task) became more attuned to the speaker’s speech over time compared to more biased participants.
Thomas St. Pierre and Jean-Pierre Koenig

Gender agreement attraction and grammatical illusion in Standard Arabic sentence comprehension

The retrieval of certain linguistic cues can lead to a failure in detecting an agreement error in a sentence causing grammatical attraction effects that reflect reduced sensitivity to ungrammaticalities. We report an eye-tracking study on gender agreement in Standard Arabic, focusing on whether attraction effects apply to a morphologically rich language and to a new dependency structure such as gender (dis)agreement between a relative-pronoun and a noun. Attraction effects can occur even when the language provides more cues for disambiguation, and when the attractor noun is retrieved earlier and is further from the RP than the noun it refers to.
Souad Kheder

Theory of Mind and Common Ground underpin mechanisms of language variation that lead to change: the Imperfective domain across three Spanish varieties

An SPR task in three Spanish varieties shows that only when interlocutors share perceptual access, Iberian and Argentinian participants allow the Simple Present to express a progressive meaning. Contrastively, in Mexican Spanish, share perceptual access no longer plays a role in improving Simple Present comprehension, and the only available option is the Present Progressive. The pattern observed across dialects is consistent with a model of variation embedded in a communicative system subject to tensions between Common Ground and Theory of Mind: while Common Ground affords the speaker greater reliance on context, Theory of Mind forces her to be linguistically explicit.
Martín Fuchs and María Mercedes Piñango

Contrastive focus constructions hurt memory for sentence processing

Good-enough processing predicts that repairs and focus work similarly. We asked whether types of focus constructions ("not X, but Y" vs. "Y, not X") had differential effects on memory for sentence content. We found that memory was worse for linguistic material that occurred in focus constructions, but that there were no differences between constructions. Given these results, we conclude that interference may arise during sentence processing or during memory retrieval, motivating future research on memory for focus.
Cassandra L. Jacobs and Fernanda Ferreira

Modeling ungrammaticality: A self-organizing model of islands

Drawing on experiments showing intermediate acceptability judgments for D(iscourse)-linked weak island violations, but clear negative judgments for Non-D-linked weak islands, and strong islands of any stripe, we propose Self-Organized Sentence Processing (SOSP). SOSP offers an independently motivated account of degrees of acceptability: linguistic structures result from continuous competitive-cooperative interaction amongst sentential elements. When combination requirements cannot be satisfied, the system forces the formation of sub-optimal structures. We claim SOSP offers a valuable new approach to grammar versus processing. Closely related to current generative theory, it differs in non-trivial ways, notably continuity, and a central role for processing in grammatical explanation.
Sandra Villata, Jon Sprouse and Whitney Tabor

Anti-locality effects without verb-final dependencies

Expectation- and memory-based sentence parsing models have both been argued to be adequate in accounting for anti-locality effects in verb-final constructions. In our study, we provide novel support for expectation-based models by investigating constructions that do not involve head-final dependencies. We conducted two self-paced reading experiments comparing the morphologically complex German determiner derjenige ‘the-jenig’, which obligatorily requires a relative clause, and the bare determiner der ‘the’, which does not. Both experiments found an anti-locality effect for derjenige, but not for der. The second experiment further found that derjenige generated an expectation that determined incremental parsing decisions on relative clause attachment.
Juliane Schwab, Ming Xiang and Mingya Liu

Not (just) any licensors cause negative polarity illusions

Initial attempts at resolving long-distance dependencies are occasionally error-prone (as in agreement attraction, anaphora, NPI licensing), and the similarities of these phenomena make highly general, memory-based explanations appealing. In a series of three speeded acceptability studies, we show that the online resolution of one dependency (NPI licensing) involves accessing and evaluating whole clausal meanings, and perhaps even richer representations. We find that the illusion’s sensitivity to the type of intrusive licensor (quantificational or sentential negation) is a consequence of clause-level meanings and its sensitivity to distance appears to be related to distance from the licensing context, not the licensor itself.
Hanna Muller, Iria de Dios Flores and Colin Phillips

Verb Surprisal in the Visual World

Is situated surprisal only manifest at nouns, and is referential entropy reduction not correlated with processing effort at all? Our study directly investigates these questions using a setup similar to Ankener et al. (2018) and interrogative German statements for a changed word order. Results from the Index of Cognitive Activity and eye movements indicate that visual context similarly affects the predictability of both verbs and nouns. We also replicate the lack of an effect on processing effort for the word that provides the constraining information. Thus, processing effort seems to correlate with situated surprisal but not with referential entropy reduction.
Christine Muljadi, Christine Ankener, Les Sikos and Maria Staudte

Embedded gapping? A cross-linguistic perspective

It is usually assumed that gapping differs from other ellipses in banning embedding. We show, based on 4 acceptability judgment tasks for Spanish, Romanian and English, that there is cross-linguistic variation wrt embedded gapping, and that two constraints seem to be at work: non factive verbs embed more easily than factive ones; no complementizer (with non factive verbs) embeds more easily than a complementizer. The difficulty of coordinating a simple and a complex clause may result from a more general parallelism constraint on coordination and the penalty on factive verbs may come from their non-assertive nature and/or from the QUD-incongruence.
Gabriela Bîlbîie, Israel de la Fuente and Anne Abeille
B57 World knowledge and the interpretation of relative and absolute adjectives Barbara Tomaszewicz and Petra B. Schumacher

Gender Effects on Gender Neutral Reflexives Resolution

Using the gender-neutral Chinese reflexive ziji as a test bed, we explored if people were sensitive to contextual gender information both in final interpretations and during the initial stages of resolution. Results showed that people were sensitive to contextual gender information and the different gender pair induced more competition between two candidates before encountering the reflexive. We hypothesize that the gender cue in the different gender condition is weighted higher because this gender information could subsequently be used to distinguish two candidates differing in gender. These expectation-based effects are above and beyond the cue-feature matching-based memory retrieval during resolution.
Yuhang Xu, Jeffrey Runner and Michael Tanenhaus

Processing information structural ambiguity: Contrastive Topics in Estonian

This study addresses the assignment of information structure during online processing in Estonian, a flexible word order language, where contrast can be grammatically encoded. A speeded acceptability judgment study showed that contrast is preferentially assigned to a clause-initial element in non-canonical Object-initial structures, but not in temporarily canonical Subject-initial structures. The findings indicate that contrast is computed and used rapidly, provided that the word order of the target clause is sufficiently constraining. In canonical clauses, where the information structure is less constrained, the processor may instead delay making commitments about discourse status.
Marju Kaps

Inferable constituents are not deaccented: Phonetic and perceptual evidence

It has been proposed that constituents can be deaccented in English when they are readily inferable from an antecedent constituent in a structurally isomorphic position. In a production study, participants read sentences containing verbs that were discourse-new, repeated, lexically entailed by an antecedent, or inferable via an accommodable implicational bridging relation. Analysis of intensity, duration, and f0, as well as a follow-up experiment collecting naive assessments of the emphasis status of the critical constituents, indicated that repeated constituents were deaccented, but discourse-new and inferable ones were not, casting doubt on the claim that readily inferable constituents can be deaccented.
Jeffrey Geiger and Ming Xiang

Prosodic Marking of Grammatical Roles in Turkish Sentences

Previous sentence processing literature focuses almost exclusively on lexical and syntactic cues used in reading in fixed word order languages. This study investigates whether prosodic cues are present in spoken language, in particular Turkish. We hypothesize that in the absence of syntactic and morphological cues, speakers of free word-order languages like Turkish provide prosodic cues that may facilitate comprehension. Our results showed a significant fundamental frequency drop difference in critical regions of sentences with and without morphological cues. We believe this reflects the utterer's effort of trying to provide additional cues in the absence of syntactic and morphological ones.
M. Yarkın Ergin and Karin Stromswold

Ellipsis and the QuD: evidence from sluicing with nominal antecedents

Sluicing is a cross-linguistically prevalent form of clausal ellipsis that occurs after interrogative wh-phrases. Traditionally, IDENTITY theories have postulated that sluices are acceptable only if the elided material is identical to its antecedent. Alternatively, QUD theories propose that sluices will be acceptable insofar as they correspond to a Question under Discussion made salient by the antecedent clause. In 3 experiments, we examine the predictions of these types of theories with respect to nominal-antecedent sluices. The results cast doubt over the validity of Identity theories and instead suggest that sluicing is affected by the availability of the relevant QUD.
Till Poppels and Andrew Kehler

The influence of prosody on (non-)restrictive relative clause interpretation

It has been argued that prosody is used to disambiguate between competing parses during spoken language comprehension for a variety of syntactic ambiguities. This study investigates whether prosody is used to disambiguate between the restrictive and non-restrictive interpretations of relative clauses. Twenty native English speakers were presented with string ambiguous context statements that contained relative clause constructions, wherein prosody was manipulated (restrictive, non-restrictive). Subjects rated the plausibility of continuation statements that biased restrictive-only interpretations of the context statements. The results we present suggest that prosody is used when developing semantic representations for context statements containing relative clauses.
Daniel Amy and Jeffrey Witzel

Frequency, predictability & lexical quality effects: individual differences in reading & proofreading

Word frequency and contextual fit have robust and independent impacts on reading time, and interact with task goals (e.g., reading for comprehension vs. proofreading for spelling errors), which influence the level of representation that must be processed (e.g., meaning vs. surface form). We investigated how individual differences in reading speed and ability to detect surface-form and semantic anomalies modulate the impact of frequency and predictability on reading times in these tasks. We find that worse readers are less sensitive to lexical properties, utilize context more than better readers, and that reading task influences these processes differently across individuals.
Sara Milligan and Elizabeth Schotter

Speech rate and language processing in older adults: Is slow speech better?

The present study explored how characteristics of “elderspeak" (exaggerated intonation, slower speech rate) affect older listeners' real-time language comprehension. Using a visual world methodology, we examined older listeners' ability to resolve ambiguities while hearing instructions relating to a visual display. The key ambiguity involved an inflected verb (“saw", past tense of “see") being temporarily interpreted as a noun (saw-tool), as reflected by fixations to a noun competitor. Manipulations included whether the sentence was spoken normally or with elderspeak, and whether the competitor was replaced with a control object. The results suggest a benefit of elderspeak on older adults' real-time comprehension.
Raheleh Saryazdi, Tamara Mostarac and Craig Chambers

Facilitation effects of QUD and event type on negative sentences

"While early investigations suggest that negative sentences are more difficult to comprehend, recent work suggests that supportive contexts can mitigate this processing cost, suggesting a pragmatic approach. In this study, we investigate how certain types of explicit QUDs can provide such supportive contexts by looking at constructions where the scope ambiguity between negation and a comparative quantiser arises, with the inverse scope reading hypothesised to be less accessible. Our findings suggest that speakers are sensitive to the inference introduced by ""how-many"" QUDs and the way it interacts with contextual information, ameliorating the comprehension difficulty of negation in inverse scope construal."
Sherry Yong Chen, Leo Rosenstein and Martin Hackl

Pronoun resolution in an ergative language: effects of case and transitivity

This study investigates pronoun resolution in Niuean, an ergative-absolutive Polynesian language. We examine how (i) case marking of the arguments and (ii) verb transitivity in an antecedent sentence affect interpretation of subject and object pronouns in a following sentence. We find an overall preference for subject antecedents (parallel to prior findings in nominative languages). For object pronouns, however, we also find effects of both case parallelism and transitivity: a subject antecedent that bears the same case as the pronoun is preferred over one that does not, and an obligatory argument is preferred as an antecedent over one that is optional.
Rebecca Tollan and Daphna Heller

Can our eyes trick our ears? An electrophysiological study of the effects of facial cues on the processing foreign-accented and native-accented speech

There is a growing number of speakers of English as a second language, which makes interactions with foreign-accented speakers very likely. In an ERP study, we examined how monolingual English listeners combine facial and linguistic cues in order to comprehend sentences produced by non-accented and Chinese-accented speakers, and to what extend these cues interact in semantic and syntactic processing. Results indicate that facial cues play an important role in activating certain stereotypes regarding a speaker and can affect the processing of both native and foreign-accented speech.
Carla Fernandez and Janet van Hell

Implicit objects in discourse: Likelihood of reference and choice of referring expression

Implicit objects (IOs) are the unrealized theme arguments of optionally transitive verbs (e.g., eat, read) in intransitive frames (Lisa ate.). Theoretical work suggests arguments are left implicit when they are low in discourse prominence. We test IOs likelihood of re-mention and likelihood of pronominalization in Experiments 1 and 2. Our findings reveal that IOs’ likelihood of re-mention is modulated by the preceding context. In contrast, rate of pronoun usage referring to IOs was not affected by prior context. Our work is the first to identify a divergence between likelihood of mention and likelihood of pronoun use with implicit arguments.
Ana Besserman and Elsi Kaiser

A systematic study of the rate of child over-irregularization errors

Overregularization errors, where a regular inflection is applied to an irregular verb, as in teach/teached or fly/flied is a well-documented and intensely studied phenomenon in English language acquisition. Conversely, over-irregularization errors, the overapplication of an irregular inflection to a regular verb like trick/truck, or the application of an incorrect irregular inflection to a non-target irregular verb, bring/brang, have been studied less thoroughly.The current study is a pilot project testing the feasibility of a novel methodology for estimating the true rate of such over-irregularization errors in child English using data from the CHILDES database.
Vanna Willerton, Graham Adachi-Kriege, Shijie Wu, Ryan Cotterell

Entertaining ungrammatical structures can succeed where surprisal struggles

Current, highly successful surprisal models of sentence processing hypothesize that a word’s processing time is proportional to how much redistribution of probability over grammatical structures it requires. We compared this approach to self-organized sentence processing (SOSP), which allows construction of subgrammatical as well as grammatical structures. We examined verb number choice (“is” vs. “are”) and reaction time after a range of pseudopartitives (e.g., “a basket (of flowers)”, “many (flowers)”). Surprisal handled the long conditions (“flowers” present) well, but, via its assumption that the short conditions involved suboptimal parse construction due to the elided NP, SOSP fit the full dataset better.
Garrett Smith and Whitney Tabor

Semantic attraction in sentence processing

We conducted two ERP studies-- a direct replication and a minimal extension of Kim & Osterhout (2005). K&O found that, in role reversals with a plausible thematic relationship between a noun phrase and verb (e.g., The hearty meal was devouring...), strong semantic attraction leads to syntactic repair (e.g., -ing to –ed) despite licit syntax, as evidenced by a semantic-thematic P600-- an effect replicated in our first study. For our second study, we removed auxiliaries (e.g., The glasses of beer drinking...) to determine whether providing fewer syntactic cues to predict a passive structure would eliminate the effect, which was found.
Kaylin Smith, Ye Ma, Yan Cong and Alan Beretta

Self-awareness matters: co-reference in German attitude clauses

This poster presents four offline experiments in German investigating how the mental attitude of an attitude holder influences the availability of a co-referential interpretation of an embedded pronoun. The experiments compared personal pronouns with a type of demonstrative pronoun in contexts that indicated the attitude holder to be either knowingly or accidentally co-referring with themself. The results showed that (i) accidental co-reference decreased the availability of co-reference overall and that (ii) participants do not differentiate between pronoun types unless they have to choose between pronouns directly, suggesting that the representation of complex mental attitudes incurs processing difficulty.
Alexander Göbel and Petra B. Schumacher

Object Who is processed differently from Subject Who, Why and How

Wh-Phrases (whPs) form a dependency with different grammatical elements: Why modifies (and thus is licensed by) the whole sentence and forms a dependency with S(entence) node, Subject who is also licensed by the sentence Those whPs that are linked to V and VP (object who & how) created additional complexity effects at the embedded NP. Once the wh-dependency is formed, these whPs are released from maintenance, thus whPs that form shorter dependencies did not create a processing complexity effect. These findings, in turn, support theories of online wh-dependency formation that involve a maintenance component.
Nayoun Kim, Alexis Wellwood and Masaya Yoshida

Listeners’ belief about the speaker and adaptation to the deviant use of prosody

Three visual-world eye-tracking experiments investigated how listeners responded to proper conventional (Exp.1) and unconventional (Exp.2-3) uses of contrastive accent. Exp.1 confirmed that conventional uses of contrastive accent lead anticipatory looks to a target object. Exp.2 showed no anticipation when contrastive accent was consistently misleading. When participants were told before the experiment that the speaker was not trustworthy in Exp.3, they learned to predict the upcoming referent with improper contrastive accent. The results suggest that listener’s beliefs about the speaker’s intentions modulates the degree of prosodic adaptation.
Chie Nakamura, Jesse Harris and Sun-Ah Jun

Fool me once: Readers Adapt to NP/Z garden paths but not ORCs

Syntactic adaptation is a key part of probabilistic expectation-based theories of comprehension, which hold that comprehenders adjust to the frequency of constructions in context. These theories suggest that frequency of a construction should be directly related to ease of comprehension. In two eyetracking experiments, we investigated adaptation when the target constructions (Object Relative Clauses and NP/Z garden paths) were either frequent or rare (32 vs 8 tokens). Adaptation rate for neither ORCs nor NP/Zs was significantly impacted by relative frequency across the experiments, suggesting that any facilitation over the course of the experiment may not be attributable to expectation-based adaptation.
Caroline Andrews, Brian Dillon and Adrian Staub

Perspectival plurality: interpreting multiple perspectival elements in one domain

Perspectival elements -- e.g. predicates of personal taste (PPTs, tasty, frightening) and logophoric anaphors (e.g. picture of herself) -- involve the point-of-view of a perspectival center/judge. However, little is known about utterances with two perspective-sensitive elements. We tested whether two perspectival elements can have distinct perspectival centers. In sentences like ‘Nora {told/heard from} Amy about the frightening photograph of {her/herself}’, does the perspectival center for the subjective adjective ‘frightening’ match the antecedent of the perspective-sensitive logophoric anaphor? Our results suggest the answer is no: We find evidence for perspectival plurality, even within the nominal domain.
Elsi Kaiser

Generalizing subjective opinions: evidence from two classes of perspectival adjectives

Our work investigates subjective perspective-taking. Adjectives called predicates of personal taste (PPTs, fun, tasty) express an individual’s (typically the speaker’s) subjective perspective/evaluation. Existing theories disagree whether utterances such as ‘The cake was tasty’ can express the opinion of people in general (vs. just the speaker). We tested whether comprehenders judge sentences with subjective adjectives to be generalizable to other evaluators beyond the speaker, and whether this depends on linguistic factors. We find that sentences with PPTs generalize beyond the speaker’s perspective, depending on linguistic context. Furthermore, we find support for semantic claims that PPTs differ from other subjective adjectives.
Elsi Kaiser and Sarah Hye-yeon Lee

Individual differences in ERP profiles to agreement attraction anomalies

Shannon McKnight, Akira Miyake and Albert Kim

Reassessing the grammaticality asymmetry in agreement attraction: An ROC analysis

Recent work has re-attributed the grammaticality asymmetry in agreement attraction to response bias, rather than the dynamics of memory access. In this paper, we use an ROC analysis of judgments to further test the relationship between bias and the grammaticality asymmetry with attractors in both PP modifier and ORC configurations.
Christopher Hammerly, Brian Dillon and Matt Wagers

Poster Session C Saturday (Day 2) 2:55 - 4:40 PM


  • East & West End Rooms: posters #C1-C41
  • The Ballroom: posters #C42-C64
  • Pre-Function Area: posters #C65-C80 

Verb bias in individuals with developmental language disorder

We examined verb bias sensitivity in children with developmental language disorder (DLD) relative to typically developing (TD) peers. Participants were 37 children, 7-9 years, 17 of whom were classified as having DLD. We employed mouse tracking in a visual world paradigm to obtain explicit and implicit measures of verb bias. Children showed sensitivity to verb bias in implicit but not explicit measures, with no differences between diagnostic groups. We found age effects when comparing results to a previous study with adults with and without DLD. Findings suggest protracted development for learning to weigh cues during sentence processing.
Jessica Hall, Amanda Owen Van Horne, and Thomas Farmer

Cross-linguistic influence on the interpretation of ambiguous wh-questions: L1 transfer but no L1 attrition

This study explores the extent to which German-English bilinguals’ interpretation of ambiguous wh-questions in German (e.g., Was jagt die Katze? ‘What chases the cat/What does the cat chase’) is influenced by the interpretation of their word-order-equivalent in English. Findings from a VWP study show such cross-linguistic influence in the final responses of L1-English L2 learners of German, but not among L1-German expats in the U.S. (aka ‘attriters’), whose on- and offline performance did not differ from L1-German speakers in Germany. These findings present evidence of resilience to L1 attrition at the level of syntactic processing even after long-term immersion.
Theres Grüter and Holger Hopp

The production of Object Relative clauses in Italian-speaking children: a syntactic priming study

"For children, Object Relative (OR) clauses with two animate noun phrases are difficult to comprehend and to produce across a number of languages. In the present study, we designed a new production task to explore the effects of syntactic priming on the production of ORs in Italian-speaking children aged 6. The results of the study demonstrate that children can be primed to produce ORs, and that they have underlying representations for ORs with two animate noun phrases. Our results are in line with previous studies suggesting that children have abstract representations for several types of syntactic structures."
Claudia Manetti and Carla Contemori

Learning subtle syntactic constraints in L2: Evidence from Norwegian-English bilinguals

Dave Kush and Anne Dahl

Effects of Misanalysed Filler-gap Dependencies in L1 and L2 Language Comprehension

Misinterpretation often lingers in garden-path sentences (Christianson et al., 2001), but how this influences filler-gap dependencies is little-known. We explored this issue by testing filler-gap dependencies like "Tom bought the novel which the girl read very happily about. vs. Tom bought the novel about which the girl read very happily.". In Experiment 1, natives/non-natives answered questions (Did the girl read the novel?). In Experiment 2, they read texts testing how the temporary ambiguity was resolved. These experiments showed that natives/non-natives often persist with misinterpretation in filler-gap dependencies due to a failure to erase the memory trace of the initially-assigned interpretation.
Hiroki Fujita and Ian Cunnings

Information structure and the processing of word order variation in the first and second language.

Presenting sentences with non-canonical word orders within contexts that meet their discourse requirements can reduce their processing cost because they are comprehended with greater ease when presented with a given-new information structure pattern. There is diverging evidence about whether the effects of information structure are the same in the processing of word order variation in the L2. This eye-tracking study explores how information structure affects the online comprehension of active (canonical) and passive (non-canonical) constructions in L2-English learners and monolinguals group. Results suggest that learners use discourse patterns in the comprehension of both canonical and non-canonical structures in their L2.
Priscila Lopez-Beltrán, Michael Johns, Paola Dussias, Cristóbal Lozano and Alfonso Palma

Processing Correlates of Verb Semantic Complexity

Predicate decomposition theory and construction grammar make distinctive assumptions of verb representations. A lexical decision task and a self-paced reading task were conducted to test verb reading times against the two hypotheses. Results from both experiments have supported the hypothesis of construction grammar.
Tun Scarlett Hao

Using word2vec to predict human language processing

Reading is sensitive to the semantic relationship between a word and upstream linguistic information (Luke & Christianson, 2016). This similarity has typically been quantified with latent semantic analysis (Landauer & Dumais, 1997). This study demonstrates the usefulness of word2vec for predicting semantic processing in eyetracking by testing how different corpus genres and model parameters affect the predictive power of word2vec-based semantic similarity scores. The best models of reading times incorporate multiple measures of semantic similarity, showing both that different models learn slightly different things from the same data and that readers use many types of semantic representations during reading.
Cassandra L. Jacobs and Katrin Erk

What Processing and Computational Modeling Can Tell Us about Syntax: The Case of Persian Relative Clauses

"In this abstract, we exploit a computationally specified parsing model as a bridge between processing results and theoretical syntax (Kobele, 2013), in order to discriminate between competing syntactic analyses of complex phenomena. We adopt a top-down parser for Minimalist grammars (Stabler 2013), combined with a set of metrics measuring memory usage --- a model which has been shown to successfully explain processing difficulty across a variety of phenomena (Graf et al., 2017) As a case study, we look at attachment ambiguities in Persian relative clauses, comparing Kayne's raising analysis of RCs (Kayne, 1994) to Karimi's base-generation approach (Karimi, 2001)."
Nazila Shafiei and Aniello De Santo

Individual differences guide pronoun interpretation in semantically constraining contexts

Syntactic and semantic cues influence pronoun interpretation. In "Ana threw the ball to Liz. She…", listeners tend to assign the pronoun to the goal (Liz), which conflicts with the bias to assign the pronoun to the subject. In two experiments, we investigate whether an individual’s print exposure (how much they read) affects the learning of these biases. Results show that print exposure correlates with use of the subject bias, but not the goal bias. This suggests that the subject bias, and not the goal, may be learned from exposure, supporting models in which referential probability explains pronoun comprehension.
Valerie Langlois and Jennifer Arnold

Reassessing the evidence for syntactic adaptation from self-paced reading studies

Recent self-paced reading (SPR) studies have found that there is a decrease in garden path effect over the course of the experiment (Fine et al 2013; Fine & Jaeger, 2016). This decrease has been interpreted as evidence for syntactic adaptation. With two SPR experiments we demonstrate that this interpretation is not necessarily valid. While we successfully replicated the decrease in garden path effect over time, we failed to find evidence that this was driven by syntactic adaptation. Rather, we demonstrate that this decrease was likely driven by asymmetric effects of task adaptation that impacts difficult sentences more than easy ones.
Grusha Prasad and Tal Linzen

Aspect is distinct from time reference: An ERP study of the perfective marker -le in Mandarin Chinese

Past neurolinguistic experiments (ERP), conducted on tense-prominent languages (Spanish, Dutch), suggest different processes for time reference and aspect. Concerning aspect-prominent languages (Mandarin), grammatical aspect violation needs syntactic repair, yielding a P600. However, the question of time reference processing in Mandarin is not clear. Precisely, the Mandarin perfective verbal morpheme -le is related (but not restricted) to past time reference. Does its violation by aspectual means need similar processes as when it is violated due to incongruent time reference? Our results revealed different patterns for processing time reference (left anterior negativity) and aspect (P600), despite the use of the same morpheme.
COLLART Aymeric and CHAN Shiaohui

Uniqueness vs. familiarity in interpreting definite descriptions

Uniqueness theories of definite descriptions claim that a description is felicitous IFF a unique referent satisfies its literal meaning (Russell, 1905; Evans, 1977; Lobner, 1985). Orthogonally, familiarity theories claim that reference succeeds IFF the referent has been made salient in the preceding discourse context (Kamp, 1981; Heim, 1982). Here, we experimentally investigate the process by which hearers interpret descriptions in English to moderate between these two theories of definite descriptions. We then implement a Rational Speech Acts model (RSA; Frank \& Goodman, 2012), and discuss in what ways hearers adhere to and diverge from rationality.
Sadhwi Srinivas and Kyle Rawlins

Convergent probabilistic cues do not trigger syntactic adaptation

Previous work has ostensibly shown that readers rapidly adapt to a priori less predicted structures after exposure to unbalanced statistical input (e.g. a high number of garden path sentences) and that these readers grow to disfavor the a priori more predicted structure after exposure. However, recent work has failed to replicate such findings. The current study uses three self-paced reading experiments to test whether co-occurring cues (preceding semantic cues and font color) can help facilitate adaptation to reduced relative/main verb garden path sentences. Results suggest an inability to rapidly overcome pre-existing predictive biases to adapt to statistically novel linguistic contexts.
Jack Dempsey, Kiel Christianson and Qiawen Liu

Non-binary gendered reference in LGBTQ+ English: Implications of singular ‘they’ for sentence processing

This study used an online Likert scale survey to measure the acceptability of third person singular ‘they’ (3SG ‘they’) in sentences for LGBTQ+ and cisgender heterosexual (cishet) participants, finding that across context conditions LGBTQ+ participants rate 3SG ‘they’ as significantly more acceptable than cishet participants, and do not conform to typical English name gender biases. Furthermore, naturalistic interview transcripts reveal consistent patterns of stance-mitigation discourse strategies such as fillers and hedging by cishet, but not LGBTQ+ speakers when referencing gender non-conforming individuals. These results may have implications for social constraints on non-binary gendered sentence production and processing.
Alecia Nichols, Elaine Chun and Amit Almor

Cloze completions reveal misinterpretation of noncanonical sentences

In two cloze experiments, we investigated online thematic role assignment in non-canonically ordered clauses. The experiments demonstrate a strong tendency in OSV clauses to produce verbs that are expected if the reader has reversed the roles of the two preverbal arguments, but are unexpected or anomalous under the correct role assignment. We take this to be evidence that verb predictions and incremental interpretations of non-canonical sentences can be guided by non-veridical parses of the input.
Jon Burnsky and Adrian Staub

The syntactic count/mass distinction in generalized classifier languages: evidence from classifier processing in Korean

Research on the count/mass distinction has investigated whether generalized classifier languages (e.g., Korean, Chinese) make the count/mass distinction in their grammars and if it is reflected in the use of the classifier system. In the current study, we seek to make headway on this debate by focusing on Korean. We found an increased N400 in the classifier-mismatched condition compared to the classifier-matched condition, but no significant P600 was observed. Our findings, consistent with the results of Kanero et al. (2015), suggest that classifiers are processed primarily semantically, eliciting the N400 effect, rather than as a syntactic (P600-eliciting) violation.
Sea Hee Choi, Si On Yoon, Andrew Armstrong, Kara Federmeier and James Yoon

Minimizing prediction errors: Comprehenders rapidly adapt to morphosyntactic violations but not to semantic violations

We conducted two event-related potential experiments that examined whether people adapt to morphosyntactically anomalous sentences and semantically anomalous sentences. The results suggest that people take into consideration not only the probability of violations but also types of prediction errors (i.e., how likely a type of error might occur) in determining whether to adapt to deviant linguistic input
Masataka Yano, Shugo Suwazono, Hiroshi Arao, Daichi Yasunaga and Hiroaki Oishi

EEG alpha power desynchronization during sentence planning is linked to partial overlap in syntactic configurations

All theories of grammar assume that linguistic expressions share partial syntactic configurations, modeled by derivation or inheritance mechanisms. Sentences like “The rabbits eat carrots” or “The mayor gives toys to the children” share the initial structure NP V… and thus exhibit partial overlap in their syntactic configurations. Are these overlaps only computational patterns or are also implemented in the brain? To explore whether overlaps are neurophysiologically detectable during sentence production, we conducted an EEG picture description experiment in Hindi. Focusing on event-related desynchronization in the alpha band, we demonstrate that overlaps are relevant for the neurocognition underlying sentence planning.
Sebastian Sauppe, Kamal K. Choudhary, Nathalie Giroud, Damián E. Blasi, Shikha Bhattamishra, Mahima Gulati, Aitor M. Egurtzegi, Ina Bornkessel-Schlesewsky, Martin Meyer and Balthasar Bickel

Asymmetry between production and comprehension at syntactic level: Evidence from Catalan-speaking children

Some studies indicate that children are able to produce morphological/grammatical forms before they are able to comprehend them (e.g., Legendre, et al 2014). Evidence from 4 studies with native Catalan-speaking children aged between 4 to 10 years is presented. Results indicate that children are able to produce and comprehend active clauses. They are able to comprehend passive clause by the age of 8-9 years, though they do not produce them. In contrast they produce object-dislocated clauses by the age of 4 years, but are not able to comprehend them until around the age of 10 years.
Merce Prat-Sala and Ulrike Hahn

Predicting across the lifespan: Evidence from the visual world paradigm

The current research used the visual world paradigm to investigate age-related differences in prediction stemming from verb selectional restrictions (Altmann & Kamide, 1999; Experiment 1) and real world knowledge (Kamide et al., 2003; Experiment 2). While prediction was observed in both, there were no significant age effects. In contrast to the ERP literature (e.g., see Federmeier, 2007), the current results reveal that predictive eye movements are strikingly stable across adulthood.
Yuki Kamide and Anuenue Kukona

Effects of prepositional phrase type on the processing of relative clauses

In an eyetracking experiment, participants read sentences containing object RCs and subject RCs in which an intervening prepositional phrase could be locative or temporal. In the locative condition, reading times at the matrix verb were significantly longer for ORCs than SRCs, but in the temporal condition reading times tended to be longer for SRCs than ORCs. The results highlight the importance of considering how intervening material in complex sentences affects both the structure and the meaning of the sentence. 
Matthew Lowder and Peter Gordon

Segmental duration as a cue to sentence structure

In order to parse speech in real time, listeners should use any informative cues available. Here we investigate the role of segmental duration. While previous work has found statistically significant differences in the mean durations of analogous segments across different lexical/syntactic structures, the goal of this work is to use production data to quantify how informative segmental duration is about syntactic/lexical structure. After implementing an ideal listener model in a Bayesian classifier, our results indicate there is indeed sufficient information contained in individual token durations so as to be useful in real-time sentence processing.
Sten Knutsen, Karin Stromswold and Dave Kleinschmidt

The time-course of rhythmic and syntactic factors in silent reading in Turkish

The study investigates syntactic and rhythmic factors in silent reading in Turkish via eye-tracking methodology. Experimental sentences had four versions manipulating phrase lengths (balanced, unbalanced) and syntactic structure (LC, EC). The analyses revealed an LC advantage in the disambiguating region and a preference for balanced lengths in the spillover region. The results support Fodor’s IPH and balanced-sisters constraint in projecting prosodic boundaries in silent reading. But as was predicted by Fodor (2002), prosodic factors (although used online) lag behind syntactic factors.
Nazik Dinçtopal Deniz

Context Effects in Irony Processing

Irony is a complex conversational phenomenon of which the successful interpretation is based on a number of linguistic and pragmatic cues. Research so far has produced mixed results on the effects of context on irony comprehension. We present evidence from two visual world eye-tracking experiments for the presence of context effects in early processing of ironic utterances. In our experiments we utilized echoic mention to manipulate context strength and found that while context did not have an effect on reference building it had a significant effect on the integration of ironic words into a previously established mental situation model.
Saskia Leymann, Verena Haser and Lars Konieczny

Effects of predictability and optionality on pronominalization

Predictability has been found to influence production at multiple levels of linguistic structure, but evidence for the influence of predictability on reduction in speakers’ choice of reference is mixed. This study tests the effect of the optional-vs-obligatory status of competitor referents on the pronominalization rate of subject referents, a potential confound from previous studies. Initial results suggest that pronominalization rates decrease when competitor referents are obligatory arguments compared to contexts with optional competitor referents, but this result failed to replicate in two follow-up experiments. As such, our study finds no evidence that the optional-vs-obligatory status of referents affects pronominalization.
Jet Hoek and Hannah Rohde

Ordering in numerals across languages supports rapid information processing

"One previously unexplained observation about numeral systems is for numerals greater than 20 to have the larger constituent number expressed before the smaller constituent. Systems that adopt the reverse ordering tend to switch order over time. To explore these phenomena, we propose the view of Rapid Information Gain, contrasting the established theory of Uniform Information Density. We find that RIG accounts for empirical patterns better than UID, suggesting an emphasis on information front-loading as opposed to smoothing in compound numerals. This shows that fine-grained generalizations about lexical compounds can be understood in information-theoretic terms in the case of numerals."
Emmy Liu and Yang Xu

Not all passives are processed equal: verb voice and word order in idiom comprehension

The processing of Italian idioms is observed in two passive constructions, i.e. with a preverbal (Passive I) and a postverbal subject (Passive II). Crucially, Passive II preserves the verb-noun order of the canonical active form. Passive II is rated as more natural and is more often completed as idiomatic in two offline questionnaires. Eye-tracking data confirms it to be read faster than Passive I, while Passive I reading is sped up by high familiarity and semantic transparency. Results thus suggest that the core issue in processing passive idioms is the violation of the canonical constituent order of the active form.
Marco Silvio Giuseppe Senaldi, Paolo Canal and Alessandro Lenci

Priming discourse structure guides pronoun comprehension

We tested the tendency to follow the subject-assignment strategy for ambiguous pronouns using 3 prime types—unambiguous 3rd person and I/you pronouns and names. We also measured individual differences (Author Recognition Task (ART)). Results showed that people were more likely to select the subject character as the referent for the ambiguous pronouns in the subject-prime than non-subject prime conditions, but most strongly when primed with unambiguous third-person pronouns, which included a main effect of ART. Name primes showed a smaller priming effect only occurring for people with high ART scores. There was no priming or ART effect with I/you pronouns.
Elyce Williams and Jennifer Arnold

Understanding over-specification: A visual-world ERP study

In a visual-world ERP experiment, we assessed the influence of specificity and reduction of ambiguity about the target (entropy) on referential processing. Participants saw visual scenes while listening to instructions that were either minimally-specified (MS) or over-specified (OS), depending on whether the target belonged to a contrast pair or was singleton. Entropy reduction was higher (HR) vs lower (LR) depending on the number of competitors. The noun showed a biphasic N400-P600 effect for OS vs MS, suggesting that listeners interpret the adjective contrastively, and an N400 effect for LR vs HR, indicating a cost for identifying more ambiguous referents.
Elli Tourtouri, Francesca Delogu and Matthew Crocker

Pupillometry reveals reduced effort in children’s sentence processing when visual speech cues are available

Processing speech can be effortful for children, particularly in classroom settings. We investigated whether the presence of visual speech cues (the speaker’s facial movements) would reduce processing effort for 7-11-year-old children, as such cues are beneficial for adults’ processing. We measured pupil dilation, an index of cognitive effort, while children completed a phoneme monitoring task with audiovisual and auditory-only stimuli in either quiet or noise. Results indicated that effort was reduced when processing audiovisual speech compared to auditory-only in both quiet and noisy conditions, suggesting that visual speech cues may facilitate language processing in the classroom and elsewhere.
Rebecca Holt, Laurence Bruggeman and Katherine Demuth

Processing subject/object asymmetries in German: case-marking and intervention

Object-gap dependencies are more difficult to process than subject-gap dependencies because of intervention. However, it is unknown if any cognitively or perceptually salient features (broad definition) or only very specific, syntactic features (narrow definition) can cause intervention. We tested this by comparing dative to accusative object extraction in a self-paced reading paradigm. The results do not provide any evidence for the broad definition of intervention, but instead suggest intervention is only caused by very specific, movement attracting features. If any, the results suggest that dative conditions are more difficult than accusative conditions.
Ankelien Schippers, Margreet Vogelzang and Esther Ruigendijk

Lexical Boost from the Subject Noun: The Influence of Task

To test the prediction that the lexical boost is driven by explicit memory (e.g., Chang et al, 2006), we tested for a subject noun boost, using prepositional object or double object structures, when participants could look back to a prime sentence while completing a target sentence (Experiment 1) or not (Experiment 2). We report a subject noun boost but only when participants could see the prime, suggesting that when available, the prime acted as a cue to boost the activation of its structure. These results indicate that the subject noun boost is affected by how explicit the word repetition is.
Laura Wakeford, Leila Kantola and Roger van Gompel

Ambiguity processing in English natives and non-natives

The similarities and differences between L1 and L2 sentence processing have been strongly debated regarding whether L2 speakers have difficulty using syntactic information. The current results from our offline and online experiments examining ambiguity processing show that L1 and L2 speakers with an L1 that prefers high attachment both preferred attaching low over high. The offline attachment preferences were modulated by relative clause postion and working memory. The findings suggest that the same parsing principle was employed in L1 and L2 processing to minimise cognitive load. Further analysis on individual differences will also be discussed.
Yesi Cheng, Ian Cunnings and Jason Rothman

Inflexible structural priming in flexible word order language

Verb semantic similarity was shown to facilitate structural repetition in English (Yi and Koenig 2016). The present study investigated whether verb meaning also modulates structural priming in a head-final and flexible word order language, i.e. a verb occurs at the end of a sentence and preverbal phrasal order is flexible. We conducted two structural priming experiments in Korean, using both the picture-description and sentence-recall paradigms. We found purely syntactic structural priming effect was robust in Korean while prior processing of (dis-)similar verb meaning does not modulate speakers’ upcoming structural choice.
Eunkyung Yi and Hongoak Yun

ERPs and artificial mini-grammars in third language transfer/learning

The role of prior linguistic knowledge is a central question in L3/Ln acquisition, where more than one previous grammar may impact learning trajectories from first exposure to the L3. Two groups of L1 Spanish-L2 English speakers were exposed to one of two artificial languages (ALs) lexically based on Spanish and English, respectively, which display both number and gender agreement between nouns, determiners and adjectives. Gender and number violations in both languages elicit broadly distributed P600 effects in an ERP experiment. This may reflect transfer from Spanish in both cases (at least for gender), or rapid acquisition of the properties anew.
Jorge González Alonso, José Alemán Bañón, Vincent DeLuca, David Miller, Sergio Miguel Pereira Soares, Eloi Puig Mayenco, Sophie Slaats and Jason Rothman

Focus scope marked by “only” influences syntactic attachment

In addition to known effects of prosodic boundaries on attachment, evidence has been mounting that accents also influence attachment (Schafer et al. 1996, Lee & Watson 2011, Carlson & Tyler 2018). But why do accents on the head of an attachment site draw the attachment of an ambiguously-attached phrase: does the accent make an attachment site more salient? Or is it that the accent indicates focus on a phrase, which makes it important to the sentence and draws attachment? We present evidence from studies with "only" supporting a focus-based theory in which the position where focus scopes actually draws attachment.
David Potter and Katy Carlson

ZWgen: A Python-based Chinese Lexical Stimuli Generator

Recent research has developed various Chinese lexicon databases focusing on different aspects. An integrated database that incorporates all of these data is lacking at this moment. Combining features of multiple databases, this study provides an attempt to automatically generate Chinese lexical stimuli. ZWgen — the python-based generator consists of a variety of lexical features that focus on orthographic, phonological, morphological and syntactic aspect, including but not limited to character stroke number, orthographic neighbourhood size, homophone density and frequency of homophones. By manipulating these lexical features, ZWgen will automatically return lexical items that meet the selection criteria.
Mingyu Yuan

(In-)definites, (anti-)uniqueness, and uniqueness expectations

Using "A" in noun phrases such as "A father of the victim" is odd, which is commonly explained by the principle "Maximize Presupposition", requiring speakers to use the alternative with the strongest presupposition (here "the", given its uniqueness presupposition). This results in an anti-uniqueness inference for "A" (clashing with stereotypical expectations in the example at hand), sometimes labelled as an 'anti-presupposition' (Percus 2006), as it derives from reasoning over the presuppositions of alternative forms. We compare these inferences to the uniqueness inferences associated with definites, while manipulating uniqueness expectations in a picture manipulation task using visual world eye-tracking.
Nadine Bade and Florian Schwarz

Personae in syntactic processing: Socially-specified agents bias expectations of verb transitivity

In this study we investigated the role of personae (speakers' "social types") in syntactic ambiguity resolution. In Experiment 1, we show that personae bias the interpretation of ambiguously transitive verbs in a sentence completion task. Experiment 2 shows that these biases also affect the perceived acceptability of the same verbs when paired with or without a direct object. Experiment 3 used these materials in an eye-tracking study which showed that transitively biased personae led readers to adopt transitive interpretations even when this possibility was ruled out by punctuation. These findings suggest a strong role for personae in sentence comprehension.
June Choe, Shayne Sloggett, Masaya Yoshida and Annette D'Onofrio

Online processing of an elided r-expression

This study employs an eye tracking experiment to investigate the processing of verb phrase ellipsis, focusing on whether the parser computes vehicle change online. A backward VPE construction is used to check the accessibility of a name within an elided VP, which is measured using a gender mismatch effect. The observation of a GMME in first pass reading times provides evidence that vehicle change has occurred. This result suggests that new material is introduced during the resolution of ellipsis, namely that the R-expression becomes pronominal during online sentence processing.
Kathleen Hall and Masaya Yoshida

Grammatical factors in morphological processing: evidence from allomorphy

Language processing is often analysed in terms of two sets of factors: quantitative-distributional and grammatical, with schools of thought differing in the postulated relation between the two. Our study investigates whether sensitivity to allomorphy - a grammatical factor - has an impact on lexical processing. Mixed effects regression models were fit to participant data from the BLP corpus with the independent variables being frequency, orthographic length, orthographic neighbourhood and inflectional entropy. A variable coding for sensitivity to allomorphy among stems was introduced as hasallos and showed a significant effect on both reaction times and accuracy scores (modelled separately).
Daniil Bondarenko, Onur Özsoy and Itamar Kastner

Cost of ungrammatical predictions during online sentence processing: evidence against surprisal

Surprisal has been successful in explaining comprehension difficulty in a number of online studies. While surprisal assumes predicted parses to be grammatical, such predictions can also be ungrammatical (Apurva & Husain, 2018). Therefore, similar to reranking cost due to incorrect (grammatical) predictions, there should be a cost for ungrammatical predictions as well. A cloze study was conducted to select two conditions that differed in the no. of ungrammatical predictions while being similar in their most frequent grammatical prediction. A self-paced reading experiment for this manipulation shows a cost of ungrammatical predictions, thereby highlighting the limitation of the surprisal metric.
Apurva Apurva and Samar Husain

Different effects of grammatical violations in Italian: an EEG study on verb-particle constructions

Using event-related brain potentials, we examined the comprehension of sentences with different types of Verb-Particle Constructions (VPCs). 44 participants read sentences containing aspectual VPCs (e.g., raschiare via “*to scrape away”) that have compositional meaning and sentences including idiomatic VPCs (e.g., fare fuori “to kill”) that have non-compositional meaning. Half of the sentences were correct and half were violated, containing a non-existing VPC. Results show violation effects (N400) in different time-windows depending on the type of VPCs, suggesting that the compositional nature of the aspectual VPCs relative to the idiomatic ones plays a crucial role in the processing of these sentences.
Federica Mantione, Patrizia Cordin and Francesco Vespignani

Multimodal prediction in text + emoji sentences: an ERP study

This ERP study investigates cross-modal lexical prediction of emojis in sentence contexts. Participants read highly and lowly constraining sentences in which a corresponding emoji was substituted for the final word (Experiment 1) or the word appeared in its written form (Experiment 2). Results indicate that participants were capable of lexical prediction of emojis in a manner qualitatively similar to that of words; follow-up analyses indicate participants’ anticipations were more lexical than conceptual or visual. This work contributes both to the nascent program of research on the language processing of emojis and to continuing research on prediction processes during language comprehension.
Benjamin Weissman, Neil Cohn and Darren Tanner

The Effect of Eventive Verbs on Children's Online Argument Representation

The current study examines 32- to 36-month-old English learners’ representation of verb arguments. In particular, we used eventive (e.g., “destroy”, “spill”) and stative verbs (e.g., “love”, ”own”) to examine whether children simply access the semantic prototypes of words or whether their representations go beyond a word’s linguistic identity and include context-specific properties (e.g., a broken rather than intact bottle upon hearing “Susan dropped the”). The results of our eye-tracking study tentatively show that toddlers possess basic verb type knowledge that they rapidly integrate in their argument representation during language comprehension. This suggests that their real-time language processing is relatively fine-grained.
Boyang Qin and Marieke van Heugten

Language dynamics across the life span: The comprehension of verbal jokes

Age-related differences in joke comprehension were studied using a paradigm that separates cognitive from affective aspects. Most verbal jokes are based on an incongruency whose revision puts demands on working memory and cognitive flexibility. Fifty older participants (70 -92 years of age) - compared to a middle aged control group - needed longer to read jokes and non-funny revision stories, they made more errors on revision stories, and their funniness ratings showed less differentiation between jokes and revision stories. Thus, despite preserved pragmatic language skills, humor comprehension in older readers seems affected by a decline of executive functions.
Evelyn Ferstl

Sustained negativities for wh-movement may not extend to other types of syntactic prediction.

We evaluated whether an ERP response (sustained anterior negativities, SAN) indexes online maintenance of a memory representation of predictively generated structures until they are confirmed by upcoming input. We assessed whether SANs for wh-dependencies extend to sentences requiring prediction of a subordinate clause (subordinating adverbials). Across 3 experiments we replicated SANs for wh-dependencies, failed to observe SANs for subordinating adverbials, and replicated this dissociation using a within-subjects design. We concluded that SANs for wh-dependencies do not necessarily extend to other constructions requiring online maintenance of syntactic predictions, suggesting that ERP correlates of structural prediction per se have not been identified.
Aura Cruz Heredia, Bethany Dickerson and Ellen Lau

Turkish “unless” is not biconditional unless the pragmatic context allows it

Ebru Evcen, Umut Özge and Duygu Özge

The Impact of Stereotypes and Noun Endings on Processing Gender in English: Comparing Native and Non-Native Performance

"This study aimed at establishing whether the perception of the noun ending “-er” as masculine carries over into English for native speakers of German. Since German is a language with a grammatical gender system, professions which end in “-er” identify a person as male in German (e.g. der Gärtner – “the (male) gardener”) while in English, they are not grammatically marked for gender. Eye-tracking data from 64 participants suggests that “-er” may slightly, but “-or” more considerably, slow down processing for German speakers when used with “women”, indicating transfer effects. Beyond that, stereotypical associations significantly influenced reading behaviour."
Julia Müller, Verena Haser and Lars Konieczny

Alternative constructions and scope ambiguities: What counts as an alternative?

Hemforth & Konieczny (2019) suggest that the existence of an accessible alternative such as “Not all my friends went to the movies.” plays a role in cross-linguistic differences in the interpretation of “all...not” constructions, such as “All my friends didn’t go to the movies.”, explaining the preference for linear scope in English and German where the “not … all” alternative exists but inverse scope in French where it does not. We extend the picture by investigating another possible alternative construction “Mes amis ne sont pas tous allés voir le film.”/ “My friends did not all go to see the movie.”
Barbara Hemforth and Lars Konieczny

Facilitation vs. inhibition as mechanisms for syntactic constraints on word recognition

We investigated whether syntactic category expectations constrain auditory word recognition by preventing competition from items that don’t fit the context, or by facilitating items that do fit the context. In a visual world experiment whose critical trials did not contain targets, we tested for increased fixations to a cohort competitor that could only be a noun, during noun and verb contexts. We found that the wrong-category lexical candidate does demonstrate phonological competition, ruling out complete inhibition as the mechanism for the syntactic category constraint.
Phoebe Gaston, Ellen Lau and Colin Phillips

Turn-taking differs between parents and therapists speaking to children with ASD

Children with Autism Spectrum Disorder (ASD) often lag behind typically-developing (TD) peers in communication skills. We evaluate differences in conversational turn-taking using transcribed LENA recordings of parent-ASD child interactions, therapist-ASD child interactions and parent-TD control pairs. We find that ASD children respond less frequently and have trouble with conversational coordination. Therapists speaking to these children emphasize their full sentences using repeated non-sentential utterances, especially discourse markers, filled pauses and exclamations, to retain the child’s attention. Therapist speech patterns do not approximate TD parental speech but are adapted to meet the attentional needs of children with ASD.
Micha Elsner, Benjamin Allen, Elizabeth Kryszak and Kiwako Ito

Same sized sisters: Relative clause attachment to conjoined NPs

RC-Attachment is ambiguous in sentences like “I am interested in novels and films that talk about love.” in that the relative clause can attach to the conjoined NP novels and films (high attachment) or to the local NP films (low attachment). Incremental processing predicts a preference for attaching the relative clause to the conjoined NP. Fodor’s SSS, however, predicts that this preference can be modified by the length of the constituents. Two acceptability studies in French and English show that the acceptability of local attachment increases with the length of the first conjoint (e.g. “19th century foreign novels”).
Antoine Hedier, Peijia Su, Fahima Maouche and Barbara Hemforth

The agent preference in sentence planning is modulated by case marking: Eye tracking evidence from Hindi, Basque and Swiss German

The agent preference is a well-established cognition principle instantiated in sentence comprehension and gesture production, i.a. We present two eye-tracking picture-description studies exploring the agent preference in sentence planning. Study 1 focused on a language-internal contrast between case-marked and unmarked agents in Hindi, conditioned by aspect; Study 2 compared planning of sentences with unconditionally marked agents in Basque to unmarked agents in German. In all three languages speakers allocated most visual attention to agents (during relational encoding). This preference is modulated in Hindi, however, by speakers' need to decide between planning marked and unmarked agents.
Aitor Egurtzegi, Sebastian Sauppe, Damian Blasi, Kamal Choudhary, Yingqi Jing, Nathalie Giroud, Shikha Bhattamishra, Mahima Gulati, Bornkessel-Schlesewsky Ina, Itziar Laka, Martin Meyer and Balthasar Bickel

Maze Made Easy: Better and easier measurement of incremental processing difficulty

The Maze task has shown promise as a way to measure incremental processing difficulty with high sensitivity and accuracy. We demonstrate that the Maze task runs reliably over the web and is substantially more sensitive than SPR on Mechanical Turk. Furthermore, we demonstrate and validate a method for automatically generating materials which dramatically reduces the preparation required while yielding the same sensitivity and accuracy. The resulting “Auto-Maze” task provides all the advantages of Maze while being as easy to prepare and run as SPR. We make our code freely available online at
Veronica Boyce, Richard Futrell and Roger Levy

EEG signatures of perceptual reversals of bistable visual and linguistic stimuli

The Reversal Negativity is an event-related potential elicited when one’s subjective perception of a bistable figure (e.g. the Necker Cube) switches from one of its possible configurations to the other. We investigated whether a similar reversal signature occurs in the linguistic domain when one’s interpretation of a “bistable” ambiguous sentence, such as “The chicken is ready to eat,” is reversed. We identified a frontal negativity occurring over a similar time course as the visual RN for reversals of bistable sentences, suggesting that similar reversal processes may occur in response to “conceptual” ambiguities, such as those present in language comprehension.
Kevin Ortego, Michael Pitts and Enriqueta Canseco-Gonzalez


In order to untangle the effect of bilingualism from the effect of weaker proficiency, we examined the effect the frequency of daily code-switching and proficiency in the non-dominant language (L2 French) on cognitive control in the Simon task. An interaction between proficiency and the frequency of code-switching showed that the high proficient bilinguals who were also frequently code-switching performed faster over time in the incongruent trials and consequently had the smallest Simon effect. The interaction between proficiency and the frequency of code-switching emphasizes the role of code-switching in exercising cognitive control and monitoring abilities in bilingual speakers.
Souad Kheder

Verbs retrieve subjects, not clausal attachment sites

Similarity-based interference can cause difficulty in processing subject-verb dependencies: when an intervening subject matches the retrieval cues of the matrix verb, the parser may erroneously retrieve the intervener. Previous studies have identified the relevant dimensions of similarity for subjects, but none have explicitly attempted to disentangle the contribution of tensed clauses from their subjects. These clauses could lead to interference if the parser retrieves the embedded clause as the attachment site for the matrix verb. In two experiments, we replicate previous findings of subject interference but fail to find evidence for interference from embedded clauses.
Netta Ben-Meir, Nick Van Handel and Matt Wagers

Computing object agreement in Georgian is easier than computing subject agreement

While subject–verb agreement is well-studied, little is known about the processing of object–verb agreement. This study investigates Georgian, a language with both flavors. In a speeded acceptability judgement experiment, participants detected errors in various kinds of agreement morphology. A Signal Detection Theory analysis reveals that participants are (i) more sensitive to agreement errors on verbs with canonical morphosyntactic properties than to errors on verbs with non-canonical properties, and (ii) more sensitive to errors in object agreement than to errors in subject agreement. Both results, we propose, follow from differing levels of certainty about the form of the sentence-final verb.
Steven Foley and Matt Wagers

Evidence for integration of noisy linguistic evidence and prior expectations depends on the task

We investigated the noisy channel model which posits a rational integration of plausibility with “noisy” input during parsing/interpretation of sentences. Experiment 1 assessed participants’ interpretations via yes/no questions. In experiment 2, to evaluate the possibility that integration of plausibility occurs when processing the question rather than during the initial parse/interpretation, participants typed repetitions of sentences. Evidence for a predicted interaction between syntax and plausibility was convincing in experiment 1, but not in experiment 2. These results suggest that reliance on plausibility is more likely when an implausible parse is highlighted, e.g. when answering a question.
Eli Kane and L. Robert Slevc

Fixation measures as a function of comprehension accuracy

This study explores eye fixations in sentences in which subjects provided correct and incorrect answers to questions about aspects of sentences meaning directly related to parsing operations postulated to be triggered by specific words. We found different patterns of fixations as a function of the accuracy of the response. We discuss the implication of this result and broader issues pertaining to the determinants of eye fixations during reading.
Maria Varkanitsa and David Caplan

Pragmatic inferences are modulated by informativity across cultures

Listeners make sophisticated pragmatic inferences during real-time comprehension. When they hear “Hand me the big…,” they anticipate the referent to be of a pair that differ in size (e.g., big cup, small cup). Here, we study pragmatic inference in the Tsimane’, an indigenous people of the Bolivian lowlands. We find that 1) Tsimane’ speakers made anticipatory looks consistent with contrastive inference while listening to size-modified references, pointing to a universal expectation that speakers typically choose their utterances to be optimally informative; and 2) the contrast effect was not significantly smaller for color modifiers relative to size (with one exception).
Rachel Ryskin, Miguel Salinas, Steven Piantadosi, Paula Rubio-Fernández and Edward Gibson

How do repeated result states fare in sentence comprehension and production?

We test how discourse level information and verbs’ lexical semantics guide expectations/predictions in comprehension and production. Specifically, we investigate the processing and production of result states after change-of-state events. These events can be described with verbs that do or do not semantically encode result states (result verbs vs. manner verbs). We find that both the comprehension and production systems have a stronger preference for a result-related linguistic expression after encountering a result verb than a manner verb. We also find that although comprehension and production may be closely related, dispreference against repetition is reflected more in production than in comprehension.
Sarah Hye-yeon Lee and Elsi Kaiser

Semantic predictability of implicit causes affects referential form choice

It is debated whether speakers are more likely to pronominalize predictable referents. With transfer events, speakers tend to pronominalize predictable goals more than sources. Thus far, the same has not been observed for predictable implicit causes versus non-causes in emotion events. This may be due to the passage continuation method used. In a paradigm more similar to natural language (i.e. with greater contextual support, where participants already knew the content of their contribution), speakers did pronominalize implicit causes more than non-causes. We hypothesize this paradigm provided greater discourse context and allowed speakers to plan responses earlier, incorporating coherence relation knowledge.
Kathryn Weatherford and Jennifer Arnold

Word Skipping in Deaf and Hearing Bilinguals: Cognitive control remains with increased perceptual span

The word-processing efficiency hypothesis (Belanger & Rayner, 2015; Belanger et al. 2012, 2013) posits that deaf readers have larger perceptual spans than comparably skilled hearing readers. We used eye-tracking to compare deaf readers with comparably skilled Native-English speakers and Chinese-English bilinguals. Between groups, deaf readers skipped words more often and had shorter first pass and total reading times, but comprehended sentences as well as the other groups. These results are broadly consistent with the word processing efficiency hypothesis.
Matt Traxler, Pilar Pinar, David Corina, Kurt Winsler, Liv Hoversten and Trevor Brothers

Syntactic Structure aids Learning of Grammatical Dependencies in Neural Networks

We assess the grammatical competence of recurrent neural networks (RNNs) by treating them as human subjects in psycholinguistics experiments, using targeted sentence stimuli to elicit evidence for latent syntactic representation. We compare two RNN language models, one that computes explicit parse trees and one that does not. We evaluate how well these models capture NPI licensing dependencies, filler-gap dependencies, and island constraints. We find that explicit representation of syntax aids the learning of structurally-adjacent dependencies, but all models have difficulty threading word expectations through embedded clauses in filler-gap dependencies, a key difference between sentence processing by humans and RNNs.
Ethan Wilcox, Peng Qian, Richard Futrell, Miguel Ballesteros and Roger Levy

Inferring sentence comprehension from eye movements in reading

How much information can be obtained about the cognitive state of a specific reader from their eye movements over an individual sentence? In this work we address this question by examining the extent to which logistic regression and Convolutional Neural Network (CNN) models can predict whether a reader will answer a comprehension question correctly based on their eye movements while reading a single sentence. We evaluate how well these models generalize to new readers and new sentences.
Yunyan Duan, Yevgeni Berzak, Klinton Bicknell and Roger Levy

Studying Morphological Computation and Storage via Lexical Decision Data

In lexical decision tasks, reaction time to certain stimuli has historically been used to infer properties of morphological processing. Reaction time has been shown to depend on a number of measures, with frequency being among the most discussed. We review modeling approaches of Lignos & Gorman, amongst others, and find that despite updated methodology, their modeling procedure is still insufficient to accurately account for the numerous autocorrelated measures of frequency in morphologically complex forms that are central to most analyses.
Gregory Theos and Timothy O'Donnell

How do readers adapt to unfamiliar syntax?: Evidence from needs+past participle

Comprehenders rapidly adapt to unfamiliar syntactic constructions. We investigate the underlying mechanisms using self-paced reading. Participants unfamiliar with the dialectal needs+past participle construction encountered either it or the conventional needs+present participle in an initial exposure phase. In a second phase, we tested the consequences of that exposure for processing other structures. Exposure to needs+past participle did not impair the conventional structure, providing evidence against probabilistic prediction (Expt. 1). But, it did impair dissimilar dialectal structures, indicating comprehenders were not just accommodating any anomalous input (Expt. 2). These results suggest comprehenders prepare for certain classes of input without specifically predicting them.
Rachel Peters and Scott Fraundorf

Context effects in the interpretation of Haddock descriptions

The interpretation of definite descriptions has been claimed to involve pragmatic context accommodation. This mechanism ensures that reference resolution is successful, even when the maximal context violates the uniqueness presupposition of the definite article. In two reference-resolution studies involving Haddock Descriptions, we show that context accommodation is less felicitous when the embedded description refers to different objects in the maximal and the restricted context, suggesting that listeners consider maximal contexts even when they are ruled out by independent linguistic constraints. An RSA model is proposed that derives the observed effects from pragmatic context coordination and uncertainty about possible adjectival thresholds.
Helena Aparicio and Elizabeth Coppock

Antecedent search for Turkish reflexive kendi: Evidence from eye tracking during reading

Previous work on reflexives reports mixed views on whether the parser implements an intelligent search. The current study addresses this with a novel reflexive form, Turkish kendi ‘self’, using object relative clauses, where candidate antecedents were manipulated by animacy. The objective was to find if processing difficulty occurred when binding accessible antecedent was inanimate. Results showed that participants had longer reading times in first pass time when accessible antecedent was inanimate. However, binding incompatible candidate antecedent influenced processing in later stages; longer fixations were made in accessible match and inaccessible match conditions than in accessible match and inaccessible mismatch conditions.
Hasan Sezer

Different Types and Qualities of Fillers: Maintenance and Retrieval

We uncover the mechanisms working behind both maintenance and retrieval components by testing what aspects of a filler are retrieved in different WhFGD (e.g. wh-filler gap dependency) constructions: "reactivated" WhFGD formation (the filler that is linked to the verb once, and is reactivated later) and "active" WhFGD formation. Information associated with the active wh-filler was well-preserved compared to the reactivated wh-filler that was linked to the gap once, released from memory and later reactivated. We argue for the position that posits both maintenance and retrieval playing a role in the processing of whFGD.
Nayoun Kim, Laurel Brehm, Patrick Sturt and Masaya Yoshida

Emotion Regulation through Irony - Evidence from Behavior and ERPs

This study investigates the effects of ironic and literal language on the perceived mental state of a speaker in either high or low emotional situations. Results from a web-based rating study show that a speaker using ironic language is perceived as being in a less negative mental state (higher in valence and lower in arousal) compared to the same speaker using literal language. In an ERP reading task, irony elicited a larger N400 effect. Furthermore, results showed enhanced P200 effects for literality in mildly negative situations, suggesting that irony is perceived differently depending on how emotional the context is.
Valeria Pfeifer and Vicky Tzuyin Lai

Grammatical and pragmatic cues guide temporal interpretation in discourse

What factors modulate the on-line comprehension of intersentential temporal relations, and how strongly? Do baseline comprehension costs vary between relations? Three self-paced reading studies compared progression (S1 event precedes S2 event) and backshift (S1 event follows S2 event), and the effects pragmatic and grammatical factors had on their comprehension. Results from all studies suggest that baseline comprehension costs of progression and backshift are not very different, if at all. Study 2 results suggest that pragmatic plausibility cues can be leveraged fairly quickly and strongly. Study 3 results suggest that aspect marker 'had' may not be leveraged very strongly.
Kelsey Sasaki

Investigating the real-world costs of intra-sentential code-switching for lexical processes

Bilinguals often engage in code-switching or switching between languages during a conversation. Prior EEG studies have found that code-switched words are harder to process; however, these tasks primarily use isolated words or single sentence contexts. We reasoned that code-switching might be facilitated by top-down expectations developed throughout a spoken discourse. In this study, we use new naturalistic EEG techniques to see how code-switches are perceived in rich contexts and how they compare to other lexical switches. We find evidence that code-switched words are initially difficult to process but are rapidly facilitated if the word makes sense in the given context.
Anthony Yacovone, Emily Moya and Jesse Snedeker

Gender mismatch and possession type effects on interpretation of VP ellipsis

"Mark walked his dog, and Larry did, too." has two possible interpretations: Larry walked Mark's dog (coreference) or Larry walked his own dog (variable binding). What influences resolution of these ambiguous constructions? We build on previous work which found that different possession types (e.g. kinship: "her mother"; ownership: "her bicycle") modulate interpretational preference in such sentences. We report four experiments testing how the effect of possession type interacts with grammatical gender and quantified nouns (e.g. "every man"), which have been claimed to influence discourse processing. Our results support a discourse-based account of how this type of ambiguity is resolved.
Jesse Storbeck and Elsi Kaiser

Contrastive informativity strength drives real-time language variation and change: an online study of the Spanish habitual in three dialectal varieties

An SPR task in three Spanish dialects shows that, with respect to the expression of the habitual meaning, besides the use of the Simple Present, Argentinian and Iberian speakers allow Present Progressive-marking when contextual information satisfies the presuppositional demands of estar, the auxiliary in the verbal periphrasis. Contrastively, Mexican speakers do not need this contextual support. This pattern is consistent with a generalization process already underway in the three varieties, with the Mexican variety appearing further along the grammaticalization path. This process of change is driven by the contrastive informativity strength of the lexico-semantic properties of the Present Progressive marker.
Martín Fuchs and María Mercedes Piñango

Plausibility in Chinese concession and causality: Evidence from self-paced reading and eye-tracking

In two experiments, we investigate the effect of plausibility on the processing of concessive and causal relations. In a self-paced reading experiment, we found the effect of plausibility at both the critical and post-critical regions in causality but only at the post-critical region in concession, showing a delayed effect in concessive relation. In an eye-tracking study, participants regressed out more at the critical regions of concession, but no such effect was found in causality, showing reprocessing of previous text and greater difficulty in concession. Together, this study suggests more complex processing and greater cognitive cost in processing concession than causality.
Siqi Lyu, Jung-Yueh Tu and Charles Lin

Prosody-meaning mismatches in PP ambiguity: Incremental processing with pupillometry

We report the results of a pupillometry experiment on the prosodic disambiguation of PP modifier sentences. We manipulated instrument plausibility ("pen" vs. "gun") and prosodic boundary placement, so that a prosodic boundary biased towards a Modifier [A] or Instrument [B] construal of the prepositional phrase ("The artist sketched [A] the man [B] with the pen/gun"). A growth curve analysis of change in pupil size suggests that listeners attend to prosodic boundary locations that disambiguate the structure, and that an online processing penalty, as indexed by increased pupil dilation, is generated when prosodic boundary information leads to an implausible construal.
Jesse Harris, Chie Nakamura, Bethany Sturman and Sun-Ah Jun

Poster Session D Sunday (Day 3) 12:05 - 2:00 PM


  • East & West End Rooms: posters #D1-D34
  • The Ballroom: posters #D35-D53
  • Pre-Function Area: posters #D54-D69 

No scope for planning -- latent mixture modelling of sentence onsets

Whether or not pre-planning a sentence involves syntax hinges, partially, on the observation that, when a sentence starts with a conjoined NP, the speech onset is delayed compared to matched sentences starting in simple NPs. An alternative view is that planning beyond the subject NP is not obligated by the linguistic encoder. We contrasted these two views directly in a series of Bayesian models. Results showed that the delay for conjoined NPs happens sporadically rather systematically. We suggest that the frequently replicated slowdown for conjoined NPs might be better explained by non-syntactic processing demands.
Jens Roeser, Mark Andrews, Mark Torrance and Thom Baguley

Speakers do prime themselves

Error-driven learning syntactic priming studies (e.g. Chang, Dell, & Bock, 2006) make the prediction that speakers cannot prime themselves. Activation-based models based on corpus data (e.g. Reitter, Keller, & Moore, 2011) predict otherwise. We show the first experimental evidence that speakers do prime themselves using a novel autocorrelation method, in addition to demonstrating the inverse frequency effect in comprehension-to-production priming.
Cassandra L. Jacobs, Duane Watson and Sun-Joo Cho

Comparing licensing processes at the interfaces of syntax, semantics and pragmatics

We compare between two licensing processes at the interface of syntax, semantics and pragmatics in German sentence comprehension. The first is the licensing of negative polarity items (NPIs), the second is the licensing of questions-sensitive discourse particles (QDiPs). We show rating, self-paced and ERP data of different licensing phenomena. Offline studies show graded acceptability reductions for inaccessible compared to absent licensers. ERP measurements show increased N400 amplitudes for unlicensed NPIs, and increased P600 amplitudes for unlicensed QDiPs. We discuss possible underlying reasons for differences between both phenomena, and how they may reflect different possible subprocesses in licensing at the interfaces.
Anna Czypionka, Mariya Kharaman, Josef Bayer, Maribel Romero and Carsten Eulitz

Testing the role of phoneme order in lexical access using transposed-phoneme priming

We report on an auditory masked priming study testing for priming by nonce strings formed by transposing two of the target’s consonant phonemes. We found that English speakers were faster to judge the lexicality of CVCCVC words when primed by nonce strings formed by initial-transposition (e.g. [sɪbkət] priming biscuit [bɪskət]), medial-transposition (e.g. [bɪksət]), and final-transposition (e.g. [bɪstək]), but not by transposing the first and last phonemes (e.g. [tɪskəb]). These results challenge the role of the order of elements in the acoustic signal for lexical access, suggesting that listeners activate candidates which contain the same set of phonemes, regardless of order.
Jonathan Geary

fMRI reveals language-specific predictive coding during naturalistic sentence comprehension

What are the neuro-cognitive mechanisms supporting predictive language processing? In particular, to what extent does prediction during comprehension recruit language-specific mechanisms, and to what extent does it rely on general cognitive mechanisms supporting prediction across domains? We present evidence from a naturalistic fMRI study that a functionally identified language network shows robust effects of surprisal while a functionally identified domain-general multiple-demand network does not. This finding suggests that predictive coding for language is primarily carried out by domain-specific language circuits, rather than by domain-general executive control.
Cory Shain, Idan Blank, Marten van Schijndel, William Schuler and Evelina Fedorenko

Rising pitch accents are more sensitive to context than falling pitch accents

The ERP study aimed to investigate effects of expectation- and signal-driven processing of prosodic prominence. We investigated cases in which i) prior context generates an expectation for a particular prosodic realization and thus leads to correlates of expected or mismatching informational processing, and ii) pitch accents differ in their perceptual prominence. rERPs indicate discrete correlates of prosodic and contextual prominence: Unexpected accents yield an expectation error (N400); prominent rising accents consume attentional resources and engender updating processes (early positivity) which is also observed for contextual highlighting for exciting contexts (late positivity) independent of the prosodic realization of the target.
Christine Roehr, Stefan Baumann, Martine Grice and Petra B. Schumacher

Input evidence vs. linguistic knowledge: Establishing new lexical entries during reading

Denisa Bordag and Andreas Opitz

Similar discourse properties guide both topicality & referential predictability judgments

Theories suggest that pronouns are assigned to predictable and/or topical referents. Here we test whether predictability and topicality are related, and whether they account for biases to assign pronouns to either a) gazed-at characters, or b) the grammatical subject. Participants viewed animated videos of a story (Ana is cleaning up with Liz), where the narrator gazed at Ana, Liz, or a neutral spot at sentence offset. Participants were either asked about predictability (Who do you think will be mentioned?) or topicality (Who do you think is the main character?). Next-mention predictions followed both gaze and subjecthood; topicality only followed subjecthood.
Sandra Zerkle and Jennifer Arnold

On the necessity of hippocampus in lexical-semantic mapping in language processing

Two studies find no evidence for impaired lexical-semantic mapping in spoken language processing following bilateral hippocampal damage and dense amnesia, pointing to a lack of necessity for hippocampus in this process. How can we reconcile these findings given previous evidence for hippocampal contributions to semantic processing? Hippocampus may contribute to lifelong tuning of lexical-semantic knowledge, predicting graded effects, with close and frequent semantic relations (tested here) remaining intact. At least for common words and concepts, our findings show that hippocampus is not necessary for engaging the lexical-semantic network during the online processing of ordinary sentences.
Sarah Brown-Schmidt, Nathaniel Klooster, Sun Joo Cho, Nazbanou Nozari and Duff Melissa

Semantic effects on processing filler-gap dependencies into adjuncts and conjuncts

In two experiments, we show that plausibility and main predicate type affect extraction from adjuncts, but not conjuncts.
Annika Kohrt, Peter O'Neill, Trey Sorensen and Dustin Chacón

Using syntactic priming to investigate how recurrent neural networks represent syntax

Recent studies have demonstrated that recurrent neural networks (RNNs) trained to process sentences encode some syntactic information. However, what specific syntactic information these networks encode remains largely unknown. Drawing on ideas from cumulative syntactic priming, we propose a novel paradigm to investigate an RNN's internal grammar. In a proof of concept experiment, we applied this paradigm to investigate how an off-the-shelf RNN language model trained on English Wikipedia represents different kinds of relative clauses. We demonstrated that the RNN was sensitive to the structural similarity between reduced and unreduced relative clauses.
Grusha Prasad, Marten van Schijndel and Tal Linzen

‘Nonetheless’ can reverse predictions immediately: evidence from ERPs

Concessive connectives such as ‘Nonetheless’ indicate that the upcoming proposition will be at odds with the current discourse. It has been shown that comprehenders can use concessive connectives to update their predictions. In the current ERP experiment we investigated whether this discourse marker can have an immediate impact on the readers’ expectations. We found that the incongruency effect as measured by the N400 was qualitatively similar for both target words that immediately followed ‘Nonetheless’, and those that were linearly removed from the concessive connective. We conclude that comprehenders can use ‘Nonetheless’ to update their expectations immediately.
Yana Arkhipova, Ryan Law, Ming Xiang and Wing Yee Chow

Timing of structural and non-structural information in processing of reflexives in the L2

This study examines how Turkish learners of English process English reflexives (himself/herself). Two eye-tracking experiments and a pen-and-paper antecedent identification task were employed to test whether Turkish learners of English were guided by syntactic requirements associated with Binding Principle A, or non-structural information (discourse prominence or linear proximity) influenced their antecedent retrieval behavior for the reflexives. Data were collected from 95 advanced learners of English. The results showed that Turkish learners of English were, to a good extent, similar to native speakers of English in their reflexive antecedent retrieval behavior both in their processing and in their ultimate interpretations.
Münir Özturhan and Nazik Dinçtopal Deniz

Spatial Language and Reference Tracking

For salient discourse referents, repeated-name anaphors are read more slowly than pronoun anaphors. fMRI research suggests that this “Repeated Name Penalty” (RNP) activates the same parietal regions engaged during tracking objects in space. We tested the hypothesis that reference tracking during reading involves spatial cognitive processes, and that pronouns and repeated names engage these processes differently. Results from three experiments showed a reduced RNP under conditions of visuo-spatial memory load compared than under conditions of non-spatial visual memory load, such that definite descriptions but not pronouns benefited from explicit spatial associations, supporting the involvement of spatial processes in reference tracking.
Ansley Potter and Amit Almor

Crosslinguistic influence modulates L2 speakers’ offline but not online processing of implicit causality

This study investigates effects of cross-linguistic activation of referential bias in L1-Korean/L2-English bilinguals’ processing of English implicit causality verbs in offline and online processing. Specifically, it provides evidence of the influence of cross-language activation at a word level on biases in subsequent sentence interpretation in offline but not online processing. These results are expected to contribute to our understanding of how cross-language activation during lexical processing influences referential choices in offline and online discourse processing.
Hyunwoo Kim and Theres Grüter

Event categorization using non-verb cues

A single verb sense can denote multiple event categories: a person's raising a glass in celebration (a toast) is a different kind of event from a crane’s raising a truck out of a lake. A rater study and similarity judgment task suggest that non-verb cues are used to distinguish such categories. A second experiment consisting of a sorting task and a feature elicitation task investigates which non-verb information is utilized for event categorization across different semantic domains. These data provide a way to expand current theories of event categorization to granularities beyond the verb sense.
Aron Marvel, Jean-Pierre Koenig and Gail Mauner

Word frequency and the elusiveness of the L2-L1 (masked) translation priming effect

We explored the masked priming asymmetry observed in lexical decision tasks (LDT) when unbalanced bilinguals are tested and non-cognate translation equivalents are used. The asymmetry consists of priming being significantly larger for L1 primes on L2 targets than vice versa (L2 primes-L1 targets). We tested L1 Spanish-L2 English speakers in a masked translation priming LDT. L2 proficiency, word frequency, and other linguistic factors were investigated. Our data showed a priming asymmetry; crucially, word frequency modulated the L2-L1 (non-significant) priming. The BIA+ model, under which slower L2-word processing prevents L2 primes from activating their L1 counterparts, can best accommodate the results.
Adel Chaouch Orozco, Jorge González Alonso and Jason Rothman

Immediate revision of disconfirmed predictions: Evidence from Eye-tracking and ERPs

Based on previous findings that comprehenders can reliably detect cues that conflict with their predictions (e.g., an adjective that mismatches the most expected noun in gender), the present study asked what cognitive processes are engaged when comprehenders detect prediction-inconsistent cues by looking at the immediate and downstream effects of prediction-inconsistent classifiers in Mandarin Chinese. We report evidence from visual world eye-tracking (Experiment 1) and ERP (Experiment 2) which suggest that comprehenders are able to use a prediction-inconsistent classifier to revise their disconfirmed predictions and pre-activate a contextually suitable (albeit originally unpredictable) noun very quickly.
Wing-Yee Chow, Di Chen and Suiping Wang

Processing of number agreement in the L2: An eye-tracking study

In sentences with a complex subject, number mismatch between the local and the head noun cause errors in production and slow-down in processing. This study, with Turkish learners of English, investigated whether linear or syntactic distance affected processing number agreement in the L2, using two eye-tracking experiments and a pen-and-paper gap-fill task. The results showed that when there are no lexical cues, L2 learners are sensitive to linear distance. However, provided with lexical information, they showed sensitivity to both linear and syntactic distance, suggesting that L2 learners can do complex syntactic processing like native speakers, when semantic cues are present.
Hilal Serin Demirler and Nazik Dinçtopal Deniz

Predictive processing is affected by linguistic complexity of cue-preceding input

In two visual world experiments, we compare predictive and regressive eye movements in Mandarin passive structures, in which the marker bèi (被) can be regarded as a reliable cue for anticipating an animate referent following it. The critical manipulation concerns the phrases preceding the cue. We contrast SRC and ORC with simple coordinate phrases. Results show fewer and delayed anticipatory and more regressive eye movements in the SRC condition, suggesting that the relative difficulty of linguistic input directly affects to what extent the comprehension system engages in prediction.
Johannes Gerwien, Yuxia Wang and Fuyun Wu

Strategies of filler-gap avoidance in Romance

Filler-gap dependencies are a central topic in sentence processing research. These involve the movement of an element (the filler) from a location at which it receives its thematic role (the gap), and in which a trace remains. We present two acceptability judgment studies and two self-paced reading studies in Italian and French that tested two temporarily ambiguous structures that only differ with respect to the presence/absence of the filler-gap dependency. We show that a structure that does not involve a gap is preferred over one that does, providing further evidence for a structural parsing strategy of gap avoicance.
Francesca Foppolo, Ingrid Konrad, Caterina Donati, Massimo Burattin, Adrian Staub and Carlo Cecchetto

Degrees of prediction: Syntactic pressures generate stronger predictions than pragmatic considerations

During sentence processing, comprehenders predictively form structural dependencies (e.g. “active filler”). This strategy was argued to be motivated by syntactic licensing, semantic integration, or pragmatic considerations. We compare formation of filler-gap (syntactic) dependencies with pragmatically-motivated dependencies, using NPs embedded in Hebrew regarding-phrases, which form discourse-prominent antecedents requiring future reference. Results of two SPR experiments suggest that pragmatic motivations can give rise to a preference for co-reference, but the parser predictively forms syntactically-motivated dependencies in a more committing fashion. In addition, the study provides evidence for a functional distinction between different measures of active dependency formation (filled-gap effects and plausibility manipulations).
Maayan Keshev and Aya Metzer-Asscher

The interaction of image salience with language: how much is language driving eye-movements in Visual World Paradigms?

Stephanie Huette and John Hollander

Availability-based production predicts speakers' real-time choices of Mandarin classifiers

We investigate how contextual predictability influences the encoding of linguistic content manifested by speaker choice in a classifier language, Mandarin, where nouns obligatorily require classifiers when modified by numerals. While different nouns take different specific classifiers, there is a general classifier that is compatible with most nouns. When the noun is unpredictable, a specific classifier would reduce the noun's surprisal, potentially facilitating comprehension (Uniform Information Density; Levy & Jaeger, 2007), but the specific classifier may be dispreferred if the general classifier is more available (Availability-Based Production; Ferreira & Dell, 2000). Our picture-naming experiment confirmed two predictions made by Availability-Based Production.
Meilin Zhan and Roger Levy

Thematic fit and grammatical constraints in good enough parsing

Recent research indicates that lexical information and semantic plausibility can influence syntactic constructions. However, the distinction between lexical association and plausibility in sentence processing is under-explored. We investigate how lexical association and plausibility each affect the processing of garden-path sentences. We manipulated ±Garden Path and Thematic Fit (Plausible Theme, Plausible Instrument, Implausible Argument) in a self-paced reading study. If the processor distinguishes between lexical association and plausibility, we predict a difference in reading times between the Plausible Instrument and Implausible Argument conditions. We found no such difference. This suggests that the processor does not distinguish between lexical association and plausibility.
Jamie Linert and Dustin Chacón

Conversational Alignment and Conversational Fluency

Three experiments investigated the automaticity of language processing during conversation. Experiments 1 and 2 measured turn transition time, speech rate, turn length, and disfluencies during spontaneous conversation with a secondary n-back task that varied in difficulty. While some trade-offs were observed between conversation and the secondary task, conversational fluency was surprisingly robust--perhaps because the interlocutors were well-aligned. Experiment 3 used a picture-description task to investigate how interlocutor alignment facilitates language production. We found a strong effect of shared topic, but no advantage for shared vocabulary.
Julie Boland

Do individual differences in fiction reading predict emotion recognition?

We examine the effect of fiction reading exposure on emotion recognition abilities. Our corpus analyses reveal systematic differences in treatment of simple (e.g. joy) and complex (e.g. relief) emotions by corpus genre: fiction is an especially pure source of complex emotion information whereas simple emotions are treated equally throughout all genres. These language statistics lend causal support to our finding that people who read more fiction are better are recognizing emotions in others, specifically complex emotions. Our results suggest far transfer of language experience to social cognition and explain why fiction language experience is uniquely beneficial.
Steven Schwering, Natalie Ghaffari-Nikou, Paula Niedenthal and Maryellen MacDonald

Do animated cues elicit similar patterns of pronoun comprehension as live cues?

Pronoun comprehension is influenced by social and syntactic cues. In “Ana is cleaning up with Liz. She needs the broom”, listeners tend to assign the pronoun to the subject character, but this effect is modulated by who the live speaker gazes/points at. This study investigates whether we can replicate this effect using an animated speaker. Results show that animated stimuli can be used to test the effects of social cues on reference comprehension, although it strengthens social-cue effects. This method allows for greater control over the stimulus while providing a platform for testing other effects on pronoun comprehension.
Valerie Langlois, Sandra Zerkle and Jennifer Arnold

Effects of emotional speaker face and event-sentence mismatches on sentence processing: an event-related brain potential study

In communication, many non-linguistic cues could enrich language processing, among them a speaker’s emotional facial expression and events. Both can (mis)match in emotional valence / reference with what a speaker is about to say. In an EEG experiment, adults inspected the emotional face of a speaker and listened to the beginning of a sentence. The face was either positive or negative. Next, participants saw a positive or negative event photo and heard the sentence continuation. The sentence content matched (vs. mismatched) the photo. ERP deviations between emotionally congruent (negative) speaker face-photo-sentence and incongruent (positive) speaker face (negative) photo-sentence trials emerged.
Katja Münster, Johanna Kißler and Pia Knoeferle

Perceptual information from word n+2 does not affect the skipping of word n+1

In an eye-tracking study we examined whether the skipping of the indefinite article "an" is affected by whether the following word begins with a vowel (e.g. "African") as is typical in English, or with a consonant (e.g. "Russian"). We used the boundary paradigm to present readers with a preview of the word following "an" which was either correct, or which started with a consonant. This preview changed to the vowel-initial target as the eye crossed a boundary in the spaced before "an". "An" skipping was unaffected by the preview manipulation, although readers did gain an n+2 preview benefit.
Michael G. Cutter, Andrea E. Martin and Patrick Sturt

Both thematic role and next-mention biases affect pronoun use in Dutch

The aim of this study was to disentangle predictability effects on pronoun use from thematic-role effects in Dutch. We conducted two web-based written continuation experiments, in which we manipulated the next-mention bias. Experiment 1 showed that this manipulation shifted the preferred referent from NP2 to NP1. Experiment 2 showed that when the referent was congruent with this bias, participants produced more personal pronouns as well as more reduced forms compared to full forms. Thematic role mainly affected demonstrative pronoun use. Thus, thematic role and predictability both seem to affect the choice of referring expression in Dutch, but in different ways.
Jorrig Vogels

Aging effect in the processing of homonymy and polysemy in Korean sentence comprehension

This study aims to explore the aging effect in the processing of homonymy and polysemy. We conducted self-paced reading comprehension and cognitive tasks (Stroop and Reading Span). Young readers’ behaviors support the existing view claiming that the semantic representation for homonym is discrete but that for polysemy is unitary. However, old readers’ behaviors did not support any specific view while they were having difficulty in resolving ambiguity. The results of cognitive models yielded that the aging effect in processing ambiguous words was more modulated by how efficiently readers could inhibit irrelevant information rather than how good readers could store/manage information.
Hongoak Yun, Sun Hyun Moon, Hyein Jeong and Soo Rim Noh

The processing of subject-predicate dependency in Japanese

In this study, three self-paced reading experiments were conducted to investigate what types of subject affects the processing of subject-predicate dependency in Japanese. Both distant-sensitive and distant-insensitive cases were found in previous studies in Japanese, so in the present study, we examined locality effects, using different types of matrix subjects. In Experiment 1, we showed the case where the greater processing difficulty occurred when the matrix subject was an NPI and contained wh phrase in nested structures. Experiments 2 and 3 showed that locality effects were observed in nested structures, regardless of the wh states and of the predicate types.
Saki Tsumura and Yuki Hirose

Falling for the grammaticality illusion? Reading span levels, individual parsing strategies and the missing-VP effect in German

"The inconsistency of sentence processing effects between studies are partially related to individual differences. In an eye-tracking study, we investigate individual parsing strategies and the missing VP effect in German. Participants had to read complex sentences involving multiple center-embeddings (grammatical <-> ungrammatical), answer a post-experiment questionnaire and also complete a reading span test. Reading times suggest that participants differ when using morpho-syntactic agreement cues (like Number) in complex sentences. Questionnaire feedback and the reading span test show a link between an individual's reading span score and their ability to keep track of structural dependencies."
Katja Suckow, Jana Häussler and Anke Holler

Do Mandarin Speakers Think about Time More Vertically?: A Revisit of Boroditsky (2001) with New Findings

How mandarin speakers conceptualize time? It is generally agreed that Mandarin speakers have two mental timelines (i.e., vertical and horizontal). As Boroditsky (2001) suggested, Mandarin speakers also have a vertical bias in the conceptualization of time due to the frequent use of vertical spatiotemporal metaphors. The present study further investigates this question with corpus-based analysis and a non-linguistic temporal sequence judgment task. Corpus analysis reveals that Mandarin speakers predominantly use temporal expressions of time rather than vertical ones. The experiment shows that Mandarin speakers do not bias towards any axis in the conceptualization of time.
Mingyu Yuan

Antecedent Retrieval during the Processing of Dutch Reciprocal Pronouns

Contrary to predictions by cue-based parsing models, processing local antecedent-anaphor dependencies seems rarely susceptible to interference effects of structurally inaccessible distractor NPs. We sought to investigate whether this is due to reactivation of the grammatical antecedent at the main clause verb when it precedes the anaphor. In this self-paced reading study on Dutch (SOV) reciprocal pronouns, no main clause verb intervened between antecedent, distractor, and reciprocal. We observed no significant distractor interference effects on reading times, suggesting that the apparent immunity of reciprocals to interference effects cannot fully be ascribed to baseline activation differences induced by a main clause verb.
Anna Giskes and Dave Kush

Computing complex events and beyond: ERP data on prediction with complex verbs

How quickly can verb-argument relations be computed? How quickly can a complex event structure impact predictions? We take advantage of the substantial differences in verb-argument structure provided by Mandarin, investigating the following two event relations: Resultative (Kid bit-broke lip, meaning: a kid bit his lip such that it broke) and Coordinate (Store owner hit-scolded employee, meaning: a store owner hit and scolded an employee). N400 evidence showed that prediction is specifically delayed at verbs instantiating the causality relation (breaking-BY-biting) relative to the coordinate relation (hitting-AND-scolding), suggesting given limited cognitive resources, not all information can be computed to update predictions immediately.
Chia-Hsuan Liao and Ellen Lau

Phonotactic markedness increases processing demands in speech production

Recent debate in speech production concerns whether putative effects of phonological markedness differ from within-language lexical statistics in their effect on speech production. In this study I examine phonotactic markedness, a type of phonological markedness which partially corresponds to language-specific distributional regularities. Using a speeded production paradigm, participants read aloud pairs of “sentences” comprised of four nonwords: one arrangement created phonotactic-violating sequences across the sentence’s word boundaries, while the other didn’t. I find violations of phonotactic markedness increase response latencies, signaling increased processing demands. This finding supports production theories which allow grammatical information to influence the course of phonological encoding.
Canaan Breiss

The role of case marking and word order in cross-linguistic priming

In the current study, we investigated the role of case marking and word order in cross-linguistic structural priming using an artificial language (AL) learning paradigm. Priming was assessed between Dutch (no case marking, SVO word order) and a) a baseline AL version with SVO word order, b) a case marking AL version, and c) an AL version with SOV word order. Cross-linguistic priming was found for the transitives, but not for the ditransitives. There was no difference between the AL versions in the priming pattern, suggesting that differences in case marking or word order do not hinder cross-linguistic priming.
Merel Muylle, Sarah Bernolet and Robert Hartsuiker

Comprehending the presupposition of too: the effects of distance and interference

"Anaphoric presupposition triggers such as too are thought to establish a dependency relation between the trigger and its presupposed content. Like other dependencies, we hypothesize that establishing presuppositional dependencies relies on memory retrieval. Our results from speeded acceptability judgement studies suggest that the processing of ""too"" shows signatures of memory retrieval, and while this retrieval process appears to be directly accessible, this retrieval of the presupposed content is not interference-prone."
Sherry Yong Chen and E. Matthew Husband

The distribution of biased versus alternating verbs affects preschoolers’ verb bias adaptation

Language learning requires balancing the productive power of language with its constraints, and verb bias is one such constraint. In a training study with familiar verbs, we found that four- and five-year-old children tracked statistics at multiple levels within their native language. Their production patterns showed that they adapted the biases of familiar dative verbs when most verbs in the study were biased toward one structure but not when most verbs alternated. Our results comport with the findings of artificial language learning studies and lend support to a role for multi-level distributional learning in syntax acquisition and verb learning.
Yi Lin, Malathi Thothathiri and Fisher Cynthia

A Rational Speech Act model of cross-linguistic differences in pronoun resolution preferences

We propose a computational model of differences in pronoun resolution preferences between English and French sentences like “The postman called the street-sweeper before/after he went home.” / “Le facteur a appelé le balayeur avant qu’il rentre à la maison/après être rentré à la maison » within the RSA framework. It takes into account a general subject bias and, importantly, differences in the inventory of syntactic alternatives (e.g. “avant de rentrer”/”before going”) based on detailed corpus studies in both languages. Model predictions are mostly confirmed in a cross-linguistic experiment while showing that base frequencies of connectors also seem to be relevant.
Miriam Schulz, Heather Burnett and Barbara Hemforth

Formal marking is redundant with lexico-semantic cues to meaning in transitive clauses

How informative are word order and case marking for distinguishing subjects and objects in usage? We answered this question for English and Russian. For both languages, we extracted triplets of verbs, subjects, and objects from corpora, retaining only head words and removing case, agreement, and order information. We presented these triplets to experimental participants and tasked them to identify the subject. In both languages, participants identified subjects in around 90% of cases, suggesting that word order and case marking are informative at most 10% of the time. This surprising redundancy constrains theories of linguistic universals based on efficient communication.
Richard Futrell, Evgeniia Diachek, Nafisa Syed, Edward Gibson and Evelina Fedorenko

Distal rhythmic patterns influence listeners’ processing of phonetic cues

Distal rhythmic alternations in pitch/duration aid listeners in word segmentation. The present study tests how they influence speech perception, investigating their effect on the perception of durational cues. Listeners categorized a “coat”~“code” vowel duration continuum, where the target syllable was grouped by distal rhythmic alternations as part of either an iamb, or trochee. Results show listeners adjust categorization based on the implied grouping of the target in either type of foot, suggesting that distal rhythmic/prosodic patterns mediate the uptake of phonetic cues. These results thus highlight the importance of temporal/rhythmic structure in speech at multiple levels of spoken language processing.
Jeremy Steffman

Reducing similarity based interference in sentence processing for individuals with agrammatic aphasia: a semantic approach

By means of Eye-tracking while listening paradigm, this study examined individuals with agrammatic Broca’s aphasia and age-matched controls in their processing of sentences containing unaccusative verbs. The aim was to determine 1) whether the structural similarity between the displaced and the intervening noun phrases results in interference during real-time processing, and 2) whether semantic manipulation of the intervening noun phrase reduces the similarity-based interference. Results revealed an on-time processing pattern for individuals with agrammatic Broca’s aphasia when the similarity-based interference was reduced via semantic manipulation (changing the animacy) of the intervening noun phrase.
Niloofar Akhavan, Carolyn Baker, Natalie Sullivan, Lewis Shapiro and Tracy Love

Effect of partial quotation and transparent free relatives on perspective shift

Perspective shift allows speakers to utter content they may not fully endorse. We explore the interpretation of perspective shifting expressions with a study on Transparent Free Relatives (Allen poured [what (is called/he calls) a beergarita]) and orthographic Partial Quotation (Allen poured a ‘beergarita’) in a Speaker/Listener production-perception experiment. While PQ and TFRs with explicit subjects (what he calls) licensed shifted interpretations for the Speaker, only TFRs without explicit subjects licensed shifted interpretations for the Listener. We propose that PQ provides a less robust cue for perspective shift, even when the speaker attempts to prosodically mark it as non-speaker oriented.
Bethany Sturman and Jesse Harris


We report results from a driving-simulator paradigm developed to test the effects of verbal-tasks on simultaneous driving performance. Two experiments (E1, E2) measured performance continually as overall distance from moving target, under different target speed conditions (fast, slow) and different conversation manipulations (listening/responding (E1), reading/responding (E2)). In both experiments, performance gradually deteriorated during speech production and improved during comprehension, with worse performance overall during speech production. Overall, our paradigm provides a useful tool for studying the rapidly changing resource requirements of language production and comprehension during concurrent driving tasks.
Jonathan Rann and Amit Almor

Sensitivity to Gender Cues during Anaphor Resolution: Evidence from Visual World Eye-tracking

We used the visual world paradigm to investigate how native English speakers interpreted argument reflexives and pronouns in real time. Results are consistent with the multiple constraints approach where both structural (binding principles) and non-structural constraints (gender) influence anaphor resolution from the beginning of the antecedent search in a form-specific sensitivity way. Also, the use of both linguistic and non- linguistic information (e.g., the gender information was encoded on visual stimuli only) during anaphor resolution supports a highly interactive and contextually sensitive model of human language processing in general.
Yuhang Xu, Nicholas Ringhoff, Rachel Coons, Lauryn Fluellen, Carly Eisen and Jeffrey Runner

Are gender and number different? Evidence from French binomials

This paper investigates the number and gender agreement in the coordination phrase in French. The French grammar dictates that only resolution rule can be used with a coordination NP. Our results show that agreement with the closest conjunct (CCA) is quite acceptable in French and number and gender agreement are processed in a different way as CCA is more acceptable for gender agreement. We argue that resolution rule causes more surprisal effects in the incremental processing and Closest Conjunct Agreement is more acceptable because it is easier to process.
Aixiu An and Anne Abeillé

Age affects the processing of subject-verb agreement (in Italian)

The study reports age-related ERP response variability to subject-verb agreement violations in native Italian speakers. We think that these results open new perspectives within the current debate on the stability versus variability of ERPs elicited by syntactic violations, suggesting that the pattern can be more complex than a specific stable pattern (P600 or LAN+600) as well as an unidimensional continuum between negative responders (N400) and positive responders (P600).
Francesco Vespignani, Luca Artesini and Francesco Pavani

Gender Bias in Picture Noun Phrase Reflexive Resolution

Using a “broken-text” paradigm, we eliminated the gender information encoded on the English reflexive himself/herself (i.e., hXXself) and tested how gender information (e.g., the gender of protentional antecedents) affected the “picture noun phrase” (PNP) reflexive resolution as well as its interaction with other constraints (e.g., verb semantics). Our results are consistent with the “multiple constraints” approach on anaphor resolution such that people are sensitive to all kinds of information even when it is unavailable. More importantly, different constraints are not used independently but are intertwined and interact with each other during resolution in a context-sensitive fashion.
Yuhang Xu, Carly Eisen, Lauryn Fluellen, Yuyi Zhou, Nicholas Ringhoff, Rachel Coons and Jeffrey Runner

Is turn prediction accuracy across Language and Music dependent on the idiosyncrasies of one’s own experience?

This study investigates whether turn prediction accuracy is influenced by the idiosyncrasies of one’s own experience across both language and music. Thirty pianists recorded 60 melodic phrases and utterances over two sessions, the pianists then predicted turn-ends of the recordings via button press or verbal/musical response. Each participant listened to a third of the recordings produced by themselves, a third produced by someone rated as similar to them in style, and a third by someone rated as dissimilar. Results suggest that people were more accurate at predicting self than other which provides support for the prediction-by-simulation account of turn-taking.
Nina Fisher, Lauren Hadley and Martin Pickering



Long filler-gap dependencies can make a main verb analysis less tempting

Race models claim that the parser adopts the first available analysis. However, there is little direct evidence for this. In this study, we provide a test of this prediction by investigating the MV/RRC ambiguity, but in contexts where the MV analysis requires resolving a filler-gap dependency. We find that making the filler-gap dependency longer reduces garden path difficulty. We take this as support for race-based models of syntactic processing, and evidence that at least part of the robust MV advantage in ambiguity resolution is rooted in how quickly/easily the parser can compute the main verb analysis
Brian Dillon

Fine-grained gender typicality systematically modulates anaphora resolution: Evidence from eye movements

The current eye-tracking study investigated the impact graded gender typicality (i.e., the gender bias associated with certain nouns) on anaphora resolution. Participants read sentences that contained role nouns with varying levels of gender typicality followed by co-referential reflexive anaphors (himself/herself/themselves). The data reveal that comprehenders are immediately sensitive to fine-grained gender typicality during anaphora resolution. Reading times during the earliest measures on the gender marked reflexives significantly covary with gender typicality. For themselves, although readers incur an early cost, they overcame it for equi-biased nouns, suggesting themselves was recognized as a gender neutral-pronoun, albeit at a delay.
Vishal Sunil Arvindam and Brian Dillon

Pragmatic relativity: Eye-tracking reference resolution in Catalan, Hindi, Hungarian and Wolof

We compared the processing of modified definite descriptions using color, material and scalar adjectives (e.g., ‘The black lamp’) across two languages with prenominal modification (PreN: Hindi and Hungarian) and two with postnominal modification (PostN: Catalan and Wolof). Eye-tracking results confirmed that speakers of PreN languages established referential contrast across categories (black objects > lamp), whereas PostN speakers established referential contrast within a category (lamps > black one). More importantly, adjective position determined whether the contrastive interpretation of an adjective resulted from a pragmatic inference (PreN languages) or incremental language processing (PostN languages). We conclude that language structure affects pragmatic reasoning.
Paula Rubio-Fernández and Julian Jara-Ettingerr

ASL signers vary adjective position to maximize efficiency: A reference production study

We ran a reference production task in American Sign Language to test the hypothesis that adjective position may be exploited to maximize efficiency. Unlike most languages, ASL has free adjective position (i.e. both ‘blue cup’ and ‘cup blue’ are acceptable). Based on previous eye-tracking data, we predicted that color adjectives would be a more efficient visual cue in prenominal position, whereas the reverse should hold for scalar adjectives, which are readily interpretable only once the noun has been processed. The results supported our predictions, with ASL signers producing color adjectives in prenominal position significantly more often than scalar adjectives.
Paula Rubio-Fernández, Anne Wienholz, Simon Kirby and Amy Lieberman

Perspective taking in newly-sighted children: Do they orient towards an interlocutor’s face?

Both children and adults follow gaze to disambiguate reference. In a study with Prakash children (newly sighted children treated from congenital cataracts), we investigated perspective taking in referential communication, starting with their orientation towards an interlocutor’s face during naturalistic interaction (Exp 1) and more controlled conditions (Exp 2). In both experiments, Prakash children made fewer fixations on their interlocutor’s face than the neurotypical controls. Interestingly, directly asking children to focus on the speaker’s face improved their gaze following. The results of this first study suggest that after years of visual deprivation, orienting towards an interlocutor’s face may not come naturally.
Paula Rubio-Fernández, Vishakha Shukla, Shlomit Ben Ami, Chetan Ralekar, Ashwini Vaidya, Michelle Oraa Ali and Pawan Sinha

Enhanced peripheral lexical processing in deaf individuals: perceptual or linguistically driven?

We investigated superior peripheral linguistic processing abilities in deaf signers compared to hearing people via three experiments testing their ability to identify or discriminate ASL handshapes in the near and far periphery. Deaf signers outperformed hearing participants, performance was always better in the near periphery, and eccentricity only interacted with hearing status for dynamically moving signs. Linguistic knowledge only facilitated discrimination for ASL signs, and only for deaf participants in the far periphery. Thus, deaf signers become particularly adept at identifying peripheral information when the task (1) contains dynamic linguistically meaningful information and (2) resembles real world processing demands.
Emily Johnson, Elizabeth Schotter and Amy Lieberman

Overspecifying state information in the production of referring expressions

Referring expressions usually include “just enough” information, but sometimes speakers “overspecify”, namely also include unnecessary information. We examine modifiers about the STATE of objects (dirty/clean, open/closed), demonstrating that they are more likely (means: 21% vs. 1%) when the object appears in a less typical state (e.g., “closed eye” is more likely than “open eye”, and vice versa for “can”). This pattern provides support for the hypothesis – originally proposed for color adjectives – that speakers tend to overspecify atypical properties. Furthermore, we observe a novel finding that overspecification is more likely when no modifiers are required in the given context.
Megan Parker and Daphna Heller

Evaluating truth: Experimental evidence from appositives and conjunctions

Three forced-choice experiments compared truth judgments of sentences containing not-at-issue (appositive) and at-issue (conjunct) content. In Experiment 1, participants judged sentences under a general QUD to be true only when each clause was also true. Experiments 2 & 3 showed that under a specific QUD, some participants disregarded clauses containing non-relevant false information. Our results suggest comprehenders can selectively attend to or upweight the content of some clauses over others. Because conjuncts and appositives were able to be backgrounded, this backgrounding property is best explained under accounts of attention-allocation and goal-directed behavior at the level of the speech act.
Margaret Kroll and Amanda Rysling

Variation in the rational interpretation of slights: Gender-based microaggressions

How can comments like “You’re actually good at math” be compliments to some, but microaggressions to others? Differences across listeners likely arise from variations in their underlying beliefs: when “actually” or “for a woman” is regarded as unnecessary given the contextual assumptions, it triggers the microaggression interpretation: the speaker did not treat the information as expected. We predicted derivations of such interpretations based on a)political party affiliation, b)beliefs about implicit sexism, and c)subjective estimates of information amount carried by the adverbials. These results present novel evidence that variations in listeners’ linguistic and non-linguistic beliefs combine to predict pragmatic interpretations.
Bethany Gardner and Chigusa Kurumada

The (non-)influence of even’s likelihood-based presupposition on lexical predictability effects

"We present the results of an eye-tracking while reading study investigating how and whether even’s likelihood-based presupposition influences lexical predictability effects. Given that even provides information to readers that upcoming material will be less likely than relevant alternatives, the presence of even in a sentence may influence how readers process predictable and unpredictable sentence continuations. The presence of even did not eliminate predictability effects. However, there was limited evidence for an effect of even on later reading time measures. This evidence suggests that the presupposition introduced by even may be computed and integrated at a later stage of semantic processing."
Erika Mayer, Brian Dillon and Adrian Staub

Inflection and Derivation in Proficient L2 Arabic Sentence Processing

The current study brings evidence of sensitivity to both inflectional and derivational sublexical structure during sentence reading in proficient L2 Arabic learners (N = 44) to bear on the controversy surrounding the source(s) of the observed differences between L1 and L2 morphological behavior. A self-paced reading task used anomalies in verbal inflection (number agreement), verbal derivation (template morpheme), and verbal semantics to test whether learners automatically access these features while reading. Results indicated that both L2 learners and native speakers alike made use of Arabic derivational and inflectional morphological structure during sentence processing.
Suzanne Freynik, Kira Gor and Polly O'Rourke

Priming children's interpretation of globally ambiguous sentences

Children (ages 4;6-6;6) were primed with a picture matching task using globally ambiguous sentences like "The sheep points at the pig with the pipe cleaner." Prime pictures disambiguated between a high (HA) and low (LA) attachment interpretation, e.g., [VP[VP points at the pig][PP with the pipe cleaner]] versus [VP points at [NP the pig with the pipe cleaner]]. As a baseline, children choose the HA interpretation 82.1% of the time. Production priming significanty increased the selection of the picture with a matching interpretation for both priming groups (HA: 93.6%, LA: 68.0%).
Emily Atkinson, Natalie Bloss, Sofia Johnson and Carolina Jones

The Effects of Word Frequency and Contextual Diversity on Word Processing during Reading across ERP and Eye-tracking Measures

The current study examined the effects of word frequency (how often a given word appears) and contextual diversity (the number of contexts in which a given word appears) using eye tracking and electrophysiological measures. The results of two experiments show that the effect of contextual diversity persists even when controlling for word frequency. However, when controlling for contextual diversity, standard word frequency effects for ERP components and eye movement measures disappear. These findings provide further evidence that contextual diversity is an important aspect of lexical representation and appears to hold more psychological relevance than mere frequency of exposure.
Patrick Plummer and Charles Perfetti

Different ways to process honorific features: Evidence from ERPs

We ask whether agreement is processed in the same manner even in a different configuration, if the nature of agreement is semantically the same. In Korean, the social status of a referent relative to that of a speaker decides which verbal form as well as which case marker to use. Our ERP results showed while a violation of a subject-verb honorific agreement elicited the P600, a violation of honorific case agreement elicited a N400. These results suggest that not semantic content itself but the way that the semantic content is used within a sentence determines the way it is processed.
Nayoung Kwon and Patrick Sturt

Garden-Path Misinterpretation in Reading While Listening

We designed a garden-path misinterpretation study using the subordinate object/main clause subject ambiguity to examine whether reading while listening (RWL) affects the rate of misinterpretation as a behavioural measure of comprehension. We hypothesized that the ability to pause, delay, or regress in reading may be disrupted by co-current speech which lack rich prosodic information, thus impairing sentence comprehension. Against our hypotheses, we found that, while our participants displayed the classic misinterpretation effect of garden-path sentences, RWL did not affect this misinterpretation negatively or positively when compared to silent reading.
Guorong Zhang and E. Matthew Husband