It's time to face a simple question: WHY is canonical form simple?

Lise Menn, University of Colorado August 1999

revised version, to appear in the Third Millennium Special Issue of Brain and Language

Simple sentences in 'canonical form' - that is, in each language's most typical main-clause surface word order for subject (S), object (O), and main verb (V) - are easier for its speakers and hearers to produce and understand, whether they are aphasic or unimpaired, children or adults, native speakers or second language learners. This has been verified for SOV Japanese and Turkish as well as for SVO English and French. It holds whether the language has a 'fixed' word order (which can be varied only if it is marked with additional grammatical morphemes, e.g. Chinese, English) or 'free' word order (which can be varied according to discourse purposes without extra morphological markers, e.g. Italian, Japanese). It holds for most Germanic languages, which require other word orders in the majority of clauses used, and in French and Italian where, as in all the Romance languages, most pronominal objects are cliticized and placed before the verb, not after it (for references, see Bates & Goodman 1999).

From the beginning of experimental work in syntax (e.g. Bever 1970), we have seen that under competence-compromising and/or performance-limiting conditions, people interpret passive sentences to mean exactly their opposites, as though only the relative order of the two nouns were being processed, with the attendant grammatical morphemes serving (at best) to cause confusion. This dependence on the order of the nouns extends to cleft and pseudo-cleft sentences (What the dog is chasing is the bird' - easy; It is the bird that the dog is chasing - hard).

The simplicity of canonical word order is embodied in transformational and relational grammars: active-voice canonical word-order sentences are those from which passives and others are derived. These multi-level grammars efficiently capture the semantics of passive and other non-canonical sentence types by equating the Subject in an underlying structure with the semantic role of Agent, and stipulating that this role remains unchanged when transformations operate to derive the surface word order. In such grammars, therefore, all agentless sentences (e.g.Chris fell), not just passives (Chris was pushed), are transformationally derived from underlying subjectless ones in which the undergoer of the action is the object (____ fell Chris, ____ pushedChris).

But why, psycholinguistically and neurolinguistically, is canonical word order simple? Is this due to its syntactic properties? In the transformational tradition, yes: canonical form is simple because sentences in this form require fewer transformations than sentences of other types, and have fewer 'traces'. Or is it because of its semantics? Bates et al. (e.g. 1988) treat it as such implicitly, referring to Agent, Action, and Object, rather than Subject, Verb, and Object.

But neither of these basic assumptions has been seriously tested. Virtually all experimental work to date is useless for evaluating them because it relies on verbs in which syntax and semantics are confounded: 'hard-core' transitive action verbs like kick, chase, and break. In the active voice of such verbs, the Agent is always the Subject, and the Object is always the Undergoer of the action; with such verbs, the processing of clauses in which the Agent is NOT the Subject (e.g. passive, pseudo-cleft) depends crucially on the comprehension or production of case markers and/or verb-form markers like be, -ed.

What about verbs whose subjects undergo or are affected by the action, like transitive receive in I received a letter, or get in He got a scolding, or smell in She smelled smoke? What about intransitive fall, trip, stumble ? What about verbs which may be either transitive or unaccusative, depending on their arguments (She dropped the apple/The apple dropped )? Is The apple dropped in fact difficult for people with aphasia? for young children? for normal adults in L1 under stress?

Based on preliminary data, Menn et al. (1998) proposed a third type of hypothesis, compatible with such linguistic approaches as Construction Grammar (Goldberg 1999) and psycholinguistic models such as the Competition Model (MacWhinney 1987). We suggest that aphasic problems with producing non-canonical structures such as passives derive principally from two sources: difficulty in retrieving the less frequent frames FOR A GIVEN VERB, and difficulty in controlling the placement of retrieved noun lemmas in the correct slot at the functional level. Therefore, common unaccusatives (fall ) and verbs of psychological state (involuntary smell ) should cause no problems except those due to choices between competing possible interpretations. If a verb is most commonly used in the passive, passive should be easy for that verb.¹

Verbs other than hard-core transitives have been excluded from the investigative paradigm because most methods require clauses which could plausibly have either of the given arguments either as subject or as object (The dog chased the bird/The bird chased the dog ). The major paper that deals explicitly with verbs whose Agents are not Subjects, Kegl (1995), claimed that they were in fact problematic in agrammatic aphasia; however, she used naturalistic story narratives, in which foreground action tends to be encoded with action verbs, background information with other types. Gottfried et al. (1997), on the other hand, showed that aphasic speakers had equal difficulties in repeating matched sentences whose Subjects were Agents and sentences whose Subjects were Undergoers.

Apparent dependence on Agent-first canonical form may therefore really be dependence on active voice verb morphology, plus the expectation that the mapping of noun positions onto semantics will follow the pattern of the individual verb’s most frequent semantic frame.

Not until we know why canonical form is simple will we be able to develop an adequate model of even the most basic aspects of on-line sentence processing. I propose that resolving this very simple question by proper studies is the most urgent aphasiological issue for the first decade of the new millenium.

Note: 1. Refinements that take the noun arguments into account may be needed: for example, long-term real-world properties of materials may affect the comprehensibility of The sun melted the ice/butter/iron. This is a separate issue from whether incidental context outside the clause affects initial semantic processing.

References:

Bates, E., Friederici, A., Wulfeck, B., & Juarez, L. A. 1988. On the preservation of word order in aphasia. Brain and Language 35:323-364.

Bates, E. & Goodman, J.C. 1999. On the emergence of grammar from the lexicon. In B. MacWhinney (ed), The Emergence of Language. Mahwah NJ: Lawrence Erlbaum Associates. pp. 29-80.

Bever, T.G. 1970. The cognitive basis for linguistic structures. In J. Hayes (ed.), Cognition and the Development of Language. New York: Wiley.

Goldberg, A. (1999). The emergence of the semantics of argument structure constructions. In Brian MacWhinney (ed), The Emergence of Language. Mahwah NJ: Lawrence Erlbaum Associates. pp. 197-212.

Gottfried, M., Menn, L., & Holland, A. 1997. Verb-argument structure and aphasic repetition. Poster presented at the 1997 meeting of the Academy of Aphasia, Philadelphia.

Kegl, J. 1995. Levels of Representation and Units of Access Relevant to Agrammatism. Brain and Language 50:151-200.

MacWhinney, B. 1987. The competition model. In B. MacWhinney (ed.) Mechanisms of Language Acquisition. Hillsdale, NJ: Erlbaum. pp. 249-308.

Menn, L., Reilly, K.F., Hayashi, M., Kamio, A., Fujita, I, & Sasanuma, S. 1998. The interaction of preserved pragmatics and impaired syntax in Japanese and English aphasic speech. Brain and Language 61: 183-225.