Processing Temporal Constraints: an Erp Study

This paper investigates how linguistic expressions of time, in particular temporal adverbs and verb tense morphemes, are used to establish temporal reference in the brain. First, a semantic analysis of tense and temporal adverbs is outlined. It is proposed that computing temporal reference amounts to solving a constraint satisfaction problem. Further, the results of an ERP study are presented which suggest that violations of verb tense (‘Last Sunday Vincent paints the window frames of his country house.’) result in larger left-anterior negative (LAN) effects, emerging around 200 ms after verb onset. Finally, the semantic analysis is combined with a computational model of parsing to provide a functional account of the ERP data. Reference to time is ubiquitous in natural language, to the point that nearly every assertion involves the location of some event within a temporal coordinate system. Much attention has been paid to the linguistic structures used to encode temporal information, which range from prepositional phrases (‘before the dawn’) to verb suffixes (‘-ed’ in the English regular simple past), to expressions borrowed from technical languages, for instance mathematics (‘10−35 seconds into the expansion phase’). Linguists have provided precise accounts of the functioning of these devices, often using the formal tools made available by modern logic. However, comparatively little is known about how expressions of time are represented and used to compute temporal reference in the brain. My aim here is to contribute to this line of research by combining a formal analysis of tense with ERP data on tense processing. 1. AN ANALYSIS OF TENSE AND TEMPORAL ADVERBS In this section I shall be concerned with a linguistic analysis of sentences in which tense is correctly or incorrectly used, such as: (1) a. Last Sunday Vincent painted the window frames of his country house. b. Last Sunday Vincent paints the window frames of his country house.* These sentences recognizably form a minimal pair of the kind customarily used in language processing experiments. The formal treatment of tense outlined below is intended to drive the interpretation of the ERP data presented in section 2. The main theoretical claim will be that processing sentences such as (1a) involves solving a constraint satisfaction problem in PROCESSING TEMPORAL CONSTRAINTS: AN ERP STUDY 3 which the relevant temporal constraints are introduced by the semantics of the adverb ‘last Sunday’ and of the verb ‘painted’. Conversely, processing a sentence such as (1b) leads to a failure to solve simultaneously such constraints. Even though the analysis developed here brings semantics center stage, tense has ramifications in morphology which can hardly be ignored. Before I turn to semantics, some notes on morphology are therefore in order. 1.1. Morphology. Comrie (1985) defined verb tense as the “grammaticalization of location in time”. Unpacked, this definition reads that tense is concerned with the location of events in time – or, possibly, a cognitive representation of time – and that the expression of time is (i) obligatory and (ii) morphologically bound on verbs. Note, however, that this can only be taken to imply that verbs obligatorily carry temporal information via morphemes, and not that there are morphological rules which determine, for any given verb form, whether tense is correctly used. Let me try to clarify this point with reference to examples (1). The verb forms ‘painted’ in (1a) and ‘paints’ in (1b) convey, via the suffixes ‘-ed’ and ’-s’, the information that the painting event is located, respectively, in the past and in the present of the moment of speech. Importantly, both verb forms are morphologically correct, in the sense in which, for instance, the over-regularized verb form ‘goed’ (past of ‘to go’) is not. What makes ‘paints’ incorrect in the context of (1b) is a mismatch between its semantics, that is, the fact that it refers (at least by default) to a present event, and that of the adverb ‘last Sunday’, which defines a temporal window in the past of the moment of speech in which the main event is to be located. From these considerations, two views on the processing consequences of the violation in (1b) ensue. On the first, (1b) is perceived by the language system as a semantic violation, and not as a morphological one, because, as we have seen, ‘paints’ is in fact morphologically well-formed. On the second, ‘paints’ in (1b) is perceived as a morphological anomaly, even though its origin is semantic. This option entails the additional hypothesis that the system is endowed with an interface component (Jackendoff 2002) mediating between semantics