The original paper on the multiple and context-independent access of word meanings during sentence comprehension is:

Swinney, David A. (1979). Lexical access during sentence comprehension: (Re)consideration of context effects. Journal of Verbal Learning & Verbal Behavior 18 (6): 645-659

Abstract: Two experiments with a total of 228 undergraduates examined the effects of prior semantic context on lexical access during sentence comprehension. In both studies, Subjects comprehended auditorily presented sentences containing lexical ambiguities and simultaneously performed a lexical decision task on visually presented letter strings. Lexical decisions for visual words related to each of the meanings of the ambiguity were facilitated when these words were presented simultaneous with the end of the ambiguity (Exp I). This effect held even when a strong biasing context was present. When presented 4 syllables following the ambiguity, only lexical decisions for visual words related to the contextually appropriate meaning of the ambiguity were facilitated (Exp II). Arguments are made for autonomy of the lexical access process, and a model of semantic context effects is offered.

Other revelant early papers:

Kintsch, Walter and Ernest F. Mross (1985). Context effects in word identification. Journal of Memory and Language 24 (3): 336-349

Abstract: Hypothesized that sense activation in word identification is affected by associative relationships among words, but not by the thematic context of a discourse. Exp I employed a cross-modal lexical decision task. 347 undergraduate Subjects listened to a discourse containing a target word and made a word/nonword decision to a visually presented test string. Results demonstrate that if the target word was a homograph, test words that were associates of the homograph were primed irrespective of the thematic context. Thematically appropriate test words that were not associatively related to the target word were not primed. This result was confirmed in Exp II, where the text was presented visually at a rapid rate. In contrast, when Subjects were given enough time to process each word in Exp III, only thematically appropriate associates were primed. No priming effects at all were obtained in Exp IV, which employed a rapid presentation rate where the test word was separated from the target word by two other interfering words. It is concluded that sense activation functions as a module independent of thematic context.

Onifer, William; Swinney, David A. (1981). Accessing lexical ambiguities during sentence comprehension: Effects of frequency of meaning and contextual bias. Memory & Cognition 9 (3): 225-236.

Abstract: Examined the exhaustive access and the terminating ordered search hypotheses of the nature of lexical access in 2 studies using a cross-modal lexical priming task. 104 undergraduates listened to sentences biased toward the primary interpretation (a meaning occurring 75% or more of the time) or a secondary interpretation (a meaning occurring less than 25% of the time) of a lexical ambiguity that occurred in each sentence. Simultaneously, Subjects made lexical decisions about visually presented words. Decisions were facilitated when presented immediately following occurrence of the ambiguity. However, when presented 1.5 sec following occurrence of the ambiguity, only visual words related to the contextually relevant meaning of the ambiguity were facilitated. Results support the exhaustive access hypothesis. It is argued that lexical access is an autonomous subsystem of the sentence comprehension routine in which all meanings of a word are momentarily accessed, regardless of the factors of contextual bias or bias associated with frequency of use.

Seidenberg, Mark S., Michael K. Tanenhaus, James M. Leiman, and Marie Bienkowski (1982). Automatic access of the meanings of ambiguous words in context: Some limitations of knowledge-based processing. Cognitive Psychology 14 (4): 489-537

Abstract: 188 undergraduates processed ambiguous words in sentences in 5 experiments. Two classes of ambiguous words (noun-noun and noun-verb) and two types of context (priming and nonpriming) were investigated using a variable stimulus onset asynchrony priming paradigm. Priming contexts contain a word highly semantically or associatively related to one meaning of the ambiguous word; nonpriming contexts favor one meaning of the word through other types of information (e.g., syntactic or pragmatic). In nonpriming contexts, Subjects consistently accessed multiple meanings of words and selected one reading within 200 msec. Lexical priming differentially affected the processing of subsequent noun-noun and noun-verb ambiguities, yielding selective access of meaning only in the former case. Results suggest that meaning access is an automatic process that is unaffected by knowledge-based ("top down") processing. Whether selective or multiple access of meaning is observed largely depends on the structure of the ambiguous word, not the nature of the context.

Related studies in brain studies using event-related potentials:

Van Petten, Cyma and Marta Kutas (1988). Tracking the Time Course of Meaning Activation. Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence. Edited by Steven L. Small and Garrison W. Cottrell and Michael K. Tanenhaus. San Mateo, CA: Morgan Kaufmann.

Van Petten, Cyma and Marta Kutas (1987). Ambiguous words in context: An event-related potential analysis of the time course of meaning activation. Journal of Memory & Language 26 (2): 188-208

Abstract: Presented 75 Subjects (aged 18-25 yrs) with sentences using words with a single spelling and pronunciation but at least 2 distinct meanings (homographs) terminating sentences of moderate contextual constraint. Target words were (1) related to the contextually biased meaning of the homograph, (2) related to the unbiased meaning, or (3) unrelated. The stimulus onset asynchrony (SOA) between homograph and target was short (200 msec) or long (700 msec). The naming latencies in Exp I and the event-related potentials elicited in Exp II showed a similar pattern of priming at the long SOA. At the short SOA, the priming effect had a later onset for contextually inappropriate than appropriate targets, indicating that both meanings were not activated at the same time.

Van Petten, Cyma and Marta Kutas (1991). Electrophysiological evidence for the flexibility of lexical processing. Understanding Word and Sentence. Series Advances in Psychology 77. Edited by Greg B. Simpson. Amsterdam: North-Holland. 129-174.

An intersting paper which addresses the question of "backward priming" (i.e. asks whether there may have been a flawed methodology in these earl studies):

Michael K. Tanenhuas and Curt Burgess and Mark Seidenberg (1988). Is Multiple Access an Artifact of Backward Priming? Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence. Edited by Steven L. Small and Garrison W. Cottrell and Michael K. Tanenhaus. San Mateo, CA: Morgan Kaufmann. 311--330.

A partial rebuttal:

Jones, Janet L. (1989). Multiple access of homonym meanings: An artifact of backward priming? Journal of Psycholinguistic Research 18 (4): 417-432.

Abstract: 42 undergraduates participated in a color-naming task to determine whether multiple homonym meanings are accessed independently of context. Subjects heard sentences ending in homonyms, then, either 0 or 200 msec later, saw target words that were appropriately related, inappropriately related, or unrelated to the preceding homonym. Results support the prediction that color-naming responses to both appropriate and inappropriate targets would be inhibited relative to unrelated targets at the 0-msec interstimulus delay. Within 200 msec, inappropriate targets were no longer inhibited, indicating that context had acted to select the appropriate meaning. Because the color-naming task eliminates backward priming, the multiple access effect obtained in this study cannot be an artifact of backward priming.

For the most recent survey of the field, see

Simpson, Greg B. (1994). Context and the Processing of Ambiguous Words. Handbook of Psycholinguistics. Edited by Morton Ann Gernsbacher. San Diego, CA: Academic Press. 359-374.

Abstract: (from the chapter) provide an updated review of recent literature in (lexical) ambiguity (the retrieval of meanings in multiple meaning words), some discussion of methodological issues, and a perspective on the future of the area; argue that the box score approach (placing studies old and new into an appropriate column) to the modularity debate (on how ambiguous words were processed) has outlived its usefulness, and that researchers in this area need to move beyond this basic question to discuss issues concerning the nature of context and a person's interaction with it, and how these may affect processing... review some of the issues that were raised in the earlier discussion of the literature; consider the bearing of recent research on those issues; discusses the nature of context and suggests that further progress will be possible only by viewing context more comprehensively than has typically been the case, considering not only local sentence context but also the context of the experimental situation itself.

Further questions about the experimental design of the multiple access studies, moving towards a minimalist inference view::

McKoon, Gail & Roger Ratcliff (1992).  Inference during reading. Psychological Review 99 (3): 440-466.

Abstract: Most current theories of text processing assume a constructionist view of inference processing. In this article, an alternative view is proposed, labeled the minimalist hypothesis. According to this hypothesis, the only inferences that are encoded automatically during reading are those that are based on easily available information, either from explicit statements in the text or from general knowledge, and those that are required to make statements in the text locally coherent. The minimalist hypothesis is shown to be supported by previous research and by the results of several new experiments. It is also argued that automatically encoded minimalist inferences provide the basic representation of textual information from which more goal-directed, purposeful inferences are constructed.

For a more general introduction to recent thinking, see also:

Graesser, A., K. K. Millis, & R.A. Zwaan (1997).  Discourse comprehension. Annual Review of Psychology 48: 163-189.

Abstract: Discusses the meaning representations that are constructed when adults read written text, such as literary stories, technical expository text, and experimenter-generated "textoids." Three phenomena that have been extensively investigated by discourse psychologists are examined: the processing of referring expressions, the connection of statements in text, and the encoding of knowledge-based inferences. Readers execute these processes in an effort to achieve coherence at local and global levels and to explain why information is mentioned in the text. These three phenomena are discussed in relation to multiple levels of discourse representation, psychological mechanisms in theories of comprehension, referring expressions, and knowledge-based inferences.

The final two citations were provided by David S. Miall (David.Miall@Ualberta.ca); the others mostly by:
 

Dan Jurafsky (jurafsky@colorado.edu)	URL: http://stripe.colorado.edu/~jurafsky
Department of Linguistics, Institute of Cognitive Science, & Department of Computer Science
University of Colorado at Boulder	PHONE: (303)492-1300
Boulder, CO 80309-0295	FAX: (303)492-4416