My research investigates the interface between language comprehension, language production, and memory representations. Though these three areas are diverse, a key theme throughout the research is the role of past experiences with particular language patterns on future comprehension abilities and production choices. This focus on distributional patterns of previous language use provides a link between adult language use and language acquisition.
The Production-Distribution-Comprehension (PDC) Account of Language Processing
When people perceive language, their interpretations of what they read or hear follow certain well-known patterns: certain kinds of sentences are harder than others, certain kinds of ambiguities are unconsciously interpreted in one way instead with other possible interpretations. Language researchers have typically viewed these biases in interpretation as emerging directly from biases built into people's language comprehension processes. For example, people tend to interpret "Mary said that John left yesterday" to mean that John's leaving was yesterday, not that Mary's saying was yesterday. To many language researchers, this result reflects something deep about the language comprehension system, and they have accounted for this interpretation bias by hypothesizing that the language comprehension system is designed to have expressions like "yesterday" modify the nearest element, in this case "left," rather than more distant elements, such as "said."
The PDC rejects the idea that comprehension
biases like this necessarily reflect inherent biases in the comprehension system. The PDC suggests instead that these biases reflect people's learning about the statistics of their language. For example, people interpret "Mary said that John left yesterday" to mean that the leaving event was yesterday because that interpretation (modifying the closest verb) is the most frequent interpretation of sentences of this sort that people hear and read in English. And the reason why these interpretation patterns exist, according to the PDC, is that there are pressures on the language production system that cause people to say sentences in a certain way. On this view, the comprehension biases are not some hardwired part of the comprehension system but are rather learned from patterns in the input. These patterns, in turn, stem from constraints on production.
The PDC account requires integrating research in production, learning and acquisition, and comprehension. Our work is described in several recent papers. Except where noted
the papers can be found on our lab publications page, http://lcnl.wisc.edu/publications/
The hypothesis about sentences like "Mary said that John left yesterday" was first described in MacDonald (1999). MacDonald and Thornton (2008) tested the hypothesis with both corpus analyses of natural language and comprehension studies.
MacDonald, M. C. (1999). Distributional information in language comprehension, production, and acquisition: Three puzzles and a moral. In B. MacWhinney. (Ed.), The Emergence of Language (pp. 177-196). Mahwah, NJ: Erlbaum.
MacDonald, M.C. & Thornton, R. (2008). When language comprehension reflects production constraints: Resolving ambiguities with the help of past experience. MANUSCRIPT SUBMITTED; INQUIRE ABOUT AVAILABILITY.
Gennari and MacDonald (2006) discussed
some PDC implications with respect to child language acquisition. We argued that some classic acquisition results, which have been interpreted to indicate the children's innate linguistic knowledge, instead reflect children's learning of the sentence patterns that adults say.
Gennari, S. P., & MacDonald, M. C. (2006). Acquisition of negation and quantification: Insights from adult production and comprehension. Language Acquisition, 13,
125-168.
Wells, Christiansen, Race, Acheson, and MacDonald (in press) investigated the learning portion of the PDC. They manipulated the language experiences of two different participant
groups over the space of about a month and found effects of experience on language comprehension patterns. In both people and in the computational model we tested, extra experience does not help all kinds of sentences equally. We trace these effects to how the language is structured and how people learn about its words and structures.
Wells, J. B., Christiansen, M. H.,
Race, D. S., Acheson, D. J., & MacDonald, M. C. (in press). Experience and sentence comprehension: Statistical learning and relative clause comprehension. Cognitive Psychology.
Gennari and MacDonald (2008a) and (2008b) pursue a PDC approach to explain how people interpret relative clauses. The (2008a) paper shows how comprehension difficulty correlates with people's expectations of how the sentence will turn out—when there is uncertainty, comprehension is more difficult. This claim is
actually quite novel, because most researchers believe that relative clauses are not ambiguous and have a very different account of why they
are difficult. The (2008b) paper traces these comprehension effects back to speakers' choices when they're producing relative clauses, as predicted by the PDC.
Gennari, S. P., & MacDonald, M. C. (2008a). Semantic indeterminacy in object relative clauses. Journal of Memory and Language, 58, 161-187.
Gennari, S.P. & MacDonald, M.C. (2008b). Linking Production and Comprehension Processes: The Case of Relative Clauses. MANUSCRIPT SUBMITTED; INQUIRE ABOUT AVAILABILITY.
Language Production and Working Memory
An important claim of the PDC is that
production pressures shape language producers' structure choices. One factor we're addressing is the role of working memory constraints on production choices.
We don't view working memory in the traditional way (that it's a system for holding information for a brief period). Instead, we see verbal working memory as actually part of the language production system. A review of that position is described in Acheson & MacDonald (in press):
Acheson, D. J., & MacDonald, M. C. (in press). Verbal Working Memory and Language Production: Common Approaches to the Serial Ordering of Verbal Information. Psychological
Bulletin.
This paper offers a different way of thinking about verbal working memory. We also turn this around and ask classic effects in working memory, such as interference between items that are going to be recalled/spoken, could explain some of the
choices speakers make during language production.
Learning within Language Production
The PDC emphasizes the role of learning in shaping language comprehension processes. People also learn about their language through production, and we investigate how people's previous learning about the statistics of a language influences future productions. So far, most of our investigations of learning and production have addressed how learning shapes the production of agreement between nouns and verbs, so that people say "The dog is sleeping" vs. "The dogs are sleeping." Thornton and MacDonald (2003) showed that plausible noun-verb relationships (presumably learned from experience) affect how people produce agreement, and Haskell and MacDonald (2005) showed how people's knowledge of word order also influences agreement. Haskell, Thornton & MacDonald (2008) investigate learning most directly in that they manipulate people's
agreement experiences and show how learning from these experiences shape subsequent language productions. Future work will tie this agreement work back to the structure choice work in the PDC by investigating how production of agreement and choice of syntactic structure influence each other.
Haskell, T. R., & MacDonald, M. C. (2005). Constituent structure and linear order in language production: Evidence from subject verb agreement. Journal of Experimental Psychology:
Learning, Memory and Cognition, 35, 891-904.
Haskell, T. R., Thornton, R., & MacDonald, M. C. (2008). The Role of Experience in Agreement Production. MANUSCRIPT SUBMITTED; INQUIRE ABOUT AVAILABILITY.
Thornton, R. & MacDonald, M. C. (2003). Plausibility and grammatical agreement. Journal of Memory and Language, 48, 740-759.
Other Work in Progress
Several other projects are ongoing but without finished papers. Here are a few of them.
Crosslinguistic studies of Production.
A better understanding of the forces that prompt speakers to choose different syntactic structures has led us to study how people produce sentences in English and other Indo-European languages and also in Asian languages such as Japanese and Korean. Though much of this work is in its early stages, we see amazing similarities between how speakers of different languages (with different word orders) seem to settle on syntactic structures to convey their ideas.
Artificial Grammar Learning and Sentence Comprehension. We investigate learning using artificial grammar learning, as simple artificial grammars can sometimes yield a clearer picture of learning than when studying the complexities of natural language.
Learning in the world vs. in language.
Using both natural and artificial languages, we ask about what exactly people are learning—are they learning about events in the world (e.g., how plausible it is for a cop to arrest someone), co-occurrences in language (e.g., the frequency that the word cop is the subject
of the verb arrest), or both?
