Generalizable distributional regularities aid fluent language processing: The case of semantic valence tendencies

Sentence processing is an extraordinarily complex and speeded process, and yet proceeds, typically, in an effortless manner. What makes us so fluent in language? Incremental models of sentence processing propose that speakers continuously build expectations for upcoming linguistic material based on partial information available at each relevant time point. In addition, statistical analyses of corpora suggest that many words entail probabilistic semantic consequences. For instance, in English, the verb provide typically precedes positive words (eg,‘to provide work’) whereas cause typically precedes negative items (eg,‘to cause trouble’; Sinclair, 1996). We hypothesized that these statistical patterns form units of meaning that imbue lexical items, and their argument structures, with semantic valence tendencies (SVTs), and that such knowledge assists fluent on-line sentence comprehension by facilitating the predictability of upcoming information. First, a sentence completion task elicited such tendencies in adults, suggesting that speakers constrain their free productions to conform to the connotative meaning of words. Second, fluent on-line reading was slowed down significantly in sentences that contained a violation of a valence tendency (eg cause optimism). Third, an automated computer algorithm assessed the pervasiveness of valence tendencies in large computerized samples of English, supporting the hypothesis that valence tendencies are a distributional phenomenon. We conclude that not only can aspects of meaning be modeled with word cooccurrence statistics, but that such statistics are likely to be computed by