JOURNAL OF MEMORY AND LANGUAGE 29,469-492 (I!?%) Measuring Inhibition and Facilitation from Pronouns MARYELLEN C.MACDONALD Massachusetts Institute of Technology AND BRIAN MACWHINNEY Carnegie Mellon University Two cross-modal experiments investigated changes in activation levels for pronominal referents and nonreferents. Subjects heard sentences with and without pronouns and re- sponded to visual probe words that appeared at variable intervals during sentence presen- tation. The gender of potential referents was controlled in Experiment 1 so that the pronoun unambiguously referred to one sentence participant, and Experiment 2 contained both un- ambiguous and ambiguous reference conditions. In both experiments, responses to probes corresponding to nonreferents were slower in the presence of an unambiguous pronoun compared to the no-pronoun condition, suggesting that pronouns inhibit nonreferents. Re- sponses to referents were speeded by unambiguous pronouns compared to the no-pronoun condition in Experiment 2. Neither of these effects appeared at probe positions immediately after the pronoun, indicating that assignment of a referent to an unambiguous pronoun took a measurable amount of time, and a substantially longer amount for an ambiguous pronoun. An additional probe condition in Experiment 1 indicated that these effects were not due to differences in overall processing load. These results were interpreted in light of discourse shifts from establishment of pronominal reference, and in terms of the sensitivity of probe response tasks to a variety of influences. 0 tsso ~~aderni~ press, fnc. An important issue in the study of pro- noun comprehension is the questionof how a pronounaffectsthe mental representation of its referent. Typically, investigationsof the effect of pronounson mental represen- tations have presentedsentences contain- ing a referent and a pronoun, such as John went to the store and he bought cookies, and then measured responses to a visual These experiments were conducted while the first author was supported by Sloan Foundation Grant #131071 to the Department of Psychology, Carnegie Mellon University. We are grateful to Bruce Chapman and Leonid Spektor, who wrote the digitizing and stimulus presentation programs for both experiments, and to Joe Danks, Morti Gemsbacher, Cathy Harris, Marcel Just, Janet McDonald, Mark Seidenberg, and several anonymous reviewers for helpful comments on an earlier version of the manuscript. Correspondence should be sent to Maryellen MacDonald, Department of Brain and Cognitive Sciences, ElO-034C, Massa- chusetts Institute of Technology, Cambridge, MA 02139. E-mail: mcm@psyche.mit.edu. probe word appearing during or after pre- sentation of the sentence.On some trials the probe corresponds to the referent, in this case John, while on other trials the probe is someother word (that is, a baseline condition). Studies using a variety of re- sponse taskshave shown that, compared to subjects’ performancein the baselinecon- dition, subjects’ responses are faster to a probe corresponding to the referent of a pronoun. This difference between referent and baseline conditions hasbeenfound in a probe namingtask (Leiman, 1982), in a lex- ical decision task (Cloitre & Bever, 1988; Nicol, 1988), andin a task in which subjects judge whether the probe word had ap- peared in the sentence (Chang, 1980; Cloitre 8z Bever, 1988; Emmorey, Norman, & O’Grady, 1989; Gernsbacher, 1989; Stevenson, 1986). Thesefindings havebeen interpreted to indicate that the processes that form a link betweena pronoun and its 469 0749-5%x/90 $3.00 Copyri& 0 19!Xl by Academic Press, Inc. All rights of reproduction in any form reserved.