Brain potentials reflect violations of gender stereotypes

LEE OSTERHOUT 0 1 MICHAEL BERSICK 0 1 JUDITH McLAUGHLIN 0 1 0 We thank Tony Greenwald, Laurie Rudman. Sean Draine, Rick McKinnon. Vicka Corey, Michael Masson, and two anonymous re viewers for helpful comments on previous versions of this paper; David Hessl and Michael Backstrom for carrying out the ratings pretest; Dawn Aiken and Linda Mobley for help with stimulus construction; and especially Marjorie Osterhout for lively discussions leading to this experiment. This research received financial support from Grant R29 DCa 1947 from the National Institute of Deafness and Other Commu nication Disorders. National Institutes of Health. Please address corre spondence and reprint requests to L. Osterhout. Department of Psy chology , Box 351525, University of Washington , Seattle. WA 98195 ( 1 University of Washington , Seattle, Washington Event-related brain potentials (ERPs) were recorded while 14 males and 14 females read sentences containing a reflexive pronoun that referred to a definitionally or stereotypically male or female antecedent noun, Pronouns that disagreed with the gender definition or gender stereotype ofthe antecedent elicited a large-amplitude positive wave. Violations of gender definitions elicited a larger positive wave than did violations of gender stereotypes. Furthermore, the positive wave elicited by stereotype violations persisted even when subjects judged these sentences to be acceptable. Finally, female subjects exhibited larger positivities than did male subjects, regardless of whether the gender mismatch involved a definitional or stereotypical antecedent. These results are taken to indicate that ERPs are sensitive to violations of gender-based occupational stereotypes and that the ERP response to stereotype violations is similar to the P600 effect elicited by a variety of syntactic anomalies. - Recent work has indicated that syntactic and pragmatic anomalies encountered during reading elicit distinct changes in the event-related brain potential (ERP). ERPs are scalp-recorded voltage changes in the electroenceph alogram that are time-locked to the onset of a sensory, motor, or cognitive event (for a review, see Rugg & Coles, 1995, or Hillyard & Picton, 1987). ERPs consist of pos itive and negative voltage deflections (or "components") that are distributed over time. Some ofthese components are sensitive to language-related events. In a series of seminal experiments, Kutas and Hillyard (1980a, 1980b, 1980c) discovered that pragmatically implausible words (e.g., "I like my coffee with cream and dog") elicit an en hanced negative-going wave with a peak amplitude at about 400 msec (the N400 effect; see Kutas & Van Petten, 1994). More recently, other researchers (Hagoort, Brown, & Groothusen, 1993; Neville, Nicol, Barss, Forster, & Gar rett, 1991; Osterhout, 1990, in press; Osterhout & Hol comb, 1992, 1993; Osterhout, Holcomb, & Swinney, 1994; Osterhout, McKinnon, Bersick, & Corey, 1996; Oster hout & Mobley, 1995; Osterhout & Nicol, 1996; for a re view, see Osterhout, 1994) have reported that a disparate set of syntactic anomalies (including anomalies involv ing phrase structure, verb subcategorization, verb tense, and constituent movement) elicits a large positive wave with an onset at about 500 msec and a duration of several hundred milliseconds. This positive wave has been vari ously labeled the P600 effect (Osterhout & Holcomb, 1992) and the syntactic positive shift (Hagoort et al., 1993). Importantly, the P600 effect is qualitatively distinct from the N400 effect.' Of particular relevance to the present study is evidence that violations of agreement between sentence constitu ents elicit a P600-like positivity (Hagoort et al., 1993; Osterhout, in press; Osterhout et al., 1996; Osterhout & Mobley, 1995). In current grammars, agreement is incor porated within a theory of syntax (Haegeman, 1991). Number, gender, and other agreement features are repre sented as a feature bundle residing under an inflectional node within the sentence. These features are mechani cally copied from the "controller" (e.g., nouns in subject position) to the "controllee" (e.g., tensed verbs and pro nouns) in a purely formal manner. Consistent with this treatment, a variety of agreement violations, including violations of subject-verb number agreement (e.g., "Most cats likes to play outside"), reflexive pronoun-antecedent number agreement (e.g., "The hungry guests helped him seljto the meal"), and reflexive pronoun-antecedent gen der agreement (e.g., "The woman blamed himseljfor the accident"), elicit a large-amplitude positive wave that is quite similar to the P600 effect elicited by a variety of syn tactic anomalies (Hagoort et al., 1993; Osterhout, in press; Osterhout & Mobley, 1995; Osterhout et al., 1996). The present study was primarily motivated by one find ing reported by Osterhout and Mobley (1995)-namely, that reflexive pronouns disagreeing in gender with a de finitionally male or female antecedent elicit a P600-like effect. In the research reported here, we examined the brain response to reflexive pronouns that refer to stereo typically male or female antecedent nouns.? Consider a sentence containing a reflexive that is inconsistent with the presumed gender bias of a noun indicating a stereotyped occupational role (e.g., "The doctor prepared her self for the operation"). Because the reflexives in such sentences would be perceived to be anomalous only if the occupational stereotypes exist, the presence of a "lin guistic anomaly" effect (i.e., the N400 or P600) in the ERPs to these reflexives would indicate that subjects (and ERPs) are sensitive to stereotype violations. Furthermore, the qualitative nature of the anomaly effect might pro vide an initial basis for speculating about the mental rep resentations and processes underlying the response to the stereotype violations. One possibility is that the per ceived anomaly results from an evaluation of the prag matic plausibility of a male playing a stereotypically fe male role or of a female playing a stereotypically male role. If so, stereotype violations might be expected to elicit a brain response similar to that elicited by prag matically inappropriate words (N400). Alternatively, the gender properties of stereo typically male or female nouns might be represented in much the way that the gender properties of definitionally male or female nouns are thought to be represented-as a grammatical feature as sociated with the word's lexical representation. Under this scenario, stereotype-violating reflexives might be per ceived as violating the grammatical constraint requiring feature agreement between the reflexive and its ante cedent, and might therefore elicit a brain response simi lar to that elicited by agreement violations and other syn tactic anomalies (P600). To summarize: the present study was designed to in vestigate two questions. First, is the ERP response to stereotype-violating reflexives distinct from (...truncated)