Adapted from Osterhout, L., Bersick, M., & McLaughlin, J. (1997). Brain potentials reflect violations of gender stereotypes. Memory and Cognition, 25, 273-285.

Humans are able to hear strings of words presented one at a time in a sentence and move beyond the meaning of each separate word to understand what the sentence means. One tool that researchers use to study how we are able to understand sentences is the record of electrical activity recorded via scalp electrodes while a research participant reads sentences. Event related brain potentials (ERPs) are scalp-recorded voltage changes that are time-locked to an event (in this case, the event is the presentation of a particular word). Researchers ask participants to read sentences displayed one word at a time on a computer screen. Researcher record electrical activity time-locked to word presentation and compare ERPs for words in different types of sentences. Records for many trials are averaged so that random electrical activity washes out and the brain activity associated with the event remains. Common practice is to record from electrodes placed at a number of different spots on the scalp. (For more detail see What are ERPs?). Both structural issues such as grammar problems and semantic issues such as word meaning confusions are known to influence ERP results.

Figure 1. A participant prepared for ERP recording.

The present research program sought to answer the question -- Do gender role stereotypes about an occupation influence the way we process words in sentences? Most college students will acknowledge that both women and men may be firefighters, nurses, housekeepers, truck drivers, dental hygienists, and the like. BUT - Will the brain process sentences that suggest the less typical gender for an occupation (e.g., female firefighter) in the same way as sentences that suggest the more typical gender (e.g., male firefighter)?

Prior ERP findings showed that pronouns that disagreed in gender with a noun antecedent (The bachelor cooked herself dinner) elicited bigger positive ERPs than pronouns that agreed in gender with a noun antecedent (The bachelor cooked himself dinner). In this case, the sentence with matching noun antecedent and pronoun is an acceptable English sentence but the sentence with the mismatched noun antecedent and pronoun is not. The violation of agreement between noun and pronoun led to an increase in the magnitude of the ERP that was positive in polarity and peaked around 600 milliseconds after the pronoun was presented, described as a larger P600 effect. The present study also tested differences in the magnitude of P600 effects for matched and mismatched nouns and pronouns. In addition, researchers added a new question. Would differences occur for sentences than included an occupation name followed later in the sentence with a stereotypical pronoun versus sentences with the same occupation name followed later in the sentence with a non-stereotypical pronoun (The firefighter ate his lunch quickly versus The firefighter ate her lunch quickly)? In the example, both sentences are grammatically acceptable English sentences. The important question is whether the ERPs elicited by his versus her will differ.

Native-English speaking undergraduates (14 men and 14 women) volunteered for the experiment. Participants read sentences presented one word at a time on a computer screen. At the end of each sentence, participants pressed a YES key if they judged the sentence an acceptable English sentence or a NO key if they judged the sentence an unacceptable English sentence. Participants were told that acceptable sentences should be semantically coherent and grammatically well formed. Researcher compared ERPs for four different types of sentences with every participant seeing many examples for each type of sentence.

The four types of sentences were (critical word emphasized here in red but of course not displayed that way in the experiment):

Click here to see an example of a sentence as it was actually presented.

As expected, participants rated most sentences like (1), (3) and (4) acceptable and rated most sentences like (2) unacceptable.

The research findings replicated the findings from prior research. As expected, there was a larger amplitude P600 effect for sentence-type (2) where the noun and pronoun did not agree than for sentence-type (1) where the noun and pronoun did agree. Similar differences, though not as large, occurred when the amplitudes of ERPs for sentence-type (3) and sentence-type (4) were compared. The P600 effect was bigger for the pronoun that did not reflect the typical gender for an occupation than for the pronoun that did reflect the typical gender stereotype. The figure below shows the averaged records (averaged across many trials and all the participants) for one of the different sites for scalp electrodes. The vertical bar at the start indicates the onset of the presentation of the pronoun in each condition. Each hashmark along the axis indicates 100 milliseconds. Positive voltage is reflected below the axis (by convention).

Figure 2. Grand-average ERPs (averaged over all participants and trials) elicited by reflexive pronouns that agreed or disagreed with the gender of definitionally or stereotypically male or female antecedent nouns.

Statistical analysis conducted on the mean ERP amplitude averaged between 500 and 800 milliseconds after pronoun presentation confirmed these observations

In sum, the researchers confirmed the prior finding that ungrammatical sentence with mismatched nouns and pronouns elicited a larger P600 effect than grammatical control sentence with matched nouns and pronouns. They also extended the previous finding to show that grammatically-correct sentences with violations of presumed occupational gender stereotypes elicit a larger P600 effect compared to sentences with pronouns that match gender stereotypes. Clearly, the participants' brains "noticed" the stereotype-violations in these grammatically-correct sentences. The researchers suggested that ERPs may prove to be an excellent direct measure of stereotypical beliefs, as they do not rely on the participants' conscious self reports and are not influenced by pressures to produce socially-desirable answers.

Questions.

1. How did prior research lead the researchers to ask the questions that they asked?

2. Did the researchers use between or within subject assignment to the four different conditions?

3. How might the four conditions described here be conceptualized as a 2 x 2 factorial design (the design the researchers used) ?

4. Suppose you had been asked to write the sentences used as stimuli for this experiment. You know that it is important to use both versions of a sentence as stimuli (e.g., The firefighter ate her lunch and The firefighter ate his lunch) but also know that two versions of the same sentence should not be presented to the same participant. How would you proceed?

5. The researchers retested two conditions used in a prior experiment (matched noun/pronoun and mismatched noun/pronoun). Testing participants and analyzing their data is time and labor intensive. So why bother retesting two conditions for which expected results are known?

Click here to read a summary of the gender-stereotype experiment written by researcher Lee Osterhout. It includes some thought-provoking additional results about difference in findings for male and female participants.

Individuals with a University of Washington NETID may wish to read a recent research article that addresses a similar topic and is available on line. See the June 1999 Issue of Language and Cognitive Processes.
Osterhout, L., & Nicol, J. (1999). On the distinctiveness, independence, and time course of the brain responses to syntactic and semantic anomalies. Language and Cognitive Processes, 14, 283-317.

If you would like to learn more about Dr. Osterhout's current research projects go to: Lee Osterhout