Information Sheet. While participants were required to provide their age, school of enrolment, pastimes, and goals in life in a questionnaire, its presentation to participants was primarily to reinforce the belief that participants would be exchanging such knowledge with their partner in the study. For subsequent analyses the age of the participant was the only measure of interest.
Implicit Knowledge Measures.
Others' Views Checklist. This measure employed by Foddy (1997) aims to tap knowledge of the content of gender stereotypes. It contains a set of traits that through pre-testing, were found to be rated as male and female and further were found to be competence-related. In Foddy's study, male and female subscales were found to correlate with each other (r = 0.37, p < 0.05). Moreover, through principal components analysis, these scale items were found to load onto the same factor as free response measures of positive and negative, male and female characteristics.
Participants administered the Others' Views Checklist (OthsVws) were required to indicate what they think others think constitutes a male thing or a female thing on a seven point Likert scale. Responses range from 1 (definitely a female thing) to 7 (definitely a male thing) with 4 (neither male nor female) the neutral midpoint. The eight male traits that were included in the scale were; independent, verbal, educated, competent, intelligent, confident, firm, and acts as a leader; and the six female traits were; empathetic, sensitive, sociable, excitable in a crisis, emotional and dependent.
Word Association Task (WrdTask). This study employed an experimental paradigm based on previously employed paradigms that aim to measure implicit knowledge (Dovidio et al., 1986; Foddy, 1997; Wittenbrink et al., 1997). This paradigm essentially involves presentation of a category label (man, woman, cloud), followed by a trait word, which is associated with the stereotype of men, women, or neither. Once exposed to the word pair (eg., woman-dependent), a response as to the association of the two words is required. In this study participants were presented with one of the following category labels; man woman, or cloud (a neutral, high frequency word). This was presented in the centre of the computer screen in lower case. A stereotypically male, stereotypically female or neutral trait word was presented directly after the category word also in the centre of the computer screen, but this time in upper case. Participants were then required to determine whether or not the trait was descriptive of the category by pressing a key labelled yes or no on a computer keyboard. After presentation of the trait word, the speed at which they responded, their response latency, was measured. This constituted one of the dependent measures.
In this paradigm the latency of response was understood to be an indicator of chronically accessible information. Assuming that stereotypic knowledge was available for activation, and that the category label was cognitively linked to the stereotype, priming with a category label should serve to activate stereotype-related information. Once activated, traits perceived as consistent with the category (eg., woman-emotional, man-assertive) should facilitate shorter response latencies, than those that are inconsistent with the category, or unrelated to it. A neutral category and gender-neutral traits were included so that each participant's overall response latency could be controlled for. The way in which category-trait pairs were considered to be consistent, inconsistent and, in the case of the neutral category and traits, neither consistent nor inconsistent, is illustrated in Table 4.
| Trait Words | |||
|---|---|---|---|
| Category Labels | Male Traits | Female Traits | Neutral Traits |
| Man | Consistent * | Inconsistent | Neither |
| Woman | Inconsistent | Consistent* | Neither |
| Cloud | Neither | Neither | Neither |
* a yes response in these pairings is treated as a correct response. In all other pairings, a no response is required for it to be a correct response.
As is implied by Table 4, an error, or disagreement with the pairing was recorded when participants answered yes to stereotype inconsistent pairs (e.g., woman-independent), or no to stereotype consistent pairs (e.g., woman-emotional). Errors were also recorded when participants answered yes to a pairing of any neutral trait word with either of the gender category labels (e.g., woman-tiring, man-lucky), and when they answered yes to a pairing of any of the three types of traits with the neutral category cloud (e.g., cloud-independent, cloud-emotional, cloud-lucky).
As participants were required to determine whether the category could be described by the trait (a semantic decision), conscious awareness of the pair was necessary (Wittenbrink et al., 1997). Hence, category labels serving as primes were exposed for a period of 500ms, a time frame which has been previously found to enable recognition of the priming word (Augostinous et al., 1994; Foddy, 1997). Furthermore, as the trait word was presented directly after the category word, the category word was therefore backwardly masked which served to reduce the residual effects involved in repeating presentation of the same primes.
To execute this task, a computer program known as dmaster (Forster, personal communication, June 6th, 1998) was employed. This program enabled presentation of the word pairs in a randomised sequence, and also recorded participants' response latencies and errors. In this study, both the category and trait words were selected from previous research on gender stereotyping (Foddy, 1997).
Three different versions of the task were employed (e.g., the 36 male traits comprised 3 sets of 12 traits each). Versions differed according to the set of traits paired with each category label. By using three versions each trait could be paired with each category, without repeating the presentation of traits to the one participant. Multiple versions were also employed to allow subsequent assessment as to whether each of the three sets of the one type of trait elicited similar responses. Table 5 shows how the three versions of the traits were constructed and assigned to the category labels.
| Category Label | Version A | Version B | Version C |
|---|---|---|---|
| Man | M: Set* 1 | M: Set 2 | M: Set 3 |
| F: Set 1 | F: Set 2 | F: Set 3 | |
| N: Set 1 | N: Set 2 | N: Set 3 | |
| Woman | M: Set 3 | M: Set 1 | M: Set 2 |
| F: Set 3 | F: Set 1 | F: Set 2 | |
| N: Set 3 | N: Set 1 | N: Set 2 | |
| Cloud | M: Set 2 | M: Set 3 | M: Set 1 |
| F: Set 2 | F: Set 3 | F: Set 1 | |
| N: Set 2 | N: Set 3 | N: Set 1 |
M = male traits, F = female traits, N = neutral traits
*Each set of traits contains 12 items.
The Sentence Completion Task. This task aims to measure the tendency to explain behaviours incongruent with stereotypic knowledge. In this, participants were required to complete a set of 22 sentence beginnings, that each described a behaviour. Of these, six comprised male and female names paired with stereotype congruent behaviours (e.g., male congruent: Bob confronted the man, female congruent: Sharon asked for help getting home). Another six comprised male and female names paired with stereotype incongruent behaviours (e.g., male incongruent: Geoff sewed the button back on, female incongruent: Helen directed the operation). The remainder described gender-neutral behaviours, (e.g., Laura ate a sandwich).
Two versions of the Sentence Completion Task (SentComp) were employed in this study (refer to Appendix C for one of the versions). This effectively counterbalanced the names paired with the gender stereotyped behaviour, so that if in one version a female name was paired with a behaviour, in the other version, a male name would be paired with this behaviour.
Explicit Beliefs Measures.
The Ambivalent Sexism Inventory. This recently developed questionnaire (Glick & Fiske, 1996) comprises 22 items concerning two aspects of attitudes towards gender roles: Hostile and Benevolent Sexism. Where Hostile Sexism can be described as sexist antipathy toward women, Benevolent Sexism refers to subjectively positive (albeit sexist) attitudes toward women. It has been found that the Benevolent Sexism subscale correlates with favourable evaluations of women, whereas the Hostile Sexism subscale correlates with negative evaluations of women (Glick & Fiske, 1996). However, a positive correlation between these two sexist attitudes is typically found (Glick & Fiske, 1996; Hadjion, 1997). As the content of these attitudes comprises opposing evaluations of women, the overarching construct measured by the scale has been termed ambivalent sexism (Glick & Fiske, 1996).
In the Ambivalent Sexism Inventory (ASI) participants were required to indicate the extent to which they agreed or disagreed with the statements on a six point Likert scale. Responses could range from strongly disagree though to strongly agree. An example of an item measuring Hostile Sexism is: Most women interpret innocent remarks or acts as being sexist. An example of a Benevolent Sexism item is: A good woman should be set on a pedestal by her man.
Women in Society Questionnaire. This scale, the WSQ, was developed by Lewis, Grieve and Kashima (1994) and contains statements pertaining to the roles and behaviours of women in society. As items in this scale tend to correlate highly with the ASI (Foddy, 1998), a subset of six items (Appendix E) that directly refer to the competency of women, was presented along with the ASI. Participants' attitudes toward women were measured in the same manner as in the ASI. An example of one of the items selected for use in this study was: Men are more financially astute than women.
Your Views Checklist (YrVws). This scale comprises the same set of features as the OthsVws (Appendix A). However, rather than aiming to assess implicit knowledge by asking respondents what they think others think constitutes a male or female thing it aims to measure explicit beliefs by asking respondents to indicate what they think constitutes a male or female thing. In a study by Foddy (1997) the male- and female-related subscales were found to correlate (r = 0.35, p < 0.05). Moreover, in the same study, this scale was found to relate to the WSQ (r = 0.36, p < 0.05).
Measuring Competence-Related Behaviour and Expectations for Competence.
Standardised Experimental Setting. This method is commonly employed in status research to assess competence-related behaviour (Foddy, 1997; Foddy & Smithson, 1996; Foschi, 1996; Pugh & Wahrman, 1983). The version employed in this study was a computerised version of this method (Foddy & Smithson, 1996). In this, participants were required to complete an ambiguous task, a pattern recognition task, on their own, and then with a simulated partner. Prior to commencing they were informed that the task was testing an important ability, pattern recognition ability (PRA), and that their results would contribute to research in the area.
Working firstly on their own, the participants completed a set of ten practise trials. Within each trial two rectangular blocks comprising red and blue sections were presented to the participants on a computer screen. Although each of the blocks contained equal areas of red and blue, they were led to believe that the colours were disproportionately represented. They were then instructed to indicate which of the two blocks, on the left or the right hand side of the screen, contained the greatest area of red. To do this, they were instructed to press a computer key labelled left if they believed the pattern with the greatest area of red was on the left or right if it was on the right. They had ten seconds to make their decision. Failure to respond meant that the trial was recorded as an error.
Working jointly with a simulated task partner, the participants were required to complete a set of 25 trials. They were informed that the goal of this exercise was to work as a team to best complete the task. They were then informed that after they had indicated which block contained the greatest area of red, their partner's response would appear on the screen. This response was generated by the computer, and in 20 (80%) of the trials, the responses were counter to the participants' responses. This protocol was adopted to ascertain how the participants' would perform when their decision was not in agreement with their partner's response. To complete the trial they were then given another ten seconds to indicate a second response; they could choose to stay with their own decision (rejection of influence) or defer to their partner's decision (acceptance of influence). The proportion of responses in which participants' stayed with their own decision (rejection of influence) was the dependent measure. This measure is referred to as P(S) or proportion of stay responses. In this, it is assumed that if the participant holds lower expectations for other's performance than for self, they will reject the other's initial answers.
After completing the joint phase, participants were provided with feedback via the computer, as to how well their team had performed. This feedback was controlled across participants, and indicated that their team had correctly responded to 20 of the 25 trials. As a breakdown of each partner's contributing score was not provided, participants were not able to deduce who in their team had best contributed to the overall score. With this in mind, they were then required to complete a series of questions via the computer that concerned their's and their partner's performance. On a Likert scale they were asked to rate the importance of taking their partner's choices into account, their estimates of how well their team had performed, their confidence in their own and in their partner's performance, their estimates of their own and partner's ability, and estimates of their own ability compared to their partner.