The problems of aging
The first part of this lab is based on the major example in the text—the Eysenck study of recall of older and younger subjects under conditions of differential processing. I chose to repeat that because you have it all laid out in the text, and may find the review helpful.
Briefly, Eysenck (1974) took subjects from two different age groups; Young (18-30) and Mature (55-65). These were further broken down into 5 subgroups, who varied in the amount of processing they performed on words they read, ranging from counting the letters to creating mental images of the words. At the end of the task, all subjects were asked to recall the words in the list.
The data can be found in a file text called Eysenck.dat, and the variables are Age, Condition, and Recall.
First of all, run the overall two-way analysis of variance, and interpret the results. Then look at the simple effects of Condition for each age group, and expand on those results with multiple comparisons tests. Why would you run the multiple comparisons on the simple effects rather than on the main effect? Is there a significant difference between the groups for the task that requires the lowest level of processing.
You can use the General Linear Model to ask for polynomial contrasts--either for main effects or simple effects. What do these tell you here? What is wrong with using polynomial contrasts for these data?
This example is based on a recent paper by Inzlicht and Ben-Zeev (2000) A threatening intellectual environment: Why females are susceptible to experiencing problem-solving deficits in the presence of males. Psychological Science, 11, 365-371.
It is well known that when you put someone in a testing situation and make a stereotype about them salient (e.g. Women can't do math), that stereotype will lead them to perform more poorly than if it had not been made salient. There are two competing theories for this result. The stereotype threat (Aronson) explanation argues that simply being placed in that situation is threatening with respect to the corresponding stereotype (it evokes the stereotype), and this, in turn, leads to poor performance. On the other hand, tokenism theory (Lord and Saenz) suggests that simply being seen as a minority or having token status should elicit cognitive deficit in all domains, not just the ones about which there are stereotypes.
Inzlicht and Ben-Zeev wondered how far you could push this--specifically, what constitutes making a stereotype salient. Is just being a minority is a situation sufficient? They also wanted to compare the stereotype threat and tokenism explanations.
They used two tasks. One was similar to the verbal portion of the SAT, while the other was similar to the Math portion. We have a stereotype that women do not do well in math, but we don't have a stereotype that they don't do well in verbal tasks (if anything, they do better). So making gender salient should lower women's performance on a math test (stereotype threat) or on both tasks (tokenism).
To make gender salient (or not) they either tested a female participant in the presence of two other females, or in the presence of two males. This was called the same-sex versus minority condition. If salience is invoked by this manipulation, women should show decreased performance when they are in the minority, relative to the same-sex condition. Depending on your theory, this decrement should either be just in math, or in both tasks.
Data very similar to those of Inzlicht and Been-Zeev can be found in Inzlicht1.sav. The first variable represents the Task (Math vs. Verbal), and the second represents Status (Same-sex vs. Minority).
Before you run this analysis, decide what analyses you need to run, and what predictions you would make.
Inzlicht and Ben-Zeev had slightly different Fs than I have. This is due to the fact that they corrected for the student's SAT scores when they applied to college. This should not make a difference, but I believe that the statistics they gave were the uncorrected ones. I had to do a bit of fudging, but the conclusions are unchanged from theirs.
Inzlicht and Ben-Zeev ran a second experiment as part of the same paper. In this study they compared males and females, but used just the math test. They also included a condition in which women were in a mixed-sex majority, rather than a same-sex majority. This represents a 2 by 2, with 2 levels of sex and two levels of same-sex vs. minority. In addition, there is a 5th group (majority) with 2 females and 1 male. The data are available in Gumby at Inzlicht2.sav.
This study allows us to ask several different questions. For example,
These questions will require more than one analysis, and may require that you look at only parts of the data at once. How would you go about answering these questions?
What do your answers tell you?
Last revised: 01/25/02