Cognitive Daily

As every high-school senior knows, many colleges and universities take “racial diversity” into account when selecting students for admission. The practice is controversial, because it could mean that qualified students are denied admission, and those who are admitted must tolerate other students with a less rigorous academic background. The institutions often argue that their admissions practices are justified because increased diversity creates a more effective learning environment.

There is some research backing these claims: schools with greater racial diversity tend to have better retention, satisfaction, and intellectual development. However, most of the data is correlational—it’s unclear whether diversity caused the positive results, or vice versa. A team of researchers led by Anthony Lising Antonio developed an experimental study to see if they could find a causal link between racial diversity and student achievement (see comments for the full citation).

They placed white student volunteers into groups based on a preliminary survey about their opinions on controversial issues. Students were placed in groups with others they agreed with. However, in each group, in addition to the three naive volunteers, there was also a “confederate”—a student who followed a script that either agreed with or contradicted the professed opinions of the others in the group. Half of the confederates were Black.

Before the groups met, each participant wrote an essay for 15 minutes on their pre-screened social issue (either child labor in developing countries or the death penalty). Then they met with their discussion group to discuss the issue. Then the participants wrote another essay on the same topic. Finally, they wrote a third essay on the topic that their group had not discussed.

Unfortunately the results of the study will probably do little to diminish the controversy over racial diversity in college admissions. The essays were rated by a panel of three judges for “integrative complexity” (IC), which is a measure of how well the essay incorporates multiple perspectives and is associated with higher achievement in college students. The post-discussion essays had significantly higher ICs when the confederate disagreed with the other members of the group, but not when the confederate was simply of a different race.

However, when students wrote on child labor for the third essay (meaning the group had been discussing capital punishment), then white students in groups with Black confederates wrote essays with higher ICs compared to those in groups with white confederates. Though the same result was not found when the third topic was the death penalty, this result does suggest that in some instances, the mere fact of racial diversity in a group can lead to improved writing.

By far the researchers’ most significant finding was one that simply matched previous research: students who had a more racially diverse group of friends and classmates outside of the study tended to write essays with higher ICs. Again, however, this finding is only a correlation, and cannot on its own show that racial diversity improves learning. Although this study is a good start, perhaps a study that provides participants more than an hour or so interaction with members of a different race will give a more definitive answer.


  1. #1 Dave Munger
    June 3, 2005

    Full citation:

    Anthony Lising Antonio, Stanford University; Mitchell J. Chang, UCLA, Kenji Hakuta, Stanford University; David Kenny, University of Connecticut; Shana Levin, Claremont McKenna College; and Jeffrey Milem, University of Maryland, “Effects of Racial Diversity on Complex Thinking in College Students,” Psychological Science, 2004

  2. #2 Dana Leighton
    June 3, 2005

    I happen to be a teacher and researcher whose specialty is integrative complexity (IC). When I read this summary, I did not remember reading the original article, but after rereading it, I remembered it. And, I remember why I disregarded it. It contained some methodological errors and mainly nonsignificant results (except, as you noted, the correlation of people with racially diverse friends and higher IC—no surprise really because IC is strongly and negatively correlated with Right Wing Authoritarianism, likely a predictor of who would choose more same-race friends).

    When we conduct IC research, we have to code written materials for complexity. As you can imagine, the coding process is fraught with potential for error, since you are subjectively rating written passages. Therefore, when the IC coding system was developed, The procedures specify that an interrater reliability (Cronbach’s alpha) of .85 be met for sufficient reliability (see Baker-Brown, Ballard, Bluck, de Vries, Suedfeld, & Tetlock, 1992).

    The reliability of these researchers only reached .62 to .70 (it did not specify what reliability measure was used – Pearson r might have been), which is too low. Realizing their problem, the authors tried to justify the low reliability through some rather dubious arguments. The low reliability will wreak havoc in the data – by increasing the variability, it may create differrences where there are none (it could work in favor or against the hypothesis).

    The correct procedure in cases of low reliability is for the raters to meet to talk over their differences, and agree on why they differed (learning why their score was too high or too low). That meeting is usually followed by a thorough recoding of the material. After the recoding, the reliability is usually sufficiently high.

    This research’s results from the IC measures were basically nonsignificant when it came to the race of the confederate. They claimed a “marginally” significant (p=.09) main effect of race of confederate on the prediscussion essay. The increase in IC was quite small (.11 on a 1-7 scale) in the presence of the black confederate. The post discussion essay and the “transfer” essay were both nonsignificant for race of collaborator.

    Overall, I was surprised Psychological Science accepted the paper. The authors are right in that it served a purpose to perhaps generate more research on the influence of racial diversity in the classroom on IC. I think a more interesting study might be to look at how stereotype threat affects IC—I would suspect that if we had boys write an essay in a coed classroom and a single sex classroom, we’d get higher IC in the single sex classroom (because of stereotypically lower scores in english tasks). I would expect that with girls we’d see little difference (stereotypically not at threat in english classes). So far as I know, this has not been done. Anyone with a IRB and a subject pool want to work with me on it? :^)

    Thanks for the great site you produce, Dave. Your research summarys are always spot on, and extraordinarily accurate. I often read them in my classes as examples.

    Dana Leighton
    Instructor of Psychology
    Tri-County Technical College
    Pendleton, SC 29670

    Baker-Brown, G., Ballard, E. J., Bluck, S., de Vries, B., Suedfeld, P., & Tetlock, P. E. (1992). The conceptual/integrative complexity scoring manual. In C. P. Smith (Ed.), Motivation and personality: Handbook of thematic content analysis (pp. 400-418). Cambridge: Cambridge University Press.

  3. disruptive

    Does racial diversity help students learn? .

  4. #4 fred lapides
    June 6, 2005

    I am no expert in this or in just about anything, but a recent piece in the NY Times suggested that learning to parse, learning syntax, learning style need to be taught and NOT an amphasis upon content, as is usually the case. What you folks seem to be discussing here is how racial diversity is reflected in subjects dealing with issues that offer perspectives on race…but ultimately, it is not the arguement and content but the merit of the writing that is the issue in writing courses.

  5. #5 Dave Munger
    June 6, 2005


    Thanks for your great insight into this research. I do think that this study shows some interesting trends and that there is a need for additional research in this area, but your points are very well taken.


    The article you’re referring to, I believe, is the Stanley Fish article about the unusual way he teaches composition. It has received a great deal of attention around the Web, for example at Conversational Reading. The point I made over there was that while his “content free” method might be effective for him, it would be difficult for others to replicate.

    The issue this study is trying to address is how to get students to represent multiple perspectives in their writing, which is something that is sorely lacking in most college writing. Fish’s answer to the problem is to remove the option of taking other perspectives at all, but slyly to force them to look at language itself from a different perspective. I think the incidental result may be that students do indeed learn to adopt new perspectives.

    But again, if it turns out that simply having more racial diversity among classmates can get the same results, then this might be easier to accomplish than changing the college curriculum so that all students have to invent a new language in their freshman composition class.

  6. #6 Dana Leighton
    June 7, 2005

    Hi Dave & Fred & et al.,

    The content-free teaching method for writing is fascinating. I generally like Stanley Fish’s articles, and this one is good. His main thesis is that through grasping that English construction is arbitrary, yet logically connected, that he can help students get beyond the “subject” of the argument, and get down to the structure of the argument—the way ideas in the argument are connected, and perhaps, interrelated. That is cogitive complexity.

    You see, in integrative complexity coding, we aim specifically/e> for coding structure, not content. The explicit recognition of multiple perspectives on an issue (regardless of the issue) earns a score of 3 (scale 1-7). The next major step in the scale is the explicit recognition that these perspectives are interrelated. That the perspectives have some kind of mutual influence. That gets a score of 5.

    It would be a fascinating study to have Fish’s students write some paragraphs for IC coding before his class, and again afterward. If we compare to a control group of “typical” writing curriculum students, do we see an increase in IC? I would guess that all other things being equal (which is a major assumption: school; teacher; selectivity; urban vs rural; prior experience with multiple races/ethnicity), that Fish’s method would produce a modest IC gain in his students, and probably more than the control.

    We are cognitive managers (a term preferred by Suedfeld over miser), and allocate our limited cognitive resources in such a manner that we have adequate reserves for challenging tasks. The more challenging the task, the more we have to carefully manage our cognitive resources. Suedfeld’s (1992) example is international decision makers who have to maintain numerous perspectives, differing goals, and multiple constituencies in mind while employing metacogniitive resources. As you can imagine, this is quite taxing cognitively. So, for some decisions, they may use fewer cognitive resources (as evidenced by lower IC in those domains).

    College students certainly are not international decision makers (although they often have to do many of the same balancing acts, on a smaller level), and so the normed average IC for college students is about 2.0 – by comparison, Supreme Court justices average around 4.0, and Osama Bin Laden showed a complexity of 1.0 (the lowest possible) before and after his September, 2001 attacks on New York and Washington (see Suedfeld & Leighton, 2002).

    If college students could have their IC increased from 2.0 to 2.5 (statistically significant) would be a major accomplishment; 2.0 to 3.0 would be a miracle!

    Dave, as regards just having a more diverse classroom, the question would be whether only the “white” students increased their complexity, or all students increased. The effect seen by Antonio, et al. can as easily, I think, be attributed to effects of stereotype threat (or perhaps self-presentation) than to any real (i.e. motivated) change in complexity. If we could see an effect in all students, I would be more easily convinced.

    More research! Onwards, lads & lasses!


    Suedfeld, P. (1992). Cognitive managers and their critics. Political Psychology, 13(3), 435-453.

    Suedfeld, P., & Leighton, D. C. (2002). Early communications in the war against terrorism: An integrative complexity analysis. Political Psychology, 23(3), 585-599.