Colleen Ganley, Leigh Mingle, Allison Ryan, Katherine Ryan, Marian Vasilyeva, and Michelle Perry write:
Stereotype threat has been proposed as 1 potential explanation for the gender difference in standardized mathematics test performance among high-performing students. At present, it is not entirely clear how susceptibility to stereotype threat develops, as empirical evidence for stereotype threat effects across the school years is inconsistent. In a series of 3 studies, with a total sample of 931 students, we investigated stereotype threat effects during childhood and adolescence. Three activation methods were used, ranging from implicit to explicit. Across studies, we found no evidence that the mathematics performance of school-age girls was impacted by stereotype threat. In 2 of the studies, there were gender differences on the mathematics assessment regardless of whether stereotype threat was activated. Potential reasons for these findings are discussed, including the possibility that stereotype threat effects only occur in very specific circumstances or that they are in fact occurring all the time. We also address the possibility that the literature regarding stereotype threat in children is subject to publication bias.
Interesting. But what’s with the numbers not being spelled out? That freaks me out.
This article by economist Wei is really eye-opening. Used data from mass administration of the NAEP where “through a design quirk, students are randomly assigned to test blocks, some of which include gender prime questions while others do not.”!
http://www.people.fas.harvard.edu/~twei/papers/sthreat_naep.pdf
Result: “Little evidence of negative stereotype threat and strong evidence of stereotype reactance with girls performing better relative to boys for some gender primes.”
Based on that, it appears the little lab studies that social psychologists do may even have gotten the *sign* of the real effect wrong.
Why? Perhaps because in the lab studies, it is obvious what the investigator is hoping to get and, helpfully, play along with it (c’mon, when you bring in college women and have them read something about how women are bad at math, and then you ask them to take a math test, that’s not exactly subtle). In the real world, when you say “hey, maybe you are bad at X!” people go “screw that!” and do better.
If the stereotype threat investigators in social psychology were serious about their work, obviously they would ask participants at the end what they thought was going on. Interesting that they don’t do that.
Big:
1. I’m pretty sure that the stereotype threat investigators in social psychology are serious.
2. You claim: “In the real world, when you say ‘hey, maybe you are bad at X!’ people go ‘screw that!’ and do better.” I’d guess this happens in some settings and not in others. Effects in social psychology are notoriously situational.
3. In any case, I appreciate that you’re reacting to the substance of the post, unlike all the other commenters who are going on and on about the rules for writing numbers in text.
(re 3.) It’s your own damn fault for pointing it out!
(re 2.) Here at NYU Physics, we had a proposed study of our new Physics Majors nixed by our IRB because it might introduce stereotype threat and differentially harm women and members of racial minorities who are thinking about majoring in physics. It is therefore of great practical importance to me to know the “sign” or “size” of the stereotype threat effect, as a function of situation and/or in the situations we care about.
Perhaps your administration could buy off the IRB members with a house on Fire Island.
I’ve been meaning to invite you for a weekend to my mansion in the Hamptons.
Andrew –
1. To clarify what I mean by ‘serious’, I am mindful here of the fact that demand and expectancy effects have been discussed in social psychology for decades, along with precautions for minimizing them. Bob Rosenthal worked on this topic in the 1970s, and showed how pernicious they can be. I would contend that if people in the field were generally serious about revealing the truth, they would have set standards that require everyone to question subjects about their theories at the end of the study, and to report the results of those queries in their article. Why don’t they do that (even at JPSP, which you mentioned as a prestigious publication)? Of course from a careerist standpoint, there is no percentage in it. It would just spoil a certain percentage of everyone’s surprising findings. If you care about the truth, you would do it–if you don’t, you wouldn’t. On reflection, my hypothesis is not exactly that they are nonserious. It is that they are not truth focused. Not good, if you are trying to be a science.
2. Well you have jumped to one of several possible conclusions here, right? Perhaps the NAEP study and it’s “reactance” effect are just one of several possible outcomes, depending in some unspecified way upon context. Or (and this was my hunch) perhaps it is the typical result you will get when you expose people to insinuations of low ability outside of the lab (whereas in the lab studies, perhaps subjects are figuring out what the experimenter is hoping for and playing along).
(As to the more general idea that social psychology is full of effects that sometimes reverse themselves depending on context, maybe so–but shouldn’t we also keep in mind the hypothesis that when that seems to be happening, maybe there is no true effect at all to speak of–just sampling error around a roughly zero effect? Not so easy to distinguish, right?)
3. Glad you were happy with my focus on the substance. I find the substance here very interesting whereas the rules for writing numbers in journal articles seems pretty boring to me.
BW
Back in 2004, I argued that “stereotype threat” was to a sizable extent a matter of hinting to certain psych majors that they shouldn’t knock themselves out working hard on this meaningless test — “priming,” as it were.
http://www.vdare.com/articles/stereotype-threat-aka-occams-butterknife
It would be unethical to try to get women or blacks to score worse on a high stakes test like the GRE, so the theory can only be tested on low stakes tests.
Why?
Do the usual rules for spelling out small numbers have any good reason?
Are there any good studies that compare texts with spell out vs numerals for:
Speed of reading?
Comprehension?
Retention?
Are there studies of human variability? For instance, do some people learn better by recalling the sound of saying a number or by its appearance? If a text has many numbers, and spelling them out lengthens a text so that a natural chunk gets larger than a page or screen, so it requires flipping back and forth, does that have any effect?
How about reading texts not in one’s native language?
Zwei-und-dreissig? Thirty-two? 32?
Do any of the usual rules for usage have a good reason? Mostly it’s convention, and I see no reason to needlessly break from it. (Nor a big deal if someone breaks a rule)
Seems to me small integers are always spelled out. But 32 is 32.
+1. This always struck me as nitpicky. There is no lack of clarity, whether you use “eighteen” or “18”.
Numbers should be written out, IMHO, when they are actually the words. The first “1” should be written as “one” because the usage “1 potential explanation” could be replaced by a word like “a” or even “the”. But the “3 studies”, etc. are counts so I have no problem with those.
The evil influence of MBA-speak/media-speak creeps onward. Also atrocious to read: “the mathematics performance … was impacted by…” Impacted? Sounds painful! Did they need to take them to the hospital?
See a set of rules.
Note odd implications of Rule 2, in which the correct usage is said to depend on other related numbers.
For example: by that rule:
NO: my 9 cats fought my neighbor’s 9 cats.
OK: my nine cats fought my neighbor’s nine cats.
OK: my nine cats fought my neighbor’s ten cats.
NO: my ten cats fought my neighbor’s ten cats.
OK: my 10 cats fought my neighbor’s 10 cats.
NO: my nine cats fought with my neighbor’s ten cats, but they were overwhelmed by one hundred and twenty-two dogs from other neighbors.
OK: my 9 cats fought with my neighbor’s 10 cats, but they were overwhelmed by 122 dogs…
It is annoying that something deemed correct usage in one place can become incorrect/awkward by including another number, possibly sentences away. Think about writing a computer program to read text and offer advice. Whether 9 or nine is “correct” cannot be known without examining much more context, and even then, it’s hard. My cats and neighbor’s cats clearly belong in same category, but what about the dogs?
Anyway, this is actually a serious suggestion for experiments in readability, if they haven’t been done. Languages change.
APA calls for numbers to only be spelled out at the beginning of sentences, except for some weird exceptions. It looks like they’re following that rule.
JVM: thanks, that’s a rule I can follow.