Tuesday, February 8, 2011

Laughable gender research: the vervet toy-preference study

Here’s my critique of one of the worst pieces of “research” ever to be trumpeted by magazines, newspapers, or blogs:

Gerianne M. Alexander and Melissa Hines, “Sex differences in response to children’s toys in nonhuman primates (Cercopithecus aethiops sabaeus).” Evolution and Human Behavior 23 (2002) 467–479.

(The full text is available free online.) I can’t believe I’ve been in another discussion with someone who can’t see what’s wrong with this study. It’s the classic example (though there are others) of the deeply flawed research that people searching for evidence to support their pre-existing notions about innate sex differences accept uncritically, to the point that they meet my substantive criticisms with claims that I’m so ideologically blinded I can’t appreciate its value. (It is heartening that some others – and I hate to link to that blog, but just search for the word “vervet” – have been able to see through it.) People want to argue that it’s merely preliminary, providing some evidence suggesting the need for more refined studies (and we now see hilarious reviews of the research like this one). No, it isn’t. It doesn’t provide evidence of anything beyond perhaps some qualities of the researchers. I’m going to offer my extended critique of the article, and I’ll allow comments as long as they demonstrate a reading and thinking about what I’ve said and aren’t utterly stupid.

Here’s the abstract:
Sex differences in children’s toy preferences are thought by many to arise from gender socialization. However, evidence from patients with endocrine disorders suggests that biological factors during early development (e.g., levels of androgens) are influential. In this study, we found that vervet monkeys (Cercopithecus aethiops sabaeus) show sex differences in toy preferences similar to those documented previously in children. The percent of contact time with toys typically preferred by boys (a car and a ball) was greater in male vervets (n = 33) than in female vervets (n = 30) (P < .05), whereas the percent of contact time with toys typically preferred by girls (a doll and a pot) was greater in female vervets than in male vervets (P < .01). In contrast, contact time with toys preferred equally by boys and girls (a picture book and a stuffed dog) was comparable in male and female vervets. The results suggest that sexually differentiated object preferences arose early in human evolution, prior to the emergence of a distinct hominid lineage. This implies that sexually dimorphic preferences for features (e.g., color, shape, movement) may have evolved from differential selection pressures based on the different behavioral roles of males and females, and that evolved object feature preferences may contribute to present day sexually dimorphic toy preferences in children.
People seem to want to suggest that the authors are simply fishing for any sex differences in object preferences. But it is obvious that they are not. They are hypothesizing that the bases for alleged human sex differences in toy preference are evolved from a time before our hominid lineage began, and they believe their results show this.

Of course, you could do research giving vervets a series of random objects, not controlling for shape, color, texture, size, and so on and with no real expectation that sex or other differences would appear. If you did, you wouldn’t know why any apparent results emerged, and these would likely be a statistical blip that wouldn’t be replicated. This is what these researchers are in fact doing, but it isn’t at all what they think they’re doing. Let’s take a look at the rationale for selecting these objects:
Boys favor construction and transportation toys, whereas girls favor toys such as dolls (Connor & Serbin, 1977; Liss, 1981). Boys are also more active (Campbell & Eaton, 1999; Eaton & Enns, 1986) and show more rough physical play than girls (DiPietro, 1981).
This is automatically problematic, as it’s improper to select some studies from a single culture at a single moment in time and suggest they represent some transcultural and transhistorical preferences of “boys” and “girls." That aside, the authors note the literature suggesting nonbiological reasons for any observed differences in toy-preference and play styles:
These sexually dimorphic play styles are thought by many to derive from learning and cognitive mechanisms associated with gender socialization (see Serbin, Poulin-Dubois, Colburne, Sen, & Eichstedt, 2001 for a recent discussion). Learning theories suggest that sex differences in play activities and toy preferences arise from modeling and reinforcement of sex-typical play (Bandura, 1977; Fagot & Hagan, 1991; Langlois & Downs, 1990). Cognitive theories suggest further that children develop an understanding of their gender identity that results in schemas or mental representations of socially defined gender appropriate behavior and a positive evaluation of toys and activities associated with this gender identification (Maccoby, 1988; Martin, 1999; Martin, Wood, & Little, 1990).
These authors, though, hypothesize that human sex differences in toy preferences reflect evolved differences. They believe they can test this hypothesis (to use the term loosely) by studying another primate species:
The present research addressed the hypothesis that toy preferences may be associated with factors other than human social and cognitive development by measuring toy preferences in a nonhuman primate, the vervet monkey (Cercopithecus aethiops sabaeus). Unlike humans, vervet monkeys are not subject to the specific social and cognitive influences proposed to explain human sex differences in toy preferences. To evaluate the possibility that sex differences in toy preferences can arise independent of these social and cognitive mechanisms, we therefore tested the hypothesis that vervet monkeys, like human beings, show sex differences in toy preferences.
In other words, they expect that vervets will exhibit the same differences in preference that humans allegedly do, for the reason that the bases for these differences evolved in some ancestor, millions of years ago, common to both vervets and humans (and indeed, all primates). It is misleading to call these objects in this study “toys” – they are not measuring play behavior, but merely contact with the objects. Furthermore, it isn’t, as the authors acknowledge later, a study of preferences: the objects are presented to the monkeys sequentially, one at a time.

But a study of sex differences in object preference for a given species should be designed with that species’ lifestyle – behaviors, including sex differences in activities; social structure - in mind. Vervets are little monkeys that, like most primates, are largely herbivorous, but for occasional insects, grubs, and small rodents. They’re known for emitting a high-pitched squeal on the approach of a predator. The authors of this paper suggest in the discussion that females are more responsible for child care. With this in mind, let’s consider the objects selected for the study:
For each trial, six toys were placed in the group cage, one at a time, in a random order. Each toy remained in the enclosure for 5 min. The six toys were a ball, a police car, a soft doll, a cooking pot, a picture book and a stuffed dog. These toys were categorized as ‘‘masculine’’ toys, ‘‘feminine’’ toys, or ‘‘neutral’’ toys on the basis of evidence that boys are more interested than girls in balls and cars (the ‘‘masculine’’ toy set), girls are more interested than boys in dolls and pots (the ‘‘feminine’’ toy set), and boys and girls are approximately equally interested in books and stuffed animals (the ‘‘neutral’’ toy set) (Berenbaum & Hines, 1992; Connor & Serbin, 1977; Maccoby & Jacklin, 1974). The six toys were selected because children in a broad age range categorize them as ‘‘masculine,’’ ‘‘feminine,’’ or ‘‘neutral’’ and, for those toys categorized as ‘‘masculine’’ or ‘‘feminine,’’ the sex difference in toy preferences is reliable and relatively large in children.
The objects were not randomly selected. They were chosen because of claimed sex differences in human children’s toy preferences. (Again, although it is sadly common, it is improper to generalize from a single culture in one moment to “boys” and “girls.” But that’s not where this research goes hilariously wrong.) Again, the suggestion is that sex differences in preferences for features of these toys evolved early in primates. They are selected as objects, not for color or size or other aspects – a ball or cooking pot could be any color, and none of these aspects is controlled for.

So, the “masculine” "toys" are a police car and a ball. Remember that they’re suggesting that a male preference for these objects evolved early in primates. A plausible argument, if innate male preference for these objects were found in humans, would be that males, because of their role in hunting, evolved an interest in objects that move, can be thrown, etc. But why would we expect male vervets to prefer such objects? Neither male nor female vervets hunt, and the authors haven’t presented any evidence of sex differences from vervet lives or behaviors, or those of their ancestors, to suggest any reason for vervet males to have evolved a relative preference for "active" objects. But in any case the authors aren’t measuring the style of play, just contact,* and there’s no way the monkeys could know prior to making contact with the objects whether or not they’re mobile. (Leaving aside that a doll or cooking pot can be a mobile toy for active play, as can really all of those included. It’s a long stretch to suggest that they somehow sense that these objects specifically would come to be associated with humans’ “rough and tumble play.”)

Of the gendered objects selected, the only one that could reasonably, based on knowledge of vervets, be expected to possibly elicit a preference from one sex would be the doll. As the authors mention in their conclusion, vervet females have primary responsibility for the young, and it’s plausible to suggest an evolved sex difference in preference for objects that resemble infants – that are “animate.” (Again, they aren’t measuring actual “play” behaviors, but merely contact.) It’s a stretch, though. There isn’t really any reason to think that vervets would see a human doll, or even a vervet doll, as animate. It reflects the authors lack of attention to the species being studied that they selected a human doll in the first place.

The inclusion of a cooking pan as a “feminine” “toy,” though, is what renders the study ridiculous. Recall that this was selected based on alleged girl preferences in late twentieth-century North America, and that the authors believe that if they observe a female preference among vervets for this object it is evidence that its foundations appeared early in our pre-hominid evolution. This is a feature of more than one bit of this sort of gender “research” that shows the extent to which the authors aren’t thinking through their premises. The only plausible argument for the alleged preference for playing with cooking pots among human girls is that they have evolved to be more nurturing or domestic or something, and these are part of their symbolic play reflecting those tendencies. It’s bizarre to include cooking in this particular sphere given that it’s far from universally associated with females (even in our own culture). But it would be completely bizarre to argue that human females evolved a preference for modern objects currently associated with cooking as objects such that they would be drawn to their form without any knowledge of their function. To suggest that female vervets or our common ancestors, who obviously lack any knowledge of cooking, evolved a preference for these as objects is simply ludicrous. Species don’t evolve in anticipation of another species millions of years later using objects in a certain way.

The neutral objects (the stuffed dog and the picture book), like the others, are random. The stuffed dog, though, unintentionally on their part, provides a sort of “control” in their terms.

Given this silly design, whatever “data” they turn up are not going to support their broader hypothesis about shared primate evolution or the evolutionary basis of alleged human sex differences. So what did they find? Generally,
The proportion of contact time with toys typically preferred by boys was greater in male vervets compared to female vervets, whereas the proportion of contact time with toys typically preferred by girls was greater in female vervets compared to male vervets. In contrast, contact time with toys preferred equally by boys and girls was comparable in male and female monkeys. These sex differences in vervets resemble the well-documented sex differences in children’s toy preferences.
Although the serial introduction of the toys does not permit a true contrast of the relative preference for ‘‘masculine’’ over ‘‘feminine’’ toys within each sex, a within-sex comparison of contact scores showed that female vervets had greater percent contact with ‘‘feminine’’ over ‘‘masculine toys,’’ P < .01, but males had similar percent contact with ‘‘masculine’’ and ‘‘feminine’’ toys, P= .19.
In contrast to these sex differences in contact with ‘‘masculine’’ or ‘‘feminine’’ toys, no sex differences in percent contact with toys in the ‘‘neutral’’ category were found, t(61) ="0.61, P=.54.
The data are presented in rather misleading ways. For example, several of the findings – as that males did not make contact with the “masculine” toys any more than they did with the “feminine” ones - are presented cursorily such that they’re easily missed, and this is abetted by the way the graphs are presented. Also, although they’re not measuring actual play, they imply that the vervets separately were engaging with these objects in ways consistent with human play, suggesting: “In some instances, we also noted that vervet monkeys [note the lack of a sex qualifier] contacted toys in ways that appeared to resemble children’s contact with them.” Figure 2 shows two pictures, with the caption:
Examples of a female and a male animal contacting toys. The female animal (left) appears to be conducting an anogenital inspection of the toy doll, similar to inspections of infant vervet monkeys. The male animal (right) appears to be moving the car along the ground in a manner similar to that a child might use.
Moving a car along the ground is pretty much to be expected for anyone touching it, and the grasping “anogenital inspection” strains credulity. Of course, the rationale for selecting this set of objects in a vervet study and expecting sex differences is so silly – it would only have been funnier if they had shown a picture of a female “appearing to be” stirring something in the pot – that these interpretations are meaningless, but it's worthwhile to note that the paper as a whole (not to mention Alexander's statements in subsequent interviews) presents the data in a rather skewed way.

The improper way they go about interpreting their data likely reflects both their biases and that they must have realized too late how ridiculous their research design was. They conclude:
In view of this evidence, our findings suggest that object features or functions associated with human sex-typed toy categories may have adaptive significance for males and females.

Although nonhuman primates can learn to categorize novel stimuli (Freedman, Riesenhuber, Poggio, & Miller, 2001), the monkeys we observed had no learning history with the individual toys used in this study. Additionally, there is no evidence that vervets have an understanding of their gender. Yet, even if they do have a gender identity, they would not have had the experiences with objects (e.g., police car, cooking pot, book) that might be necessary to form categories based on associations between toys and gender in humans. Sex differences in toy preferences in a species lacking relevant social and cognitive experiences suggest, therefore, that other determinants of sex-typed object categorization exist.
It’s nothing short of amazing that they would suggest that the fact that vervets have no knowledge or understanding of these toys and showed some differences suggests innateness of alleged human preferences. This fact is what makes their design so crazily bad. They go on to argue that
Children’s toys, therefore, appear to have differential value for males and females of at least two primate species, vervets and humans. This new finding provides additional support for the hypothesis that sex differences in toy preferences can arise independent of the social and cognitive mechanisms thought by many to be the primary influences on toy preferences in human beings.
But as I’ve discussed, sex differences in these object “preferences” in vervets, should they appear, are likely to be a fluke, an artifact of a single study. They acknowledge: “Information is not yet adequate to know what low-level perceptual properties may contribute to the responses to sex-typed toys by vervet monkeys and this study was not designed to evaluate these hypothetical feature preferences.” But this is a tremendous understatement. The idea that female vervets would have evolved a preference for cooking pots is absurd on its face. In the conclusion they pull novel “explanations” out of the air.

Again, the doll was probably the only of the gendered objects for which a plausible rationale exists for thinking it might elicit sex differences. They appeared to find them in that females made more contact with it than males. So they launch into a discussion of female nurturing (all of a sudden talking about vervet behaviors!):
[T]oys preferred by girls have been described as objects that afford opportunities for nurturance (Campbell, Shirley, & Heywood, 2000; Eisenberg, Murray, & Hite, 1982; Miller, 1987), and selection pressures may favor responsiveness to object cues (e.g., an animate-like form) that signal maternal behavior because these cues enhance infant survival. A doll, for example, may be of greater interest to females than males in primate species where females interact with infants more than males do. Such species include humans (Blakemore, 1981) and vervet monkeys (Lancaster, 1971; Meaney, Lozos, & Stewart, 1990).
As I said, it’s questionable whether vervets would recognize a human doll as “animate-like.” Interestingly, one of the “neutral” objects was also considered by the authors to be “animate-like.” And they note:
However, as we found no sex differences in response to toy categories based on an animate-like (doll, dog) or inanimate-like (car, ball, book, pan) distinction, it appears that other characteristics contributed to the female object preferences we observed.
Indeed, the males touched the “animate-like” furry dog more than both of the “masculine” objects. So the alleged nurturing-based attraction to “animate-like” objects among females appears to be thrown in there to promote readings that aren’t, in their terms, based on the data. (As with the doll, I’m not convinced the furry dog was "animate" to vervets, though it's also possible that it was and the doll was not.)

And of course this ignores the pan. So they switch to speculation about color which they can’t support with their data. (I wonder if it wasn’t until they were writing this up that they realized how profoundly dumb had been the inclusion of the cooking pot…) They try to get around this by selectively suggesting – in a way not at all in keeping with the study’s rationale or design – that the female “preference” for the "feminine" objects might have been related to their color:
Color may also provide an important cue for female interest…A preference for red or reddish pink has been proposed to elicit female behaviors to infants that enhance infant survival, such as contact (Higley, Hopkins, Hirsch, Marra, & Suomi, 1987). The hypothesis that reddish pink or red may be a cue signaling opportunities for nurturance and thus eliciting female responsiveness could explain our finding of greater female contact with both the doll (with a pink face) and the pot (colored red).
So this “hypothesis” could provide a post hoc “explanation.” Leaving aside the fact that the “masculine” ball was orange, it should be noted that the pot was also hard and metallic, and potentially noisy, but the authors don’t suggest that any of these features – inconsistent with human gender stereotypes – could have been the basis for this preference. Of course, without controlling for other aspects of the objects, any of these speculations are basically worthless.

In any event, any preference for reddish objects, should it exist, is not the same as a preference for “feminine toys,” which is the whole basis for the study and its conclusions. Recall: This is a study of object preferences, and the cooking pot was selected because human “girls are more interested than boys in…pots.” They are arguing that "object features or functions associated with human sex-typed toy categories" shape the preferences of vervets, indicating something about human evolution. But while hardness and metallicness are inherent features of modern cooking pots, redness is not – a cooking pot could be any color at all. The random suggestion that “Well, females may have been drawn to this cooking pot because it was red" is totally inconsistent with the whole rationale of the study. They're suggesting that females would prefer red objects regardless of their sex-typing. (No doubt this would have been their post hoc "reasoning" had, for example, the car been red and this been the case. There is no finding too inconsistent with gender stereotypes that those who want to cannot find a way to interpret in terms of gender stereotypes.)

On to the males. More speculation:
Toys preferred by boys, such as the ball and police car used in this research, have been characterized as objects with an ability to be used actively (O’Brien & Huston, 1985) or objects that can be propelled in space (Benenson, Liroff, Pascal, & Cioppa, 1997). Preferences for such objects may exist because they afford greater opportunities for engaging in rough or active play. In humans, these characteristics have in turn been suggested to relate to targeting or navigating abilities (for discussion, see Alexander, in press) that might be particularly useful for males for purposes of hunting or locating food or mates (Eals & Silverman, 1994; McBurney, Gaulin, Devineni, & Adams, 1997; Silverman & Eals, 1992). As suggested for females in regard to objects that signal nurturance, males may therefore have evolved preferences for objects that invite movement.
The various problems with this have been discussed above: vervets do not hunt, they’ve proposed nothing about vervets that would make anyone expect males to be relatively drawn to objects that "invite movement," and they aren’t studying what the vervets actually did with the objects (again, all of the objects they selected have an ability to be "used actively” or “propelled in space,” and there’s no logical reason to think that vervets would perceive these particular ones as “active”). But the fact is that the males didn’t prefer (note the change from “prefer” to “appear to prefer”) the “masculine” toys to the “feminine” toys. "[A]lthough female vervets preferred ‘feminine’ toys over ‘masculine’ toys,” they concede,
male vervets did not appear to prefer ‘‘masculine’’ toys over ‘‘feminine’’ toys. This difference between male vervets and boys may indicate that toy preferences in boys are directed by gender socialization to a larger degree than are toy preferences in girls.
It doesn’t, because the study is ridiculous to begin with. It doesn’t show anything about the evolution of alleged sex differences in human toy preferences. None of this stops them from making proclamations like:
We...suggest that children’s toy preferences reflect innate object preferences that are elaborated in typical human development by subsequent gender socialization. We found differences between male and female vervet monkeys that resemble the well-established differences in the toy preferences of boys and girls, consistent with the proposed existence of innate object preferences.
It’s a joke, and no amount of "But look! The female vervets touched the eyeshadow significantly more than the Super Bowl tickets!" is going to salvage it.

* "Contact was coded when an animal (or animals) approached a toy and made physical contact with it. Approach was coded only when an animal moved within 2 m of a toy without contacting it physically..."


  1. I typed a long and exasperated response but now your platform won't accept blockquotes and only wants 4096 characters.
    Got other stuff to do and will not be reformatting at this time.

    later I guess

  2. I thought that was all of Blogger.

    You can post it at Pharyngula, but, seriously, before you get all exasperated and write some long thing, please step back and read this again and think about it some more. They're saying that "The results suggest that sexually differentiated object preferences arose early in human evolution," on the basis of this study. On what basis did they expect that a finding that female vervets "preferred" cooking pots as objects would show this? Don't say they weren't hypothesizing sex differences in object preference paralleling alleged human ones in the sense that they have the same evolutionary basis, because they very obviously were - it's the basis for this crazy conclusion.

    It's garbage, Sven. Just think about it. Cooking pots. "Selected because" human "girls are more interested than boys in...pots." Think about this evolutionarily.

  3. I thought that was all of Blogger.


    It's already wrote, so I'll probably just e-mail it to you; nobody else gives a shit anyway so why post it publicly.

    As long as I'm here, though, let me attempt a brief response to your comment:

    What I think is your basic point of misinterpretation: By saying that "The results suggest that sexually differentiated object preferences arose early in human evolution" the authors were not saying "The results suggest that sexually differentiated preferences for these specific objects arose early in human evolution."

    Don't say they weren't hypothesizing sex differences in object preference paralleling alleged human ones in the sense that they have the same evolutionary basis, because they very obviously were

    Look, I don't pretend to know what was in their heart of hearts, but that is not the hypothesis they claimed to be testing in the paper being critiqued, and it's also not a hypothesis that could have been tested using their methods. I don;t think the authors were confused about that. You are.

    it's the basis for this crazy conclusion.

    yeah, that was not their conclusion, it was, explicitly, instead a suggestion, an extrapolation from the data. And, as indicated above, you've misinterpreted it anyway.

    It's garbage

    It's an inconsequential little paper in a second-tier journal. It may have some methodological issues (I'm not competent to judge), and much of the Discussion may well be bullshit (I think so anuyway), but it's not "garbage". The data are not bereft of meaning, and the meaning that can be--and actually is--gleaned from the data is not nearly so ridiculous as you insist.

    And your last couple of sentences indicate once again that you're profoundly confused about it.

    But I'll e you my screed with more detail and then I never want to hear about the stupid study again.

  4. Look, I don't pretend to know what was in their heart of hearts, but that is not the hypothesis they claimed to be testing in the paper being critiqued,

    It is. They're testing another primate species to see if sex differences in object preferences parallel observed human ones in order to see whether the human differences have evolutionary roots. It's why another primate species was selected, and it's why these toys were selected. There is no plausible basis on which to surmise that human and vervet females have a shared evolutionary preference for cooking pots.

    yeah, that was not their conclusion, it was, explicitly, instead a suggestion, an extrapolation from the data.

    No. It was the whole rationale for the study. To deny this you have to willfully misread what they've said and done. (The argument that human differences are based in socialization does not rest on any claim that no sex differences in object preferences could exist in any other species. Of course they could. Testing that wouldn't necessarily say anything about humans.)

    It's an inconsequential little paper in a second-tier journal. It may have some methodological issues (I'm not competent to judge), and much of the Discussion may well be bullshit (I think so anuyway),

    It got a lot of press at the time, and continues to be cited and referred to.

    but it's not "garbage". The data are not bereft of meaning, and the meaning that can be--and actually is--gleaned from the data is not nearly so ridiculous as you insist.


  5. any particular reason my last comment ain't here?

  6. Yes, Sven. First, I hadn't checked my comments until today. Second, read the last sentence of the first full paragraph of my post. (And note that I responded to windy about their handwaving on TET and that no response was forthcoming. I think she's trying to spare your feelings.) I already violated my own policy for your earlier comment, and I'm not going to continue to.

    I would advise you to read the article again, but I've concluded that it's likely futile. I can only assume that in this case your sexism has rendered you stupid. I hope you do read the Cordelia Fine book, but I'm not too optimistic about that, either. The epistemic prejudice is extreme in your case.

  7. Interesting dissection of this piece. For us laity it would have been interesting to include an explanation of what the data was (the counts of preference). I'm going to go look myself =), but the conclusion that female vervets like red pots has anything to do with our common evolutionary origigins is chortleworthy.