Anim Cogn (2001) 4:269–280 DOI 10.1007/s100710100084 REVIEW Brian Hare Can competitive paradigms increase the validity of experiments on primate social cognition? Received: 8 September 2000 / Accepted after revision: 8 April 2001 / Published online: 2 June 2001 © Springer-Verlag 2001 Abstract Experiments vary in their ability to distinguish what limits of difficulty and in what functions chim- between competing hypotheses. In tests on primate cogni- panzee can possibly show insight: negative or con- tion the majority of this variation is due to an experimenter’s fused results from complicated and accidentally-cho- ability to test primates in valid settings while providing the sen test-material, have obviously no bearing upon the adequate amount of experimental control. While experi- fundamental question, and, in general, the experi- menters studying primate cognition can use methods of menter should recognize that every intelligence test is control perfected in captivity, it is still very unclear how a test, not only of the creature examined but also of the to design and then objectively evaluate the external valid- experimenter himself. I have said that to myself quite ity of new experimental paradigms. I recommend that more often, and yet I have remained uncertain whether the effort be allocated to specify how to create relevant test experiments I performed may be considered ‘satisfac- settings for primates. Primate social life is highly compet- tory’ in this respect: without theoretical foundations, and itive. This means that all aspects of primates themselves, in unknown territory, methodological mistakes may quite including their cognitive abilities, have likely been shaped well have occurred; anyone who continues this work by the need to out-compete conspecifics. Based on this hy- will be able to prevent them more easily” (Koehler 1925, pothesis, sophisticated cognitive abilities of primates might p. 265). best be demonstrated in competitive contexts. Thus, it is suggested that one possible measure of validity is whether Any observed behavior can have multiple explanations as investigators integrate a competitive component into their to how and why it might have occurred (Hauser 2000). experimental designs. To evaluate this methodological Therefore, cognitive ethologists design and conduct ex- prediction I review the literature on chimpanzee perspec- periments for the purpose of distinguishing between dif- tive-taking as a case study including several recent studies ferent possible mechanisms underlying behavior, and that include a competitive component in their experimen- more specifically problem-solving behavior (Shettleworth tal designs. 1998). The validity of an experiment (also referred to as experimental sensitivity) is the measure of an experiment’s Keywords Chimpanzees · Perspective-taking · probability of being able to reject a null hypothesis or dis- Ecological validity · Competitive paradigms · Social tinguish between competing hypotheses (Gottsdanker 1978; cognition Shaughnessy and Zechmeister 1994). Cognitive ethologists strive to maximize experimental sensitivity. In addition, they wish to objectively evaluate the sensitivity of each Introduction other’s experiments. Unfortunately, as Koehler (1925) eloquently pointed out, “...the success of the intelligence tests in general will depending on the quality of design, experiments vary in be more likely endangered by the person making the their ability to distinguish between competing hypotheses. experiment than by the animal. One must learn and, if In experiments on animal cognition this often means that necessary, establish by preliminary observation, within experiments vary in their ability to demonstrate that a be- havioral strategy is guided by a sophisticated cognitive mechanism as opposed to more parsimonious ones. Two B. Hare (✉) components of design account for much of this variation: Department of Anthropology, Harvard University, 11 Divinity Avenue, Cambridge, MA 02138, USA (1) the proper use of experimental control (internal valid- e-mail: [email protected], ity), and (2) the appropriateness of setting or situation in Tel.: +1-617-4955243, Fax: +1-617-4968041 which the experimenter poses her experimental question to 270 subjects (external validity) (Gottsdanker 1978; Shaughnessy familiar settings (e.g., Oden et al. 1990; Boysen and and Zechmeister 1994). Therefore, experimenters inter- Berntson 1995; Tomasello et al. 1998). Therefore, in the ested in either designing or evaluating the validity of an case of cognitive tests on primates, ecological validity, de- experiment must take into account both the controls and fined as the degree to which an experiment is able to sim- setting of a paradigm. ulate a relevant aspect of subjects’ individual ontogenies Historically psychologists have placed a major empha- and/or their species evolutionary histories, is the measure sis on the utility of experimental control for the purpose of that becomes most important for evaluating external va- increasing the sensitivity of experiments (Shettleworth lidity (Hare et al. 2000, 2001). In summary, experimental 1998). Experimental control refers to the ability to pro- control alone does not allow one to maximize the validity duce a setting(s) in which one can isolate and manipulate of experiments on primates. Instead, the combination of a variable(s) relevant to deciding between two (or more) (1) proper control(s) and (2) an ecologically valid setting competing hypotheses (Martin and Bateson 1993). Many allows one to maximize the validity of cognitive experi- methods of control have been developed and standard- ments on primates. ized, such as controlling for order effects, removing biases using blind experimenters, and counterbalancing. The merit of the different methods of control is that, when used ap- propriately and in concert, they are extremely powerful Captive experimenters and can be applied to any species or paradigm allowing investigators to objectively evaluate each other’s experi- Most studies of primate cognition are experiments, which ments. In addition, if an experiment is controlled well, a are conducted in captivity (Tomasello and Call 1997; but positive result (rejection of the null hypothesis) reflects see Cheney and Seyfarth 1990a). Captivity affords com- the ability of the test subjects and should be replicable. parative psychologists the control, which is essential for iso- Less emphasis has been given to the limitations of ex- lating and manipulating variable(s). For example, the beauty perimental control. First, control can not make the results of the Skinner box in the eyes of the behaviorist is that the of an experiment generalizable to situations outside the test researcher has eliminated any competing variables that situation itself (Gottsdanker 1978). Second, no amount of might contaminate results. experimental control can help interpret negative results (i.e., Unfortunately, creating a valid test situation in captiv- an animal is unable to perform a task or performs in a way ity that is relevant to either the individual ontogeny or evo- that is consistent with the null hypothesis) “which seldom lutionary history of a test species is much more difficult. lend themselves to diagnosis” (Premack 1988, p. 179). No For example, everyone can agree that regardless of species matter how well an experiment has been counterbalanced or paradigm the number of pieces of food hidden left or or randomized, it is impossible to hypothesize whether a right must be counterbalanced. At the same time it is ex- null result was obtained because of the animal or because of tremely difficult to generate a set of objective principles the experiment. This is why the primate cognitive literature which might help distinguish between experiments with is replete with examples of how methodological changes high and low amounts of ecological validity for species as can dramatically change previously accepted results and in- diverse as aye-ayes, cotton-top tamarins, and chimpanzees terpretations (e.g., Oden et al. 1990; Boysen and Berntson (not to mention non-primates). The challenge is only 1995; Tomasello et al. 1998). Therefore, it is only after magnified when one considers the multitude of questions dozens of experiments and dozens of different approaches that experimenters wish to ask using an infinite number of have been utilized that one can feel confident that null re- paradigms. In addition, the problem may be exacerbated sults might accurately reflect the ability of a test species by the likelihood that the majority of comparative psy- (Heyes 1998). chologists studying primates never get the opportunity to Although methods of experimental control cannot help observe their test(s) species in their natural habitat. As a null results become more generalizable or interpretable, result of these difficulties, relative to the effort that has the second major component of validity potentially can. gone into developing methods of experimental control, lit- Assessing the external validity of an experiment, one puts tle attention has been focused onto the specifics of how to test subjects in experimental situations that somehow sim- maximize or even objectively evaluate the ecological va- ulate aspects of the real world that are relevant to the in- lidity of captive experiments on primate cognition. In fact, dividual or species (Gottsdanker 1978). If subjects are some of the most intrepid and successful cognitive exper- tested in settings with high amounts of