Jones & Love (J&L) provide important theoretical contributions to psychology and Bayesian modeling. Especially illuminating is their discussion of whether Bayesian models are agnostic about psychology, serving mainly as useful scientific and mathematical tools, or instead make substantive claims about cognition.
Despite its many strengths, the target article succumbs to some common misconceptions about evolutionary psychology (EP) (Confer et al. Reference Confer, Easton, Fleischman, Goetz, Lewis, Perilloux and Buss2010). The first is an erroneous characterization of EP's approach to optimality and constraints. Although the article acknowledges the importance of constraints in evolutionary theory, it lapses into problematic statements such as “evolutionary pressures tune a species’ genetic code such that the observed phenotype gives rise to optimal behaviors” (sect. 5, para. 3). J&L suggest that evolutionary psychologists reinterpret behavioral phenomena as “optimal” by engaging in a post hoc adjustment of their view of the relevant selection pressures operating in ancestral environments.
These statements imply that a key goal of EP is to look for optimality in human behavior and psychology. On the contrary, the existence of optimized mechanisms is rejected by evolutionary psychologists, as this passage from Buss et al. (Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998) illustrates:
[T]ime lags, local optima, lack of genetic variation, costs, and limits imposed by adaptive coordination with other mechanisms all constitute major constraints on the design of adaptations. . . . Adaptations are not optimally designed mechanisms. They are . . . jerry-rigged, meliorative solutions to adaptive problems . . ., constrained in their quality and design by a variety of historical and current forces. (Buss et al. Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998, p. 539)
J&L argue that “it is not [simply] any function that is optimized by natural selection, but only those functions that are relevant to fitness” (sect. 5, para. 4). We agree with the implication that psychologists must consider the fitness-relevance of the mechanisms they choose to investigate. Identifying adaptive function is central. Nonetheless, natural selection is better described as a “meliorizing” force, not an optimizing force (see Dawkins Reference Dawkins1982, pp. 45–46) – and thus even psychological mechanisms with direct relevance to fitness are not optimized. As J&L correctly note elsewhere, selection does not favor the best design in some global engineering sense, but rather features that are better than competing alternatives extant in the population at the time of selection, within existing constraints (Buss et al. Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998; Dawkins Reference Dawkins1982).
Despite occasional problems with the target article's depiction of EP's views on optimality, we fully agree with J&L that (a) adaptationist accounts place significant constraints on explanation, (b) evolution proceeds by “survival of the best current design, not survival of the globally optimal design” (sect. 5.3, para. 3), (c) human cognition is not optimally designed, and (d) the “rational program” in Bayesian modeling has an overly narrow focus on optimally functioning adaptations.
J&L present a partly accurate and partly inaccurate characterization of the relevance of mechanism in evolutionary approaches. They correctly acknowledge the importance of elucidating the specific mechanistic workings of adaptations. However, the target article compares EP to Bayesian Fundamentalism and Behaviorism by claiming that all three approaches eschew the investigation of mechanism. We disagree with this latter assessment.
In our view, it is difficult or impossible to study function without investigating form or mechanism. The central logic of adaptationism makes the inextricable link between form (or mechanism) and function clear: An adaptation must necessarily be characterized by a good fit between form and function – between an adaptation and the adaptive problem it was “designed” to solve .The key point is that evolutionary approaches to psychology necessarily involve the joint investigation of mechanism and function. Evolutionary psychology generates hypotheses about “design features,” or particular mechanistic attributes, that adaptations either must have or might have in order to successfully solve the adaptive problems that they evolved to solve. Indeed, mechanism is one of Tinbergen's (Reference Tinbergen1963) four explanatory levels – mechanism, ontogeny, function, and phylogeny. Ideally, all should be analyzed in order to achieve a complete understanding of any behavior or psychological phenomenon, and all are central to core aims of EP. Of course, not every scientist explores all four questions; every empirical study has delimited aims; and the field is certainly far from a complete understanding of all of the design features of any mechanism, whether it be the human visual system or incest-avoidance adaptations.
As a single example of mechanistic EP research, adaptationist analyses of fear have uncovered social inputs that elicit the emotion, nonsocial inputs that trigger the emotion, the adaptive behavioral output designed to solve the problem, the perceptual processes involved in detecting threats and reacting fearfully, the developmental trajectory of human fears, and the physiological and endocrinological mechanisms driving the fear response (see, e.g., Bracha Reference Bracha2004; Buss Reference Buss2011; Neuhoff Reference Neuhoff2001; Öhman et al. Reference Öhman, Flykt and Esteves2001). Analogous progress has been made in understanding other evolved mechanisms, such as mating adaptations, perceptual biases, and adaptive social inference biases (Buss Reference Buss2011).
Most human adaptations are only just beginning to be subjected to scientific investigation, and many mechanistic details have certainly not yet been elucidated. EP could profitably increase its use of formal mechanistic modeling in this endeavor. Fusing the strengths of mathematical and computational modelers with those of evolutionary psychologists would enrich both fields.
Finally, the target article depicts EP as occasionally falling into “backward-looking” hypotheses (sect. 5.2, para. 3) or engaging in “just so” storytelling (sect. 5.2, para. 1; Gould & Lewontin Reference Gould and Lewontin1979). By this, the authors mean that evolutionary psychologists sometimes note a behavior or psychological mechanism, and then construct a conceivable function for it and simply stop there. We agree with J&L that this practice would be highly problematic if it were the end point of scientific analysis.
Fortunately, leading work in EP proceeds using both the forward method in science (theory leads directly to hypothesis, which then leads to empirical predictions, which are then tested) as well as the backward method (observed phenomenon leads to hypothesis, which in turn leads to novel empirical predictions, which are then tested) (see Buss Reference Buss2011; Tooby & Cosmides Reference Tooby, Cosmides, Barkow, Cosmides, Tooby, Barkow, Cosmides and Tooby1992). Much of evolutionary psychology uses the forward method, and here it is not even possible to level the “just-so story” criticism. When evolutionary psychologists employ the backward method, they typically avoid the problem by taking the additional necessary step of deriving novel and previously untested predictions from the hypothesis (for numerous examples, see Buss Reference Buss2011). We concur with the implication that there are better and poorer practitioners of the rigors of science, and that all should be held to the highest standards for more rapid progress.
In sum, we view an accurately characterized modern evolutionary psychology as largely avoiding the conceptual pitfalls J&L note, and we look forward to a richer and more successful integration of Bayesian modeling and evolutionary psychology.
Jones & Love (J&L) provide important theoretical contributions to psychology and Bayesian modeling. Especially illuminating is their discussion of whether Bayesian models are agnostic about psychology, serving mainly as useful scientific and mathematical tools, or instead make substantive claims about cognition.
Despite its many strengths, the target article succumbs to some common misconceptions about evolutionary psychology (EP) (Confer et al. Reference Confer, Easton, Fleischman, Goetz, Lewis, Perilloux and Buss2010). The first is an erroneous characterization of EP's approach to optimality and constraints. Although the article acknowledges the importance of constraints in evolutionary theory, it lapses into problematic statements such as “evolutionary pressures tune a species’ genetic code such that the observed phenotype gives rise to optimal behaviors” (sect. 5, para. 3). J&L suggest that evolutionary psychologists reinterpret behavioral phenomena as “optimal” by engaging in a post hoc adjustment of their view of the relevant selection pressures operating in ancestral environments.
These statements imply that a key goal of EP is to look for optimality in human behavior and psychology. On the contrary, the existence of optimized mechanisms is rejected by evolutionary psychologists, as this passage from Buss et al. (Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998) illustrates:
[T]ime lags, local optima, lack of genetic variation, costs, and limits imposed by adaptive coordination with other mechanisms all constitute major constraints on the design of adaptations. . . . Adaptations are not optimally designed mechanisms. They are . . . jerry-rigged, meliorative solutions to adaptive problems . . ., constrained in their quality and design by a variety of historical and current forces. (Buss et al. Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998, p. 539)
J&L argue that “it is not [simply] any function that is optimized by natural selection, but only those functions that are relevant to fitness” (sect. 5, para. 4). We agree with the implication that psychologists must consider the fitness-relevance of the mechanisms they choose to investigate. Identifying adaptive function is central. Nonetheless, natural selection is better described as a “meliorizing” force, not an optimizing force (see Dawkins Reference Dawkins1982, pp. 45–46) – and thus even psychological mechanisms with direct relevance to fitness are not optimized. As J&L correctly note elsewhere, selection does not favor the best design in some global engineering sense, but rather features that are better than competing alternatives extant in the population at the time of selection, within existing constraints (Buss et al. Reference Buss, Haselton, Shackelford, Bleske and Wakefield1998; Dawkins Reference Dawkins1982).
Despite occasional problems with the target article's depiction of EP's views on optimality, we fully agree with J&L that (a) adaptationist accounts place significant constraints on explanation, (b) evolution proceeds by “survival of the best current design, not survival of the globally optimal design” (sect. 5.3, para. 3), (c) human cognition is not optimally designed, and (d) the “rational program” in Bayesian modeling has an overly narrow focus on optimally functioning adaptations.
J&L present a partly accurate and partly inaccurate characterization of the relevance of mechanism in evolutionary approaches. They correctly acknowledge the importance of elucidating the specific mechanistic workings of adaptations. However, the target article compares EP to Bayesian Fundamentalism and Behaviorism by claiming that all three approaches eschew the investigation of mechanism. We disagree with this latter assessment.
In our view, it is difficult or impossible to study function without investigating form or mechanism. The central logic of adaptationism makes the inextricable link between form (or mechanism) and function clear: An adaptation must necessarily be characterized by a good fit between form and function – between an adaptation and the adaptive problem it was “designed” to solve .The key point is that evolutionary approaches to psychology necessarily involve the joint investigation of mechanism and function. Evolutionary psychology generates hypotheses about “design features,” or particular mechanistic attributes, that adaptations either must have or might have in order to successfully solve the adaptive problems that they evolved to solve. Indeed, mechanism is one of Tinbergen's (Reference Tinbergen1963) four explanatory levels – mechanism, ontogeny, function, and phylogeny. Ideally, all should be analyzed in order to achieve a complete understanding of any behavior or psychological phenomenon, and all are central to core aims of EP. Of course, not every scientist explores all four questions; every empirical study has delimited aims; and the field is certainly far from a complete understanding of all of the design features of any mechanism, whether it be the human visual system or incest-avoidance adaptations.
As a single example of mechanistic EP research, adaptationist analyses of fear have uncovered social inputs that elicit the emotion, nonsocial inputs that trigger the emotion, the adaptive behavioral output designed to solve the problem, the perceptual processes involved in detecting threats and reacting fearfully, the developmental trajectory of human fears, and the physiological and endocrinological mechanisms driving the fear response (see, e.g., Bracha Reference Bracha2004; Buss Reference Buss2011; Neuhoff Reference Neuhoff2001; Öhman et al. Reference Öhman, Flykt and Esteves2001). Analogous progress has been made in understanding other evolved mechanisms, such as mating adaptations, perceptual biases, and adaptive social inference biases (Buss Reference Buss2011).
Most human adaptations are only just beginning to be subjected to scientific investigation, and many mechanistic details have certainly not yet been elucidated. EP could profitably increase its use of formal mechanistic modeling in this endeavor. Fusing the strengths of mathematical and computational modelers with those of evolutionary psychologists would enrich both fields.
Finally, the target article depicts EP as occasionally falling into “backward-looking” hypotheses (sect. 5.2, para. 3) or engaging in “just so” storytelling (sect. 5.2, para. 1; Gould & Lewontin Reference Gould and Lewontin1979). By this, the authors mean that evolutionary psychologists sometimes note a behavior or psychological mechanism, and then construct a conceivable function for it and simply stop there. We agree with J&L that this practice would be highly problematic if it were the end point of scientific analysis.
Fortunately, leading work in EP proceeds using both the forward method in science (theory leads directly to hypothesis, which then leads to empirical predictions, which are then tested) as well as the backward method (observed phenomenon leads to hypothesis, which in turn leads to novel empirical predictions, which are then tested) (see Buss Reference Buss2011; Tooby & Cosmides Reference Tooby, Cosmides, Barkow, Cosmides, Tooby, Barkow, Cosmides and Tooby1992). Much of evolutionary psychology uses the forward method, and here it is not even possible to level the “just-so story” criticism. When evolutionary psychologists employ the backward method, they typically avoid the problem by taking the additional necessary step of deriving novel and previously untested predictions from the hypothesis (for numerous examples, see Buss Reference Buss2011). We concur with the implication that there are better and poorer practitioners of the rigors of science, and that all should be held to the highest standards for more rapid progress.
In sum, we view an accurately characterized modern evolutionary psychology as largely avoiding the conceptual pitfalls J&L note, and we look forward to a richer and more successful integration of Bayesian modeling and evolutionary psychology.