Although Rahnev & Denison (R&D) addressed the question of (sub)optimalities in perception, they mentioned evolutionary thinking only briefly. However, the theory of evolution can shed light on this important issue, suggesting that suboptimalities in perception (and other biological functions) are to be expected. The reason for this is twofold.
First, evolution by natural selection works if and only if three conditions are met: There is variability among the members of a species; the variability is hereditable; and there is differential reproduction based on this variability (e.g., Lewontin Reference Lewontin1970). If we apply this line of thinking to the perceptual realm, it implies that the perceptual systems of members of a species vary and differ in degree of fitness – some members perceive the environment more adaptively than other members. And although the power of “natural selection in the wild” (Endler Reference Endler1986) should not be underestimated, it is unlikely that it will eliminate all the variability in the evolving population (e.g., Mayr Reference Mayr2004). After all, many perceptual tasks are not crucial for survival and reproduction, allowing for a fair amount of variability and suboptimality among the members of a species (e.g., Withagen & Chemero Reference Withagen and Chemero2009).
A second reason why suboptimalities are to be expected is that there are always multiple constraints that have a severe influence on evolving organisms. As Gould and Lewontin (Reference Gould and Lewontin1978) noted, organisms are “integrated wholes, with Baupläne […] constrained by phyletic heritage, pathways of development, and general architecture” (p. 581; emphasis in original). These constraints (e.g., Dawkins Reference Dawkins1982; Gould & Lewontin Reference Gould and Lewontin1978; Mayr Reference Mayr2002) imply that it is unlikely that the best possible (i.e., optimal) perceptual system is always available in the population that natural selection is working on (Withagen Reference Withagen2004).
Although an evolutionary analysis of perception predicts the suboptimalities that R&D discussed and defended to some extent, it questions the overall Bayesian approach that they adopted. As Darwin (Reference Darwin1859/1985) had already claimed toward the end of On the Origin of Species, his theory of evolution by natural selection provides psychology with “a new foundation” (p. 458). Yet when it comes to theorizing about perception, many post-Darwinian theories of perception follow pre-Darwinian theories both in the conceptualization of what perception is (i.e., a mental state residing in the head) and how it comes about (i.e., by means of inferential processes). The Bayesian account that R&D adopted is no exception. Although this account is often connected to Helmholtz's perspective, ultimately the gist of both accounts can be traced back to Descartes’ theory of perception—the stimulus information that reaches the senses is impoverished; hence, inferential processes are needed to gain knowledge of the environment (Reed Reference Reed1982). In fact, when theorists of perception adopt evolutionary thinking, they often consider it as an “afterthought,” one that does not affect the fundamental principles of their theorizing (Heft Reference Heft2007, p. 92).
Gibson was one of the first to complain about this situation. As he put it in his book The Senses Considered as Perceptual Systems, “[t]he classics of vision were unaffected by evolutionary considerations or by knowledge of animal behavior but nevertheless they dominate the theories of perception” (Gibson Reference Gibson1966, p. 155). Gibson took evolutionary considerations seriously and argued that they require us to rethink the fundamentals of the science of perception. We agree. After all, from an evolutionary perspective, the main function of perception is not to gain knowledge of the environment, as many theories of perception hold, but to guide our actions in the environment (e.g., Anderson Reference Anderson2014; Chemero Reference Chemero2009). Animals have to perceive what Gibson called affordances, the action possibilities the environment provides them.
Crucially for present purposes, Gibson claimed that an evolutionary approach questions the dominant assumption of impoverished stimulus information that, as mentioned earlier, also underlies the Bayesian approach. Fundamentally, this assumption implies that the animal is in a state of uncertainty about the environment. As Müller (Reference Müller1837–1840/1938), Helmholtz’ mentor, put it, “[i]n our intercourse with external nature it is always our own sensations that we become acquainted with, and from them we form conceptions of the properties of external objects, which may be relatively correct” (p. 1068; emphases added). If animals are in such a state of uncertainty about what is “out there” in the environment, it is hard to explain how animals can generally adaptively cope with the environment. Hence, Gibson replaced the representational theory of perception with a so-called contact theory of perception (Dreyfus & Taylor Reference Dreyfus and Taylor2015). In his view, perception is not a mental state in the head, but a “keeping-in-touch with the world” (Gibson Reference Gibson1979/1986, p. 239). Gibson argued that this direct contact can be established because there is information in the ambient energy arrays that specifies the affordances, that is, that informs about them. And if animals detect these specifying variables, a direct and adaptive perceptual contact with the affordances in their environment is established. Notably, the availability of such rich information obviates the need for inference, be it Helmholtzian or Bayesian.
One might argue that this focus on specifying variables is not in keeping with the suboptimal performances that have been demonstrated in empirical studies and are implied by the above evolutionary analysis. Indeed, to be a truly evolutionary view, the Gibsonian perspective needs to recognize that animals (occasionally) behave and perceive suboptimally (e.g., Chemero Reference Chemero2009; Withagen Reference Withagen2004). However, suboptimality (or optimality, for that matter) need not be explained in Bayesian terms of uncertainty, but can also be accounted for in terms of the informational variables that animals detect (e.g., Jacobs & Michaels Reference Jacobs and Michaels2007). The numerous patterns in the ambient energy arrays that animals might possibly detect differ in degree of usefulness, and the adaptiveness of the perceptual grip on, and actions in the environment is determined by which of these variables is picked up (e.g., de Wit et al. Reference de Wit, van der Kamp and Withagen2015; Withagen Reference Withagen2004; Withagen & Chemero Reference Withagen and Chemero2009). Contrary to the Bayesian approach that R&D adopted, such a perspective stays closer to the fundamental change in thinking that Darwin's evolutionary program implies.
You have
Access
Although Rahnev & Denison (R&D) addressed the question of (sub)optimalities in perception, they mentioned evolutionary thinking only briefly. However, the theory of evolution can shed light on this important issue, suggesting that suboptimalities in perception (and other biological functions) are to be expected. The reason for this is twofold.
First, evolution by natural selection works if and only if three conditions are met: There is variability among the members of a species; the variability is hereditable; and there is differential reproduction based on this variability (e.g., Lewontin Reference Lewontin1970). If we apply this line of thinking to the perceptual realm, it implies that the perceptual systems of members of a species vary and differ in degree of fitness – some members perceive the environment more adaptively than other members. And although the power of “natural selection in the wild” (Endler Reference Endler1986) should not be underestimated, it is unlikely that it will eliminate all the variability in the evolving population (e.g., Mayr Reference Mayr2004). After all, many perceptual tasks are not crucial for survival and reproduction, allowing for a fair amount of variability and suboptimality among the members of a species (e.g., Withagen & Chemero Reference Withagen and Chemero2009).
A second reason why suboptimalities are to be expected is that there are always multiple constraints that have a severe influence on evolving organisms. As Gould and Lewontin (Reference Gould and Lewontin1978) noted, organisms are “integrated wholes, with Baupläne […] constrained by phyletic heritage, pathways of development, and general architecture” (p. 581; emphasis in original). These constraints (e.g., Dawkins Reference Dawkins1982; Gould & Lewontin Reference Gould and Lewontin1978; Mayr Reference Mayr2002) imply that it is unlikely that the best possible (i.e., optimal) perceptual system is always available in the population that natural selection is working on (Withagen Reference Withagen2004).
Although an evolutionary analysis of perception predicts the suboptimalities that R&D discussed and defended to some extent, it questions the overall Bayesian approach that they adopted. As Darwin (Reference Darwin1859/1985) had already claimed toward the end of On the Origin of Species, his theory of evolution by natural selection provides psychology with “a new foundation” (p. 458). Yet when it comes to theorizing about perception, many post-Darwinian theories of perception follow pre-Darwinian theories both in the conceptualization of what perception is (i.e., a mental state residing in the head) and how it comes about (i.e., by means of inferential processes). The Bayesian account that R&D adopted is no exception. Although this account is often connected to Helmholtz's perspective, ultimately the gist of both accounts can be traced back to Descartes’ theory of perception—the stimulus information that reaches the senses is impoverished; hence, inferential processes are needed to gain knowledge of the environment (Reed Reference Reed1982). In fact, when theorists of perception adopt evolutionary thinking, they often consider it as an “afterthought,” one that does not affect the fundamental principles of their theorizing (Heft Reference Heft2007, p. 92).
Gibson was one of the first to complain about this situation. As he put it in his book The Senses Considered as Perceptual Systems, “[t]he classics of vision were unaffected by evolutionary considerations or by knowledge of animal behavior but nevertheless they dominate the theories of perception” (Gibson Reference Gibson1966, p. 155). Gibson took evolutionary considerations seriously and argued that they require us to rethink the fundamentals of the science of perception. We agree. After all, from an evolutionary perspective, the main function of perception is not to gain knowledge of the environment, as many theories of perception hold, but to guide our actions in the environment (e.g., Anderson Reference Anderson2014; Chemero Reference Chemero2009). Animals have to perceive what Gibson called affordances, the action possibilities the environment provides them.
Crucially for present purposes, Gibson claimed that an evolutionary approach questions the dominant assumption of impoverished stimulus information that, as mentioned earlier, also underlies the Bayesian approach. Fundamentally, this assumption implies that the animal is in a state of uncertainty about the environment. As Müller (Reference Müller1837–1840/1938), Helmholtz’ mentor, put it, “[i]n our intercourse with external nature it is always our own sensations that we become acquainted with, and from them we form conceptions of the properties of external objects, which may be relatively correct” (p. 1068; emphases added). If animals are in such a state of uncertainty about what is “out there” in the environment, it is hard to explain how animals can generally adaptively cope with the environment. Hence, Gibson replaced the representational theory of perception with a so-called contact theory of perception (Dreyfus & Taylor Reference Dreyfus and Taylor2015). In his view, perception is not a mental state in the head, but a “keeping-in-touch with the world” (Gibson Reference Gibson1979/1986, p. 239). Gibson argued that this direct contact can be established because there is information in the ambient energy arrays that specifies the affordances, that is, that informs about them. And if animals detect these specifying variables, a direct and adaptive perceptual contact with the affordances in their environment is established. Notably, the availability of such rich information obviates the need for inference, be it Helmholtzian or Bayesian.
One might argue that this focus on specifying variables is not in keeping with the suboptimal performances that have been demonstrated in empirical studies and are implied by the above evolutionary analysis. Indeed, to be a truly evolutionary view, the Gibsonian perspective needs to recognize that animals (occasionally) behave and perceive suboptimally (e.g., Chemero Reference Chemero2009; Withagen Reference Withagen2004). However, suboptimality (or optimality, for that matter) need not be explained in Bayesian terms of uncertainty, but can also be accounted for in terms of the informational variables that animals detect (e.g., Jacobs & Michaels Reference Jacobs and Michaels2007). The numerous patterns in the ambient energy arrays that animals might possibly detect differ in degree of usefulness, and the adaptiveness of the perceptual grip on, and actions in the environment is determined by which of these variables is picked up (e.g., de Wit et al. Reference de Wit, van der Kamp and Withagen2015; Withagen Reference Withagen2004; Withagen & Chemero Reference Withagen and Chemero2009). Contrary to the Bayesian approach that R&D adopted, such a perspective stays closer to the fundamental change in thinking that Darwin's evolutionary program implies.