Adaptationism Carves Emotions at Their Functional Joints

We appreciate Agnes Moors’s (this issue) history of the debates among classical and constructionist emotion researchers and her attempt at integration. We have been pursuing an alternative perspective that emerges from evolutionary psychology, which integrates many aspects of the theories she discusses but embraces a universalism that Moors rejects. An adaptationist approach to the emotions takes an engineering perspective: Each emotion was designed by selection to solve problems that arose in a particular domain. That is, the detailed architecture of the emotion (the problem solver) was engineered by natural selection so that its design features functionally engaged the detailed structure of its particular adaptive problem (i.e., fear of predators reflects the dangerous properties of predators; sexual jealousy reflects the decision structure of mate choice). This means that a theory of what computational circuits each emotion will embody can be derived from analyzing the nature and properties of the emotion’s associated adaptive problem. For examples involving sexual jealousy, anger, shame, pride, gratitude, pathogen disgust, and sexual disgust, see Buss (2000); Fessler, (2001); Lieberman, Tooby, and Cosmides (2007); Lim (2012); Sell, Tooby, and Cosmides (2009); Sznycer Tooby, Cosmides, Porat, Shalvi, and Halperin (2016); Sznycer et al. (2016); Tooby, Cosmides, Sell, Lieberman, and Sznycer (2008); Tyber, Lieberman, Kurzban, and DeScioli (2013); and Weisfeld and Dillon (2012). Paradoxically, this view has led us to make many of the same points that Moors raises (Cosmides & Tooby, 2000; Tooby & Cosmides, 1990a, 2008). But they flow from the view that emotion is a fruitful scientific concept, not a mishmash of folk notions ready for the dustbin of history. We highlight five examples here. 1. It is true: Folk concepts, such as “emotion,” need not pick out natural kinds. That’s why it is better to study the evolved design of the computational systems that generate behavior and regulate physiology. Arguing about criteria for classifying these systems as “emotions” is pointless; questions like “What is a concept?” and “What is an emotion?” are matters of discovery, not stipulation. The systems that give rise to anger, shame, pride, gratitude, fear, jealousy, love, lust (etc.) will not have a uniform architecture defined by necessary and sufficient features. The architecture of each evolved system should reflect the computational requirements of the adaptive problem that selected for its design. Discovering the design of systems that give rise to phenomena that people think of as emotions is valuable. A good theory does not reify common sense, it explains it. Indeed, Moors’s point about the infinite variety of colors can be turned on its head. Understanding the evolved design of rods and cones, the psychophysics of similarity for color, and the visual system’s color constancy mechanisms illuminates why we see grass as green at both noon and sunset (when it is bathed in “red” [long wavelength] light); why red and violet— associated with the longest and shortest wavelengths of the visual spectrum—are seen as more similar than red and green; why we experience cherries and leaves as differing in color; and why it is easier to learn color labels for categories organized around focal colors (e.g., saturated red) than nonfocal ones (such as maroon or salmon). Universality and invariance are found at the level of evolved computational design. Folk ideas and experiences about color are thereby explained, not jettisoned as irrelevant to the psychology of color (Rosch, 1973; Shepard, 1992). 2. Verbal labels can indeed be misleading: Fear of predators and fear of losing a valued relationship are probably caused by distinct computational systems, each designed to solve a different ancestral adaptive problem. The odds of discovering this are low unless your research program is guided by theories of adaptive function. By saying that different emotions may be activated by both threats, we are not suggesting that researchers should endlessly multiply emotions of fear by adding a new label for each stimulus that can elicit it. If both situations flip mechanisms regulating attention, perception, inference, memory, goals, learning, behavior, physiology (etc.) into the same configuration, it would be more sensible to invoke a single emotion—fear—that can be activated by two distinct monitoring systems: one that sifts for cues that you are in danger from predators and another that sifts for cues that your relationships are in jeopardy. In our view, emotions are superordinate programs that evolved to solve problems of mechanism coordination (Cosmides & Tooby, 2000; Tooby & Cosmides, 1990a, 2008). The mind is a crowded zoo of functionally specialized programs, many endowed with content-rich procedures that are good at solving one kind of adaptive problem, but useless or even counterproductive if activated in response to others. Visual attention includes a category-specific system that monitors the location and state of nonhuman animals (New, Cosmides, & Tooby,