Henrich et al. argue – correctly, in our view – that there is great diversity both across and within domains of human cognition. They then argue for the stronger claim that there is no underlying rhyme or reason for this diversity. This latter claim grounds their central methodological point: For all (interesting and not obviously culturally induced) cognitive phenomena, psychologists ought not automatically conclude, from samples of Western undergraduates, that any particular cognitive phenomenon is uniform across people. This methodological conclusion obviously depends crucially on the stronger descriptive claim. If a pattern of uniformity emerged with respect to a particular type of cognitive phenomenon, then psychologists could justifiably generalize (in only those particular domains, of course). We argue here that such a pattern does emerge, if one carves up the space of cognitive phenomena appropriately. More precisely, we suggest that there is diversity in cognitive “contents” (broadly construed), but uniformity in learning “processes” (broadly construed).

Cognitive “phenomena” do not form a unitary type, as there is a natural, defensible distinction between the cognitive “objects” of the mind, and dynamic mental “processes.” Cognitive objects include representations, knowledge structures, and so on. Mental processes influence both those cognitive objects, and also other processes. As one example of this distinction, there is a natural difference between a concept and the processes by which one learns that concept. Concepts are particular representations that can be deployed in a range of circumstances; concept learning processes are the means by which those representations emerge. Importantly, the word process in our argument should explicitly not be read as a synonym for “specific causal mechanism”; the distinction between a process and its “target” requires only a thin notion of “process” (i.e., something like “if-then” rules, not a specific neural mechanism). Furthermore, our argument does not depend on the particular ways in which learning processes are instantiated in different individuals, as long as those processes have the same computational/psychological profile.

We have no doubt that Henrich et al. are correct that there is great diversity in representations, including people's folk-biological concepts, notions of “fairness” and “agency,” and moral concepts. It is unsurprising that people who develop in different environments – physical, mental, or social – develop different representations of those environments. But diversity in representation does not imply diversity in learning processes. For example, diversity in our learned concepts is arguably due to differences in inputs, not differences in concept learning processes. After all, the whole point of learning is that the same process can produce very different outputs depending on the environment, and thereby enable the organism to adapt to the demands of an unpredictable or ever-changing environment. There are thus prima facie reasons – genetic and evolutionary – to think that there is likely to be uniformity in learning processes. Of course, part of the point of Henrich et al.'s article is that we ought not rest with the assertion that “there really ought to be uniformity in learning processes”; instead, it is an empirical matter as to whether there actually is such uniformity.

Determining whether learning processes are uniform across both WEIRD and non-WEIRD individuals requires discovering the learning processes in each individual, which is inevitably a tricky matter. We can rarely (if ever) directly observe a learning process; instead, we must find its characteristics by determining how the relevant cognitive objects change in response to different inputs (either within- or between-participants). Despite these difficulties, it is nonetheless often possible to learn about different individuals' learning processes. Unfortunately, as Henrich et al.'s descriptive survey shows, there have been relatively few cross-cultural studies of learning processes, as opposed to representations or (relatively) automatic processing. By our reading, essentially all of the phenomena that Henrich et al. discuss are representations/contents, rather than learning processes. There are many psychological studies of learning processes, but almost entirely restricted to WEIRD participants. We must therefore look to more indirect evidence.

It is suggestive that many of the cases of cognitive uniformity that Henrich et al. identify are also cases in which the environment is plausibly (relatively) uniform. On the “uniformity in learning processes” view we advocate, one would predict this sort of representational uniformity just when there is relative uniformity in the relevant environment. For example, it is (in our view) unsurprising that the ability to pass the False Belief Task emerges across cultures, since (presumably) all developing individuals have social environments that require interaction with (and prediction of the behavior of) individuals with false beliefs.

Moreover, Henrich et al. themselves seem to assume uniformity of learning processes in some of their explanations for diversity. For example, they seem to endorse an explanation of diversity in the Müller-Lyer illusion in terms of variation in the visual environments of developing individuals. That is, Henrich et al. seem to argue that differences in adult visual representations are due to differences in infant/child visual environments. But implicit in this argument is an assumption that there is an “infant environment→adult visual representation” mapping that is shared between both WEIRD and non-WEIRD individuals. More generally, all explanations of diversity in terms of environmental variation presuppose that there is relative uniformity in learning processes (in our thin sense). Explanations based in environmental variation require some shared process by which that variation can lead to behavioral or cognitive differences. If there is no shared learning process, then an appeal to environmental differences is simply a non sequitur.

Learning processes (in a thin sense) are different from the objects they influence, and this distinction can provide a basis for when to expect uniformity and diversity in human cognition. We suggest that uniformity should be expected for learning processes, and diversity should be expected for objects whenever the relevant learning environments differ (and there are no relevant, direct selection pressures on the objects). We do not doubt that people are more diverse than contemporary cognitive psychology admits, and Henrich et al. do a great service in providing a descriptive characterization of diversity. They simply paint with too broad a methodological brush.