Sinking In: The Peripheral Baldwinisation of Human Cognition. Cecilia Heyes, Nick Chater & Dominic Michael Dwyer. Trends in Cognitive Sciences, 2020.

Some theories have proposed that humans have evolved to experience some stimuli (e.g. snakes, spiders) as more potentially frightening, so that a fear for these entities is learned faster than a fear for more neutral things. In evolutionary psychology, there has been talk about modules for a fear of snakes, for example. However, research suggests that rather than “the fear system” itself having innate biases towards picking up particular kinds of fears, humans are evolutionarily biased towards paying extra attention to things like spiders and snakes. Because of these stimuli being more attended than others, it also becomes more probable that a fear response gets paired with them.

The authors call the attention system “peripheral” and the fear system “central”, in that the attention system brings in information for the fear system to process. (This is in analogy to the peripherals of a computer, where e.g. the keyboard and mouse are used to deliver information to the central processor.) They argue that in general, while it is possible for responses to specific environmental stimuli to become genetic as sensitivty for those stimuli is selected for, this learning will be more likely to get encoded into “peripheral” than “central” systems.

One of their other examples is that the central mechanisms of language learning seem theoretically and empirically unlikely to be affected by the environment – there are no genes for learning English grammar better than Chinese grammar. However, there are indications that the peripheral mechanisms of language have been more affected. E.g. some languages use lexical tone (where word identities are partly defined by pitch contours), and genes that seem to make lexical tone easier to perceive seem to be more common among speakers of those languages.

Seligman’s account suggested that specialised, central mechanisms of fear learning more readily connect aversive events, such as electric shock, with fear-relevant stimuli, such as snakes – which presented genuine threats to our evolutionary ancestors – than with ‘fear-irrelevant’ stimuli such as geometric shapes or flowers. This account predicts that fear of fear-relevant objects should be learned faster, and be extinguished more slowly when shock no longer occurs, as well as being resistant to topdown modification, for example, by instructions indicating that shocks will not occur.

The results of early experiments were consistent with some of these predictions (e.g., [50,51]), but none has withstood extended experimental investigation. Faster or better conditioning with fear-relevant stimuli has rarely been observed, and there is ample evidence that, like most associative learning (e.g., [52]), it can be modified by instruction (reviewed in [53,54]). Initially it seemed that responses to fear-relevant stimuli might extinguish more slowly. However, a recent systematic review [55] found that most positive findings came from a single laboratory, and a large majority of the full set of studies had failed to find differences between fear-relevant and fear-irrelevant stimuli in the rate of extinction.

These results suggest that fear of snakes and other fear-relevant stimuli is learned via the same central mechanisms as fear of arbitrary stimuli. Nevertheless, if that is correct, why do phobias so often relate to objects encountered by our ancestors, such as snakes and spiders, rather than to objects such as guns and electrical sockets that are dangerous now [10]? Because peripheral, attentional mechanisms are tuned to fear-relevant stimuli, all threat stimuli attract attention, but fear-relevant stimuli do so without learning (e.g., [56]). This answer is supported by evidence from conditioning experiments demonstrating enhanced attention to fear-relevant stimuli regardless of learning (Box 2), studies of visual search [57–59], and developmental psychology [60,61]. For example, infants aged 6–9 months show a greater drop in heart rate – indicative of heightened attention rather than fear – when they watch snakes than when they watch elephants [62].

In sum: early research on taste-aversion and fear learning launched the idea that animal minds are populated by adaptively specialised central learning mechanisms – that were later cast by evolutionary psychologists as ‘modules’. Over the past 50 years careful experimental work with rodents, using the original methods, has confirmed the occurrence of adaptive specialisation, and more recent studies on Drosophila have shown that it is likely to have occurred via a Baldwinian process. However, the rodent work has also shown that the changes are in peripheral rather than in central cognitive mechanisms. […]

Why might selection operate primarily at the cognitive periphery? A parallel with the evolution of other biological mechanisms is suggestive: internal physiological processes and anatomical structures are remarkably well-conserved. The organisation of the digestive, circulatory, and respiratory systems is similar across vertebrate species, and they are so deeply interconnected that modifications beyond changes of size and shape may be difficult without causing substantial collateral damage. Moreover, even such modest changes to central systems will impact on a wide variety of functions and may therefore not be under strong selection from any one function. By contrast, interfaces with the external environment (jaws, teeth, digestive enzymes, bone and muscle structure) can be adapted to local circumstances (e.g., food sources) without interfering with central systems. The central machinery of cognition is less well understood, but may be equally interlocking, with widespread functional ramification, and a consequent resistance to evolutionary change.

Alternatively, it is possible that central cognitive processes are fully evolvable, but, at least in the human case, tend to be adaptively specialised by cultural rather than by genetic selection [101]. In domains such as language, imitation, mathematics, and ethics, changes to central mechanisms can be acquired through cultural learning. Cognitive skills that are taught, and those that are learned from others through more informal social interaction, do not need to sink in. Baldwinisation would bring little if any fitness advantage for skills that are reliably inherited via a non-genetic route [17], and specialised central mechanisms may be more teachable than specialised peripheral mechanisms. Plausibly, it is easier to learn grammatical constructions than vocal control through conversation, and, in the case of imitation, easier to learn sensorimotor mappings than intrinsic motivation through non-vocal social interaction.