Measurement of physiological responses has long been an important tool for ethologists, comparative psychologists, and behavioral neuroscientists. From blood hormone assays, to electrophysiology, to the use of brain scanning methods, a variety of approaches of varying degrees of invasiveness have been applied to understanding psychological phenomena ranging from emotion and perception to decision making.

Now, Brügger et al. (2021) have demonstrated the utility of infrared thermography to study how common marmosets process socially relevant signals. A previous study from the same lab used thermography of the exposed skin of the monkeys’ noses in a study of emotional arousal (Ermatinger et al., 2019). In the new study, they measured temperature changes across situations of varying social significance. Specifically, they investigated thermal responses to playbacks of calls related to food sharing. Subjects heard single calls of three types: (i) an infant’s food-begging call, (ii) an unrelated adult’s food call, or (iii) an unrelated adult’s “chatter” call (typically used when there is a conflict); or they heard one of two pairs of calls: a (i)(ii) sequence representative of a positive interaction (food sharing), or a (i)(iii) sequence representative of a negative interaction (no sharing).

Brügger and her colleagues found, compared to baseline and controls, a decrease in temperature in nonbreeding female marmosets who heard a positive interaction playback. Nonbreeding males showed an increase for negative interactions, food calls alone, and chatter call alone. Breeding animals showed an increase in the chatter-alone condition and a decrease for negative interactions and for food calls alone. Modeling of these results showed that the thermal responses to playbacks of dyadic interactions could not be explained as an additive combination of the responses to each call component alone, indicating that the dyadic calls were processed “holistically” by the marmosets.

The pattern of variation in thermal responses between the three different categories of subjects was interpreted by the researchers as indicating different levels of arousal appropriate to the social structures surrounding breeding and food sharing. Nonbreeding marmosets assist breeding pairs, sharing food with the infants, providing a context in which interactions over food are potentially significant to them. A decrease in nasal temperature corresponds to blood flow being diverted away from the periphery in preparation for action, and may reflect both the motivational and emotional status of the animal. The cognitive interpretation was validated by measuring the latency of marmoset subjects to look into the chamber from which the playbacks emanated, establishing a preference for interaction with cooperative individuals. (I have omitted some details of the experimental setup that are not important to the points I make below.) They interpreted different responses among nonbreeding marmosets by sex and breeding marmosets irrespective of sex in light of the different implications of the different interactions for breeding opportunities or cooperation.

The experimental use of playbacks to support inferences about animal cognition stretches back over four decades (Seyfarth et al., 1980). The use of sequences for which an appropriate response to the whole is not the sum of the appropriate responses to the parts has also been used for nearly three of those decades (Cheney et al., 1995). But, as Brügger et al. point out, earlier studies were limited by their reliance on overt behavioral responses, such as looking towards the sound source, which may not always be present. At the same time, other physiological methods are either not feasible with freely moving subjects, or, as in the case of hormonal assays, may require animals to be handled in ways that may also affect their physiology and emotional state. Blood assays also lack good temporal resolution.

Passive thermography seems like an excellent solution to several of these issues. It is non-invasive, and can be carried out from a distance. It has a potential temporal resolution that is theoretically as good as the camera frame rate, although much less in practice because the temperature readings require subjects to be oriented towards the camera. To deal with this and other sources of noise and variability, Brügger et al. discarded trials for which they had too few frames and aggregated data into 1-second bins.

The use of skin temperature to investigate cognitive processes opens up several interesting lines of research, at least in species that have sufficiently large areas of bare skin, and a similar tendency to shunt blood from periphery to core. In the title of their paper, Brügger et al. ask “Do marmosets understand others’ conversations?” Unpacking the inferences that support their answer will help to illustrate both the potential of the approach and some places where more experiments are indicated.

The seemingly most straightforward inference is that the sequence of an infant’s begging call followed by an adult’s response produces thermal changes in some subjects that are not the sum of either piece of the sequence alone and are also not explained by the actual activity levels of the focal subjects during the experiment. The discovery of these differences depended on partitioning the marmoset subjects into three categories: breeders (male and female combined), nonbreeding females, and nonbreeding males (initially seven animals in each category for a total of 21 subjects). Not stated in the report is whether the decision to partition the data into to these three categories was made before the data were collected. Nevertheless, the partition could be justified on the grounds that male and female non-breeders have different overall reproductive interests from each other, and from breeders.

I mentioned above that initially there were seven animals in each category. Inspection of Fig. 2 (Brügger et al., 2021, p. 6) of the report reveals, however, that for one of the experimental conditions the number of breeding subjects was eight. This is because one of the females switched status during the study (Rahel Brügger, personal communication, 7 March 2021). This opens up some interesting research questions, such as how the pattern of responses changes when breeding status changes. Of course, pregnancy is likely to have measurable effects on blood flow between the core and the periphery, so measurements during the transition may represent several factors. Still, the prospect of investigating not just how interactions among others are processed, but how this processing might relate to an individual’s understanding of her own situation, is intriguing.

To this point I had avoided using the term “understanding” even though it is in the title of the article. Brügger et al. never define the term, but they consistently contrast understanding with the possibility that “the subjects independently, but simultaneously, reacted to the separate elements of the interaction” (2021, pp. 9-10). Cheney et al. (1995) introduced the idea that baboons process call sequences holistically and this has been pursued in other species (e.g., meerkats; Townsend et al., 2012). But what this says about understanding, if that term is taken to refer to explicit, concept-based cognition, is less clear. Townsend and colleagues argued that their results suggested meerkats do have a concept of conspecifics as individuals (2012, p. 183), but they were cautious not to make any claim about whether this concept might involve integration across different sensory modalities.

Some of the strength of the inferences in these earlier studies comes from using call sequences that are anomalous (Cheney et al., 1995) or impossible (Townsend et al., 2012), provoking enhanced looking times towards the source of the sound. This strategy contrasts with the present experiment where the social-ecological validity of the sequence is important to the interpretation of the physiological responses, and the claim of understanding is bolstered by the appropriateness of the marmoset responses given their breeding status within the species’ natural history. Further testing with both natural and unnatural sequences will be useful.

The use of thermography being pioneered in Judith Burkart’s lab is a welcome step towards more detailed, integrative theories of the cognitive and emotional processes underlying complex social behavior.