Do Monkeys and Young Children Understand Exclusive “Or” Relations? A Commentary on Ferrigno et al. (2021)

Claims that nonhuman animals can reason about “or” relations date back to antiquity (see Rescorla, 2009). In the 3rd century B.C.E., Chrysippus offered an anecdote of a dog that chased a scent to a juncture of multiple roads. After failing to find the scent in all but one of the roads, the dog proceeded to run down the final road without bothering to sniff it, apparently because it inferred that the source of the scent must have taken that road. Such reasoning involves the disjunctive syllogism: first representing the “or” relation between two dependent possibilities (A or B) and then when one possibility is ruled out (not A), inferring that the other must be true (therefore B). In their recent article, Ferrigno et al. (2021) claim to provide empirical evidence that animals, in this case baboons, can indeed reason using the disjunctive syllogism and therefore understand such “or” relations. Here, we draw attention to the results of a similar study we recently conducted with human children (Gautam et al., 2021) and explain why Ferrigno et al.’s conclusion is incomplete and misleading.

Adapting earlier paradigms (Call, 2004; Mody & Carey, 2016), Ferrigno et al. first hid rewards within two pairs of containers—hiding one reward within each pair (A or B; C or D)—before giving baboons two sequential choices between all four containers. Given that each pair of containers concealed exactly one reward, the experiment was essentially a test of the baboons’ understanding of exclusive “or” relations. That is, if the first choice was shown to be empty, then the other container within that pair must have concealed a reward and should have been the next logical choice via the disjunctive syllogism (not A, therefore B; disregard C and D). If the first choice was shown to contain a reward, however, then the other container within that pair must have been empty and should have been logically ruled out as an option for the next choice (A, therefore not B; guess between C and D). Across individuals, the baboons performed significantly above the a priori chance level of 33.3% when the first container was revealed to be empty but no differently from the a priori chance level of 66.7% when the first container was revealed to be baited.

Commenting on the monkeys’ poor performance when the first container was baited, Ferrigno et al. write that “it is currently unknown whether young children could pass these types of trials because only trials in which the first cylinder was empty were included in previous work testing young children (Mody & Carey, 2016)” (Ferrigno et al., 2021, p. 297). We are happy to point out that we independently conducted a near-identical experiment with young children (Gautam et al., 2021) that was published around the same time as Ferrigno et al.’s study. The only major difference in our setup was that a sock puppet controlled by an experimenter, rather than the subjects themselves, revealed the initial empty or baited container. We found that 2.5-, 3-, and 4-year old children—like the monkeys—performed significantly above the a priori chance level of 33.3% when the first container was empty but no differently from the a priori chance level of 66.7% when the first container was baited. By contrast, 5-year-olds performed significantly above these chance levels in both conditions.

The convergence in results between 2.5- to 4-year-olds and baboons suggests that similar cognitive processes may underlie the responses of both populations. Unlike Ferrigno et al., however, we do not believe that the performance of the younger children, nor of the baboons, amounts to compelling evidence that they understood the “or” relations central to the task. In the disjunctive syllogism, the relation “or” in the premise “A or B” can be inclusive (at least one of A and B is true) or exclusive (exactly one of A and B is true). Yet critically, as outlined above, in the studies of both Ferrigno et al. and Gautam et al. (2021), the relation was in fact exclusive “or,” as there was exactly one baited container in each pair. Therefore, if the monkeys and young children correctly appreciated the logical structure of the task, they should have been able to just as easily rule out a container with 0% chance of reward (after initially seeing a baited container) as to rule in a container with 100% chance of reward (after initially seeing an empty container). But they did not.

Ferrigno et al. propose that the baboons’ failure on trials in which the initial container was baited can simply be attributed to the high chance level on these trials. Yet across individuals, the baboons’ performance (66.3%) almost exactly matched the chance level (66.7%). We do not think that the authors’ suggestion that this response pattern represents ceiling performance is compelling. After all, even a monkey that did not observe the hiding events at all, or any subject that picked randomly, would be expected to produce precisely the same outcome. In other words, the subjects appeared to treat the container with 0% chance of reward as if it had just as much a chance of concealing a reward as the other two containers not chosen first (C and D).

Although the response patterns of the monkeys and young children fail to demonstrate an understanding of exclusive “or” relations, they are nonetheless consistent with an incorrect understanding of the relation between the containers within each pair as inclusive “or” (Gautam et al., 2021). That is, these populations may have some understanding that each pair contains at least one reward while failing to appreciate that each pair contains exactly one reward. Only the 5-year-olds in the Gautam et al. study provided compelling evidence that they appreciated the exclusive “or” relation by consistently passing both versions of the task. This interpretation is consistent with other recent data (Redshaw & Suddendorf, 2016; Suddendorf et al., 2020) and aligns with theory (Leahy & Carey, 2020; Redshaw & Suddendorf, 2020) suggesting that young children and nonhuman primates—unlike older children—may not represent mutually exclusive possibilities as such (see Laland & Seed, 2021, for a related discussion).

In their conclusion, Ferrigno et al. suggest that the monkeys’ responses reflect “combinatorial . . . or logical thought” (p. 299) and that verbal labels such as “or” are not necessary for understanding logical relations. We agree that approaching the problem with a nonverbal approximation of inclusive “or” would require combinatorial thought, as the pairs of containers (A and B; C and D) must be demarcated and kept separate in working memory. This approach to the problem would also imply that participants do not simply simulate a single possibility in situations of uncertainty and treat that simulation as reality (cf. Leahy & Carey, 2020). But whatever the cognitive processes underlying monkeys’ and young children’s responses, the patterns across conditions fail to provide evidence that these populations understand the exclusive “or” relations central to the task.