Simultaneous learning of directional and non-directional stimulus relations in baboons (Papio papio)

Abstract

SummaryWhile humans exposed to a sequential stimulus pairing A-B are commonly assumed to form a bidirectional mental relation between A and B, evidence that non-human animals can do so is limited. Careful examination of the animal literature suggests possible improvements in the test procedures used to probe such effects, notably measuring transfer effects on the learning of B-A pairings, rather than direct recall of A upon cuing with B. We developed such an experimental design and tested 20 Guinea baboons (Papio papio). Two pairings of visual shapes were trained (A1-B1, A2-B2) and testing was conducted in a reversed order, either with conserved pairings (B1-A1, B2-A2) or broken ones (B1-A2, B2-A1). We found baboons’ immediate test performance to be above chance level for conserved pairings and below chance level for broken ones. Moreover, baboons needed less trials to learn conserved pairings compared to broken ones. These effects were apparent for both pairings on average, and separately for the best learned pairing. Baboons’ responding on B-A trials was thus influenced by their previous A-B training. Performance level at the onset of testing, however, suggests that baboons did not respond in full accordance with the hypothesis of bidirectionality. To account for these data, we suggest that two competing types of relations were concomitantly encoded: a directional relation between A and B, which retains the sequential order experienced, and a non-directional relation, which retains only the co-occurrence of events, not their temporal order.

Brain areas for reversible symbolic reference, a potential singularity of the human brain

“…In line with this, a retroactive gradient has been shown in memory storage in humans, where memory is strongest for stimuli that were presented close to the reward but preceding it (Braun et al, 2018). This memory trace may explain the slight facilitation observed in baboons when they learn reversed congruent pairs relative to reversed incongruent pairs (Chartier et al, 2022). Although neuronal replay in both forward and reverse directions exists in non-human animals, it might be that this mechanism has selectively expanded to symbol-related areas of the human brain -a clear hypothesis for future work.…”

“…Interpretation must remain cautious, as there are also some occasional behavioral reports of spontaneous reversal of learned associations, for instance in one well-trained California sea lion and a Beluga whale (Kastak et al, 2001; Murayama et al, 2017; Schusterman and Kastak, 1998) and possibly in 1 out of 20 baboons in Medam et al (2016). These studies may indicate that, with sufficient training, symbolic representation might eventually emerge in some animals, as also suggested by the small reversal trend in a recent behavior study in baboons (Chartier and Fagot, 2022). However, they may also merely show that animals may begin to spontaneously reverse new associations once they have received extensive training with bidirectional ones (Kojima, 1984).…”

“…A small number of behavioral studies, spread over four decades, report that non-human animals such as bees and pigeons, but also macaques, baboons and chimpanzees, struggle to reverse the associations that they learned in one direction (Imai et al, 2021; Kojima, 1984; Lipkens et al, 1988; Medam et al, 2016; Sidman et al, 1982; Howard et al, 2019; see Chartier and Fagot, 2022, for a review and discussion). In a recent experiment, Chartier and Fagot (2022) explored this question in 20 free-behaving baboons.…”

Evidence for compositionality in baboons (Papio papio) through the test case of negation

“…This suggests that they learned the meaning of atomic cues and learned how meaning is systematically modified by the negation morpheme rather than holistically learning each complex cue as a single meaningful unit. In the future, one could try to show that baboons immediately and spontaneously understand new complex cues in a compositional manner, although one-shot learning of this type is hard to measure, as it may be masked by noisy experimental responses and independent biases (such as the strong iconicity bias present here) (see also 45 for additional evidence that stronger effects can be observed during learning than during generalization in matching-to-sample tasks in baboons). What our results do suggest is that the ability to combine abstract representations is not unique to humans.…”

Do Backward Associations Have Anything to Say About Language?

“…There are, however, three reasons to believe that this failure is not suggestive of a general incapacity of nonhuman animals to spontaneously encode bidirectional relations between items. First, nonhuman animals show indirect evidence for backward associations in transfer tasks in which A‐B training facilitates subsequent B‐A learning (Chartier & Fagot, 2022b; Richards, 1988; Soares Filho, Silva, Velasco, Barros, & Tomanari, 2016; Velasco, Huziwara, Machado, & Tomanari, 2010), suggesting that at least some B‐A association is formed while learning A‐B. Second, such a failure could simply be specific to the Matching‐to‐Sample task and indeed, evidence for backward associations was found in rat conditioning experiments (Arcediano, Escobar, & Miller, 2003; Matzel, Held, & Miller, 1988).…”