Rey, A. 1 , Tosatto, L. 1 , Németh, D. 2, 3 & Fagot, J. 1
1 CNRS & Aix-Marseille University, France
2 CRNL, Université de Lyon, France
3 Eötvös Loránd University, Budapest, Hungary
Associative learning is certainly the fundamental mechanism supporting regularity extraction. Using a serial response time paradigm, previous studies have shown that non-human primates can chunk up to three elements of a sequence in a row, even if the transitional probabilities between these elements were not always deterministic. In the present experiment, twenty Guinea baboons (Papio papio) were repeatedly exposed to the same deterministic sequence composed of 9 elements during 600 successive trials. The evolution of response times for each element of the sequence revealed chunking patterns that differed across monkeys. Surprisingly, we also found that these patterns evolved over time without any concatenation of chunks but with a reconfiguration of the chunks. This unexpected result is not predicted by any existing model of statistical learning.