Figure 4: Emergence of theta pattern similarity after learning.

(a) The learning curves of novel PA learning in monkeys K and M. Initial- and late-learning periods are shown by grey and purple bands, respectively. The horizontal dotted lines indicate a chance level of 33.3%. Error bars: s.d. (b) Comparisons of the pair-decoding accuracy between initial- and late-learning periods. The decoding accuracy in the late-learning period is significantly above chance in A36, TE and ALL (all channels) in monkeys K and M; in PH in monkey M (n=2 monkeys, permutation test, ***P<0.001, **P<0.01, *P<0.05, corrected for multiple comparison). Horizontal grey lines indicate the chance level (33.3%). Short green lines indicate 5% significance levels in the permutation tests. Also, decoding accuracy is significantly higher in the late- than in the initial-learning period in A36 (monkey K) or in TE (monkey M) (n=2 monkeys, permutation test, *: P<0.05). (c) Development of the theta pattern similarity. d’ for each pair-decoding are arranged in confusion matrices (see Methods). The elements along the left-to-right diagonal axis correspond to correct predictions. The confusion matrices changed into diagonal patterns in the late period of learning in the two monkeys. (d) Comparisons of average d’ between the initial- and late-learning periods. Lines indicate each of six pairs in monkeys K and M. Diagonal (within pair) and non-diagonal (across pair) elements of the confusion matrices in (c) were separately averaged, and compared between the initial- and late-learning periods. The averaged d’ of within-pair decoding is significantly greater in the late- than in the initial-learning period (n=2 monkeys, two-tailed paired t-test, P<0.05).