Fig. 5: Algorithm selection adapted to complexity.

a, Scatter plot of the proportion of solutions consistent with high-complexity algorithms, for matched instances in monkeys G and B. Horizontal error bars are s.e.m. across instances within bins, for monkey B. Vertical error bars are s.e.m. across instances within bins, for monkey G (Methods). Solid and dashed lines are the best fitting and unity lines, respectively. The statistic in the main text is calculated using all n = 462 instances. b, Box plots showing the proportion of low-complexity solutions instances when the maximum value item was 0.7 ml or less than 0.7 ml. Orange and brown boxes show data from monkeys G and B, respectively. n = 462 instances. c, Box plots showing the proportion of solutions that were consistent with a k = 2 algorithm under instances with complexity k < 2 or k = 2. Orange and brown bars show data from monkeys G and B, respectively. n = 462 instances. d, Bar plots showing the standardized regression coefficients (SRC) of independent variables reported in the text. Error bars are standard errors derived from the mixed-effects logistic regression model. n = 11511 and n = 8507 trials for monkeys G and B, respectively. Box plots showing the median (line), quartiles (boxes), range (whiskers) and outliers (+).