Fig. 5
Adaptation pathway dependence on violation magnitude. (a) Fraction of networks (\(n=100\)) adapting via relational module (\(\theta\) sign change) for rule reversal \(\alpha \rightarrow -\alpha\), fitted with logistic function (green) to estimate inflection point \(\bar{\alpha }\). Arrows mark the \(\alpha\) values from Fig. 4. The logistic function fit has two parameters, the inflection point, \(\bar{\alpha }\sim 0.34\), and the slope parameter, \(\sim 14.10\) (see Methods). Error bars: 95% CI. (b) Adaptation pathways for general rule changes \(\alpha _1\rightarrow -\alpha _2\). Color indicates fraction (out of \(n=50\)) adapting via relational module (\(\theta\)). Squares: representational module (Z) dominant (ratio \(<50\%\)); triangles: relational module dominant (ratio \(\ge 50\%\)).
