Figure 1.

Avoiding confounds between value and difficulty. (a) Foraging frequency (left) and difficulty, as indexed by log(RT) (right), as a function of RVF in an experiment claiming value signals and difficulty have been confounded in ACC [28]. The black line indicates behavioural data and the red one the corresponding model fit. The grey bars are the sample sizes and the dotted lines are the two indifference points (red = subjective or empirical and black = objective indifference point, i.e. where the value of searching and engaging are objectively equal; ‘RV_forage = 0’). The participants tended not to forage and to be inaccurate. For example, foraging frequency barely reaches 80% even on the right hand side of the left panel and the participants’ empirical indifference points were far from the objective indifference point. (b) After adequate task training and instruction in a version of the task employing a balanced and evenly sampled range of search and engage values in which decisions are non-trivial and require value comparison [19] several features of the experiment, participant performance, and data are notable: (i) participants balance all the factors that should influence decision-making in an approximately rational manner and the point of empirical indifference is close to the objective indifference point meaning that ii) data are sampled from both left and right of decision space ensuring foraging values and difficulty decorrelation; iii) Log(RT) decreases either side of the empirical indifference point in an approximately similar manner confirming foraging values and log(RT) are not correlated. Foraging decisions plotted (similar format to a) as a function of RVF (based on all three variables that should influence behaviour: search value, engage value, and an additional factor related to the cost of foraging). Adapted from [19, 28].