a, Average choice behavior when choosing between the left and right options plotted as a function of the value of the unavailable option (low value, blue; high value, red) value for each animal separately (data are averaged across sessions). Curves plot logistic functions fit to the choice data. b, A partial regression plot shows the uncontaminated effect of the unavailable option’s value on accuracy (y axis, accuracy residuals; x axis, residuals of the unavailable option’s value) for each animal during the fMRI sessions after partialling out the difference in value between the better and worse options in any given trial (which should rationally be the main determinant of decision-making). There was a linear relationship between the residual accuracy and the residual value of the unavailable option in three out of four animals (mixed-effect linear models, animal 1: t1160 = 2.28, P = 0.02; animal 2: t1358 = 3.85, P = 0.0001; animal 3: t1329 = 0.30, P = 0.76; animal 4, t950 = 5.85, P = 6.8 × 10–9). Each bin contains 25% of averaged data across sessions (±s.e.m.). c,d, Extra data from four animals (c) show the same effect as in the main fMRI experiment (d). e, The impact of the unavailable option on the current decision—an impact that seems to be mediated by vmPFC—was not significantly altered by the application of TUS to ACC; there was no difference between the ACC-TUS sessions (n = 18) and the baseline sessions.