5) ( Figures 2C and 2D). In addition, buy Erastin sex, age, or education covariates did not explain a significant proportion of variance in any of the reversal error scores (R2 < 0.01, F(3,678) < 1.8; p > 0.1). In summary, the present data set reveals a double dissociation between effects of the SERT and DAT1 genotypes on reversal learning, with SERT altering global lose-shifting and DAT1 altering postreversal perseveration. In a final ANOVA, we ascertained that the relative difference in lose-shift and perseveration Z scores was predicted by the difference in SERT and DAT1 genotype (R2 = 0.16, F(5,676) < 25.5; p = 0.009). This significant interaction confirms
the double dissociation between the two effects, with SERT affecting lose-shifting but not perseveration, and DAT1 affecting perseveration but not lose-shifting. We http://www.selleck.co.jp/products/tenofovir-alafenamide-gs-7340.html next used computational models to investigate the mechanisms that might underlie the DAT1 genotype results. Although DAT1 shows robust effects in our data set, the measure of perseveration to which it is related is relatively opaque, in contrast to the more direct measure of trial-by-trial switching with which SERT was associated. This opaqueness results from the fact that (perseveration) error scores require some form of “topdown” definition or knowledge by the experimenter, e.g., when the reversal, unbeknownst to the subject, has occurred. This has hampered comparison
of previous studies of reversal learning studies, which have reported a veritable zoo of reversal error measures, such as errors to criterion, total reversal errors, maintenance errors, perseverative errors, learning errors, and chance errors. Models of reinforcement learning can provide a more principled approach to assessing behavior, because they are independent of such external definitions that the subject is unaware of (learning criterion, point of reversal). Instead, like for win-stay/lose-shift measures, they take into account only past choices and observed outcomes. We aimed
to understand the process or mechanism underlying the effect of DAT1 on perseveration using a reinforcement learning model to examine how perseveration Etomidate can arise from a learning process integrating reward over a longer timescale. For simplicity, we do not consider the more transparent SERT effects on lose-shift behavior here, although we have verified in simulations not reported here that our model captures them when it is augmented with an additional parameter that directly controls switching after losses, without affecting long-term value integration. In the context of reinforcement learning models, two features of the DAT1 effects are puzzling. First, the effect is selective to the reversal phase, and second, the relationship between performance in the acquisition and reversal phases reverses sign depending on genotype.