Prior expectation modulates the interaction between sensory and prefrontal regions in the human brain

Abstract

How do expectations about the identity of a forthcoming visual stimulus influence the neural mechanisms of perceptual decision making in the human brain? Previous investigations into this issue have mostly involved changing the subjects' attentional focus or the behavioral relevance of certain targets but rarely manipulated subjects' prior expectation about the likely identity of the stimulus. Also, because perceptual decisions were often paired with specific motor responses, it has been difficult to dissociate neural activity that reflects perceptual decisions from motor preparatory activity. Here we designed a task in which we induced prior expectations about the direction of a moving-dot pattern and withheld the stimulus-response mapping until the subjects were prompted to respond. In line with current models of perceptual decision making, we found that subjects' performance was influenced by their expectation about upcoming motion direction. The integration of such information into the decision process was reflected by heightened activity in the dorsolateral prefrontal cortex. Activity in this area reflected the degree to which subjects adjusted their decisions based on the prior expectation cue. Furthermore, there was increased effective connectivity between sensory regions (motion-sensitive medial temporal area MT+) and dorsolateral prefrontal cortex when subjects had a prior expectation about the upcoming motion direction. Dynamic causal modeling suggested that stimulus expectation modulated both the feedforward and feedback connectivity between MT+ and prefrontal cortex. These results provide a mechanism of how prior expectations may affect perceptual decision making, namely by changing neural activity in, and sensory drive to, prefrontal areas.

Figure 1.

Task design. In each trial, subjects were asked to press keys to indicate whether a patch of dots was moving in an expanding or contracting direction. A cue in the form of a simple geometric shape indicated the likely direction of the motion. In half of the trials, the cue was predictive of the motion direction (75% valid, 25% invalid), whereas in the other half of the trials, a neutral cue was presented, which gave no information about the likely direction of the upcoming motion (i.e., a non-predictive cue). To perform optimally when the cue was predictive, subjects needed to combine the information from the cue and stimulus. The response mapping was only shown after the offset of the stimulus, so that subjects could not prepare for a motor response before the end of the motion presentation. Each trial lasted between 3.5 and 4.5 s (chosen from a uniform distribution), and subjects had up to 1 s to give an answer after the offset of the stimulus. ITI, Intertrial interval.

Figure 2.

Behavioral results. A, Accuracy and reaction times are plotted as a function of motion coherence (low, medium, high) and cue type (valid, neutral, invalid). Overall, higher motion coherence led to higher accuracy and lower reaction times. Similarly, invalid cues decreased performance, whereas valid cues improved it. Error bars show the SEM. B, The SDT measures d′ and c were computed independently for predictive (i.e., valid and invalid) and non-predictive (i.e., neutral) cues to gauge subjects' ability to do the task with/without the expectation provided by the cues. Surprisingly, d′ was lower for predictive than for neutral cues. Subjects were unbiased (c ∼0) when doing the task with neutral cues but shifted their criterion toward the expected percept when predictive cues were presented. Error bars show the SEM.

Figure 3.

Neural activation differences induced by expectation. A, Larger activation was found bilaterally in both IPS and bilateral DLPFC for trials in which subjects had an expectation than for trials in which subjects had no expectation (shown in yellow). Bilateral MT+ (shown in purple) was functionally localized using an independent localizer for each subject. B, Percentage signal change is plotted for each of the three cue types (invalid, valid, and neutral) for left DLPFC, IPS, and MT+ (left column) and its right hemisphere counterpart (right column). DLPFC showed larger activity for invalidly cued trials compared with validly cued trials (both p values <0.02). No such difference was found for IPS (both p values >0.8). There were no differences in MT+ for the differently cued trials (all p values >0.2). C, Time courses for each of the six regions of interest are plotted for each trial type. Error bars represent the SEM.

Figure 4.

Brain-behavior correlation of expectation-induced bias. We observed a significant positive across-subjects correlation between the behavioral criterion shift induced by the prediction cue and the neural activity increase for predictive trials in left DLPFC (Spearman's ρ = 0.50, p = 0.01) and left IPS (Spearman's ρ = 0.61, p = .002). Only trends were found in right DLPFC and right IPS. Conversely, this correlation was negative in left and right MT+, with the effect being significant in right MT+ (Spearman's ρ = −0.38, p = 0.04). The activity increase was computed from the contrast predictive (valid and invalid) > non-predictive (neutral) trials. Criterion shift is a measure of the degree to which subjects adjusted their decision bias (for details, see Materials and Methods). We used Spearman's rank correlation, a nonparametric test that is insensitive to extreme values in the variables. All significant correlations remain significant if Pearson's product-moment correlation was used.

Figure 5.

Effective connectivity between bilateral MT+ and DLPFC. We used a PPI analysis to look at the effective connectivity between left DLPFC and left (left panel) and right (right panel) MT+. Left DLPFC was chosen because it was sensitive to the presence and validity of the cue (Fig. 3) and correlated with the extent to which subjects shifted their criterion based on the predictive cues (Fig. 4). We tested whether the connectivity between MT+ and left DLPFC depended on the cue identity [computed from the contrast predictive (valid + invalid) > non-predictive (neutral)]. Connectivity between MT+ and left DLPFC was higher when subjects had a prior expectation about upcoming motion direction, as revealed by the difference in the slope of regression between fMRI activity. The regression lines reflect averages across subjects. The shaded regions correspond to one SEM.

Figure 6.

Dynamic causal modeling of connectivity between bilateral MT+ and left DLPFC. A, Three families of models tested whether expectation modulated only the feedforward, the feedback, or both connections between MT+ and DLPFC. In all models, there were recurrent intrinsic connections between bilateral MT+ and left DLPFC. The contribution of the stimulus varied between models and is not depicted in this figure (for details, see Materials and Methods). B, Bayesian model comparison was used to compute the exceedance probability for each of the three families of models. The exceedance probability was largest for the recurrent family of models, suggesting that expectation modulated both the feedforward and feedback connections between MT+ and DLPFC. C, All connections and their values are shown for the best-fitting model (the recurrent model in which the stimulus modulated only MT+). Across subjects, expectation significantly modulated the feedforward connection from left MT+ to left DLPFC (p = 0.003) and from right MT+ to left DLPFC (p = 0.006).