Action-Specific Disruption of Perceptual Confidence

Theoretical models of perception assume that confidence is related to the quality or strength of sensory processing. Counter to this intuitive view, we showed in the present research that the motor system also contributes to judgments of perceptual confidence. In two experiments, we used transcranial magnetic stimulation (TMS) to manipulate response-specific representations in the premotor cortex, selectively disrupting postresponse confidence in visual discrimination judgments. Specifically, stimulation of the motor representation associated with the unchosen response reduced confidence in correct responses, thereby reducing metacognitive capacity without changing visual discrimination performance. Effects of TMS on confidence were observed when stimulation was applied both before and after the response occurred, which suggests that confidence depends on late-stage metacognitive processes. These results place constraints on models of perceptual confidence and metacognition by revealing that action-specific information in the premotor cortex contributes to perceptual confidence.


Meta-d' estimation
Meta-d' provides a response-bias free measure of how well confidence ratings track task accuracy (Maniscalco & Lau, 2012). In order to estimate meta-d' for each response separately (congruent, incongruent), we employed a response-specific meta-d' model (Maniscalco & Lau, 2014). Response-specific meta-d' is estimated by fitting the distribution of confidence ratings for each subject conditional on the discrimination response being correct or incorrect, separately for congruent and incongruent responses.
A comprehensive overview of meta-d' and its response-specific variant is provided in Maniscalco & Lau (2014).
The fitting of meta-d' rests on calculating the likelihood of the confidence rating data given a particular "type 2" signal detection theoretic model. While conventional "type 1" SDT considers how well an observer can discriminate objective states of the world, such as stimulus present or absent, type 2 SDT characterises an observer's ability to discriminate her own correct or incorrect responses. Consider the simple case where the observer rates confidence as either "high" or "low." We can then distinguish 4 possible outcomes in the type 2 task: high confidence correct trials, low confidence correct trials, low confidence incorrect trials, and high confidence incorrect trials. By direct analogy with the type 1 analysis, we may refer to these outcomes as type 2 hits, type 2 misses, type 2 correct rejections, and type 2 false alarms, respectively.
Type 2 hit rate (HR) and type 2 false alarm rate (FAR) summarize an observer's type 2 performance and may be calculated as Maniscalco & Lau, 2012). Describing the observed type 2 ROC in terms of these type 1 SDT parameters underpins the meta-d' model. By convention, the prefix "meta-" is added to each type 1 SDT parameter in order to indicate that the parameter is being used to fit type 2 ROC curves. Thus, the type 1 SDT parameters d' c, and c 2 , when used to characterize type 2 ROC curves, are named meta-d' meta-c, and meta-c 2 .
The equations below describe the calculation of type 2 probabilities from the type 1 SDT model for both S1 and S2 responses, e.g. congruent and incongruent in our experiment.
For notational convenience, below we express these probabilities in terms of the standard SDT model parameters, omitting the "meta" prefix.
For a discrete confidence scale ranging from 1 to H, H -1 type 2 criteria are required to rate confidence for each response type. Define type 2 confidence criteria for S1 and S2 responses as: Next we consider the procedure for finding the parameters of the type 1 SDT model that maximize the likelihood of the response-specific type 2 data for a particular response, "S1" (e.g. congruent in our experiments). The same procedure can be applied to estimate meta-d' for "S2" (e.g. incongruent) responses. The likelihood of the type 2 confidence data can be characterized using the multinomial model as Maximizing likelihood is equivalent to maximizing log-likelihood, and in practice it is typically more convenient to work with log-likelihoods. The log-likelihood for type 2 data is given by θ "S1" is the set of parameters for the response-specific meta-SDT model: is a count of the number of times in the data a confidence rating of y was provided when the stimulus was s and response was "S1". y and s are indices ranging over all possible confidence ratings and stimulus classes, respectively.