Abstract Background The predictive coding framework of perception postulates that we automatically infer what is around us by combining our sensory input with our prior beliefs. Mathematical models based in… Click to show full abstract
Abstract Background The predictive coding framework of perception postulates that we automatically infer what is around us by combining our sensory input with our prior beliefs. Mathematical models based in Bayesian statistics can describe this process, elucidating both typical as well as atypical brain processes, such as emergence of hallucinations. A previous study using a Pavlovian conditioning task showed that hallucinations are a result of over-weighting of prior beliefs over incoming sensory evidence. Participants with Auditory Verbal Hallucinations (AVH) were more susceptible to auditory conditioned hallucinations (ACH) than individuals without AVH, regardless of a diagnosis of psychosis. This suggests a common underlying mechanism for the emergence of hallucinations irrespective of functional status. To further investigate these mechanisms, we developed a visual conditioned hallucinations (VCH) task, modeled after the original ACH task. Together, these tasks can elucidate the specificity of conditioned hallucinations with regards to sensory modality. In addition to the deployment of the VCH, we also tested the feasibility of the first known online deployment of QUEST, an adaptive psychometric thresholding method, in a targeted population. Methods Task Methodology. Individual trials of the task consist of simultaneous presentation of a faintly-presented (low-contrast) visual stimulus (Gaussian stripes embedded in visual white noise) and a salient auditory stimulus (loud tone). An association between visual and auditory stimuli is established, followed by testing of the strength of this association by presenting the auditory stimulus alone. Trials where the visual stimulus is not presented, but nevertheless reported to be perceived are considered as CH trials. Data Collection and Analysis. Subjects were recruited on two platforms, the Amazon-powered Mechanical Turk (MTurk) and the COPE Project an initiative targeted at individuals with unusual experiences. Analyses were conducted by pooling across the four groups and grouping based on presence of visual or auditory hallucinations, creating four groups: AH+/VH+, AH+/VH-, AH-/VH+, AH-/VH-. Between-group differences were analyzed using two-sample t-tests. Results The responses of participants across the board, both in the control sample and in those with perceptual experiences, validates the efficacy of using the QUEST thresholding paradigm to predict response to threshold levels. Participants who were assigned VH+ groups showed a significantly higher rate of reporting CHs, compared to the control group with no hallucinatory experiences (p < 0.01). Similarly, participants assigned to AH+ groups also showed a significantly higher rate of reporting CHs (p < 0.01). Differential parameter values for a computational model assessing perceptual inference were computational signatures of CH were also examined between groups, as were neural signatures encoding visual versus auditory-mediated conditioned hallucinations. Discussion In this study, we show the feasibility of psychometric experiments in a targeted online sample of individuals with unusual experiences. We also demonstrate that the differential parameter values of the computational modeling of visually conditioned hallucinations differentiate between individuals with and without hallucinations, leading to better understanding of the neural signatures of hallucinations and potential therapeutic targets in clinical populations.
               
Click one of the above tabs to view related content.