Humans are sub-optimal observers: An investigation of multimodal perception

Petrie, Kirstie (2015). Humans are sub-optimal observers: An investigation of multimodal perception Honours Thesis, School of Psychology, The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
hons2015_petrie_kirstie.pdf Thesis full text application/pdf 1.12MB 0
Author Petrie, Kirstie
Thesis Title Humans are sub-optimal observers: An investigation of multimodal perception
School, Centre or Institute School of Psychology
Institution The University of Queensland
Publication date 2015-10-07
Thesis type Honours Thesis
Supervisor Derek Arnold
Total pages 86
Language eng
Subjects 1701 Psychology
Formatted abstract
The precision of sensory encoding can be enhanced when information is encoded in two, rather than just one, sensory modality (Ernst & Bülthoff, 2004). This is called a bimodal encoding advantage. In contemporary literature, the dominant account of bimodal encoding advantages is Maximum Likelihood Estimation (MLE). According to this, humans combine multisensory information in a statistically optimal fashion. This thesis investigated a novel, alternate account – that bimodal encoding advantages are governed by accurate feelings of confidence in each unimodal stimulus encoding. It is well established that confidence in perceptual decisions correlates well with objective measures of task performance, even when no feedback is provided. This suggests humans have instantaneous insight into how well they have encoded sensory information. Accordingly, it was hypothesised that feelings of confidence could govern bimodal encoding advantages, by allowing participants to rely more on better-encoded information eliciting a greater level of confidence on any given trial. This was assessed by examining the precision of spatial judgments on audio, visual, and audiovisual trials. Across three experiments, overall bimodal performance was always superior. However, consistent with the research hypothesis, this advantage was lost when comparing data matched in terms of elicited confidence. In addition, in the final experiment inconsistent bimodal trials were included. These involved audio and visual signals, matched in terms of elicited sensitivity, signalling opposite spatial offsets. Statistically optimal MLE summation predicts chance performance on such trials – instead a robust bias to base judgments on visual information was evident. In sum, project results discredit the dominant account of bimodal encoding advantages, and are instead consistent with a novel hypothesis presented in this thesis.
Keyword Multimodal

Citation counts: Google Scholar Search Google Scholar
Created: Thu, 17 Mar 2016, 11:45:27 EST by Lisa Perry on behalf of School of Psychology