Judging rapid resolutions: Insight into the information requirements for accurate categorisation using exposure duration and resolution
Graham, Madeleine (2017) Judging rapid resolutions: Insight into the information requirements for accurate categorisation using exposure duration and resolution. Honours thesis, Murdoch University.
Abstract
Humans are capable of making incredibly rapid visual categorisations. The speed of these non-analytic categorisations is the result of processing broad, global properties of an image, rather than by using specific features. It is unknown just how much (or little) information is actually required for rapid categorisations. We systematically reduced the amount of information in hundreds of photographs of birds, faces, and paintings, and asked people to decide whether two simultaneously presented images were from the same category, or two different categories. We flashed these images on screen for either 50, 250, 500, or 2500 milliseconds. This novel experimental design is the first to combine these two traditionally separate techniques of investigating human categorisation decisions. We found that people's accuracy in categorising images depended both on image resolution and exposure duration, but that exposure duration only affected accuracy at higher resolutions. Interestingly, categorisation accuracy peaked at 500ms, with no significant change beyond that exposure duration, suggesting that there may be a hard limit for accurate image categorisation. We conclude that the novel combination of resolution and exposure duration is a promising way to investigate human categorisation, and our findings provide a promising avenue for future research into the minimum exposure duration requirements of categorisation.
Keywords: categorisation, non-analytic, analytic, resolution, exposure duration
Item Type: | Thesis (Honours) |
---|---|
Murdoch Affiliation(s): | School of Psychology and Exercise Science |
Supervisor(s): | Thompson, Matthew |
URI: | http://researchrepository.murdoch.edu.au/id/eprint/40890 |
![]() |
Item Control Page |
Downloads
Downloads per month over past year