Implicit learning of perceptual comparisons with extended training.
Timothy Y. Chen · UC Research Repository (University of Canterbury) · 2019
How do we compare stimuli that vary in magnitude? An overview of relevant literature shows that perceptual comparisons have been a longstanding topic of interest in various subfields of psychology, and there currently is a mixed view in regards to how cognitive systems encode and compare stimulus magnitudes. In particular, a conjecture put forth by Torgerson (1961) states observers use a single quantitative relation to compare magnitudes, and that relation is either a difference or ratio. Torgerson’s (1961) conjecture has stimulated considerable research using stimuli of various modalities; these studies have typically used experimental designs such as magnitude estimation in which responses are symbolic (i.e., numbers) and observers can use their mathematical knowledge. Recently Grace et al. (2018) developed a novel, nonsymbolic task – an ‘artificial algebra’ – to study perceptual comparison, and found results inconsistent with Torgerson (1961). Their results showed significant control by both difference and ratio relations over responding, which contradicts Torgerson’s (1961) prediction that observers use a single operation. The research in this thesis used Grace et al.’s (2018) artificial algebra task with the primary aims of replicating and extending their results. We conducted two experiments, with brightness and line length as the stimulus modalities. For each experiment, eight observers were randomly assigned to receive training with difference or ratio relations. They completed four sessions on separate days, so we could assess the effect of extended training. On each trial, observers saw a pair of stimuli and made a response on an analogue bar based on comparing the stimulus magnitudes. Feedback was provided based on the difference or ratio of the nominal stimulus values. Results were similar to those of Grace et al. (2018) and showed that observers responded accurately in both experiments. Although there was relatively little improvement in measures of accuracy over sessions, multiple regressions showed that control by the trained relation increased while control by the untrained relation decreased. This confirms that feedback provided over multiple sessions was effective in terms of training observers to respond based on either difference or ratio relations. Results also showed that for the majority of observers and sessions, there was significant control by the untrained relation. These results are consistent with those of Grace et al. and show that responding in their task is based on two operations rather than one, contrary to Torgerson’s conjecture.