It is generally understood that individuals have a limited cognitive capacity, hence an appropriate metric of capacity is necessary for the purpose of measuring capacity allocated to represent different forms of information and the final decision itself. Previous research has shown that the sample size model serves as a fitting metric of capacity that characterises capacity precisely and quantifiably. However recent research has reported ambiguous findings - performance in a visual task was better than what the sample size model predicted. Results on the visual task resembled something closer to a capacity-unlimited process rather than a capacity-limited sample-size process. Hence, the present study attempts to make some changes to the visual task and investigate whether the performance is consistent with the sample size model.
Research Questions / Hypotheses
This study attempts to investigate whether the adapted visual stimuli will demonstrate performance consistent with the predictions of the sample size model.
Fifty individuals participated in the experiment, however only 13 participants completed all four sessions and presented complete datasets. Thirty-seven participants were excluded from the analysis due to incompleteness of all four sessions.
This experiment is a four session experiment. Each session takes no longer than one hour. The visual task had two 300ms pedestal intervals with the intensity of colour in the stimuli centre either upwards or downwards. Three different set sizes (1, 2, 4) of offset Gabor patches were presented simultaneously or sequentially in a two-interval, two-alternative forced choice task.
The proportion of correct responses across set size was modelled using two different computational models, namely the sample size model and the power law model. Two participants were excluded from the analysis, so only 11 complete datasets were included in the analysis of model fit. The analysis using AIC found that the sample size model was the preferred model at the group level. In terms of participant-by-participant model comparison, the single exponent power law model best fit the data for 6 out of 11 participants with the exponent fairly close to sample size.
The findings of our study show that performance across set size was consistent with the predictions of the sample size model. The results of this study will be communicated in the form of an honours thesis and fourth-year conference presentation.