| 研究実績の概要 |
To test the empirical (i.e., behavioural and neurocognitive) implications of our universal constructions explanation for systematicity, we conducted two experiments that measured changes in behavioural and neurophysiological variables with product arity, a kind of universal construction. In the first experiment, subjects performed a visual search task. Target and non-targets were rectangular bars in one of four colours, orientations, and spatial frequencies. Search was conducted in three product arity conditions, where the target was uniquely identifiable along one, two, or three visual feature dimensions (unary, binary, and ternary arity conditions, respectively). The first experiment replicated our earlier result in showing increased EEG synchrony with product arity in the lower gamma-band region: i.e., greater frontal-parietal electrode synchrony with higher arity. In the second experiment, subjects performed a match-to-sample task: i.e., determine whether given object (rectangular bar) is the same as the previously displayed target object. Matching was done in three product arity conditions. We did not observe a systematic increase in behavioral EEG responses with product arity.
In addition to empirical work, we pursued methodological and theoretical developments of our universal construction account of systematicity. On the methodological side, we adapted Empirical Bayes for the analysis of EEG synchrony, used to assess universal constructions. On the theoretical side, we provided a category theory explanation for the (second-order) systematicity of learning.
|
| 今後の研究の推進方策 |
In a follow-up study, we tested the hypothesis that systematicity (or lack thereof) depends on the subject's perceived cost-benefit assessment of the task at hand, by explicitly manipulating the cost-benefit perception variable. In a pilot study, 6 subjects were required to learn a set of cue-target associations: i.e., learning the target associated with each cue. Learning cue-target associates was done in a product and a non-product condition for small to large set sizes. In the product condition, awareness and utilization of the product rule afforded greater target predictability (benefit), but at the cost of slower response via the product construction and having to find (learn) the target construction. Conversely, circumventing a product construction by utilizing direct cue-to-target associations affords faster response times, but at the cost of inability to predict testing trials and having to learn more direct cue-target associations. When the number of associations (set size) is low, we expect subjects to utilize the (apparently) low cost non-product strategy, but when number of associations is high, we expect them to employ the (apparently) high cost product strategy. Consistent with these expectations, response errors decreased with set size in the product condition compared to the non-product condition, and response times correspondingly increased. The plan is to run a new experiment, based on this pilot study to further test our cost-benefit hypothesis and associated in neurophysiological (EEG) measures.
|
