Cognitive modelling of visual configuration reproduction tasks

<p dir="ltr">The process of reproducing visual patterns from memory raises questions about processing and encoding of perceived visual information in memory and strategies to overcome visual memory limitations. The thesis investigates the cognitive processes of drawing visual configurations from memory. It investigates the role of chunking in visual pattern reproduction and how it is affected by stimulus viewing time, features of the stimulus, or existing knowledge of participants. Two empirical datasets of visual configuration reproduction tasks were reanalysed. And cognitive models were developed to evaluate the results of analyses.</p><p dir="ltr">The first disc location memorisation task concerns reproducing a number of randomly placed discs with various presenting durations. Analysis focusing on temporal and spatial measures to examine the presence of chunking in the disc production procedure of participants. The analyses indicate that in short trials (<i>≤ </i>200 <i>ms</i>), participants are likely to retain the stimulus as a mental image. However, in long exposure trials (<i>≥</i><i> </i>1 <i>s</i>), participants were also likely to build chunks based on salient geometric patterns which contain a larger number of discs. Two models about the disc location memorisation task were developed within the ACT-R cognitive architecture to better understand the cognition by comparing the models’ performance to that of human data. Moreover, the testing of models which only use the standard ACT-R showed that the task cannot be modelled without adding additional capabilities. Thus, a snapshot model with a fovea-peripheral based activation mechanism was developed, which simulates how more attention and processing resources are given to the centre of the visual field for short stimulus exposure trials. For long exposure trials, a chunking model was developed based on the snapshot model by adding chunking processes which can encode geometric patterns.</p><p dir="ltr">The second complex diagram drawing task is reproducing a single complex abstract diagram in four types of tasks including tracing, copying, immediate recall and delayed recall across 10 sessions. Cluster size and response strategy analyses found that participants used chunking to aid the memorising processes, and most of them changed their strategy from an initial generic nearest neighbour first strategy to a chunk dependent strategy with increasing familiarity of the stimulus. An ACT-R model was developed following the procedure of drawing the complex diagram. The model includes a pattern identifying mechanism based on the visibilities of lines, a cognitive processes competition between visual information acquisition and memory retrieval processes, and a symmetry based visual feature encoding process. The model simulated the effects of strategy changes and chunking hierarchy formation on human participants’ performance in this drawing task.</p><p dir="ltr">The overall studies found that chunking plays a substantial role in visual configuration re-production tasks. The implications of the cognitive models are discussed with regard to existing accounts of diagrammatic reasoning and problem solving.</p>