Graduation Year
2020
Document Type
Dissertation
Degree
Ph.D.
Degree Name
Doctor of Philosophy (Ph.D.)
Degree Granting Department
Psychology
Major Professor
Chad Dubé, Ph.D.
Committee Member
Kenneth Malmberg, Ph.D.
Committee Member
David Melcher, Ph.D.
Committee Member
Geoffrey Potts, Ph.D.
Committee Member
Elizabeth Schotter, Ph.D.
Keywords
ensemble representation, working memory
Abstract
Human observers benefit from encoding summary statistical information from multiple similar stimuli, also known as ensemble coding. Empirical studies on ensemble coding have adopted paradigms with vastly different task requirements, raising questions about ensemble coding’s underlying mechanisms in different tasks. Are ensemble coding mechanisms task-dependent? How to model ensemble coding in different tasks? The current paper aims to answer these questions by systematically reviewing the ensemble coding tasks and models and proposing a computational model framework of ensemble coding in different tasks. The task review categorizes the tasks into two main types: Summary tasks (explicitly require observers to summarize ensemble features of multiple stimuli) and Isolate tasks (no explicit requirements to summarize). The model review shows that a unified task-dependent framework could model both Summary and Isolate tasks. The task and model reviews inspire the proposal of a task-dependent fidelity-based integration model (FIM). Three studies tested the FIM framework. FIM outperformed alternative models in successfully accounting for typical data patterns across different paradigms (single item estimation, mean estimation, and member identification). The FIM analyses critically address individual and ensemble representations’ fidelity differences in different tasks. The three studies support ensemble coding’s task-dependent hypothesis and offer detailed FIM implementations for ensemble coding tasks. As a model framework, FIM can model cognitive psychology paradigms beyond the ensemble coding literature, suggesting broad future applications.
Scholar Commons Citation
Tong, Ke, "A Fidelity-based Integration Model for Explicit and Implicit Ensemble Coding" (2020). USF Tampa Graduate Theses and Dissertations.
https://digitalcommons.usf.edu/etd/8489
