A linking hypothesis for eyetracking and mousetracking in the visual world paradigm

For a linking hypothesis in the visual world paradigm to clearly accommodate existing findings and make unambiguous predictions, it needs to be computationally implemented in a fashion that transparently draws the causal connection between the activations of internal representations and the measured output of saccades and reaching movements. Quantitatively implemented linking hypotheses provide an opportunity to not only demonstrate an existence proof of that causal connection but also to test the fidelity of the measuring methods themselves. When a system of interest is measured one way (e.g., ballistic dichotomous outputs) or another way (e.g., smooth graded outputs), the apparent results can differ substantially. What is needed is one linking hypothesis that can produce both types of outputs. The localist attractor network simulation of spoken word recognition demonstrated here recreates eye and mouse movements that capture key findings in the visual world paradigm, and especially relies on one particularly powerful theoretical construct: feedback from the action-perception cycle. Visual feedback from the eye position enhancing the cognitive prominence of the fixated object allows the simulation to fit a wider range of findings, and points to predictions for new experiments. When that feedback is absent, the linking hypothesis simulation no longer fits human data as well. Future experiments, and improvements of this network simulation, are discussed.