Semantic Dynamics and Sense Uncertainty Efects – Predictive Power and Limitations of the SSD Model

Abstract

The Semantic Settling Dynamics (SSD) model was proposed to account for the large body of inconsistent fndings on processing of lexical ambiguity (Armstrong & Plaut, 2016). However, the evaluation of this model is still at its roots. We focused on polysemes (words with multiple related senses) and tested the prediction that an increase in time spent in processing would be followed by a decrease in polysemy advantage in a lexical decision task (LDT). We described polysemy by using a continuous measure of sense uncertainty (entropy) and focused on the slope of entropy effect in relation to the time spent in processing. In the standard visual LDT (baseline) we replicated the finding that polysemous words with higher entropy (high number of balanced senses) were processed faster (Filipović Đurđević, 2007). An attempt to prolong the processing by reducing the contrast between the stimuli and the background did not affect the time spent in processing, but neither did it affect the slope of the entropy effect. Finally, presenting the same stimuli in auditory LDT prolonged the processing and cancelled out the entropy effect. Our results speak in favor of the SSD model, but also point to certain limitations which we will discuss in detail.