Different Classes of Words Are Learned in Different Ways

What determines vocabulary growth patterns? The research presented here examines the growth pattern of words listed in the McArthur-Bates Communicative Development Inventory using a computational model. Our model characterizes vocabulary growth curves based on the sampling of learning relevant events and a threshold (the number of such events needed) for acquisition of the word. Using this general class of models, fits of vocabulary growth curves suggests a transition from one in which acquisition is primarily limited by the threshold for acquisition to one in which acquisition is primarily limited by sampling speed. Further analyses suggest that these parameters of the learning model link to meaningful psychological factors: specifically the acquisition of threshold limited (and earlier learned) words are correlated with frequency whereas sampling-speed-limited words are correlated with imageability of the word in the input.