Language, Visual Working Memory, and Dot Subtraction: What Counts?

To investigate cognitive factors affecting subtraction of visual objects, we adapted the dot subtraction task developed by Pica, Lemer, Izard, and Dehaene (2004), who used it to investigate calculation by the Munduruku, an indigene group in Brazil that has a limited number word vocabulary. In the dot subtraction task, briefly displayed arrays of moving dots are used to represent the quantities for subtraction. We tested 40 Canadian university students' dot enumeration, Arabic digit subtraction, visual working memory, and performance on the dot subtraction task with dot display durations of 2, 1.5, 1, and .5 s. In the 2 s condition, error rates were uniformly low, whereas in the .5 s condition, error rates increased sharply as the minuend increased from 4 to 8, as was observed with the Munduruku. Individual differences in dot subtraction accuracy were predicted by dot enumeration skill with longer dot display durations but were predicted by visual working memory efficiency with shorter durations. Pica et al. (2004) attributed the Munduruku participants' very poor subtraction to the absence of counting words, but our results show that a shift to reliance on visual working memory is a nonlinguistic factor that comes into play in the dot subtraction task when time to encode the dot arrays is limited.