Abstract
Numerical cognition is widespread among animal species (Brannon & Park, 2015) and present from birth in humans (Izard, Sann, Spelke, & Streri, 2009). Laboratory studies show that longer stimulus duration improves numerical discrimination accuracy (Inglis & Gilmore, 2013; Wood & Spelke, 2005). Inglis & Gilmore (2013) suggested that longer durations allow subjects to resample the stimulus image multiple times, resulting in a more accurate final estimate. The current study tested this “multiple sampling” hypothesis alongside two competing hypotheses (“sequential enumeration”, “longer processing time”) to determine why stimulus duration relates to accuracy. Adult subjects (N=14 and 15, in E1 and E2, respectively) completed a fully within-comparison, 2AFC task in which they judged which of two arrays had more dots; accuracy was the dependent measure. Display time was kept brief to prevent counting. Experiment 1 found higher accuracy in the 500ms stimulus condition than the 100ms condition (M100ms=0.57, M500ms=0.65, F(1,13)=30.63, p<.001). Although post hoc Helmert comparison revealed significant difference between extra small group and other three larger groups (Mextra_small=0.64, Mlater=0.60, F(1,13)=49.21, p<.001), no difference was found between three larger set size groups (all Fs<3.00, all ps>0.008), suggesting that items were not enumerated sequentially. Data from Experiment 2 extended these findings. When stimulus duration was held constant at 100ms, adding a 400ms delay between stimulus offset and mask onset improved accuracy (M100+0msDelay=0.61, M100+400msDelay=0.64, F(1,13)=19.84, p=.001). There was no difference in accuracy between the 500ms stimulus duration condition (E1) and the 100ms duration + 400ms mask delay condition (E2) (F(1,28)=0.16, p=.70, >.025), indicating that overall processing time rather than stimulus duration per se improves accuracy. This is contrary to the multiple sampling hypothesis forwarded by Inglis & Gilmore, but we note that we cannot exclude the possibility that sampling continues in iconic memory. These findings provide insights into the processes underlying number representation.