Search

Humans can estimate numerosities-such as the number sheep in a flock-without deliberate counting. A number of biases have been identified in these estimates, which seem primarily rooted in the spatial organization of objects (grouping, symmetry, etc). Most previous studies on the number sense used static stimuli with extremely brief exposure times. However, outside the laboratory, visual scenes are often dynamic and freely viewed for prolonged durations (e.g., a flock of moving sheep). The purpose of the present study is to examine grouping-induced numerosity biases in stimuli that more closely mimic these conditions. To this end, we designed two experiments with limited-dot-lifetime displays (LDDs), in which each dot is visible for a brief period of time and replaced by a new dot elsewhere after its disappearance. The dynamic nature of LDDs prevents subjects from counting even when they are free-viewing a stimulus under prolonged presentation. Subjects estimated the number of dots in arrays that were presented either as a single group or were segregated into two groups by spatial clustering, dot size, dot color, or dot motion. Grouping by color and motion reduced perceived numerosity compared to viewing them as a single group. Moreover, the grouping effect sizes between these two features were correlated, which suggests that the effects may share a common, feature-invariant mechanism. Finally, we find that dot size and total stimulus area directly affect perceived numerosity, which makes it difficult to draw reliable conclusions about grouping effects induced by spatial clustering and dot size. Our results provide new insights into biases in numerosity estimation and they demonstrate that the use of LDDs is an effective method to study the human number sense under prolonged viewing.

The study investigated potential effects of the presentation order of numeric information on retrospective subjective judgments of descriptive statistics of this information. The studies were theoretically motivated by the assumption in the naïve sampling model of independence between temporal encoding order of data in long-term memory and retrieval probability (i.e. as implied by a "random sampling" from memory metaphor). In Experiment 1, participants experienced Arabic numbers that varied in distribution shape/variability between the first and the second half of the information sequence. Results showed no effects of order on judgments of mean, variability or distribution shape. To strengthen the interpretation of these results, Experiment 2 used a repeated judgment procedure, with an initial judgment occurring prior to the change in distribution shape of the information half-way through data presentation. The results of Experiment 2 were in line with those from Experiment 1, and in addition showed that the act of making explicit judgments did not impair accuracy of later judgments, as would be suggested by an anchoring and insufficient adjustment strategy. Overall, the results indicated that participants were very responsive to the properties of the data while at the same time being more or less immune to order effects. The results were interpreted as being in line with the naïve sampling models in which values are stored as exemplars and sampled randomly from long-term memory.