Familiarity with an Object's Size Influences the Perceived Size of Its Image

Abstract

It is known that judgments about objects' distances are influenced by familiar size: a soccer ball looks farther away than a tennis ball if their images are equally large on the retina. We here investigate whether familiar size also influences judgments about the size of images of objects that are presented side-by-side on a computer screen. Sixty-three participants indicated which of two images appeared larger on the screen in a 2-alternative forced-choice discrimination task. The objects were either two different types of balls, two different types of coins, or a ball and a grey disk. We found that the type of ball biased the comparison between their image sizes: the size of the image of the soccer ball was over-estimated by about 5% (assimilation). The bias in the comparison between the two balls was equal to the sum of the biases in the comparisons with the grey disk. The bias for the coins was smaller and in the opposite direction (contrast). The average precision of the size comparison was 3.5%, irrespective of the type of object. We conclude that knowing a depicted object's real size can influence the perceived size of its image, but the perceived size is not always attracted towards the familiar size.

Keywords: Bayesian prior; bias; cue combination; expectations; precision; size perception; visual size.

Figure 1

Overview of the stimuli that we used in the four conditions of the experiment. In the two Direct conditions, two images were presented (either Balls or Coins); in the two Indirect conditions, a ball was presented together with a grey reference disk. Note that in the analysis of the Coins condition, test and reference are switched when interpreting the size difference (see for instance insets of Figure 2) to make it easier to compare the curves.

Figure 2

Results of an example participant. The probability of seeing the right image as larger is plotted as a function of the difference in size. Dots are the data; the curve is the cumulative Gaussian that we fitted. The biases are indicated by a solid vertical line, and the horizontal error bars are its corresponding 95% confidence intervals. The inset on the lower right of each panel shows an image corresponding to a 10% size difference in that condition.

Figure 3

Participants’ individual biases in the four conditions. A negative bias corresponds to an assimilation effect of familiar size. The green outlines identify the data of the example participant of Figure 2. (A) Individual participants’ biases are plotted for the four conditions, superimposed on a boxplot. (B–G): the bias in one condition as a function of that in another condition. Dots with error bars represent individual biases with their 95% confidence intervals. Colour coding reflects the mean precision of the participant across all four conditions (see Figure 4). The dashed identity lines indicate the points with equal values for both conditions. The grey lines with shaded area indicate perpendicular regressions with their 95% confidence interval.

Figure 4

Participants’ individual precision in the four conditions. (A) Individual participants’ data are plotted for the four conditions, superimposed on a boxplot. (B–G): the precision in one condition as a function of that in another condition. Dots with error bars represent individual precision with their 95% confidence intervals. See Figure 3 for further details.

Figure 5

The bias in the comparison of the tennis ball and the soccer ball. We plotted the bias obtained from comparing the balls separately with neutral grey discs (Indirect) as a function of the bias obtained by comparing simultaneously presented images of the two objects (Direct). See Figure 3 for further details.

Figure 6

The precision of judgements for the two groups of participants. Each datapoint represents one participant in one condition. Filled symbols indicate students that participated in exchange for course credit; open symbols indicate volunteers who did not receive any compensation. (A) Precision for each condition. (B) Precision as a function of age for both groups, with for each a linear fit (line) with 95% confidence interval (grey area’s).