Judgments regarding the physical properties of stimuli are influenced by a multitude of factors, including primary and secondary features of the stimulus under scrutiny and their relative salience, the context in which the stimulus is perceived, as well as top-down influences such as preconceived cognitions regarding the stimulus under investigation. The theory of information integration put forth by Anderson (1981) accounts for evaluative judgments based upon physical features of the target that are collectively integrated into an overall judgment. Contextual events, as well as internal states, including mood, dictate which features we attend to and remember (see also Winkielman, Schwarz, Fazendeiro, & Reber, 2003). For example, when judging how tall someone is, one takes into account the person’s physical height, which may be influenced by context (e.g., adjacent people or objects), secondary attributes (e.g., person’s weight), and potentially even one’s schema for the person’s profession or hobbies (e.g., a person’s height may be overestimated if identified as an athlete versus a nonathlete). 

Regarding evaluative judgments of size, systematic investigations reveal that people (and animals) are quite astute at estimating relative and absolute size, and yet there are predictable misperceptions, giving way to size illusions. One notable example is the Delboeuf illusion, which occurs when the size of a central dot is overestimated when encircled by a small outer ring and underestimated when encircled by a large outer ring (Delboeuf, 1892). Adult humans readily perceive this illusion in two-dimensional format, which is attributed to contrast and assimilation mechanisms in which items more distally positioned are contrasted with one another, leading to underestimates of size, whereas objects more proximally positioned to one another are perceptually integrated, leading to overestimates of size (Oyama, 1960; see Nicolas, 1995and Parrish, 2020, for a review). The Delboeuf illusion appears within real-world decisional settings as well, such that adult humans overestimate food portions (e.g., amounts of cereal) presented in small dishware (assimilation effect between food and small bowl) and underestimate food portions presented in large dishware (contrast effect between food and large bowl; e.g., Van Ittersum & Wansink, 2012). Similarly, nonhuman primates, including great apes and monkeys, are sensitive to the Delboeuf illusion, including for the two-dimensional array (Parrish, Brosnan, & Beran, 2015) and when presented with a relative food choice task (Parrish & Beran, 2014), along with a number of nonprimate species (e.g., fish: Fuss & Schluessel, 2017; reptiles: Santacà, Miletto Petrazzini, Agrillo, & Wilkinson, 2019; cats: Szenczi, Velázquez-López, Urrutia, Hudson, & Bánszegi, 2019).

Beyond contextual influences on size judgments like those described for the Delboeuf illusion, estimates of size also are influenced by secondary attributes inherent to the stimuli themselves. For example, color and contrast impact the perception of object size, including perceived weight. In a behavioral consumerism task, adult humans judged 3D objects of greater color saturation as larger in size than identical items of less color saturation, and in turn rated these products as more favorable and valuable (measured in terms of willingness to pay; Hagtvedt & Brasel, 2017). Relatedly, darker objects are perceived to be heavier than identical objects of lighter coloration in the so-called brightness–weight illusion (Walker, Francis, & Walker, 2010). The impact of color on size judgments is not limited to humans; one example is the expansion–contraction color effect in which warm colors, such as yellow, are perceived to be larger and closer than cool colors, such as blue, which emerges similarly for nonhuman species (e.g., Albertazzi, Rosa-Salva, Da Pos, & Sovrano, 2015). And size illusions are affected by varying the contrast between the illusory array and background stimuli; for example, the magnitude of the Delboeuf illusion increases under high-contrast conditions (e.g., Van Ittersum & Wansink, 2012; Weintraub & Cooper, 1972).

In a recent study with adult humans, the impact of figure–ground contrast on perceived stimulus height was explored. Participants judged high-contrast stimuli (black stimuli on a white background) to be taller than low-contrast stimuli (gray stimuli on a white background), including for letter pairs and pseudoletter pairs (Barra, Pallier, & New, 2020). Relatedly, for literate adults, letters appeared taller than identically sized pseudoletters, and words appeared taller or greater in size than pseudowords of the same size (New, Doré-Mazars, Cavézian, Pallier, & Barra, 2016; Reber et al. 2004a). The interactive activation model (IAM) proposes that stimuli are coded first for their features, which receive bottom-up activation of the stimulus properties themselves (e.g., contrast), and subsequently at the level of letters and then words, which receive top-down activation regarding unit meaning (McClelland & Rumelhart, 1981). In the study by Barra et al. (2020), a cumulative effect of contrast and letter was reported, such that mixed sets of variably sized letters versus pseudoletters yielded the best performance if a high-contrast black letter was compared against a low-contrast gray pseudoletter. Presumably, black letters received greater activation at the feature level (higher figure–ground contrast) as well as the letter level, above and beyond the gray pseudoletter, which yielded less activation at the feature level (lower figure–ground contrast) and no activation at the letter unit.

These results also are discussed in light of perceptual fluency, in which stimuli that are processed faster are more easily identified, preferred, and judged as more memorable in metacognition tasks (e.g., Jacoby & Dallas, 1981; Reber, Winkielman, & Schwarz, 1998; Wänke, Schwarz, & Bless, 1995). Manipulations of stimulus features, including quality, contrast, size, duration, and repetition influence “ease of processing” or perceptual fluency (see Jacoby, 1983; Roediger, 1990; Tulving & Schacter, 1990; Winkielman et al., 2003, for reviews). For example, words printed in larger fonts are judged to be more memorable than words printed in smaller fonts (Rhodes & Castel, 2008). Increasing font size for early readers (children ages 5–7 years old) and dyslexic children increased reading rates and comprehension (Hughes & Wilkins, 2002; Katzir, Hershko, & Halamish, 2013; O’Brien, Mansfield, & Legge, 2005). The effects of perceptual fluency also have been observed in nonhuman primates, including rhesus monkeys, which judged high-contrast stimuli to be more memorable than low-contrast stimuli in a meta-memory task, although there was no difference in memory performance across contrast levels (Ferrigno, Kornell, & Cantlon, 2017).

Barra et al. (2020) suggested that bottom-up or featural influences (e.g., variable contrast) impact judgments early in processing via differential signal-to-noise ratios, which then interact with down-stream processing of verbal stimuli or for higher order judgments such as metamemory estimates. In the current study, we extended the investigation of differential feature activation on size judgments to two populations, including a nonhuman primate species (rhesus macaques) and preschool children (3 to 5 years old) to isolate the role of contrast in size estimates devoid of verbal comprehension among a nonverbal animal species and preliterate or early literate human children. This work contributes to our understanding of the role of secondary characteristics such as figure–ground contrast on size estimation. Despite the comparative studies on the brightness illusion (e.g., Agrillo, Miletto Petrazzini, & Bisazza, 2016; Hodos & Leibowitz, 1978; Huang, MacEvoy, & Paradiso, 2002; Kinoshita, Takahashi, & Arikawa, 2012) and various size illusions (e.g., Byosiere, Chouinard, Howell, & Bennett, 2020; Fuss & Schluessel, 2017; Nakamura, Watanabe, & Fujita, 2008, 2014; Parrish et al., 2015; Parrish & Beran, 2014; Parron & Fagot, 2007; Qadri & Cook, 2019), there has not been an investigation of size judgment performance as a function of figure–ground contrast in animals or children to our knowledge. These results would be informative to the discussion of the level at which these phenomena emerge (early vs. late visual processing) and the evolutionary and developmental trajectory of such effects. Thus, we extended the height superiority effect as established with variable levels of figure-ground contrast by Barra et al. (2020) to rhesus monkeys and children. In the current study, we focused efforts at the feature level, presenting pairs of shapes of the same and different size and varying contrast (high contrast—black/white shapes on white/black backgrounds, and low contrast—gray shapes on white/black backgrounds) to assess the role of contrast on misperceptions of size. Continuity in performance for this phenomenon across children and a nonhuman primate species would indicate that the mechanisms underlying perceptual fluency likely are widespread, emerging at the level of bottom-up mechanisms and further amplified via top-down mechanisms.

Human children and rhesus macaques engaged a size-discrimination task in which they were rewarded for selecting the larger of two shapes. To establish baseline performance, we presented control trials with two different-sized shapes on identical backgrounds (e.g., black vs. black shapes or gray vs. gray shapes on white background). In critical test trials, we introduced congruent trials in which the larger shape also was higher in contrast (black shape on white background or white shape on black background) relative to the low-contrast shape (gray shape on white or black background). This condition was predicted to facilitate size discrimination performance akin to what has been documented among adult humans (Barra et al., 2020). Incongruent trials presented the smaller shape in high-contrast (black shape on white background or white shape on black background) relative to the low-contrast, larger shape (gray shape on white or black background). The incongruent condition was predicted to disrupt size discrimination performance if subjects overestimated the smaller, high-contrast shape, particularly when faced with difficult discriminations or identically sized shapes.

Experiment 1

Method

Subjects

We tested five adult male rhesus macaques (Macaca mulatta) from Georgia State University’s (GSU) Language Research Center, ranging in age from 17 to 28 years. All monkeys had extensive experience in computerized cognition testing. Monkeys were individually housed, with daily access to outdoor enclosures in which they could choose to spend time with a companion monkey. Indoor and outdoor enclosures included a variety of enrichment items (e.g., toys, foraging puzzles) as well as access to a computer in the monkeys’ indoor enclosure that allowed them to engage in cognitive testing. Monkeys received a daily diet of chow and fruits/vegetables, with continuous access to water. This study was approved by the Georgia State University IACUC. Georgia State University is accredited by American Association for Laboratory Animal Science.

Human participants included 34 children (20 females) between the ages of 38 and 65 months at the start of the experiment (M = 55.5 months, SD = 7.5 months). Children were recruited for participation in the current task and other cognitive testing through preschool centers in the Atlanta, GA, area. Children were tested individually in the lunchroom or playroom area of the preschool and received a small toy or sticker for participation. Children were instructed that they could stop the task and return to their classroom whenever they chose. Parental permission was obtained prior to participation, children always assented to working with the researchers, and this study was approved by Georgia State University Institutional Review Board.

Apparatus

Monkeys were tested using the Language Research Center’s Computerized Test System, which included a personal computer, digital joystick, 17-inch color monitor, and food pellet dispenser (Evans, Beran, Chan, Klein, & Menzel, 2008; Richardson, Washburn, Hopkins, Savage-Rumbaugh, & Rumbaugh, 1990). Computer programs were written using Microsoft Visual Basic 6.0. Monkeys were not restrained during testing and viewed the monitor from approximately 30 cm to 40 cm. Children were tested using PC laptop computers, with a 17-inch color monitor. Children responded via touch-screen presses rather than joystick responding. They viewed the monitor from approximately 40 cm to 50 cm.

Design and procedure

The procedure was similar for children and for monkeys. Participants engaged a two-choice size-discrimination task, with the objective of selecting the larger of two simultaneously presented stimuli. Children responded through touch-screen presses to stimuli, whereas monkeys moved a cursor on-screen by deflecting a joystick with their hand to control movement of that joystick. Children received happy cheers when correct and short negative sounds when incorrect as well as a brief time-out period (4 seconds). Monkeys also received auditory feedback when correct and a food reward (45 mg Bio-Serv food pellets). When incorrect, monkeys heard a buzz tone and experienced a brief timeout period (20 seconds).

Monkeys learned the task contingencies via trial-and-error learning, and children were provided minimal instructions for the task (“Choose a shape, and if you choose the right one, you will hear a cheer sound; otherwise, there will be a buzz sound”). To minimize cuing for the children, experimenters did not monitor the screen during the task, instead providing general feedback to keep the child engaged, and experimenters deflected any questions regarding specific task performance or task rules. This ensured that children, like monkeys, had to learn to choose the larger stimulus through computer-controlled feedback.

On each trial, children and monkeys saw two stimuli presented on the computer screen, to the left and right of the screen’s top area. Stimuli remained on-screen until a selection was made, and thus this task was not performed under speeded judgment, given the need to use different response modes (touch screen versus joystick) across species. During training trials, those stimuli were either both gray or were both presented as contrast colors against the opposite background (i.e., white shapes on a black background or black shapes on a white background). Gray shapes were presented on the same background color as the contrast condition (i.e., gray and white shapes on a black background or gray and black shapes on a white background). For monkeys, the contrast was always black shapes on a white background and gray shapes on a white background. Children were randomly assigned to either white on black or black on white trials presentations for this condition.

Stimulus design

On each trial, the software generated two rectangular stimuli onscreen. Both stimuli always had a width of 40 mm. The first stimulus was given a height randomly chosen between 60 and 70 mm. The second shape was increased in size relative to the first shape by adding 0–10 additional mm to its height. These additions reflected what we called Level 0 to Level 10 trials. This range of sizes of stimuli gave these rectangles visual angles ranging from 16 to 33 degrees, depending on the distance the child or monkey was from the screen. The larger the level for a trial, the objectively easier the discrimination. On each trial, the larger shape was randomly assigned to the left or the right side of the screen. Note that Level 0 presented two equal-sized shapes for the truest test of the illusory effect of contrast. To ensure that absolute position onscreen could not be used as a cue, on each trial both shapes were positioned randomly from 12 to 30 mm from the top screen edge. The left stimulus was randomly positioned from 12 to 30 mm from the left screen edge and the right stimulus was randomly positioned from 120 to 138 mm from the left screen edge.

Training

During training, all trials were Levels 8, 9, or 10 (indicating that the larger shape was 8, 9, or 10-mm taller than the smaller shape). The training criterion was at least 9 of 10 most recent trials correct for children. For monkeys, the training criterion was at least 34 of the most recent 40 trials correct. Once this criterion was met, participants moved directly to the testing phase.

Testing

In the testing phase, children completed 60 additional trials in a single session that lasted approximately 10 to 15 minutes. Monkeys completed 1,200 test trials, and they worked on this task at highly variable rates. The experiment required three test sessions for Chewie, 19 for Han, three for Lou, four for Murph, and four for Obi.

During test trials, there were four trial types (see Fig. 1 for an example of trial types). For the two control trial types, both shapes were gray (“control gray”) or both shapes were in the contrast color (“control contrast”) such that two black shapes were on a white background or two white shapes were on a black background. Again, monkeys always saw two gray shapes on a white background (“control gray”) or two black shapes on a white background (“control contrast”). For control trials, there was no chance for a contrast effect to occur between the two stimuli to be discriminated, but these trials allowed us to determine whether subjects understood the “choose bigger” rule and whether greater contrast in general led to better perceptual acuity. We predicted that children and monkeys would perform equivalently in choosing the larger of two gray shapes and in choosing the larger of two highly contrasted shapes (white or black).

Fig. 1
figure 1

Example of test trials, including control contrast trials (top left panel), control gray trials (top right panel), contrast larger trials (in which the higher contrast black shape was the larger of the two; bottom left panel), and contrast smaller trials (in which the lower contrast gray shape was the larger of the two; bottom right panel)

For the two test trial types, we presented a contrast larger condition, in which the higher contrasted shape was truly larger (congruency between size and contrast). The contrasting color (black when the background was white, or white when the background was black) was the larger shape compared with the gray shape. For the contrast smaller condition, the more highly contrasted shape was truly smaller than the gray shape (incongruency between size and contrast). The prediction was that performance should be higher in the congruent contrast larger condition than the incongruent contrast smaller condition if monkeys and children show a similar effect as human adults, in which stimuli with greater contrast to their background are perceived as larger.

For children, levels during the test phase ranged from 0 (shapes were same size) to 4, and equal numbers of trials with each level of each condition were presented to each child. For monkeys, levels ranged from 0 to 9, so that in each block of 40 trials a monkey saw one trial at each level for each trial type. This differed from the children because monkeys were completing many more trials than was feasible with children, and because we wanted to have both species perform numerous trials at their perceptual threshold rather than with discriminations that were too easy, because that likely would mask the perceptual phenomenon of interest. On trials where the stimuli were the same size (Level 0), one stimulus was randomly chosen as the “correct” option, and if it was selected positive feedback was given (cheers for children and pellets for monkeys), whereas if the other stimulus was selected, negative feedback was given (negative sound and 4-second timeouts for children and 20-second timeouts for monkeys).

Results

Children

Children who were shown contrast colors in white required an average of 22.41 trials (SD = 6.58) to complete the training phase, and children who were shown contrast colors in black required an average of 28.76 trials (SD = 16.04) to complete the training phase. This difference in training trials was not significant, t(32) = 1.51, p = .14. A general linear model (GLM) univariate analysis was conducted with contrast color (black or white) as a between-subjects independent factor, trial type and trial level (1 to 4) as within-subjects factors, and percentage correct as the dependent measure. Because Level 0 trials did not have a correct larger shape, they were excluded from this analysis. Mauchly’s test of sphericity indicated that the assumption of sphericity had been violated for level only, χ2(5) = 11.85, p = .037; thus, we report the Greenhouse–Geisser correction for main effects of level. There was no main effect of contrast color, F(1, 32) = 0.0, p = 1.00, ηp2 = .000. There was not a main effect of level, F(2.46, 78.84) = 1.36, p = .264, ηp2 = .04. There was a main effect of trial type, F(3, 96) = 8.43, p < .001, ηp2 = .21. The only significant interaction was of contrast color and trial type, F(3, 96) = 2.87, p = .041, ηp2 = .08.

Figure 2 presents performance as a function of trial type and trial level (1 to 4) for each contrast color. We conducted two planned contrast analyses using paired t tests. Because of the interaction of contrast color and trial type, we made these comparisons for each contrast color group, collapsing across trial level. We compared performance on contrast larger trials, where the larger shape was in the contrast color (i.e., black on white background, or white on black background) versus the smaller gray shape to the contrast smaller trials, in which the larger shape was gray and the contrast shape was smaller. Children in both color conditions performed significantly better on trials in the contrast larger trials, in which the contrast color was the larger shape (white background, M = 69.61%; black background, M = 70.10%), compared with contrast smaller trials, in which the contrast color was the smaller shape (white background, M = 46.57%; black background, M = 57.84%). For both color conditions, this difference was significant, white background t(16) = 2.77, p = .014; black background t(16) = 2.71, p = .016. Thus, congruency between contrast and size, such that the larger shape also was presented in higher contrast, increased size judgment performance for preschoolers in the present task.

Fig. 2
figure 2

Performance by children in Experiment 1 as a function of trial type and trial level (1 to 4) for each contrast color (white background and black background). For children, trial level did not affect performance. The two contrast conditions differed from each other, but the two control conditions did not. Error bars indicate standard errors of the means

We also compared performance on control trials in which both shapes were in the contrast color (white background, M = 61.76%; black background, M = 51.47%) to trials in which both shapes were in gray (white background, M = 62.75%; black background, M = 61.27%). There was no difference in performance for these trial types for children who were presented with a white background, t(16) = 0.22, p = .829. However, for children presented with a black background, performance was better for trials where both stimuli were gray compared with white, t(16) = 2.31, p = .034. This last effect explained the interaction of background color and trial type, but was not related to the question of the illusory effect of stimulus contrast.

For Level 0 trials, in which the shapes were the same size, we examined the number of times (out of six trials per child) that the contrast color was selected over the gray shape. The children selected that contrast shape on an average of 3.44 trials (SD = 1.28), and a one-sample t test showed that this did not reach statistical significance, t(33) = 2.00, p = .053.

Monkeys

Monkeys required an average of 556 trials (SD = 367) to complete the training phase. A general linear model (GLM) univariate analysis was conducted, with trial type and trial level (1 to 9) as within-subjects factors, and percentage correct as the dependent measure (see Fig. 3). Mauchly’s test of sphericity indicated that the assumption of sphericity had been violated for trial type only, χ2(5) = 19.23, p = .003; thus, we report the Greenhouse–Geisser correction for main effects of condition. Level 0 trials were excluded from this analysis. There was a main effect of level, F(8, 32) = 18.25, p < .001, ηp2 = .82. There was not a main effect of trial type, F(1.05, 4.20) = 2.13, p = .15, ηp2 = .35. There was not a significant interaction, F(24, 96) = 1.48, p = .09, ηp2 = .27. Planned comparisons using paired t tests across level indicated that the monkeys did not perform significantly better on trials in the contrast larger trials, in which the contrast color was the larger shape (M = 76.62%), compared with the contrast smaller trials when the gray shape was larger (M = 53.53%), t(4) = 1.41, p = .23. There also was no difference in performance when both shapes were the contrast color (M = 70.59%) compared with when both shapes were gray (M = 69.37%), t(4) = 0.98, p = .39.

Fig. 3
figure 3

Performance by monkeys in Experiment 1 (white background, top two panels) and Experiment 2 (black background, bottom two panels) as a function of trial type and trial level. In both experiments, the two contrast conditions did not differ significantly, nor did the two control conditions. Error bars indicate standard errors of the means

Because of the small sample size, we also analyzed each monkey’s data individually (see Fig. 4 and Table 1). For each monkey, a general linear model (GLM) univariate analysis was conducted, with trial type as a fixed factor, trial level (1 to 9) as a co-variate, and percentage correct as the dependent measure. Chewie showed a significant effect of level, F(1, 28) = 21.41, p < .001, ηp2 = .43, and a main effect of trial type, F(3, 28) = 14.69, p < .001, ηp2 = .61. There was no interaction, F(3, 28) = 1.19, p = .33, ηp2 = .11. Lou showed a significant effect of level, F(1, 28) = 27.63, p < .001, ηp2 = .50, but no main effect of trial type, F(3, 28) = 1.62, p = .207, ηp2 = .15. There was no interaction, F(3, 28) = .34, p = .794, ηp2 = .04. Murph showed a significant effect of level, F(1, 28) = 22.15, p < .001, ηp2 = .44, and a main effect of trial type, F(3, 28) = 21.10, p < .001, ηp2 = .69. There was no interaction, F(3, 28) = 0.93, p = .439, ηp2 = .09. Obi showed a significant effect of level, F(1, 28) = 34.11, p < .001, ηp2 = .55, and a main effect of trial type, F(3, 28) = 3.49, p = .029, ηp2 = 27. There was no interaction, F(3, 28) = 1.05, p = .385, ηp2 = .10. As with the other monkeys, Han showed a significant effect of Level, F(1, 28) = 41.62, p < .001, ηp2 = .60. Han also showed a main effect of trial type, F(3, 28) = 7.47, p = .001, ηp2 = .45, and there was an interaction, F(3, 28) = 3.00, p = .047, ηp2 = .24. The same planned comparisons as for the group data also were performed (see Table 1). Four of five monkeys showed better performance when the larger shape was in the higher contrast color versus gray. The exception, Han, showed the opposite result. None of the monkeys performed differently on control trials when both shapes were in gray or in the higher contrast color.

Fig. 4
figure 4

Performance of individual monkeys in Experiment 1 as a function of trial type and trial level (1 to 9) for the white background condition

Table 1 Performance of each monkey on each condition in Experiment 1 and Experiment 2

Because the monkeys did many more Level 0 trials than did the children in which the shapes were of equal size, we could examine their individual proportions of choice of the contrast color over the gray color for this level (see Fig. 5). Chewie and Murph chose the contrasting shape over the gray shape (ps < .001, binomial test). Lou (p = .44) and Obi (p = .155) did not show a significant difference in choice. Han chose the contrasting shape significantly less often than the gray shape (p = .006).

Fig. 5
figure 5

Performance of individual monkeys in Experiment 1 and Experiment 2 on Level 0 trials. Binomial tests assessed performance against 50% chance responding: * indicates a significant bias to choose the higher contrast stimulus, and ** indicates a significant bias to choose the lower contrast stimulus

Discussion

Preschool children showed the expected overestimation of size for higher-contrast stimuli relative to lower-contrast stimuli, and this matched reports from adult humans. Consistent with the children, four of the five monkeys (excluding Han) performed better at the size-discrimination task when there was congruency between shape size and contrast, such that larger shapes in higher contrast facilitated performance. These same four monkeys either showed a significant preference for high-contrast shapes relative to same-sized gray shapes or showed no bias for either shape in Level 0 trials with equal-sized shapes. Again, Han showed an opposite pattern of the bias from what is typical in human adults and the children and remaining monkeys tested here for Level 0 trials, with a significant preference for gray shape relative to the same-sized high-contrast shape.

In Experiment 2, we replicated Experiment 1, with monkeys now presented with white shapes on a black background, to understand better how these results generalized to the opposite contrast. Recall that, for monkeys, the contrast condition always was black shapes on a white background and gray shapes on a white background for Experiment 1. Furthermore, Experiment 2 was included to extend the comparability of comparative and developmental testing, given that children in Experiment 1 were presented with black shapes on a white background or the reverse, white shapes on a black background.

Experiment 2

Method

The monkey subjects, stimulus design, training, and testing procedures for Experiment 2 were identical to those outlined for Experiment 1, with one notable exception. In Experiment 1, for monkeys, the contrast condition was always black versus gray shapes on a white background. We now presented monkeys with the opposite contrast condition, such that they were presented with white and gray shapes on a black background.

Results

Monkeys required an average of 110 trials (SD = 33.74) to complete the training phase. A general linear model (GLM) univariate analysis was conducted with trial type and trial level (1 to 9) as within-subjects factors, and percentage correct as the dependent measure (see Fig. 3). Level 0 trials were excluded from this analysis. There was a main effect of level, F(8, 32) = 18.72, p < .001, ηp2 = .82. There was a main effect of trial type, F(3, 12) = 4.30, p = .028, ηp2 = .52. There was not a significant interaction, F(24, 96) = 0.63, p = .90, ηp2 = .14. Planned comparisons using paired t tests across level indicated that the monkeys did not perform significantly better on trials in the contrast larger trials, in which the contrast color was the larger shape (M = 79.17%) compared with the contrast smaller trials when the gray shape was larger (M = 59.70%), t(4) = 2.22, p = .09. There also was no difference in performance when both shapes were the contrast color (M = 71.04%) compared with when both shapes were gray (M = 71.18%), t(4) = −.21, p = .84.

Because of the small sample size, we again analyzed each monkey’s data individually (see Fig. 6 and Table 1). For each monkey, general linear model (GLM) univariate analysis was conducted, with trial type as a fixed factor, trial level (1 to 9) as a co-variate, and percentage correct as the dependent measure. Chewie showed a significant effect of level, F(1, 28) = 24.63, p < .001, ηp2 = .47, and a main effect of trial type, F(3, 28) = 3.42, p = .031, ηp2 = .27. There was no interaction, F(3, 28) = 2.50, p = .08, ηp2 = .21. Lou showed a significant effect of level, F(1, 28) = 31.00, p < .001, ηp2 = .53, and an effect of trial type that approached, but did not reach, statistical significance, F(3, 28) = 2.86, p = .055, ηp2 = .23. There was no interaction, F(3, 28) = 0.23, p = .87, ηp2 = .02. Murph showed a significant effect of level, F(1, 28) = 9.18, p = .005, ηp2 = .25, and a main effect of trial type, F(3, 28) = 14.69, p < .001, ηp2 = .61. There was no interaction, F(3, 28) = 0.70, p = .56, ηp2 = .07. Obi showed a significant effect of level, F(1, 28) = 92.81, p < .001, ηp2 = .77. Obi also showed a main effect of trial type, F(3, 28) = 25.67, p < .001, ηp2 = .73, and there was an interaction, F(3, 28) = 7.50, p = .001, ηp2 = .45. Han showed a significant effect of level, F(1, 28) = 68.16, p < .001, ηp2 = .71, but no main effect of trial type, F(3, 28) = 1.41, p = .26, ηp2 = .13. There was no interaction, F(3, 28) = 0.95, p = .43, ηp2 = .09. The same planned comparisons as for the group data also were performed (see Table 1).

Fig. 6
figure 6

Performance by individual monkeys in Experiment 2 as a function of trial type and trial level (1 to 9) for the black background condition

We again examined the monkeys’ proportions of choice of the contrast color over the gray color for the Level 0 trials (where both shapes were the same size). Those data are shown in Fig. 5. Lou (p = .025, binomial test), Murph (p = .014), and Obi (p < .001) chose the contrasting shape significantly more often than the gray shape. Chewie (p = .52) and Han (p = .52) did not choose the contrasting shape over the gray shape.

General discussion

Preschool-aged children and rhesus macaques overestimated stimulus size for shapes presented in high contrast (e.g., black shapes on white background) relative to shapes presented in low contrast (e.g., gray shapes on white background). In control trials comparing two shapes of the same contrast, monkeys and children performed well in selecting the truly larger shape. In test trials, most children and monkeys performed better in the contrast larger condition, in which there was congruency between size and contrast (the higher contrast shape was truly larger), relative to the contrast smaller condition, in which there was incongruency between size and contrast (the higher contrast shape was smaller). When presented with equal-sized high-contrast and low-contrast shapes (Level 0), all monkeys demonstrated a significant bias to select the higher contrast shape to the lower contrast gray shape in at least one experiment. Level 0 trials were considered the truest test of the contrast effect on size estimates in that there was no correct answer, and a bias to select higher to lower contrast stimuli of equal size reveals, at minimum, a preference for high contrast and possibly a true misperception of size. An important follow-up would be to examine this phenomenon using a size adjustment task (i.e., adjusting of one stimulus until perceived size equality is met) or a size estimate task (i.e., reporting of perceived percent difference) to determine the degree to which these stimuli are overestimated or underestimated. Children, like monkeys, performed better for the contrast larger congruent trials than the contrast smaller trials, and this pattern held for both experiments (Experiment 1 - black and gray shapes on white background as well as Experiment 2 - white and gray shapes on black background). For children, their pattern of performance for Level 0 trials was similar to monkeys (high contrast > low contrast), although not at a statistically significant level.

These results lend developmental and comparative support to the human adult literature in which participants overestimated the size of high-contrast stimuli relative to low-contrast stimuli. In Barra et al. (2020), adult participants perceived higher contrast (black figure/white ground) stimuli to be taller than lower contrast (gray figure/white ground) stimuli, including for letter pairs and pseudoletter pairs. Barra et al. (2020) discussed their results in light of perceptual fluency, in which stimuli that are processed more quickly are more easily recognized, preferred, or perceived as more frequent/longer lasting (e.g., Goldinger, Kleider, & Shelley, 1999; Jacoby & Dallas, 1981; Reber et al., 1998; Reber et al., 2004b; Wänke et al., 1995; Whittlesea, Jacoby, & Girard, 1990; but see Reber, Christensen, & Meier, 2014). High-contrast stimuli likely are more easily processed (and subsequently overestimated or preferred) as they create a higher signal-to-noise ratio at the feature level. These effects were additive in Barra et al. (2020) in that most errors for adult humans occurred when black letters were pitted against gray pseudoletters, suggesting that black letters received greater activation at both the feature level via contrast and the letter unit level, driving size overestimates (see McClelland & Rumelhart, 1981, for more information on the interactive activation model). Future comparative work might further assess this interaction by manipulating the role of stimulus familiarity alongside variations at the feature level, such as contrast. Specifically, the use of symbolic stimuli (e.g., Arabic numerals or lexigram symbols) that some primates have a long-standing training history with (e.g., Washburn & Rumbaugh, 1991) would allow for further investigations of the role of bottom-up and top-down influences on perceptual fluency in psychophysical and conceptual discriminations. Furthermore, the use of eye-tracking data with comparative and developmental populations would be useful in exploring the role of attention in this paradigm.

Reber, Wurtz, et al. (2004a) suggested that the speed with which stimuli are processed is influenced by manipulations to stimulus meaning or perceptual features, which can enhance subjective fluency (e.g., feelings of ease for processing akin to metacognitive feelings of knowing or tip-of-the-tongue states) as well as objective fluency (e.g., speed, recall). In the current study, manipulations of perceptual fluency via figure–ground contrast led to measurable changes in objective performance for size discrimination among children and monkeys. In a comparative study by Ferrigno and colleagues (2017), perceptual fluency also was manipulated via figure–ground contrast. Interestingly, rhesus monkeys’ metacognitive judgments (in the form of confidence wagers) were affected by perceptual fluency, although memory accuracy for high-contrast versus low-contrast stimuli was not similarly affected. Metacognitive judgments by human adults also are affected by the perceptual fluency of stimuli such that stimuli that are easier to process or retrieve are judged as easier to remember (e.g., Benjamin, Bjork, & Schwartz, 1998; Busey, Tunnicliff, Loftus, & Loftus, 2000; Rhodes & Castel, 2008, 2009), suggesting convergence in the impact of fluency on metacognitive processes across nonhuman and human primates. Intriguingly, these comparative results may suggest that the effects of perceptual fluency occur outside of conscious awareness, such that stimuli that are more easily processed or of higher perceptual fluency are judged as more memorable (e.g., Begg, Duft, Lalonde, Melnick, & Sanvito, 1989; see also Undorf & Erdfelder, 2011). The current study contributes to our understanding of perceptual fluency on judgment and decision-making, disassociating subjective (e.g., metacognitive judgments, as shown by Ferrigno et al. 2017) and objective (i.e., performance-based measures, as shown here) effects on cognitive performance, particularly for nonverbal species and young children. Furthermore, it is unlikely that these effects are limited to variations in contrast; the influence of perceptual fluency on evaluative judgments like size should extend to other stimulus features (e.g., clarity, saturation, hue). The relative influence of varying stimulus features should interact with the natural history of the species under investigation (e.g., manipulations of figure–ground contrast are adaptive for visually oriented species) as well as the salience of the stimulus under investigation, which may further interact with top-down influences, such as symbolic processing or experience.

Furthermore, the present results underscore the influence of secondary stimulus attributes (figure–ground contrast) on evaluative judgments of size, akin to what is observed among adult humans (Barra et al., 2020) and young children (current study). Research on size illusions typically focus on a subset of illusory arrays known as geometric illusions, in which the target under scrutiny is misperceived on the basis of contextual stimuli (e.g., Delboeuf illusion: target dot is misperceived as a function of inducing ring size). Nonhuman primates as well as other animal species similarly experience a variety of geometric illusions, shedding light on similarities in the perceptual systems of closely and distantly related species (see Kelley & Kelley, 2014, and Parrish & Beran, 2020, for reviews). Contrast effects on size misperception extend previous comparative findings on the expansion–contraction color effects (e.g., Albertazzi et al., 2015) as well as the brightness illusion, in which a stimulus embedded in a darker background appears to be lighter than an identical object embedded in a lighter background (e.g., Agrillo et al., 2016; Hodos & Leibowitz, 1978; Huang et al., 2002; Kinoshita et al., 2012). These results suggest that brightness constancy, or the tendency to maintain similar perceptions of brightness across changes in illumination, is widespread across the animal kingdom and further underscores contrast as a critical perceived dimension.

Future comparative and developmental work should investigate the impact of not only perceptual fluency (bottom-up featural manipulations) on judgments but also conceptual fluency (top-down semantic manipulations) on perception and attention as well as higher order processes like metamemory, as the bulk of this research is constrained to adult human populations. To do this would require training animals to recognize some types of stimuli as having values that could then be used to assess top-down processes on perception (e.g., using Arabic numerals representing food amounts in a size-judgment task). Monkeys have shown effects of symbolic stimuli impacting discrimination performance in a version of the Stroop task in which relative quantities of stimuli were affected when the smaller quantity was made up of numerals of a larger given value. In that study (Washburn, 1994), the monkeys already had learned the ordinal sequence of numerals and then struggled more to choose a larger quantity of a smaller numeral (e.g., choosing five 2s over two 4s). This suggests that monkeys also may show equivalent top-down contributions to perception of stimulus size based on previously learned information about numeral value. Whether these bottom-up and top-down processes interact cumulatively to influence size judgments alongside other evaluations among nonverbal or preverbal species and populations is unknown at present.