Circles Are Detected Faster Than Downward-Pointing Triangles in a Speeded Response Task

Abstract

Simple geometric shapes are associated with facial emotional expressions. According to previous research, a downward-pointing triangle conveys the threatening perception of an angry facial expression, and a circle conveys the pleasant perception of a happy facial expression. Some studies showed that downward-pointing triangles have the advantage to capture attention faster than circles. Other studies proposed that curvature enhances visual detection and guides attention. We tested a downward-pointing triangle and a circle as target stimuli for a speeded response task. The distractors were two stimuli that resulted from the mixture of both targets to control for low-level features’ balanced presentation. We used 3 × 3, 4 × 4, and 5 × 5 matrices to test whether these shapes led attention to an efficient response. In Experiment 1, participants responded faster to the circle than to the downward-pointing triangle. They also responded slower to both targets as the number of distractors increased. In Experiment 2, we replicated the main findings of Experiment 1. Overall, the circle was detected faster than the downward-pointing triangle with small matrices, but this difference decreased as the matrix size increased. We suggest that circles capture attention faster because of the influence of low-level features, that is, curvature in this case.

Keywords

shape low-level features perception speeded response task curvature

Angry faces would capture our attention faster according to an anger superiority effect (Dickins & Lipp, 2014; Fox et al., 2000; Hansen & Hansen, 1988; Pinkham et al., 2010). Happy faces would capture our attention faster according to a happiness superiority effect (Becker et al., 2011; Calvo & Nummenmaa, 2008; Craig et al., 2014). Savage et al. (2013) found that previous anger and happiness superiority effects emerged depending on the specific face database used. They indicated the unbalanced presentation of low-level features between stimulus sets as the cause of either one of the two effects, but this imbalance was sometimes also found within the same stimulus set. The low-level features that configure the stimulus sets must be balanced. Therefore, angry and happy superiority effects may not be based on pure facial emotional content, but rather on the role of low-level features that compose the face (Coelho et al., 2011; Savage et al., 2013; Wolfe, 2018). These low-level features would guide our attention (Wolfe, 2010, 2018; Wolfe & Utochkin, 2019).

According to Larson et al. (2007, 2009, 2012) and Lobue and Larson (2010), we can evoke emotional meaning with simple nonrepresentational geometric shapes underlying facial expressions. A downward-pointing triangle would convey the threat of an angry facial expression, and a circle would convey the pleasantness of a happy facial expression (Aronoff, 2006; Aronoff et al., 1988, 1992). Larson et al. (2007) hypothesized that downward-pointing triangles capture attention rapidly because they elicit an adaptive perception of threat. They used 4 × 4 matrices in different experiments in which participants had to determine whether a target stimulus was present for each matrix. They found that downward-pointing triangles were detected slightly faster than circles in one experiment. However, they found no differences in another experiment. They concluded that the downward-pointing triangle captured attention faster than the circle. They suggested that downward-pointing triangles are similar to the geometric configuration of angry faces and that these shapes activate the amygdala (Larson et al., 2009), a neural structure related to the detection of potential threat.

According to Wolfe and Utochkin (2019), low-level visual features guide our attention efficiently. Simple shapes may elicit faster visual detection because of the advantage of some of these features to capture attention. However, Wolfe (2018) suggested that threat perception was a probable nonguiding attentional attribute if other low-level visual features were controlled for. He proposed several guiding attentional attributes such as line termination, closure, topological properties, or even curvature. Curvature has been suggested as a basic feature for visual search (Treisman & Gormican, 1988; Wolfe et al., 1992), and to elicit a quick visual detection over rectilinear shapes (LoBue, 2014; Lobue & Deloache, 2010). Given that a triangle is a sharp-angled and rectilinear shape, whereas a circle is a completely curved shape, the proposal that curvature is a guiding attentional attribute does not match the findings reported by Larson et al. (2007). Moreover, in a study about speed of processing shape, Bertamini et al. (2019) reported four experiments using abstract shapes, smoothed (curved) and angular. In the four tasks, responses for curved shapes were faster. According to the authors, there was evidence that smoothed shapes with continuous change in curvature along the contour are processed more efficiently, and they tend to be classified as targets.

From the studies about the anger versus happiness superiority effects, Frischen et al. (2008) proposed that the set size should be varied to assess linear search slopes to calculate search efficiency. Savage et al. (2013) also indicated that the procedure that uses the same stimuli as distractor and target confounds the effects because we do not know whether the speed is a consequence of the target processing or the distractor processing. Larson et al. (2007) used triangles and circles as both target and distractor stimuli. Interestingly, participants responded faster to only circle distractor matrices than to only triangle distractor matrices. Therefore, it is difficult to know whether the reported superiority of the downward-pointing triangle was due to the targets or to the distractors.

Our study aims to shed light on the apparent contradiction between the results of Larson et al. (2007) that downward-pointing triangles are detected faster than circles and the proposal that curvature is a guiding attentional attribute (Wolfe, 2018) and is processed more efficiently (Bertamini et al., 2019). In two experiments, we examined whether downward-pointing triangles or circles capture attention faster when distractors are different from both targets and the matrix size is varied. We designed a similar speeded response task to the Larson’s et al. (2007) where participants had to detect the presence of a target stimulus (either the downward-pointing triangle or the circle) or its absence (matrices with only distractors) in matrices with different number of elements. As the triangle and the circle were the target stimuli, we used two different distractor stimuli. The distractors resulted from the mixture of the two target stimuli: a half-circle and a half downward-pointing triangle. The combination of the distractor stimuli controlled for target stimulus similarity and balanced presentation of low-level features. We tried to ensure that the faster visual response was due to the target stimulus. We also examined whether the downward-pointing triangle or the circle led to a fast and accurate response. A fast and accurate visual response for a target stimulus among different matrix sizes would suggest that the low-level features of that shape guide attention (Wolfe & Utochkin, 2019). We hypothesized that circles would capture attention faster than downward-pointing triangles (LoBue, 2014; Lobue & Deloache, 2010; Treisman & Gormican, 1988; Wolfe, 1998, 2018; Wolfe et al., 1992).

Experiment 1

Methods

Participants

Larson et al. (2007) reported a large effect size of the matrix factor (downward-pointing triangle target, circle target, all downward-pointing triangles, and all circles), using eta-squared (η² > .25). Based on a statistical power of .95, and an alpha error of .05, we calculated our sample size according to both a large and medium effect size. It resulted in a sample size of 18 participants for a large effect and 40 participants for a medium effect. Thus, 57 undergraduate students (11 male) from the University of the Balearic Islands took part in the experiment in exchange for course credits (M_age = 20.46; SD_age = 5.28). All participants reported having normal or corrected-to-normal vision and provided written consent before the experiment. The study received ethical approval from the Committee for Ethics in Research of the Balearic Islands (IB 3828/19 PI), and it was conducted following the Declaration of Helsinki (2008).

Materials

A circle and a downward-pointing triangle were created as the target stimuli. We combined the two targets to create the distractor stimuli: a half-circle and a half-downward-pointing triangle. The combination of the distractors controlled for target stimulus similarity and balanced presentation of low-level features. Two distractors (Distractor 1 and Distractor 2) were created, changing the sides of the two halves (i.e., the half-circle either on the left or on the right) to control for the presentation of the half sides. Distractor 1 or “Circangle” had a half-circle on the left and a half-triangle on the right, whereas Distractor 2 or “Tricircle” had a half-triangle on the left and a half-circle on the right. The four figures were 91 pixels high and 93 pixels wide. They were created using Adobe Photoshop CS6 (Figure 1).

Figure 1.

Simple Geometric Shapes. C represents the circle shape; T represents the downward-pointing triangle shape; D1 represents the Circangle Distractor; and D2 represents the Tricircle Distractor.

We arranged the geometric shapes into matrices. The task was designed using OpenSesame 3.2 software (Mathôt et al., 2012). There were six matrix types: a circle target surrounded by Circangle Distractors, a circle target surrounded by Tricircle Distractors, a downward-pointing triangle surrounded by Circangle Distractors, a downward-pointing triangle surrounded by Tricircle Distractors, a matrix with only Circangle Distractors, and a matrix with only Tricircle Distractors. We used three matrix sizes in which the geometric shapes were equally separated: 3 × 3 (411 pixels high × 414 pixels wide), 4 × 4 (571 × 574 pixels), and 5 × 5 (731 × 734 pixels; Figure 2).

Figure 2.

Examples of the Six Matrix Types and the Three Matrix Sizes. (A) Circle target matrices; (B) Downward-pointing triangle matrices; and (C) Distractor matrices. From top to bottom: 3 × 3, 4 × 4, and 5 × 5 matrix sizes. The task was structured in three blocks regarding the three matrix sizes. In each block, Circangle and Tricircle Distractors were separated into two subblocks.

Participants carried out the task using individual computers in isolated cabins. The distance between the participant and the computer screen was 45 cm. Computers were equipped with Intel i5 processors and 21-inch screens set at 1,920 × 1,080 pixels resolution and 60 Hz.

Procedure

We organized the speeded response task in three blocks: a 3 × 3 matrix size block, a 4 × 4 matrix size block, and a 5 × 5 matrix size block. Each block was separated into two subblocks. Circangle Distractor was used in one subblock, and Tricircle Distractor was used in the other subblock. In each subblock, three matrix types were randomly presented. That is, the first subblock consisted of matrices with a circle target surrounded by Circangles, matrices with a downward-pointing triangle surrounded by Circangles, and matrices with only Circangles, whereas the second subblock consisted of the same matrix types with Tricircle Distractors. Thus, we controlled for the influence of the change of the distractor while participants performed the task (Frischen et al., 2008).

Participants carried out 400 experimental trials: 72 trials in the 3 × 3 block, 128 trials in the 4 × 4 block, and 200 trials in the 5 × 5 block. We used a different number of trials per block because the target was presented once in each matrix position. In each block, there was the same number of matrices with and without targets. There were also eight practice trials at the beginning of the task.

Participants received verbal instructions before starting the session and written instructions before the task. They were informed that geometric shapes arranged in matrices would be presented on the screen. Some of the matrices would all have the same geometric shapes, but others would have a different geometric shape from the others. Their task was to press an “equal” key when all the matrices had the same geometric shapes and a “not equal” key when the matrix had one geometric shape different from the others. They had to respond as fast as they could.

Each trial began with a central fixation cross presented for 500 ms, followed by a white screen for 100 ms. Then, a matrix was presented in the center of the screen until participants responded or for a maximum of 2,000 ms. We counterbalanced block sequence, subblock sequence, and keys for response: “equal” and “not equal,” left-side or right-side. Trials were randomized. The experiment took about 25 minutes.

Analysis

We defined errors as incorrect responses or nonresponses for the 2,000-ms stimulus presentation time. Anticipated responses were defined as faster than 200 ms. From 57 participants, we collected 22,800 trials. One participant was eliminated because of an error rate greater than 25%. Delayed (98) and anticipated (3) responses were excluded from the trial set. A total of 575 trials were eliminated because of incorrect responses. Also, 329 trials were eliminated because of statistical cleaning (extreme values), using the box plot figures. A total of 21,395 trials from 56 participants remained for the analysis of response time (RT) averages.

Results

Analyses were conducted using SPSS 25.0.0 (SPSS Inc., Chicago, IL, USA) and R environment for statistical computing (R Core Team, 2019) with an alpha level of .05. Post hoc tests used Bonferroni correction. Confidence intervals (CIs) for effect sizes were calculated using an ad hoc script (http://daniellakens.blogspot.com.es/2014/06/calculating-confidence-intervals-for.html), ascertaining the obtained values with the “MBESS” package from the R statistical software. First, we carried out a five-factor mixed analysis of variance (ANOVA) on RT with Target (circle and triangle), Matrix (3 × 3, 4 × 4 and 5 × 5) and Distractor (Circangle and Tricircle) as within-subject factors. Block sequence (i.e., random sequence of 3 × 3, 4 × 4 and 5 × 5 blocks) and subblock sequence (i.e., random sequence of distractors) were included as between-subject factors. The effects of block sequence, F(5, 55) = .6, p = .7, η_p² = .06, 90% CI [0, 0.08], and subblock sequence, F(1, 55) = 1.4, p = .24, η_p² = .026, 90% CI [0, 0.12], were nonsignificant. Hence, we conducted a three-factor repeated-measures ANOVA on RT averages with Target, Matrix, and Distractor as within-subject factors.

Results yielded a significant main effect of Target, F(1, 55) = 7.63, p = .008, η_p² = 0.12, 90% CI [0.019, 0.26]. Participant responses were faster for the matrices with the circle (M = 668 ms, SD = 87) rather than the downward-pointing triangle (M = 680 ms, SD = 90). This result suggests that they detected circles faster than triangles. Matrix also showed a significant main effect, F(2, 55) = 41.16, p < .001, η_p² = .43, 90% CI [0.44, 0.68]. Bonferroni-corrected comparisons showed that participants responded significantly faster to the 3 × 3 matrices (M = 638 ms, SD = 89) than to the 4 × 4 matrices (M = 666 ms, SD = 87), t(55) = –3.4, p = .004, g = –.32, 95% CI [–48.31, –7.6], and also faster than to the 5 × 5 matrices (M = 716 ms, SD = 90), t(55) = –9.6, p < .001, g = –.87, 95% CI [–97.64, –57.6]. They also responded faster to the 4 × 4 matrices than to the 5 × 5 matrices, t(55) = –5.2, p < .001, g = –.56, 95% CI [–73.3, –26].

Target × Matrix interaction was significant, F(2, 55) = 5.73, p = .004, η_p² = .09, 90% CI [0.03, 0.3]. As Figure 3 shows, RTs in the 4 × 4 and 5 × 5 matrices were quite similar for the matrices with the circle and triangle targets. However, the difference between circles (M = 623 ms, SD = 83) and triangles (M = 655 ms, SD = 95) was significant in the 3 × 3 matrices, t(55) = 4.36, p < .001, g = –.36, 95% CI [–53.9, –10.14]. Due to this difference, we analyzed the data only considering the nine central positions (the ones corresponding to the 3 × 3 matrices) in the 5 × 5 matrices. In this case, there was no significant difference between participant responses for the matrices with the circle (M = 671 ms, SD = 90) and triangle targets (M = 677 ms, SD = 87) according to the Wilcoxon signed-rank test, Z = 623.5, p = .16, r = .22, 95% CI [–0.48, 0.08].

Figure 3.

Experiment 1: Response Time According to Target × Matrix Size Interaction (***p < . 001). Error bars represent 95% CI.

We also found a significant effect of Distractor, F(1, 55) = 6.6, p = .013, η_p² = .11, 90% CI [0.01, 0.25]. Participants responded faster to Tricircle matrices (M = 666 ms, SD = 87) than to Circangle matrices (M = 680 ms, SD = 90). However, none of the interactions related to Distractor was significant, either Target × Distractor, F(1, 55) = 1.28, p = .26, η_p² = .02, Matrix × Distractor, F(1, 55) = 1.43, p = .24, η_p² = .02, or the triple interaction, F(1.76, 55) = .49, p = .6, η_p² = .009.

Three linear regression analyses were carried out to examine the slope of the function relating RT to matrix size. We built a model for each type of matrix, that is, matrices with a circle, matrices with a triangle, and matrices with only distractors. In each model, RT was the dependent variable. Matrix size was the predictor variable with three levels: the number of items of the matrix (i.e., 9, 16, and 25). Results showed a significant linear relationship between RT and matrix size in the three regression analyses. With the circle as a target, we found a significant but weak relationship between the RT and the number of items in the matrix, β = 5.34, t(166) = 5.12, p < .001 (R²_adj = .13). With the triangle as a target, we also found a significant and weak relationship between RT and the number of items, β = 3.82, t(164) = 3.84, p < .001 (R²_adj = .077). Finally, with only distractors (target-absent matrices), the number of items was a significant predictor of slower RTs, β = 13.04, t(162) = 8.05, p < .001 (R²_adj = .28).

We also conducted a three-factor repeated-measures ANOVA on accuracy. We defined accuracy as the proportion of the correct responses about the presence (matrices with a circle or a downward-pointing triangle) or absence (matrices with only distractors) of a target stimulus in the total number of trials. The ANOVA included Target (circle and triangle), Matrix (3 × 3, 4 × 4 and 5 × 5) and Distractor (Circangle and Tricircle) as within-subject factors. Target was nonsignificant, F(1, 51) = 2.3, p = .13, η_p² = .04, 90% CI [0, 0.16]. In contrast, matrix showed a significant main effect, F(1.8, 51) = 4.45, p = .02, η_p² = .08, 90% CI [0.012, 0.26]. Participants were less accurate with the 5 × 5 matrices (M = .956, SD = .05) than with the 4 × 4 ones (M = .967, SD = .045), t(51) = 3.2, p = .007, g = –.23, 95% CI [0.003, 0.024]. There were no significant difference between 5 × 5 matrices and 3 × 3 matrices (M = .965, SD = .05), t(51) = 2, p = .27, g = .18, 95% CI [–0.003, 0.025], nor between 4 × 4 matrices and 3 × 3 matrices, t(51) = –.5, p = 1, g = .04, 95% CI [0.003, 0.024]. Interestingly, the Target × Matrix interaction was significant, F(1.7, 51) = 5.8, p = .007, η_p² = .1, 95% CI [0.01, 0.32]. This effect was caused by the significant difference between circles (M = .979, SD = .042) and triangles (M = .956, SD = .044) in the 3 × 3 matrices, t(51) = 3.66, p = .005, g = .53, 95% CI [0.007, 0.04] (Figure 4). All other effects were nonsignificant.

Figure 4.

Experiment 1: Accuracy According to Target × Matrix Size Interaction (**p < . 01). Error bars represent 95% CI.

Discussion

Experiment 1 showed that participants responded to the matrices with a circle faster than the matrices with a downward-pointing triangle, which suggests that they detected circles faster than triangles. This result was due mainly to the difference in the 3 × 3 matrices. Participants were also more accurate when responding for matrices with a circle than for matrices with a triangle in the 3 × 3 matrices. Although the difference was not large, this result partially supported our hypothesis. In contrast, this result did not match Larson et al.’s (2007) findings. These authors indicated that a downward-pointing triangle, similar to the geometric configuration of an angry face, captured attention more rapidly than a circle, similar to the geometric configuration of a happy face. However, our results indicated that the circle shape captured attention slightly faster when we introduced different distractor stimuli and controlled for the balanced presentation of low-level features.

Both targets showed an increase in RT as the matrix size increased. These findings indicated that the response for these simple geometric shapes loses speed to guide attention as the number of distractor stimuli increases.

Last, participants responded faster to matrices with Tricircle Distractor than to matrices with Circangle Distractor, both with the circle target and the downward-pointing triangle target. This tendency was similar for the matrices with only distractors. The faster processing of Tricircle Distractors may be related to a directional preference in visual perception (Nachshon, 1985). Olivers et al. (2014) suggested that reading and writing direction may influence nonlinguistic tasks such as visual search ones. Nachson et al. (1999) showed that people scan visual stimuli in a direction that is consistent with the acquired reading/writing habits. If this was the case, it would be easier to process the stimuli when participants first process a half-triangle and then a half-circle (Tricircle) than first a half-circle and then a half-triangle (Circangle) in the same stimulus. However, we cannot determine whether this ease of processing was due to the initial processing of a half-triangle or the final processing of a half-circle. Moreover, this hypothesis would not be plausible if the stimulus were processed holistically.

In summary, Experiment 1 showed three main findings: (a) Participants responded to matrices with a circle and matrices with a downward-pointing triangle similarly fast in a speeded response task, except for 3 × 3 matrices, where they responded to the matrices with a circle significantly faster than matrices with a triangle; (b) the response slopes (i.e., the RT × Matrix Size functions) seem similar both in circles and triangles; and (c) the matrices with the Tricircle Distractor composed of a half-triangle on the left and a half-circle on the right were solved faster.

Experiment 2

Experiment 1 results did not match either our hypothesis (completely) or the findings of Larson et al. (2007). Moreover, the difference between the two distractors was unexpected, as both were composed of the same halves. Consequently, we replicated the same experiment at the University of Seville. Following Experiment 1, we expected that participants would respond slightly faster to the matrices with a circle than to matrices with a downward-pointing triangle.

Methods

Participants

The calculation for the sample size was the same as for Experiment 1. Sixty undergraduate students (12 male) from the University of Seville took part in the experiment in exchange for course credits (M_age = 21.22, SD_age = 4.93). All participants reported having normal or corrected-to-normal vision and provided written consent. The experiment received approval by the Ethics Committee of the University of Seville, and it was conducted following the Declaration of Helsinki (2008).

Materials and Procedure

We used the same speeded response task as in Experiment 1. Participants carried out the task in a room using individual computers. They received the same verbal and written instructions before the task. The distance between the participants and the computer screen was 45 cm. Computers were equipped with Intel i5 processors and 21.5-in. screens set at 1,920 × 1,080 pixels resolution and 60 Hz.

Analysis

From 60 participants, we collected 24,000 trials. Three participants were eliminated because of an error rate greater than 25%. Delayed (211) and anticipated (0) responses were excluded from the analyses. A total of 716 trials were eliminated because of incorrect responses, and 232 trials were eliminated because of extreme value statistical cleaning as in Experiment 1. A total of 21,641 trials from 57 participants remained for the analysis of RT averages.

Results

Analyses were carried out as in Experiment 1. The effects of block sequence, F(5, 56) = 0.52, p = .76, η_p² = .054, 90% CI [0, 0.07], and subblock sequence, F(1, 56) = .02, p = .9, η_p² = .0001, 90% CI [0, 0.03], on RT were nonsignificant. Hence, we conducted a three-factor repeated-measures ANOVA on RT with Target (circle and triangle), Matrix (3 × 3, 4 × 4, and 5 × 5), and Distractor (Circangle and Tricircle) as within-subject factors.

Target showed a significant main effect, F(1, 56) = 22.14, p < .001, η_p² = .28, 90% CI [0.12, 0.42]. As in Experiment 1, participant responses were significantly faster for the matrices with the circle (M = 711 ms, SD = 107) rather than the triangle (M = 741.7 ms, SD = 116). Matrix also showed a significant main effect, F(2, 56) = 30.95, p < .001, η_p² = .36, 90% CI [0.36, 0.62]. Participants responded significantly faster to the 3 × 3 matrices (M = 690 ms, SD = 115) than to the 4 × 4 matrices (M = 720 ms, SD = 109), t(56) = –2.73, p = .025, g = –.27, 95% CI [–56.54, –2.85], and faster than to the 5 × 5 matrices (M = 770 ms, SD = 112), t(56) = –7.93, p < .001, g = –.7, 95% CI [–104.5, –55]. They also responded significantly faster to the 4 × 4 matrices than to the 5 × 5 matrices, t(56) = –5.1, p < .001, g = –.45, 95% CI [–74, –26].

Target × Matrix interaction was significant, F(2, 56) = 4.48, p = .01, η_p² = .07, 90% CI [0.002, 0.2]. In the 3 × 3 matrices, participants were faster responding for the matrices with a circle (M = 669 ms, SD = 106) than for the matrices with a triangle (M = 711 ms, SD = 123), t(56) = –4.87, p < .001, g = –.36, 95% CI [–67.46, –16.15]. In the 4 × 4 matrices, participants were also faster detecting circles (M = 702 ms, SD = 103) than triangles (M = 738 ms, SD = 114), t(56) = –4.18, p < .001, g = –.33, 95% CI [–61.5, –10.2]. However, in the 5 × 5 matrices, the difference between circles (M = 762 ms, SD = 112) and triangles (M = 777 ms, SD = 111) was nonsignificant, t(56) = –1.66, p = .1, g = –.13, 95% CI [–39.9, 11.4] (Figure 5). As in Experiment 1, we analyzed the data from the nine central positions in the 5 × 5 matrices. The difference between the matrices with the circle as target (M = 716 ms, SD = 111) and the matrices with the triangle as target (M = 727 ms, SD = 88) was not significant, t(53) = –1.05, p = .3, d = –.14, 95% CI [–0.41, 0.12].

Figure 5.

Experiment 2: Response Time According to Target × Matrix Size Interaction (***p < . 001). Error bars represent 95% CI.

As in Experiment 1, the effect of distractor was significant, F(1, 56) = 10.97, p = .002, η_p² = .16, 90% CI [0.04, 0.3]. Participants responded faster to Tricircle matrices (M = 717 ms, SD = 104) than to Circangle matrices (M = 736 ms, SD = 119). However, none of the interactions related to the Distractor factor was significant, either Target × Distractor, F(1, 56) = 3.95, p = .052, η_p² = .066, Matrix × Distractor, F(1, 56) = 1.3, p = .27, η_p² = .02, or the triple interaction, F(1, 56) = .35, p = .7, η_p² = .006.

Three linear regression analyses were carried out to examine the slope of the function relating RT to matrix size as in Experiment 1. Results showed a positive linear relationship between RT and matrix size in the three regression analyses. With the circle as a target, we found a significant but weak relationship between RT and number of items, β = 4.98, t(163) = 4.3, p < .001 (R²_adj = .1). With the triangle as a target, we also found a significant and weak relationship between RT and number of items, β = 3.5, t(163) = 2.87, p = .004 (R²_adj = .04). Finally, with only distractors (target-absent matrices), the number of items was a strong significant predictor of slower RTs, β = 14.96, t(170) = 7.1, p < .001 (R²_adj = .22).

We also conducted a three-factor repeated-measures ANOVA on accuracy as in Experiment 1. Results showed that, although participants were less accurate with the matrices with a triangle (M = 0.953, SD = 0.055) than with the matrices with a circle (M = 0.961, SD = 0.054), Target was not significant, F(1, 51) = 2.56, p = .1, η_p² = .05, 90% CI [0, 0.17]. In contrast, Matrix showed a significant main effect, F(2, 51) = 6.37, p = .002, η_p² = .1, 90% CI [0.04, 0.33]. Participants were less accurate in the 5 × 5 matrices (M = 0.948, SD = 0.05) than in the 3 × 3 ones (M = 0.965, SD = 0.02), t(51) = 3.54, p = .002, g = –.44, 95% CI [0.005, 0.03]. There were no significant differences either between 5 × 5 matrices and 4 × 4 matrices (M = 0.957, SD = 0.05), t(51) = 2, p = .15, g = .18, 95% CI [–0.002, 0.021], or between 4 × 4 matrices and 3 × 3 matrices, t(51) = 1.6, p = .36, g = .21, 95% CI [0.004, 0.020]. In this case, the Target × Matrix interaction was nonsignificant, F(2, 51) = 0.54, p = .58, η_p² = .01, 90% CI [0, 0.09] (Figure 6).

Figure 6.

Experiment 2: Accuracy According to Target × Matrix Size Interaction. Error bars represent 95% CI.

Discussion

Experiment 2 replicated the main findings of Experiment 1. First, participants detected the matrices with a circle faster than the matrices with a downward-pointing triangle. We found significant differences between the two targets in the 3 × 3 and 4 × 4 matrices. Overall, participants were also more accurate responding to the matrices with a circle than the matrices with a downward-pointing triangle, but the difference was nonsignificant. These results suggest that the circle had the advantage to capture attention faster than the downward-pointing triangle when we controlled for the balanced presentation of the low-level features. Second, both targets lost speed as the number of distractors increased. This suggests that the interference from distractors led participants to greater distraction, or involved a serial search for the target across the elements of the matrices. Third, as in Experiment 1, participants responded faster to matrices with Tricircle Distractor than to matrices with Circangle Distractor. This advantage occurred in matrices with and without targets.

General Discussion

We examined whether a downward-pointing triangle or a circle had the advantage to capture attention faster in a speeded response task. Participants had to detect whether a matrix contained all the same distractor shapes or whether there was a target stimulus. We used the circle and the downward-pointing triangle as target stimuli and combined them to create two new distractor stimuli. This was aimed to control for target stimuli similarity and the balanced presentation of low-level features. We arranged these four stimuli in 3 × 3, 4 × 4, and 5 × 5 matrices to create the speeded response task. We used the task in two experiments. Experiment 1 was carried out at the University of the Balearic Islands, and Experiment 2 was carried out at the University of Seville. The objective of Experiment 2 was to replicate the unexpected results of Experiment 1.

In both experiments, participants responded faster to the matrices with a circle than the matrices with a downward-pointing triangle. However, there was a special pattern of results as a function of the matrix size. Responding to a circle was significantly faster than responding to a downward-pointing triangle with the 3 × 3 matrices in both experiments. With the 4 × 4 matrices, the difference only was significant in Experiment 2. With the 5 × 5 matrices, there was no difference. Similarly, although participants were highly accurate in both experiments, the differences revealed an overall better performance with the circle as target than with the downward-pointing triangle as target, especially in the 3 × 3 matrices of Experiment 1. This pattern shows that the difference between circles and triangles is significant in small matrices, but it decreases and becomes nonsignificant in large matrices with mixed results in the medium matrices. Consequently, we analyzed the RT in the nine central positions of the 5 × 5 matrices, that is, the nine positions that were the same as in the 3 × 3 matrices. Although participants also showed faster responses for circles than for triangles in both experiments, these differences were nonsignificant. This result showed that the faster RT for circles in the 3 × 3 matrices comes from the number of items and not from their position. It may be related to the crowding phenomenon (Whitney & Levi, 2011).

The slope of the RT by matrix size functions represents the mean time consumption of each additional item in the target-present and target-absent matrices. In the target-present matrices, the functions with downward-pointing triangles as target showed slopes lower than 4.0, whereas the functions with circles as target showed slopes around 5.0. However, the slope difference between the two targets seems to be due to the higher initial value of the triangle function in the 3 × 3 matrices. Furthermore, the adjustment of the data to the triangle function is weaker than to the circle function.

Our findings partially support that curved shapes enhance visual response over straight and sharp-angled shapes (Bertamini et al., 2019; LoBue, 2014; Lobue & Deloache, 2010; Treisman & Gormican, 1988; Wolfe et al., 1992). Other studies indicated that downward-pointing triangles evoked more rapid responses than circles because of a threat perception advantage (Larson et al., 2007, 2009). However, they used a noncontrolled target-distractor similarity, and three stimuli indistinctly as targets and distractors: circles, downward-pointing triangles, and upward-pointing triangles. It was not possible to know whether the advantage to detect downward-pointing triangles emerged from a true advantage of a shape or from the fast processing of the other shapes as distractors. Larson et al. (2007) based their conclusions in a slight difference in Experiment 3. Moreover, in three of their experiments, matrices with only circles had lower RTs than matrices with only upward-pointing triangles and matrices with only downward-pointing triangles. They also used a grid to insert the shapes. The triangles and the angles in the grid could have influenced the performance differently across conditions.

Hansen and Hansen (1988) suggested that threatening faces pop out of crowds and lead to a parallel visual search. However, this result was due to low-level visual confounds (Purcell et al., 1996; Savage et al., 2016), and facial expressions do not pop out of crowds (Cave & Batty, 2006; Fox et al., 2000; Nothdurft, 1993). With simple geometric shapes that could evoke anger (triangles) and pleasantness (circles), we found no advantage for responding to downward-pointing triangles. On the contrary, our findings showed a tendency for faster response to the circles which disappeared as matrix size increases. One of the possible explanations of this result is that the faster response to the circles appears only in small sets because the situation is more comfortable, but it disappears when the situation is crowded, as if comfort also disappears.

Low-level stimulus features can lead to differences in search efficiency (Savage et al., 2013). Several studies found an increased sensitivity to detect curved features in visual search displays (Andrews et al., 1973; Treisman & Gormican, 1988; Wilson et al., 1997; Wolfe et al., 1992). Pasupathy showed that the explicit representation of curvature in area V4 might provide a physiological basis for increased sensitivity to curvature (Pasupathy, 2006; Pasupathy & Connor, 1999, 2001, 2002). Other studies suggested a perceptual bias for the detection of simple curvilinear shapes because they are “snake-like” stimuli (Isbell, 2006, 2009; LoBue, 2014; Lobue & Rakison, 2013). However, Van Strien et al. (2016) suggested that the superior threat detection of “snake-like” stimuli was not only driven by the curvature of snakes, but most probably also by other threat-relevant physical and contextual cues. Subra et al. (2018) supported the snake-like hypothesis, but they found that modern threats captured attention faster than ancient threats. They suggested a relevance-based explanation rather than an evolutionary-based explanation of threat detection. Interestingly, Coelho et al. (2019) reported several challenges to evolutionary-based explanations of the snake detection theory and highlighted that the low-level features of the visual stimulus could have affected previous findings. Wolfe (2018) also suggested threat perception as a probable nonguiding attribute in visual search when the other basic features are controlled for. Instead, curvature was proposed as a probable guiding aspect of low-level shape features that leads to an efficient visual search (Sakai et al., 2007; Treisman & Gormican, 1988; Wolfe, 2018; Wolfe et al., 1992). That is, curvature might elicit similar RTs independently of the number of items in a matrix. Some of these studies highlight the role of low-level features in visual processing. We used simple geometric shapes and introduced distractor stimuli to control for the balanced presentation of low-level features and target stimulus similarity in a speeded response task. Although our findings are constrained to geometric shapes related to emotional expressions, the use of our neutral stimuli could contribute to understanding the role of low-level features in attentional and perceptual experience.

From the viewpoint of the anger versus happiness superiority effects, our results support the latter. Previous studies reported a relationship between curvilinearity and faces. Yue et al. (2014) presented curved versus sharp-angled natural stimuli (e.g., faces and objects) and computer-generated shapes (e.g., spheres and pyramids) to macaque monkeys in a functional magnetic resonance imaging scanner. They found three cortical patches hierarchically organized processing simple curvature, moderately complex curved features, and shapes mainly composed of curved features, respectively. Interestingly, this curvature-processing network was adjacent to a well-known face-processing network. Therefore, they suggested a possible functional link between curvature and face processing. Moreover, people associate curved features with positive valence and sharp-angled features with negative valence. Palumbo et al. (2015) pointed out the role of affective processes underlying preference for curvature. They reported that the implicit association between positive valence and curvature was stronger than the association between negative valence and sharp-angled features. We prefer curved shapes over sharp-angled ones (Corradi & Munar, 2020; Corradi et al., 2018; Gómez-Puerto et al., 2016). Altogether, the affective association between curvature and positive valence could be interpreted from the happiness superiority effect, which may explain that the circle evokes rapid capture of attention.

However, we believe that known geometric shapes are not the best stimuli to test this hypothesis. Some geometric shapes are associated not only with an affective value (Watson et al., 2011, 2012) but also with semantic meaning. Such semantic meaning could impact preference for shapes (Leder et al., 2011). Moreover, an efficient visual search or speeded response requires the target to be sufficiently different from the distractor stimulus (Duncan & Humphreys, 1989). This raises the possibility that our distractor stimuli were too similar to the targets. Thus, we need to test more basic stimuli and to introduce different distractor stimuli to investigate the role of low-level visual features guiding attention.

A related limitation is that our distractors were homogeneous within each matrix (Becker et al., 2011). However, we wanted to minimize the impact of low-level features so we combined both target shapes in the distractors to control for target stimuli similarity (Savage et al., 2016). We also tried to control distractor stimulus directionality; hence, we created the two distractor versions. We used them separately within each subblock to hold distractors and participants’ expectations constant across conditions. Nevertheless, we found faster processing with Tricircle Distractor than with Circangle Distractor. A possible explanation of this finding is the influence of reading and writing directional preference role in visual perception (Nachshon, 1985). Olivers et al. (2014) showed that literacy affects the way people sample the visual world and when they are seeking target objects from competing information. Thus, they suggested that reading and writing direction might also influence other visual tasks. People scan visual stimuli in the same direction that they acquired reading/writing habits (Nachson et al., 1999). We could expect that our participants visually processed distractor stimuli from left to right. When the half-triangle was on the left (Tricircle), it may have led to faster discrimination of the circle target, whereas it may have led to slower discrimination of the triangle. On the other hand, when the half-circle was on the left (Circangle), it may have led to more difficult discrimination of the circle target, but it may also have led to even more difficult discrimination of the triangle target because the half-circle could have captured participants’ attention. Consequently, the influence of the faster processing of Tricircle was similar in both target stimuli.

In summary, we used a speeded response task in two experiments, finding that the circles captured attention faster than the downward-pointing triangles. However, this circle’s superiority to capture attention disappears as the number of distractors increases. We suggest that circles may capture attention faster because of the low-level feature of curvature (Wolfe, 2018). However, circles and triangles are associated with semantic meaning and affective values. Therefore, they may not be the most suitable stimuli to test the natural propensity of their features to guide attention. We also highlight the need to balance the presentation of low-level features to explore visual attentional processing. Then, we could conclude that a given stimulus could drive attentional and perceptual experience.

Footnotes

Data Accessibility Statement

The data sets generated and analyzed during the current study is available on the Open Science Framework:

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the project PSI2016-77327-P of the Spanish Government (AEI/ERDF, EU) granted by the Agencia Estatal de Investigación (AEI) and the European Regional Development Funds (ERDF). E. G. C. acknowledges the predoctoral contract FPU18/00365 granted by the Ministerio de Ciencia, Innovación y Universidades.

ORCID iD

Erick G. Chuquichambi

References

Andrews

D. P.

Butcher

A. K.

Buckley

B. R.

(1973). Acuities for spatial arrangement in line figures: Human and ideal observers compared. Vision Research, 13, 599–620. https://doi.org/10.1016/0042-6989(73)90026-6

Aronoff

(2006). How we recognize angry and happy emotion in people, places, and things. Cross-Cultural Research, 40, 83–105. https://doi.org/10.1177/1069397105282597

Aronoff

Barclay

A. M.

Stevenson

L. A.

(1988). The recognition of threatening facial stimuli. Journal of Personality and Social Psychology, 54, 647–655. https://doi.org/10.1037/0022-3514.54.4.647

Aronoff

Woike

B. A.

Hyman

L. M.

(1992). Which are the stimuli in facial displays of anger and happiness: Configurational bases of emotion recognition. Journal of Personality and Social Psychology, 62, 1050. https://doi.org/10.1037/0022-3514.62.6.1050

Becker

D. V.

Anderson

U. S.

Mortensen

C. R.

Neufeld

S. L.

Neel

(2011). The face in the crowd effect unconfounded: Happy faces, not angry faces, are more efficiently detected in single- and multiple-target visual search tasks. Journal of Experimental Psychology: General, 140, 637–659. https://doi.org/10.1037/a0024060

Bertamini

Palumbo

Redies

(2019). An advantage for smooth compared with angular contours in the speed of processing shape. Journal of Experimental Psychology: Human Perception and Performance, 45, 1304–1318. https://doi.org/10.1037/xhp0000669

Calvo

M. G.

Nummenmaa

(2008). Detection of emotional faces: Salient physical features guide effective visual search. Journal of Experimental Psychology: General, 137, 471–494. https://doi.org/10.1037/a0012771

Cave

K. R.

Batty

M. J.

(2006). From searching for features to searching for threat: Drawing the boundary between preattentive and attentive vision. Visual Cognition, 14, 629–646. https://doi.org/10.1080/13506280500193107

Coelho

C. M.

Cloete

Wallis

(2011). The face-in-the-crowd effect: When angry faces are just cross(es). Journal of Vision, 10, 7. https://doi.org/10.1167/10.1.7

10.

Coelho

C. M.

Suttiwan

Faiz

A. M.

Ferreira-Santos

Zsido

A. N.

(2019). Are humans prepared to detect, fear, and avoid snakes? The mismatch between laboratory and ecological evidence. Frontiers in Psychology, 10, 2094. https://doi.org/10.3389/fpsyg.2019.02094

11.

Corradi

Munar

(2020). The curvature effect. In Nadal

Vartanian

(Eds.), The Oxford handbook of empirical aesthetics. Oxford University Press. https://doi.org/10.1093/oxfordhb/9780198824350.013.24

12.

Corradi

Rosselló

Vañó

Chuquichambi

Bertamini

Munar

(2018). The effects of presentation time on preference for curvature of real objects and meaningless novel patterns. British Journal of Psychology, 110, 670–685. https://doi.org/10.1111/bjop.12367

13.

Craig

B. M.

Becker

S. I.

Lipp

O. V.

(2014). Different faces in the crowd: A happiness superiority effect for schematic faces in heterogeneous backgrounds. Emotion, 14, 794–803. https://doi.org/10.1037/a0036043

14.

Dickins

D. S. E.

Lipp

O. V.

(2014). Visual search for schematic emotional faces: Angry faces are more than crosses. Cognition and Emotion, 28, 98–114. https://doi.org/10.1080/02699931.2013.809331

15.

Duncan

Humphreys

G. W.

(1989). Visual search and stimulus similarity the efficiency of visual selection. Psychological Review, 96, 433–458. https://doi.org/10.1037/0033-295X.96.3.433

16.

Fox

Lester

Russo

Bowles

R. J.

Pichler

Dutton

(2000). Facial expressions of emotion: Are angry faces detected more efficiently? Cognition & Emotion, 14, 61–92. https://doi.org/10.1080/026999300378996

17.

Frischen

Eastwood

J. D.

Smilek

(2008). Visual search for faces with emotional expressions. Psychological Bulletin, 134, 662–676. https://doi.org/10.1037/0033-2909.134.5.662

18.

Gómez-Puerto

Munar

Nadal

(2016). Preference for curvature: A historical and conceptual framework. Frontiers in Human Neuroscience, 9, 1–8. https://doi.org/10.3389/fnhum.2015.00712

19.

Hansen

C. H.

Hansen

R. D.

(1988). Finding the face in the crowd: An anger superiority effect. Journal of Personality and Social Psychology, 54, 917–924. https://doi.org/10.1037/0022-3514.54.6.917

20.

Isbell

(2006). Snakes as agents of evolutionary change in primate brains. Journal of Human Evolution, 51, 1–35. https://doi.org/10.1016/j.jhevol.2005.12.012

21.

Isbell

(2009). The fruit, the tree, and the serpent. Harvard University Press.

22.

Larson

C. L.

Aronoff

Sarinopoulos

I. C.

Zhu

D. C.

(2009). Recognizing threat: A simple geometric shape activates neural circuitry for threat detection. Journal of Cognitive Neuroscience, 21, 1523–1535. https://doi.org/10.1162/jocn.2009.21111

23.

Larson

C. L.

Aronoff

Stearns

J. J.

(2007). The shape of threat: Simple geometric forms evoke rapid and sustained capture of attention. Emotion, 7, 526–534. https://doi.org/10.1037/1528-3542.7.3.526

24.

Larson

C. L.

Aronoff

Steuer

E. L.

(2012). Simple geometric shapes are implicitly associated with affective value. Motivation and Emotion, 36, 404–413. https://doi.org/10.1007/s11031-011-9249-2

25.

Leder

Tinio

P. P. L.

Bar

(2011). Emotional valence modulates the preference for curved objects. Perception, 40, 649–655. https://doi.org/10.1068/p6845

26.

LoBue

(2014). Deconstructing the snake: The relative roles of perception, cognition, and emotion on threat detection. Emotion, 14, 701–711. https://doi.org/10.1037/a0035898

27.

Lobue

Deloache

J. S.

(2010). Superior detection of threat-relevant stimuli in infancy. Developmental Science, 13, 221–228. https://doi.org/10.1111/j.1467-7687.2009.00872.x

28.

Lobue

Larson

C. L.

(2010). What makes an angry face look so … angry ? Examining visual attention to the shape of threat in children and adults. Visual Cognition, 18, 1165–1179. https://doi.org/10.1080/13506281003783675

29.

Lobue

Rakison

D. H.

(2013). What we fear most: A developmental advantage for threat-relevant stimuli. Developmental Review, 33, 285–303. https://doi.org/10.1016/j.dr.2013.07.005

30.

Mathôt

Schreij

Theeuwes

(2012). OpenSesame: An open-source, graphical experiment builder for the social sciences. Behavior Research Methods, 44, 314–324. https://doi.org/10.3758/s13428-011-0168-7

31.

Nachshon

(1985). Directional preferences in perception of visual stimuli. International Journal of Neuroscience, 25, 161–174. https://doi.org/10.3109/00207458508985369

32.

Nachson

Argaman

Luria

(1999). Effects of directional habits and handedness on aesthetic preference for left and right profiles. Journal of Cross-Cultural Psychology, 30, 106–114. https://doi.org/10.1177/0022022199030001006

33.

Nothdurft

H.-C.

(1993). Faces and facial expressions do not pop out. Perception, 22, 1287–1298. https://doi.org/10.1068/p221287

34.

Olivers

C. N. L.

Huettig

Singh

J. P.

Mishra

R. K.

(2014). The influence of literacy on visual search. Visual Cognition, 22, 74–101. https://doi.org/10.1080/13506285.2013.875498

35.

Palumbo

Ruta

Bertamini

(2015). Comparing angular and curved shapes in terms of implicit associations and approach/avoidance responses. PLoS One, 10, e0140043. https://doi.org/10.1371/journal.pone.0140043

36.

Pasupathy, A. (2006). Neural basis of shape representation in the primate brain. In S. Martínez-Conde, S. L. Macknik, L. M. Martínez, J.-M. Alonso, & P. U. Tse (Eds.), Progress in Brain Research (Vol. 154, Issue SUPPL. A, pp. 293–313). Elsevier. https://doi.org/10.1016/S0079-6123(06)54016-6

37.

Pasupathy

Connor

C. E.

(1999). Responses to contour features in macaque area V4. Journal of Neurophysiology, 82, 2490–2502. https://doi.org/10.1152/jn.1999.82.5.2490

38.

Pasupathy

Connor

C. E.

(2001). Shape representation in area V4: Position-specific tuning for boundary conformation. Journal of Neurophysiology, 86, 2505–2519. https://doi.org/10.1152/jn.2001.86.5.2505

39.

Pasupathy

Connor

C. E.

(2002). Population coding of shape in area V4. Nature Neuroscience, 5, 1332–1338. https://doi.org/10.1038/nn972

40.

Pinkham

Griffin

Baron

Sasson

N. J.

Gur

R. C.

(2010). The face in the crowd effect: Anger superiority when using real faces and multiple identities. Emotion, 10, 141–146. https://doi.org/10.1037/a0017387

41.

Purcell

D. G.

Stewart

A. L.

Skov

R. B.

(1996). It takes a confounded face to pop out of a crowd. Perception, 25, 1091–1108. https://doi.org/10.1068/p251091

42.

R Core Team. (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing. http://www.R-project.org

43.

Sakai

Morishita

Matsumoto

(2007). Set-size effects in simple visual search for contour curvature. Perception, 36, 323–334. https://doi.org/10.1068/p5511

44.

Savage

R. A.

Becker

S. I.

Lipp

O. V.

(2016). Visual search for emotional expressions: Effect of stimulus set on anger and happiness superiority. Cognition and Emotion, 30, 713–730. https://doi.org/10.1080/02699931.2015.1027663

45.

Savage

R. A.

Lipp

O. V.

Craig

B. M.

Becker

S. I.

Horstmann

(2013). In search of the emotional face: Anger versus happiness superiority in visual search. Emotion, 13, 758–768. https://doi.org/10.1037/a0031970

46.

Subra

Muller

Fourgassie

Chauvin

Alexopoulos

(2018). Of guns and snakes: Testing a modern threat superiority effect. Cognition and Emotion, 32, 81–91. https://doi.org/10.1080/02699931.2017.1284044

47.

Treisman

Gormican

(1988). Feature analysis in early vision: Evidence for search asymmetries. Psychological Review, 95, 15–48. https://doi.org/10.1037/0033-295X.95.1.15

48.

Van Strien

J. W.

Christiaans

Franken

I. H. A.

Huijding

(2016). Curvilinear shapes and the snake detection hypothesis: An ERP study. Psychophysiology, 53, 252–257. https://doi.org/10.1111/psyp.12564

49.

Watson

D. G.

Blagrove

Evans

Moore

(2012). Negative triangles: Simple geometric shapes convey emotional valence. Emotion, 12, 18–22. https://doi.org/10.1037/a0024495

50.

Watson

D. G.

Blagrove

Selwood

(2011). Emotional triangles: A test of emotion-based attentional capture by simple geometric shapes. Cognition and Emotion, 25, 1149–1164. https://doi.org/10.1080/02699931.2010.525861

51.

Whitney

Levi

D. M.

(2011). Visual crowding: A fundamental limit on conscious perception and object recognition. Trends in Cognitive Sciences, 15, 160–168. https://doi.org/10.1016/j.tics.2011.02.005

52.

Wilson

H. R.

Wilkinson

Asaad

(1997). Concentric orientation summation in human form vision. Vision Research, 37, 2325–2330. https://doi.org/10.1016/S0042-6989(97)00104-1

53.

Wolfe

J. M.

(1998). Visual search. Attention, Perception, & Psychophysics, 20, 13–73. https://doi.org/10.1016/j.tics.2010.12.001

54.

Wolfe

J. M.

(2010). Visual search. Current Biology, 20, R346–R349. https://doi.org/10.1016/j.cub.2010.02.016

55.

Wolfe, J. M. (2018). Visual Search. In J. T. Wixted (Ed.), Stevens' Handbook of Experimental Psychology and Cognitive Neuroscience (4th ed., pp. 1–55). John Wiley & Sons, Inc. https://doi.org/10.1002/9781119170174.epcn213

56.

Wolfe

J. M.

Utochkin

I. S.

(2019). What is a preattentive feature? Current Opinion in Psychology, 29, 19–26. https://doi.org/10.1016/J.COPSYC.2018.11.005

57.

Wolfe

J. M.

Yee

Friedman-Hill

S. R.

(1992). Curvature is a basic feature for visual search tasks. Perception, 21, 465–480.

58.

Yue, X., Pourladian, I. S., Tootell, R. B. H., & Ungerleider, L. G. (2014). Curvature-processing network in macaque visual cortex. Proceedings of the National Academy of Sciences of the United States of America, 111, E3467–E3475. https://doi.org/10.1073/pnas.1412616111