Foveal crowding modifies a target's properties under a brief presentation time

Abstract

The perception of chromatic and achromatic visual information is combined and processed in the parvocellular stream; however, they are separate processes at the early stage of the visual cortex. In our previous study, we noted that there is difficulty discriminating the color of a letter target presented at the fovea under a crowded presentation for a short time. Visual crowding occurs when an easily identified isolated stimulus becomes very difficult to identify when it is surrounded by stimuli with similar properties. One opinion is that crowding reduces the ability to identify the target but not its features (e.g., color and texture); however, some studies indicated that the ability to recognize features is also impaired under peripheral crowding conditions. Here, we investigated whether the processing of chromatic information can be impaired at the fovea using a classic crowding experiment when tested at brief presentation times (20, 40, and 120 ms). The participants reported both the target's identity and chromaticity (dual task). We found that the target's identification and color discrimination are impaired when presented for 20–40 ms but that they recover for longer presentation times. This effect is increased when temporal backward masking is added. This finding suggests that crowding resembles masking under brief presentation times and occurs at a later processing stage, after an initial masking stage.

Introduction

Contextual modulation refers to the effect of the surrounding distractors on the perception of a target embedded within the distractors (Kingdom, Angelucci, & Clifford, 2014). Two main related phenomena are visual masking and visual crowding. Visual masking occurs when one visual stimulus interferes with the detection of another visual stimulus (Felsten & Wasserman, 1980). Visual crowding occurs when an easily identified isolated stimulus becomes very difficult to identify when surrounded by stimuli with similar properties (Balas, Nakano, & Rosenholtz, 2009; Bouma, 1970; Flom, Heath, & Takahashi, 1963; Whitney & Levi, 2011a). Despite the seemingly similar features of masking and crowding, there is an ongoing debate about whether they are the same or are different (Chung, Levi, & Legge, 2001; Doron, Spierer, & Polat, 2015; Lev & Polat, 2015; Pelli, Palomares, & Majaj, 2004). It was suggested that the main parameter that can discriminate between them is the effect on the target: in masking, the target's detection was impaired and may disappear, whereas in crowding, the appearance of objects (identification) was impaired (Greenwood, Bex, & Dakin, 2010; Parkes, Lund, Angelucci, Solomon, & Morgan, 2001). 

Most of the research on crowding was performed in the peripheral vision field, based on the observation that it is very difficult to produce crowding conditions in the center of the visual field (Levi, 2008; Spillmann, 2014; Strasburger, Harvey, & Rentschler, 1991). However, recent studies have successfully produced crowding at the fovea (Malania, Herzog, & Westheimer, 2007; Manassi, Sayim, & Herzog, 2012; Pelli et al., 2016; Sayim, Westheimer, & Herzog, 2008; Sayim, Westheimer, & Herzog, 2010). Most of these studies examined crowding using Vernier discrimination configurations. Another way to increase crowding is by presenting the stimulus for a very short time in the periphery (Tripathy, Cavanagh, & Bedell, 2014; Tripathy & Cavanagh, 2002) or fovea (Lev, Yehezkel, & Polat, 2015; Siman-Tov, Lev, & Polat, 2021; Siman-Tov, Lev, & Polat, 2024). A very recent study that used adaptive optics indicated that crowding can occur in the fovea (Coates, Levi, Touch, & Sabesan, 2018) when the target size is 1.85′–4′ and the critical spacings are 0.75 to 1.3 arcminutes edge-to-edge, within the limit of visual acuity. This very small letter spacing was also noted in the study of Flom (Flom et al., 1963). However, our recent data show that crowding with letter spacing larger than 4 arcmin (Lev et al., 2015) can take place in the fovea only when the stimulus is presented for a very short time, about 40 ms, and also for a small size (2 arcmin) (Siman-Tov et al., 2021). When a stimulus is presented under crowding conditions (0.4 letter-spacing) at the fovea for presentation times from 30 to 240 ms, the ability to identify the target decreases as the presentation time decreases (Lev et al., 2015). In other words, the crowding effect in the fovea decreases as the presentation time increases. This effect is consistent with the findings in peripheral crowding under backward masking (BM) conditions, showing that the crowding effect decreases as the inter-stimulus interval (ISI) increases (Yeshurun, Rashal, & Tkacz-Domb, 2015). 

The results of recent studies confirmed the finding that crowding is stronger when similarity exists between the target and flankers; this includes shape and size (Andriessen & Bouma, 1976; Levi, Toet, Tripathy, & Kooi, 1994), orientation (Andriessen & Bouma, 1976; Hariharan, Levi, & Klein, 2005), spatial frequency (Chung et al., 2001), depth (Levi et al., 1994), and color (Levi et al., 1994). These studies concluded that similarity reinforces the effect of crowding. In addition, the unequivocal finding obtained from studies in the periphery or at the fovea is that crowding is greatly reduced when the target is distinct from the flankers, such as in color, polarity, or depth, creating a “segmentation”, i.e., a visual separation between the target letter and the background (Levi et al., 1994; Põder, 2006; Põder, 2007; Polat, Bonneh, & Sagi, 2010; Sayim et al., 2008; Siman-Tov et al., 2021; Whitney & Levi, 2011a). 

Currently, there is a continuous debate regarding whether crowding and masking are related (Lev & Polat, 2015; Lev et al., 2015) or are distinct (Levi, 2008; Pelli et al., 2004; Sahar & Yeshurun, 2024). Over the last few years, we have explored the relationships between crowding and masking (Lev & Polat, 2015; Lev et al., 2015; Doron, Lev, Wygnanski-Jaffe, Moroz, & Polat, 2020; Benhaim-Sitbon, Lev, & Polat, 2023). The working assumption in our lab is that a relationship exists between masking and crowding, but it is not general and it can be found for a combination of parameters in time and space (Lev & Polat, 2015) (such as the size of the perceptive field (PF), the temporal presentation, and the retinal location). Moreover, our assumption is based on accumulating data from our lab, starting from early development (Doron et al., 2020) to adulthood (Lev & Polat, 2011; Lev & Polat, 2015) and amblyopia (Bonneh, Sagi, & Polat, 2004; Bonneh, Sagi, & Polat, 2007; Lev et al., 2015). 

A possible explanation that was recently suggested is that crowding and masking are related to the size of the PF. This proposal is consistent with the suggestion of (Flom et al., 1963) and was demonstrated in our recent studies in participants with typical visual development (Lev & Polat, 2011; Lev, Yehezkel, & Polat, 2015), children (Doron et al., 2015), and participants with subtypical binocular development (Benhaim-Sitbon et al., 2023). These results are also consistent with the Scale-shift model of crowding (Flom et al., 1963; Lev & Polat, 2011; Lev & Polat, 2015). The Scale-shift model posits that the size of the PF increases with increasing eccentricity and is correlated with an increasing crowding effect in the periphery. Essentially, the critical distance for crowding (i.e., the minimum separation required to identify a target without interference from distractors) is assumed to scale with eccentricity, in line with the size of the PF responsible for target recognition (Lev & Polat, 2011; Lev & Polat, 2015). A very relevant recent study by Bondarko, Chikhman, Danilova, and Solnushkin (2024) confirmed this proposal. They tested foveal crowding for a brief presentation time of 40 ms using a target Landolt C surrounded by two identical distractors located symmetrically along the horizontal or surrounded by a single distractor. Their results show that recognition of the test Landolt C was significantly impaired when the test and distractors were adjacent, with performance improving as the separation increased. They concluded that these findings support the hypothesis that crowding at the resolution limit occurs when both the test and distractors fall within the same smallest receptive field (RF) responsible for target recognition (Bondarko et al., 2024). This model can account for the smallest crowding effect at the fovea where the PF is small, and the larger crowding effect at the periphery due to a larger PF or at the fovea of amblyopic adults, where the PF is assumed to be large (Bonneh et al., 2004; Bonneh et al., 2007; Flom et al., 1963; Lev & Polat, 2011; Lev & Polat, 2015). 

Several studies indicated that the ability to recognize colors is also impaired under crowding conditions in the periphery (Greenwood & Parsons, 2020; Kennedy & Whitaker, 2010; van den Berg, Roerdink, & Cornelissen, 2007; Yashar, Wu, Chen, & Carrasco, 2019). However, to date, there is no evidence that this also occurs in the fovea. In our previous study (Siman-Tov et al., 2021) we found that tagging (differentiating the target from the matrix by a different color) the target in red abolished the crowding; however, the tagging effect is reduced at 20 ms (i.e., the segmentation is weaker for shorter presentation times). Note that studies consider crowding as a consequence of grouping, i.e., when a similarity exists between targets and flankers, crowding occurs (Banks, Larson, & Prinzmetal, 1979; Chung et al., 2001; Livne & Sagi, 2007). These studies indicate that crowding increases when the target is similar to the flankers and that it decreases when the target and flankers are dissimilar. Taken together, in considering the notion that crowding and grouping may be related, and that crowding may behave like masking under some spatial-temporal conditions (such as a brief presentation time, the target flanker's distance, and fovea vs. periphery), we hypothesized that color perception, hence, color discrimination, is impaired by crowding at very brief presentation times. Therefore, in this study, we aimed to better understand whether letter identification and color discrimination behave similarly under foveal crowding at brief presentation times. 

Methods

Participants

Ten adults (Experiments 1 and 2) and five adults (Experiments 3 and 4) with normal or corrected-to-normal vision and with no known neurological disorders participated in the study. Using the Ishihara test, we ensured that the participants did not have red/green (protan and deutan) color processing deficiencies. The participants were between 21 and 34 (26 ± 3.22; mean ± STD). Visual functions were evaluated (refraction, visual acuity, and binocular vision) by a certified optometrist before participation in the study. The participants signed a consent form that was approved by the Internal Review Board of Bar-Ilan University, and all procedures were performed according to the relevant guidelines and regulations and each participant was included only after “informed” consent was obtained. All the study protocols were approved by the Ethics Committee of Bar-Ilan University. 

Stimuli and procedures

The viewing distance was fixed at 185 cm. We measured the luminance intensity of the stimulus (E letter-isolated/crowding) and the background (a white screen) using a luminance meter (LS-100; Konica Minolta, Tokyo, Japan). The luminance for both the black and red of the target letter and matrix was 9.5 cd/m²; presented on a white screen of luminance 65 cd/m². The contrasts were 85% (calculated for static stimulation, \(\frac{{I - {{I}_b}}}{{{{I}_b}}}\), where I and I_b represent the luminance of the target letter and the background, respectively). The target letter was an E letter presented at the center of the screen (presented on the central fovea), and it was marked by a fixation point. The letter E was displayed at a size of 2.7 mm, corresponding to a visual acuity of 20/20 (i.e., the stroke width is one minute of arc and the overall size of the letter is five minutes of arc). All stimuli were viewed binocularly. The target letter was shown in black or red, isolated or under crowded conditions (Figure 1). For crowded conditions, there was a matrix of E letters (black or red) around the target letter, and the orientation of the Es was arranged randomly (including a situation where all Es had the same orientation). Each matrix letter's size was identical to the target letter's size. The size of the matrix was 5 × 5 letters with 0.4 letter spacing between letters. Note that 0.4 letter spacing represents 0.4 times the letter size (i.e., 2 minutes of arc); in addition, the spacing was measured from edge to edge. In addition, we followed a very common standard in the crowding literature, and in several studies from our lab, namely, that letter spacing of 0.4 letters is a critical distance that evokes crowding in the fovea in neurotypical participants (Lev et al., 2015). This procedure provides a general “unit” in the research by overcoming the variability that may emerge from using different letter sizes in different studies. The dual task of the participants (clicking on the mouse key, right or left) was to indicate the direction of the E target (the right click for E or the left click for ∃) and to discriminate between black and red (the right click for red or the left click for black). All experiments were administered in a dark room and were performed on the same day. The different presentation times were interleaved and presented in random order (20, 40, and 120 ms for Experiments 1 and 2, and 20 and 40 ms for Experiments 3 and 4), and repeated six times for each participant: 10 trials for each of the conditions in Figure 1 for a given duration. Thus Experiments 1 and 2 consisted of 1,080 trials per participant (6 conditions × 3 presentation times × 5 jitter directions × 2 trials (for each session) X 6 repetitions), Experiment 3 consisted of 1200 trials per participant (10 conditions × 2 presentation times × 5 jitter directions × 2 trials × 6 repetitions), and Experiment 4 consisted of 3840 trials per participant (64 conditions × 5 jitter directions × 2 trials × 6 repetitions). The minimum number of trials per data point was 60, which is a reliable data collection when measuring several data points. Each block lasted about four to six minutes, continuously without a break, but participants were allowed to take a break without any time limit between the blocks. We used a jitter of 0.4 letter spacing (to avoid adaptation to the target location), that is, all the stimuli were jittered together and presented in the center of the screen and also in this small jitter in five directions (middle, up, down, right, and left) at random. 

Figure 1.

 

Stimuli: (A) 1. A black isolated letter. 2. A red isolated letter. (B) 1. A black target letter in a black matrix with crowding 0.4 letter spacing (black crowding). 2. A red target letter among black flankers; a matrix with crowding 0.4 letter spacing (red tagging). (C) 1. A red target letter in a red matrix with crowding 0.4 letter spacing (red crowding). 2. A black target letter among red flankers; matrix with crowding 0.4 letter spacing (black tagging). The orientation of the Es’ flankers (in all four orientations) was arranged randomly; their orientation in the figure is for illustration purposes only. Note that the target could only be a rightward or leftward E.

Experiment 1: Identification then discrimination

The dual task of the participants was first to indicate the direction of the E target and second to discriminate between black and red. 

Experiment 2: Discrimination then identification

To double-check whether the effect of reduced color discrimination was indeed affected by the crowding conditions and not because this was the second answer in the dual task, we performed a control experiment (Experiment 2) in which the order of responses in the participants dual task was reversed (i.e., the dual task of the participants was first to discriminate between black and red and then to indicate the direction of the E target). 

Experiment 3: The effect of contrast reduction

In Experiment 3, the luminance intensity of the matrix systematically increased: 9.5, 20.6, 33.5, and 47.6 cd/m² (i.e., the contrasts decreased, respectively: 85%, 68%, 48%, and 26%). Isolated E means 0%. We included only a black matrix in this experiment. The target letter maintained the same luminance/contrast. The dual task of the participants was first to indicate the direction of the E target and second to discriminate between black and red. 

Experiment 4: Backward masking

In Experiment 4, an isolated E (black or red) was presented for 20 or 40 ms, alone and followed by an ISI of 20, 40, 80, 120, and 160 ms; a matrix (black or red) was presented for 40 ms. (In statistical analysis: there are six ISIs, but for isolated E, ISI = NaN). 

Apparatus

Stimuli were displayed on a 23.5″ (53.3 × 30 cm) LCD monitor (ASUS VG248QE; ASUS, Taipei, Taiwan) with 1920 × 1080 pixel resolution and at a 120 Hz refresh rate using an NVIDIA GeForce GT 730 graphic card. The visual angle of the LCD monitor was 16.4° × 9.3°. The monitor was designed for gaming and was found suitable for visual psychophysics due to its high temporal accuracy. The stimuli were presented using an in-house-developed platform for psychophysical experiments (PSY) developed by Yoram Bonneh, running on a Windows PC (Bonneh, Adini, & Polat, 2015). 

Data analysis

Three-way and two-way repeated measures analysis of variance (ANOVA) were performed to test the effect of two or three nominal variables (e.g., condition, presentation time, target color, and contrast). Regarding the three-way analyses, the follow-up analyses were performed based on significant interactions. A significant two-way interaction was further analyzed by averaging the data over the third variable. A significant main effect (without significant interactions) was further analyzed by averaging the other two variables. Finally, post hoc analysis (for pairwise comparisons) was performed as paired t-tests followed by Bonferroni correction. The effect size was calculated by Generalized Eta2 (η2G) for ANOVA and Cohen's d for t-tests. 

Results

An intriguing question from our previous study (Siman-Tov et al., 2021) is whether the ability to discriminate between the appearance of the color levels of targets, black or red, is also impaired under foveal crowding conditions. Here, the participants dual task was to identify the E target (a typical crowding task) and to discriminate between black and red targets having the same luminance (see the Methods). Note that despite the accepted interpretation of these terms (identification/discrimination), in this paper, identification refers to letter orientation, and discrimination refers to the color task. 

Figures 2A and 2B (identification response first) and Figures 3A and 3B (discrimination response first) present the identification and discrimination results. For Experiments 1 and 2's Identification task, there was no significant three-way interaction between Condition (isolated, crowding, and tagging), Target color, and Presentation time (Experiment 1: F(4,36) = 1.714, p = 0.168, ES = 0.015; Experiment 2: F(4,36) = 0.97, p = 0.436, ES = 0.008, three-way repeated measures ANOVA), but a significant two-way interaction was found between Condition and Target color (Experiment 1: F(2,18) = 5.091, p = 0.018, ES = 0.031; Experiment 2: F(2,18) = 3.845, p = 0.041, ES = 0.011); reflecting the fact that identification was slightly higher with red than black targets in the tagging condition but slightly lower in the crowding condition. Importantly, there was also a significant 2two-way interaction between Conditions and Presentation time (Experiment 1: F(4,36) = 12.3, p < 0.0001, ES = 0.152; Experiment 2: F(4,36) = 6.729, p < 0.001, ES = 0.051) (Table 1); although tagging did not improve performance in comparison to crowding with short presentation times (20 ms for Experiment 1 and 20–40 ms in Experiment 2), it effectively reduced the crowding effect at 120 ms [Tables 2, 3]. However, there was no significant interaction between Target color and the Presentation time, and no significant main effect of Target color (Table 1). 

Figure 2.

 

Experiment 1, Identification then discrimination. (A) Identification. (B) Color discrimination. The dual task of the participants was to indicate the direction of the E target and to discriminate between black and red. Error bars represent ± standard error of the mean.

Figure 3.

 

Experiment 2, Discrimination then identification. (A) Identification. (B) Color discrimination. The dual task of the participants was to indicate the direction of the E target and to discriminate between black and red. Error bars represent ± SE (standard error) of the mean.

Table 1.

 

Repeated measures three-way ANOVA for experiments 1 and 2 identification and discrimination tasks. Note: ES = effect size.

 

Table 2.

 

Post hoc for Experiment 1 identification and discrimination task. Notes: ES = effect size, p adj = p value after Bonferroni correction.

 

Table 3.

 

Post hoc for Experiments 2 identification and discrimination task. Notes: ES = effect size, p adj = p value after Bonferroni correction.

 

For the discrimination task, a similar three-way repeated measures ANOVA revealed a significant two-way interaction between condition and target color (Experiment 1: F(2,18) = 4.830, p = 0.021, ES = 0.025; Experiment 2: F(2,18) = 4.621, p = 0.024, ES = 0.048); in Experiment 1, red targets were discriminated more accurately than black targets in the crowding and tagging conditions but not in the isolated condition. In Experiment 2, discrimination was worse for the red than black targets in all conditions but more so for the isolated condition. This might reflect a bias to respond with “black” in the isolated condition. There was also a 2-way interaction between Conditions and Presentation time (Experiment 1: F(4,36) = 6.639, p < 0.001 ES = 0.099; Experiment 2: F(4,36) = 7.632, p < 0.001 ES = 0.093) (Table 1); In Experiment 1, the only significant pairwise comparison (after Bonferroni correction) was lower discrimination in the tagging than isolated condition with 20 ms presentation time, whereas in Experiment 2 lower discrimination with tagging than isolated emerged for both 20 ms and 40 ms presentation times, but not with 120 ms presentation times (Tables 2, 3). This might suggest that when presentation time was short participants tended to base their responses on the matrix color, which increases accuracy in the crowded condition in comparison to the tagging condition. In Experiment 2, we also found a two-way interaction between target color and presentation time (F(2,18) = 10.284, p = 0.001, ES = 0.061); discrimination was lower with red than black target with the shorter presentation times (20–40 ms) but not with the longest (120 ms). 

Note that for the identification task, the percentage of correct answers increases (improves) under isolated and crowded conditions as the presentation time increases, consistent with the study of Lev et al. (2015). Additionally, with identification in both experiments, tagging the target in red reduced crowding, particularly for 120 ms (Tables 2, 3), which is consistent with our previous study (Siman-Tov et al., 2021). For the isolated E in the discrimination task, the performance (percentage correct) was lower for 20 ms but still higher than the chance level (75%) (Figures 2B, 3B) (i.e., the ability to discriminate starts from the accepted threshold and not from a chance level). 

The results of Experiments 1 and 2 indicated that tagging improved identification in comparison to the crowded conditions only with long presentation times, and lower discrimination with tagging emerged particularly with 20 ms presentation times. This may suggest that the chromatic processing is influenced by the presentation time; at shorter presentation times the color information may not be fully processed. Alternatively, the chromatic processing possibly receives interference from the surrounding letters suggested to have a quick effect of masking during the first phase of crowding (Lev & Polat, 2015). Our previous studies (Lev et al., 2015; Lev & Polat, 2015) also explain why foveal crowding is effective for short presentation times but that it diminishes and disappears for longer presentation times. Here we explored the effect of crowding at the fovea, which differs from the effect of crowding at the periphery, which can reveal crowding for longer presentation times. This can be explained by the temporal dynamics of inhibition and excitation and on the excitation inhibition balance (E/I). At the fovea, results show that the inhibition is fast and acts first to reduce the activity of the network (Lev & Polat, 2015; Sterkin, Sterkin, & Polat, 2008), thus for short presentation time the E/I balance is dominated by inhibition resulting in suppression or crowding; however, since the excitation is slower (Lev & Polat, 2015; Polat & Sagi, 2006; Sterkin et al., 2008), it delayed and combined with the inhibition at longer presentation time shifting the E/I balance to reduce or even cancel the inhibition (hence reduced crowding). At the periphery, it was shown (Lev & Polat, 2011) that the inhibition (suppression) is larger and the excitation (facilitation) is smaller than at the fovea. Thus the E/I balance is dominated by inhibition; hence, the crowding effect at the periphery is larger and is found for longer presentation times than at the fovea. 

However, an alternative opinion suggests that crowding differs from masking, and that the ability to recognize color could be attributed to the difficulty of the task (Chakravarthi & Cavanagh, 2009; Danilova & Bondarko, 2007; Levi, 2008; Levi & Carney, 2011a; Pelli & Tillman, 2008; Strasburger & Malania, 2013; Whitney & Levi, 2011b), but under masking conditions, the ability to recognize the stimulus properties is eliminated (Pelli et al., 2004; Thomas, 1985). 

Surround suppression (lateral masking as in our matrix stimulus) acts when the mask and target are not presented in the same spatial location (Cavanaugh, Bair, & Anthony Movshon, 2002; DeAngelis, Freeman, & Ohzawa, 1994). Likewise, studies suggest that crowding can be explained as surround suppression (Levi, 2008), but see (Petrov, Popple, & McKee, 2007). Thus, in crowding, surround suppression may inhibit the ability to identify the target (Cavanaugh et al., 2002). Some studies show that the flanker's contrast affects peripheral (Chung et al., 2001; Rashal & Yeshurun, 2014) and foveal (Chung et al., 2001) crowding (i.e., the flanker's inhibition can be reduced if the target's contrast is higher than the flanker's contrast) (Rashal & Yeshurun, 2014). Because the crowding effect in the periphery and fovea increases with increasing mask contrast (Pelli et al., 2004; Rashal & Yeshurun, 2014), we wanted to determine, under foveal conditions, whether the ability to discriminate between colors is also affected by the contrast of the matrix (Experiment 3). The dual task of the participants was to identify the E direction and to discriminate between black and red (the same luminance) under four different matrix contrasts (26%, 48%, 68%, and 85%). 

For Experiment 3's Identification task, there was a main effect of contrast (F(4,16) = 23.992, p < 0.0001, ES = 0.521, three-way repeated-measures ANOVA, Table 4); Identification accuracy decreased with increasing matrix contrast. We also found a significant main effect of presentation time (F(1,4) = 78.958, p < 0.001, ES = 0.492); Identification accuracy was higher for the 40 than 20 ms presentation times. However, none of the interactions reached statistical significance. For the discrimination task, there were similar significant main effects of contrast (F(4,16) = 11.062, p < 0.001, ES = 0.120) and presentation times (F(1,4) = 81.351, p < 0.001, ES = 0.154), but here there was a two-way interaction of target color and presentation times (F(1,4) = 70.585, p = 0.001, ES = 0.078) that was qualified by a significant three-way interaction (Contrast × Target color × Presentation times; F(4,16) = 6.806, p = 0.002, ES = 0.03); in general discrimination was higher for black than red targets, but this difference was larger in the harder conditions—when presentation times were short and matrix contrast was high. This three-way interaction is consistent with the bias to report “black” that was also apparent in Experiments 1 and 2, and because this bias is not critical for the goals of this study, we present in Figure 4 the data averaged across target color. Note that we included only a black matrix in this experiment. 

Table 4.

 

Three-way ANOVA for experiments 3 identification and discrimination task. Note: ES = effect size.

 

Figure 4.

 

Experiment 3: (A) Identification. (B) Discrimination. Error bars represent ± standard error of the mean.

The main result of this experiment is that E identification improved as the matrix contrast decreased, as known in previous crowding studies (Rashal & Yeshurun, 2014). This result is consistent with the notion that the surround suppression is high for high matrix contrast and decreases with decreasing contrast. Here, we also show a similar effect of surround contrast on color discrimination, showing that color discrimination is improved as the matrix contrast decreases (Figure 4B). Therefore the results show that both identification and discrimination are affected by surround contrast, suggesting that the masking component may be involved in these processing. 

The results of the above experiments may support our hypothesis that crowding may have a spatial masking component. In addition, an earlier study indicated that the foveal crowding effect increased under temporal masking (BM conditions; Lev et al., 2015) when masking was presented after the crowding stimulus, which may have a complex effect on both the target and the surrounding letters. Thus, in Experiment 4, using a paradigm similar to Chung (2016), we sought to determine whether BM would have a comparable effect on both the identification and discrimination tasks of the isolated E. Thus here the BM (matrix) followed a target that was presented alone without flankers (Figure 5). Moreover, because color discrimination of an isolated E is much less affected by short presentation times compared with crowding conditions (Table 1 [Condition × Presentation time], Figures 2B, 3B), we introduced an experimental condition (the temporal BM paradigm, known to induce suppression), which may reveal the duration effect on isolated target color discrimination. 

Figure 5.

 

The letter E appeared in black or red color, alone or before a black or red matrix. The dual task of the participants was to indicate the direction of the E target and to discriminate between black and red.

For Experiment 4's Identification task, there was no significant interaction between ISI, target color, and presentation time (Black matrix: F(5,20) = 1.527, p = 0.226, ES = 0.016; Red matrix: F(5,20) = 0.930, p = 0.483, ES = 0.007, three-way repeated measures ANOVA), nor was there any 2-way interaction with Target color [Table 5]. Therefore, the results in Figure 6 present the average of the two target colors. The analysis did reveal significant main effects of ISI (Black: F(5,20) = 51.652, p < 0.0001, ES = 0.636; Red: F(5,20) = 31.158, p < 0.0001, ES = 0.483) and presentation times (Black: F(1,4) = 330.001, p > 0.0001, ES = 0.540; Red: F(1,4) = 20.5, p = 0.011, ES = 0.315). Identification was better with longer presentation times and ISIs. Because a previous study indicated that foveal crowding increased under BM conditions, more so with shorter ISIs (Lev et al., 2015), we also examined the effects of BM (i.e., compared accuracy with and without BM, Table 5), for each ISI after averaging across presentation time (because there was no ISI × Presentation time interaction). As expected, a significant difference in identification occurred at short ISIs (20 and 40 ms), but it disappeared at longer ISIs (80 ms and above for the red matrix, and 120 ms and above for the black matrix). 

Table 5.

 

Experiment 4 Identification task, 3-way repeated measures ANOVA and target alone versus BM comparisons.

 

Figure 6.

 

Experiment 4, Backward masking, identification, and color discrimination. Error bars represent ± standard error of the mean. Note that the X-axis represents SOA, i.e., the isolated E presentation time + ISI. BM = backward masking.

For Experiment 4's discrimination task, when the matrix was black, there was a main effect of target color that was qualified by the two-way interactions Target color × ISI (F(5,20) = 5.68, p = 0.002, ES = 0.154, Table 6): Discrimination accuracy was higher for black than red targets, more so when the ISIs were short; and Target color × Presentation times (F(1,4) = 56.747, p = 0.002, ES = 0.16): Discrimination accuracy was higher for black than red targets, more so when presentation time was short. When the matrix was red, the two-way interactions Target color × ISI and Target color × Presentation times were also significant but they were qualified by the three-way interaction (Target color × ISI × Presentation times, F(5,20) = 3.407, p = 0.022, ES = 0.029, Table 6); Apart for the no BM condition, in which black targets were more easily discriminated, discrimination with the red matrix was better for red than black targets, and this was more pronounced when the task was harder—presentation time and ISI were shorter. These various effects that involve target color, once more reflect the two biases that were already observed in our previous experiments: the participants tend to respond with “black” and report the matrix color. As before, because these biases are not part of this study’s goals, the results in Figure 6 are averaged across target color also for the discrimination task. More relevant to our goals, similar to Identification, we found a significant main effect of ISI (Black: F(5,20) = 11.694, p < 0.001, ES = 0.228; Red: F(5,20) = 6.439, p = 0.001, ES = 0.394) and presentation times (Black: F(1,4) = 431.026, p < 0.001, ES = 0.463; Red: F(1,4) = 21.931, p = 0.009, ES = 0.347). Discrimination was better with longer presentation times and ISIs. Finally, when we examined the effects of BM (i.e., compared accuracy with and without BM, Table 6), for each ISI after averaging across presentation time (because there was no ISI × Presentation time interaction), with a black matrix a significant difference in discrimination occurred only at short ISIs (20 and 40 ms, although after Bonferroni correction it was only marginally significant). With a red matrix none of the comparison reached statistical significance. 

Table 6.

 

Experiment 4 discrimination task, three-way repeated measures ANOVA and target alone versus BM comparisons.

 

Discussion

The main research question in this article was to determine whether crowding can also affect the ability to discriminate between chromatic colors, such as black and red targets. To this end, we asked the participants to identify the direction of the black or red target and to discriminate between the black and red colors. An additional question was whether crowding modulates letter identification and color discrimination differently under very brief presentation times, and if so, it may suggest the existence of a masking component. 

We found that, as expected, on increasing the presentation time, the ability to identify the E target increased under all conditions, regardless of the matrix color (black/red). Importantly, as we hypothesized, the color discrimination ability decreased significantly under crowded conditions more so at short presentation times (20–40 ms) (Figures 2, 3). For the identification conditions, note that there is still a reduction of 17% (the average between Experiments 1 and 2) under crowding conditions, compared with an isolated condition at a presentation time of 120 ms (Figures 2, 3). Note also that for identification with long presentation time, tagging produced a significant improvement in the percent correct (i.e., it reduced the crowding effect) (Tables 2, 3). 

The identification results are in agreement with our previous investigation of foveal crowding (Siman-Tov et al., 2021) that indicated that tagging the target with a red color improves (reduces crowding) the target's identification for longer presentation times (80, 120, and 240 ms), but the effect was smaller in reducing crowding at short presentation times (40 ms). Thus a red target is “ungrouped” from the black flanks and vice versa. Similarly, less grouping was found when the target and surround had different colors (Baylis & Driver, 1992; Levi et al., 1994; Sayim et al., 2008). 

To confirm the hypothesis that the flankers (crowding) have a major effect on impairing the discrimination ability between black and red, we performed another two experiments to test the effect of increasing the spatial suppression (contrast), and temporal (BM) on color discrimination. Previous studies considered crowding as a suppressive effect (Levi, 2008; Polat & Sagi, 1993), and that high-contrast flankers reduce performance (Lev & Polat, 2015; Rashal & Yeshurun, 2014). In addition, reducing the contrast of the matrix was shown to decrease suppression (Lev & Polat, 2015). We found that contrast reduction improved the ability of both E identification and color discrimination. This finding reinforces the hypothesis that crowding affects the ability to discriminate between black and red. In addition, in the BM experiment, we found that both tasks, E identification, and color discrimination, were affected similarly by BM, under an ISI of 20 ms. It is suggested that spatial masking is more pronounced at brief presentation times (Lev & Polat, 2015). Temporal masking (BM) reduced the discrimination and the identification of the target, mainly for the brief SOAs (see Figure 6). Thus our results suggest that the temporal masking effect (BM) is combined with strong local spatial suppression that dominates at short presentation times (Lev & Polat, 2015; Polat & Sagi, 1993, Sterkin et al., 2008). 

There is a phenomenon in which color perception is influenced by the surrounding colors; this is known as color assimilation (Monnier & Shevell, 2003). The target stimulus is perceived as the average of the stimuli surrounding it. This theory is based on the Helson findings (Beck, 1966). However, this phenomenon is less likely to explain the current effect because it also occurs for static stimuli, whereas our effect disappears at a presentation time of 120 ms. 

Magnocellular and parvocellular systems have different optimal ranges for processing visual inputs. The magnocellular system processes information rapidly (transient), with low contrast, low spatial frequency, and high temporal frequency, and is not sensitive to color. The parvocellular system processes information slowly (sustained), with high contrast, high spatial frequency, and low temporal frequency, but it is sensitive to color. Thus, in our study, the only different feature is the target's color. Longer presentation times may be sufficient for the parvocellular system to process the target's color. However, a brief presentation time (transient) is not optimal for the parvocellular system; thus information that is processed by a magnocellular system may take over; consequently, color information may be impaired. However, this explanation is insufficient to account for our data because color perception is intact for an isolated target under the same brief presentation time and is impaired only under crowding conditions. 

Models of masking postulate that reciprocal inhibition exists between parvocellular and magnocellular systems driven by cortical cells, referred to as inter-channel inhibition (Blake, 1985; Ogmen, Breitmeyer, & Melvin, 2003). Thus a brief presentation time may be an advantage for the magnocellular system, which in turn, may exert inhibition on the parvocellular system, thus reducing the color perception. This inhibition may be larger when adding a surround matrix. Note that this model is one of the leading models in BM. 

There are several spatial-temporal ways to increase the masking (suppression), for example, increasing the contrast (Polat, Mizobe, Pettet, Kasamatsu, & Norcia, 1998; Polat & Norcia, 1996) and increasing the surround size (Lev & Polat, 2015; Levi & Carney, 2011b; Solomon & Morgan, 2000). Changing the spatial configuration also adds a masking effect (Lev & Polat, 2016). It was shown that spatial interactions are asymmetric. Collinear configurations produce facilitation, but non-collinear configurations do not produce it and may even produce suppression (Adini, Sagi, & Tsodyks, 1997; Lev & Polat, 2016; Polat & Sagi, 1993). Interestingly, it was shown that when measuring responses to moving side-by-side configuration vs. collinear configuration, it was found that responses are slower for side-by-side than for collinear configuration by 10–20 ms, both in humans (Paradis, Morel, Seriès, & Lorenceau, 2012) and cortical neurons of cats V1 (Gerard-Mercier, Carelli, Pananceau, Troncoso, & Frégnac, 2016). Thus, when combining the two configurations to produce one texture (a cross, used as crowding stimuli), the effect of suppression increases (Lev & Polat, 2016; Levi & Carney, 2011b). Levi & Carney (2011b) conducted experiments to understand the factors that limit task performance in the presence of flankers in the visual systems. They found that for flanked targets, in normal foveal vision, the critical distance is more or less proportional to the target size and the effect of suppression is higher in a cross than in a collinear configuration, probably due to suppression from the side flankers or by the temporal mismatch (10–20 ms) between the collinear and non-collinear configurations (Lev & Polat, 2016). Therefore it was suggested that the temporal mismatch is reminiscent of BM. Here, it is possible that the matrix of crowding (masking) at the brief presentation time produced a mismatch between the chromatic (parvocellular, processing the target) and the achromatic (magnocellular, processing the matrix) systems, thus impairing the color perception. Therefore, in this study, we increased the masking effect by using BM. 

We suggest that an early masking component may exist in crowding because we believe that it is the most plausible interpretation that emerged from the supporting data in this study and our ongoing lab research. Below are the main factors that support our interpretation: (i) We consider crowding as a suppressive effect (Lev & Polat, 2015; Polat & Sagi, 1993). (ii) The spatial and temporal masking effect is effective only at a short presentation time, which is considered to be an advantage for inhibition (suppression) (Lev & Polat, 2015). (iii) The masking effect is more effective at high contrast, which is believed to produce suppression (Lev & Polat, 2015; Norcia, Kasamatsu, Polat, Mizobe, & Pettet, 1998). (iv) The masking effect that emerged in this study is consistent with our previous results (Lev et al., 2015; Lev & Polat, 2015). (v) A previous study considered the “crowding” effect at the fovea as masking (Levi, Klein, & Hariharan, 2002). However, the findings observed here for the crowding Experiments (1–3) could also be explained by crowding conditions that do not involve masking. 

Conclusions

In conclusion, we found that color discrimination is also impaired by foveal crowding when the target is presented for 20–40 ms, but that it recovers for longer presentation times. In addition, color discrimination is impaired when temporal masking is added, which may be consistent with our recent study indicating that there is an early and rapid masking component in crowding and that crowding may be processed at a later stage after an earlier masking stage (Siman-Tov et al., 2024). Moreover, our recent study (Lev & Polat, 2016) has shown that high perceptual sensitivity to “cross” produced by Gabor elements is impaired when imposing a temporal mismatch of 10–20 ms between the presentation of the vertical collinear and the side horizontal Gabor elements. This temporal mismatch disrupted the synchronization and produced a perceptual separation of two spatial components consisting of the global object, which consequently impaired its visual perception (Lev & Polat, 2016). Thus the impairment of color discrimination in our study by BM (which produced temporal separation) may represent the effect of a temporal mismatch. Here, it is possible that masking at a brief presentation time, such as 20 ms, produced a mismatch between the chromatic and achromatic processing, thus delaying their binding. 

Acknowledgments

Commercial relationships: none. 

Corresponding author: Uri Polat. 

Email: [email protected]. 

Address: School of Optometry and Vision Sciences, Bar-Ilan University, Ramat Gan 5290002, Israel.