Effect of colour pop-out on the recognition of letters in crowding conditions

Transcription

1 Psychological Research (07) 71: DOI 10.7/s ORIGINAL ARTICLE Endel Põder Effect of colour pop-out on the recognition of letters in crowding conditions Received: 2 March 05 / Accepted: 24 January 06 / Published online: 23 May 06 Springer-Verlag 06 Abstract The crowding evect of adjacent objects on the recognition of a target can be reduced when target and Xankers diver in some feature, that is irrelevant to the recognition task. In this study, the mechanisms of this evect were explored using targets and Xankers of the same and diverent colours. It was found that facilitation nearly equal to that of diverently coloured targets and Xankers can be observed with a diverently coloured background blob in the location of the target. The diverent-colour evect does not require advance knowledge of the target and Xanker colours, but the evect increases in the course of three trials with constant mapping of colours. The results are consistent with the notion of exogenous attention that facilitates the processing at the most salient locations in the visual Weld. Introduction When a target stimulus is surrounded by other (irrelevant) stimuli, its perception is impaired. This is known as lateral masking or crowding. Spatial extent of crowding depends on the eccentricity of stimuli in the visual Weld: adverse interaction disappears when the distance between target and Xanking stimuli is larger than approximately 0.5E (E is eccentricity of the target stimulus) (Bouma, 1970; Toet & Levi, 1992). Several studies have reported that the extent of crowding depends on target Xanker similarity (e.g. Andriessen & Bouma, 1976; Kooi, Toet, Tripathy, & Levi, 1994; Nazir, 1992). There are two main kinds of explanations proposed for crowding. The Wrst one is based on inhibitory interactions between feature detectors located close to each other (e.g., Bjork & Murray, 1977; Wolford & Chambers, E. Põder Department of Psychology, University of Tartu, 78 Tiigi Street, Tartu, 50410, Estonia endel.poder@ut.ee Tel.: ). The other assumes a compulsory pooling of feature signals over some area, while the access to the individual signals within this area is impossible (He, Cavanagh, & Intriligator, 1996; Parkes, Lund, Angelucci, Solomon, & Morgan, 01). Several recent studies seem to favour the second account (Levi, Hariharan, & Klein, 02; Pelli, Palomares, & Majaj, 04). However, these two mechanisms are not necessarily incompatible with each other. In addition, the distribution of attention and response competition may play some role in the crowding evect (Wolford & Chambers, 1983). The feature inhibition account could quite naturally explain the evect of target Xanker similarity on crowding. There is much evidence, both psychophysical and neurobiological, for inhibition between adjacent mechanisms tuned to similar visual features (e.g. Blakemore & Tobin, 1972; Cannon & Fullenkamp, 1991; Cavanaugh, Bair, & Movshon, 02; Petrov, Carandini, & McKee, 05). The more features the Xankers share with the target, the stronger should their inhibitory evect on the target be. However, the exact relations of these mechanisms with crowding are not clear; and it is hard to say to what extent they can explain the target Xanker similarity evects in crowding conditions. As an alternative to the direct feature-based inhibition, a mechanism of exogenously controlled attention (Itti & Koch, 00; Koch & Ullman, 1985; Nothdurft, 00) can be considered. According to such a model, a target with a unique visual feature, presented among homogenous distractors, attracts attention automatically, and its processing is facilitated. There are also top-down accounts for the target Xanker similarity evect. Styles and Allport (1986) have suggested that priming of the target set (or category) could reduce interference from Xankers that do not belong to this primed set. Intriligator and Cavanagh (01) have noted that non-spatial feature-based selection can reduce crowding with targets that pop out among distractors. The purpose of this study is to explore possible mechanisms of the reduction of crowding in the condition of

2 642 target pop-out. In three experiments, diverent colour manipulations were used to produce pop-out of the target. The goal was to distinguish direct feature-based inhibition from the mechanism of salience-based exogenous attention. Also, the roles of bottom-up and top-down mechanisms in the target Xanker similarity evect on crowding were analysed. According to a simple feature-based inhibition model, the responses to visual features within the target are suppressed by similar features within Xankers. For example, a vertical line contained in the target is inhibited by similar lines present in the Xankers, and this inhibition is stronger if the lines have the same colour as well. In the Wrst experiment, I tried to produce a colour pop-out at the target location without a colour diverence between the target and the Xankers themselves. If the reduction of crowding persists, at least this kind of feature inhibition model must be wrong. Experiment 1 The Wrst experiment compares crowding evects in three conditions: (1) target and Xankers of the same colour, (2) target and Xankers of diverent colours, and (3) target and Xankers of the same colour, but the target on top of a diverent-coloured blob. Stimuli were letter conwgurations typical to crowding studies (Fig. 1). They consisted of a central target letter and four irrelevant Xanker letters surrounding the target. In every trial, all Wve letters were drawn randomly (with replacement) from 26 uppercase letters of the English alphabet. The height of letters was 0.5 cm. From the viewing distance of approximately cm, the letters subtended about 0.5. The letters were presented randomly either to the left or right from the Wxation point. Eccentricity of the target letter was 3.3. The distance between the Xankers and the target (measured from centre to centre) was varied from 0.58 to 2.3, in steps of 0.58 (from 0.18 to 0.70 relative to the target eccentricity). The background was light grey (approximately cd/m 2 ). In the same-colour condition, all letters were black. In the diverent-colour condition, the target letter was red; the Xankers were black. In the coloured-blob condition, all letters were black, but the background at the location of the target was changed from grey to equiluminant yellow. The yellow blob had a circular Gaussian prowle with sigma of The observer Wxated a small black cross, which was permanently present on the screen. He/she initiated a new trial by clicking the computer mouse on the nexttrial button. After a delay of 0 ms, a conwguration of Wve letters was exposed for ms to either the left or right of the Wxation point. The observer had to identify the central (target) letter and insert his /her answer using Fig. 1 Examples of stimuli used in this study. a Ordinary crowding stimuli. In diverent conditions, target and Xanker letters could be coloured diverently. b Coloured blob stimuli are illustrated by diverent grey levels (actually, coloured blobs were equiluminant with the background) the keyboard. There was no feedback about the correctness of responses. Colour condition and target Xanker distance were held constant within blocks of 50 trials. The order of blocks with diverent conditions was counterbalanced. Seven observers (two male, Wve female; age 30, one 56) with normal or corrected-to-normal vision took part in the experiment. One observer (the author) had previous experience in similar experiments; others had very little or no experience with this type of tasks. The observers were paid for their participation. Each of them ran 1,0 trials ( trials per data point). Figure 2 shows the percentage of correct responses as a function of target Xanker distance for the three colour conditions. Both coloured target and coloured blob decrease the evect of crowding. An ANOVA with observer as random factor conwrmed signiwcant evects of target Xanker distance [F(3, 18) = 61.5, P < 0.001] and colour condition [F(2, 12) = 9.1, P < 0.005]. Post hoc comparisons revealed that all three colour conditions divered from each other (P < 0.05). Thus, the red letter is more evective than the yellow blob in reducing the crowding evect. This result may indicate that with coloured letters, there is feature-based interaction in addition to salience-based facilitation. Alternatively, the red letters may simply be more salient than the yellow blob and therefore more evective in directing attention to the target location.

3 643 Target-flanker distance (eccentricity units) Fig. 2 Results of Experiment 1. Percentage of correctly recognized targets, as a function of target Xanker distance, for three colour conditions: same colour (target and Xankers black), diverent colour (target red, Xankers black), and coloured blob (yellow blob at the target location) Experiment 2 In this experiment, an attempt was made to Wnd more evective blob stimuli. Here, target location was indicated by a saturated red blob, which should be more salient than the yellow blob used in the Wrst experiment. were identical to those of Experiment 1 except for the following diverences. The background was green with luminance of cd/m 2. The Gaussian blob at the target location was equiluminant red. All letters (target and Xankers) were white (approximately cd/m 2 ). Two conditions were run: one with the red blob at the target location, and another without it (all letters on the green background). Six new (except the author) observers (4 male, 2 female; age 56) participated in this experiment with 0 trials (0 for both conditions). Same colour Different colour Coloured blob The results are depicted in Fig. 3. The evect of the red blob was larger than the evect of the yellow blob and comparable with the evect of coloured letters in Experiment 1. [The ANOVA conwrmed the signiwcant evect of target Xanker distance (F(3, 15) = 41.7, P < 0.001) and blob condition (F(1, 5) = 17.6, P < 0.01)]. The evect of blob was also more consistent across observers, as compared with Experiment 1. This experiment shows that a coloured blob can be as eycient as a diverent colour of the target letter. The results of this experiment cannot be explained by a model that assumes interaction of feature detectors only (because in both conditions, all features of the letters themselves were identical), and are consistent with the Fig. 3 Results of Experiment 2. Percentage of correctly recognized targets, as a function of target Xanker distance, with and without red blob at the target location account of salience-based attention that facilitates the processing at the location of a colour singleton. It is parsimonious to assume the same mechanism also for the explanation of reduced crowding-evect with diverently coloured target and Xankers. Also, it is hard to explain the reduction of the crowding-evect in the coloured blob condition by non-spatial colour-based selection (because target and Xanker letters had the same colour). Experiment 3 In the Wrst two experiments, a reduction in crowding was observed when either the target or its local background had a unique colour, which supposedly attracted attention to the target. It would be important to know the roles of bottom-up and top-down mechanisms in this facilitation. In the third experiment, the colours varied unpredictably, which removed the top-down evect of knowledge of the target (and Xankers ) colour. Target-flanker distance (eccentricity units) were similar to those used in the conditions with red and black letters in Experiment 1. However, the colours of target and Xankers were selected randomly for each trial. Both the target and the Xankers could be independently either red or black (on any trial, the four Xankers had the same colour). Five observers (1 male, 4 female; age 56) from those who took part the Experiment 1 participated in this experiment; each of them ran 0 trials (in blocks of 50 trials, with constant target Xanker distance). No blob Red blob The results are shown in Fig. 4. It seems that even in the random condition diverent-colour targets can be identiwed

4 644 T black, F black T red, F black T black, F red T red, F red Same colour Different colour Target-flanker distance (eccentricity units) Fig. 4 Results of Experiment 3. Percentage of correctly recognized targets, as a function of target Xanker distance, for diverent combinations of target (T) and Xanker (F) colours more eyciently. An ANOVA conwrmed signiwcant evect of target Xanker distance [F(3, 12) = 36.4, P < 0.001] and colour condition (same-diverent) [F(1, 4) = 19.4, P < 0.05]; the evect of actual target colour (red black) was not signiwcant. Several studies have reported that attention to feature singletons can be modiwed by the properties of the stimuli in previous trials (priming or automatic top-down mechanism) (e.g. Maljkovic & Nakayama, 1994). In order to test for this evect here, trials were broken down according to the number of preceding trials with the same assignment of colours to the target and Xankers. The analysis revealed that colour speciwcations of preceding trials had no evect when target and Xankers were of the same colour. However, targets that had colours diverent from their Xankers were identiwed more accurately when the same mapping of colours occurred in one or two preceding trials. An ANOVA conwrmed a signiwcant linear evect of the number of preceding trials with identical colour mapping [F(1, 3987) = 4.3, P < 0.05], and interaction with colour condition (same vs. diverent) of the current trial [F(1, 3987) = 6.2, P < 0.05]. These results are depicted in Fig. 5. We can see that the advantage of the identiwcation of diverent-colour targets accumulates when the colours of target and Xankers remain constant during three trials. This pattern is similar to that found by Maljkovic and Nakayama (1994) in visual search, and suggests that the same (or similar) attention deployment mechanism is functioning in the present crowding experiments. General discussion The Wrst two experiments of this study show that nearly equal facilitation of crowded letters can be observed with diverently coloured targets and Xankers, and with a diverently coloured background blob in the location of Number of preceding trials with identical colours Fig. 5 EVect of colour assignment on preceding trials on target recognition, when the target colour on the current trial was the same as or diverent from that of the Xankers the target. The most reasonable conclusion seems to be that exogenously controlled attention is attracted to the location of a salient colour singleton (either a target itself or a coloured blob), and facilitates visual processing in that location. The third experiment shows that this facilitation of singleton processing functions mostly in bottom-up fashion, even without knowledge of the colours of forthcoming target and Xankers. However, facilitation accumulates in the course of several trials with the same colour mapping. This suggests a close relationship with the automatic attention guidance mechanism studied by Kristjansson, Mackeben, and Nakayama (01) and Wolfe, Butcher, Lee, and Hyle (03), and others. Although this mechanism is automatic and independent of the observer s knowledge, it is not strictly bottom-up, because it is avected by an implicit memory trace that originates from previous experience. Some studies have shown that the evect of irrelevant Xankers on performance can be avected by grouping factors that are even more complex than just feature-based pop-out of the target (Banks & White, 1984; Baylis & Driver, 1992). At present, it is hard to say whether those results can be explained by the same or similar mechanism of automatic allocation of attention as used here. If the evect of feature diverence between target and Xankers in crowding displays is mediated by exogenous attention, other forms of exogenous manipulation of attention should also reduce the amount of crowding. At present, empirical evidence is inconsistent: some studies have found spatial pre-cueing evects with crowding stimuli (Huckauf & Heller, 03; Morgan, Ward, & Castet, 1998) while others have reported no such evects (Nazir, 1992; Wilkinson, Wilson, & Ellemberg, 1997). Intriligator and Cavanagh (01) have suggested that crowding is a consequence of insuycient spatial resolution of attention, which is determined by the large receptive Welds at some higher level of visual processing. The present study suggests that salience-based exogenously controlled attention can reduce the crowding evect with colour singleton targets. The mechanisms of saliencebased modulation are supposedly located at relatively

5 645 low levels of the visual system, possibly at V1 (Li, 00; Nothdurft, 1997). Thus, it is possible that several kinds of attention at diverent levels of processing play a role in determining the observed extent of crowding. Acknowledgments I would like to thank Talis Bachmann, one anonymous reviewer, and especially Dirk Vorberg for suggestions that helped to improve this article. References Andriessen, J. J., & Bouma, H. (1976). Eccentric vision: Adverse interactions between line segments. Vision Research 16, Banks, W. P., & White, H. (1984). Lateral interference and perceptual grouping in visual detection. Perception & Psychophysics, 36, Baylis, G. C., & Driver, J. (1992). Visual parsing and response competition: The evect of grouping factors. Perception & Psychophysics, 51, Bjork, E. L., & Murray, J. T. (1977). On the nature of input channels in visual processing. Psychological Review, 84, Blakemore, C., & Tobin, E. A. (1972). Lateral inhibition between orientation detectors in the cat s visual cortex. Experimental Brain Research, 15, Bouma, H. (1970). Interaction evects in parafoveal letter recognition. Nature, 226, Cannon, M. W., & Fullenkamp S.C. (1991). Spatial interactions in apparent contrast: inhibitory evects among grating patterns of diverent spatial frequencies, spatial positions and orientations. Vision Research, 31, Cavanaugh, J. R., Bair, W., & Movshon, J. A. (02). Selectivity and spatial distribution of signals from the receptive Weld surround in macaque V1 neurons. Journal of Neurophysiology, 88, He, S., Cavanagh, P., & Intriligator, J. (1996). Attentional resolution and the locus of visual awareness. Nature, 383, Huckauf, A., & Heller, D. (03). Spatial selection in peripheral letter recognition: In search of boundary conditions. Acta Psychologica, 111, Intriligator, J., & Cavanagh, P. (01). The spatial resolution of visual attention. Cognitive Psychology, 43, Itti, L., & Koch, C. (00). A saliency-based search mechanism for overt and covert shifts of visual attention. Vision Research,, Koch, C., & Ullman, S. (1985). Shifts in selective visual attention: Towards the underlining neural circuitry. Human Neurobiology, 4, Kooi, F. L., Toet, A., Tripathy, S. P., & Levi, D. (1994). The evect of similarity and duration on spatial interaction in peripheral vision. Spatial Vision, 8, Kristjansson, A., Mackeben, M., & Nakayama, K. (01). Rapid, object-based learning in the deployment of transient attention. Perception, 30, Levi, D. M., Hariharan, S. & Klein, S. A. (02). Suppressive and facilitatory spatial interactions in peripheral vision: Peripheral crowding is neither size invariant nor simple contrast masking. Journal of Vision 2, Li, Z. (00). Pre-attentive segmentation in the primary visual cortex. Spatial Vision, 13, Maljkovic, V. & Nakayama, K. (1994). Priming of pop-out: I. Role of features. Memory & Cognition, 22, Morgan, M. J., Ward, R. M., & Castet, E. (1998). Visual search for tilted target: Tests of spatial uncertainty models. The Quarterly Journal of Experimental Psychology 51A, Nazir, T. A. (1992). EVects of lateral masking and spatial precueing on gap-resolution in central and peripheral vision. Vision Research, 32, Nothdurft H. C. (1997). DiVerent approaches to the coding of visual segmentation. In M. Jenkins, & L. Harris, Computational and biological mechanisms of visual coding (pp. 43). New York: Cambridge University Press. Nothdurft, H. C. (00). Salience from feature contrast: additivity across dimensions. Vision Research,, Parkes, L., Lund, J., Angelucci, A., Solomon, J. A., & Morgan, M. (01). Compulsory averaging of crowded orientation signals in human vision. Nature Neuroscience, 4, Pelli, D. G., Palomares, M. and Majaj, N. J. (04). Crowding is unlike ordinary masking: Distinguishing feature detection and integration. Journal of Vision 4, Petrov, Y., Carandini, M., & McKee, S. (05). Two distinct mechanisms of suppression in human vision. Journal of Neuroscience, 25, Styles, E. A., & Allport, D. A. (1986). Perceptual integration of identity, location and colour. Psychological Research, 48, Toet, A., & Levi, D. M. (1992). The two-dimensional shape of spatial interaction zones in the parafovea. Vision Research, 32, Wilkinson, F., Wilson, H. R., & Ellemberg, D. (1997). Lateral interactions in peripherally viewed texture arrays. Journal of the Optical Society of America A, 14, Wolfe, J. M., Butcher, S. J., Lee, C., & Hyle, M. (03). Changing your mind: On the contributions of top-down and bottom-up guidance in visual search for feature singletons. Journal of Experimental Psychology: Human Perception and Performance, 29, Wolford, G., & Chambers, L. (1983). Lateral masking as a function of spacing. Perception & Psychophysics, 33, Wolford, G., & Chambers, L. (1984). Contour interaction as a function of retinal eccentricity. Perception & Psychophysics, 36,