Abstract

Most investigations of visual search have focused on the discrimination between a search target and other task-irrelevant distractor objects (selection). The attentional limitations that arise when multiple target objects in the same display have to be processed simultaneously (access) remain poorly understood. Here, we employed behavioral and electrophysiological measures to investigate the factors that determine whether multiple target objects can be accessed in parallel. Performance and N2pc components were measured for search displays that contained either a single target or two target objects. When two target objects were present, they either had the same or different target-defining features. Participants reported whether search displays contained a single target, two targets with shared features, or two targets with different features. There were performance costs as well as reduced N2pc amplitudes for two-target/different relative to two-target/same displays, suggesting that access to multiple target objects defined by different features was impaired. These behavioral and electrophysiological costs were also observed in a task where all search display objects were physically different, but not during color or shape singleton search, confirming that they do not reflect a low-level perceptual grouping of physically identical targets. These results demonstrate strong feature-specific limitations of visual access, as proposed by the Boolean map theory of visual attention. They suggest that multiple target objects can be accessed in parallel only when they share task-relevant features and demonstrate that mechanisms of visual access can be studied with electrophysiological markers.

INTRODUCTION

Visual environments usually contain one or several objects that are important for thought and behavior, among others that can be safely ignored. In such situations, visual attention has to perform two tasks. First, it has to distinguish potentially task-relevant objects from other currently irrelevant objects. The ability to detect possible target objects and to separate them from distractors that are simultaneously present in a visual scene is usually referred to as “selection.” Selection is a core function of top–down attentional control, and a large number of visual search studies have investigated the factors that are involved in such target selection processes. In such experiments, participants have to report the presence or absence of a particular target object under conditions where the separation of targets and distractors is either easy or more challenging. The difficulty of target selection is manipulated by contrasting feature and conjunction search (e.g., Treisman & Gelade, 1980), by varying target–distractor similarity (e.g., Duncan & Humphreys, 1989) or the number of possible target-defining features (e.g., single-color vs. multiple-color search; Irons, Folk, & Remington, 2012). Current theories of visual search, such as the Guided Search model (e.g., Wolfe, 2007), have the explicit goal to identify the mechanisms that are responsible for differences in the efficiency of target selection processes.

However, there is another equally important function of visual attention that has been investigated much less intensively. Once target objects have been distinguished from irrelevant distractors, they have to be made accessible to higher level cognitive processes responsible for object recognition, conscious awareness, and action control. This second attentional function has been described as attentional engagement (e.g., Zivony & Lamy, 2016; Posner & Petersen, 1990) or visual access (Huang & Pashler, 2007). These different labels tend to be linked to different aspects of the attentional processing of target objects, with “engagement” emphasizing the processing of target features and their integration into object representations and “access” the conscious detection and recognition of target objects. Here, we will use the term “visual access” to refer to this attentional function. This selection–access distinction is often obscured in the attention literature, where generic references to attentional selection processes can refer to the discrimination of targets and distractors, the subsequent in-depth processing of target objects, or both of these mechanisms (but see Duncan, 2006; Desimone & Duncan, 1995, for an exception). In fact, it is relatively straightforward to distinguish experimental studies of selection and visual access. To investigate limitations in visual access, a critical manipulation concerns the number of target objects that are simultaneously present in a visual display. Experiments that manipulated this factor (e.g., Huang, Treisman, & Pashler, 2007; Duncan, 1980) found strong performance impairments when multiple target objects had to be processed concurrently, even when these targets could be easily distinguished from distractors (i.e., the demands on selection were low). In contrast, studies of visual selection typically employ search tasks where only a single target object can be present in individual displays, thereby minimizing the demands on visual access.

The fact that there are substantial performance costs when multiple target objects are presented simultaneously demonstrates that visual access is subject to severe capacity limitations. However, the factors that are responsible for these limitations have so far not been identified. A series of experiments by Huang et al. (2007) suggested that visual access might be specifically impaired when simultaneously presented multiple target objects are defined by different visual features (e.g., by different colors or shapes). In these experiments, match-to-sample tasks were employed where briefly presented and then masked sample objects (e.g., a blue and a red square) either appeared in the same sample display or in two successive displays. Participants had to match one of these sample colors to the color of a subsequent probe object. Accuracy was strongly impaired when the two targets were presented simultaneously as compared with successively, indicating that it is difficult to access two different colors of two different target objects at the same time. Importantly, no such simultaneous presentation costs were observed in another task where object locations had to be matched, which suggests that this limitation was present only for visual access to nonspatial features of objects, but not their locations.

Based on these observations, Huang and Pashler (2007) proposed a “Boolean map” account as a general theory of visual access. According to this theory, visual access is mediated by representations (Boolean maps), which are generated on the basis of particular feature values (e.g., “red”). In these Boolean maps, visual input is segregated into selected regions (e.g., all red objects in a display) and nonselected regions (all other regions of the visual field). Selected regions can be spatially discontinuous (i.e., contain two or more nonadjacent objects) and can be associated with feature labels (e.g., “red” or “square”). According to Huang and Pashler (2007), the content of a currently active Boolean map determines which visual objects can be visually accessed at any given instant. Because any given Boolean map can only contain a single feature label, it is not possible to simultaneously access two or more target objects in a search display when these objects are defined by different features (e.g., a red and a green target). In such displays, these objects have to be accessed sequentially (see Huang & Pashler, 2012; Huang, 2010, for further details).

Although the Boolean map theory aims to provide a comprehensive perspective on the access functions of visual attention, its impact on research in this field has so far been limited. This may be due to the fact that Boolean maps are described in terms of the properties of abstract representational structures, which makes it challenging to link them to other more neuroscientifically inspired accounts of visual attention and visual search, such as the biased competition framework (e.g., Desimone & Duncan, 1995) or the Guided Search model (e.g., Wolfe, 2007). In addition, many predictions of the Boolean map theory refer to the phenomenology of visual perception (see Huang & Pashler, 2007, for examples) and therefore hard to test with standard experimental paradigms commonly used in attention research. However, the single-feature access principle that is at the heart of the Boolean map theory provides a clear and readily testable hypothesis about the source of the capacity limitations for visual access. According to this principle, access to multiple target objects in the same display is impaired when these targets are defined by different features. In contrast, parallel access to multiple target objects should be possible when these objects share the same target-defining feature. The goal of this study was to test these predictions.

We studied visual access to target objects in search displays that contained either one or two of these objects. In two-target search displays, the target objects could either be defined by two different features (e.g., one red and one green object) or by a single shared feature (e.g., two red objects). The single-feature access principle predicts that access to multiple targets should be impaired in the former but not in the latter case. To investigate this claim, we measured visual search performance and also recorded ERP correlates of attentional target processing based on EEG waveforms that were recorded continuously while participants performed the search task. Visual access is assumed to take place during a brief “attentional episode” (e.g., Wyble, Potter, Bowman, & Nieuwenstein, 2011) that is initiated once a search display has been encountered and one or more target objects have been located among distractors. These access processes should in principle be measurable with brain activity markers that provide the required temporal resolution. One possible ERP marker of visual access is the N2pc component, which has been employed in many previous visual search experiments. This ERP component typically emerges between 180 and 200 msec poststimulus at posterior electrodes contralateral to the side of target objects that are presented together with distractors in visual search displays (e.g., Woodman & Luck, 1999; Eimer, 1996; Luck & Hillyard, 1994a, 1994b; see Luck, 2012, for a review). The N2pc is assumed to be generated in extrastriate ventral visual cortex (Hopf et al., 2000) and to reflect the enhancement of neural responses for objects with target-defining features at specific locations within visual cortical maps. It is elicited in search tasks where targets are defined by a specific feature and also during search for targets that are defined by feature conjunctions (e.g., Berggren & Eimer, 2018; Fuggetta, Pavone, Walsh, Kiss, & Eimer, 2006; Luck & Ford, 1998; Luck & Hillyard, 1994a).

Although the N2pc is usually regarded as an ERP marker of attentional target selection (Luck, 2012; see also Eimer, 2014, 2015, for further discussion), this interpretation is complicated by the fact that the selection and access functions of visual attention are often not clearly distinguished. Thus, the question remains whether the N2pc is exclusively associated with the discrimination between target and distractor objects (selection), with the subsequent in-depth processing of selected target objects (access), or with both of these processes. Previous N2pc studies of conjunction search (e.g., Berggren & Eimer, 2018) have shown that the early versus late phase of target N2pc components can reflect different aspects of attentional control (feature-based vs. object-based guidance). Such observations suggest that this component may not be linked exclusively to a single stage of attentional processing (see also Zivony, Allon, Luria, & Lamy, 2018, for recent evidence that the N2pc can be associated with the attentional engagement to target objects).

To test more directly whether the N2pc can be employed as a marker of visual access, this component needs to be measured in search tasks where the number of target objects in the same display is manipulated. This was done in several recent studies (e.g., Mazza, Pagano, & Caramazza, 2013; Mazza & Caramazza, 2011, 2012; Pagano & Mazza, 2012; Drew & Vogel, 2008), where between one and four color-defined targets were presented in the same hemifield. N2pc amplitudes increased with the number of target objects in the same display, and this was interpreted as evidence that the N2pc reflects an object individuation mechanism that operates by binding target-defining features to particular locations (e.g., Mazza et al., 2013). Such links between N2pc amplitudes and the number of targets were only found in tasks where multiple target individuation was required (e.g., when the exact number of targets in a display had to be reported), but not when observers simply had to decide whether a search display contained at least one color-defined target or only distractors (Mazza & Caramazza, 2011), suggesting that the underlying processes are goal-sensitive rather than mandatory. It is clear that the object individuation mechanism postulated by Mazza and colleagues reflects a relatively late stage of attentional processing that follows the initial discrimination between target and distractor objects (selection), but the question remains whether this mechanism is similar or identical to the visual access or attentional engagement processes proposed by others (Zivony et al., 2018; Huang & Pashler, 2007).

The present experiments were designed to investigate links between visual access and the N2pc component more directly and to specifically test the single-feature access principle postulated by the Boolean map theory. In Experiment 1, circular search displays contained either one or two color-defined target objects among multiple distractor objects in different nontarget colors (see Figure 1, top). Participants searched for one of two possible colored targets (two-color search; see Grubert & Eimer, 2016; Irons et al., 2012). There were single-target displays, which included one target object that was equally likely to have either of these colors, and two-target displays with two target objects. In half of these displays, both target objects shared the same color (e.g., both red; two-target/same displays). In the other half, they had two different colors (e.g., red and green; two-target/different displays). Both target objects always appeared on the same side of a search display, but never at adjacent positions. Participants performed a three-alternative choice task. They reported on each trial whether the search display contained a single target, two identical color targets, or two different color targets.

Figure 1. 

Examples of search displays shown across experiments (not to scale). Top: In Experiments 1 and 2A, search displays included eight squares. One or two of these were targets defined by their color. The examples show a two-target/same search display with two red targets and a two-target/different display with a red and a green target. Middle: In Experiments 2B and 2C, targets were color or shape singletons among uniform distractors (gray squares). The examples show two-target/different search displays. Bottom: In Experiment 3, search displays contained six colored shapes, with targets defined by two possible colors and two possible shapes (e.g., red, green, circle, square). The examples on the upper row show two-target/match search displays, containing two red targets or two circle targets. The examples on the lower row show two-target/mismatch search displays. Here, two targets could contain mismatching target features from the same feature dimension (e.g., one red and one green target) or mismatching target features from different dimensions (e.g., one red and one circle target).

Figure 1. 

Examples of search displays shown across experiments (not to scale). Top: In Experiments 1 and 2A, search displays included eight squares. One or two of these were targets defined by their color. The examples show a two-target/same search display with two red targets and a two-target/different display with a red and a green target. Middle: In Experiments 2B and 2C, targets were color or shape singletons among uniform distractors (gray squares). The examples show two-target/different search displays. Bottom: In Experiment 3, search displays contained six colored shapes, with targets defined by two possible colors and two possible shapes (e.g., red, green, circle, square). The examples on the upper row show two-target/match search displays, containing two red targets or two circle targets. The examples on the lower row show two-target/mismatch search displays. Here, two targets could contain mismatching target features from the same feature dimension (e.g., one red and one green target) or mismatching target features from different dimensions (e.g., one red and one circle target).

If only a single feature from any given dimension can be accessed at any moment, simultaneous access to two target objects in the same display should be possible when they share the same target-defining feature but should be impaired when their features are different. This should result in behavioral costs for these displays relative to two-target/same displays. Importantly, if the N2pc component was sensitive to the number of objects that can be simultaneously accessed, there should also be systematic N2pc amplitude differences between the different types of search displays in Experiment 1. If both target objects can be accessed simultaneously in two-target/same displays, N2pc components elicited by these displays should be larger than N2pcs triggered by single-target displays, reflecting access to two versus just one target object. Such a result would be in line with previous N2pc studies that manipulated the number of same-color target objects in the same display (e.g., Mazza & Caramazza, 2011; Drew & Vogel, 2008). The critical question concerns the N2pc components elicited by two-target/different displays. If access to both targets is impaired in these displays, N2pc components should be reduced in size relative to two-target/same displays. In fact, the Boolean map theory makes the specific prediction that only one of the two targets should be accessible in two-target/different displays. If this was the case, N2pc components for these displays should be identical in size to the N2pc elicited by single-target displays, reflecting access to a single target object in both cases. An alternative possibility is that N2pc components elicited in response to visual search displays generally do not reflect visual access to targets but are instead generated by earlier selection mechanisms that dissociate between target and distractor objects (e.g., by inhibiting distractors; e.g., Luck & Hillyard, 1994b). In this case, there should be no systematic N2pc amplitude differences between two-target/same and two-target/different displays.

EXPERIMENT 1

Methods

Participants

Thirteen individuals took part in Experiment 1. One participant was excluded from analysis due to chance-level accuracy. Of the final sample of 12 participants (M age = 34 years, SD = 7; six men; all right-handed), all reported normal or corrected-to-normal vision.

Stimuli and Procedure

The experimental task was programmed and executed using E-Prime 2.0 software (Psychology Software Tools, Inc.). Stimuli were presented on a 24-in. BenQ monitor (60 Hz; 1920 × 1080 screen resolution) attached to a SilverStone PC, with viewing distance at approximately 90 cm. Manual responses were registered via standard keyboard button response. Stimuli were shown on a black background. A gray fixation cross (0.25 × 0.25 degrees of visual angle) was present constantly throughout each experimental block. Each search display contained eight colored squares (0.95 × 0.95°) presented equidistant from fixation at an eccentricity of 1.78°, measured from fixation to the center of each object. These stimuli were arranged around an imaginary circle, with four items in the left and four in the right visual hemifield (see Figure 1, top). Squares could appear in the following six colors: red (CIE coordinates: .605/.322), orange (.543/.409), yellow (.405/.470), green (.296/.604), blue (.169/.152), magenta (.270/.134), or gray (.305/.325). All colors were matched for luminance (14 cd/m2).

Participants performed a two-color search task where targets were defined by one of two possible colors (e.g., red and green). These two target colors were specified at the start of the experiment, remained constant for each individual participant, and were rotated between participants. Gray was never used as a target color for any participant. There were three different types of search displays. Single-target displays contained one square in one of the two possible target colors. Two-target displays included two-target color squares. In half of these displays, these two targets had the same color (e.g., two red items; two-target/same displays). In the other half, the two targets had two different colors (e.g., one red, one green; two-target/different displays). In these two-target displays, both target objects always appeared at two randomly selected positions in the same hemifield, with the restriction that they were never presented at immediately adjacent locations (to avoid color grouping effects for two-target/same displays). For all search displays, the colors of the distractor items were selected randomly from the remaining nontarget colors, with the restriction that each display always contained one pair of color-matching items on either side that never appeared at immediately adjacent locations. In single-target and two-target/different displays, two distractor squares on the left and right side matched in their color, and the matching colors on both sides always differed. In two-target/same displays, two color-matching targets on one side were accompanied by two distractors in the same nontarget color on the opposite side. This was done to ensure that participants could not infer the presence of two target objects from the presentation of two color-matching items in the displays. Search displays were presented for 100 msec, followed by a 1400-msec intertrial interval before the next search display onset that also acted as the response window. A 1000-msec blank interval was presented at the start of each block before the onset of the first search display.

Participants' task was to report whether each search display contained a single target, two identical targets, or two different targets by pressing the “1,” “2,” or “3” keys on the numeric keypad with their right index, middle, or ring finger. The single-target response was always “1,” whereas two-target/same and two-target/different responses were counterbalanced within-subject halfway through the experiment. Following practice, participants completed six experimental blocks of 72 trials each. Single-target, two-target/same, and two-target/different displays were equiprobable, and the target(s) were equally likely to be presented on the left or right side.

EEG Data Recording and Analysis

EEG was DC recorded at 27 scalp electrodes, mounted on an elastic cap at sites Fpz, F7, F8, F3, F4, Fz, FC5, FC6, T7, T8, C3, C4, Cz, CP5, CP6, P9, P10, P7, P8, P3, P4, Pz, PO7, PO8, PO9, PO10, and Oz. A 500-Hz sample rate was utilized, and a 40-Hz low-pass filter was applied during recording. An electrode attached to the left earlobe was used as an online reference, with the average of both earlobes used for offline re-referencing. No other filters were applied offline. Trials with eye blinks (exceeding ±60 μV at Fpz), horizontal eye movements (exceeding ±30 μV in the horizontal EOG channels), and muscle movement artifacts (exceeding ±80 μV at all other channels) were rejected from EEG analyses, as were trials where an incorrect response was recorded. All other trials were segmented into epochs for each search display ranging from −100 to +500 msec relative to search display onset. EEG epochs were averaged relative to a 100-msec prestimulus baseline, computed separately for all three search display types, and for displays where target(s) appeared in the left or right visual field. N2pc components were quantified based on ERP mean amplitudes obtained at lateral posterior electrode sites PO7/PO8 between 200 and 300 msec poststimulus onset.

Results

Behavioral Performance

RTs on trials with correct responses were fastest for two-target/same displays (M = 754 msec), slowest for two-target/different displays (M = 813 msec), and intermediate for single-target displays (M = 792 msec). A one-way ANOVA showed a significant main effect of Display Type, F(2, 22) = 6.74, p = .005, ηp2 = .38. A paired t test confirmed that RTs to two-target displays were reliably faster when these displays contained two identical as compared with two different color targets, t(11) = 6.12, p < .001. This difference is shown in Figure 2 (top left). The RT differences between single-target and two-target/same displays and between single-target and two-target/different displays were not reliable (t(11) = 1.91 and 1.20, p = .083 and .25, respectively). Error rates also differed between the three search display types, F(2, 22) = 7.91, p < .005, ηp2 = .42. As shown in Figure 2 (bottom left), errors on two-target trials were more frequent when target colors differed than when they were identical (M = 19% vs. 9%), t(11) = 3.65, p = .004. For single-target displays, errors were recorded on 11% of all trials. This error rate was lower than for two-target/different displays, t(11) = 2.59, p = .03, but did not differ from two-target/same displays, t(11) = 1.14, p = .28. On trials with two-target/different displays, participants were more likely to incorrectly report the presence of a single color target than the presence of two identical color targets (55% vs. 40% of all error trials; 5% were no-response trials). For two-target/same displays, reports of single targets were less frequent than reports of two different targets (31% vs. 64%).

Figure 2. 

RTs (top) and error rates (bottom) observed in Experiments 1, 2A, 2B, 2C, and 3, for two-target displays with identical/feature-matching targets versus two-target displays with different/feature-mismatching targets.

Figure 2. 

RTs (top) and error rates (bottom) observed in Experiments 1, 2A, 2B, 2C, and 3, for two-target displays with identical/feature-matching targets versus two-target displays with different/feature-mismatching targets.

N2pc Components

Figure 3 (top) shows ERPs elicited by the three different search display types at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s) in the 350-msec interval following search display onset, together with the corresponding contralateral–ipsilateral difference waveforms. Clear N2pcs were triggered by all search displays. Critically, N2pc amplitudes elicited by two-target/different displays were considerably smaller than N2pcs to two-target/same displays and similar in size to N2pcs triggered by single-target displays.

Figure 3. 

Top: Grand-averaged ERPs obtained in Experiment 1 in the 350-msec interval following search display onset at posterior electrode sites PO7/PO8 contralateral and ipsilateral to the side of the target object(s) in the search displays. ERPs are shown separately for single-target displays, two-target/same displays, and two-target/different displays. The corresponding difference waveforms computed by subtracting ipsilateral from contralateral ERPs are shown on the right. Bottom: Contralateral and ipsilateral ERPs and corresponding difference waves measured for these three types of search displays in Experiment 2A.

Figure 3. 

Top: Grand-averaged ERPs obtained in Experiment 1 in the 350-msec interval following search display onset at posterior electrode sites PO7/PO8 contralateral and ipsilateral to the side of the target object(s) in the search displays. ERPs are shown separately for single-target displays, two-target/same displays, and two-target/different displays. The corresponding difference waveforms computed by subtracting ipsilateral from contralateral ERPs are shown on the right. Bottom: Contralateral and ipsilateral ERPs and corresponding difference waves measured for these three types of search displays in Experiment 2A.

Mean amplitude data obtained 200–300 msec poststimulus onset were entered into a 3 × 2 repeated-measures ANOVA with the factors Display Type (single-target, two-target/same, two-target/different) and Laterality (contralateral, ipsilateral). A significant main effect of Laterality, F(1, 11) = 84.04, p < .001, ηp2 = .88, confirmed the reliable presence of N2pc components. There was also a significant Condition × Laterality interaction, F(2, 22) = 14.93, p < .001, ηp2 = .58. Comparisons of contralateral versus ipsilateral ERPs showed that reliable N2pc components were elicited by all three search display types (ts > 6.99, ps < .001). To compare these N2pcs, difference values were computed by subtracting ipsilateral from contralateral ERPs. N2pc components elicited by two-target/same displays (M diff = −2.51 μV, SD = .91) were significantly larger relative to single-target displays (M diff = −1.50 μV, SD = .64; t(11) = 4.97, p < .001) and, crucially, also relative to two-target/different displays (M diff = −1.79 μV, SD = .89; t(11) = 4.67, p = .001). N2pc components triggered by two-target/different and single-target displays showed no reliable amplitude difference, t(11) = 1.39, p = .19.

Discussion of Experiment 1

The results of Experiment 1 show that access to two target objects in the same search display is impaired when these objects are defined by two different colors than when they share the same color. RTs were delayed for two-target/different as compared with two-target/same displays, and response errors were much more frequent on trials where search displays contained two different color targets (19% vs. 9%). Critically, these behavioral differences were accompanied by corresponding N2pc differences. N2pc amplitudes were reliably smaller for two-target/different displays relative to two-target/same displays, suggesting an impairment of visual access to both target objects in the former displays. According to the single-feature access principle proposed by the Boolean map theory, observers can access two search targets simultaneously when these targets share the same color. In line with this assumption, N2pc components were larger for two-target/same relative to single-target displays (confirming previous observations; e.g., Mazza & Caramazza, 2011; Drew & Vogel, 2008), which is likely to reflect parallel visual access to two targets versus just a single target object. The single-feature access principle also assumes that the behavioral costs for two-target/different displays are due to the fact that only a single target object can be accessed in these displays. In this case, the N2pc elicited should be identical to the N2pc triggered by single-target displays. Although there was a numerical trend for slightly enhanced N2pc amplitudes in response to two-target/different as compared with single-target displays, this difference was not significant, which suggests, in both types of search displays, only a single target object was accessed.

These ERP results suggest that the N2pc component is sensitive to visual access to target objects during visual search and to the number of targets that can be simultaneously accessed in a search display. The N2pc amplitude difference between two-target/same and two-target/different displays provides clear evidence that visual access to two targets in the same display is impaired when these targets are defined by two different colors, as suggested by the single-feature-access principle. However, an alternative interpretation of this amplitude difference in terms of relatively low-level visual grouping effects needs to be considered. Two physically identical target objects in the same hemifield may be perceptually grouped, whereas no such grouping is possible for pairs of different color targets, and larger N2pc amplitudes for two-target/same displays may be a direct result of such visual grouping processes. A previous study (Mazza & Caramazza, 2012) that measured N2pc components to multiple target objects in the same hemifield during an enumeration task also found larger N2pc amplitudes for displays where these targets shared the same color relative to displays where they differed in color and attributed this difference to the perceptual grouping of same-color targets. Because such visual grouping effects are likely to occur at relatively early stages of perceptual processing and presumably before visual access, it is critical to rule out the possibility that the N2pc differences observed in Experiment 1 between two-target/same and two-target/different displays primarily reflect perceptual grouping. In contrast to Mazza and Caramazza (2012), who presented all target objects next to each other in the same hemifield (thereby increasing the probability that same-color targets would be perceptually grouped), target pairs in the two-target displays of Experiment 1 never appeared at adjacent positions but were always separated by a nontarget color distractor. These displays also always included a pair of physically identical distractor objects on the other side. This was done to minimize any perceptual grouping of same-color targets, but further evidence that the critical N2pc amplitude differences observed in Experiment 1 were not due to such grouping effects is clearly needed. One goal of Experiments 2 and 3 was to provide such evidence.

EXPERIMENT 2

In Experiment 1, target objects appeared among multiple distractors in different nontarget colors and could therefore only be discriminated from distractors on the basis of their color. The behavioral and N2pc results observed in this experiment showed that visual access to two target objects in the same search display was strongly impaired when these targets differed in their color and suggest that only a single target object could be accessed in these displays. These conclusions should not be based the results of a single experiment and require confirmation in a replication experiment. This was done in Experiment 2A, which used the same stimulus parameters and analysis procedures as Experiment 1.

If the performance and N2pc differences observed in Experiment 1 between the three types of search displays reflect differences in visual access to color-defined target objects, a different pattern of effects should be observed in search tasks where access is not mediated by specific target features. To test this, Experiments 2B and 2C employed tasks where targets were salient feature singletons (e.g., the only colored objects among homogeneous gray distractors). Such singletons can be detected by registering local discontinuities in search displays, without the need to discriminate specific target features (singleton detection mode; Bacon & Egeth, 1994). The Boolean map theory assumes that, in such singleton search tasks, visual access does not operate in a feature-specific fashion but on the basis of bottom–up salience values (Huang & Pashler, 2007). Because visual access is not controlled by specific feature labels but operates in a purely location-based fashion in such tasks, observers should be able to simultaneously access two singleton target objects in the same display, regardless of whether these two objects are identical or differ in their features. If this is correct, there should be no behavioral or electrophysiological differences between two-target/same and two-target/different displays in singleton search tasks.

This was tested in Experiments 2B and 2C, separately for color singleton and shape singleton search. Search displays contained one or two target singletons among homogeneous distractors (see Figure 1). Targets were defined by one of two possible colors (Experiment 2B) or one of two possible shapes (Experiment 2C). In two-target/same displays, both targets were identical, whereas in two-target/different displays, they differed in their color or shape. Participants again had to report the presence of a single target, two identical targets, or two different targets. If observers can access multiple target objects irrespective of their features during singleton search, there should no longer be any behavioral and electrophysiological access costs for two-target/different displays. Therefore, the performance and N2pc differences between these displays and two-target/same displays that were observed in Experiment 1 search should be eliminated in Experiments 2B and 2C. In contrast and importantly, if the N2pc amplitude differences between these displays were due to the perceptual grouping of physically identical target objects, they should again be present in Experiments 2B and 2C.

Methods

Participants

Thirteen individuals took part in Experiment 2. Of these, one participant was excluded from analysis due to a high proportion of data contaminated by horizontal eye movements in all three subexperiments. Of the remaining 12 participants (M age = 28 years, SD = 8; four men; one left-handed), all reported normal or corrected-to-normal vision.

Stimuli and Procedure

These matched Experiment 1, with the following exceptions. Experiment 2A was a direct replication of Experiment 1, with the exception that the set of two target colors for each participant were chosen from three possible colors (red, blue, or green) and that this target color set was fully counterbalanced across participants. Experiment 2B (color singleton search) was identical to Experiment 2A, expect that all distractor squares were uniformly gray (see Figure 1, middle). The same set of three possible target colors was used, but for each participant, one of the two target colors used in Experiment 2A was replaced by the remaining third color. In Experiment 2C (shape singleton search; Figure 1, right middle), all search display items appeared in gray. Targets were now defined by their shape. For each participant, two target shapes were chosen from a set of three shapes (circle, triangle, or hexagon), and this was counterbalanced across participants. All distractor items were uniform gray squares. In all other respects, Experiment 2C was identical to Experiments 2A and 2B. The order in which Experiments 2A, 2B, and 2C were performed was randomized across participants.

EEG Data Recording and Analysis

These were identical to Experiment 1. For each subexperiment, ERPs were computed separately for single-target, two-target/same, and two-target/different displays with the target(s) in the left or right visual field. N2pc mean amplitudes were again computed within a 200–300 msec poststimulus time window.

Results

Experiment 2A (Replication of Experiment 1)

Behavioral performance.

As expected, results were very similar to Experiment 1 (see Figure 2). RTs on trials with correct responses were slower for two-target/different displays (M = 791 msec) than for two-target/same displays (M = 739 msec) and single-target displays (M = 725 msec), as reflected by resulting in a main effect of Display Type, F(2, 22) = 11.76, p < .001, ηp2 = .52. Paired t tests confirmed that this RT delay for two-target/different displays was reliable relative to two-target/same displays, t(11) = 6.24, p < .001, and also relative to single-target displays, t(11) = 3.88, p = .003. There was no significant RT difference between single-target and two-target/same displays (t < 1). Error rates also differed between display types, F(2, 22) = 8.33, p = .002, ηp2 = .43. Errors were more frequent in response to two-target/different displays (M = 15%) as compared with two-target same displays (M = 7%), t(11) = 3.04, p = .01, and also relative to single-target displays (M = 4%), t(11) = 3.05, p = .01. Error rates did not differ reliably between single-target and two-target/same displays, t(11) = 1.45, p = .17. On trials with two-target/different displays, participants were more likely to erroneously report the presence of a single target object than two identical targets (58% vs. 39% of all error trials). The opposite tendency was observed for errors in response to two-target/same displays (25% vs. 69%).

N2pc components.

Figure 3 (bottom) shows ERPs elicited by the three different search display types at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s) and the corresponding contralateral–ipsilateral difference waveforms. Results were generally similar to Experiment 1 (Figure 3, top). Crucially, N2pc components were again much larger for two-target/same as compared with two-target/different displays. The N2pcs for single-target displays was slightly attenuated relative to N2pcs to two-target/different displays. An analysis of N2pcs mean amplitudes obtained 200–300 msec poststimulus onset with the factors Display Type and Laterality revealed a significant main effect of Laterality, F(1, 11) = 35.01, p < .001, ηp2 = .76, and a reliable Condition × Laterality interaction, F(2, 22) = 13.55, p < .001, ηp2 = .55. All three types of search displays elicited reliable N2pc components (ts > 4.52, ps ≤ .001). N2pcs for two-target/same displays (M diff = −2.50 μV, SD = 1.66) were larger than N2pcs to single-target displays (M diff = −0.96 μV, SD = .73), t(11) = 4.19, p < .001, and also, critically, larger than N2pcs for two-target/different displays (M diff = −1.53 μV, SD = .78), t(11) = 4.67, p = .001. N2pc amplitudes to two-target/different displays were now reliably larger than N2pcs to single-target displays, t(11) = 3.15, p = .009.

Experiment 2B (Color Singleton Search)

Behavioral performance.

For RTs, an effect of Display Type was present, F(2, 22) = 45.91, p < .001, ηp2 = .81. RTs were faster for single-target displays (M = 534 msec) relative to two-target/same displays (M = 645 msec), t(11) = 6.76, p < .001, and two-target/different displays (M = 656 msec), t(11) = 7.53, p < .001. The RT difference between two-target/same and two-target/different displays was not significant, t(11) = 1.37, p = .20. For error rates, the same pattern was observed. There was an effect of Display Type, F(2, 22) = 11.33, p < .001, ηp2 = .51. Errors were less frequent for single-target displays (M = 1%) relative to both two-target/same displays (M = 6%), t(11) = 4.20, p = .001, and two-target/different displays (M = 5%), t(11) = 3.13, p = .01, which did not differ in terms of error rates, t(11) = 1.25, p = .24. In contrast to the feature-guided color search tasks (Experiments 1 and 2A), the majority of errors on trials with two-target/different displays were reports of two identical targets (87% of all error trials), whereas incorrect single-target responses were very infrequent (10%). This pattern of errors was identical to that found for two-target/same displays (86% vs. 10%).

N2pc components.

Figure 4 (top) shows contralateral and ipsilateral ERPs and the corresponding N2pc difference waveforms elicited in Experiment 2B for the three different search display types. N2pcs were smallest for single-target displays and slightly larger for two-target/same as compared with two-target/different displays. For N2pc mean amplitudes, a main effect of Laterality, F(1, 11) = 82.96, p < .001, ηp2 = .88, was accompanied by a Laterality × Display Type interaction, F(2, 22) = 21.63, p < .001, ηp2 = .66. Reliable N2pcs were present for all three display types (ts > 8.04, ps < .001). N2pc amplitudes were larger for two-target/same displays (M diff = −3.41 μV, SD = 1.40) and two-target/different displays (M diff = −2.80 μV, SD = 1.21) relative to single-target displays (M diff = −1.62 μV, SD = .68; both ts > 5.12, both ps < .001). Although the N2pc was numerically larger for two-target/same displays, the amplitude difference relative to two-target/different displays only approached statistical significance, t(11) = 1.93, p = .08.

Figure 4. 

Top: Grand-averaged ERPs obtained in Experiment 2B (color singleton search) at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s) in single-target, two-target/same, and two-target/different displays, shown together with the corresponding contralateral-ipsilateral difference waveforms. Bottom: Contralateral and ipsilateral ERPs and N2pc difference waves measured for these three types of search displays in Experiment 2C (shape singleton search).

Figure 4. 

Top: Grand-averaged ERPs obtained in Experiment 2B (color singleton search) at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s) in single-target, two-target/same, and two-target/different displays, shown together with the corresponding contralateral-ipsilateral difference waveforms. Bottom: Contralateral and ipsilateral ERPs and N2pc difference waves measured for these three types of search displays in Experiment 2C (shape singleton search).

Experiment 2C (Shape Singleton Search)

Behavioral performance.

An effect of Display Type for RTs, F(2, 22) = 46.23, p < .001, ηp2 = .81, again reflected faster responses for single-target displays (M = 619 msec) relative to two-target/same displays (M = 749 msec), t(11) = 7.14, p < .001, and two-target/different displays (M = 758 msec), t(11) = 7.57, p < .001. RTs for the two different types of two-target displays did not differ (t < 1). Analogous results were found for error rates. An effect of Display Type, F(2, 22) = 19.09, p < .001, ηp2 = .63, reflected the fact that participants made fewer errors for single-target displays (M = 2%) relative to both two-target/same displays (M = 8%), t(11) = 4.21, p = .001, and two-target/different displays (M = 11%), t(11) = 4.73, p < .001. The error rate difference between both two-target search displays was also reliable, t(11) = 3.34, p = .007. For two-target/different displays, participants reported two identical targets on 85% of all error trials and a single target on only 11% of these trials. A similar pattern of errors was found for two-target/same displays (79% vs. 16%).

N2pc components.

Figure 4 (bottom) shows contralateral and ipsilateral ERPs and the corresponding N2pc difference waveforms elicited in Experiment 2C for all search display types. N2pcs were much smaller for single-target relative to two-target displays and appeared virtually identical for two-target/same and two-target/different displays. A main effect of Laterality, F(1, 11) = 90.35, p < .001, ηp2 = .89, and a Laterality × Display Type interaction, F(2, 22) = 27.88, p < .001, ηp2 = .72, were present for N2pc mean amplitudes. Reliable N2pcs were elicited by all three display types (ts > 5.61, ps < .001). N2pc amplitudes were larger for two-target/same displays (M diff = −3.59 μV, SD = 1.34) and two-target/different displays (M diff = −3.41 μV, SD = 1.32) relative to single-target displays (M diff = −1.48 μV, SD = .91; both ts > 5.61, both ps < .001). Notably, there was no evidence for N2pc amplitude differences between two-target/same and two-target/different displays (t < 1).

Discussion of Experiment 2

The results of Experiment 2A confirmed key observations of Experiment 1. Again, performance was impaired for two-target/different as compared with two-target/same displays, and target N2pc components were reliably attenuated when search displays contained two different versus two identical targets. These findings again demonstrate that visual access to multiple target objects is impaired when these targets are defined by two different as compared with a single shared feature. However, and in contrast to Experiment 1, there was now also a small but reliable N2pc amplitude enhancement for two-target/different relative to single-target displays. This result is not entirely consistent with the single-feature access principle of the Boolean map theory, which predicts that only a single target object can be accessed in both displays. Its implications will be further considered in the General Discussion section.

A very different pattern of behavioral and N2pc results was observed in the color and shape singleton search tasks of Experiments 2B and 2C. Here, RTs did not differ between two-target/same and two-target/different search displays (see Figure 2), indicating that the two target objects could be accessed equally well in both types of displays. There were also no differences in error rates between these displays in the color singleton task (Experiment 2B) and only a small albeit reliable difference in the shape singleton task (Experiment 2C). These results strongly suggest that, in contrast to search for color-defined targets (Experiments 1 and 2A), both target objects could be accessed simultaneously during singleton search, regardless of whether their features were identical or not. This was further confirmed by the pattern of target N2pc components measured in these tasks. During shape singleton search, virtually identical N2pc components were triggered in response to two-target/same and two-target/different displays. During color singleton search, there was a numerical tendency for a larger N2pc for two-target/same displays, but this difference only approached significance.

According to the Boolean map theory, the observation that both singleton target objects in Experiments 2B and 2C could be accessed simultaneously, regardless of whether their features were identical or not, implies that access was not based on feature labels but operated in an entirely location-based fashion. However, this cannot be entirely correct. Because participants had to report whether the singleton targets in two-target displays were identical or different, some information about their colors or shapes had to be available. This suggests an important modification to the single-feature access principle: Even though individual feature values may not have been explicitly represented in these experiments, the identity relationship between both targets (same vs. different) must still have been accessible in parallel. In other words, observers are able to directly access the presence of a task-relevant feature difference in search displays without being able to simultaneously access the exact values of these features. This apparent dissociation between visual access to individual feature labels and to the identity relationship between features needs to be investigated more systematically in future research.

Another goal of Experiment 2 was to rule out the possibility that the N2pc differences between two-target/same and two-target/different displays observed in Experiment 1 were not due to differences in visual access, but instead the result of a low-level visual grouping of two physically identical targets (as suggested by Mazza & Caramazza, 2012). No such N2pc amplitude differences were observed in the color and shape singleton search tasks of Experiments 2B and 2C, in spite of the fact that such perceptual grouping effects should have been equally if not more pronounced in color and shape singleton displays. This provides additional evidence that such N2pc amplitude modulations are indeed associated with differences in the efficiency of visual access processes. Further support for this conclusion will be provided by Experiment 3.

EXPERIMENT 3

In the color-guided search tasks of Experiments 1 and 2A, two-target/same displays always contained a pair of identical target objects. It is possible that the performance benefits and N2pc enhancements observed for these displays relative to two-target/different displays were a result of this fact. Multiple feature-defined target objects in the same display may be simultaneously accessible only when they are physically identical, but not when they differ in some of their features. The Boolean map theory does not assume that physical identity is necessary for multiple-object access. According to this theory, multiple targets can be accessed in parallel, as long as they share one feature in one dimension (e.g., color), even when they differ in another dimension (e.g., shape).

We tested this prediction in Experiment 3, where search displays contained six colored shapes (see Figure 1). As before, there were single-target and two-target displays. Target objects were now defined by four possible features from two dimensions (two colors and two shapes; e.g., red, green, circle, and square). Each target object only matched one of these four features (either a target-defining color or a target-defining shape) and had a task-irrelevant feature in the other dimension. When two targets were present in the same display, their task-irrelevant features were always different. Thus, and in contrast to Experiments 1 and 2A, these targets were never physically identical. The critical manipulation was whether target objects in the same display had a matching target-defining feature (e.g., red star and red pentagon; two-target/match displays) or not (e.g., red star and blue pentagon; two-target/mismatch displays). If two target objects can be accessed simultaneously when they share a feature, Experiment 3 should find behavioral and N2pc differences between two-target/match and two-target/mismatch displays that mirror those found between two-target/same and two-target/different displays in Experiments 1 and 2A.

The single-feature access principle of the Boolean map theory applies to features from the same dimension, such as color. It states that two targets defined by two different colors cannot be accessed simultaneously. This raises the question whether this is also the case when two targets are defined by features from different dimensions (e.g., a color-defined and a shape-defined target object). For example, the Guided Search model (Wolfe, 2007) postulates an asymmetry between the control of visual search within and across dimensions. According to this model, search can be guided in parallel by multiple features from different dimensions, but only by a single feature within each dimension. The Boolean map theory does not predict such an asymmetry for visual access to multiple target objects. According to Huang and Pashler (2007), Boolean maps can include feature labels for different dimensions, but only when these labels apply to the same object. When these features are distributed across different target objects, access to these objects requires the sequential creation of multiple Boolean maps, one for each feature, which can subsequently be combined via Boolean operations (intersection and union; see also Huang & Pashler, 2012). This implies that visual access to multiple target objects always has to be mediated by a single shared feature and that features from different dimensions cannot be accessed simultaneously.

We tested this hypothesis in Experiment 3 by contrasting two different types of two-target/mismatch displays. Half of these displays contained two targets that were defined within the same dimension (i.e., two different colors or shapes; same-dimension displays). The other displays included one shape-defined and one color-defined target (different-dimension displays). Performance and N2pc components were measured separately for these two types of displays to identify possible differences in visual access to different features within and across dimensions. If it was possible to simultaneously access two features from different dimensions (i.e., one color and one shape) even when they belong to different objects, but only one feature from the same dimension, there should be an access benefit for different-dimension displays, with better performance and larger N2pc components relative to same-dimension displays.

Methods

Participants

Twelve individuals took part in Experiment 3 (M age = 28 years, SD = 6; four men; one left-handed), all reporting normal or corrected-to-normal vision.

Stimuli and Procedure

Participants now performed a four-feature search task where targets were defined by one of two possible colors or two possible shapes, respectively. Pilot testing revealed that this task was very challenging and produced much slower RTs and higher error rates than the tasks used in Experiments 1 and 2. For this reason, search display duration was extended from 100 to 150 msec, and the interval between two successive search displays from 1400 to 1850 msec. In addition, the number of search display objects was reduced from eight to six in Experiment 3. Each search display contained six colored shapes arranged around an imaginary circle at an eccentricity of 2.10° (three objects in the left and three in the right visual field; see Figure 1, bottom). The distance between these objects was increased (by increasing stimulus eccentricity from 1.78° to 2.10°) to minimize visual crowding, which was less of an issue in Experiments 1 and 2 where search display objects only differed in a single dimension. Objects could appear in eight possible colors (red, orange, yellow, green, blue, magenta, gray, and, additionally for this experiment, pink; CIE coordinates: .315/.178) and eight possible shapes (square, circle, triangle, cross, heart, star, pentagon, and hexagon). For each participant, two colors and two shapes were chosen as target features, from a rotated choice of red/blue/green and circle/square/triangle.

Participants' task was to search for target objects that matched one of the four designated target features (e.g., red, green, circle, or square). Single-target displays contained one object with a target-defining feature, which was equally likely to be a color or a shape. In two-target/match displays, the two target objects were defined by the same color or shape (e.g., red pentagon and red star; blue circle and orange circle). In two-target/mismatch displays, the two targets were defined by two different features. For half of these displays, these features were drawn from the same dimension (i.e., two different target colors or two different target shapes; e.g., red pentagon and green heart, or green circle and yellow square; same-dimension displays). For the other half of two-target/mismatch displays, they belonged to different dimensions (i.e., one target color and one target shape; e.g., red pentagon and blue circle; different-dimension displays). Single-target, two-target/match and two-target/mismatch displays were equally likely and randomly distributed within each block.1 In all search displays, the colors and shapes of the distractor objects and the other irrelevant attribute of the target object(s) were unique features. As in Experiments 1 and 2, both target objects in two-target displays were always presented in the same hemifield. Because search displays now only contained three objects on each side, the locations of these two targets were selected randomly, and they could therefore be presented at adjacent locations.

Response instructions were analogous to Experiments 1 and 2. Participants had to report on each trial whether a search display contained one target, two targets with a matching task-relevant feature, or two targets with mismatching task-relevant features. Because this task was difficult, it required considerable training. All participants completed a set of practice blocks before formal EEG testing, until they achieved a fixed performance threshold (above 75% response accuracy). This practice phase lasted approximately 15–20 min and was followed by 16 experimental blocks of 72 trials each.

EEG Recording and Analysis

These were similar to Experiments 1 and 2. ERPs were computed separately for all different types of search displays containing target object(s) in the left or right visual field. In the case of single-target displays, two-target/match displays, and two-target/mismatch displays where both targets were drawn from the same dimension, separate ERPs were computed for displays that contained color-defined or shape-defined target objects. Separate ERPs were also computed for two-target/mismatch displays that contained one shape-defined and one color-defined target (different-dimension displays). N2pc mean amplitudes were again computed within a 200–300 msec poststimulus time window.

Results

Behavioral Performance

An initial ANOVA compared RTs for single-target displays, two-target/match displays (both collapsed across displays with color-defined and shape-defined targets), and two-color/mismatch displays (collapsed across displays with same-dimension and different-dimension targets). A significant effect of Display Type was observed, F(2, 22) = 73.75, p < .001, ηp2 = .87. RTs were much faster for two-target/match displays (M = 865 msec) relative to two-target/mismatch displays (M = 1093 msec), t(11) = 12.55, p < .001, and single-target displays (M = 1113 msec), t(11) = 8.89, p < .001, which did not differ in terms of RTs (t < 1). RTs in response to two-target/mismatch displays were slower when targets were defined in different dimensions as compared with the same dimension (M = 1013 vs. 1173 msec), t(11) = 11.22, p < .001. However, RTs for both types of displays were reliably slower than RTs to two-target/match displays (both ts > 10.08, both ps < .001).

To assess any RT differences between displays with color-defined and shape-defined targets, an additional analysis was conducted for single-target, two-target/match, and two-target/mismatch displays with the factors Display Type and Target Dimension (color, shape). This analysis excluded search displays with two targets from different dimensions. A main effect of Display Type, F(2, 22) = 60.83, p < .001, ηp2 = .85, was accompanied by an effect of Target Dimension, F(1, 11) = 38.38, p < .001, ηp2 = .78, as RTs were generally faster for displays with color-defined relative to shape-defined targets (M = 967 vs. 1027 msec). These two factors interacted, F(2, 22) = 17.36, p < .001, ηp2 = .61. RT benefits for color-defined targets were present for single-target displays (M = 1097 vs. 1129 msec), t(11) = 4.52, p = .001, and for two-target/mismatch displays (M = 948 vs. 1077 msec), t(11) = 5.44, p < .001, but not for two-target/match displays (M = 857 vs. 874 msec), t(11) = 1.75, p = .11.

Parallel analyses were performed for error rate data. The initial ANOVA revealed an effect of Display Type, F(2, 22) = 11.78, p < .001, ηp2 = .52. Error rates were lower for two-target/match displays (M = 5%) as compared with two-target/mismatch displays (M = 19%), t(11) = 6.60, p < .001, and single-target displays (M = 15%), t(11) = 2.84, p = .02, which did not differ reliably, t(11) = 1.31, p = .22. Errors in response to two-target/mismatch displays were more frequent when the two targets were drawn from different dimensions than when they came from the same dimension (M = 27% vs. 11%), t(11) = 6.97, p < .001. However, error rates were higher for both types of displays relative to two-target/match displays (both ts > 4.45, both ps ≤ .001). Across all two-target/mismatch displays, participants reported the presence of a single target on 77% of all error trials and two identical targets on only 23% of these trials. For two-target/match displays, the opposite pattern of response errors was observed (33% single-target responses vs. 65% different-target responses). The additional ANOVA including the factor Target Dimension obtained main effects of Display Type, F(2, 22) = 6.37, p = .007, ηp2 = .37, and Target Dimension, F(1, 11) = 12.77, p = .004, ηp2 = .54, as errors were generally less frequent for color-defined targets (M = 9% vs. 12%). There was a marginal interaction, F(2, 22) = 3.33, p = .054, ηp2 = .23, between both factors. A color benefit for response accuracy was present for two-target/mismatch displays (error rates: M = 8% vs. 15%), t(11) = 3.49, p = .005, but not for single-target displays (M = 16% vs. 14%; t < 1) or two-target/match displays (M = 6% vs. 5%), t(11) = 1.27, p = .23.

N2pc Components

Figure 5 (top) shows contralateral and ipsilateral ERPs and the corresponding N2pc difference waveforms elicited in Experiment 3 for single-target displays and two-target/match displays (collapsed across displays with color-defined and shape-defined targets) and two-target/mismatch displays (collapsed displays with same-dimension and different-dimension target objects). The overall pattern of N2pc components was similar to the results observed in Experiments 1 and 2A (see Figure 3). N2pc amplitudes were larger for two-target/match as compared with N2pcs triggered by two-target/mismatch displays, which was only slightly larger than N2pcs in response to single-target displays. In an ANOVA of N2pc mean amplitudes with the factors Display Type and Laterality, a main effect of Laterality, F(1, 11) = 47.67, p < .001, ηp2 = .81, was accompanied by a significant Display Type × Laterality interaction, F(2, 22) = 41.75, p < .001, ηp2 = .79. All three display types triggered reliable N2pc components (ts > 5.09, ps < .001). Critically, N2pc amplitudes were significantly larger for two-target/match displays as compared with two-target/mismatch displays (M diff = −2.62 vs. 1.52 μV, SD = 1.15 vs. 1.03), t(11) = 13.96, p < .001. The N2pc for single-target displays (M diff = −.98 μV, SD = .47) was smaller than for two-target/match displays, t(11) = 6.98, p < .001, and was also attenuated relative to two-target/mismatch displays, t(11) = 2.73, p = .02.

Figure 5. 

Top: Grand-averaged ERPs obtained in Experiment 3 at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s). ERPs are shown separately for single-target displays, two-target/match displays, and two-target/mismatch displays, together with the corresponding N2pc difference waveforms. Bottom: Contralateral and ipsilateral ERPs and corresponding difference waves measured in Experiment 3 for two-target/mismatch displays with targets from the same or different dimension.

Figure 5. 

Top: Grand-averaged ERPs obtained in Experiment 3 at electrodes PO7/PO8 contralateral and ipsilateral to the side of the target object(s). ERPs are shown separately for single-target displays, two-target/match displays, and two-target/mismatch displays, together with the corresponding N2pc difference waveforms. Bottom: Contralateral and ipsilateral ERPs and corresponding difference waves measured in Experiment 3 for two-target/mismatch displays with targets from the same or different dimension.

Figure 5 (bottom) shows contralateral and ipsilateral ERPs as well as N2pc difference waveforms elicited in Experiment 3 for two-target/mismatch displays that included two target objects defined by two different features from the same dimension (two colors or two shapes) or from different dimensions (one color, one shape). N2pc components were reliably present and similar in size for both types of displays (same-dimension displays: M diff = −1.62 μV, SD = .85, t(11) = 6.60, p < .001; different-dimension displays: M diff = −1.37 μV, SD = 1.35, t(11) = 3.54, p = .005). A direct comparison of N2pc amplitudes between these displays obtained no reliable difference, t(11) = 1.21, p = .25.

To assess any N2pc differences between search displays with color-defined versus shape-defined target objects, N2pcs for one-target, two-target/match, and two-target/mismatch displays were computed separately for color and shape targets. Figure 6 shows the resulting contralateral–ipsilateral N2pc difference waves. Note that for two-target/mismatch displays (right), only displays containing two different target objects from the same dimension were included. An omnibus 3 × 2 × 2 ANOVA of N2pc mean amplitudes with the factors Display Type, Target Dimension, and Laterality obtained a significant three-way interaction, F(2, 22) = 4.48, p = .02, ηp2 = .29. As can be seen in Figure 6, there were no N2pc amplitude differences between color-defined and shape-defined targets in single-target and two-target/match displays, and this was confirmed by the absence of any Target Dimension × Laterality interactions for these displays (both F < 1). In contrast, this interaction was significant for two-target/mismatch displays, F(1, 11) = 9.81, p = .01, ηp2 = .47, with larger N2pc amplitudes for displays containing two different shape-defined targets relative to displays with two different color targets (M diff = −1.99 vs. −1.27 μV, SD = 1.14 vs. .71; see Figure 6, right).

Figure 6. 

N2pc difference waveforms computed by subtracting ipsilateral from contralateral ERPs measured in Experiment 3 for search displays that contained either color-defined or shape-defined targets. Difference waves are shown separately for single-target displays, two-target/match displays, and two-target/mismatch displays with two different color-defined or shape-defined targets.

Figure 6. 

N2pc difference waveforms computed by subtracting ipsilateral from contralateral ERPs measured in Experiment 3 for search displays that contained either color-defined or shape-defined targets. Difference waves are shown separately for single-target displays, two-target/match displays, and two-target/mismatch displays with two different color-defined or shape-defined targets.

Discussion of Experiment 3

Experiment 3 tested the single-feature access principle of the Boolean map theory in a search task where targets were defined by one of four possible features from two different dimensions (color and shape), and target objects in the same display were never physically identical. We compared behavioral performance and N2pc components for two-target search displays with target objects that either had the same or two different target-defining features. The critical question was whether the differences between these two-target/match and two-target/mismatch displays would be similar to those observed in Experiments 1 and 2A between two-target/same and two-target/different displays. This was indeed the case. RTs were more than 200 msec faster in response to displays with two feature-matching targets than for displays with mismatching targets, and error rates were much lower (5% vs. 19%; see Figure 2). Importantly, N2pc components elicited by two-target/match displays were also larger than N2pcs for two-target/mismatch displays. These observations are in line with the single-feature access principle of the Boolean map theory and show that it also applies to nonidentical target objects. Similar differences between two-target/match and mismatch displays were observed regardless of whether these displays contained color-defined or shape-defined targets, demonstrating that they are not specific to color-guided attentional control processes but generalize across feature dimensions. They mirrored the differences observed in Experiments 1 and 2A between displays with two identical versus two different color targets, in spite of the fact that target objects were always physically different in Experiment 3.

As in Experiment 2A, N2pc amplitudes to two-target/mismatch displays were slightly but reliably larger than N2pcs for single-target displays. This is not entirely in line with the prediction of the Boolean map theory that only a single target object can be accessed in both types of displays. However, as will be discussed below, this difference may have been the result of factors that are not directly linked to visual access. It is notable that N2pc components to two-target/mismatch displays were virtually identical when these displays contained two targets defined in the same versus different dimensions (see Figure 5, bottom). This observation shows that the limitation in visual access to targets defined by different features do not differ as a function of whether these target-defining features come from the same or from different dimensions. In both cases, access is limited to a single feature. Although N2pcs to search displays with two target objects defined by different features in the same versus different dimensions were statistically equivalent, clear behavioral costs (slower RTs and more errors) were found for displays containing one color-defined and one shape-defined target. The absence of N2pc differences between these displays suggests that these costs are generated at processing stages that follow visual access. For example, a decision that two targets are different may be reached faster when it can be based on comparing these targets within the same dimension than when different dimensions are involved.

The fact that the behavioral and N2pc effects observed in Experiment 3 were similar to those found in Experiments 1 and 2A also provides further evidence against the perceptual grouping hypothesis. If these effects were due to the perceptual grouping of identical target objects in the same display, they should have been absent or considerably smaller in Experiment 3, where these target objects always differed in a least one feature (color or shape). Although the presence of the other shared feature could still have resulted in some perceptual grouping, any such tendency should have been smaller than in Experiments 1 and 2A for physically identical target objects. The results of Experiment 3 are thus not in line with a low-level perceptual grouping account. They instead suggest that simultaneous visual access to target objects is impaired when they are defined by different features as compared with a single shared feature, even in search displays that do not contain any physically identical objects.

GENERAL DISCUSSION

Selection and access are two distinct attentional functions during visual search, but the vast majority of research in this area has studied the factors that affect the efficiency of dissociating target and distractor objects (i.e., selection). In this study, we focused instead on visual access and investigated the factors that are responsible for the limitations of visual access to multiple target objects in the same search display. According to the Boolean map theory (Huang & Pashler, 2007), multiple targets can be accessed simultaneously when they share the same target-defining feature but not when these features are different. In three experiments, we manipulated the number and the properties of target objects in visual search displays and measured behavioral and electrophysiological correlates of visual access. These experiments found converging evidence from performance and ERPs that multiple-target access is indeed strongly impaired when target objects are defined by different features.

In Experiment 1, performance was better and N2pc amplitudes were larger for search displays where two target objects shared the same color than for displays that contained two different color-defined target objects. This demonstrates that visual access to different-color targets was impaired relative to access to same-color targets and suggests that, in the former case, only a single target object could be accessed at any given moment. These observations were confirmed in Experiment 2A. In contrast, no such differences between two-color/same and two-color/different displays were observed in Experiments 2B and 2C, where targets were color or shape singletons, and could therefore be accessed in a feature-independent fashion (Bacon & Egeth, 1994). In line with the predictions of the Boolean map theory, these results suggest that two color or shape singleton target objects can be accessed in parallel even when they are physically different. Experiment 3 demonstrated that the impairment of access to targets defined by different as compared with shared features remained present in a search task where target objects varied across two dimensions (color and shape), and two targets in the same display were never physically identical.

The costs for search performance and target N2pc amplitudes observed in this study for search displays containing two targets defined by different features as compared with a single shared feature could, in principle, be due not to differences in the efficiency of visual access, but instead to low-level perceptual grouping processes. In Experiments 1 and 2A, two-color/same displays contained two identical target objects, whereas two-target/different displays included two physically different targets. Even though these targets were never presented at immediately adjacent positions, the possibility remains that two identical targets were visually grouped and that this resulted in better performance and larger N2pc components for these targets (see Mazza & Caramazza, 2012, for such a proposal). However, other results of this study are not in line with this hypothesis. There were no reliable performance or N2pc amplitude differences between two-target/same and two-target/different displays when targets were feature singletons (in Experiments 2B and 2C), even though perceptual grouping should have been equally if not more effective under these conditions. Furthermore, essentially the same costs for search displays with two targets that did not share a feature were observed in Experiment 3, where all search display objects were physically different, and no perceptual grouping of identical target objects was possible. These observations rule out the possibility that these costs were primarily associated with low-level visual grouping, and thus strongly suggest that they reflect differences in visual access.

If this conclusion is correct, it also implies that the N2pc component is sensitive to visual access processes and can be employed to investigate factors that modulate the efficiency of access to multiple targets in the same search display. Although it has previously been argued that the N2pc is primarily associated with earlier selection-related functions of attention (such as distractor suppression; e.g., Luck & Hillyard, 1994b), the current results firmly connect this component to attentional target processing stages that follow the initial discrimination between task-relevant and irrelevant objects in visual displays. At a more general level, the question whether the N2pc reflects either selection or access may be misplaced, because there is no a priori reason to assume that this component has to be exclusively associated with a single discrete stage of attentional processing. Previous research has shown that the N2pc is sensitive to different aspects of attentional control that emerge gradually in real time (see Eimer, 2014, for further discussion). For example, the early and late phase of the target N2pc in conjunction search tasks appear to reflect dissociable feature-based versus object-based attentional guidance mechanisms (Berggren & Eimer, 2018). Thus, the current evidence for links between the N2pc and visual access does not imply that this component cannot also be sensitive to factors that modulate target–distractor selection processes.

Overall, the results of this study provide strong support for the claim of the Boolean map theory that visual access to multiple target objects in the same search display is strongly impaired when these objects are defined by different features. However, this theory makes more specific claims about the nature of this impairment. It assumes that access to visual target objects is governed by maps that separate the visual field into selected and nonselected regions and can contain only a single feature label for all target objects. When search displays contain two or more objects that differ with respect to their target-defining features, a Boolean map can only represent one of these targets with its corresponding feature label. In this case, sequential feature-specific selection processes are required, each resulting in a different Boolean map, which can then be combined (Huang & Pashler, 2007). Thus, the single-feature access principle of the Boolean map theory predicts that only a single target object can be immediately accessed in two-target/different displays, and the serial-access hypothesis postulates that access to each target in these displays operates in a sequential fashion.

The current results provide only partial support for these two specific claims. The pattern of errors observed for two-target/different displays is in line with the hypothesis that only one of the two target objects was immediately accessed in these displays. When access was feature-based (in Experiments 1, 2A, and 3), participants were more likely to incorrectly report the presence of a single target object than the presence of two matching objects. This indicates that they failed to register the presence of the other target in the same display, consistent with the single-object access hypothesis. In contrast, such single-target reports were very infrequent for displays with two feature-matching targets and for all two-target displays during color or shape singleton search, suggesting that both targets were visually accessible under these conditions. However, the pattern of N2pc results observed in this study was not entirely in line with the single-feature access principle proposed by the Boolean map theory. In Experiments 2A and 3, small but reliable N2pc amplitude differences were found between single-target displays and two-target/different or two-target/mismatch displays, respectively, with smaller N2pcs for single-target displays. A similar nonsignificant tendency in the same direction was present in Experiment 1. If access was limited to a single target object in both displays, the corresponding N2pc components for these displays should have been identical in size. The fact that N2pc amplitudes tended to be larger for two-target/different displays suggests that even though visual access to both targets was strongly impaired relative to two-target/same displays, it may not have been completely restricted to a single target object on all trials.

It is possible that the N2pc amplitude differences between single-target and two-target/different displays were due to the fact that the target objects in this study were defined by different possible features (two equally likely colors in Experiments 1 and 2A; two possible colors and two possible shapes in Experiment 3). This was necessary to be able to include two-target search displays with two different target-defining features. As a result, participants did not know which target feature would be encountered in any particular search display and thus had to activate preparatory attentional task sets for all possible target features. Previous research has shown that such multiple-feature task sets can be employed (e.g., Grubert & Eimer, 2016; Irons et al., 2012). However, the guidance of search is generally less effective than in tasks where targets are defined by a single constant feature, as reflected by slower RTs and smaller target N2pc components. These multiple-feature costs suggest that observers fail to activate task sets for all possible target attributes on some trials, resulting in an impaired allocation of attention to targets defined by an insufficiently prepared feature. It is likely that such multiple-feature costs were also present in the two-color search tasks of Experiments 1 and 2A and especially in the four-feature search task of Experiment 3. Importantly, these costs will affect single-target search displays more strongly than displays with two different targets. When only one task set for one target color is fully prepared during two-color search (as in Experiments 1 and 2A), selecting and accessing a target that matches the other color in single-target displays will be impaired, resulting in attenuated target N2pcs. In contrast, because two-target/different displays always contain both target colors, the target that happens to match the currently active task set can be selected and accessed on all trials, thereby reducing or eliminating any multiple-feature costs at the level of target N2pc components. The same applies to the four-feature search task of Experiment 3, where a failure to activate task sets for all possible target features on some trials will produce larger costs for N2pc amplitudes for single-target as compared with two-target/mismatch displays. In this case, the N2pc amplitude differences observed between these two types of displays would not reflect differences in the number of objects that can be accessed simultaneously, but instead limitations in the guidance of attentional selection (i.e., the discrimination between targets and distractors) during multiple-feature search that specifically affect single-target displays. If this was correct, these N2pc differences would not be inconsistent with the single-feature access principle of the Boolean map theory.

According to the serial-access hypothesis postulated by this theory, two successive access processes are required when different target objects with mismatching features appear in the same search display. In principle, this should give rise to two temporally separated N2pc components in response to these displays. The N2pc measured during the 200–300 msec poststimulus time window should reflect the initial access to one of these targets and a second N2pc elicited at a longer poststimulus latency access to the other target object. However, no ERP support for two serial-access processes triggered by two-target/different displays was obtained in this study. Comparisons of ERPs elicited by two-target/same versus two-target/different displays beyond 300 msec poststimulus did not find any evidence for an enhanced contralateral negativity in response to the latter displays that would reflect a delayed second N2pc component. The apparent absence of such effects does not necessarily imply that the serial-access claim of the Boolean map theory is incorrect, as it may instead reflect methodological limitations of this study. Both target objects in two-target displays always appeared in the same hemifield and thus will both have elicited contralateral negativities in the same hemisphere. It is generally difficult to dissociate serial attentional allocation processes by temporally separating lateralized ERP components with the same polarity. For this reason, previous N2pc studies of serial selection (e.g., Eimer & Grubert, 2014; Woodman & Luck, 2003) employed different procedures where one target was presented laterally and the other on the vertical midline (above/below fixation), thus ensuring that only the lateral target would trigger an N2pc. Similar procedures will be required in future work to test the serial-access hypothesis of the Boolean map theory more decisively.

Overall, the current study has provided novel insights into the factors that are responsible for the limitations of visual access to multiple target objects in the same display. We have provided behavioral and electrophysiological evidence demonstrating that such access processes are strongly affected by whether these targets share task-relevant features or not. Multiple-target access is strongly impaired when targets are defined by different features, as predicted by the Boolean map theory (Huang & Pashler, 2007). Although specific assumptions (i.e., the single-feature access and serial-access hypotheses) have to await further tests, the more general claim of this theory that visual access mechanisms are constrained by the match versus mismatch of target-defining features was fully confirmed by the present results. Simultaneous access to multiple target objects appears to be subject to feature-selective inhibitory interactions, suggesting that the attentional control processes that mediate the encoding and recognition of these objects are feature-specific and operate in a competitive fashion. Identifying the nature of these competitive interactions and the functional architecture of the top–down control networks that generate them will be an important goal for future research on the mechanisms of visual access.

Acknowledgments

This work was supported by grant ES/L016400/1 from the Economic and Social Research Council (ESRC), United Kingdom. N. B. is also supported by a New Investigator Grant ES/R003459/1 from the ESRC.

Reprint requests should be sent to Nick Berggren, Department of Psychological Sciences, Birkbeck College, University of London, Malet Street, London WC1E 7HX, United Kingdom, or via e-mail: nbergg01@mail.bbk.ac.uk.

Note

1. 

Due to a programming error, the total number of two-target displays with two identical versus two different color-defined targets was not identical for the first six participants tested. These participants were presented with 168 identical-color displays and 216 different-color displays. The additional trials with different-color displays were removed from analysis. This error was corrected for the other six participants, who received each of these two types of displays on 192 trials.

REFERENCES

REFERENCES
Bacon
,
W. F.
, &
Egeth
,
H. E.
(
1994
).
Overriding stimulus-driven attentional capture
.
Perception & Psychophysics
,
55
,
485
496
.
Berggren
,
N.
, &
Eimer
,
M.
(
2018
).
Object-based target templates guide attention during visual search
.
Journal of Experimental Psychology: Human Perception and Performance
,
44
,
1368
1382
.
Desimone
,
R.
, &
Duncan
,
J.
(
1995
).
Neural mechanisms of selective visual attention
.
Annual Review of Neuroscience
,
18
,
193
222
.
Drew
,
T.
, &
Vogel
,
E. K.
(
2008
).
Neural measures of individual differences in selecting and tracking multiple moving objects
.
Journal of Neuroscience
,
28
,
4183
4191
.
Duncan
,
J.
(
1980
).
The locus of interference in the perception of simultaneous stimuli
.
Psychological Review
,
87
,
272
300
.
Duncan
,
J.
(
2006
).
EPS mid-career award 2004: Brain mechanisms of attention
.
Quarterly Journal of Experimental Psychology
,
59
,
2
27
.
Duncan
,
J.
, &
Humphreys
,
G. W.
(
1989
).
Visual search and stimulus similarity
.
Psychological Review
,
96
,
433
458
.
Eimer
,
M.
(
1996
).
The N2pc component as an indicator of attentional selectivity
.
Electroencephalography and Clinical Neurophysiology
,
99
,
225
234
.
Eimer
,
M.
(
2014
).
The neural basis of attentional control in visual search
.
Trends in Cognitive Sciences
,
18
,
526
535
.
Eimer
,
M.
(
2015
).
EPS Mid-Career Awarded 2014: The control of attention in visual search: Cognitive and neural mechanisms
.
Quarterly Journal of Experimental Psychology
,
68
,
2437
2463
.
Eimer
,
M.
, &
Grubert
,
A.
(
2014
).
Spatial attention can be allocated rapidly and in parallel to new visual objects
.
Current Biology
,
24
,
193
198
.
Fuggetta
,
G.
,
Pavone
,
E. F.
,
Walsh
,
V.
,
Kiss
,
M.
, &
Eimer
,
M.
(
2006
).
Cortico-cortical interactions in spatial attention: A combined ERP/TMS study
.
Journal of Neurophysiology
,
95
,
3277
3280
.
Grubert
,
A.
, &
Eimer
,
M.
(
2016
).
All set, indeed! N2pc components reveal simultaneous attentional control settings for multiple target colours
.
Journal of Experimental Psychology: Human Perception and Performance
,
42
,
1215
1230
.
Hopf
,
J. M.
,
Luck
,
S. J.
,
Girelli
,
M.
,
Hagner
,
T.
,
Mangun
,
G. R.
,
Scheich
,
H.
, et al
(
2000
).
Neural sources of focused attention in visual search
.
Cerebral Cortex
,
10
,
1233
1241
.
Huang
,
L.
(
2010
).
What is the unit of visual attention? Object for selection, but Boolean map for access
.
Journal of Experimental Psychology: General
,
139
,
162
179
.
Huang
,
L.
, &
Pashler
,
H.
(
2007
).
A Boolean map theory of visual attention
.
Psychological Review
,
114
,
599
631
.
Huang
,
L.
, &
Pashler
,
H.
(
2012
).
Distinguishing different strategies of across-dimension attentional selection
.
Journal of Experimental Psychology: Human Perception and Performance
,
38
,
453
464
.
Huang
,
L.
,
Treisman
,
A.
, &
Pashler
,
H.
(
2007
).
Characterizing the limits of human visual awareness
.
Science
,
317
,
823
825
.
Irons
,
J. L.
,
Folk
,
C. L.
, &
Remington
,
R. W.
(
2012
).
All set! Evidence of simultaneous attentional control settings for multiple target colors
.
Journal of Experimental Psychology: Human Perception and Performance
,
38
,
758
775
.
Luck
,
S. J.
(
2012
).
Electrophysiological correlates of the focusing of attention within complex visual scenes: N2pc and related ERP components
. In
S. J.
Luck
&
E. S.
Kappenman
(Eds.),
The Oxford handbook of event-related potential components
(pp.
329
360
).
New York
:
Oxford University Press
.
Luck
,
S. J.
, &
Ford
,
M. A.
(
1998
).
On the role of selective attention in visual perception
.
Proceedings of the National Academy of Sciences, U.S.A.
,
95
,
825
830
.
Luck
,
S. J.
, &
Hillyard
,
S. A.
(
1994a
).
Electrophysiological correlates of feature analysis during visual search
.
Psychophysiology
,
31
,
291
308
.
Luck
,
S. J.
, &
Hillyard
,
S. A.
(
1994b
).
Spatial filtering during visual search: Evidence from human electrophysiology
.
Journal of Experimental Psychology: Human Perception and Performance
,
20
,
1000
1014
.
Mazza
,
V.
, &
Caramazza
,
A.
(
2011
).
Temporal brain dynamics of multiple object processing: The flexibility of individuation
.
PLoS One
,
6
,
e17453
.
Mazza
,
V.
, &
Caramazza
,
A.
(
2012
).
Perceptual grouping and visual enumeration
.
PLoS One
,
7
,
e50862
.
Mazza
,
V.
,
Pagano
,
S.
, &
Caramazza
,
A.
(
2013
).
Multiple object individuation and exact enumeration
.
Journal of Cognitive Neuroscience
,
25
,
697
705
.
Pagano
,
S.
, &
Mazza
,
V.
(
2012
).
Individuation of multiple targets during visual enumeration: New insights from electrophysiology
.
Neuropsychologia
,
50
,
754
761
.
Posner
,
M. I.
, &
Petersen
,
S. E.
(
1990
).
The attention system of the human brain
.
Annual Review of Neuroscience
,
13
,
25
42
.
Treisman
,
A. M.
, &
Gelade
,
G.
(
1980
).
A feature-integration theory of attention
.
Cognitive Psychology
,
12
,
97
136
.
Wolfe
,
J. M.
(
2007
).
Guided Search 4.0: Current progress with a model of visual search
. In
W.
Gray
(Ed.),
Integrated models of cognitive systems
(pp.
99
119
).
New York
:
Oxford
.
Woodman
,
G. F.
, &
Luck
,
S. J.
(
1999
).
Rapid shifts of attention during visual search
.
Nature
,
400
,
867
869
.
Woodman
,
G. F.
, &
Luck
,
S. J.
(
2003
).
Serial deployment of attention during visual search
.
Journal of Experimental Psychology: Human Perception and Performance
,
29
,
121
138
.
Wyble
,
B.
,
Potter
,
M. C.
,
Bowman
,
H.
, &
Nieuwenstein
,
M.
(
2011
).
Attentional episodes in visual perception
.
Journal of Experimental Psychology: General
,
140
,
488
505
.
Zivony
,
A.
,
Allon
,
A. S.
,
Luria
,
R.
, &
Lamy
,
D.
(
2018
).
Dissociating between the N2pc and attentional shifting: An attentional blink study
.
Neuropsychologia
,
121
,
153
163
.
Zivony
,
A.
, &
Lamy
,
D.
(
2016
).
Attentional capture and engagement during the attentional blink: A “camera” metaphor of attention
.
Journal of Experimental Psychology: Human Perception and Performance
,
42
,
1886
1902
.