Differential effects of exogenous and endogenous attention on second-order texture contrast sensitivity

Abstract

Abstract

Abstract:

Abstract  Thevisual system can use a rich variety of contours to segment visual scenes into distinct perceptually coherent regions. However, successfully segmenting an image is a computationally expensive process. Previously we have shown that exogenous attention—the more automatic, stimulus-driven component of spatial attention—helps extract contours by enhancing contrast sensitivity for second-order, texture-defined patterns at the attended location, while reducing sensitivity at unattended locations, relative to a neutral condition. Interestingly, the effects of exogenous attention depended on the second-order spatial frequency of the stimulus. At parafoveal locations, attention enhanced second-order contrast sensitivity to relatively high, but not to low second-order spatial frequencies. In the present study we investigated whether endogenous attention—the more voluntary, conceptually-driven component of spatial attention—affects second-order contrast sensitivity, and if so, whether its effects are similar to those of exogenous attention. To that end, we compared the effects of exogenous and endogenous attention on the sensitivity to second-order, orientation-defined, texture patterns of either high or low second-order spatial frequencies. The results show that, like exogenous attention, endogenous attention enhances second-order contrast sensitivity at the attended location and reduces it at unattended locations. However, whereas the effects of exogenous attention are a function of the second-order spatial frequency content, endogenous attention affected second-order contrast sensitivity independent of the second-order spatial frequency content. This finding supports the notion that both exogenous and endogenous attention can affect second-order contrast sensitivity, but that endogenous attention is more flexible, benefitting performance under different conditions.

Introduction

Each time we open our eyes, we are confronted with far more information than our visual system, limited by the high metabolic cost of cortical computations, can effectively process. Given these limits, we need mechanisms to optimally allocate processing resources according to task demands (Lennie, 2003). Visual attention is one such mechanism that helps manage the expenditure of cortical computation (for a review, see Carrasco, 2011). Usually, we foveate the location in space to which we wish to attend, but we can also direct our attention to a region in our visual field without directing our eyes to that location (Posner, 1980). The selection of information based on its spatial location in the absence of eye movements is referred to as spatial covert attention. Spatial covert attention selectively grants priority in processing to part of the otherwise overwhelming amount of information in our environment and processes it in a privileged way, at the expense of information at unattended locations (Carrasco, 2011; Lu & Dosher, 1998; Luck et al., 1994; Montagna, Pestilli, & Carrasco, 2009; Pestilli & Carrasco, 2005). By trading off processing resources between attended and unattended locations in the visual field, attention allows one to optimize performance in visual tasks while overcoming the visual system's limited capacity. 

Spatial covert attention can be deployed exogenously (involuntarily) and endogenously (voluntarily). Exogenous attention is stimulus-driven, automatically activated by the sudden onset of a stimulus in the visual field. Its effect on perception is fast and transient, peaking at approximately 100–120 ms and decaying again quickly. Endogenous attention is conceptually driven, voluntarily allocated to a location in the visual field within approximately 300–500 ms and can be sustained for several seconds (Carrasco, 2011; Cheal & Lyon, 1991; Herrmann, Montaser-Kouhsari, Carrasco, & Heeger, 2010; Jonides & Irwin, 1981; Ling & Carrasco, 2006a; Liu, Stevens, & Carrasco, 2007; Müller & Findlay, 1988; Müller & Rabbitt, 1989; Nakayama & Mackeben, 1989). 

Whereas the shifts of attention prompted by central, sustained cues are under voluntary control and observers can allocate resources according to cue validity (Giordano, McElree, & Carrasco, 2009; Kinchla, 1980; Mangun & Hillyard, 1990; Sperling & Melchner, 1978), it is extremely difficult for observers to ignore peripheral, transient cues (Cheal, Lyon, & Hubbard, 1991; Giordano et al., 2009; Jonides, 1981; Nakayama & Mackeben, 1989; Yantis & Jonides, 1996). Involuntary transient shifts of attention occur even when the cues are known to be uninformative and irrelevant (Montagna et al., 2009; Pestilli & Carrasco, 2005; Pestilli, Viera, & Carrasco, 2007; Prinzmetal, McCool, & Park, 2005; Yeshurun & Rashal, 2010), and even when responding to the cues may impair performance (Carrasco, Loula, & Ho, 2006; Hein, Rolke, & Ulrich, 2006; Talgar & Carrasco, 2002; Yeshurun, 2004; Yeshurun & Carrasco, 1998, 2000; Yeshurun & Levy, 2003; Yeshurun, Montagna, & Carrasco, 2008). 

Attention can affect sensitivity at both attended and unattended locations of the visual field. For instance, attention can increase contrast sensitivity to patterns defined by changes in luminance contrast (i.e., first-order information), improving behavioral performance in various tasks. Specifically, attention enhances contrast sensitivity for first-order information at the attended location and decreases sensitivity at unattended locations, relative to a neutral condition (Cameron, Tai, & Carrasco, 2002; Carrasco, Penpeci-Talgar, & Eckstein, 2000; Dosher & Lu, 2000; Ling & Carrasco, 2006b; Lu & Dosher, 1998; Pestilli & Carrasco, 2005; Pestilli, Ling, & Carrasco, 2009). These effects of attention on first-order processing occur across a wide range of spatial frequencies (Carrasco et al., 2000) and early in the visual stream (Brefczynski & DeYoe, 1999; Gandhi, Heeger, & Boynton, 1999; Herrmann et al., 2010; Huk & Heeger, 2000; Liu, Pestilli, & Carrasco, 2005; McAdams & Reid, 2005; Reynolds & Chelazzi, 2004; Störmer, McDonald, & Hillyard, 2009). Much of the early visual system from the retina to primary visual cortex supports the detection of first-order luminance changes (DeValois & DeValois, 1988). Linear spatial filters that are selective for spatial frequency and orientation, akin to simple cells in V1, are effective for signaling first-order information (Graham, 1989) and attention has been shown to boost neuronal activity at the attended location (Reynolds & Chelazzi, 2004). 

Humans are sensitive to changes in visual attributes other than luminance that cannot be detected by such linear mechanisms. Natural environments contain a rich variety of contours that the visual system extracts to parse the retinal image into figures and backgrounds. For example, boundaries defined by changes in textural attributes (e.g., local orientation, contrast, and spatial frequency) that are not accompanied by luminance changes on either side of the texture boundary cannot be detected by linear mechanisms. Texture is a property that is statistically defined. The visual system can segment regions of the visual field based on local textural properties, such as separating “vertically oriented textures” from “horizontally oriented textures.” Consider the scene in Figure 1. The borders between the zebras and the grass involve primarily differences in average luminance, which can be detected by first-order filters. In contrast, the border between the two zebras on the right does not involve a difference in average luminance. Nevertheless, we still perceive a smooth, continuous occlusion boundary between the two animals, as we are also sensitive to changes in textural attributes (e.g., local orientation). Such textural patterns are commonly referred to as second-order information, to distinguish them from first-order luminance-defined patterns. 

Figure 1

 

Natural scene containing boundaries signaled by differences in luminance (first-order information) and/or by differences in textural properties (second-order information). Picture obtained from http://www.pdphoto.org.

The processing of first- and second-order patterns is supported by different neural mechanisms (Ellemberg, Allen, & Hess, 2004; Larsson, Landy, & Heeger, 2006; Morgan, Mason, & Baldassi, 2000; Schofield & Georgeson, 1999; Scott-Samuel & Georgeson, 1999). Similar to first-order filters, second-order filters are tuned for orientation (Arsenault, Wilkinson, & Kingdom, 1999; Dakin, Williams, & Hess, 1999; Graham & Wolfson, 2001) and spatial frequency (Landy & Oruç, 2002; Scott-Samuel & Georgeson, 1999), but have wider bandwidths (Landy & Oruç, 2002). Whereas representations in V1 are sensitive to first-order statistics, neurophysiologic evidence suggests that second-order representations for static stimuli involve additional processing after V1, building up throughout the ventral visual stream from V1 to V4 (El-Shamayleh & Movshon, 2011; Hallum, Landy, & Heeger, 2011; Larsson et al., 2006; Montaser-Kouhsari, Landy, Heeger, & Larsson, 2007). Similarly, for dynamic stimuli, second-order motion representations involve additional processing along the dorsal stream (Ashida, Lingnau, Wall, & Smith, 2007; Baker, 1999; Dumoulin, Baker, Hess, & Evans, 2003). 

Although the effects of attention on first-order information are well documented, little is known about the effects of attention on second-order processing. Texture perception has been considered to be pre-attentive (Braun & Sagi, 1990; Julesz, 1981; Schubö & Meinecke, 2007), but many studies have shown that attention can benefit performance in various texture segmentation tasks (Casco, Grieco, Campana, Corvino, & Caputo, 2005; Talgar & Carrasco, 2002; Yeshurun & Carrasco, 1998, 2000, 2008; Yeshurun et al., 2008). Texture patterns are predominant in natural scenes and are vital for the reliable detection and identification of object boundaries. Naturally occurring edges are typically defined by spatially coincident changes in both first- and second-order information (Johnson & Baker, 2004; Johnson, Kingdom, & Baker, 2005). Nevertheless, first- and second-order information can carry different information about the visual scene, even when spatially correlated (Schofield, 2000). Furthermore, many first-order luminance edges do not signal the presence of object boundaries as they often originate from non-uniform surface illuminations (Kingdom, 2003; Schofield, Hesse, Rock, & Georgeson, 2006; Schofield, Rock, Sun, Jiang, & Georgeson, 2010; Sun & Schofield, 2011). For instance, in Figure 1, the shadow cast by the tree does not signal an object boundary. Therefore, second-order information provides a more reliable signal of discontinuities between adjacent surfaces. Successfully segmenting an image is a computationally expensive process. By enhancing the sensitivity to second-order information, attention can serve to highlight region boundaries in visual scenes, efficiently reducing the complexity of interpreting the retinal input. 

Recently, we have documented that exogenous attention helps extract texture-defined boundaries by enhancing contrast sensitivity for second-order information at the attended location, while reducing sensitivity at unattended locations, relative to a neutral condition (Barbot, Landy, & Carrasco, 2011a). However, these effects of attention were a function of the second-order spatial frequency of the stimulus and its eccentricity. At parafoveal locations, exogenous attention enhanced second-order contrast sensitivity to relatively high, but not to low, second-order spatial frequencies. At more peripheral locations, exogenous attention enhanced second-order contrast sensitivity for low second-order spatial frequencies. Barbot et al. (2011a) showed that these effects were consistent with the resolution hypothesis (Carrasco et al., 2006; Carrasco & Yeshurun, 2009; Yeshurun & Carrasco, 2000), whereby exogenous spatial attention increases resolution by increasing the sensitivity of the smallest second-order filters at the attended location. 

Endogenous attention and exogenous attention show some common (Hikosaka, Miyauchi, & Shimojo, 1993; Montagna et al., 2009; Suzuki & Cavanagh, 1997) and some unique perceptual effects. For instance, transient attention improves performance in a texture segmentation task at peripheral locations, but impairs it at central locations (Carrasco et al., 2006; Talgar & Carrasco, 2002; Yeshurun & Carrasco, 1998; Yeshurun et al., 2008), whereas sustained attention improves performance at all eccentricities (Yeshurun et al., 2008). 

In the present study we investigate whether endogenous attention—the more controlled, top-down component of spatial attention—affects second-order contrast sensitivity at both attended and unattended locations and, if so, whether its effects are similar to those of exogenous attention. To that end, we manipulated either exogenous or endogenous attention, maintaining task and identical stimuli, and compared their effects on contrast sensitivity to second-order, orientation-defined, texture patterns. 

Method

Participants

Eight undergraduate and graduate students (two females; age range: 22–32) participated as observers in the exogenous attention experiment. Seven of them also participated in the endogenous attention experiment. Four were experienced psychophysical observers, but all (except one author) were naive as to the purposes of the experiment. All participants had normal or corrected-to-normal vision. The Institutional Review Board at New York University approved the experimental procedures and all participants gave informed consent. 

Apparatus

Stimuli were generated using Matlab (MathWorks, Natick, MA) and MGL (http://gru.brain.riken.jp/mgl). Observers viewed the stimuli on a 21-inch Sony GDM-F520 CRT monitor (resolution: 1600 × 1200, 100 Hz) at a viewing distance of 57 cm with the head stabilized using a chin rest. The displays were calibrated using a Photo Research (Chatworth, CA) PR650 SpectraColorimeter to produce linearized lookup tables. To ensure that all observers were able to maintain steady fixation, eye position was monitored using an infrared video camera system for endogenous-attention (EyeLink 1000, SR Research, Kanata, Ontario, Canada). In the exogenous-attention experiment, observers did not have time to move their eyes between cue onset and stimulus presentation. Nevertheless, we ensured that all observers were able to maintain steady fixation during the first session using an infrared video camera (ISCAN, Burlington, MA). 

Stimuli

Texture patterns were constructed by spatially modulating two orthogonal sine wave gratings (first-order carriers C₁ and C₂ oriented ±45°, 4 cycle/deg) using a second horizontal or vertical sine wave grating with lower spatial frequency (second-order modulator M). The resulting stimulus was defined as follows (Figure 2A): 

 

where L₀ is the mean luminance of the display and E is a circular stimulus envelope with raised-cosine edges (4° diam., raised-cosine width: 0.2°). There was a fixed orientation difference of 45° between the modulator and each of the carriers. For each stimulus, the value of C was set to 70% peak luminance contrast to prevent luminance clipping. Nine second-order modulator contrast levels were used in each attention experiment, which ranged from 8% to 96% (exogenous) and 16% to 96% (endogenous). The square root ensured the root mean square (RMS) contrast was constant across the display (Landy & Oruç, 2002). Stimuli were centered at 5° of eccentricity in each quadrant because contrast sensitivity and spatial resolution do not differ across those locations (Abrams, Nizam, & Carrasco, 2012; Carrasco, Talgar, & Cameron, 2001). There were two different stimulus conditions (Figure 2B): modulator spatial frequencies of 0.5 (‘low') and 1 cycle/deg (‘high'). Note that second-order spatial frequencies separated by an octave or more should activate separate channels (Landy & Oruç, 2002). Similar patterns have been used in psychophysical and neuroimaging studies of texture segmentation (Barbot et al., 2011a; El-Shamayleh & Movshon, 2011; Landy & Oruç, 2002; Larsson et al., 2006). 

Figure 2.

 

(A) Stimulus construction. Texture patterns were computed by modulating two orthogonal luminance gratings (the carriers, with ± 45° orientations) with a third vertical or horizontal modulator grating of lower spatial frequency. The contrast-modulated carrier patterns were summed, and the result was multiplied by a circular window with raised-cosine edges. Mean luminance was constant across the stimulus. (B) Stimulus conditions. Two spatial frequencies were used for the second-order modulator (‘low': 0.5 cycle/deg; ‘high': 1 cycle/deg), while the first-order carrier was of higher spatial frequency (4 cycle/deg).

Procedure

Figure 3 depicts the trial sequence for the exogenous- and endogenous-attention experiments. Each trial began with a 300-ms fixation cross at the center of the screen. Observers were instructed to fixate the cross throughout each trial. Next, exogenous or endogenous attention was manipulated either via peripheral or central precues, respectively, which induced observers to direct their attention to the cued location. The order of the attention experiments (exogenous or endogenous) was counterbalanced across observers. Within each attention experiment, each experimental session only contained one of the two different stimulus conditions (i.e., low or high spatial frequency); the order was counterbalanced. 

Figure 3.

 

Trial sequence. The trial sequence was identical for the exogenous and endogenous attention conditions except for the location and timing of the peripheral and central cues, which differed to maximize the effects of each type of attention.

Exogenous attention was manipulated by a peripheral cue presented adjacent to the target location. In two-thirds of the trials (cued trials), one black rectangle (width: 0.3°, length 1.6°, orientation: 45°) was flashed for 60 ms 3.5° from one stimulus location (i.e., 1.5° border-to-border from the upcoming stimulus). In the remaining one-third of the trials (neutral trials), four black rectangles were presented next to the four possible stimulus locations, distributing observers' attention across space. After an inter-stimulus-interval (ISI) of 40 ms, two second-order texture stimuli were presented simultaneously for 100 ms. The two stimuli could be presented at any two of the four possible locations. Orientation and phase of the second-order modulator were independently randomized across trials for each second-order stimulus in a pair. Then, 100 ms after stimulus display, a white response cue was presented near the fixation cross, pointing to one of the two stimuli (the target). Response cues considerably reduce location uncertainty by indicating the exact target location (Kinchla, Chen, & Evert, 1995; Lu & Dosher, 2004; Luck et al., 1994; Pestilli & Carrasco, 2005; Yeshurun et al., 2008). 

The task was a two-alternative, forced-choice, second-order-orientation-discrimination task. Observers were instructed to report the orientation (vertical or horizontal) of the second-order sine wave grating presented at the target location indicated by the response cue. Auditory feedback was given. About 2,530 trials per stimulus condition were collected for each observer in 16 blocks (in approximately four experimental sessions)—an equal number of trials for each of nine second-order modulator contrast levels. 

For the endogenous-attention experiment, the procedure and the task were the same as for the exogenous condition except that attention was manipulated via symbolic, central cues presented at fixation preceding stimulus presentation. In two-thirds of the trials (cued trials), a black line pointing at one possible location was presented for 300 ms near the fixation cross. In the remaining one-third of the trials (neutral trials), four black lines were presented near fixation, pointing at the four possible stimulus locations. After an ISI of 300 ms, the two second-order texture stimuli were presented simultaneously for 100 ms, and then a response cue was presented 100 ms after stimulus display. About 1,620 trials per stimulus condition were collected for each observer in 12 blocks (in approximately three experimental sessions)—an equal number of trials for each of nine second-order modulator contrast levels. 

As illustrated in Figure 4, the relationship between the attentional cue and the response cue defined the validity of the precue (Downing, 1988; Lu & Dosher, 2004; Luck et al., 1994; Montagna et al., 2009; Pestilli & Carrasco, 2005). On valid trials, the location indicated by the precue and the response cue matched, and observers reported the orientation of the cued (‘attended') stimulus. On invalid trials, the precue and response cue did not match, and observers reported the orientation of the uncued (‘unattended') stimulus. On neutral trials, all four potential stimulus locations were precued. 

Figure 4

 

Cue validity for exogenous and endogenous attention.

For exogenous attention, we used non-predictive precues. The target was equally likely to be the cued (a valid trial) or uncued (an invalid trial) stimulus, and the precue gave no information about target orientation; the precue only indicated stimulus onset, as was the case in the neutral trials (Lu & Dosher, 2004; Lu, Liu, & Dosher, 2000; Montagna et al., 2009; Pestilli & Carrasco, 2005; Pestilli et al., 2007). Observers were told that the precue was uninformative and that it would not be advantageous to move their eyes to the cued location. Note that goal-directed saccades require about 250 ms (Leigh & Zee, 1991; Mayfrank, Kimmig, & Fischer, 1987). Thus, no eye movements to the stimulus could occur between precue onset and stimulus onset (100 ms). Peripheral cues cannot be voluntarily ignored, even if observers are instructed to do so (Jonides & Irwin, 1981), and their effect is not influenced by cue validity (Giordano et al., 2009), given that the peripheral cue automatically draws attention to the cued location (Jonides & Yantis, 1988; Yantis & Jonides, 1984). 

For endogenous attention, we used informative precues. Effects of central cues depend on cue validity and might not occur when the central cues are uninformative (Giordano et al., 2009; Jonides & Irwin, 1981; Jonides & Yantis, 1988; Kinchla, 1980; Sperling & Melchner, 1978). In two-thirds of the cued trials, the cued location corresponded to the target location (valid trials). In one-third of the cued trials, the cued location corresponded to the distractor location (invalid trials). Observers were told that the precue was informative and that it would be advantageous to deploy their attention to the cued location without moving their eyes. Eye position was monitored to ensure that observers were not breaking fixation at any time from the precue onset to the stimulus offset. 

Analysis

For each observer, performance was assessed separately for each stimulus condition (i.e., low or high second-order spatial frequency), second-order modulation contrast and cueing condition (valid, invalid and neutral). We used signal detection theory, treating the vertical second-order stimulus as a signal-present and the horizontal second-order stimulus as a signal-absent trial. Performance was evaluated as d′ = z(hit rate) − z(false alarm rate). d′ values were averaged over observers. The data were fit with a Naka-Rushton function: 

 

using a least-squares criterion, where d′(c) represents performance as a function of contrast, d′_max is the asymptotic performance at high contrast values, c₅₀ is the contrast at which the observer achieves half the asymptotic performance, and n determines the slope of the psychometric function. For each stimulus condition, we fit the data from the three attention conditions, allowing distinct values of d′_max, c₅₀ and n for each attention condition. 

Confidence intervals and p-values were computed by bootstrapping. Specifically, individual psychophysical trials were randomly resampled with replacement to generate resampled data sets, which were refitted using the same procedure. We repeated this procedure of resampling and refitting 10,000 times to generate bootstrap distributions of the fitted parameters. Confidence intervals for the fitted parameters and p-values were based on these bootstrap distributions, e.g., to test if there was a benefit for valid and a cost for invalid cues, compared to neutral cues either in d′_max, c₅₀ or n. We assembled the bootstrap distribution of the differences between the conditions (e.g., valid minus neutral trials) and performed statistical tests by assessing the percentage of the values in the tail of the distribution of the differences greater than zero for changes in d′_max or n, or lower than zero for changes in c₅₀. Performance in this orientation-discrimination task improves with increasing second-order contrast. Thus, if attention increases second-order contrast sensitivity at the attended location and decreases it at unattended locations, it should improve performance at the attended location and performance should be degraded at unattended locations. 

For endogenous attention, eye positions were measured using an infrared eye tracker (EyeLink CL, SR Research, Kanata, Ontario, Canada) with 1000 Hz sampling rate in the main experimental sessions. Eye positions were analyzed offline. For analysis, raw data were converted to eye position in degrees of visual angle. Eye position samples around the time of blinks (100 ms preceding and following a blink) were excluded from further analysis. The mean eye position during the fixation interval at the beginning of each trial served as a baseline and was subtracted from the mean eye position in each following interval (cue presentation, ISI, stimulus presentation) to compensate for any slow drift in the measurements during each block. We used the standard Eyelink detection algorithm (combined velocity [30°/s] and acceleration [8000°/s²] criteria to detect saccades; below, we report their frequency for each condition). 

Results

Exogenous attention

Figure 5 shows the performance data and psychometric function fits averaged across observers (N = 8) for the exogenous-attention experiment. The effects of exogenous attention depended on the spatial frequency of the second-order modulator, as previously reported (Barbot et al., 2011a). When the spatial frequency of the second-order modulator was low (Figure 5A), the three psychometric functions (valid, neutral and invalid) were indistinguishable and no effect of cueing was observed on d'_max (p_{valid-invalid} > 0.1), c₅₀ (p_{valid-invalid} > 0.1) or n (p_{valid-invalid} > 0.1). However, when the spatial frequency of the second-order modulator was high (Figure 5B), performance was consistent with response gain, i.e., attention modulated the value of d′_max (p_{valid-invalid} < 0.01), but not c₅₀ (p_{valid-invalid} > 0.1). Compared to the neutral condition, a valid cue to the target enhanced second-order contrast sensitivity, as indicated by an increase in d′_max, whereas an invalid cue decreased sensitivity, reducing d′_max. No change of n (p_{valid-invalid} > 0.1) was observed. 

Figure 5.

 

(A,B) Effects of exogenous attention on performance (d′) as a function of second-order modulator contrast for the low (A) and high (B) second-order spatial frequency conditions (Figure 2). Each panel shows psychometric functions and parameter estimates (c₅₀: second-order contrast yielding half-maximum performance; d′_max: asymptotic performance) for each cueing condition (valid, neutral and invalid). Each data point represents the mean across observers (N = 8). Error bars correspond to ±1 SEM for data points and 68%-confidence intervals obtained by bootstrapping for parameter estimates. All r² > .98. (C, D) Effects of exogenous attention on individual observers' parameter estimates in the low (C) and high (D) second-order spatial frequency conditions (c₅₀: open circles; d′_max: open squares). Each plot displays individual observers' parameter estimates in the valid (red symbols) and invalid (blue symbols) cue conditions normalized by the corresponding values in the neutral-cue condition. Filled symbols indicate mean across observers.

Figure 5C and D shows the values of c₅₀ and d′_max for individual observers in the valid and invalid attention conditions in which each value is normalized by (i.e., divided by) the corresponding parameter values from the neutral attention condition. The effect of attention on second-order contrast sensitivity was consistent across observers. For the low second-order spatial frequency condition (Figure 5C), there was no clear pattern across valid and invalid conditions, reflecting the absence of attentional modulation. In contrast, for the high second-order spatial frequency condition (Figure 5D), valid d′_max values were higher than the neutral d′_max values (i.e., red squares above the unity line), and invalid d′_max values were lower than the neutral d′_max values (i.e., blue squares below the unity line). These results show that exogenous attention affects contrast sensitivity for second-order, orientation-defined patterns in a manner that depends on second-order spatial frequency. 

Endogenous attention

Figure 6 shows the performance data and psychometric function fits averaged across observers (N = 7) for the endogenous-attention experiment. Although the effects of exogenous attention depended on the spatial frequency of the second-order modulator, endogenous attention affected performance independent of the second-order spatial frequency content. For the low second-order spatial frequency condition (Figure 6A), changes in performance were consistent with a mixture of contrast gain and response gain, as indicated by a decrease in c₅₀ (p_{valid-invalid} < 0.05) and by an increase in d′_max (p_{valid-invalid} < 0.05) with attention. For the high second-order spatial frequency conditions (Figure 6B), changes in performance were consistent with response gain, as indicated by an increase in d′_max (p_{valid-invalid} < 0.05) with attention. A marginal increase in c₅₀ was also observed (p_{neutral-invalid} = 0.058). No change in n was observed for either the high- or the low-frequency stimulus condition (p_{valid-invalid} > 0.1). 

Figure 6

 

Effects of endogenous attention on performance (d′) as a function of second-order modulator contrast for the low (A) and high (B) second-order spatial frequency conditions (N = 7). All r² > .98. All plotting conventions as in Figure 5.

Figure 6C-D shows the values of c₅₀ and d′_max for individual observers in the valid and invalid attention conditions in which each value is normalized by (i.e., divided by) the corresponding parameter values from the neutral attention condition. For both second-order spatial frequency conditions, valid d′_max values were higher than the neutral d′_max values (i.e., red squares above the unity line), and invalid d′_max values were lower than the neutral d′_max values (i.e., blue squares below the unity line). In addition, we observe that valid c₅₀ values were generally lower and invalid c₅₀ values higher than the neutral c₅₀ values for both stimulus conditions. These results indicate that unlike exogenous attention, endogenous attention can increase second-order contrast sensitivity regardless of the spatial frequency content. 

Eye tracking

Observers were able to voluntarily deploy their attention to the cued location without breaking fixation. Less than 5% of the eye-movement data were missing or unable to be analyzed due to problems with the calibration. Less than 1% of all recorded gaze positions were outside the central fixation area subtending 2° of visual angle. Observers did not move their eyes toward the cued location during cued trials. For each cued trial, we computed average gaze position from cue onset to stimulus offset. We rotated the coordinate system trial-by-trial to one in which the fixation point was at the origin and cued location was displaced 5° to the right. In this coordinate system, in the low second-order spatial frequency condition, the horizontal and vertical gaze positions averaged over cued trials were 0.049° ± 0.29° and 0.004° ± 0.36° respectively, and the x-value was not significantly greater than zero (one-tailed t-test, p > 0.4); i.e., gaze stayed at the fixation mark. The corresponding values for the high second-order spatial frequency condition were 0.039° ± 0.27° and 0.0002° ± 0.28° (x-value: one-tailed t-test, p > 0.4). Saccade frequency during the interval between cue onset and stimulus offset was also extremely low (0.23% and 0.32% in the low and high second-order spatial frequency conditions respectively). Thus, our results are not due to eye movements toward the cued location. 

Discussion

We measured attentional trade-offs in second-order contrast sensitivity at parafoveal locations (5° eccentricity). We compared the effects of exogenous and endogenous attention for second-order orientation-defined patterns of either relatively high or low second-order spatial frequency. The results show that both types of attention affect contrast sensitivity to second-order texture-defined information, increasing sensitivity at the attended location and decreasing it at unattended locations. Moreover, these findings reveal that exogenous attention and endogenous attention differentially affect second-order contrast sensitivity. 

Covert attention enhances second-order processing

Previous work has shown that attention increases contrast sensitivity to first-order information (Carrasco, 2006, 2011; Reynolds & Chelazzi, 2004). The present findings indicate that spatial covert attention can enhance sensitivity to second-order information as well. This finding challenges the idea that second-order processing is purely pre-attentive and does not benefit from attention (Braun & Sagi, 1990). Although attention is not necessary for texture segmentation to occur, texture perception can benefit from the allocation of attentional resources. Performance in our second-order orientation-discrimination task is unaffected by modest changes in first-order carrier contrast (Barbot et al., 2011a). Thus, the attentional effect cannot be due to an effect of attention on effective first-order contrast. Moreover, had the attention effects been due to changes in first-order contrast sensitivity, they should have been the same regardless of the second-order frequency in the exogenous attention experiment. The changes in sensitivity with attention observed in the present study can only be attributed to a change in second-order contrast sensitivity. When covert attention is directed to a given location, our ability to discriminate patterns defined by changes in either first-order luminance or second-order textural attributes improves at that location and worsens elsewhere. 

Attentional trade-offs for both exogenous and endogenous attention emerged for simple and non-cluttered displays in which only two stimuli were competing for processing, supporting the notion of limited resources. This finding challenges the idea that perceptual processes are of unlimited capacity (Palmer, Verghese, & Pavel, 2000) or that selective attention is required only once the perceptual load exceeds the capacity limit of the system (Lavie, 1995). Attention helps manage metabolic consumption in the brain between attended and unattended locations by biasing competition in favor of information at the attended area, at the cost of information processing at other unattended areas. Our findings support the notion that trade-offs are a ubiquitous property of attentional selection that can affect various visual attributes (Abrams, Barbot, & Carrasco, 2010; Carrasco, 2011; Montagna et al., 2009; Pestilli & Carrasco, 2005; Pestilli, Carrasco, Heeger, & Gardner, 2011; Pestilli et al., 2007). 

Segmentation of the visual scene into distinct perceptually coherent regions is crucial for the reliable detection and identification of objects. Object segmentation begins with the detection of discontinuities representing boundaries between adjacent regions, rather than immediate detection of objects per se (Appelbaum, Wade, Pettet, Vildavski, & Norcia, 2008; Li, 2003). Access to multiple visual cues, such as first- and second-order contours, considerably improves performance. When both first- and second-order cues are present, texture segmentation improves (Smith & Scott-Samuel, 2001), but this benefit only occurs when the two cues are correlated in an ecologically valid manner (Johnson, Prins, Kingdom, & Baker, 2007). Perceived contour location is a compromise between the position signaled by second-order texture-defined cues and by other cues such as luminance or motion (Rivest & Cavanagh, 1996). When an edge is defined by multiple cues, the cues are combined using a weighted average, with greater weight given to the more reliable cues (Landy & Kojima, 2001). Considering that many first-order luminance cues originate from non-uniform surface illumination, second-order information appears to represent a more reliable signal of discontinuities between adjacent surfaces (Kingdom, 2003; Schofield et al., 2006; Schofield et al., 2010; Sun & Schofield, 2011). By enhancing sensitivity to first-order cues, but more importantly to the more reliable second-order cues, attention can efficiently improve segmentation of visual scenes into distinct regions, significantly improving object detection and identification. 

On the flexibility and automaticity of covert attention

Exogenous attention increases second-order contrast sensitivity as a function of the second-order spatial frequency content; it affects sensitivity for patterns with relatively high, but not low, second-order spatial frequency. These results for exogenous attention are consistent with our previous findings (Barbot et al., 2011a). As in the present study, for stimuli at parafoveal locations, exogenous attention improved performance only for the high-spatial frequency patterns, but at more eccentric locations, it also improved performance for the low-spatial frequency patterns. Thus, the effects of exogenous attention on second-order contrast sensitivity were a function of both the second-order spatial frequency of the stimulus and its eccentricity. We interpreted these findings in terms of the resolution hypothesis, whereby exogenous spatial attention increases resolution by increasing the sensitivity of the smallest second-order filters at a particular eccentricity (Carrasco, 2011; Carrasco & Yeshurun, 2009). Given that the average filter size, and thus the resolution, decreases with eccentricity, the results were consistent with the fact that the smallest filters at parafovea mediate the high but not the low spatial frequency condition. Conversely, at more eccentric locations, the filters with highest spatial resolution are larger and coarser, and thus are used for the low second-order spatial frequency. Hence, a mechanism affecting the sensitivity of the smallest filters at a given eccentricity could explain why the effect of exogenous attention depended on the modulator frequency of the second-order texture and its eccentricity. 

In the present study, however, endogenous attention increased second-order contrast sensitivity regardless of the second-order spatial frequency content. Taken together, the present findings reveal that both exogenous attention and endogenous attention modulate second-order contrast sensitivity, but via different mechanisms, and that endogenous attention is more flexible. 

Whereas endogenous attention and exogenous attention often modulate perception in similar ways, this study provides further evidence that they can differentially affect perception. Endogenous attention is more flexible, being able to adjust its operation depending on goals and task demands (Carrasco, 2011). For instance, in a texture segmentation task, constrained by spatial resolution, exogenous attention improves performance at peripheral locations but impairs it at central locations (Carrasco et al., 2006; Talgar & Carrasco, 2002; Yeshurun & Carrasco, 1998, 2000, 2008). In contrast, endogenous attention improves performance at all eccentricities (Yeshurun et al., 2008), by adjusting the sensitivity of selective second-order spatial frequency filters to optimize spatial resolution based on task demands (Barbot, Montagna, & Carrasco, 2011b), consistent with a more flexible endogenous-attention system. 

There are also differences in the effects of endogenous and exogenous attention on temporal processing. For instance, endogenous allocation of attention improves temporal order judgments, whereas exogenous allocation of attention impairs judgments (Hein et al., 2006). Furthermore, exogenous attention and endogenous attention have different effects when cue validity is manipulated, consistent with a more flexible endogenous-attention system. Whereas both the benefits (valid trials) and costs (invalid trials) due to endogenous attention increase with cue validity, the effects of exogenous attention are similar across cue-validity conditions (Giordano et al., 2009). Whereas endogenous attention can be flexibly allocated according to cue validity, exogenous attention is automatic and unaffected by cue validity. 

Conclusion

Exogenous and endogenous attention can affect one's ability to discriminate second-order information: Contrast sensitivity to second-order texture-defined patterns increases at the attended location, while decreasing elsewhere. Together with the effects of covert attention on first-order contrast sensitivity, our study suggests that attention aids in the segmentation of the retinal image by increasing both first- and second-order sensitivity at the attended location. Moreover, this study revealed that whereas the effects of exogenous attention are a function of the second-order spatial frequency content, endogenous attention affected second-order contrast sensitivity regardless of the second-order spatial frequency content. These findings support the idea that exogenous and endogenous attention affects perceptual processing via different mechanisms. Endogenous attention is under the observer's voluntary control and can flexibly adjust to optimize performance depending on goals and task demands. 

Acknowledgments

This work was supported by grants NIH R01-EY016200 to MC and R01-EY16165 to MSL.The authors thank members of the Carrasco Lab for comments on an earlier draft of this paper. 

Commercial relationships: none. 

Corresponding author: Antoine Barbot. 

Email: [email protected]. 

Address: Department of Psychology, New York University, New York, NY, USA. 

Footnotes

¹  Data of three of the eight observers for the exogenous attention experiment were published previously (Barbot et al., 2011a).