Exp Brain Res (2008) 189:411–420 DOI 10.1007/s00221-008-1438-1
R ES EA R C H A R TI CLE
The perceptual consequences of the attentional bias: evidence for distractor removal Matthias Niemeier · Vaughan V. W. Singh · Matthew Keough · Nadine Akbar
Received: 20 March 2008 / Accepted: 18 May 2008 / Published online: 7 June 2008 © Springer-Verlag 2008
Abstract A fundamental question of attentional research concerns the perceptual consequences of attention. Spatial attention can enhance stimuli within the focus of attention relative to stimuli outside; or attention can remove the inXuence of distracting stimuli and other forms of external noise inside the focus of attention. It is known that both strategies apply depending on how attention is cued to a location in space. Here we asked which strategy applies in an uncued situation in which people show a spontaneous bias of attention to the left side. To measure bias, we used a gratingscales task with stimuli corrupted by pixel noise. If biased attention resulted in biased stimulus enhancement its eVect should be largest when there is little noise or few distractors within the attended region, and bias should decline with increasing noise. If, however, bias caused distractors to be removed asymmetrically, larger bias should show up with noisy stimuli. We found that bias rose exponentially as noise increased, in agreement with the external noise removal model, and we found evidence that noise modiWed interhemispheric competition between attentional systems. Our data oVer new insights into the neural mechanisms of the right-hemisphere dominance in spatial and attentional tasks. Keywords Spatial neglect · Pseudoneglect · Spatial attention · Interhemispheric competition · Lateralization
M. Niemeier (&) · V. V. W. Singh · M. Keough · N. Akbar Department of Psychology, University of Toronto at Scarborough, 1265 Military Trail, Toronto M1C 1A4Canada e-mail:
[email protected] M. Niemeier Centre for Vision Research, York University, Toronto, Canada
Introduction Spatial attention improves performance. This may happen at multiple levels from decisional stages (e.g., Palmer 1994; Shiu and Pashler 1994) to perceptual mechanisms. At the perceptual stage, attention improves the signal-to-noise ratio, that is, stimuli relevant for the behavioural goal at hand are strengthened by biasing competition with irrelevant things such as distracting stimuli, Xaws in the stimulus display, or errors committed by the sensory system (Desimone and Duncan 1995). The signal-to-noise ratio could increase through attention strengthening signals (but also noise) inside the focus of attention relative to stimuli presented outside (e.g., Cameron et al. 2002; Carrasco et al. 2000; Lu and Dosher 1998), or attention could reduce the inXuence of external noise and, indistinguishable from it, errors in sensory processes (Baldassi and Burr 2000; Dosher and Lu 2000; Morgan et al. 1998; Shiu and Pashler 1994). The two strategies can be tested experimentally as they diVer in the degree to which noise added to the stimulus display modulates the eVect of attention on task performance (e.g., Lu and Dosher 1998). Attention-based stimulus enhancement should yield eVects in situations with few distractors or little external noise inside the focus of attention but disappear as noise increases. External distractor exclusion, on the other hand, should show larger attentional eVects the more the stimulus display is cluttered with unwanted visual information. Which attentional strategy applies depends on the condition. Known is that central cueing of attention is associated with external noise exclusion; peripheral cueing results in both signal enhancement and external distractor exclusion (Lu and Dosher 2000). Here we asked, what are the perceptual consequences of attention under uncued, spontaneous conditions? In
123
412
attentionally neutral situations, without cues, attention is often spontaneously biased to the left side due to righthemisphere dominance for spatial and attentional functions. This has been concluded from neuropsychological studies in patients demonstrating that right-brain damage, more so than left-brain damage, causes severe deWcits of spatial behaviour such as spatial neglect and related dysfunctions (Hillis et al. 2005; Karnath et al. 2001, 2002; Leibovitch et al. 1998; Mort et al. 2003; Rorden et al. 2006; Vallar and Perani 1986). Furthermore, functional studies have conWrmed right-brain dominance for spatial tasks in the intact brain (Corbetta and Shulman 2002; Coull et al. 2001; Fink et al. 2001, 2002; Foxe et al. 2003). Behaviourally, this hemispherical asymmetry manifests as biases in perceptual judgments on horizontally arranged stimuli. For example, in the line bisection task, a classic neglect test, some right-brain damaged patients exhibit pathological rightward biases in their size or midline judgments (e.g., Binder et al. 1992; Heilman and Valenstein 1979; Milner and Harvey 1995; Schenkenberg et al. 1980), and comparable misperceptions occur with luminance (Mattingley et al. 1994, 2004) and numerosity (Luh 1995; Nicholls et al. 1999). Healthy individuals show smaller biases, mostly to the left (Jewell and McCourt 2000, for a review) that vary as a function of visual properties much like what has been found for pathological biases (McCourt and Jewell 1999). Perceptual biases are believed to result from asymmetries in the distribution of spatial attention rather than asymmetries of other visuo-spatial functions because attentional cueing experiments increase or decrease perceptual biases asymmetrically depending on whether they direct attention to the left or to the right side (Bultitude and Aimola Davies 2006; Harvey et al. 1995, 2000; Ishiai et al. 1995; McCourt et al. 2005; Mennemeier et al. 1997; Nichelli et al. 1989; Nicholls and Roberts 2002; Reuter-Lorenz et al. 1990; Reuter-Lorenz and Posner 1990; Riddoch and Humphreys 1983). Further support comes from a recent study that tested uncued biases in spatial frequency judgments of a gratingscales task (Fig. 1a; Niemeier et al. 2008). Participants exhibited a “cross-over” of bias: a leftward bias when examining gratingscales for high spatial frequencies and a rightward bias when attending to low spatial frequencies, precisely as expected based on the observation that attention increases apparent spatial frequencies (Gobell and Carrasco 2005). In the present study we used the gratingscales task, a sensitive measure of attentional bias (Niemeier et al. 2007), combined with diVerent degrees of distracting noise. If spontaneously biased attention improved perception through stimulus enhancement, perceptual biases should be stronger for noiseless gratingscales than for noisy ones. If, however, attentional bias reduced the inXuences of external
123
Exp Brain Res (2008) 189:411–420
Fig. 1 Samples of the gratingscales task. Participants were asked to judge which of the two bars had “more of the thinner stripes”. Black dashed boxes (not presented during testing) indicate the area in which gratings transformed from low spatial frequency to high spatial frequency and vice versa. Stimuli were combined with diVerent degrees of pixel noise. a 0% pixel noise. b 50% noise. c 75% noise. d 87.5% noise
distractors the bias in the gratingscales task should increase with stimulus noise. We found support for the latter. Furthermore, it remains unclear at which perceptual stage noise removal occurs. Early processes might show diVerences for stimuli that favour the magnocellular or parvocellular system (e.g., Morrone et al. 2002; Pitzalis et al. 2005; Spinelli et al. 1996). This we tested in the second experiment. Later, rather unspeciWc mechanisms might involve attentional eVort (Bradshaw et al. 1987; McCourt and Jewell 1999). Experiment 3 was designed to disentangle eVort and stimulus noise. Lastly, we have previously observed that the gratingscales biases show a certain inXuence of presentation time (Niemeier et al. 2007) consistent with the notion that the right-dominance of spatial and attentional functions arises from interhemispheric competition (e.g., Hilgetag et al. 2001; Kinsbourne 1970). Therefore, Experiment 4 investigated the inXuence of noise on the presentation time eVect.
Exp Brain Res (2008) 189:411–420
Methods Participants One hundred and three undergraduate students gave their informed and written consent prior to their participation in the present study and obtained a course credit. Twenty participated in the Wrst experiment (12 females, median age: 18 years), 21 in the second experiment (16 females, median age: 18 years), 21 in the third experiment (13 females, median age: 19 years), and 41 in the fourth experiment (23 females, median age: 18 years). All procedures were approved by the Human Participants Review Sub-Committee of the University of Toronto and have therefore met the ethical standards laid out in the 1964 Declaration of Helsinki. All participants were healthy, had normal or corrected to normal vision, and were right handed as conWrmed with the Edinburgh handedness inventory (OldWeld 1971). Apparatus and procedure Participants sat in front of a 19-in. monitor (Viewsonic E90fb) at a distance of 60 cm. A chin rest was used to keep head movements to a minimum. Eye movements were not restricted though behavioural observations suggested that participants Wxated the centre of the screen. We wrote our experiments in Matlab (MathWorks) using the Psychophysics Toolbox extensions (Brainard 1997; Pelli 1997). In all four experiments we used diVerent versions of the gratingscales task, a novel and sensitive measure of attentional bias (Niemeier et al. 2007). Samples of the gratingscales are given in Fig. 1. Each stimulus consisted of two horizontal bars Wlled with rows of luminance- or colour-deWned wavelets. Spatial frequency of the pattern increased from left to right in one bar, half the time in the upper and half the time in the lower bar, and from right to left in the respective other bar. For example, in the upper bar in Fig. 1a, b spatial frequency increases from left to right and vice versa in the lower bar, and in Fig. 1c, d spatial frequency increases in opposite directions. In the present study, spatial frequencies always ranged from 0.6 to 2 cycles per degree (cpd). Within each gratingscales bar, spatial frequency changed smoothly from low to high within a central area (dashed rectangles in Fig. 1) following a half-cycle of a cosine function (see Niemeier et al. 2007). Left and right of the central area spatial frequency remained constant. This way it was possible to shift the central area to create asymmetrical stimuli that always spanned the same range of spatial frequencies. For example, in Fig. 1b the central area is shifted ¡12.5% leftward relative to stimulus width, and in Fig. 1c it is shifted +12.5% rightward, while the stimuli in Fig. 1a, d are symmetrical. In total there were 11
413
equally spaced degrees of stimulus asymmetry from ¡12.5 to +12.5% (e.g., Fig. 1c). Per condition each of these stimuli was presented 16 times, and every time participants were asked to choose the bar that contained “more of the thinner stripes”. Based on the participants’ responses we calculated probabilities of choosing the bar with the high spatial frequency component on the right side as a function of stimulus asymmetry, and we Wtted sigmoid (Weibull) functions to the participants’ responses to determine the point at which they had no preference for bars with high spatial frequency components on the left or right side. We used this point of subjective equality (PSE) as our measure of perceptual bias. The other free parameter of the Weibull function as used here is its slope which reXects a measure of sensitivity, given the performance of two or more participants in one task. Alternatively, slope can be regarded as an estimate of task diYculty given the performance of one participant in two or more tasks (a more detailed description of the method can be found in Niemeier et al. 2007). We measured PSEs and slopes as a function of noise level. Therefore a certain percentage of random pixels of each gratingscales stimulus was set to a random luminance value between white and black (Exp. 1, 3 and 4) or to a random isoluminant colour value between red and green (Exp. 2). A total of four noise levels were tested: 0, 50, 75, and 87.5% (Fig. 1a–d). In Experiment 1 we tested all 4 noise levels sorted into blocks consisting of 88 trials, 2 for each condition. The order of blocks was pseudorandom for the Wrst half and mirror-reversed for the second. Presentation time was 150 ms. Experiment 2 was similar to Experiment 1, except gratingscales and noise were isoluminant (red–green; the blue gun of the monitor was turned oV). To measure perceived isoluminance we used a common Xicker fusion test. For six combinations of red and green participants adjusted luminance relative to an olive-green reference of intermediate brightness, and we Wtted a Weibull function to the data to interpolate any combination of red and green. This function then served to create isoluminant noise and gratingscales that varied from red to green on an olive-green background similar to the stimuli in Fig. 1. We used noise levels 50, 75, and 87.5%, and there were two presentation times, 150 and 500 ms. The six conditions were tested twice, Wrst in a random and then in the mirror-reversed order. To look at potential inXuences of attentional eVort, Experiment 3 used mixed blocks with 352 trials in which stimuli with a noise level of 50% appeared with a probability of 25%. Three times more frequent were stimuli, called “context stimuli”, with noise levels of 0 or 87.5% to create “easy” and “diYcult” blocks, respectively. The order of
123
414
Exp Brain Res (2008) 189:411–420
conditions was AABBBBAA where A and B represent “easy” and “diYcult” or vice versa. Experiment 4 presented luminance deWned gratingscales with presentation times 75, 150, 300, and 600 ms, and we chose noise levels of 0 and 75% (noisier stimuli might have caused fatigue and raised variability in performance given the larger number of noisy trials). The resulting eight conditions were tested twice in the usual half random, half mirror-reversed order.
Results Experiment 1: inXuence of pixel noise on the attentional bias Experiment 1 presented gratingscales stimuli that diVered in noisiness. We observed that on average leftward bias increased as a function of pixel noise (Fig. 2a). A repeated measures ANOVA revealed a signiWcant eVect of noise level [F(1.76, 33.52) = 5.07, P = 0.014 after Greenhouse– Geisser correction of the degrees of freedom, “GGc”]; and a trend analysis suggested this inXuence to be exponential [F(1, 19) = 6.39, P = 0.021]. Subsequent t tests indicated a trend for a stronger leftward bias for 87.5% noise relative to the other conditions [ts(21) ¸ 2.51, Ps · 0.021, not signiWcant after Holm’s correction]. Noise also showed a trend for an eVect on task diYculty in that diYculty increased (and slopes of the psychometric functions decreased) numerically with rising noise (Fig. 2b). Though the respective ANOVA conducted on slopes did not conWrm this eVect to be signiWcant [F(1.87, 35.60) = 1.69, P = 0.200 after GGc], we repeated the Wrst ANOVA of bias in the form of an ANCOVA with covariate slope (averaged across noise levels) to rule out the possibility that the increase in bias was due to greater task demands. Still the inXuence of noise on bias remained signiWcant [F(1.81, 32.55) = 4.66, P = 0.019 after GGc] consistent with the idea that perceptual bias is governed by mechanisms of noise reduction.
Fig. 2 Results of Experiment 1. a Group average of perceptual bias as a function of pixel noise. b Group average of the slope of the psychometric functions. Error bars represent standard errors
Experiment 2: isoluminant gratingscales As one possible interpretation, the noise eVect on bias could depend on early visual mechanisms of noise removal within the magnocellular system. If noisy stimuli favoured the magnocellular system somehow causing bias to rise, then isoluminant stimuli and isoluminant noise should favour the parvocellular system, and there should be no increase of bias. However, in Experiment 2 we did Wnd biases to increase as isoluminant noise increased (Fig. 3a) much like what we had observed in Experiment 1. We conWrmed this with a 2-way ANOVA. Factor “noise level” was signiWcant [F(1.09, 21.84) = 14.93, P < 0.001 after GGc], and had an
123
exponential plus quadratic-exponential inXuence as revealed through a trend analysis [F(1, 20) = 15.78; P < 0.001 and F(1, 20) = 11.02, P = 0.003, respectively] with signiWcant t tests between the three noise levels (ts(20) ¸ 2.39, Ps · 0.027, all signiWcant after Holm’s correction). In contrast, factor “presentation time” and the interaction were insigniWcant [“time”: F(1, 20) = 0.33, P = 0.574; “interaction”: F(1.19, 23.80) = 0.08, P = 0.819 after GGc]. Also, similar to Experiment 1, a 2-way ANOVA on slope suggested an increase in task diYculty as a function
Exp Brain Res (2008) 189:411–420
415
inXuence of noise reduced to an insigniWcant trend [F(1.08, 20.49) = 2.92, P = 0.100]. However, the interaction between “noise level” and covariate slope was not signiWcant either [F(1.08, 20.49) = 0.21, P = 0.668]. So including a covariate might simply have reduced statistical power. Indeed, slopes and biases for all six conditions correlated only poorly (rs · 0.259, Ps ¸ 0.258), and pooling across presentation times reduced correlations further (rs · 0.009, Ps ¸ 0.970). Experiment 3: inXuence of eVort on the attentional bias As an indirect consequence of task diYculty, participants might have increased their attentional eVort during very noisy conditions which could have caused stronger biases. So far we might have overlooked the relationship because our analyses on slope examined between-subject eVects and also did not manipulate eVort systematically. Therefore, in Experiment 3 we used stimuli with medium noise (50%) either in a very noisy context where 75% of the time the gratingscales had 87.5% noise or in a low-noise context where 75% of the stimuli had 0% noise. However, context did not cause diVerences in bias [t(20) = ¡0.67, P = 0.511; bars in Fig. 4a]. If any the trend was in the opposite direction. Likewise, context did not inXuence task diYculty [t(20) = ¡0.37, P = 0.712; bars in Fig. 4b]. Experiment 4: presentation time and visual noise
Fig. 3 Results of Experiment 2. a Group average of perceptual bias. b Group average of slope. Black squares 150 ms presentation time, grey squares 500 ms presentation time. Error bars represent standard errors
of noise [F(1.28, 25.59) = 39.51, P < 0.001; Fig. 3b] reXecting an exponential eVect [F(1, 20) = 50.61, P < 0.001] with signiWcant diVerences between all noise levels [ts(20) ¸ 3.46, Ps · 0.002, all tests signiWcant after Holm’s correction]. In addition, shorter presentation time increased task diYculty [F(1, 20) = 13.50, P = 0.002]. But there was no interaction with noise level [F(1.86, 37.16) = 2.02, P = 0.150]. To estimate the extent to which bias resulted from task diYculty, we repeated the ANOVA of bias with “slope” (averaged across conditions) as covariate, and we found the
Our data so far support the idea that perceptual bias is associated with mechanisms of noise reduction; and we have ruled out early visual mechanisms as well as unspeciWc cognitive inXuences of attentional eVort. Also, there was no (simple) relationship between bias and noise-dependent task diYculty. However, we have recently found that participants performing the gratingscales task with more or less diYculty diVered in perceptual bias across presentation times (Niemeier et al. 2007). Poor performers showed stronger biases for 75 and 300 ms and were less biased for 150 and 600 ms, and good performers showed the opposite pattern, perhaps indicating competition between two neural systems associated with perceptual bias. Here we reasoned, if diVerences in task diYculty were due to individual diVerences in noise reduction then increasing stimulus noise should push everybody’s performance towards that of poor performers resulting in a similar presentation-time dependent pattern of bias. Therefore, in Experiment 4 we chose the same presentation times as used before (Niemeier et al. 2007) and two noise conditions. As predicted, bias varied as a function of presentation time, especially for noisy stimuli (Fig. 5a), showing about the same peaks and minima observed previously in poor performers only (Niemeier et al. 2007). We
123
416
Fig. 4 Results of Experiment 3. a Group average of perceptual bias. b Group average of slope. Bars indicate data for the 50% noise stimuli presented either in a context with mostly noise-less stimuli (white bars and circles) or with mostly noisy stimuli (grey bars and circles). Circles indicate averages for context inducing stimuli. Error bars represent standard errors
conWrmed this observation in a series of analyses. First, a 2-way ANOVA yielded signiWcant inXuences of time [F(2.47, 98.73) = 3.27, P = 0.032 after GGc] and noise [F(1, 40) = 13.42, P < 0.001], though there was no interaction [F(3, 120) = 1.28, P = 0.283]. For a planned comparison of time eVects we then averaged across peaks and minima, respectively, and found the diVerence to be signiWcant [t(40) = 2.84, P = 0.007] while neither peaks [t(40) = 1.18, P = 0.246] nor minima [t(40) = 0.87, P = 0.388] diVered from each other.
123
Exp Brain Res (2008) 189:411–420
Fig. 5 Results of Experiment 4. a Group average of perceptual bias as a function of presentation time. b Group average of slope. Black circles data for 0% noise, white circles 75% noise. Error bars represent standard errors
Next, we repeated the initial ANOVA of bias as an ANCOVA using slope averaged across conditions as covariate to reduce inXuences of task diYculty. This resulted in a signiWcant eVect of time [F(2.48, 96.69) = 4.53, P = 0.008 after GGc]. Noise level had no inXuence [F(1, 39) = 1.25, P = 0.271]. More importantly, however, the time-by-noise interaction was signiWcant [F(3, 117) = 3.25, P = 0.024], and subsequent 1-way ANCOVAs revealed that this reXected time-dependent variations of bias in the noisy condition [75% noise: F(3, 117) = 4.56,
Exp Brain Res (2008) 189:411–420
P = 0.005; 0% noise: F(2.34, 91.30) = 1.149, P = 0.327 after GGc] with peaks, again, diVering signiWcantly from minima [F(1, 39) = 13.53, P < 0.001] and no diVerence among peaks or minima, respectively [peaks: F(1, 39) = 2.24, P = 0.142; minima: F(1, 39) = 0.01, P = 0.913]. An additional 2-way ANOVA conducted on slope revealed all three eVects to be signiWcant [time: F(2.60, 104.16) = 19.28, P < 0.001; noise: F(1, 40) = 91.84, P < 0.001; interaction: F(2.19, 87.49) = 5.68, P = 0.004, Fig. 5b]. The time eVect reXected an increase in slopes with presentation time. This was signiWcant for 600 ms relative to other times [ts(40) ¸ 4.39, Ps · 8.17e¡5) but missed signiWcance for 150 and 300 ms relative to 75 ms [ts(40) ¸ 2.11, Ps · 0.041, not signiWcant after Holm’s correction]. Slopes for 150 and 300 ms were essentially identical [t(40) = 0.11, P = 0.916]. Lastly, the interaction could be explained by signiWcant time eVects [0%: F(2.33, 93.16), P < 0.001; 75%: F(2.25, 89.91) = 8.59, P < 0.001] that diVered depending on noise level; for the no-noise condition it was cubic and for the noisy condition quadratic [Fs(1, 40) ¸ 6.42, Ps · 0.015] in addition to linear trends in both cases [Fs(1, 40) ¸ 7.46, Ps · 0.009].
Discussion What are the perceptual consequences of the attentional bias? In the present study we tested two recently discussed models (e.g., Dosher and Lu 2000). One suggests that attention enhances sensory input, composed of relevant and irrelevant information, within the focus of attention relative to other areas in the visual Weld. The second model proposes that attention reduces the inXuence of distractors within the focus of attention. To measure attentional bias we used the gratingscales task as a sensitive test of bias (Niemeier et al. 2007), and we added noise to the stimuli. According to the Wrst model this should impact the perceptual beneWts of attention in attended relative to less attended regions of the stimulus. In other words, bias should decline. However, according to the second model, bias should increase with noise. We observed bias to increase, in support of the idea that attention reduces the inXuence of distractors together with noise in early sensory processes. Could other sources of noise and noise reduction play a role? Of course, errors and noise occur at any processing stage of the brain, and attention might reduce such internal noise, for example by allocating additional neural resources to reduce imperfections in the neural representation of the stimulus. Depending on how those representations are implemented, internal noise could either be constant or stimulus-dependent. In the Wrst case, Lu and Dosher (1999) have shown that attentional eVects would decline with rising external noise just like predicted for the signal enhance-
417
ment model, and therefore bias would decline too. In the second case, attentional eVects would remain of constant size regardless of external noise (Lu and Dosher 1999). Both possibilities are in disagreement with the present data. Could our results be due to eVects speciWc to the magnocellular system? This might be expected because the magno- and parvocellular pathways are independently involved in some attentional mechanisms (Morrone et al. 2002), and diVerences between the two systems have been reported for some patients with right-brain deWcits (Pitzalis et al. 2005; Spinelli et al. 1996). However, when we used isoluminant stimuli in Experiment 2 we observed the same noise eVect on bias as found in the Wrst experiment. This could indicate that magno- and parvocellular pathways perform similar attentional functions. Yet, a more parsimonious explanation is that the attentional inXuence relevant for the attentional bias is conveyed through mechanisms downstream from early visual areas where the separation between the magno- and parvocellular systems is lost. One kind of cognitive, yet unspeciWc inXuence could be attentional eVort. That is, participants might have increased their eVorts when examining diYcult, noisy stimuli, and somehow that could have biased attention further. We tested this possibility in Experiment 3. We presented gratingscales with medium noise either in a context with very noisy stimuli or with stimuli with no added noise. However, on average context did not aVect bias. Perhaps, diYcult contexts discouraged some of the participants such that they made less eVort with noisy stimuli. This could explain why we found no eVect in Experiment 3, but not why we did Wnd an eVect in Experiments 1 and 2; noise should have discouraged some of those participants too. Alternatively, Experiment 3 might have manipulated eVort too weakly. But then, Experiments 1 and 2 do not seem to have manipulated eVort more strongly. So, our data do not support the hypothesis that the eVect of noise is caused by increased attentional processing. Two previous studies have reported evidence for an eVect of attentional eVort. In variants of the line bisection task they manipulated contrast on one side of the stimulus only and found that bias shifts left- or rightward depending on which stimulus side had lower contrast, potentially demanding more attention or inducing covert shifts of attention (Bradshaw et al. 1987; McCourt and Jewell 1999). Our Wndings do not contradict those previous ones because we manipulated visibility of the entire stimulus, not only one side. However, the observed inXuence of distracting noise on bias might be at variance with one Wnding regarding luminance contrasts. Foxe et al. (2003) presented pre-bisected lines at diVerent contrasts (of the entire stimulus) and observed no behavioural change in bias. This could, of course, reXect that Foxe et al.’s (2003) data would have
123
418
required further testing. However, the eVect that we observed was quite strong and not easily overlooked. For example, observed power in Experiment 1 was 0.75, and even when we analyzed only 10 randomly chosen data sets out of the 20 in total, the eVect was close to signiWcant (median P value = 0.059 based on 60 repetitions of the ANOVA with diVerent random draws of participants). Could the disagreement with Foxe et al. (2003) be explained by the fact that they randomized the order of contrast levels to minimize contrast adaptation while we sorted trials into blocks of diVerent distractor noise levels? Noisy gratingscales had reduced contrast for luminance if sampled locally across neighbouring pixels. So in our study contrast adaptation might have been more likely. However, Okubo and Nicholls (2005) have found that interhemispheric diVerences in contrast gain control show up in mixed contrast conditions, not in blocked conditions. Given this, Foxe et al. should have observed an eVect equal to or even greater than ours. More likely is that reducing luminance contrast is a weaker manipulation of signal-to-noise ratio than adding pixel noise to the stimulus. Of course, no stimulus is entirely noise free and it still has to pass through noisy sensors, but both noise sources might be much smaller than the noise that we added to the gratingscales (also note that bias rose only for noise of more than 50%, Figs. 1a, 2a). Therefore, the null eVect of low contrast is in agreement with the conclusion that attention removes distractors and external noise in a spatially biased way. Our results allow for novel insights into the mechanisms underlying attentional bias. First, we argue that attentional bias probably reXects sustained rather than transient forms of attention given that distractor removal is the signature of the former (Lu and Dosher 2000). Evidence for an independence of the two forms of attention comes from behavioural (Deubel 1995; Ling and Carrasco 2006), neuropsychological (Niemeier and Karnath 2003), and imaging studies (e.g., Corbetta and Shulman 2002), and there is reason to assume that for temporal tasks sustained attention is rather associated with the right hemisphere as well (Okubo and Nicholls 2008). Second, our data might shed light on the neural processes that bring about the attentional bias. That is, we have previously proposed that the attentional bias involves two competing systems, one imposing a stronger leftward bias than the other. We concluded this from the Wnding that participants performing the gratingscales task with more or less diYculty diVered in perceptual bias across presentation times in a mirror-symmetric, yet consistently left-biased manner: “poor” or “task-insensitive” performers were more biased at all times, and they showed peaks of bias at times when “good” or “task-sensitive” performers had minimum bias and vice versa (Niemeier et al. 2007).
123
Exp Brain Res (2008) 189:411–420
Our present data oVer an explanation for the inXuence of task performance. Poor performers compared to good performers might have noisier sensors resulting in higher levels of incoming noise. This would make the task more diYcult and, at the same time, it should favour the attentional system involved in external noise removal. In contrast, good performers would have to handle less incoming noise with little need for noise removal resulting in a smaller bias in the gratingscales task. So then, good performers should behave like poor performers as levels of external noise increase. This is consistent with what we observed in Experiment 4. We found that with noisy stimuli all participants, on average, showed peaks and minima like those of poor performers before, consistent with the idea that noise sways competition in favour of the left-biasing system. The competition could arise from interactions between systems in the left and the dominant right hemisphere (Hilgetag et al. 2001), or competition could occur within the right hemisphere between one system that is more dominantly represented in the right hemisphere than the other. For example, Corbetta and Shulman (2002) have suggested that attentional tasks involve one dorsal frontoparietal attentional network and one ventral frontoparietal attentional network, the latter being more right-dominant than the former. Third, one way to perform attentionally biased noise removal would be to Wne-tune the properties of early sensory Wlters, either individual neurons or neural populations, such that they respond more robustly to irrelevant noise and become more sensitive to the relevant signal (e.g., Lu and Dosher 1998). As one example, if a task requires perception of gratings (e.g., the gratingscales task) the best detection mechanism would be sensitive to a certain range of spatial frequencies (of course, more complex mechanisms would be necessary to detect more complex stimuli yet they would still be regarded as Wlters). Narrowing the Wlters’ sensitivity range for spatial frequencies (or other features) would eVectively remove incoming noise. As one potential consequence, apparent spatial frequency of stimuli might increase. There is some evidence that this actually happens (Gobell and Carrasco 2005; Lee et al. 1999; but see Lu and Dosher 2004). Given this, the notion of attentional bias removing external noise in a biased fashion is also consistent with our previous Wnding that, paradoxically, bias crosses over to the right side when participants compare low spatial frequencies instead of high spatial frequencies. Further research will be required to conWrm this idea. In conclusion, in the present study we report evidence that the perceptual consequence of the attentional bias is removal of external noise from distractors, imperfect stimuli and errors in early sensors, preferably in the left relative to the right visual Weld. Other perceptual consequences of
Exp Brain Res (2008) 189:411–420
signal enhancement or internal noise reduction seem to be of less importance. Further research will be required to test whether other kinds of distractors yield the same eVects and whether other non-attentional asymmetries contribute to the distractor eVects. For now we rule out alternative explanations such as processing diVerences in early, magno- and parvocellular systems or later inXuences of attentional eVort. Our data support the notion that attentional bias arises from mutually competitive neural systems. For the future, the question as to where and how this competition occurs will reveal important insights into the functions and dysfunctions of the right hemisphere.
References Baldassi S, Burr DC (2000) Feature-based integration of orientation signals in visual search. Vision Res 40:1293–1300 Binder J, Marshall R, Lazar R, Benjamin J, Mohr JP (1992) Distinct syndromes of hemineglect. Arch Neurol 49:1187–1194 Bradshaw JL, Nathan G, Nettleton NC, Wilson L, Pierson J (1987) Why is there a left side underestimation in rod bisection? Neuropsychologia 25:735–738 Brainard DH (1997) The Psychophysics Toolbox. Spat Vis 10:433– 436 Bultitude JH, Aimola Davies AM (2006) Putting attention on the line: investigating the activation–orientation hypothesis of pseudoneglect. Neuropsychologia 44:1849–1858 Cameron EL, Tai JC, Carrasco M (2002) Covert attention aVects the psychometric function of contrast sensitivity. Vision Res 42:949– 967 Carrasco M, Penpeci-Talgar C, Eckstein M (2000) Spatial covert attention increases contrast sensitivity across the CSF: support for signal enhancement. Vision Res 40:1203–1215 Corbetta M, Shulman GL (2002) Control of goal-directed and stimulus-driven attention in the brain. Nat Rev Neurosci 3:201– 215 Coull JT, Nobre AC, Frith CD (2001) The noradrenergic alpha2 agonist clonidine modulates behavioural and neuroanatomical correlates of human attentional orienting and alerting. Cereb Cortex 11:73–84 Desimone R, Duncan J (1995) Neural mechanisms of selective visual attention. Annu Rev Neurosci 18:193–222 Deubel H (1995) Separate adaptive mechanisms for the control of reactive and volitional saccades. Vision Res 35:3529–3540 Dosher BA, Lu ZL (2000) Noise exclusion in spatial attention. Psychol Sci 11:139–146 Fink GR, Marshall JC, Weiss PH, Zilles K (2001) The neural basis of vertical and horizontal line bisection judgments: an fMRI study of normal volunteers. Neuroimage 14:S59–S67 Fink GR, Marshall JC, Weiss PH, Toni I, Zilles K (2002) Task instructions inXuence the cognitive strategies involved in line bisection judgements: evidence from modulated neural mechanisms revealed by fMRI. Neuropsychologia 40:119–130 Foxe JJ, McCourt ME, Javitt DC (2003) Right hemisphere control of visuospatial attention: line-bisection judgments evaluated with high-density electrical mapping and source analysis. Neuroimage 19:710–726 Gobell J, Carrasco M (2005) Attention alters the appearance of spatial frequency and gap size. Psychol Sci 16:644–651 Harvey M, Milner AD, Roberts RC (1995) An investigation of hemispatial neglect using the landmark task. Brain Cogn 27:59–78
419 Harvey M, Pool TD, Roberson MJ, Olk B (2000) EVects of visible and invisible cueing procedures on perceptual judgments in young and elderly subjects. Neuropsychologia 38:22–31 Heilman KM, Valenstein E (1979) Mechanisms underlying hemispatial neglect. Ann Neurol 5:166–170 Hilgetag CC, Theoret H, Pascual-Leone A (2001) Enhanced visual spatial attention ipsilateral to rTMS-induced ‘virtual lesions’ of human parietal cortex. Nat Neurosci 4:953–957 Hillis AE, Newhart M, Heidler J, Barker PB, Herskovits EH, Degaonkar M (2005) Anatomy of spatial attention: insights from perfusion imaging and hemispatial neglect in acute stroke. J Neurosci 25:3161–3167 Ishiai S, Seki K, Koyama Y, Okiyama R (1995) EVects of cueing on visuospatial processing in unilateral spatial neglect. J Neurol 242:367–373 Jewell G, McCourt ME (2000) Pseudoneglect: a review and meta-analysis of performance factors in line bisection tasks. Neuropsychologia 38:93–110 Karnath HO, Ferber S, Himmelbach M (2001) Spatial awareness is a function of the temporal not the posterior parietal lobe. Nature 411:950–953 Karnath HO, Himmelbach M, Rorden C (2002) The subcortical anatomy of human spatial neglect: putamen, caudate nucleus and pulvinar. Brain 125:350–360 Kinsbourne M (1970) A model for the mechanism of unilateral neglect of space. Trans Am Neurol Assoc 95:143–146 Lee DK, Itti L, Koch C, Braun J (1999) Attention activates winner-takeall competition among visual Wlters. Nat Neurosci 2:375–381 Leibovitch FS, Black SE, Caldwell CB, Ebert PL, Ehrlich LE, Szalai JP (1998) Brain-behavior correlations in hemispatial neglect using CT and SPECT: the Sunnybrook stroke study. Neurology 50:901–908 Ling S, Carrasco M (2006) Sustained and transient covert attention enhance the signal via diVerent contrast response functions. Vision Res 46:1210–1220 Lu ZL, Dosher BA (1998) External noise distinguishes attention mechanisms. Vision Res 38:1183–1198 Lu ZL, Dosher BA (1999) Characterizing human perceptual ineYciencies with equivalent internal noise. J Opt Soc Am A Opt Image Sci Vis 16:764–778 Lu ZL, Dosher BA (2000) Spatial attention: diVerent mechanisms for central and peripheral temporal precues? J Exp Psychol Hum Percept Perform 26:1534–1548 Lu ZL, Dosher BA (2004) Spatial attention excludes external noise without changing the spatial frequency tuning of the perceptual template. J Vis 4:955–966 Luh KE (1995) Line bisection and perceptual asymmetries in normal individuals: what you see is not what you get. Neuropsychology 9:435–448 Mattingley JB, Bradshaw JL, Nettleton NC, Bradshaw JA (1994) Can task speciWc perceptual bias be distinguished from unilateral neglect? Neuropsychologia 32:805–817 Mattingley JB, Berberovic N, Corben L, Slavin MJ, Nicholls ME, Bradshaw JL (2004) The greyscales task: a perceptual measure of attentional bias following unilateral hemispheric damage. Neuropsychologia 42:387–394 McCourt ME, Jewell G (1999) Visuospatial attention in line bisection: stimulus modulation of pseudoneglect. Neuropsychologia 37:843–855 McCourt ME, Garlinghouse M, Reuter-Lorenz PA (2005) Unilateral visual cueing and asymmetric line geometry share a common attentional origin in the modulation of pseudoneglect. Cortex 41:499–511 Mennemeier M, Vezey E, Chatterjee A, Rapcsak SZ, Heilman KM (1997) Contributions of the left and right cerebral hemispheres to line bisection. Neuropsychologia 35:703–715
123
420 Milner AD, Harvey M (1995) Distortion of size perception in visuospatial neglect. Curr Biol 5:85–89 Morgan MJ, Ward RM, Castet E (1998) Visual search for a tilted target: tests of spatial uncertainty models. Q J Exp Psychol A 51:347–370 Morrone MC, Denti V, Spinelli D (2002) Color and luminance contrasts attract independent attention. Curr Biol 12:1134–1137 Mort DJ, Malhotra P, Mannan SK, Rorden C, Pambakian A, Kennard C, Husain M (2003) The anatomy of visual neglect. Brain 126:1986–1997 Nichelli P, Rinaldi M, Cubelli R (1989) Selective spatial attention and length representation in normal subjects and in patients with unilateral spatial neglect. Brain Cogn 9:57–70 Nicholls MER, Roberts GR (2002) Can free-viewing perceptual asymmetries be explained by scanning, pre-motor or attentional biases? Cortex 38:113–136 Nicholls MER, Bradshaw JL, Mattingley JB (1999) Free-viewing perceptual asymmetries for the judgement of brightness, numerosity and size. Neuropsychologia 37:307–314 Niemeier M, Karnath HO (2003) Stimulus-driven and voluntary saccades are coded in diVerent coordinate systems. Curr Biol 13:585–589 Niemeier M, Stojanoski B, Greco AL (2007) InXuences of time and spatial frequency on the perceptual bias: evidence for competition between hemispheres. Neuropsychologia 45:1029–1040 Niemeier M, Stojanoski B, Singh V, Chu E (2008) Paradoxical crossover due to attention to high or low spatial frequencies. Brain Cogn (in press) Okubo M, Nicholls MER (2005) Flexible contrast gain control in the right hemisphere. Brain Cogn 59:269–276 Okubo M, Nicholls MER (2008) Hemispheric asymmetries for temporal information processing: Transient detection versus sustained monitoring. Brain Cogn 66:168–175
123
Exp Brain Res (2008) 189:411–420 OldWeld RC (1971) The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9:97–113 Palmer J (1994) Set-size eVects in visual search: the eVect of attention is independent of the stimulus for simple tasks. Vision Res 34:1703–1721 Pelli DG (1997) The VideoToolbox software for visual psychophysics. Transforming numbers into movies. Spat Vis 10:437–442 Pitzalis S, Di Russo F, Spinelli D (2005) Loss of visual information in neglect: The eVect of chromatic- versus luminance-contrast stimuli in a “what” task. Exp Brain Res 163:527–534 Reuter-Lorenz PA, Posner MI (1990) Components of neglect from right-hemisphere damage: an analysis of line bisection. Neuropsychologia 28:327–333 Reuter-Lorenz PA, Kinsbourne M, Moscovitch M (1990) Hemispheric control of spatial attention. Brain Cogn 12:240–266 Riddoch MJ, Humphreys GW (1983) The eVect of cueing on unilateral neglect. Neuropsychologia 21:589–599 Rorden C, Fruhmann Berger M, Karnath HO (2006) Disturbed line bisection is associated with posterior brain lesions. Brain Res 1080:17–25 Schenkenberg T, Bradford DC, Ajax ET (1980) Line bisection and unilateral visual neglect in patients with neurologic impairment. Neurology 30:509–517 Shiu LP, Pashler H (1994) Neglible eVect of spatial precuing on identiWcation of single digits. J Exp Psychol Hum Percept Perform 20:1037–1054 Spinelli D, Angelelli P, De Luca M, Burr DC (1996) VEP in neglect patients have longer latencies for luminance but not for chromatic patterns. Neuroreport 7:815–819 Vallar G, Perani D (1986) The anatomy of unilateral neglect after right-hemisphere stroke lesions. A clinical/CT-scan correlation study in man. Neuropsychologia 24:609–622