Several studies with significant C1 attention effects survive critical analysis

Transcription

1 COGNITIVE NEUROSCIENCE, 2018 VOL. 9, NOS. 1 2, AUTHOR RESPONSE Several studies with significant C1 attention effects survive critical analysis Scott D. Slotnick Department of Psychology, Boston College, Chestnut Hill, MA, USA ABSTRACT In a discussion paper (Slotnick, this issue), I conducted a selective review of spatial attention studies to compare experimental parameters and determine whether particular stimulus, task, or analysis conditions were more likely to produce significant attentional modulation of the eventrelated potential (ERP) C1 component. It was concluded that to maximize C1 attention effects, stimuli should be in the upper visual field, there should be distractors, conditions should be high perceptual or attentional load, there should be exogenous cuing, and effects should be measured at midline parietal-occipital electrodes POz, Pz, and CPz. Commentaries were received by Fu (this issue), Qu and Ding (this issue), Zani and Proverbio (this issue), Pitts and Hillyard (this issue), Di Russo (this issue), and Mohr and Kelly (this issue). Comments included additional ideas to amplify C1 attention effects, support for some conclusions, and challenges to some conclusions. The challenges led to a more in depth analysis of many issues pertaining to C1 attention effects including optimal electrode and stimulus locations, null V1 source localization attention effects, whether all significant C1 attention effects can be discounted, and the number of studies with null versus significant C1 attention effects. Analysis of the studies that survived critical analysis, which included several that observed significant C1 attention effects, led to the same conclusions as Slotnick (this issue). Lines of future research include replicating studies that have observed C1 attention effects using identical experimental parameters and systematically manipulating parameters to determine the impact of each parameter on C1 spatial attention effects. ARTICLE HISTORY Recieved 15 October 2017 Revised 19 October 2017 Published online 8 November 2017 KEYWORDS Attention; spatial attention; C1; P1; V1; ERP In Slotnick (this issue), I conducted a selective review of spatial attention studies to compare experimental parameters and determine whether particular stimulus, task, or analysis conditions were more likely to produce significant attentional modulation of the event-related potential (ERP) C1 component. It was concluded that to maximize C1 attention effects, stimuli should be in the upper visual field, there should be distractors, conditions should be high perceptual or attentional load, there should be exogenous cuing, and effects should be measured at midline parietal-occipital electrodes POz, Pz, and CPz. Six commentaries on that paper were received by Fu (this issue), Qu and Ding (this issue), Zani and Proverbio (this issue), Pitts and Hillyard (this issue), Di Russo (this issue), and Mohr and Kelly (this issue). All of the commentaries were excellent and included new ideas, support for some of my conclusions, and criticism of some of my conclusions. All of the comments are considered below, which led to a more critical analysis of many studies in the selective review and the addition of a couple other studies. Although the results of the slightly revised review led to the same conclusions as Slotnick (this issue), the commentaries and my responses to them have clarified many issues on the topic of C1 attentional effects and provided some future research directions. Additional ideas to amplify C1 attention effects Three of the commentaries provided suggestions for additional experimental parameters or analysis techniques not considered in Slotnick (this issue) that may amplify C1 attention effects. Fu proposed that stimulus parameters should be adjusted such that the C1 amplitude is not too low or too high such that attention effects can be observed (i.e., floor and ceiling effects should be avoided). Fu also proposed that certain participants could be excluded (e.g., those with a small magnitude or no C1, i.e., low signal, or strong alpha, i.e., high noise) or included (e.g., athletes) to maximize C1 attention effects. With regard to studies that employ exogenous cuing, Fu highlighted that the optimal stimulus onset CONTACT Scott D. Slotnick sd.slotnick@bc.edu Department of Psychology, Boston College, McGuinn Hall, Chestnut Hill, MA 02467, USA 2017 Informa UK Limited, trading as Taylor & Francis Group

2 76 S. D. SLOTNICK asynchrony (SOA) to maximize C1 attention effects has yet to be identified. Also with regard to exogenous cuing studies, Qu and Ding suggested subtracting ERP effects associated with upper visual field stimulation and lower visual field stimulation to cancel out cue related activity that might contaminate valid trial activity (exogenous cue effects are discussed in more detail below). This method assumes that cue-related activity is identical for upper visual field stimuli and lower visual field stimuli and thus could be subtracted out; however, this assumption is questionable as the magnitude and topography of ERP components are known to vary for upper versus lower visual field stimuli (see Figure 1 in Slotnick, this issue). Zani and Proverbio correctly pointed out that the conclusion that exogenous cueing should be employed to maximize C1 attention effects was based on only spatial attention studies. They proposed that endogenous cueing may maximize C1 attention effects for non-spatial (e.g., object-based) attention studies. Providing some counterevidence for this proposal, one study that employed exogenous cuing during non-spatial/surface-based attention observed significant C1 attention effects (Khoe, Mitchell, Reynolds, & Hillyard, 2005). All of the suggestions above are topics of future research. Optimal electrode locations to measure C1 attention effects In Slotnick (this issue), studies were only included in the primary analysis that measured C1 attention effects at or near midline parietal-occipital electrodes in an effort to minimize the influence of early P1 attention effects, which are maximal at lateral occipital electrodes. Bases on the electrode locations that showed significant C1 attention effects in these studies, it was concluded that such effects should be measured at midline parietal-occipital electrodes POz, Pz, and CPz. In their commentaries, Pitts and Hillyard along with Di Russo correctly pointed out that upper visual field stimuli typically produce a C1 maximum that is ipsilateral to the stimulus and lower visual field stimuli typically produce a C1 maximum that is contralateral to the stimulus. Of importance, the C1 maximum is only slightly lateral to the midline (see Figure 1 in Slotnick, this issue). As can be seen in Table 1 of Slotnick (this issue), these slightly lateral electrodes (i.e., PO1, PO2, P1, P2, CP1, CP2) were considered in the selective review. For upper visual field stimuli, which produce an ipsilateral C1 maximum, slightly off-midline electrodes may be more optimal to measure C1 attention effects than midline electrodes. For instance, for left upper visual field stimuli, ipsilateral electrodes PO1, P1, and CP1, may be more sensitive than electrodes POz, Pz, and CPz. For lower visual field stimuli, caution must be taken, as measuring C1 attention effects at contralateral electrode locations will increase the likelihood of picking up contralateral early P1 attention effects. In Slotnick (this issue), studies were not included in the primary analysis that measured C1 attention effects at lateral occipital electodes in an effort to avoid early P1 attention effects. Of particular relevance, the studies of Kelly, Gomez-Ramirez, and Foxe (2008) and Baumgartner, Graulty, Hillyard, and Pitts (this issue) were excluded from the primary analysis. In these studies, upper visual field C1 attention effects were measured at ipsilateral occipital electrodes, which I argued may have reflected the ipsilateral voltage sink corresponding to the contralateral early P1 voltage source, and lower visual field C1 attention effects were measured at contralateral occipital electrodes, which I argued may have reflected the contralateral early P1 to some degree. There were a number of arguments in the commentaries against the claim that upper visual field C1 effects measured at ipsilateral occipital electrodes may have reflected the early P1 voltage sink. Critically, the issue is not whether the upper visual field ipsilateral C1 reflects only the early contralateral P1 sink, but rather whether the C1 reflects the P1 sink to some degree, such that significant V1 attention effects may actually reflect extrastriate attention effects. Said another way, the burden of proof is on those who propose to measure C1 attention effects at lateral occipital electrodes to convincingly show that these effects do not reflect any P1 attention effects. Pitts and Hillyard stated that dipole modeling has shown that the upper visual field C1 source is separate from the diffuse far-field sink of the P1 source (citing Clark, Fan, & Hillyard, 1994, and Di Russo, Martínez, & Hillyard, 2003). The dipole modeling results from these studies (see Figure 9 in Clark et al., 1994, and Figures 7 and 8 in Di Russo et al., 2003) show the C1 dipole is in medial cortex (corresponding to V1) and the P1 dipole is in contralateral occipital

3 COGNITIVE NEUROSCIENCE 77 cortex (corresponding to extrastriate cortex). However, that the C1 and P1 dipoles are spatially separate only shows that the C1 and P1 voltage topographies are not identical, and does not rule out that the C1 is influenced by the P1 to some degree. As the ipsilateral P1 sink is generated by a deeper source (further from the scalp), it would be expected to produce a more diffuse topography than the ipsilateral C1, as proposed. However, the separate C1 and P1 dipole voltage topographies were not shown, which would be necessary to assess the degree to which the ipsilateral C1 and ipsilateral P1 sink overlap. Mohr and Kelly argued that the ipsilateral C1 and P1 sink were separable because Baumgartner et al. found significant P1 effects and non-significant C1 effects. Although this shows the ipsilateral C1 and P1 sink are not identical, it is still possible that the ipsilateral C1 reflects the P1 sink to some degree. Mohr and Kelly also argued that the ipsilateral C1 did not reflect the early contralateral P1 sink because the ipsilateral C1 was diminished when the early contralateral P1 peaked in magnitude (citing Di Russo, Martínez, Sereno, Pitzalis, & Hillyard, 2002; anddi Russo et al., 2005). This is a correct description (see Figures 5 and 6 in Di Russo et al., 2002, andfigure3 in, 2005). However, when the early contralateral P1 peaks in these studies, there is still a robust ipsilateral negativity that appears to reflect the corresponding P1 voltage sink and largely overlaps with the ipsilateral C1. The contralateral P1 onset can be as early as 72 milliseconds after stimulus onset and both Kelly et al. and Baumgartner et al. measured C1 attention effects from milliseconds after stimulus onset, thus it seems possible that the C1 attention effects in these studies may have reflected the early contralateral P1 sink to some degree. In a final argument, Mohr and Kelly highlighted that the C1 attention effects in their study began at 57 milliseconds after stimulus onset, well before the onset of the early contralateral P1. This argument is compelling, as there is no way for a P1 sink to influence the ipsilateral C1 before P1 onset. Therefore, based on this argument, Kelly et al. s rapid C1 attention effects (before 72 milliseconds after stimulus onset) reflect evidence of significant V1 attentional modulationandareconsideredinthereanalysisbelow. Pitts and Hillyard stated that it would be preferable to measure C1 where its amplitude is maximal on an individual participant basis. Although this would seem to be ideal, requiring individual participant mapping would reduce the possibility of replicating the previous studies that observed significant C1 attention effects at or just lateral to electrodes POz, Pz, and CPz (see Table 1 in Slotnick, this issue). Moreover, based on the findings of Kelly et al. and Baumgartner et al., individual participant C1 mapping yields selection of lateral occipital electrodes that may overlap with the early contralateral P1 source or ipsilateral P1 sink to some degree. One solution might be to conduct individual participant C1 mapping and analyze C1 attention effects from milliseconds after stimulus onset (i.e., before P1 onset). However, given that this early time range is before the C1 typically peaks, the combination of individual participant C1 mapping and assessment of C1 attention effects in an early time window (before P1 onset) would be expected to have limited sensitivity. Therefore, measuring C1 attention effects at or just lateral to POz, Pz, and CPz (approximately when the C1 peaks) is still recommended, as this has been shown to be sensitive to C1 attention effects and avoids contamination by P1 attention effects. Optimal stimulus locations to produce C1 attention effects Based on the studies in the primary analysis of Slotnick (this issue), it was concluded that to maximize C1 attention effects stimuli should be in the upper visual field. This conclusion was qualified in Slotnick (this issue) by acknowledging that many studies that observed significant C1 attention effects tested only upper visual field locations and there was some evidence for marginally significant C1 attention effects for lower visual field locations. Of particular relevance, Rauss, Pourtois, Vuilleumier, and Schwartz (2009) employed stimuli in the upper visual field or lower visual field and observed significant C1 attention effects for upper visual field stimuli (F(1,13) = 6.17, p =.027) and null C1 attention effects for lower visual field stimuli (F < 1). Although these are the results of only one study, they provide some evidence that C1 attention effects may be stronger in the upper visual field than the lower visual field. Related to the previous section, another reason to employ upper visual field stimuli is that the C1 topography will have negative polarity that is maximum over the ipsilateral

4 78 S. D. SLOTNICK hemisphere, making it more spatially separable from the early P1 topography that has positive polarity over the contralateral hemisphere (see Figure 1 in Slotnick, this issue). By contrast, the lower visual field C1 topography will have the same polarity and overlap with the early contralateral P1 topography; thus, it is more difficult to disentangle lower visual field contralateral C1 attention effects from early contralateral P1 attention effects. Mohr and Kelly pointed out that C1 topography and polarity vary as a function of stimulus position and questioned the assumption that upper visual field stimuli produce a C1 that is negative in magnitude and lower visual field stimuli produce a C1 that is positive in magnitude. Figure 1 illustrates the group average variation in C1 topography and polarity (adapted from Figure 6 in Clark et al., 1994). Although there can be individual participant variability in C1 topography and polarity (see Figure 7 in Clark et al., 1994), the group average visual field polarity map illustrates the most commonly observed pattern of activity. Mohr and Kelly are correct that it is theoretically possible for the upper visual field-negative polarity assumption to be Figure 1. Voltage topography as a function of stimulus position from milliseconds after stimulus onset. C1 and P1 are labeled in the right upper visual field and the left upper visual field. Stimuli were presented 8 in visual angle from fixation. Voltage topographies as a function of polar angle from the horizontal meridian are labeled in the right visual field and the analogous voltage topographies are shown in the left visual field (voltage magnitude key at the bottom left). Reprinted from Human Brain Mapping, Volume 2, Vincent P. Clark, Silu Fan, and Steven A. Hillyard, Identification of early visual evoked potential generators by retinotopic and topographic analysis, Pages , Copyright (1994), with permission from John Wiley and Sons.

5 COGNITIVE NEUROSCIENCE 79 violated near the upper vertical meridian. However, this assumption was not violated in the studies included in the primary analysis of Slotnick (this issue) as all of them employed stimuli close to 45 of polar angle from the horizontal meridian. That is, based on the general pattern of activity across participants (Figure 1), the studies included in the primary analysis of Slotnick (this issue) employed stimulus locations that were well within the safe zone, far from where polarity reversals occur. In support of this point, if the topography and polarity reversals were not consistent across participants in the studies included in the primary analysis of Slotnick (this issue), significant C1 attention effects would not have been observed (since several studies reported significant C1 attention effects, this is not of concern). It is notable that the stimuli employed by Rauss et al. (2009) and Rauss, Pourtois, Vuilleumier, and Schwartz (2012) extended from near the horizontal meridian to near the vertical meridian within both upper visual field quadrants. Based on the visual field C1 polarity map (Figure 1), this means that the stimuli in those studies would primarily produce a C1 of negative polarity but a portion of the stimuli near the upper vertical meridian would produce a C1 with positive polarity, thus muting the upper visual field C1 response to some degree. Of importance, in accordance with the upper visual field-negative polarity assumption, a robust upper visual field C1 response with negative polarity was observed and significant C1 attention effects were reported in Rauss et al. (2009, 2012), so this is not of concern. Pitts and Hillyard took issue with the claim in Slotnick (this issue) that the horizontal meridian stimuli in Martínez et al. (1999, 2001) would produce a muted C1 response that would limit sensitivity to detect attention effects. They correctly point out that the C1 reverses in polarity approximately 20 below the horizontal meridian. Note that when the term muted was used in Slotnick (this issue) to describe the C1 response in studies that presented stimuli at the horizontal meridian, this meant largely diminished (like a trombone mute rather than a TV mute). Figure 1 shows that stimuli presented at the horizontal meridian are largely diminished when compared to stimuli presented at or just below 45 of polar angle from the horizontal meridian. As such, studies that employed stimuli along the horizontal meridian were not considered in the primary analysis of Slotnick (this issue), as they would be expected to produce a largely diminished C1 response such that it would be difficult to observe significant C1 attention effects. Regarding optimal stimulus location, based on the above considerations and the visual field C1 polarity map (Figure 1), it is suggested that stimuli in the upper visual field should be presented at or just below the line 45 of polar angle from the horizontal meridian, stimuli in the lower visual field should be presented just below the line 45 of polar angle from the horizontal meridian, and stimuli should not be so large that they extend to visual field regions where the C1 is diminished (i.e., near the upper vertical meridian or near/just below the horizontal meridian). If resources were unlimited, it would be ideal, as suggested by Mohr and Kelly along with Pitts and Hillyard, to map out optimal stimulus locations for each participant, but this does not seem necessary as presenting stimuli at approximately 45 from the horizontal meridian produces a robust C1 response that is sensitive to attention effects. Be wary of null V1 cortical source localization attention effects Qu and Ding questioned whether the C1 attention effects of Rauss et al. (2009) and Rauss et al. (2012) reflected modulation of V1 given that their cortical sources were primarily localized to the prefrontal and frontal/temporal/parietal regions, respectively. To weight these null V1 source localization attention effects, cortical source localization methods must be considered in some detail. Cortical source localization involves converting a scalp voltage topography into a limited number of cortical sources. For example, within the millisecond time window, the voltage magnitudes at 128 electrodes might be converted into one V1 dipole current source that is fixed in location and orientation with a magnitude that varies over time. Source localization typically requires modeling the head (e.g., the brain, skull, cerebrospinal fluid, and scalp; such head models are universally poor), modeling cortical activity, and adjusting the cortical dipole(s) using an iterative model fitting procedure to minimize the difference between the observed scalp voltage topography and the

6 80 S. D. SLOTNICK model scalp voltage topography (for a review, see Slotnick, 2004). Of particular relevance, cortical source localization is a massive data reduction, where activity from a large number of electrodes is typically converted into one or a few cortical sources. Said another way, each cortical source represents the activity from a large number of electrodes. Given that significant C1 attentional modulation typically occurs at one or a few electrodes (see Table 1 in Slotnick, this issue) and cortical source localization depends on activity from a large number of electrodes, cortical source localization techniques would be expected to produce null attention effects in V1. In short, a significant attention effect at one or a few electrodes will be washed out by the non-significant attention effects at the numerous other electrodes distilled into the cortical source (unless the attention effect is very robust). As such, the null C1/V1 source localization attention effects in Rauss et al. (2009) are not informative, while the marginally significant C1/V1 source localization attention effects in Rauss et al. (2012; see their Figure 4B, region 4) do providesomeevidenceforattentionalmodulation within this region. That is, there appears to be some degree of V1 attentional modulation in Rauss et al. (2012) despite the largely null electophyisiological activity included in the source localization analysis. The expectation of null V1 source localization attention effects has potential implications for Hillyard and colleagues findings that dipole source localization C1/V1 attention effects are initially non-significant (< 90 milliseconds after stimulus onset) and then become significant later in time, presumably due to feedback from extrastriate cortex (Di Russo et al., 2003; Martínez et al., 1999, 2001; Noesselt et al., 2002). The initial null V1 attention effects may be due to limited sensitivity of cortical source localization coupled with the small magnitude of C1 attentional modulation. A magnetoencephalography (MEG) spatial attention study by Poghosyan and Ioannides (2008) provided evidence that the dipole source localization techniques employed by Hillyard and colleagues had limited sensitivity to detect attentional modulation in the initial wave of V1 activity. Before employing the dipole modeling procedure of Hillyard and colleagues, Poghosyan and Ioannides observed significant V1 spatial attention effects beginning 55 to 60 milliseconds after stimulus onset and peaking at approximately 70 milliseconds after stimulus onset. After employing thesamedipolemodelingproceduresashillyard and colleages, Poghosyan and Ioannides observed null V1 attention effects 55 to 70 milliseconds after stimulus onset and attentional enhancement later in time. One of my spatial attention ERP source localization studies (Slotnick, Hopfinger, Klein, & Sutter, 2002) provided evidence that the initial V1 attention effects are much smaller in magnitude than subsequent attention effects in this region. This study employed a multi-stimulus array where each of 60 checkerboards were independently modulated (i.e., reversed in luminance) tens of thousands of times according to a binary m-sequence, which produced a very high signal-to-noise ratio. A previous study using the same stimulus and source localization procedures showed that the dipole locations varied systematically as stimulus position moved from the top to the bottom of each hemifield (Slotnick, Klein, Carney, Sutter, & Dastmalchi, 1999). This is diagnostic of a V1 cortical source, where the cortical surface is continuous from the upper vertical meridian to the lower vertical meridian of each hemifield (smoothly traversing the horizontal meridian representation near the base of the calcarine sulcus), and inconsistent with an extrastriate (e.g., V2, V3) cortical source, where the cortical surface is disjoint near the horizontal meridian (see Figure 2 in Slotnick, Thompson, & Kosslyn, 2005, which illustrates detailed functional magnetic resonance imaging, fmri, retinotopic maps of early visual regions). Moreover, another study that used the same stimulus and source localization procedures as Slotnick et al. (2002) showed that the cortical magnification factor computed using dipole locations were similar to V1 cortical magnification estimates from psychophysics, cortical stimulation, and fmri (Slotnick, Klein, Carney, & Sutter, 2001). Slotnick et al. (2002) observed significant attentional modulation of the V1 dipole magnitude corresponding to the stimulus at the attended location that did not change over time from the early epoch (50 80 milliseconds after stimulus onset) to the late epoch ( milliseconds after stimulus onset). By contrast, for V1 dipoles corresponding to stimuli in a broader region

7 COGNITIVE NEUROSCIENCE 81 extending from the attended location to fixation, there was no significant attentional modulation in the early epoch but significant attentional modulation in the late epoch, which is reminiscent of the findings of Hillyard and colleagues. The overall magnitude of the attention effects within the broader facilitatory region were over five times larger in magnitude than at the attended location. This larger magnitude of V1 attentional modulation surrounding the attended location can be attributed to feedback from extrastriate cortex, given that extrastriate regions have larger receptive fields than V1 and are active within the late epoch. These results suggest that attention produces a small change in V1 dipole magnitudes during the initial V1 response, which will typically be dominated by noise and produce null findings. C1/V1 dipole source localization attention effects are usually highly influenced by noise and therefore difficult to trust. It is important to identify the optimal conditions to maximize C1 attention effects, such that experiments (and dipole modeling) can have a higher signal-to-noise ratio and produce more convincing results. cuing paradigms, a significant C1 attention effect can be considered reliable (i.e., not due to differential cue effects) if the baseline level of activity is close to zero and there is no differential activity in valid versus invalid trials before C1 onset. Figure 2 shows that in both Fu et al. (2009) and Dassanayake, Michie, and Fulham (2016), the baseline magnitudes of activity were close to zero and there was no differential valid versus invalid activity before C1 onset. Thus, the C1 attention effects in these two studies appear to be real as there was no evidence of cue related Discounting significant C1 attention effects Qu and Ding along with Pitts and Hillyard highlighted that in attention studies with exogenous cuing, valid trials are preceded by a cue in the target location while invalid trials are not preceded by a cue in the target location. They correctly pointed out that C1 attention effects, corresponding to a higher magnitude of activity for valid than invalid trials, might reflect an increase in activity due to the preceding cue for valid trials. With regard to the studies considered in the primary analysis of Slotnick (this issue), Pitts and Hillyard noted that in Fu, Fedota, Greenwood, and Parasuraman (2010), valid trials had a higher magnitude of activity than invalid trials that was developed approximately 30 to 40 milliseconds after stimulus onset, which is before C1 onset. Such an early onset does suggest a cue-related effect occurred in this study that may have artifactually produced a significant C1 effect. As such, the study by Fu et al. (2010) did not present convincing evidence of a significant C1 attention effect and is not considered in the reanalysis below. Following the same logic as Pitts and Hillyard, for exogenous Figure 2. Significant C1 attention effects in two exogenous cuing studies. (a) Valid and invalid activation timecourses at electrode Pz are shown in red and green, respectively (key to the right), corresponding to stimulation in the upper visual field. C1, P1m, and N1 components are labeled. Each tick on the horizontal axis represents 50 milliseconds and each tick on the vertical axis represents 1 microvolt (negative is up and the axes cross at 0). Reprinted from NeuroImage, 48/1, Shimin Fu, Yuxia Huang, Yuejia Luo, Yan Wang, John Fedota, Pamela M. Greenwood, and Raja Parasuraman, Perceptual load interacts with involuntary attention at early processing stages: Eventrelated potential studies, , 2009, with permission from Elsevier. (b) Valid and invalid activation timecourses at electrode POz are shown as solid lines and dotted lines, respectively (key to the right). C1 and P1m components are labeled. Each tick on the horizontal axis represents 100 milliseconds and each tick on the vertical axis represents 1 microvolt (negative is up and the axes cross at 0). Activation timecourses corresponding to stimulation in the upper visual field (thicker lines) produced C1 components with negative polarity and were used to test attention effects. Activation timecourses corresponding to stimulation in the lower visual field (thinner lines) produced C1 components with positive polarity and were shown to illustrate polarity reversal. Reprinted from International Journal of Psychophysiology, 105, Tharaka L. Dassanayake, Patricia T. Michie, and Ross Fulham, Effect of temporal predictability on exogenous attentional modulation of feedforward processing in the striate cortex, 9 16, 2016, with permission from Elsevier.

8 82 S. D. SLOTNICK activity before C1 onset. Although a theoretical argument can be made that all exogenous cueing studies should be discounted because the effects might be due to differential valid versus invalid cueing, baseline/pre-c1 activity can simply be evaluated to show that this is not the case. Related to this, Qu and Ding mentioned that Rauss et al. (2012) had a baseline shift; however, this is not of concern as the magnitude of pre-c1 activity was less than one-fourth of the magnitude of the C1 attention effect. Pitts and Hillyard also pointed to Dassanayake et al. s interpretation that their C1 and P1 modulation could be described as a third negative component that spanned C1 and P1. Figure 2(b) provides some evidence against this interpretation as the magnitude of the P1 attention effect is four times larger than the magnitude of the C1 attention effect, while a protracted negativity that spanned both components would be expected to produce effects of approximately the same magnitude. Still, it is possible that a third negative component produced differential C1 and P1 effects, which should be investigated further. Along the same lines, Pitts and Hillyard argued that attended and unattended C1 effects should have the same voltage topography/ source, with attention effects corresponding to amplification of this source. I agree it would be ideal to show the voltage topography in the attended and unattended conditions to illustrate a common neural generator that is amplified by attention. However, this is often not done because it is reasonable to assume that an ERP component at a particular electrode where the attended activity and the unattended activity begin, increase, peak, and decrease with nearly identical timing can be attributed to the same cortical source. In addition, excluding studies that did not provide voltage topographies would remove many studies that have reported both significant C1 attention effects and null C1 attention effects. It is notable that Hopfinger and West (2006) did not show voltage topographies for attended and unattended stimuli, but Pitts and Hillyard considered their null findings as evidence against the hypothesis that exogenous cueing modulates C1. Mohr and Kelly aimed to discount the studies that manipulated attentional load (Ding, Martínez, Qu, & Hillyard, 2014; Rauss et al., 2009, 2012) by claiming that they manipulated load rather than spatial attention. This is not a valid argument as these studies manipulated attentional load at central fixation which, as noted in Slotnick (this issue), changes the allocation of spatial attention at other visual field locations. Thus, although spatial attention was not manipulated directly, it was manipulated indirectly, as indicated by the significant C1 spatial attention effects to peripheral distractors in Rauss et al. (2009, 2012). Null C1 attention effects versus significant C1 attention effects Slotnick (this issue) only considered papers in the primary analysis that employed parameters that were sensitive to C1 attention effects, including measuring these effects at or near midline parietal-occipital electrodes (to avoid P1 contamination) and stimulating within each quadrant (to avoid a muted C1 response). Mohr and Kelly claimed that Table 1 in Slotnick (this issue) left out a number of studies that have shown null C1 attention effects (citing ten studies). In fact, four of these studies (Fu, Greenwood, & Parasuraman, 2005; Fu et al., 2008; Martínez et al., 1999, 2001) were intentionally excluded in the Other spatial attention studies section of Slotnick (this issue) because they employed non-optimal methods. Of the remaining six studies referenced by Mohr and Kelly, four studies fit the exclusion criteria of Slotnick (this issue): one study because they presented stimuli along the horizontal meridian and did not measure C1 attention effects (Curran, Hills, Patterson, & Strauss, 2001), one study because all statistical tests of C1 attention effects included occipital electrodes (Fu, Fan, Chen, & Zhuo, 2001; the C1ʹ component also peaked after 100 milliseconds), one study because they only measured C1ʹ attention effects at lateral electodes (Johannes, Münte, Heinze, & Mangun, 1995), and one study because it was a review paper (Hillyard, Teder-Sälejärvi, & Münte, 1998; the only potentially new empirical paper that was referenced in this review, Gomez Gonzales, Clark, Fan, Luck, & Hillyard, 1994, presented stimuli along the horizontal meridian). The remaining two studies (Clark & Hillyard, 1996; Wijers, Lange, Mulder, & Mulder, 1997), which both employed endogenous cuing with no distractors, did report null C1 attention effects and are considered in the reanalysis below.

9 COGNITIVE NEUROSCIENCE 83 Di Russo took issue with my characterizing C1 attention effects with p values greater than 0.05 as marginally significant. In Slotnick (this issue), it was stated that the attention effect in the right lower visual field of Di Russo et al. (2003) would likely be significant if tested alone. Di Russo kindly conducted this statistical test in his commentary, which was a non-significant F-test (F(1,29) = 4.06, p =.0563). It is notable that this is equivalent to a significant onetailed t-test (t(29) = 2.015, p =.027); a one-tailed test is justified as it is known that spatial attention produces a greater magnitude of activity in V1 for attended than unattended stimuli (e.g., Slotnick, Schwarzbach, & Yantis, 2003; Tootell et al., 1998). Still, it is reasonable to take a more conservative statistical approach and characterize marginally significant findings as non-significant, as will be done in the reanalysis below. The key point is that such marginally significant results should not be taken to support the null hypothesis. As stated by Wagenmakers, Fisherian p values are not designed to quantify support in favor of the null hypothesis. A p value indicates the evidence against the null hypothesis. It is not possible to observe the data and corroborate the null hypothesis; one can only fail to reject it (Wagenmakers, 2007, p. 795). Regarding C1 attention effects, null findings can be observed for numerous reasons such as an insufficient number of participants, participant variability, non-optimal electrode (or reference) selection, nonoptimal stimulus location, other non-optimal experimental parameters (e.g., low load conditions or no distractors), or insensitive analysis procedures. Unless optimal experimental conditions and analysis techniques are employed, null findings provide little informational value and should never be taken to support the null hypothesis. If a scientist claims their results provide some support for null C1 attention effects, the burden of proof is on them to show that optimal experimental conditions and analysis techniques were employed. Based on the commentaries and my responses to them, Table 1 illustrates an update of the spatial attention studies that employed methods that were sensitive to C1 attention effects and survived critical analysis. The original conclusions of Slotnick (this issue) were that to maximize sensitivity to C1 attention effects, stimuli should be in the upper visual field, there should be distractors, there should be high perceptual or attentional load, there should be exogenous cuing, and effects should be measured at midline parietal-occipital electrodes POz, Pz, and CPz. Table 1 continues to support all of these suggestions. It is notable that three of the nine studies that employed endogenous cuing observed significant C1 attention effects (and only one of these studies directly manipulated spatial attention), while two of the three studies that employed exogenous curing observed significant C1 attention effects. Five of the eleven studies in Table 1 observed significant C1 attention effects. As eloquently stated by William James, Ifyouwishtoupsetthelaw that all crows are black, you mustn t seek to show that no crows are; it is enough if you prove one single crow to be white (James, 1986, p. 121). Here, we have a murder of eleven crows, and five of them are white. One can pretend that the white crows don t existbuttheydo,onecantrytokill the white crows but they survive, and one can add more black crows but there will always be white crows. It often seems as if we are looking through Table 1. Spatial attention study parameters and C1 significance. Study VF Distractors Load Cue Electrode(s) C1 significance Clark & Hillyard (1996) LUVF/RUVF No Low per/att End IPz Not sig Wijers et al. (1997) LUVF/RUVF No Low per/att End POz Not sig Di Russo et al. (2003) LUVF/RUVF/LLVF/RLVF No Low per/att End POz Not sig Hopfinger and West (2006) LUVF/RUVF No Low per/att End/Exo Pz/POz Not sig Kelly et al. (2008) LUVF/RUVF/LLVF/RLVF No Low per/att End Lateral occ Sig starting at 57 ms Fu et al. (2009) LUVF/RUVF Yes Low/Med/High per Exo POz/Pz/CPz High load sig Pz Rauss et al. (2009) UVF/LVF Yes Low/High att End CP1/CPz/CP2/P1/Pz/P2 UVF low vs. high load sig Di Russo et al. (2012) LUVF/RUVF/LLVF/RLVF No Low per/att End PO1/PO2/P1/P2 Not sig Rauss et al. (2012) UVF Yes Low/High att End CP1/CPz/CP2/P1/Pz/P2 High vs. low load sig Ding et al. (2014) UVF/LVF Yes Low/High att End POz Not sig Dassanayake et al. (2016) LUVF/RLVF Yes High per Exo Oz/POz/Pz UVF sig at POz L = left, R = right, UVF = upper visual field, LVF = lower visual field, Per = perceptual, Att = attentional, End = endogenous, Exo = exogenous, Sig = significant. Significant and non-significant effects are shown in gray and white, respectively. Only low/med/high load and low/high load studies manipulated load. Low perceptual/attentional load studies were classified as such because there were no distractors and the tasks were simple target detection.

10 84 S. D. SLOTNICK thick brush and there are disagreements about whether there is a white crow (C1) or a white pigeon (P1), which highlights the importance of replicating studies that have reported significant C1 attention effects. However, there are now several convincing studies that have shown spatial attention can modulate the C1 component. The C1 spatial attention effect is small in magnitude and requires optimal experimental parameters to uncover, but it exists. Future directions The selective review of Slotnick (this issue) and a slight revision of that review in the present paper based on the commentaries and my responses to them (Table 1) aimed to identify the experimental parameters that affect attentional modulation of the ERP C1 component. It must be underscored that the conclusions drawn were based on the limited number of studies that employed techniques that were sensitive to C1 attentional modulation. I agree with Mohr and Kelly that which parameters affect the C1 attention effect and why those parameters affect the C1 attention effect are largely open questions. The present conclusions should serve as a set of hypotheses that should be investigated in future experiments. Where do we go from here? As noted by Pitts and Hillyard, there have been almost no replications of previous significant C1 attention findings. Specifically, Ding et al. (2014) failed to replicate Rauss et al. (2009) and Baumgartner et al. (this issue) failed to replicate Kelly et al. (2008). Although the experimental parameters in Ding et al. and Baumgartner et al. were not identical to the original studies, they were very similar, which suggests C1 attention effects can be easily perturbed. Although Dassanayake et al. (2016) did replicate Fu et al. (2009), more replications are needed particularly those that employ the identical experimental parameters to further illustrate the conditions under which attention can modulation the C1 component. It is critically important to reproduce findings across laboratories, which is one line of future research. As noted by Mohr and Kelly, many of the studies that observed significant C1 attentional modulation employed the same experimental parameters, such as exogenous cuing with distractors. As suggested by Slotnick (this issue), another line of future research will be to systematically manipulate experimental parameters to determine the impact of each parameter on C1 spatial attention effects. Disclosure statement No potential conflict of interest was reported by the author. References Clark, V. P., Fan, S., & Hillyard, S. A. (1994). Identification of early visually evoked potential generators by retinotopic and topographic analysis. Human Brain Mapping, 2, Clark, V. P., & Hillyard, S. A. (1996). Spatial selective attention affects early extrastriate but not striate components of the visual evoked potential. Journal of Cognitive Neuroscience, 8, Curran, T., Hills, A., Patterson, M. B., & Strauss, M. E. (2001). Effects of aging on visuospatial attention: An ERP study. Neuropsychologia, 39, Dassanayake, T. L., Michie, P. T., & Fulham, R. (2016). Effect of temporal predictability on exogenous attentional modulation of feedforward processing in the striate cortex. International Journal of Psychophysiology, 105, Di Russo, F., Martínez, A., & Hillyard, S. A. (2003). Source analysis of event-related cortical activity during visuo-spatial attention. Cerebral Cortex, 13, Di Russo, F., Martínez, A., Sereno, M. I., Pitzalis, S., & Hillyard, S. A. (2002). Cortical sources of the early components of the visual evoked potential. Human Brain Mapping, 15, Di Russo, F., Pitzalis, S., Spitoni, G., Aprile, T., Patria, F., Spinelli, D., & Hillyard, S. A. (2005). Identification of the neural sources of the pattern-reversal VEP. NeuroImage, 24, Di Russo, F., Stella, A., Spitoni, G., Strappini, F., Sdoia, S., Galati, G.,... Pitzalis, S. (2012). Spatiotemporal brain mapping of spatial attention effects on pattern-reversal ERPs. Human Brain Mapping, 33, Ding, Y., Martínez, A., Qu, Z., & Hillyard, S. A. (2014). Earliest stages of visual cortical processing are not modified by attentional load. Human Brain Mapping, 35, Fu, S., Fan, S., Chen, L., & Zhuo, Y. (2001). The attentional effects of peripheral cueing as revealed by two event-related potential studies. Clinical Neurophysiology, 112, Fu, S., Fedota, J., Greenwood, P. M., & Parasuraman, R. (2010). Early interaction between perceptual load and involuntary attention: An event-related potential study. Neuroscience Letters, 468, Fu, S., Greenwood, P. M., & Parasuraman, R. (2005). Brain mechanisms of involuntary visuospatial attention: An event-related potential study. Human Brain Mapping, 25, Fu, S., Huang, Y., Luo, Y., Wang, Y., Fedota, J., Greenwood, P. M., & Parasuraman, R. (2009). Perceptual load interacts with

11 COGNITIVE NEUROSCIENCE 85 involuntary attention at early processing stages: Eventrelated potential studies. NeuroImage, 48, Fu, S., Zinni, M., Squire, P. N., Kumar, R., Caggiano, D. M., & Parasuraman, R. (2008). When and where perceptual load interacts with voluntary visuospatial attention: An eventrelated potential and dipole modeling study. NeuroImage, 39, Gomez Gonzales, C. M., Clark, V. P., Fan, S., Luck, S. J., & Hillyard, S. A. (1994). Sources of attention-sensitive visual event-related potentials. Brain Topography, 7, Hopfinger, J. B., & West, V. M. (2006). Interactions between endogenous and exogenous attention on cortical visual processing. NeuroImage, 31, Hillyard, S. A., Teder-Sälejärvi, W. A., & Münte, T. F. (1998). Temporal dynamics of early perceptual processing. Current Opinion in Neurobiology, 8, James, W. (1986). Essays in psychical research (the works of William James). Cambridge, MA: Harvard University Press. Johannes, S., Münte, T. F., Heinze, H. J., & Mangun, G. R. (1995). Luminance and spatial attention effects on early visual processing. Cognitive Brain Research, 2, Kelly, S. P., Gomez-Ramirez, M., & Foxe, J. J. (2008). Spatial attention modulates initial afferent activity in human primary visual cortex. Cerebral Cortex, 18, Khoe, W., Mitchell, J. F., Reynolds, J. H., & Hillyard, S. A. (2005). Exogenous attentional selection of transparent superimposed surfaces modulates early event-related potentials. Vision Research, 45, Martínez, A., Anllo-Vento, L., Sereno, M. I., Frank, L. R., Buxton, R. B., Dubowitz, D. J.,... Hillyard, S. A. (1999). Involvement of striate and extrastriate visual cortical areas in spatial attention. Nature Neuroscience, 2, Martínez,A.,DiRusso,F.,Anllo-Vento,L.,Sereno,M.I.,Buxton,R.B., & Hillyard, S. A.(2001). Putting spatial attention on the map: Timing and localization of stimulus selection processes in striate and extrastriate visual areas. Vision Research, 41, Noesselt, T., Hillyard, S. A., Woldorff, M. G., Schoenfeld, A., Hagner, T., Jäncke, L.,... Heinze, H. J. (2002). Delayed striate cortical activation during spatial attention. Neuron, 35, Poghosyan, V., & Ioannides, A. A. (2008). Attention modulates earliest responses in the primary auditory and visual cortices. Neuron, 58, Rauss, K. S., Pourtois, G., Vuilleumier, P., & Schwartz, S. (2009). Attentional load modifies early activity in human primary visual cortex. Human Brain Mapping, 30, Rauss, K., Pourtois, G., Vuilleumier, P., & Schwartz, S. (2012). Effects of attentional load on early visual processing depend on stimulus timing. Human Brain Mapping, 33, Slotnick, S. D. (2004). Source localization of ERP generators. In T. C. Handy (Ed.), Event-related potentials: A methods handbook (pp ). Cambridge, MA: The MIT Press. Slotnick, S. D., Hopfinger, J. B., Klein, S. A., & Sutter, E. E. (2002). Darkness beyond the light: Attentional inhibition surrounding the classic spotlight. NeuroReport, 13, Slotnick, S. D., Klein, S. A., Carney, T., & Sutter, E. E. (2001). Electrophysiological estimate of human cortical magnification. Clinical Neurophysiology, 112, Slotnick, S. D., Klein, S. A., Carney, T., Sutter, E., & Dastmalchi, S. (1999). Using multi-stimulus VEP source localization to obtain a retinotopic map of human primary visual cortex. Clinical Neurophysiology, 110, Slotnick, S. D., Schwarzbach, J., & Yantis, S. (2003). Attentional inhibition of visual processing in human striate and extrastriate cortex. NeuroImage, 4, Slotnick, S. D., Thompson, W. L., & Kosslyn, S. M. (2005). Visual mental imagery induces retinotopically organized activation of early visual areas. Cerebral Cortex, 15, Tootell, R. B., Hadjikhani, N., Hall, E. K., Marrett, S., Vanduffel, W., Vaughan, J. T., & Dale, A. M. (1998). The retinotopy of visual spatial attention. Neuron, 21, Wagenmakers, E. J. (2007). A practical solution to the pervasive problems of p values. Psychonomic Bulletin & Review, 14, Wijers, A. A., Lange, J. J., Mulder, G., & Mulder, L. J. (1997). An ERP study of visual spatial attention and letter target detection for isoluminant and nonisoluminant stimuli. Psychophysiology, 34,