- Email: [email protected]
Two distinct neural mechanisms in early visual cortex determine subsequent visual processing
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Available online at www.sciencedirect.com
ScienceDirect Journal homepage: www.elsevier.com/locate/cortex
Research report
Two distinct neural mechanisms in early visual cortex determine subsequent visual processing Christianne Jacobs a,b,c,*,1, Tom A. de Graaf b,c,1 and Alexander T. Sack b,c a
Department of Psychology, FST, University of Westminster, London, UK Department of Cognitive Neuroscience, FPN, Maastricht University, Maastricht, The Netherlands c Maastricht Brain Imaging Center (M-BIC), Maastricht, The Netherlands b
article info
abstract
Article history:
Neuroscience research has conventionally focused on how the brain processes sensory
Received 2 October 2013
information, after the information has been received. Recently, increased interest focuses
Reviewed 18 December 2013
on how the state of the brain upon receiving inputs determines and biases their subse-
Revised 26 February 2014
quent processing and interpretation. Here, we investigated such 'pre-stimulus' brain
Accepted 19 June 2014
mechanisms and their relevance for objective and subjective visual processing. Using non-
Action editor Jason Barton
invasive focal brain stimulation [transcranial magnetic stimulation (TMS)] we disrupted
Published online 16 July 2014
spontaneous brain state activity within early visual cortex (EVC) before onset of visual stimulation, at two different pre-stimulus-onset-asynchronies (pSOAs). We found that
Keywords:
TMS pulses applied to EVC at either 20 msec or 50 msec before onset of a simple orientation
Visual perception
stimulus both prevented this stimulus from reaching visual awareness. Interestingly, only
Early visual cortex
the TMS-induced visual suppression following TMS at a pSOA of 20 msec was reti-
Transcranial magnetic stimulation
notopically specific, while TMS at a pSOA of 50 msec was not. In a second experiment, we
Suppression
used more complex symbolic arrow stimuli, and found TMS-induced suppression only
State-dependence
when disrupting EVC at a pSOA of ~ 60 msec, which, in line with Experiment 1, was not retinotopically specific. Despite this topographic unspecificity of the 50 msec effect, the additional control measurements as well as tracking and removal of eye blinks, suggested that also this effect was not the result of an unspecific artifact, and thus neural in origin. We therefore obtained evidence of two distinct neural mechanisms taking place in EVC, both determining whether or not subsequent visual inputs are successfully processed by the human visual system. © 2014 Elsevier Ltd. All rights reserved.
1.
Introduction
The occipital lobe of the human brain is dedicated to the processing of visual inputs. Early cortical stages of visual
information processing occur approximately 60e100 msec after stimulus presentation in visual areas V1, V2 and V3 of the occipital brain, together commonly referred to as early visual cortex (EVC). Much research has focused on the specific
* Corresponding author. Department of Psychology, Faculty of Science and Technology, University of Westminster, 309 Regent Street, W1B 2HW London, United Kingdom. E-mail address: [email protected] (C. Jacobs). 1 Equal contribution. http://dx.doi.org/10.1016/j.cortex.2014.06.017 0010-9452/© 2014 Elsevier Ltd. All rights reserved.
2
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
visual properties these early visual areas process, their neuronal tuning to these properties, and the information flow within and/or between these regions of EVC (e.g., Cardin, Friston, & Zeki, 2011; Downing, Chan, Peelen, Dodds, & Kanwisher, 2006; Nandy, Sharpee, Reynolds, & Mitchell, 2013). These are examples of how research conventionally focuses on the brain's response to inputs from the environment. Yet, how inputs are processed may depend not only on nature of the information, but also on the prior state of the brain (Arieli, Sterkin, Grinvald, & Aertsen, 1996; Busch, Dubois, & VanRullen, 2009; Hesselmann, Kell, Eger, & Kleinschmidt, 2008; Mathewson, Gratton, Fabiani, Beck, & Ro, 2009). Recent studies have addressed the potential modulatory role of neural state prior to visual input. Electrophysiological studies have shown, for example, that under conditions of near-threshold stimulus visibility, the variability in stimulus perception is indeed reflected in pre-stimulus brain activity (Busch, et al., 2009; van Dijk, Schoffelen, Oostenveld, & Jensen, 2008; Dugue, Marque, & VanRullen, 2011; Hanslmayr et al., 2007; Mathewson, et al., 2009; Romei et al., 2008; Romei, Gross, & Thut, 2010; Toscani, Marzi, Righi, Viggiano, & Baldassi, 2010). So far, the power (van Dijk, et al., 2008; Romei, et al., 2008; Romei, et al., 2010; Toscani, et al., 2010) and phase (Busch, et al., 2009; Dugue, et al., 2011; Mathewson, et al., 2009) of pre-stimulus parieto-occipital oscillations in the alpha frequency band (i.e., 8e12 Hz), and phase-locking in the beta (16e30 Hz) and gamma (>30 Hz) frequency bands (Hanslmayr, et al., 2007) have been related to this perceptual modulation. Next, researchers reversed the line of reasoning and showed that they could affect stimulus visibility by externally manipulating parieto-occipital alpha oscillations (e.g., de Graaf et al., 2013; Mathewson et al., 2012; Romei, et al., 2010), establishing a causal relation between stimulus visibility and pre-stimulus neural state. Together, these studies demonstrate that neural processing that occurs prior to sensory input can play a functional role in perception. By means of Transcranial Magnetic Stimulation (TMS), a brain interference tool that allows temporal disruption of neuronal activity, the contribution of pre-stimulus EVC state to visual perception can be investigated. If applied over occipital cortex, TMS can lead to a complete abolishment of conscious perception of suprathreshold visual stimuli. This effect is well-established for TMS applied at stimulus onset asynchronies (SOAs) around 70e130 msec (e.g., Amassian et al., 1989; Corthout, Uttl, Ziemann, Cowey, & Hallett, 1999; Sack, van der Mark, Schuhmann, Schwarzbach, & Goebel, 2009, see Kammer, 2007, for review). Several chronometric TMS studies investigating the temporal profile of EVC involvement in visual perception have revealed additional time windows prior to stimulus onset at which TMS over EVC disrupts multiple aspects of visual perception: visual discrimination (Corthout, Hallett, & Cowey, 2003; Corthout, Uttl, Juan, Hallett, & Cowey, 2000; Laycock, Crewther, Fitzgerald, & Crewther, 2007), subjective visibility, and priming (Jacobs, de Graaf, Goebel, & Sack, 2012). The problem with pre-stimulus masking is that TMS pulses prior to visual stimuli may elicit all kinds of non-neural or non-specific effects, such as attentional priming, multisensory priming/integration (due to the ‘click’ of the pulse), and most importantly the induction of eye blinks. In previous
work, we already addressed several of these alternative explanations of pre-stimulus masking effects, for instance by removing trials with eye blinks (Jacobs, Goebel, & Sack, 2012), controlling for sound with Sham TMS (de Graaf, Cornelsen, Jacobs, & Sack, 2011; Jacobs, Goebel, et al., 2012), and controlling for sensory stimulation of the skin with vertex TMS (Jacobs, Goebel, et al., 2012). Therefore, we concluded that a neural mechanism underlies (at least part of) the obtained pre-stimulus TMS masking effects. We proposed that prestimulus TMS exerts its effects by putting EVC in a suboptimal state (de Graaf, Cornelsen, et al., 2011), perhaps one of rhythmic neuronal firing at an ineffective frequency and/or phase, thereby hampering subsequent visual processing (Jacobs, Goebel, et al., 2012). In Jacobs, Goebel, et al. (2012), we reported a rather broad time interval in which TMS could negatively influence visual perception of symbolic arrow stimuli ranging from 80 to 40 msec at group level, but showing more narrow effective time windows in single participants. Another study by our group showed impaired visual discrimination and subjective visibility of bar stimuli for high-intensity (>65% maximal stimulator output) EVC-TMS at 25 msec (de Graaf, Cornelsen, et al., 2011). Since the latter was not a chronometric TMS study, we cannot exclude the possibility that both prestimulus time windows are part of a broader period of EVC relevance that stretches from 25 to 80 msec. Yet, other studies have reported multiple, separate time windows of visual suppression by TMS within the pre-stimulus time frame (Corthout, et al., 2003; Corthout, et al., 1999). The identification of two distinct time periods of EVC perceptual relevance leaves room for multiple interpretations: it could imply a single neural mechanism that comes into play repetitively, or two separate neural mechanisms which independently occur in EVC, but which are both necessary for accurate visual perception. Here, our aim is to investigate these two alternatives, and as such, to shed new light on the role of pre-stimulus processes in EVC for visual perception. In the current project, we investigated the relevance of prestimulus brain state in EVC in two separate experiments. In Experiment 1, we used a paradigm of increasing magnetic stimulation strength (de Graaf, Cornelsen, et al., 2011) to test possible TMS masking effects at 50 msec, and at 20 msec. In contrast to our previous explorations, we presented stimuli at the TMS-targeted visual field location and in a control location, allowing us to evaluate retinotopic specificity of potential masking effects. Moreover, we compared these results to the pattern of masking effects for two post-stimulus masking windows (þ90 and þ120 msec), always measuring both objective (forced-choice stimulus orientation determination) and subjective (stimulus visibility rating) visual processing. Looking ahead, we found retinotopic TMS masking at 20 msec and non-retinotopic TMS masking at 50 msec, supporting a separation of two pre-stimulus masking windows with fundamentally different underlying mechanisms. Previous work using symbolic arrow stimuli (Jacobs, Goebel, et al., 2012) indeed found masking effects at 50 msec, but found no suppression at 20 msec. To elucidate these matters, in Experiment 2 we measured a range of SOAs, using symbolic arrow stimuli. Moreover, we measured EoG simultaneously, to later evaluate the influence of eye blinks
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
3
Fig. 1 e Two examples of trials in Experiment 1. A horizontal or vertical bar stimulus was presented on each trial in either the lower left quadrant (experimental stimulus location) or in the upper right quadrant (control stimulus location) of the visual field. The stimulation intensity varied across trials and ranged from 35 to 80% maximal stimulator output. On each trial a single TMS pulse was delivered at one of four possible time windows: 2 pre-stimulus time windows (¡50 msec and ¡20 msec), and two post-stimulus time windows (þ90 msec and þ120 msec). In Trial 1 (Tr 1), the TMS pulse precedes a horizontal bar stimulus presented at the experimental visual field location by 20 msec. In Trial 2 (Tr 2), the TMS pulse follows a vertical bar stimulus presented at the control visual field location by 90 msec.
on visual performance. Also here, we tested performance when stimuli were presented in the TMS-targeted visual field location, and in a control location. Looking ahead, there was no visual suppression of arrow stimuli at 20 msec, in contrast to the orientation stimuli of Experiment 1. But TMS masking at 50 msec was replicated, and it was again nonretinotopic. However, crucially, when we excluded trials that were contaminated by eye blinks we still found TMSsuppression of visual stimuli across the visual field. Thus, taken together, the experiments detailed below suggest that at least two pre-stimulus masking windows exist both of which are of neural origin. One of these is retinotopic and obtained with one set of stimuli yet absent with another set of stimuli. The other is non-retinotopic, not stimulus-dependent, but nevertheless neural in origin, or at least not caused by TMSinduced eye blinks. We discuss implications and potential mechanisms below.
2.
Experiment 1
2.1.
Materials and methods
2.1.1.
Participants
Eleven participants volunteered for this study. Three (including authors T.G. and C.J.) were aware of the goals and purposes of this study, while eight were fully naı¨ve. No participants had abnormal vision or a history of neuropsychiatric disorders. Participants were screened by a medical supervisor prior to participation and supplied written informed consent. Experimental protocols were approved by the local medicalethical committee.
2.1.2.
Stimuli, tasks, design
Stimuli involved small rectangular bars presented at two possible locations: a TMS-targeted location to the lower left (LL), and a control location to the upper right (UR) of a central fixation cross. Bars were either horizontal or vertical, and greyscale. Their luminance was not fixed, but based on an initial subject-specific stimulus calibration measurement. Participants performed a visual discrimination task on the
horizontal versus vertical (2-alternative forced-choice: 2AFC) bars (using keyboard button presses: left index finger on ‘Z’ key for vertical bars, right index finger on ‘/’ key for horizontal bars). Twenty trials for nine stimulus brightness levels allowed us to plot performance separately for each stimulus location. Per location, the stimuli leading to performance closest to each other and to 90% correct discrimination were selected. We thus at the same time ensured equal difficulty of the tasks for both stimulus locations, and prevented ceiling effects on performance. All stimuli, in all phases of the experiment, were presented for 2 frames at a refresh rate of 60 Hz on a standard TFT monitor (Samsung SyncMaster): 33.4 msec, using Presentation software (Neurobehavioral Systems, CA, USA). In the actual experiment, participants performed the same 2AFC discrimination task (see Fig. 1), and a subjective stimulus orientation visibility judgment task e rating the visibility of the orientation on a scale of 1e4. We emphasized that it was not a confidence rating, but explicitly a subjective visibility rating. We indicated that 1 was equivalent to “did not see the stimulus at all”, 4 was equivalent to “clearly saw the stimulus”, and 2 and 3 parametrically in-between. During analysis, it appeared that two participants had inverted the visibility rating scale. We inverted their average visibility ratings so that the pattern of results matched the other subjects and their objective performance. Trial duration was fixed and jittered around 6 sec on average from the range 5e7 sec. There were three breaks in the session, in which participants were invited to move and rest their eyes.
2.1.3.
TMS parameters
TMS was administered using a figure-8 coil (MC-B70) on a MagPro X100 stimulator. Single biphasic TMS pulses were administered to occipital cortex with the handle pointing outwards. All conditions were measured in a single experimental session, consisting first of the stimulus calibration phase, then a localization phase, then the experimental measurement. Localization was done using the phosphene localization method (as in de Graaf, Cornelsen, et al., 2011; de Graaf, Herring, & Sack, 2011). It was assumed that positioning the coil to elicit phosphenes in a particular visual field location
4
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 2 e Results Experiment 1: Objective visual processing. Proportion of correct responses per level of TMS intensity as quantified in percentage of maximum stimulator output (MSO). Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL).Separate graphs correspond with the four different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse: ¡50 ms (A), ¡20 ms (B), þ90 ms (C), and þ120 ms (D), respectively. Error bars represent standard error of mean.
would allow disruption of visual processing in the same visual field location with appropriate TMS intensity. TMS thus targeted the LL region of the visual field where visual targets appeared on half of the trials. To prevent order-effects, yet prevent also undesirable surprise-effects, trials with visual targets in the targeted LL region were presented in short blocks, as were targets appearing in the non-targeted UR region. These blocks were always announced with a short instruction screen, indicating the upcoming target location. Subjects were then instructed to maintain fixation but covertly attend the thus cued location throughout the block. Aside from the stimulus position condition (StimPos), we implemented an SOA condition with four levels: TMS pulses administered at 50, 20, þ90 msec, þ120 msec (see Fig. 1), and a stimulation intensity condition (Int) with five levels (35%, 50%, 60%, 70%, 80% of machine output). Stimulus-TMS SOAs were tested and confirmed in a separate calibration measurement session using a photodiode and oscilloscope
recording the pulses to the TMS device and the actual appearance of stimuli in our experiment on screen, thus allowing us to correct for the difference in response delay at the two locations on the monitor. In contrast to during a previous experiment (de Graaf, Cornelsen, et al., 2011), our setup now also allowed us to use serial communication between the stimulus PC and the TMS device to automatically set the TMS intensity for every trial, using Presentation software (Neurobehavioral Systems, CA, USA). The intensity was thus pseudo-randomized throughout the experiment. In total, for all 40 condition cells (StimPos SOA Int), 16 trials were measured per participant, leading to a total of 640 trials per participant per session.
2.1.4.
Analysis
Only trials on which participants gave both an orientation and visibility judgment after the stimulus/TMS pulse and before the onset of the next trial were included (in total 99.2% of all trials). We performed a full-model repeated-measures ANOVA (RM-
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
5
Fig. 3 e Results Experiment 1: Subjective visual processing. Average subjective visibility rating per level of TMS intensity as quantified in percentage of maximum stimulator output (MSO). Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL).Separate graphs correspond with the four different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse: ¡50 msec (A), ¡20 msec (B), þ90 msec (C), and þ120 msec (D), respectively. Error bars represent standard error of mean.
ANOVA) with factors StimPos, SOA, and Int, separately for the accuracy (hit rate: proportion correct) over participants, and subjective visibility ratings, using SPSS v19 software (IBM, New York, NY). Follow-up tests were RM-ANOVAs as reported in Results (all RM-ANOVA results with conservative GreenhouseGeisser correction). “Trend analyses” reported below reflect default tests of within-subject contrasts provided with RMANOVAs by SPSS, focusing on linear and quadratic contrasts. In addition, RM-ANOVA analyses were performed on the signal detection measure of d0 , which we computed from the recoded discrimination responses (i.e., stimulus ‘horizontal’ & response ‘horizontal’ ¼ hit; stimulus ‘horizontal’ & response ‘vertical’ ¼ miss; stimulus ‘vertical’ & response ‘horizontal’ ¼ false alarm; stimulus ‘vertical’ & response ‘vertical’ ¼ correct rejection). The outcomes of these analyses were identical to the results from the ANOVA analyses on raw accuracy scores as reported in the section below. Please refer to the Supplementary Material for an additional results
section on the analyses of the d0 measure and a graphical representation of d’ across conditions.
2.2.
Results
2.2.1.
Objective visual processing
On the 2AFC orientation discrimination task, the full-model RMANOVA with factors SOA, StimPos, and Int revealed a three-way interaction (F(5.3,53.4) ¼ 2.8; p ¼ .025), indicating that the effects of stimulus position and intensity were dependent on SOA. We thus performed follow-up RM-ANOVAs separately for each time window. Shown in Fig. 2 is average proportion of trials correct over participants, plotted over levels of stimulation intensity, separately for the LL and UR visual stimulus positions. Fig. 2C suggests that for the ‘classical’ masking window around þ90 msec there was a clear masking effect of TMS pulses, with increasingly disrupted visual task performance with increasing TMS intensity. This effect seems specific to
6
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 4 e Two examples of trials in Experiment 2. A left or rightward directed horizontal arrow stimulus was presented on each trial in either the lower left quadrant (experimental stimulus location) or in the upper right quadrant (control stimulus location) of the visual field. The stimulation intensity was fixed to 70% of maximal stimulator output. On each trial a single TMS pulse was delivered at one of five possible pre-stimulus time windows: ¡80, ¡60, ¡40, ¡20 or 0 msec. In Trial 1 (Tr 1), the TMS pulse precedes a leftward pointing arrow stimulus presented at the experimental visual field location. In Trial 2 (Tr 2), the TMS pulse precedes a rightward pointing arrow stimulus presented at the control visual field location.
stimuli on the LL. For UR, increasing TMS intensity appears to have no effect. Indeed, the RM-ANOVA showed a significant interaction StimPos Int (F(3.1,31.1) ¼ 3.1; p ¼ .042). Follow-up RM-ANOVA confirmed that Int has an effect for LL stimuli (F(2.9,29.3) ¼ 7.2; p ¼ .001), but not for UR stimuli (p ¼ .474). Moreover, trend analysis revealed a significant linear component for LL (F(1,10) ¼ 21.7; p ¼ .001), suggesting that increasing intensity disrupts performance linearly. Fig. 2D suggests that while there seems to be an effect of TMS at SOA þ120 msec at the highest intensities, specifically on LL, this is not as clear or pronounced as for the classical window. Statistically, there was no significant interaction between StimPos and Int (p ¼ .163), but follow-up analysis did yield a significant effect of Int on LL trials (F(2.5,24.7) ¼ 3.8; p ¼ .030), though not at all on UR trials (p ¼ .427). Interestingly, trend analysis results were in line with the observation that pulses at þ120 msec only successfully masked visual stimuli at the highest intensities. There was no linear (p ¼ .124) but only a quadratic trend (F(1,10) ¼ 18.6; p ¼ .002). The two pre-stimulus time windows paint a very dissimilar picture. For the 20 msec SOA, there appears to be a clear masking effect that is specific to LL (Fig. 1B). Indeed, RMANOVA revealed a significant interaction between StimPos and Int (F(2.4,23.6) ¼ 5.3; p ¼ .010). Follow-up tests showed a strong significant effect of Int for LL trials (F(2.3,22.6) ¼ 7.3; p ¼ .003), that moreover was significantly linear (F(1,10) ¼ 13.3; p ¼ .004). There was no effect of Int for UR (p ¼ .589). This suggests that TMS masking at this pre-stimulus time window constitutes ‘true’ masking, that is robust, strong, and retinotopically specific. Interestingly, the latter is not the case for the 50 msec window. Fig. 2A does show strong disruptions of visual task performance, as in 20 msec, but it does not seem to be specific to the targeted visual field location. Statistically, there was actually a significant interaction between StimPos and Int (F(3.1,31.1) ¼ 3.5; p ¼ .025), but follow-up analysis showed that while Int indeed had an effect on LL (F(2.8,27.7) ¼ 8.1; p ¼ .001), it
also had a strong effect on UR (F(2.7,26.8) ¼ 11.1; p < .001). This suggests that the effects of TMS in this time window, with these stimuli, are not specific.
2.2.2.
Subjective visual processing
For subjective visibility ratings (scale of 1e4), the full-model RM-ANOVA revealed a very strong 3-way interaction between SOA StimPos Int (F(4.2,42.3) ¼ 4.9; p ¼ .002). We thus proceeded, similarly to the analysis for the objective discrimination performance, to analyze separately per SOA. The patterns for these SOA conditions are shown in Fig. 3. Fig. 3C shows the visibility ratings averaged over participants, plotted over levels of TMS intensity separately for stimuli presented on the LL and UR with pulses applied at SOA þ90 msec. It appears that, analogously to the objective task performance results, there was a specific effect of Int for LL trials. Statistically, indeed there was an interaction between StimPos and Int (F(1.6,16.0) ¼ 11.4; p ¼ .001), and follow-up tests revealed a significant effect of Int for LL (F(1.5,15.3) ¼ 22.6; p < .001) that was linear (F(1,10) ¼ 28.5; p < .001, although now a quadratic trend was also supported; F(1,10) ¼ 5.6; p ¼ .040), and not for UR (p ¼ .676). Also in the other time windows, the effects of TMS on subjective visibility ratings followed closely the patterns of results on behavioral discrimination performance. For poststimulus SOA þ120 msec (Fig. 3D), the interaction between StimPos and Int was significant (F(3.2,32.4) ¼ 4.6; p ¼ .007) (even though this interaction did not reach significance for the objective measure, see above). Again, there was a significant effect of Int for LL (F(2.1,21.1) ¼ 5.0; p ¼ .015) that was linear (F(1,10) ¼ 9.3; p ¼ .012), but not for UR (p ¼ .795). A main reason to include the þ120 msec SOA in this study was to evaluate whether e in contrast to þ90 msec e there would be a dissociation between subjective and objective visual processing (Koivisto, Railo, & Salminen-Vaparanta, 2011). The data appear inconclusive on this point, since on the one hand there was a significant interaction between TMS intensity and
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
stimulus location for the subjective measure only, yet on the other hand the patterns of results for subjective and objective processing seemed highly similar. We therefore leave this question for future studies. For the pre-stimulus windows, there was a significant interaction StimPos Int for SOA 20 msec (F(2.0;20.4) ¼ 13.9; p < .001), and a significant effect of Int for LL (F(1.6,16.3) ¼ 15.2; p < .001) that was linear (F(1,10) ¼ 20.6; p ¼ .001) but no effect for UR (p ¼ .662). For visibility ratings in SOA 50 msec there was no interaction StimPos Int (p ¼ .208), but a main effect of Int (F(1.9,18.5) ¼ 18.1; p < .001). This again confirms a disruption of visual processing that is non-retinotopic.
3.
Experiment 2
3.1.
Materials and methods
3.1.1.
Participants
Ten participants volunteered for this study, 4 of which also participated in Experiment 1. No participants had abnormal vision or a history of neuropsychiatric disorders. One participant was later excluded from data-analysis as her data contained a disproportional amount of eye blinks. Participants were screened by a medical supervisor prior to participation, and supplied written informed consent. Experimental protocols were approved by the local medical-ethical committee.
3.1.2.
Stimuli, tasks, and design
Horizontal arrow stimuli pointing left or right and subtending 1.04 by .44 visual angle were presented at 4 eccentricity from a central fixation cross either in the LL quadrant of the visual field, or in the UR quadrant. Arrow stimuli consisted of multiple black arrow shapes on a white background (see Fig. 4). All stimuli were presented on a standard TFT monitor (Samsung SyncMaster; refresh rate 60 Hz) for a duration of 16.7 msec. The experiment consisted of two sessions; an experimental TMS session and a Sham TMS control session. The Sham session was included to control for possible baseline performance differences across the two visual field locations, and to control for non-specific TMS effects related to the clicking sound of the magnetic pulses. Each session started with the application of electrodes near participants' eyes for the purpose of recording electrooculographical (EoG) activity in the eye muscles. Ag/AgCl electrodes were prepared above and below participants' left eye to measure eye-movement related potential differences and to the mastoid behind their right ear, which served as a neutral reference. EoG data were recorded with Vision Recorder (BrainProducts, Gilching (Mu¨nich), Germany) software at a sampling rate of 500 Hz (for more details on EoG data acquisition, see Jacobs, Goebel, et al., 2012). Next, participants completed a practice run consisting of 48 trials to acquaint them with the forced-choice visual discrimination task. On each trial they were presented with an arrow stimulus and they had to subsequently indicate by button press 1) in which direction the arrow was pointing, and 2) to what extent they consciously perceived the arrow stimulus. Response buttons were 4 keys on a standard QWERTYkeyboard: the Z, the X, the dot and the slash keys. Response
7
coding for Response 1) was such that a button press of Z or X with the left hand corresponded to ‘left’ and a button press of any of the other two keys with the right hand corresponded to ‘right’. Response coding for Response 2) was such that the Zkey corresponded with complete awareness of the stimulus, the X-key with almost complete awareness, the dot-key with almost no unawareness and the slash-key with no awareness at all. During practice trial-by-trial feedback was given on discrimination performance by a color change of the fixation cross. After the practice run, the TMS coil was first positioned at the correct location over the right EVC (see TMS parameters). During the experimental phase of the session, the task was identical to the practice run (see Fig. 4), except for the now absent feedback. Data was acquired in blocks of 24 trials. Within blocks the location of the visual stimuli was fixed, but their location would alternate between LL and UR across blocks. Half of the participants started with a block of stimuli in LL visual field, whereas the other half would start with a block of stimuli in UR visual field. The first two blocks consisted of a no TMS baseline measurement. During the next 10 TMS blocks, single pulses of TMS were administered at varying SOAs. This resulted in a total number of 24 trials per TMS time window per visual field location. Trial order was randomized within blocks. Trials were separated by a jittered inter-trial-interval of ~5250 msec. Stimulus presentation and behavioral response recording was performed with the Presentation (NeuroBehavioral Systems Inc, Albany, CA) software package.
3.1.3.
TMS parameters
For the experimental EVC-TMS session, identical TMS equipment, pulse type and coil orientation were employed as in Experiment 1. Again, coil positioning was based on the phosphene localization procedure. Stimulation intensity was fixed to a maximum of 70% maximum stimulator output for each participant. In each trial a single TMS pulse was delivered to right EVC at any of 5 possible pSOAs: 80, 60, 40, 20 or 0 msec relative to the onset of the visual stimulus (see Fig. 4). In the Sham-TMS session, the TMS coil was replaced with the MC-P-B70 placebo coil, which elicits a clicking sound comparable to the MC-B70 coil, but in absence of the magnetic stimulation of the underlying brain tissue. Stimulator intensity was set to a value that matched the Sham TMS pulses to the real TMS pulses with regard to clicking sound volume. As in the EVC-TMS session, single TMS pulses at either 80, 60, 40, 20 or 0 msec SOAs were applied. Also in Experiment 2 we measured and corrected for the different presentation delays for the two stimulus locations, by means of the calibration procedure described earlier.
3.1.4.
Analysis
The EoG data was preprocessed in VisionAnalyzer (BrainProducts, Gilching (Mu¨nich), Germany) and exported to Matlab vR2012a (The MathWorks Inc, Natick, MA) for further custom analyses. Based on the EoG amplitude in the period running from 200 to þ100 msec relative to the visual stimulus trials were classified as either ‘blink’ or ‘no blink’ trials (see Jacobs, Goebel, et al., 2012; Skotte, Nojgaard, Jorgensen, Christensen, & Sjogaard, 2007, for further details). Blink trials were
8
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 5 e Results Experiment 2. Top panels (A) depict accuracy data, i.e., proportion of correct responses for the different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse for the EVC (left panel) and Sham (right panel) TMS session. Bottom panels (B) depict subjective visibility rating per SOA for EVC (C) and Sham (D) TMS session. Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL). Error bars represent standard error of mean.
excluded from the behavioral dataset and statistical analyses were all conducted on this reduced dataset. Using the SPSS v19 statistical software package (IBM, New York, NY), we performed a three-factorial repeated-measures ANOVA (RM-ANOVA) with factors TMS Type (2 levels), StimPos (2 levels) and SOA (6 levels), separately for the accuracy (hit rate: proportion correct) over participants, and subjective visibility ratings (1- to 4-point scale). Follow-up tests were RMANOVAs per level of TMS Type and pairwise comparisons as reported in Results (Greenhouse-Geisser correction in case of violation of sphericity assumptions). Here too, RM-ANOVA analyses were performed additionally on the signal detection measure of d0 , which we computed from the recoded discrimination responses (i.e., stimulus ‘left’ & response ‘left’ ¼ hit; stimulus ‘left’ & response ‘right’ ¼ miss; stimulus ‘right’ & response ‘left’ ¼ false alarm; stimulus ‘right’ & response ‘right’ ¼ correct rejection). The outcomes of these analyses were identical to the results from the ANOVA analyses on raw accuracy scores as reported in the section below. Please refer to the Supplementary Material for an additional
results section on the analyses of the d’ measure and a graphical representation of d’ across conditions.
3.2.
Results
We excluded one participant based on the disproportional amount of eye blinks (>40%) she displayed, which did not allow valid statistical analyses on the proportion of remaining trials. The other nine participants blinked within the crucial temporal window on 4.1% of trials on average. After removal of all trials containing eye blinks and after exclusion of one participant's data, we conducted a three-way RM-ANOVA on the remaining data.
3.2.1.
Objective visual processing
The three-way ANOVA analysis revealed significant main effects of the factors TMS Type (F(1, 9) ¼ 10.4; p ¼ .012) and SOA (F(1.6, 13.1) ¼ 7.1; p ¼ .011), and a significant interaction effect between these two factors (F(1.8, 14.7) ¼ 7.6; p ¼ .006). No main or interaction effects for the factor StimPos appeared (all p > .5).
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Based on these outcomes, we pooled the data for the factor StimPos and performed two one-way RM-ANOVAs on the factor SOA: one for the EVC-TMS data and one for the ShamTMS data. The EVC-TMS data revealed a main effect of SOA (F(1.7, 13.2) ¼ 7.8; p ¼ .008). Post-hoc tests revealed significant differences between the No TMS baseline condition (mean proportion correct ¼ .98) and the 80 msec (mean ¼ .79; p ¼ .008) and 60 msec (mean ¼ .79; p ¼ .006) TMS time windows (see Fig. 5A). The Sham-TMS data did not show an SOA effect. Thus, pre-stimulus TMS applied 80 or 60 msec prior to visual stimulation impairs discrimination performance in a nonretinotopic fashion.
3.2.2.
Subjective visual processing
Results of the three-way ANOVA on the subjective visibility ratings were similar to those of the accuracies. Again, main effects of both TMS Type (F(1, 8) ¼ 26.7; p ¼ .001) and SOA (F(1.5, 12.1) ¼ 7.1; p ¼ .013) were revealed as well as a Type SOA (F(1.6, 12.7) ¼ 10.8; p ¼ .003) interaction effect. Because this analysis showed that visual field location did not affect subjective visibility, we again pooled the factor StimPos, and conducted a one-way RM-ANOVA on the factor SOA for each level of TMS Type. The SOA of real EVC-TMS, but not of Sham-TMS, significantly affected visibility ratings (F(1.5, 11.9) ¼ 9.6; p ¼ .005). Posthoc analyses showed a broader effective TMS period for the visibility ratings than for accuracy. EVC-TMS significantly reduced subjective visibility when applied at 80 msec (mean ¼ 2.2 on a 1- to 4-point scale; p ¼ .001), 60 msec (mean ¼ 2.1; p ¼ .001) and 40 msec (mean ¼ 2.4; p ¼ .011) compared to No TMS baseline (mean ¼ 3.2; see Fig. 5B). Summarizing, both visual discrimination ability and subjective visual awareness are affected by pre-stimulus EVCTMS, but not by Sham TMS. This effect is time-specific and ranges from 80 msec to 60 msec for discrimination performance, and broadened into the 40 msec time window for visibility ratings. In contrast to our findings from Experiment 1, no significance decrease in any of the two measures of visual processing occurred when TMS was applied at the 20 msec time point. None of the detected effects proved to be influenced by the position in the visual field at which the visual stimuli were presented.
4.
Discussion
The current study consists of two experiments in which we explored whether ongoing neural activity (brain state) in EVC modulates the perception of future visual information. To this end, we applied chronometric focal TMS to EVC at various SOAs before and after visual stimulus onset. Moreover, we used different stimuli and visual field locations. Experiment 1 revealed non-retinotopic TMS-induced perceptual impairment at a pSOA of 50 msec. Yet, a TMS pulse administered to EVC at 20 msec selectively disrupted visual perception at the corresponding topographic location in the visual field. These results match a study by Corthout et al. (2003) who described in three subjects a pattern of current-direction (CD)-dependent visual suppression at 20 msec and CD-independent
9
visual suppression at 50 msec, and Corthout et al. (1999) showing in three subjects a coil location dependent suppression effect at 10 msec but coil location independent effect at 60, 50 msec. In the current study, TMS over EVC also resulted in topographically impaired visual perception at two post-stimulus timings (90 and 120 msec), an effect which has previously been reported and attributed to direct interference with stimulus-related neural activity in EVC (e.g., Amassian, et al., 1989; Corthout, et al., 1999; Sack, et al., 2009). In contrast to Experiment 1, the results of Experiment 2 showed a single broad window, centered around 60 msec, in which pre-stimulus activity in EVC influenced objective and subjective visual perception in a non-topographic way. This masking effect was obtained after removal of trials with eye blinks, demonstrating a neural origin. Yet, while this window thus converged with the 50 msec TMS effect observed in Experiment 1, disruption of EVC activity 20 msec before stimulus onset did not affect visual perception at all. And indeed no corresponding topographically specific pre-stimulus effect was found at any other pSOAs either. Taken together, these two experiments thus revealed a clear dissociation between two pre-stimulus mechanisms: one pSOA (20 msec) which is retinotopically specific, and one pSOA (~50 msec) which is retinotopically unspecific. It remains an empirical question why the late retinotopic TMS effect was absent in Experiment 2, since the experiments differed in more than one-way. We had no clear a priori hypothesis expecting one type of stimulus to generate qualitatively different effects as compared to the other stimulus type. Based on our current data, it is difficult to infer where the stimulus-related difference in results comes from, since the stimuli differ across several dimensions, from low-level properties like contrast and size, to timing (stimulus duration), to high-level properties like the presence or absence of symbolic meaning in the stimulus. Previous work has started to compare TMS masking effects for different types of stimuli (de Graaf, Goebel, & Sack, 2012; Koivisto et al., 2011) for conventional post-stimulus SOAs. So far, to our knowledge no studies reported systematic and significant stimulus-driven differences in results for pre-stimulus SOAs (though see Kammer, Puls, Strasburger, Hill, & Wichmann, 2005, concerning luminance). It would be useful to conduct systematic investigations in future work. To guide such future work, we would briefly point out two possible causes for the potential stimulus-dependence of the 20 msec masking effect obtained here. On the one hand, since the arrow stimuli were larger and higher-contrast, they could have been more robust and less susceptible to the 20 msec TMS masking effect. But since there was not even a hint of masking for the arrow stimuli, the difference in results across experiments could alternatively signify something more fundamental. For example, compared to the simple orientation stimuli of Experiment 1, the symbolic arrows of Experiment 2 were more complex, higher in informational/symbolic content, and therefore possibly processed (at least in part) in cortical regions hierarchically further up the visual stream. Orientation information is actually predominantly processed in EVC, the region directly targeted by TMS here. That could account for higher susceptibility of the used bar stimuli to disruptions of local cortical processing.
10
4.1.
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Local pre-stimulus processing at 20 msec
At least for orientation stimuli in Experiment 1, ongoing EVC processing occurring some 20 msec prior to visual stimulation is crucial to unimpaired visual perception. This appears to be relatively local processing, because the TMS effect at 20 msec is retinotopic. Retinotopy is most prominent in LL visual cortices EVC, simple orientation processing is focused in these regions, and EVC was the site directly targeted by TMS in this study. It thus seems plausible that this particular pre-stimulus suppression effect originated from the cortical area right underneath the coil, although we cannot completely rule out spreading of the TMS to higher-level visual areas within the same hemisphere which contain cruder retinotopic maps. The nature of such local processing, and the circuitry involved, remain an open question. A question made all the more interesting because the effects of a single TMS pulse need to last quite some time, when taking into account retino-striate transmission time on top of the pSOA. From paired-pulse studies in the motor system, we know that single TMS pulses can have inhibitory and facilitative effects (Chen, 2004) outlasting the duration of the pulse itself. Suprathreshold TMS causes a decrease in motor-evoked potentials (MEPs) after a long inter-pulse interval of 50e200 msec: an effect presumably modulated by intracortical GABAergic activity on slower GABAb receptors (McDonnell, Orekhov, & Ziemann, 2006). If similar cortical circuitry holds in early visual regions, the suprathreshold TMS pulses we applied here might have recruited intracortical inhibitory networks within occipital cortex, which reduce cortical excitability with a delay of at least 50 msec, but potentially up to 200 msec. In line with this, Moliadze, Zhao, Eysel, and Funke (2003) found that strong single TMS pulses to occipital cortex can have inhibitory effects on single visual cortex cells for some 100e200 msec. It will be interesting for future work to evaluate to what extent such mechanisms apply to occipital processing in the context of pre-stimulus TMS masking.
over vertex). It thus seems that the 50 msec suppression effect involves a time-specific, relatively global (because non-retinotopic and stimulus-independent), but nevertheless neural mechanism. One physiological explanation within these constraints would be that the visual cortex is continuously alternating between states of high and low excitability and that a TMS pulse affects this alternation. For example, recent evidence shows that both the power and phase of prestimulus oscillations in the alpha frequency (i.e., ~8e12 Hz) range have been associated with stimulus visibility (Busch, et al., 2009; van Dijk, et al., 2008; Dugue, et al., 2011; Hanslmayr, et al., 2007; Mathewson, et al., 2009; Romei, et al., 2008; Toscani, et al., 2010). In this context, a TMS pulse might phase-lock such oscillations, inducing a period of decreased excitability at a fixed later stage. Previous research suggests at least that a set of several rhythmic pulses can have such phase-locking effects (Thut et al., 2011), which can be attentionally/ perceptually relevant (Romei et al., 2010). One alternative explanation might be a spreading of the physical TMS stimulation from one hemisphere to the other through transcallosal connections, leading to visual suppression in both visual hemifields. This seems unlikely, since in Experiment 1 we showed a very similar pattern of TMS masking for both visual field locations across intensities. A ‘spreading explanation’ of the induced electric activity would presumably be reflected by a weaker masking effect for the suboptimal visual field location. Relatedly, it is also an empirical question whether TMS pulses over EVC could spread to higher visual or attentional areas (e.g., parietal regions). An indirect effect on such functionally relevant higher cortical areas, whether directly or through backpropagation of the pulse energy to a wider area of early visual cortices, would also be in line with the neural yet non-retinotopic and stimulus-independent suppression effect. For the moment, however, this seems rather speculative.
4.3. 4.2.
Conclusion
Non-local pre-stimulus processing at 50 msec
The even earlier pre-stimulus suppression effect must in any case involve a different mechanism, since it is not retinotopically specific and seems independent of stimulus parameters. The temporal specificity of the observed suppression effect moreover implies that visibility depends not just on a general on/off state in a broad time period, but rather on the state of the visual cortex at a particular time point prior to the arrival of the visual input. Previous studies have ascribed pre-stimulus TMS effects to non-specific TMSinduced eye blinks (Beckers & Homberg, 1991; Corthout, et al., 2000). However, when taking together this study with our previous work, we have aimed to control for all possible non-neural effects of TMS pulses. Here and previously (Jacobs, Goebel, et al., 2012), blinking could not fully account for the pre-stimulus TMS effect around 50 msec. In Jacobs, Goebel, et al., 2012; Jacobs, de Graaf, et al., 2012, we also controlled for auditory effects (Sham TMS over occipital cortex) and somatosensory stimulation (real TMS
We here provided evidence for two distinct occipital mechanisms involved in biasing successful processing of upcoming visual inputs. Since these mechanisms must both involve functionally relevant neuronal processing, future studies can more directly assess their cortical loci, their computational dynamics, and potentially their functional interplay.
Acknowledgments The work presented here was supported by European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement n [263472]) awarded to A.T.S. T.A.G. is supported by the Netherlands Organization for Scientific Research (NWO 45113-024). The authors would like to thank Azadeh Seyed Tafreshiha for assistance with Experiment 1.
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.cortex.2014.06.017.
references
Amassian, V. E., Cracco, R. Q., Maccabee, P. J., Cracco, J. B., Rudell, A., & Eberle, L. (1989). Suppression of visual perception by magnetic coil stimulation of human occipital cortex. Electroencephalography and Clinical Neurophysiology, 74(6), 458e462. Arieli, A., Sterkin, A., Grinvald, A., & Aertsen, A. (1996). Dynamics of ongoing activity: explanation of the large variability in evoked cortical responses. Science, 273(5283), 1868e1871. Beckers, G., & Homberg, V. (1991). Impairment of visual perception and visual short term memory scanning by transcranial magnetic stimulation of occipital cortex. Experimental Brain Research, 87(2), 421e432. Busch, N. A., Dubois, J., & VanRullen, R. (2009). The phase of ongoing EEG oscillations predicts visual perception. Journal of Neuroscience, 29(24), 7869e7876. Cardin, V., Friston, K. J., & Zeki, S. (2011). Top-down modulations in the visual form pathway revealed with dynamic causal modeling. Cerebral Cortex, 21(3), 550e562. Chen, R. (2004). Interactions between inhibitory and excitatory circuits in the human motor cortex. Experimental Brain Research, 154(1), 1e10. Corthout, E., Hallett, M., & Cowey, A. (2003). Interference with vision by TMS over the occipital pole: a fourth period. NeuroReport, 14(4), 651e655. Corthout, E., Uttl, B., Juan, C. H., Hallett, M., & Cowey, A. (2000). Suppression of vision by transcranial magnetic stimulation: a third mechanism. NeuroReport, 11(11), 2345e2349. Corthout, E., Uttl, B., Ziemann, U., Cowey, A., & Hallett, M. (1999). Two periods of processing in the (circum)striate visual cortex as revealed by transcranial magnetic stimulation. Neuropsychologia, 37(2), 137e145. van Dijk, H., Schoffelen, J. M., Oostenveld, R., & Jensen, O. (2008). Prestimulus oscillatory activity in the alpha band predicts visual discrimination ability. Journal of Neuroscience, 28(8), 1816e1823. Downing, P. E., Chan, A. W., Peelen, M. V., Dodds, C. M., & Kanwisher, N. (2006). Domain specificity in visual cortex. Cerebral Cortex, 16(10), 1453e1461. Dugue, L., Marque, P., & VanRullen, R. (2011). The phase of ongoing oscillations mediates the causal relation between brain excitation and visual perception. Journal of Neuroscience, 31(33), 11889e11893. de Graaf, T. A., Cornelsen, S., Jacobs, C., & Sack, A. T. (2011). TMS effects on subjective and objective measures of vision: stimulation intensity and pre- versus post-stimulus masking. Consciousness and Cognition, 20, 1244e1255. de Graaf, T., Goebel, R., & Sack, A. T. (2012). Feedforward and quick recurrent processes in early visual cortex revealed by TMS? NeuroImage, 61(3), 651e659. de Graaf, T. A., Gross, J., Paterson, G., Rusch, T., Sack, A. T., & Thut, G. (2013). Alpha-band rhythms in visual task performance: phase-locking by rhythmic sensory stimulation. PLoS One, 8(3), e60035. de Graaf, T. A., Herring, J., & Sack, A. T. (2011). A chronometric exploration of high-resolution ‘sensitive TMS masking’ effects on subjective and objective measures of vision. Experimental Brain Research, 209(1), 19e27.
11
Hanslmayr, S., Aslan, A., Staudigl, T., Klimesch, W., Herrmann, C. S., & Bauml, K. H. (2007). Prestimulus oscillations predict visual perception performance between and within subjects. NeuroImage, 37(4), 1465e1473. Hesselmann, G., Kell, C. A., Eger, E., & Kleinschmidt, A. (2008). Spontaneous local variations in ongoing neural activity bias perceptual decisions. Proceedings of the National Academy Sciences of the United States of America, 105(31), 10984e10989. Jacobs, C., Goebel, R., & Sack, A. T. (2012). Visual awareness suppression by pre-stimulus brain stimulation; a neural effect. NeuroImage, 59(1), 616e624. Jacobs, C., de Graaf, T. A., Goebel, R., & Sack, A. T. (2012). The temporal dynamics of early visual cortex involvement in behavioral priming. PLoS One, 7(11), e48808. Kammer, T. (2007). Masking visual stimuli by transcranial magnetic stimulation. Psychological Research, 71(6), 659e666. Kammer, T., Puls, K., Strasburger, H., Hill, N. J., & Wichmann, F. A. (2005). Transcranial magnetic stimulation in the visual system. I. The psychophysics of visual suppression. Experimental Brain Research, 160(1), 118e128. Koivisto, M., Railo, H., & Salminen-Vaparanta, N. (2011). Transcranial magnetic stimulation of early visual cortex interferes with subjective visual awareness and objective forced-choice performance. Consciousness and Cognition, 20(2), 288e298. Laycock, R., Crewther, D. P., Fitzgerald, P. B., & Crewther, S. G. (2007). Evidence for fast signals and later processing in human V1/V2 and V5/MTþ: a TMS study of motion perception. Journal of Neurophysiology, 98(3), 1253e1262. Mathewson, K. E., Gratton, G., Fabiani, M., Beck, D. M., & Ro, T. (2009). To see or not to see: prestimulus alpha phase predicts visual awareness. Journal of Neuroscience, 29(9), 2725e2732. Mathewson, K. E., Prudhomme, C., Fabiani, M., Beck, D. M., Lleras, A., & Gratton, G. (2012). Making waves in the stream of consciousness: entraining oscillations in EEG alpha and fluctuations in visual awareness with rhythmic visual stimulation. Journal of Cognitive Neuroscience, 24(12), 2321e2333. McDonnell, M. N., Orekhov, Y., & Ziemann, U. (2006). The role of GABA(B) receptors in intracortical inhibition in the human motor cortex. Experimental Brain Research, 173(1), 86e93. Moliadze, V., Zhao, Y., Eysel, U., & Funke, K. (2003). Effect of transcranial magnetic stimulation on single-unit activity in the cat primary visual cortex. Journal of Physiology, 553(Pt 2), 665e679. Nandy, A. S., Sharpee, T. O., Reynolds, J. H., & Mitchell, J. F. (2013). The fine structure of shape tuning in area V4. Neuron, 78(6), 1102e1115. Romei, V., Brodbeck, V., Michel, C., Amedi, A., Pascual-Leone, A., & Thut, G. (2008). Spontaneous fluctuations in posterior alphaband EEG activity reflect variability in excitability of human visual areas. Cerebral Cortex, 18(9), 2010e2018. Romei, V., Gross, J., & Thut, G. (2010). On the role of prestimulus alpha rhythms over occipito-parietal areas in visual input regulation: correlation or causation? Journal of Neuroscience, 30(25), 8692e8697. Sack, A. T., van der Mark, S., Schuhmann, T., Schwarzbach, J., & Goebel, R. (2009). Symbolic action priming relies on intact neural transmission along the retino-geniculo-striate pathway. NeuroImage, 44(1), 284e293. Skotte, J. H., Nojgaard, J. K., Jorgensen, L. V., Christensen, K. B., & Sjogaard, G. (2007). Eye blink frequency during different computer tasks quantified by electrooculography. European Journal of Applied Physiology, 99(2), 113e119. Thut, G., Veniero, D., Romei, V., Miniussi, C., Schyns, P., & Gross, J. (2011). Rhythmic TMS causes local entrainment of natural oscillatory signatures. Current Biology, 21(14), 1176e1185. Toscani, M., Marzi, T., Righi, S., Viggiano, M. P., & Baldassi, S. (2010). Alpha waves: a neural signature of visual suppression. Experimental Brain Research, 207(3e4), 213e219.
Available online at www.sciencedirect.com
ScienceDirect Journal homepage: www.elsevier.com/locate/cortex
Research report
Two distinct neural mechanisms in early visual cortex determine subsequent visual processing Christianne Jacobs a,b,c,*,1, Tom A. de Graaf b,c,1 and Alexander T. Sack b,c a
Department of Psychology, FST, University of Westminster, London, UK Department of Cognitive Neuroscience, FPN, Maastricht University, Maastricht, The Netherlands c Maastricht Brain Imaging Center (M-BIC), Maastricht, The Netherlands b
article info
abstract
Article history:
Neuroscience research has conventionally focused on how the brain processes sensory
Received 2 October 2013
information, after the information has been received. Recently, increased interest focuses
Reviewed 18 December 2013
on how the state of the brain upon receiving inputs determines and biases their subse-
Revised 26 February 2014
quent processing and interpretation. Here, we investigated such 'pre-stimulus' brain
Accepted 19 June 2014
mechanisms and their relevance for objective and subjective visual processing. Using non-
Action editor Jason Barton
invasive focal brain stimulation [transcranial magnetic stimulation (TMS)] we disrupted
Published online 16 July 2014
spontaneous brain state activity within early visual cortex (EVC) before onset of visual stimulation, at two different pre-stimulus-onset-asynchronies (pSOAs). We found that
Keywords:
TMS pulses applied to EVC at either 20 msec or 50 msec before onset of a simple orientation
Visual perception
stimulus both prevented this stimulus from reaching visual awareness. Interestingly, only
Early visual cortex
the TMS-induced visual suppression following TMS at a pSOA of 20 msec was reti-
Transcranial magnetic stimulation
notopically specific, while TMS at a pSOA of 50 msec was not. In a second experiment, we
Suppression
used more complex symbolic arrow stimuli, and found TMS-induced suppression only
State-dependence
when disrupting EVC at a pSOA of ~ 60 msec, which, in line with Experiment 1, was not retinotopically specific. Despite this topographic unspecificity of the 50 msec effect, the additional control measurements as well as tracking and removal of eye blinks, suggested that also this effect was not the result of an unspecific artifact, and thus neural in origin. We therefore obtained evidence of two distinct neural mechanisms taking place in EVC, both determining whether or not subsequent visual inputs are successfully processed by the human visual system. © 2014 Elsevier Ltd. All rights reserved.
1.
Introduction
The occipital lobe of the human brain is dedicated to the processing of visual inputs. Early cortical stages of visual
information processing occur approximately 60e100 msec after stimulus presentation in visual areas V1, V2 and V3 of the occipital brain, together commonly referred to as early visual cortex (EVC). Much research has focused on the specific
* Corresponding author. Department of Psychology, Faculty of Science and Technology, University of Westminster, 309 Regent Street, W1B 2HW London, United Kingdom. E-mail address: [email protected] (C. Jacobs). 1 Equal contribution. http://dx.doi.org/10.1016/j.cortex.2014.06.017 0010-9452/© 2014 Elsevier Ltd. All rights reserved.
2
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
visual properties these early visual areas process, their neuronal tuning to these properties, and the information flow within and/or between these regions of EVC (e.g., Cardin, Friston, & Zeki, 2011; Downing, Chan, Peelen, Dodds, & Kanwisher, 2006; Nandy, Sharpee, Reynolds, & Mitchell, 2013). These are examples of how research conventionally focuses on the brain's response to inputs from the environment. Yet, how inputs are processed may depend not only on nature of the information, but also on the prior state of the brain (Arieli, Sterkin, Grinvald, & Aertsen, 1996; Busch, Dubois, & VanRullen, 2009; Hesselmann, Kell, Eger, & Kleinschmidt, 2008; Mathewson, Gratton, Fabiani, Beck, & Ro, 2009). Recent studies have addressed the potential modulatory role of neural state prior to visual input. Electrophysiological studies have shown, for example, that under conditions of near-threshold stimulus visibility, the variability in stimulus perception is indeed reflected in pre-stimulus brain activity (Busch, et al., 2009; van Dijk, Schoffelen, Oostenveld, & Jensen, 2008; Dugue, Marque, & VanRullen, 2011; Hanslmayr et al., 2007; Mathewson, et al., 2009; Romei et al., 2008; Romei, Gross, & Thut, 2010; Toscani, Marzi, Righi, Viggiano, & Baldassi, 2010). So far, the power (van Dijk, et al., 2008; Romei, et al., 2008; Romei, et al., 2010; Toscani, et al., 2010) and phase (Busch, et al., 2009; Dugue, et al., 2011; Mathewson, et al., 2009) of pre-stimulus parieto-occipital oscillations in the alpha frequency band (i.e., 8e12 Hz), and phase-locking in the beta (16e30 Hz) and gamma (>30 Hz) frequency bands (Hanslmayr, et al., 2007) have been related to this perceptual modulation. Next, researchers reversed the line of reasoning and showed that they could affect stimulus visibility by externally manipulating parieto-occipital alpha oscillations (e.g., de Graaf et al., 2013; Mathewson et al., 2012; Romei, et al., 2010), establishing a causal relation between stimulus visibility and pre-stimulus neural state. Together, these studies demonstrate that neural processing that occurs prior to sensory input can play a functional role in perception. By means of Transcranial Magnetic Stimulation (TMS), a brain interference tool that allows temporal disruption of neuronal activity, the contribution of pre-stimulus EVC state to visual perception can be investigated. If applied over occipital cortex, TMS can lead to a complete abolishment of conscious perception of suprathreshold visual stimuli. This effect is well-established for TMS applied at stimulus onset asynchronies (SOAs) around 70e130 msec (e.g., Amassian et al., 1989; Corthout, Uttl, Ziemann, Cowey, & Hallett, 1999; Sack, van der Mark, Schuhmann, Schwarzbach, & Goebel, 2009, see Kammer, 2007, for review). Several chronometric TMS studies investigating the temporal profile of EVC involvement in visual perception have revealed additional time windows prior to stimulus onset at which TMS over EVC disrupts multiple aspects of visual perception: visual discrimination (Corthout, Hallett, & Cowey, 2003; Corthout, Uttl, Juan, Hallett, & Cowey, 2000; Laycock, Crewther, Fitzgerald, & Crewther, 2007), subjective visibility, and priming (Jacobs, de Graaf, Goebel, & Sack, 2012). The problem with pre-stimulus masking is that TMS pulses prior to visual stimuli may elicit all kinds of non-neural or non-specific effects, such as attentional priming, multisensory priming/integration (due to the ‘click’ of the pulse), and most importantly the induction of eye blinks. In previous
work, we already addressed several of these alternative explanations of pre-stimulus masking effects, for instance by removing trials with eye blinks (Jacobs, Goebel, & Sack, 2012), controlling for sound with Sham TMS (de Graaf, Cornelsen, Jacobs, & Sack, 2011; Jacobs, Goebel, et al., 2012), and controlling for sensory stimulation of the skin with vertex TMS (Jacobs, Goebel, et al., 2012). Therefore, we concluded that a neural mechanism underlies (at least part of) the obtained pre-stimulus TMS masking effects. We proposed that prestimulus TMS exerts its effects by putting EVC in a suboptimal state (de Graaf, Cornelsen, et al., 2011), perhaps one of rhythmic neuronal firing at an ineffective frequency and/or phase, thereby hampering subsequent visual processing (Jacobs, Goebel, et al., 2012). In Jacobs, Goebel, et al. (2012), we reported a rather broad time interval in which TMS could negatively influence visual perception of symbolic arrow stimuli ranging from 80 to 40 msec at group level, but showing more narrow effective time windows in single participants. Another study by our group showed impaired visual discrimination and subjective visibility of bar stimuli for high-intensity (>65% maximal stimulator output) EVC-TMS at 25 msec (de Graaf, Cornelsen, et al., 2011). Since the latter was not a chronometric TMS study, we cannot exclude the possibility that both prestimulus time windows are part of a broader period of EVC relevance that stretches from 25 to 80 msec. Yet, other studies have reported multiple, separate time windows of visual suppression by TMS within the pre-stimulus time frame (Corthout, et al., 2003; Corthout, et al., 1999). The identification of two distinct time periods of EVC perceptual relevance leaves room for multiple interpretations: it could imply a single neural mechanism that comes into play repetitively, or two separate neural mechanisms which independently occur in EVC, but which are both necessary for accurate visual perception. Here, our aim is to investigate these two alternatives, and as such, to shed new light on the role of pre-stimulus processes in EVC for visual perception. In the current project, we investigated the relevance of prestimulus brain state in EVC in two separate experiments. In Experiment 1, we used a paradigm of increasing magnetic stimulation strength (de Graaf, Cornelsen, et al., 2011) to test possible TMS masking effects at 50 msec, and at 20 msec. In contrast to our previous explorations, we presented stimuli at the TMS-targeted visual field location and in a control location, allowing us to evaluate retinotopic specificity of potential masking effects. Moreover, we compared these results to the pattern of masking effects for two post-stimulus masking windows (þ90 and þ120 msec), always measuring both objective (forced-choice stimulus orientation determination) and subjective (stimulus visibility rating) visual processing. Looking ahead, we found retinotopic TMS masking at 20 msec and non-retinotopic TMS masking at 50 msec, supporting a separation of two pre-stimulus masking windows with fundamentally different underlying mechanisms. Previous work using symbolic arrow stimuli (Jacobs, Goebel, et al., 2012) indeed found masking effects at 50 msec, but found no suppression at 20 msec. To elucidate these matters, in Experiment 2 we measured a range of SOAs, using symbolic arrow stimuli. Moreover, we measured EoG simultaneously, to later evaluate the influence of eye blinks
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
3
Fig. 1 e Two examples of trials in Experiment 1. A horizontal or vertical bar stimulus was presented on each trial in either the lower left quadrant (experimental stimulus location) or in the upper right quadrant (control stimulus location) of the visual field. The stimulation intensity varied across trials and ranged from 35 to 80% maximal stimulator output. On each trial a single TMS pulse was delivered at one of four possible time windows: 2 pre-stimulus time windows (¡50 msec and ¡20 msec), and two post-stimulus time windows (þ90 msec and þ120 msec). In Trial 1 (Tr 1), the TMS pulse precedes a horizontal bar stimulus presented at the experimental visual field location by 20 msec. In Trial 2 (Tr 2), the TMS pulse follows a vertical bar stimulus presented at the control visual field location by 90 msec.
on visual performance. Also here, we tested performance when stimuli were presented in the TMS-targeted visual field location, and in a control location. Looking ahead, there was no visual suppression of arrow stimuli at 20 msec, in contrast to the orientation stimuli of Experiment 1. But TMS masking at 50 msec was replicated, and it was again nonretinotopic. However, crucially, when we excluded trials that were contaminated by eye blinks we still found TMSsuppression of visual stimuli across the visual field. Thus, taken together, the experiments detailed below suggest that at least two pre-stimulus masking windows exist both of which are of neural origin. One of these is retinotopic and obtained with one set of stimuli yet absent with another set of stimuli. The other is non-retinotopic, not stimulus-dependent, but nevertheless neural in origin, or at least not caused by TMSinduced eye blinks. We discuss implications and potential mechanisms below.
2.
Experiment 1
2.1.
Materials and methods
2.1.1.
Participants
Eleven participants volunteered for this study. Three (including authors T.G. and C.J.) were aware of the goals and purposes of this study, while eight were fully naı¨ve. No participants had abnormal vision or a history of neuropsychiatric disorders. Participants were screened by a medical supervisor prior to participation and supplied written informed consent. Experimental protocols were approved by the local medicalethical committee.
2.1.2.
Stimuli, tasks, design
Stimuli involved small rectangular bars presented at two possible locations: a TMS-targeted location to the lower left (LL), and a control location to the upper right (UR) of a central fixation cross. Bars were either horizontal or vertical, and greyscale. Their luminance was not fixed, but based on an initial subject-specific stimulus calibration measurement. Participants performed a visual discrimination task on the
horizontal versus vertical (2-alternative forced-choice: 2AFC) bars (using keyboard button presses: left index finger on ‘Z’ key for vertical bars, right index finger on ‘/’ key for horizontal bars). Twenty trials for nine stimulus brightness levels allowed us to plot performance separately for each stimulus location. Per location, the stimuli leading to performance closest to each other and to 90% correct discrimination were selected. We thus at the same time ensured equal difficulty of the tasks for both stimulus locations, and prevented ceiling effects on performance. All stimuli, in all phases of the experiment, were presented for 2 frames at a refresh rate of 60 Hz on a standard TFT monitor (Samsung SyncMaster): 33.4 msec, using Presentation software (Neurobehavioral Systems, CA, USA). In the actual experiment, participants performed the same 2AFC discrimination task (see Fig. 1), and a subjective stimulus orientation visibility judgment task e rating the visibility of the orientation on a scale of 1e4. We emphasized that it was not a confidence rating, but explicitly a subjective visibility rating. We indicated that 1 was equivalent to “did not see the stimulus at all”, 4 was equivalent to “clearly saw the stimulus”, and 2 and 3 parametrically in-between. During analysis, it appeared that two participants had inverted the visibility rating scale. We inverted their average visibility ratings so that the pattern of results matched the other subjects and their objective performance. Trial duration was fixed and jittered around 6 sec on average from the range 5e7 sec. There were three breaks in the session, in which participants were invited to move and rest their eyes.
2.1.3.
TMS parameters
TMS was administered using a figure-8 coil (MC-B70) on a MagPro X100 stimulator. Single biphasic TMS pulses were administered to occipital cortex with the handle pointing outwards. All conditions were measured in a single experimental session, consisting first of the stimulus calibration phase, then a localization phase, then the experimental measurement. Localization was done using the phosphene localization method (as in de Graaf, Cornelsen, et al., 2011; de Graaf, Herring, & Sack, 2011). It was assumed that positioning the coil to elicit phosphenes in a particular visual field location
4
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 2 e Results Experiment 1: Objective visual processing. Proportion of correct responses per level of TMS intensity as quantified in percentage of maximum stimulator output (MSO). Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL).Separate graphs correspond with the four different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse: ¡50 ms (A), ¡20 ms (B), þ90 ms (C), and þ120 ms (D), respectively. Error bars represent standard error of mean.
would allow disruption of visual processing in the same visual field location with appropriate TMS intensity. TMS thus targeted the LL region of the visual field where visual targets appeared on half of the trials. To prevent order-effects, yet prevent also undesirable surprise-effects, trials with visual targets in the targeted LL region were presented in short blocks, as were targets appearing in the non-targeted UR region. These blocks were always announced with a short instruction screen, indicating the upcoming target location. Subjects were then instructed to maintain fixation but covertly attend the thus cued location throughout the block. Aside from the stimulus position condition (StimPos), we implemented an SOA condition with four levels: TMS pulses administered at 50, 20, þ90 msec, þ120 msec (see Fig. 1), and a stimulation intensity condition (Int) with five levels (35%, 50%, 60%, 70%, 80% of machine output). Stimulus-TMS SOAs were tested and confirmed in a separate calibration measurement session using a photodiode and oscilloscope
recording the pulses to the TMS device and the actual appearance of stimuli in our experiment on screen, thus allowing us to correct for the difference in response delay at the two locations on the monitor. In contrast to during a previous experiment (de Graaf, Cornelsen, et al., 2011), our setup now also allowed us to use serial communication between the stimulus PC and the TMS device to automatically set the TMS intensity for every trial, using Presentation software (Neurobehavioral Systems, CA, USA). The intensity was thus pseudo-randomized throughout the experiment. In total, for all 40 condition cells (StimPos SOA Int), 16 trials were measured per participant, leading to a total of 640 trials per participant per session.
2.1.4.
Analysis
Only trials on which participants gave both an orientation and visibility judgment after the stimulus/TMS pulse and before the onset of the next trial were included (in total 99.2% of all trials). We performed a full-model repeated-measures ANOVA (RM-
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
5
Fig. 3 e Results Experiment 1: Subjective visual processing. Average subjective visibility rating per level of TMS intensity as quantified in percentage of maximum stimulator output (MSO). Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL).Separate graphs correspond with the four different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse: ¡50 msec (A), ¡20 msec (B), þ90 msec (C), and þ120 msec (D), respectively. Error bars represent standard error of mean.
ANOVA) with factors StimPos, SOA, and Int, separately for the accuracy (hit rate: proportion correct) over participants, and subjective visibility ratings, using SPSS v19 software (IBM, New York, NY). Follow-up tests were RM-ANOVAs as reported in Results (all RM-ANOVA results with conservative GreenhouseGeisser correction). “Trend analyses” reported below reflect default tests of within-subject contrasts provided with RMANOVAs by SPSS, focusing on linear and quadratic contrasts. In addition, RM-ANOVA analyses were performed on the signal detection measure of d0 , which we computed from the recoded discrimination responses (i.e., stimulus ‘horizontal’ & response ‘horizontal’ ¼ hit; stimulus ‘horizontal’ & response ‘vertical’ ¼ miss; stimulus ‘vertical’ & response ‘horizontal’ ¼ false alarm; stimulus ‘vertical’ & response ‘vertical’ ¼ correct rejection). The outcomes of these analyses were identical to the results from the ANOVA analyses on raw accuracy scores as reported in the section below. Please refer to the Supplementary Material for an additional results
section on the analyses of the d0 measure and a graphical representation of d’ across conditions.
2.2.
Results
2.2.1.
Objective visual processing
On the 2AFC orientation discrimination task, the full-model RMANOVA with factors SOA, StimPos, and Int revealed a three-way interaction (F(5.3,53.4) ¼ 2.8; p ¼ .025), indicating that the effects of stimulus position and intensity were dependent on SOA. We thus performed follow-up RM-ANOVAs separately for each time window. Shown in Fig. 2 is average proportion of trials correct over participants, plotted over levels of stimulation intensity, separately for the LL and UR visual stimulus positions. Fig. 2C suggests that for the ‘classical’ masking window around þ90 msec there was a clear masking effect of TMS pulses, with increasingly disrupted visual task performance with increasing TMS intensity. This effect seems specific to
6
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 4 e Two examples of trials in Experiment 2. A left or rightward directed horizontal arrow stimulus was presented on each trial in either the lower left quadrant (experimental stimulus location) or in the upper right quadrant (control stimulus location) of the visual field. The stimulation intensity was fixed to 70% of maximal stimulator output. On each trial a single TMS pulse was delivered at one of five possible pre-stimulus time windows: ¡80, ¡60, ¡40, ¡20 or 0 msec. In Trial 1 (Tr 1), the TMS pulse precedes a leftward pointing arrow stimulus presented at the experimental visual field location. In Trial 2 (Tr 2), the TMS pulse precedes a rightward pointing arrow stimulus presented at the control visual field location.
stimuli on the LL. For UR, increasing TMS intensity appears to have no effect. Indeed, the RM-ANOVA showed a significant interaction StimPos Int (F(3.1,31.1) ¼ 3.1; p ¼ .042). Follow-up RM-ANOVA confirmed that Int has an effect for LL stimuli (F(2.9,29.3) ¼ 7.2; p ¼ .001), but not for UR stimuli (p ¼ .474). Moreover, trend analysis revealed a significant linear component for LL (F(1,10) ¼ 21.7; p ¼ .001), suggesting that increasing intensity disrupts performance linearly. Fig. 2D suggests that while there seems to be an effect of TMS at SOA þ120 msec at the highest intensities, specifically on LL, this is not as clear or pronounced as for the classical window. Statistically, there was no significant interaction between StimPos and Int (p ¼ .163), but follow-up analysis did yield a significant effect of Int on LL trials (F(2.5,24.7) ¼ 3.8; p ¼ .030), though not at all on UR trials (p ¼ .427). Interestingly, trend analysis results were in line with the observation that pulses at þ120 msec only successfully masked visual stimuli at the highest intensities. There was no linear (p ¼ .124) but only a quadratic trend (F(1,10) ¼ 18.6; p ¼ .002). The two pre-stimulus time windows paint a very dissimilar picture. For the 20 msec SOA, there appears to be a clear masking effect that is specific to LL (Fig. 1B). Indeed, RMANOVA revealed a significant interaction between StimPos and Int (F(2.4,23.6) ¼ 5.3; p ¼ .010). Follow-up tests showed a strong significant effect of Int for LL trials (F(2.3,22.6) ¼ 7.3; p ¼ .003), that moreover was significantly linear (F(1,10) ¼ 13.3; p ¼ .004). There was no effect of Int for UR (p ¼ .589). This suggests that TMS masking at this pre-stimulus time window constitutes ‘true’ masking, that is robust, strong, and retinotopically specific. Interestingly, the latter is not the case for the 50 msec window. Fig. 2A does show strong disruptions of visual task performance, as in 20 msec, but it does not seem to be specific to the targeted visual field location. Statistically, there was actually a significant interaction between StimPos and Int (F(3.1,31.1) ¼ 3.5; p ¼ .025), but follow-up analysis showed that while Int indeed had an effect on LL (F(2.8,27.7) ¼ 8.1; p ¼ .001), it
also had a strong effect on UR (F(2.7,26.8) ¼ 11.1; p < .001). This suggests that the effects of TMS in this time window, with these stimuli, are not specific.
2.2.2.
Subjective visual processing
For subjective visibility ratings (scale of 1e4), the full-model RM-ANOVA revealed a very strong 3-way interaction between SOA StimPos Int (F(4.2,42.3) ¼ 4.9; p ¼ .002). We thus proceeded, similarly to the analysis for the objective discrimination performance, to analyze separately per SOA. The patterns for these SOA conditions are shown in Fig. 3. Fig. 3C shows the visibility ratings averaged over participants, plotted over levels of TMS intensity separately for stimuli presented on the LL and UR with pulses applied at SOA þ90 msec. It appears that, analogously to the objective task performance results, there was a specific effect of Int for LL trials. Statistically, indeed there was an interaction between StimPos and Int (F(1.6,16.0) ¼ 11.4; p ¼ .001), and follow-up tests revealed a significant effect of Int for LL (F(1.5,15.3) ¼ 22.6; p < .001) that was linear (F(1,10) ¼ 28.5; p < .001, although now a quadratic trend was also supported; F(1,10) ¼ 5.6; p ¼ .040), and not for UR (p ¼ .676). Also in the other time windows, the effects of TMS on subjective visibility ratings followed closely the patterns of results on behavioral discrimination performance. For poststimulus SOA þ120 msec (Fig. 3D), the interaction between StimPos and Int was significant (F(3.2,32.4) ¼ 4.6; p ¼ .007) (even though this interaction did not reach significance for the objective measure, see above). Again, there was a significant effect of Int for LL (F(2.1,21.1) ¼ 5.0; p ¼ .015) that was linear (F(1,10) ¼ 9.3; p ¼ .012), but not for UR (p ¼ .795). A main reason to include the þ120 msec SOA in this study was to evaluate whether e in contrast to þ90 msec e there would be a dissociation between subjective and objective visual processing (Koivisto, Railo, & Salminen-Vaparanta, 2011). The data appear inconclusive on this point, since on the one hand there was a significant interaction between TMS intensity and
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
stimulus location for the subjective measure only, yet on the other hand the patterns of results for subjective and objective processing seemed highly similar. We therefore leave this question for future studies. For the pre-stimulus windows, there was a significant interaction StimPos Int for SOA 20 msec (F(2.0;20.4) ¼ 13.9; p < .001), and a significant effect of Int for LL (F(1.6,16.3) ¼ 15.2; p < .001) that was linear (F(1,10) ¼ 20.6; p ¼ .001) but no effect for UR (p ¼ .662). For visibility ratings in SOA 50 msec there was no interaction StimPos Int (p ¼ .208), but a main effect of Int (F(1.9,18.5) ¼ 18.1; p < .001). This again confirms a disruption of visual processing that is non-retinotopic.
3.
Experiment 2
3.1.
Materials and methods
3.1.1.
Participants
Ten participants volunteered for this study, 4 of which also participated in Experiment 1. No participants had abnormal vision or a history of neuropsychiatric disorders. One participant was later excluded from data-analysis as her data contained a disproportional amount of eye blinks. Participants were screened by a medical supervisor prior to participation, and supplied written informed consent. Experimental protocols were approved by the local medical-ethical committee.
3.1.2.
Stimuli, tasks, and design
Horizontal arrow stimuli pointing left or right and subtending 1.04 by .44 visual angle were presented at 4 eccentricity from a central fixation cross either in the LL quadrant of the visual field, or in the UR quadrant. Arrow stimuli consisted of multiple black arrow shapes on a white background (see Fig. 4). All stimuli were presented on a standard TFT monitor (Samsung SyncMaster; refresh rate 60 Hz) for a duration of 16.7 msec. The experiment consisted of two sessions; an experimental TMS session and a Sham TMS control session. The Sham session was included to control for possible baseline performance differences across the two visual field locations, and to control for non-specific TMS effects related to the clicking sound of the magnetic pulses. Each session started with the application of electrodes near participants' eyes for the purpose of recording electrooculographical (EoG) activity in the eye muscles. Ag/AgCl electrodes were prepared above and below participants' left eye to measure eye-movement related potential differences and to the mastoid behind their right ear, which served as a neutral reference. EoG data were recorded with Vision Recorder (BrainProducts, Gilching (Mu¨nich), Germany) software at a sampling rate of 500 Hz (for more details on EoG data acquisition, see Jacobs, Goebel, et al., 2012). Next, participants completed a practice run consisting of 48 trials to acquaint them with the forced-choice visual discrimination task. On each trial they were presented with an arrow stimulus and they had to subsequently indicate by button press 1) in which direction the arrow was pointing, and 2) to what extent they consciously perceived the arrow stimulus. Response buttons were 4 keys on a standard QWERTYkeyboard: the Z, the X, the dot and the slash keys. Response
7
coding for Response 1) was such that a button press of Z or X with the left hand corresponded to ‘left’ and a button press of any of the other two keys with the right hand corresponded to ‘right’. Response coding for Response 2) was such that the Zkey corresponded with complete awareness of the stimulus, the X-key with almost complete awareness, the dot-key with almost no unawareness and the slash-key with no awareness at all. During practice trial-by-trial feedback was given on discrimination performance by a color change of the fixation cross. After the practice run, the TMS coil was first positioned at the correct location over the right EVC (see TMS parameters). During the experimental phase of the session, the task was identical to the practice run (see Fig. 4), except for the now absent feedback. Data was acquired in blocks of 24 trials. Within blocks the location of the visual stimuli was fixed, but their location would alternate between LL and UR across blocks. Half of the participants started with a block of stimuli in LL visual field, whereas the other half would start with a block of stimuli in UR visual field. The first two blocks consisted of a no TMS baseline measurement. During the next 10 TMS blocks, single pulses of TMS were administered at varying SOAs. This resulted in a total number of 24 trials per TMS time window per visual field location. Trial order was randomized within blocks. Trials were separated by a jittered inter-trial-interval of ~5250 msec. Stimulus presentation and behavioral response recording was performed with the Presentation (NeuroBehavioral Systems Inc, Albany, CA) software package.
3.1.3.
TMS parameters
For the experimental EVC-TMS session, identical TMS equipment, pulse type and coil orientation were employed as in Experiment 1. Again, coil positioning was based on the phosphene localization procedure. Stimulation intensity was fixed to a maximum of 70% maximum stimulator output for each participant. In each trial a single TMS pulse was delivered to right EVC at any of 5 possible pSOAs: 80, 60, 40, 20 or 0 msec relative to the onset of the visual stimulus (see Fig. 4). In the Sham-TMS session, the TMS coil was replaced with the MC-P-B70 placebo coil, which elicits a clicking sound comparable to the MC-B70 coil, but in absence of the magnetic stimulation of the underlying brain tissue. Stimulator intensity was set to a value that matched the Sham TMS pulses to the real TMS pulses with regard to clicking sound volume. As in the EVC-TMS session, single TMS pulses at either 80, 60, 40, 20 or 0 msec SOAs were applied. Also in Experiment 2 we measured and corrected for the different presentation delays for the two stimulus locations, by means of the calibration procedure described earlier.
3.1.4.
Analysis
The EoG data was preprocessed in VisionAnalyzer (BrainProducts, Gilching (Mu¨nich), Germany) and exported to Matlab vR2012a (The MathWorks Inc, Natick, MA) for further custom analyses. Based on the EoG amplitude in the period running from 200 to þ100 msec relative to the visual stimulus trials were classified as either ‘blink’ or ‘no blink’ trials (see Jacobs, Goebel, et al., 2012; Skotte, Nojgaard, Jorgensen, Christensen, & Sjogaard, 2007, for further details). Blink trials were
8
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Fig. 5 e Results Experiment 2. Top panels (A) depict accuracy data, i.e., proportion of correct responses for the different stimulus-onset asynchronies (SOAs) between visual stimulus and TMS pulse for the EVC (left panel) and Sham (right panel) TMS session. Bottom panels (B) depict subjective visibility rating per SOA for EVC (C) and Sham (D) TMS session. Dotted lines represent the results for stimuli presented in the upper right (UR) visual field, solid lines represent the results for stimuli presented in lower left visual field (LL). Error bars represent standard error of mean.
excluded from the behavioral dataset and statistical analyses were all conducted on this reduced dataset. Using the SPSS v19 statistical software package (IBM, New York, NY), we performed a three-factorial repeated-measures ANOVA (RM-ANOVA) with factors TMS Type (2 levels), StimPos (2 levels) and SOA (6 levels), separately for the accuracy (hit rate: proportion correct) over participants, and subjective visibility ratings (1- to 4-point scale). Follow-up tests were RMANOVAs per level of TMS Type and pairwise comparisons as reported in Results (Greenhouse-Geisser correction in case of violation of sphericity assumptions). Here too, RM-ANOVA analyses were performed additionally on the signal detection measure of d0 , which we computed from the recoded discrimination responses (i.e., stimulus ‘left’ & response ‘left’ ¼ hit; stimulus ‘left’ & response ‘right’ ¼ miss; stimulus ‘right’ & response ‘left’ ¼ false alarm; stimulus ‘right’ & response ‘right’ ¼ correct rejection). The outcomes of these analyses were identical to the results from the ANOVA analyses on raw accuracy scores as reported in the section below. Please refer to the Supplementary Material for an additional
results section on the analyses of the d’ measure and a graphical representation of d’ across conditions.
3.2.
Results
We excluded one participant based on the disproportional amount of eye blinks (>40%) she displayed, which did not allow valid statistical analyses on the proportion of remaining trials. The other nine participants blinked within the crucial temporal window on 4.1% of trials on average. After removal of all trials containing eye blinks and after exclusion of one participant's data, we conducted a three-way RM-ANOVA on the remaining data.
3.2.1.
Objective visual processing
The three-way ANOVA analysis revealed significant main effects of the factors TMS Type (F(1, 9) ¼ 10.4; p ¼ .012) and SOA (F(1.6, 13.1) ¼ 7.1; p ¼ .011), and a significant interaction effect between these two factors (F(1.8, 14.7) ¼ 7.6; p ¼ .006). No main or interaction effects for the factor StimPos appeared (all p > .5).
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Based on these outcomes, we pooled the data for the factor StimPos and performed two one-way RM-ANOVAs on the factor SOA: one for the EVC-TMS data and one for the ShamTMS data. The EVC-TMS data revealed a main effect of SOA (F(1.7, 13.2) ¼ 7.8; p ¼ .008). Post-hoc tests revealed significant differences between the No TMS baseline condition (mean proportion correct ¼ .98) and the 80 msec (mean ¼ .79; p ¼ .008) and 60 msec (mean ¼ .79; p ¼ .006) TMS time windows (see Fig. 5A). The Sham-TMS data did not show an SOA effect. Thus, pre-stimulus TMS applied 80 or 60 msec prior to visual stimulation impairs discrimination performance in a nonretinotopic fashion.
3.2.2.
Subjective visual processing
Results of the three-way ANOVA on the subjective visibility ratings were similar to those of the accuracies. Again, main effects of both TMS Type (F(1, 8) ¼ 26.7; p ¼ .001) and SOA (F(1.5, 12.1) ¼ 7.1; p ¼ .013) were revealed as well as a Type SOA (F(1.6, 12.7) ¼ 10.8; p ¼ .003) interaction effect. Because this analysis showed that visual field location did not affect subjective visibility, we again pooled the factor StimPos, and conducted a one-way RM-ANOVA on the factor SOA for each level of TMS Type. The SOA of real EVC-TMS, but not of Sham-TMS, significantly affected visibility ratings (F(1.5, 11.9) ¼ 9.6; p ¼ .005). Posthoc analyses showed a broader effective TMS period for the visibility ratings than for accuracy. EVC-TMS significantly reduced subjective visibility when applied at 80 msec (mean ¼ 2.2 on a 1- to 4-point scale; p ¼ .001), 60 msec (mean ¼ 2.1; p ¼ .001) and 40 msec (mean ¼ 2.4; p ¼ .011) compared to No TMS baseline (mean ¼ 3.2; see Fig. 5B). Summarizing, both visual discrimination ability and subjective visual awareness are affected by pre-stimulus EVCTMS, but not by Sham TMS. This effect is time-specific and ranges from 80 msec to 60 msec for discrimination performance, and broadened into the 40 msec time window for visibility ratings. In contrast to our findings from Experiment 1, no significance decrease in any of the two measures of visual processing occurred when TMS was applied at the 20 msec time point. None of the detected effects proved to be influenced by the position in the visual field at which the visual stimuli were presented.
4.
Discussion
The current study consists of two experiments in which we explored whether ongoing neural activity (brain state) in EVC modulates the perception of future visual information. To this end, we applied chronometric focal TMS to EVC at various SOAs before and after visual stimulus onset. Moreover, we used different stimuli and visual field locations. Experiment 1 revealed non-retinotopic TMS-induced perceptual impairment at a pSOA of 50 msec. Yet, a TMS pulse administered to EVC at 20 msec selectively disrupted visual perception at the corresponding topographic location in the visual field. These results match a study by Corthout et al. (2003) who described in three subjects a pattern of current-direction (CD)-dependent visual suppression at 20 msec and CD-independent
9
visual suppression at 50 msec, and Corthout et al. (1999) showing in three subjects a coil location dependent suppression effect at 10 msec but coil location independent effect at 60, 50 msec. In the current study, TMS over EVC also resulted in topographically impaired visual perception at two post-stimulus timings (90 and 120 msec), an effect which has previously been reported and attributed to direct interference with stimulus-related neural activity in EVC (e.g., Amassian, et al., 1989; Corthout, et al., 1999; Sack, et al., 2009). In contrast to Experiment 1, the results of Experiment 2 showed a single broad window, centered around 60 msec, in which pre-stimulus activity in EVC influenced objective and subjective visual perception in a non-topographic way. This masking effect was obtained after removal of trials with eye blinks, demonstrating a neural origin. Yet, while this window thus converged with the 50 msec TMS effect observed in Experiment 1, disruption of EVC activity 20 msec before stimulus onset did not affect visual perception at all. And indeed no corresponding topographically specific pre-stimulus effect was found at any other pSOAs either. Taken together, these two experiments thus revealed a clear dissociation between two pre-stimulus mechanisms: one pSOA (20 msec) which is retinotopically specific, and one pSOA (~50 msec) which is retinotopically unspecific. It remains an empirical question why the late retinotopic TMS effect was absent in Experiment 2, since the experiments differed in more than one-way. We had no clear a priori hypothesis expecting one type of stimulus to generate qualitatively different effects as compared to the other stimulus type. Based on our current data, it is difficult to infer where the stimulus-related difference in results comes from, since the stimuli differ across several dimensions, from low-level properties like contrast and size, to timing (stimulus duration), to high-level properties like the presence or absence of symbolic meaning in the stimulus. Previous work has started to compare TMS masking effects for different types of stimuli (de Graaf, Goebel, & Sack, 2012; Koivisto et al., 2011) for conventional post-stimulus SOAs. So far, to our knowledge no studies reported systematic and significant stimulus-driven differences in results for pre-stimulus SOAs (though see Kammer, Puls, Strasburger, Hill, & Wichmann, 2005, concerning luminance). It would be useful to conduct systematic investigations in future work. To guide such future work, we would briefly point out two possible causes for the potential stimulus-dependence of the 20 msec masking effect obtained here. On the one hand, since the arrow stimuli were larger and higher-contrast, they could have been more robust and less susceptible to the 20 msec TMS masking effect. But since there was not even a hint of masking for the arrow stimuli, the difference in results across experiments could alternatively signify something more fundamental. For example, compared to the simple orientation stimuli of Experiment 1, the symbolic arrows of Experiment 2 were more complex, higher in informational/symbolic content, and therefore possibly processed (at least in part) in cortical regions hierarchically further up the visual stream. Orientation information is actually predominantly processed in EVC, the region directly targeted by TMS here. That could account for higher susceptibility of the used bar stimuli to disruptions of local cortical processing.
10
4.1.
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Local pre-stimulus processing at 20 msec
At least for orientation stimuli in Experiment 1, ongoing EVC processing occurring some 20 msec prior to visual stimulation is crucial to unimpaired visual perception. This appears to be relatively local processing, because the TMS effect at 20 msec is retinotopic. Retinotopy is most prominent in LL visual cortices EVC, simple orientation processing is focused in these regions, and EVC was the site directly targeted by TMS in this study. It thus seems plausible that this particular pre-stimulus suppression effect originated from the cortical area right underneath the coil, although we cannot completely rule out spreading of the TMS to higher-level visual areas within the same hemisphere which contain cruder retinotopic maps. The nature of such local processing, and the circuitry involved, remain an open question. A question made all the more interesting because the effects of a single TMS pulse need to last quite some time, when taking into account retino-striate transmission time on top of the pSOA. From paired-pulse studies in the motor system, we know that single TMS pulses can have inhibitory and facilitative effects (Chen, 2004) outlasting the duration of the pulse itself. Suprathreshold TMS causes a decrease in motor-evoked potentials (MEPs) after a long inter-pulse interval of 50e200 msec: an effect presumably modulated by intracortical GABAergic activity on slower GABAb receptors (McDonnell, Orekhov, & Ziemann, 2006). If similar cortical circuitry holds in early visual regions, the suprathreshold TMS pulses we applied here might have recruited intracortical inhibitory networks within occipital cortex, which reduce cortical excitability with a delay of at least 50 msec, but potentially up to 200 msec. In line with this, Moliadze, Zhao, Eysel, and Funke (2003) found that strong single TMS pulses to occipital cortex can have inhibitory effects on single visual cortex cells for some 100e200 msec. It will be interesting for future work to evaluate to what extent such mechanisms apply to occipital processing in the context of pre-stimulus TMS masking.
over vertex). It thus seems that the 50 msec suppression effect involves a time-specific, relatively global (because non-retinotopic and stimulus-independent), but nevertheless neural mechanism. One physiological explanation within these constraints would be that the visual cortex is continuously alternating between states of high and low excitability and that a TMS pulse affects this alternation. For example, recent evidence shows that both the power and phase of prestimulus oscillations in the alpha frequency (i.e., ~8e12 Hz) range have been associated with stimulus visibility (Busch, et al., 2009; van Dijk, et al., 2008; Dugue, et al., 2011; Hanslmayr, et al., 2007; Mathewson, et al., 2009; Romei, et al., 2008; Toscani, et al., 2010). In this context, a TMS pulse might phase-lock such oscillations, inducing a period of decreased excitability at a fixed later stage. Previous research suggests at least that a set of several rhythmic pulses can have such phase-locking effects (Thut et al., 2011), which can be attentionally/ perceptually relevant (Romei et al., 2010). One alternative explanation might be a spreading of the physical TMS stimulation from one hemisphere to the other through transcallosal connections, leading to visual suppression in both visual hemifields. This seems unlikely, since in Experiment 1 we showed a very similar pattern of TMS masking for both visual field locations across intensities. A ‘spreading explanation’ of the induced electric activity would presumably be reflected by a weaker masking effect for the suboptimal visual field location. Relatedly, it is also an empirical question whether TMS pulses over EVC could spread to higher visual or attentional areas (e.g., parietal regions). An indirect effect on such functionally relevant higher cortical areas, whether directly or through backpropagation of the pulse energy to a wider area of early visual cortices, would also be in line with the neural yet non-retinotopic and stimulus-independent suppression effect. For the moment, however, this seems rather speculative.
4.3. 4.2.
Conclusion
Non-local pre-stimulus processing at 50 msec
The even earlier pre-stimulus suppression effect must in any case involve a different mechanism, since it is not retinotopically specific and seems independent of stimulus parameters. The temporal specificity of the observed suppression effect moreover implies that visibility depends not just on a general on/off state in a broad time period, but rather on the state of the visual cortex at a particular time point prior to the arrival of the visual input. Previous studies have ascribed pre-stimulus TMS effects to non-specific TMSinduced eye blinks (Beckers & Homberg, 1991; Corthout, et al., 2000). However, when taking together this study with our previous work, we have aimed to control for all possible non-neural effects of TMS pulses. Here and previously (Jacobs, Goebel, et al., 2012), blinking could not fully account for the pre-stimulus TMS effect around 50 msec. In Jacobs, Goebel, et al., 2012; Jacobs, de Graaf, et al., 2012, we also controlled for auditory effects (Sham TMS over occipital cortex) and somatosensory stimulation (real TMS
We here provided evidence for two distinct occipital mechanisms involved in biasing successful processing of upcoming visual inputs. Since these mechanisms must both involve functionally relevant neuronal processing, future studies can more directly assess their cortical loci, their computational dynamics, and potentially their functional interplay.
Acknowledgments The work presented here was supported by European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement n [263472]) awarded to A.T.S. T.A.G. is supported by the Netherlands Organization for Scientific Research (NWO 45113-024). The authors would like to thank Azadeh Seyed Tafreshiha for assistance with Experiment 1.
c o r t e x 5 9 ( 2 0 1 4 ) 1 e1 1
Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.cortex.2014.06.017.
references
Amassian, V. E., Cracco, R. Q., Maccabee, P. J., Cracco, J. B., Rudell, A., & Eberle, L. (1989). Suppression of visual perception by magnetic coil stimulation of human occipital cortex. Electroencephalography and Clinical Neurophysiology, 74(6), 458e462. Arieli, A., Sterkin, A., Grinvald, A., & Aertsen, A. (1996). Dynamics of ongoing activity: explanation of the large variability in evoked cortical responses. Science, 273(5283), 1868e1871. Beckers, G., & Homberg, V. (1991). Impairment of visual perception and visual short term memory scanning by transcranial magnetic stimulation of occipital cortex. Experimental Brain Research, 87(2), 421e432. Busch, N. A., Dubois, J., & VanRullen, R. (2009). The phase of ongoing EEG oscillations predicts visual perception. Journal of Neuroscience, 29(24), 7869e7876. Cardin, V., Friston, K. J., & Zeki, S. (2011). Top-down modulations in the visual form pathway revealed with dynamic causal modeling. Cerebral Cortex, 21(3), 550e562. Chen, R. (2004). Interactions between inhibitory and excitatory circuits in the human motor cortex. Experimental Brain Research, 154(1), 1e10. Corthout, E., Hallett, M., & Cowey, A. (2003). Interference with vision by TMS over the occipital pole: a fourth period. NeuroReport, 14(4), 651e655. Corthout, E., Uttl, B., Juan, C. H., Hallett, M., & Cowey, A. (2000). Suppression of vision by transcranial magnetic stimulation: a third mechanism. NeuroReport, 11(11), 2345e2349. Corthout, E., Uttl, B., Ziemann, U., Cowey, A., & Hallett, M. (1999). Two periods of processing in the (circum)striate visual cortex as revealed by transcranial magnetic stimulation. Neuropsychologia, 37(2), 137e145. van Dijk, H., Schoffelen, J. M., Oostenveld, R., & Jensen, O. (2008). Prestimulus oscillatory activity in the alpha band predicts visual discrimination ability. Journal of Neuroscience, 28(8), 1816e1823. Downing, P. E., Chan, A. W., Peelen, M. V., Dodds, C. M., & Kanwisher, N. (2006). Domain specificity in visual cortex. Cerebral Cortex, 16(10), 1453e1461. Dugue, L., Marque, P., & VanRullen, R. (2011). The phase of ongoing oscillations mediates the causal relation between brain excitation and visual perception. Journal of Neuroscience, 31(33), 11889e11893. de Graaf, T. A., Cornelsen, S., Jacobs, C., & Sack, A. T. (2011). TMS effects on subjective and objective measures of vision: stimulation intensity and pre- versus post-stimulus masking. Consciousness and Cognition, 20, 1244e1255. de Graaf, T., Goebel, R., & Sack, A. T. (2012). Feedforward and quick recurrent processes in early visual cortex revealed by TMS? NeuroImage, 61(3), 651e659. de Graaf, T. A., Gross, J., Paterson, G., Rusch, T., Sack, A. T., & Thut, G. (2013). Alpha-band rhythms in visual task performance: phase-locking by rhythmic sensory stimulation. PLoS One, 8(3), e60035. de Graaf, T. A., Herring, J., & Sack, A. T. (2011). A chronometric exploration of high-resolution ‘sensitive TMS masking’ effects on subjective and objective measures of vision. Experimental Brain Research, 209(1), 19e27.
11
Hanslmayr, S., Aslan, A., Staudigl, T., Klimesch, W., Herrmann, C. S., & Bauml, K. H. (2007). Prestimulus oscillations predict visual perception performance between and within subjects. NeuroImage, 37(4), 1465e1473. Hesselmann, G., Kell, C. A., Eger, E., & Kleinschmidt, A. (2008). Spontaneous local variations in ongoing neural activity bias perceptual decisions. Proceedings of the National Academy Sciences of the United States of America, 105(31), 10984e10989. Jacobs, C., Goebel, R., & Sack, A. T. (2012). Visual awareness suppression by pre-stimulus brain stimulation; a neural effect. NeuroImage, 59(1), 616e624. Jacobs, C., de Graaf, T. A., Goebel, R., & Sack, A. T. (2012). The temporal dynamics of early visual cortex involvement in behavioral priming. PLoS One, 7(11), e48808. Kammer, T. (2007). Masking visual stimuli by transcranial magnetic stimulation. Psychological Research, 71(6), 659e666. Kammer, T., Puls, K., Strasburger, H., Hill, N. J., & Wichmann, F. A. (2005). Transcranial magnetic stimulation in the visual system. I. The psychophysics of visual suppression. Experimental Brain Research, 160(1), 118e128. Koivisto, M., Railo, H., & Salminen-Vaparanta, N. (2011). Transcranial magnetic stimulation of early visual cortex interferes with subjective visual awareness and objective forced-choice performance. Consciousness and Cognition, 20(2), 288e298. Laycock, R., Crewther, D. P., Fitzgerald, P. B., & Crewther, S. G. (2007). Evidence for fast signals and later processing in human V1/V2 and V5/MTþ: a TMS study of motion perception. Journal of Neurophysiology, 98(3), 1253e1262. Mathewson, K. E., Gratton, G., Fabiani, M., Beck, D. M., & Ro, T. (2009). To see or not to see: prestimulus alpha phase predicts visual awareness. Journal of Neuroscience, 29(9), 2725e2732. Mathewson, K. E., Prudhomme, C., Fabiani, M., Beck, D. M., Lleras, A., & Gratton, G. (2012). Making waves in the stream of consciousness: entraining oscillations in EEG alpha and fluctuations in visual awareness with rhythmic visual stimulation. Journal of Cognitive Neuroscience, 24(12), 2321e2333. McDonnell, M. N., Orekhov, Y., & Ziemann, U. (2006). The role of GABA(B) receptors in intracortical inhibition in the human motor cortex. Experimental Brain Research, 173(1), 86e93. Moliadze, V., Zhao, Y., Eysel, U., & Funke, K. (2003). Effect of transcranial magnetic stimulation on single-unit activity in the cat primary visual cortex. Journal of Physiology, 553(Pt 2), 665e679. Nandy, A. S., Sharpee, T. O., Reynolds, J. H., & Mitchell, J. F. (2013). The fine structure of shape tuning in area V4. Neuron, 78(6), 1102e1115. Romei, V., Brodbeck, V., Michel, C., Amedi, A., Pascual-Leone, A., & Thut, G. (2008). Spontaneous fluctuations in posterior alphaband EEG activity reflect variability in excitability of human visual areas. Cerebral Cortex, 18(9), 2010e2018. Romei, V., Gross, J., & Thut, G. (2010). On the role of prestimulus alpha rhythms over occipito-parietal areas in visual input regulation: correlation or causation? Journal of Neuroscience, 30(25), 8692e8697. Sack, A. T., van der Mark, S., Schuhmann, T., Schwarzbach, J., & Goebel, R. (2009). Symbolic action priming relies on intact neural transmission along the retino-geniculo-striate pathway. NeuroImage, 44(1), 284e293. Skotte, J. H., Nojgaard, J. K., Jorgensen, L. V., Christensen, K. B., & Sjogaard, G. (2007). Eye blink frequency during different computer tasks quantified by electrooculography. European Journal of Applied Physiology, 99(2), 113e119. Thut, G., Veniero, D., Romei, V., Miniussi, C., Schyns, P., & Gross, J. (2011). Rhythmic TMS causes local entrainment of natural oscillatory signatures. Current Biology, 21(14), 1176e1185. Toscani, M., Marzi, T., Righi, S., Viggiano, M. P., & Baldassi, S. (2010). Alpha waves: a neural signature of visual suppression. Experimental Brain Research, 207(3e4), 213e219.