Chromatic information is processed by the visual system both at an unconscious level and at a level that results in conscious perception of color. It remains unclear whether both conscious and unconscious processing of chromatic information depend on activity in the early visual cortex or whether unconscious chromatic processing can also rely on other neural mechanisms. In this study, the contribution of early visual cortex activity to conscious and unconscious chromatic processing was studied using single-pulse TMS in three time windows 40–100 msec after stimulus onset in three conditions: conscious color recognition, forced-choice discrimination of consciously invisible color, and unconscious color priming. We found that conscious perception and both measures of unconscious processing of chromatic information depended on activity in early visual cortex 70–100 msec after stimulus presentation. Unconscious forced-choice discrimination was above chance only when participants reported perceiving some stimulus features (but not color).
Visual stimuli are processed by the brain and may influence behavior even if they are not consciously perceived (Weiskrantz, 1997). A classical example of this is blindsight, where patients who are clinically blind in certain visual field locations because of damaged visual cortex nevertheless have a capacity to discriminate features of visual stimuli they do not consciously perceive (Weiskrantz, Watrrington, Sanders, & Marshall, 1974; Pöppel, Held, & Frost, 1973; for a review, see Cowey, 2010). The neural mechanisms that underlie the conscious and unconscious modes of processing are not, however, well understood. Roughly speaking, they could rely on different anatomical and physiological systems, but they might also depend completely or partly on overlapping neural mechanisms. In this study, we directly tested the contrasting accounts whether activity in the early visual areas (V1/V2) is (Breitmeyer, Ro, & Singhal, 2004) or is not (Boyer, Harrison, & Ro, 2005) necessary for unconscious processing of chromatic information.
Conscious visual perception of color is based on the processing of chromatic information by the visual system, most importantly, the discrimination of different wavelengths (Gegenfurtner & Kiper, 2003). However, processing of chromatic information does not necessarily result in conscious perception of color. For instance, patients suffering from achromatopsia (Heywood & Cowey, 1999) or visual field defects caused by lesions to early visual areas (Stoerig & Cowey, 1989) may behaviorally respond to chromatic information in the absence of conscious color vision. Remarkably, in some of the patients with damaged V1, the spectral sensitivity curve of the blind visual field locations is similar to the sensitivity curve found in healthy controls (Stoerig & Cowey, 1989).
One of the key questions in explaining unconscious chromatic processing concerns the pathways that relay chromatic information to cortex. Conscious color vision is mainly mediated by the LGN of thalamus and by ventral visual cortical areas (Claeys et al., 2004; Heywood & Cowey, 1999; Zeki et al., 1991; Meadows, 1974). In blindsight patients with V1 lesions, the geniculate pathways that bypass V1 and project to extrastriate visual areas are possible mediators of unconscious chromatic processing, but unconscious chromatic processing has also been proposed to rely on extrageniculate inputs to cortex (Weiskrantz, 1997; Cowey & Storeig, 1989; Stoerig & Cowey, 1989; Stoerig, 1987). The superior colliculus (SC) has been found to contribute to the residual visual abilities of blindsight patients (e.g., Danckert & Rossetti, 2005), even for chromatic stimuli (Tamietto et al., 2009; Leh, Mullen, & Ptito, 2006). There is neural evidence from studies on monkeys that the SC responds to chromatic information, but the delayed onset latency of these color signals suggests a cortical rather than retinal origin (White, Boehnke, Marino, Itti, & Munoz, 2009). There is also some evidence that chromatic information is relayed to cortex via the pulvinar in monkeys (Felsten, Benevento, & Burman, 1983).
In addition to neuroanatomy, conscious and unconscious processing may differ with respect to their functional characteristics. Prominent accounts assert that stimulus-driven feedforward activity spreading to higher cortical areas is sufficient for unconscious processing of visual features, but that conscious visual perception of stimulus features requires recurrent feedback activity between cortical areas (Lamme, 2010; Dehaene, Changeux, Naccache, Sackur, & Sergent, 2006). In healthy participants, TMS can be employed to temporarily disturb cortical processing (Walsh & Pascual-Leone, 2003). Conscious perception of visual stimuli can be partially or completely suppressed (e.g., Koivisto, Railo, & Salminen-Vaparanta, 2011; Amassian et al., 1989), and color recognition threshold was elevated (Paulus, Korinth, Wischer, & Teragu, 1999) by TMS of early visual areas 60–100 msec after the visual stimulus onset. The visibility of a stimulus can also be suppressed by a subsequent spatially nonoverlapping visual mask presented some tens of milliseconds after the first stimulus. This method called “metacontrast masking” is believed to spare early feedforward signals but inhibit later consciousness-related signals (Breitmeyer & Öğmen, 2006; Lamme & Roelfsema, 2000; Macknik & Livingstone, 1998). Both TMS (Boyer et al., 2005) and metacontrast masking (Breitmeyer, Ro, et al., 2004) have been applied to study the functional neuroanatomy of conscious and unconscious chromatic processing, but the methods have yielded contrasting results.
Colored stimuli that are rendered completely invisible by a subsequent metacontrast mask produce unconscious priming when participants respond to the color of the mask (Schmidt, 2000). Breitmeyer, Ro, et al. (2004) reported that, in unconscious color priming, the functioning of the prime was best characterized by wavelength, not by the subjectively experienced prime color. Thus, the results are consistent with the view that later processing stages in the ventral visual areas are responsible for conscious perception of color, but unconscious color priming depends on an early, stimulus-related response of neurons located in the early visual cortex (see also Breitmeyer, Öğmen, & Chen, 2004).
In contrast, Boyer et al. (2005) showed that color discrimination remained above chance although conscious visibility of stimulus color was suppressed by occipital TMS, applied between 86 and 114 msec after stimulus onset. They concluded that unconscious processing of chromatic information is possible without V1. However, as noted by the authors, it is possible that, in their study, the TMS pulses did not eliminate the initial stimulus-driven cortical activity (Boyer et al., 2005), which reaches early visual cortical areas approximately 60 msec following stimulus onset (Vanni, Tanskanen, Seppä, Uutela, & Hari, 2001; Wilson, Babb, Halgren, & Crandall, 1983). Furthermore, in the study by Boyer et al. (2005) the participants merely indicated with a dichotomous scale whether they were conscious of the stimulus color. Overgaard (2011) has proposed that at least some reports of unconscious perception could be methodological artifacts caused by the use of dichotomous measures of conscious vision although visual experience would often be best characterized by a gradual scale.
The aim of this study was to clarify the contrasting findings on whether V1/V2 is crucial (Breitmeyer, Ro, et al., 2004) or not (Boyer et al., 2005) for unconscious processing of chromatic information. We used behavioral measures of conscious color recognition (subjective perception of color), unconscious color recognition (forced-choice guessing of stimulus color when it was reported not consciously perceived), and unconscious priming by invisible (masked) chromatic information. In each condition, TMS was applied to interfere with visual processing in early visual areas 40, 70, or 100 msec after stimulus onset. If unconscious processing of chromatic information depends on activity in the early visual areas and is not mediated by projections that bypass them, TMS should reduce the magnitude of priming and render forced-choice guessing to chance level.
Thirteen neurologically healthy, right-handed participants (age = 21–32 years, seven women) took part in the experiment (including author H.R.). All had normal or corrected-to-normal visual acuity. A written informed consent was obtained before the study. The study was conducted in accordance with the declaration of Helsinki, and it was approved by the ethics committees of University of Turku (TMS) and the Hospital District of Helsinki and Uusimaa (fMRI). The participants received study credits or a small fee for participating.
Visual stimuli were presented on a white background (54 cd/m2) on a 19 in. CRT monitor set to 75 Hz (13.3 msec/frame). Viewing distance was 150 cm, and head position was controlled with a chin rest. There were two kinds of stimuli, a round disk and an annulus, each presented 2.1° from the fixation in either the bottom left or top right quadrant of the screen (see Figure 1). In each trial, the disk and the annulus were always presented in the same quadrant. Each stimulus was either blue or red, with equal luminance (9 cd/m2). The disk, which was 0.3° in diameter, was presented for one screen refresh, and it served as a prime stimulus in the color priming condition. The annulus (inner diameter = 0.3°, outer diameter = 0.5°), functioning as a metacontrast mask, was presented 53.3 msec after the onset of the prime for 186.6 msec. These masking parameters are known to strongly suppress the conscious visibility of the disk, while the mask annulus remains clearly visible (e.g., Schmidt, 2000). Visual stimuli were presented using E-Prime software (Psychology Software Tools, Sharpsburg, PA).
The participants took part in three experimental conditions: color priming, masked prime visibility, and conscious color recognition. All conditions were performed with and without occipital TMS. Visual stimulation (Figure 1) was identical in color priming and masked prime visibility conditions, but the latter served as a control condition to investigate whether conscious perception of the prime stimulus (disk) was successfully suppressed by the mask. The conscious color recognition condition was employed to study how TMS affects the conscious and unconscious processing of an unmasked disk stimulus.
In the color priming condition, the participants were asked to report the color of the annulus (the mask) as fast and accurately as possible. They pressed a button on a gamepad with their right index finger when they saw a blue annulus and with their left index finger when they saw a red annulus. The priming condition (without TMS) included 72 trials divided into three blocks. In each block, the disk and the annulus were presented in same color (congruent; 50% of trials) or in different color (incongruent; 50% of trials) randomly in either the left or right quadrant. Because metacontrast masking is known to be strongest when the masked stimulus remains unattended (Ramachandran & Cobb, 1995) and because we wanted to minimize the visibility of the prime, the participants were not informed about the presentation of the prime stimulus. None of the participants spontaneously reported noticing the prime disk during the color priming condition, and they were later surprised to find out that it was presented.
To examine how well the annulus masked the visibility of the prime disk, the experiment included a control condition where the participants were asked to report the color of the prime disk as accurately as possible without time pressure (masked prime visibility condition). In addition to reporting the color of the prime (i.e., the disk), the participants were asked to rate how well they perceived it by pressing a corresponding key on the gamepad with their right thumb. There were four alternatives: 0 = I did not see the disk at all, 1 = I saw the disk faintly, but I did not see its color, 2 = I saw the disk faintly, and I also believe I saw its color, and 3 = I saw the disk and its color clearly. It was stressed that the two lowest alternatives should only be used when the participant did not see the color of the disk. This control condition was physically identical with the priming condition but included (without TMS) 96 trials, which were completed in two blocks of trials. Each block included equal numbers of congruent and incongruent trials presented randomly in the left or the right quadrant. This control condition was always completed at the end of the experiment.
The experiment also included a conscious color recognition condition, where the prime disk was presented alone without the masking annulus. In this condition, the participants' task was to report the color of the disk as accurately as possible (unspeeded forced-choice response) and to rate the visibility of the disk using the same four alternatives as in the masked prime visibility condition. Note that objective measures such as color recognition performance might also reflect unconscious processing and gradual measures of visibility force participants to adopt a stricter criterion for reporting conscious contents than dichotomous measures (Overgaard, 2011; Overgaard, Rote, Mouridsen, & Ramsøy, 2006). The conscious color recognition condition (without TMS) consisted three blocks of 24 stimuli. The target disk was presented in either the left (50%) or the right (50%) quadrant in random order. The conscious color recognition condition was always completed before the masked prime visibility condition, but the order of the priming and the conscious color recognition conditions was counterbalanced.
In addition to the behavioral baseline (no TMS) conditions (see Behavioral Procedure), the three conditions were all performed with TMS. The color priming condition was divided into four blocks of 48 trials, the masked prime visibility condition was divided into two blocks of 48 trials, and the conscious color recognition condition (without metacontrast masking) was divided into three blocks of 48 trials. The TMS blocks were identical with the behavioral baseline blocks with the exception that a single TMS pulse was applied to early visual areas 40, 70, or 100 msec after the onset of the disk stimulus (SOA) on each trial (see Figure 1). The aim was to disturb the activity in V1/V2 associated with the processing of the disk while minimizing the interference of mask processing (see Ro, Breitmeyer, Burton, Singhal, & Lane, 2003). The 40, 70, and 100 msec disk-TMS SOAs correspond to mask-TMS SOAs of −13, 17, or 47 msec, which, together with the relatively long presentation time of the mask stimulus, should minimize the TMS suppression of the mask. During the experiment, the participants performed behavioral and TMS blocks in turns, with each condition performed consecutively. Before each condition, the participants were familiarized to the task by completing a practice block. The participants took small breaks between the blocks every now and then. Each participant was tested in one experimental session that lasted approximately for 2 hr.
A Nexstim eXimia (Helsinki, Finland) stimulator fitted with a biphasic 70-mm figure-of-eight coil was used for TMS. The coil was fixed on a tripod during stimulation so that current direction was from lateral to medial during second pulse phase. The intensity of TMS was 65% of the stimulator's maximal output during TMS conditions.
Our aim was to stimulate the V1/V2 of right hemisphere with TMS. Hence, we aimed to interfere with the processing of the contralateral, left hemifield target using TMS while the ipsilateral, right hemifield targets served as controls for possible nonspecific effects of TMS. The stimulation site was determined either by using fMRI or by using a phosphene mapping method. Six participants whose visual areas had earlier been mapped with a multifocal fMRI procedure (Henriksson, Karvonen, Salminen-Vaparanta, Railo, & Vanni, submitted) agreed to participate in the current TMS study. Although we anticipated that a sample of six would probably lack the statistical power to reveal many of the effects we were interested in, we decided to include these participants to the current study because they would allow a precise stimulation of V1/V2. In these participants, the TMS pulses were directed to the retinotopic region of V1 corresponding to the location where the left hemifield stimulus was presented during the experiment (see Figure 2A and B). The relationship between the brain and the TMS coil was continuously registered using an MRI-guided eXimia Navigated Brain Stimulation system (Ruohonen & Karhu, 2010), which estimates the TMS-induced electric field by using a spherical conductor model (Ilmoniemi, Ruohonen, & Karhu, 1999; Hämäläinen & Sarvas, 1989).
For the remaining seven participants, the stimulation site was mapped using a widely used phosphene method in the beginning of the session (e.g., Sack, van der Mark, Schuhmann, Schwarzbach, & Goebel, 2009; Silvanto, Lavie, & Walsh, 2005). These participants wore EEG caps (electrodes removed), which had a coordinate system drawn on the occipital part of the cap. This coordinate system allowed us to monitor the position of the coil. Pulses were first aimed 2 cm above inion, after which, the coil was systematically moved to right and left in the occipital sites. TMS intensity was 65–70% of maximal output during the phosphene mapping. Throughout the phosphene mapping, four black rectangles (numbered 1–4) were presented on the computer screen, each in one quadrant 2.1° from the fixation. The bottom left rectangle thus corresponded to the location we aimed to stimulate in the main experiment. The participants were asked to close their eyes, imagine that they are looking at the center of the screen, remember the locations of the four rectangles, and report if they noticed phosphenes and where they perceived them. Only the participants who perceived phosphenes contralaterally to the stimulation site after left and right hemisphere stimulation were allowed to take part in the experiment. The location over which TMS produced phosphenes that overlapped with the bottom left rectangle was chosen as the stimulation site (on average 3.6 cm above and 2.5 cm to the right from the inion in these participants). A comparison of the results of the MRI and phosphene guided groups is reported in Modeling of the TMS-induced Electric Field in the fMRI Group.
MRI Imaging and Analysis
The fMRI measurements of the participants (n = 6) whose visual cortical areas were mapped were carried out with a 3-T MRI scanner (Signa HDxt, General Electric, Inc., Milwaukee, WI) with a phased array, eight-channel head coil. Visual stimuli were presented with a three-micromirror data projector (Christie X3, Christie Digital Systems Ltd., Cypress, CA) using Presentation software (Neurobehavioral Systems, Inc., Albany, CA). The major imaging parameters were repetition time of 1.8 sec, echo time of 30 msec, flip angle of 60°, field of view of 20 cm, matrix of 64 × 64, and slice thickness of 3 mm. Twenty-nine (29) slices were acquired in interleaved order. The retinotopic visual areas were mapped using a modified version (Henriksson et al., submitted) of the multifocal stimulus method described by Vanni, Henriksson, and James (2005). Four runs, 4 min each, comprised 32 miniblocks of 7.3-sec duration, and during each miniblock, a subset of 24 regions were stimulated. For each participant, a high-resolution T1-weighted anatomical image (voxel size = 1 mm × 1 mm × 1 mm) was acquired with 3-D SPGR BRAVO sequence (ASSET calibration). Standard preprocessing with slice-time and motion correction was followed by estimation of general linear model with SPM8 Matlab toolbox. Functional areas were determined from 3-D images using functional and anatomical landmarks with SPM8 and Freesurfer softwares (Figure 2A and B).
Color priming was analyzed using a three-way repeated measures ANOVA: SOA (4: no TMS, 40 msec, 70 msec, 100 msec) × Congruency (2: congruent, incongruent) × Hemifield (2: stimulated or control hemifield). The dependent variables were median RTs (calculated from correct responses) and mean color recognition accuracies. The masked prime visibility condition and the conscious color recognition condition were analyzed by a SOA (4) × Hemifield (2) ANOVA. Color discrimination accuracies and visibility ratings were the dependent variables. Greenhouse–Geisser-corrected p values were used whenever the sphericity assumption was violated (Mauchly's test, p < .05). Multiple comparisons were performed following Fisher's procedure using paired samples t tests. All p values are two-tailed, unless otherwise noted.
Mean RTs in the priming condition are shown in Figure 3A. The main effect of Congruency indicated robust priming across conditions (F1, 12 = 59.86, p < .001). A statistically significant SOA × Congruency × Hemifield interaction showed that priming was modulated by TMS (F3, 36 = 3.23, p = .03). This interaction was further analyzed by computing SOA × Congruency ANOVAs separately for the stimulated and control hemifield. In the control hemifield (ipsilateral with respect to TMS), only the main effect of Congruency reached significance (F1, 12 = 12.00, p < .001), and the absence of other effects (ps ≥ .2) verified that TMS did not modulate priming in the control hemifield. However, in the stimulated hemifield, TMS reduced the magnitude of priming (SOA × Congruency: F3, 36 = 3.26, p = .03), as shown in Figure 4. Pairwise comparisons indicated that priming was reduced during 70 msec (p = .04) and 100 msec (p = .02) TMS SOAs when compared with the amount of priming when no TMS was applied. The amount of priming was not different between the 70 and 100 msec SOAs (p = .3). Despite the reduction of priming magnitude, statistically significant priming was observed in all TMS SOA conditions (ps ≤ .03) when the RTs in the congruent and incongruent conditions were compared in the stimulated hemifield. The results are the same when only the participants in the fMRI group are considered (Figure 2C, bars).
Unconscious processing mediated by the SC have been reported to disappear when short wavelength stimuli (e.g., blue), to which the SC is insensitive to, are employed (Tamietto et al., 2009; Leh et al., 2006). Hence, if unconscious priming is mediated by the SC, TMS of early visual areas should more strongly affect priming by blue disks than red disks, as blue stimuli cannot rely on collicular projections. However, a Disk Color × SOA ANOVA with the magnitude of priming as a dependent variable indicated that priming by blue or red stimuli was not differently affected by TMS (ps ≥ .18), suggesting that priming by both colors rely on early visual cortex.
Concerning color recognition accuracies in the priming condition (see Figure 3B), ANOVA showed that, across conditions, response accuracy was lower in incongruent trials than in congruent trials (congruency: F1, 12 = 12.14, p = .005). The main effect of SOA (F3, 36 = 4.01, p = .02) demonstrated that response accuracies were, in general, influenced by TMS: Accuracies were higher during 100 msec SOAs than other TMS conditions (ps ≤ .05). However, the absence of other effects (ps ≥ .2) showed that TMS did not have any retinotopic effects on response accuracies.
Masked Prime Visibility
Our aim in this study was to investigate unconscious color priming. The control condition where the participants attempted to recognize the color of the masked prime and rated its visibility was used to verify that the participants did not consciously perceive the prime disk. As Figure 5B shows, when no TMS was applied, the participants, on average, reported having extremely faint percepts of the prime and did not consciously perceive the color of the prime. The mean visibility ratings are smaller than 1 (I saw the disk faintly, but I did not see its color) in all conditions (ps ≤ .04), which demonstrates that the participants were unconscious of the prime color and, often, of the stimulus altogether. Moreover, in the control condition, the participants were solely attending on the prime, which means that they perceived it more clearly than in the proper priming condition (Ramachandran & Cobb, 1995). The analysis of visibility data only showed an interaction between SOA and hemifield (F3, 36 = 4.28, p = .03), which reflected the result that in the stimulated hemifield TMS pulses during 70 msec (p = .02) and 100 msec (p = .03) SOAs suppressed the visibility of the prime even more when compared with the no-TMS condition of the same hemifield (Figure 5B). TMS did not affect visibility ratings in the control hemifield (p ≥ .4).
In the masked prime visibility task, forced-choice color discrimination accuracy was near chance level, and ANOVA produced no significant effects (ps ≥ .6). In summary, these results suggest that the participants did not consciously perceive the color of the masked disk prime and that TMS disturbed the processing of the prime without suppressing the masking effectiveness of the metacontrast annulus mask.
Conscious Color Recognition
Figure 6A shows the participants' performance when they were asked to recognize the color of the prime disk when it was presented alone, without the following annulus as a visual mask. Color recognition accuracy remained high for stimuli presented in the control hemifield, but in the stimulated hemifield, color recognition performance decreased during TMS. This was verified by the interaction between SOA and Hemifield (F3, 36 = 6.37, p = .001). The main effect of Hemifield (F1, 12 = 10.50, p = .007) was also statistically significant. Pairwise comparisons showed that TMS decreased recognition performance in the contralateral field during 70 msec (p = .007) and 100 msec (p = .02) SOAs when compared with performance in the same hemifield without TMS (40 msec SOA: p = .09). Performance was not statistically significantly different between the 70 and 100 msec SOAs in the stimulated hemifield (p = .3). TMS did not have any statistically significant effect on conscious recognition of the stimuli in the ipsilateral field.
Visibility ratings followed a similar pattern as recognition accuracies, as shown in Figure 6B. ANOVA yielded main effects of Hemifield (F1, 12 = 9.44, p = .01) and SOA (F3, 36 = 5.42, p = .02) and, more importantly, their interaction (F3, 36 = 7.03, p = .007). In the stimulated hemifield, prime visibility was reduced at all three TMS SOAs (ps ≤ .03) when compared with the visibility of targets presented in the same hemifield without TMS. Visibility ratings did not differ between 70 and 100 msec SOAs in the stimulated hemifield (p = .1). When no TMS was applied or the stimulus was presented in the control hemifield, participants on average reported perceiving the disk and its color clearly. The results remain the same when only the participants of the fMRI group are considered (Figure 2C, line).
We further examined the relationship between visibility ratings and color recognition accuracies during the most effective TMS SOA (70 msec) using a linear regression model. As Figure 3C illustrates, the model showed that visibility ratings strongly predicted (Ra2 = .68) color recognition accuracy (F1, 12 = 26.54, p < .001). The intercept of the model was .45, suggesting that when the target was not perceived at all, color recognition accuracy was at chance level. Unfortunately, we were not able to perform similar analysis at other SOAs or in the no-TMS condition, because the stimulus was rarely completely suppressed in these conditions.
To see whether the participants could unconsciously discriminate the target color, we examined forced-choice discrimination performance during the TMS trials in which the participants reported either not seeing the disk at all or reported being unconscious of the color of the disk. All TMS SOAs were included in the analysis to maximize the number of trials in the analysis. Only those participants who had at least 15 applicable trials were included in the analysis (n = 7, mean = 36 trials per participant, SD = 13 trials per participant). Mean color recognition accuracy was 59% (SD = 18%), which is not statistically significantly different from chance (t6 = 1.25, p = .3). However, when only the trials where the participants reported being unconscious of the color of the stimulus, but not of the stimulus altogether, were included (n = 6, mean = 31 trials per participant, SD = 16 trials per participant), performance (66% correct, SD = 17%) was statistically significantly above chance (t6 = 2.05, one-tailed p = .04). There were only three participants with enough data (17 trials on average, range = 15–19) to examine only the trials where the stimulus was completely suppressed from consciousness: their mean color recognition accuracy was 49% (range = 37–60%).
Modeling of the TMS-induced Electric Field in the fMRI Group
Across the six participants with fMRI-based localization of the stimulation area, the estimated TMS-induced electric field strength was 119 V/m (SD = 33 V/m) in V1, 122 V/m (SD = 22 V/m) in V2d, and 103 V/m (SD = 27 V/m) in V3d (calculated for each area from the retinotopic cortical location corresponding to the visual field location where the left hemifield stimuli were presented; Figures 1 and 2B). The estimated electric field was statistically significantly different only between areas V2d and V3d (one-tailed p = .03; other comparisons, one-tailed p ≥ .1). In each participant, the electric field was highest in either V1 or V2d. Thus, consistent with other reports (Salminen-Vaparanta, Noreika, Revonsuo, Koivisto, & Vanni, in press), when the pulse was directed to V1, TMS influenced all three early visual areas (V1–V3), although the emphasis was on V1 and V2d. Possible differences between the fMRI and phosphene mapping groups in conscious and unconscious performance were examined by including a Group (2: fMRI, phosphene mapping) factor to the ANOVAs described above. None of the retinotopic effects of TMS including the Group factor approached significance (ps > .21). Figure 2C shows that in the fMRI-group, TMS modulated stimulus visibility and priming magnitude in a similar way as in the whole sample. This result, together with the fact that “suppression-guided,” lower visual field TMS suppression has been shown to preferably affect V2d (Thielscher, Reichenbach, Uğurbil, & Uludağ, 2010), suggests that TMS affected V1–V3 also in the group where stimulation site was determined by phosphene mapping.
We studied the contribution of activity in early visual cortex to conscious and unconscious processing of chromatic information. Unconscious processing was assessed using unconscious metacontrast-masked color priming and by the participants' color discrimination performance during the trials when they reported not being aware of the stimulus or its color. Our results showed that TMS to early visual areas 70–100 msec after stimulus onset decreases, not only conscious perception of the stimulus color, but also unconscious color priming and color discrimination performance. The results suggest that activity in early visual cortex during this time window underlies both conscious and unconscious color processing. Thus, the findings provide causal support for the view that unconscious color processing depends on activity in early visual cortex (Breitmeyer, Ro, et al., 2004).
Single TMS pulses applied 70 or 100 msec after stimulus onset to early visual areas significantly reduced the subjective visibility and recognition of the stimulus and its color. This is in line with the finding that activity in V1 and V2 strongly correlate with the subjective experience of color (Maeda et al., 2010; Engel, Zhang, & Wandell, 1997). Our observation that color recognition accuracy was strongly predicted by visibility ratings is consistent with previous reports on the relationship between objective and subjective measures of vision in TMS studies (de Graaf, Cornelsen, Jacobs, & Sack, 2011). It is likely that during the 70 msec SOA, TMS interfered with feedforward activity, which has been estimated to begin approximately 60 msec after stimulus onset (Vanni et al., 2001; Wilson et al., 1983). Local cortical feedback activity also begins rapidly after the first sensory-driven signals (Bullier, 2003; Hupe et al., 2001), although the feedback signals, which have been linked to the formation of visual consciousness, are assumed to begin after 100 msec following stimulus onset (Railo, Koivisto, & Revonsuo, 2011; Supèr, Spekreijse, & Lamme, 2001). TMS also slightly decreased conscious visibility when TMS was applied 40 msec after the stimulus, which corresponds to some estimates of the very first stimulus-driven activation (Wilson et al., 1983). During the 40 msec SOA, TMS might also have influenced stimulus-related signals indirectly, for example, by modulating prestimulus visual cortical activity.
In the current study, color discrimination performance remained above chance level when the participants reported being unconscious of the stimulus color, but not of the stimulus altogether. In this respect, our results replicate the results of Boyer et al. (2005). Our results also extend this finding and suggest that, if activity in the early visual cortex is successfully disturbed during its initial processing stages, even unconscious color discrimination becomes impossible: When the trials where the participants reported not seeing the stimulus at all were also included in the analyses, color recognition performance was not statistically significantly above chance level.
Our result suggests that residual stimulus-related activity in early visual cortex, which might not be sufficient for conscious perception of color, may still contribute to unconscious processing of chromatic information. Radoeva, Prasad, Brainard, and Aguirre (2008) reached a similar conclusion by observing that, in a cortically blind patient (who demonstrated blindsight), the damaged early visual cortical areas nevertheless showed residual retinotopic activation. Crucially, the patient's performance in detecting stimuli of varying contrast levels in the impaired visual field was reflected in the neural response function of V1 in the lesioned hemifield (Radoeva et al., 2008). This type of residual activity might also underlie Type 2 blindsight, which is associated with “feelings” or faint subjective percepts concerning the stimuli presented in the blind field (Ffytche & Zeki, 2010; Weiskrantz, 1998).
Our finding emphasizes the importance of gradual measures of consciousness when studying unconscious processing (Overgaard, 2011). Nevertheless, it may still be asked whether the better-than-chance unconscious discrimination performance was based on purely unconscious processing of chromatic information. An alternative interpretation is that the response criterion for conscious color recognition was simply higher than for some other stimulus features. Thus, the observers may have had a brief conscious glimpse of the color, too, but because of a higher criterion did not report seeing the color, but only some other stimulus features. This would explain why the implicit color recognition took place only when some other features of the stimulus still crossed the subjective perceptual threshold but disappeared when the stimulus did not cross the subjective perceptual threshold at all. Previous TMS studies using gradual measures of consciousness have shown that, despite complete visual suppression by TMS over early visual areas, participants are nevertheless able to discriminate the location where a stimulus was presented (Railo & Koivisto, in press; Koivisto et al., 2011).
Because this study focused on short TMS SOAs, it remains possible that only conscious color perception might be suppressed at later time windows (e.g., during feedback phases), whereas unconscious processing of color or other stimulus features might be suppressed only during a more narrow early time window (e.g., during feedforward phase). Recent TMS studies that have measured both conscious and unconscious processes and manipulated SOA at wider ranges than this study, however, have failed to find any late V1/V2 time windows specific to conscious perception of stimulus features such as form (Koivisto et al., 2011; Sack et al., 2009) or motion (Koivisto, Mäntylä, & Silvanto, 2010).
Unconscious metacontrast masked color priming (Breitmeyer, Ro, et al., 2004; Schmidt, 2000) and other types of unconscious processing of metacontrast masked features (Railo & Koivisto, in press; Breitmeyer & Öğmen, 2006) have been explained by assuming that metacontrast masking spares the earliest stimulus-related signals. Our study provides support for this by showing that TMS suppression of the early stimulus-related signals decreases the magnitude of unconscious color priming. Moreover, the timing of the suppression of unconscious priming was comparable to the TMS suppression of a similar, unmasked stimulus. At group level, TMS did not completely eliminate priming, which can be explained by the fact that TMS did not always successfully disrupt stimulus-related activity. In addition, the TMS SOA, during which priming was maximally suppressed, varied between participants, which also explains why complete suppression of priming was not observed at group level. Individual differences are probably explained by differences in retinocortical transmission latencies (Stufflebeam et al., 2008) and by slight differences in TMS-induced electric fields. It could be argued that the activation of collicular (e.g., Tamietto et al., 2009) or geniculate projections to extrastriate areas (e.g., Cowey & Storeig, 1989) explains why priming was not completely suppressed, but this conflicts with the observation that TMS did completely eliminate unconscious priming in many individual participants in this study (priming was smaller than 10 msec in six participants during one of the TMS SOAs).
Our result that unconscious color discrimination and priming depend on early visual areas is at odds with the observation that blindsight patients can discriminate wavelengths in total absence of conscious perception (Stoerig & Cowey, 1989, 1992), and that unconscious chromatic stimuli, presented to the blind visual field, speed up responses to a visible stimulus (Tamietto et al., 2009). The disagreement might be explained by methodological differences, as in “TMS-induced blindsight” the interference is not limited to V1, as is the cortical damage in typical blindsight cases who demonstrate preserved chromatic discrimination capacities (Stoerig & Cowey, 1989; Stoerig, 1987). It is also possible that blindsight patients might have learned to base their decisions on weak, but totally unconscious chromatic signals. Finally, changes in connectivity, because of neural plasticity, might explain performance differences between healthy subjects and patients with blindsight (Bridge, Thomas, Jbabdi, & Cowey, 2008).
To conclude, our results provide evidence that, in healthy subjects, unconscious perception of color, whether measured by forced-choice discrimination or by unconscious priming, is mediated by early cortical activation in the same time window as conscious vision.
This work was supported by the Academy of Finland (grants 125175 and 12469). N. S.-V. was supported by the National Doctoral Programme of Psychology in Finland, and L. H. was supported by the Finnish Cultural Foundation. The authors would like to thank Simo Vanni for help with the fMRI measurements and Teemu Laine for technical assistance.
Reprint requests should be sent to Henry Railo, Department of Psychology, University of Turku, 20014 Turku, Finland, or via e-mail: email@example.com.