SEARCH

SEARCH BY CITATION

Keywords:

  • consciousness;
  • human brain;
  • perception;
  • primary visual cortex;
  • transcranial magnetic stimulation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

It has been suggested that unconscious visual processing of some stimulus features might occur without the contribution of early visual cortex (V1/V2). In the present study, the causal role of V1/V2 in unconscious processing of simple shapes in intact human brain was studied by applying transcranial magnetic stimulation (TMS) on early visual cortex or lateral occipital cortex (LO) while observers performed a metacontrast-masked response priming task with arrow figures as visual stimuli. Magnetic stimulation of V1/V2 impaired masked priming 30–90 ms after the onset of the prime. Stimulation of LO reduced the magnitude of masked priming at 90–120 ms, but this effect occurred only in the early parts of the priming experiment. A control task measuring the visibility of masked primes indicated that the orientation of masked primes could not be consciously discriminated and that TMS did not influence the conscious visibility of the primes indirectly by reducing the effectiveness of the mask in the critical time windows. We conclude that feedforward sweep of processing from V1/V2 (30–90 ms) to LO (90 ms and above) is necessary for unconscious priming of shape, whereas conscious perception requires also the contribution of recurrent (feedback) processing.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

An important topic in visual neuroscience is the difference between conscious and unconscious processing (e.g. Baars, 1988; Dehaene et al., 2006; Lamme, 2010). Activation along the ventral visual pathway from V1 to temporal cortex correlates with the contents of visual awareness (Vanni et al., 1996; Logothetis, 1998; Tong et al., 1998; Bar et al., 2001; Beck et al., 2001; Moutoussis & Zeki, 2002; Pins & ffytche, 2003), but the same visual areas are activated, although usually less strongly, also during unconscious processing (Kouider & Dehaene, 2007; Rees, 2007).

The contribution of V1 to conscious and unconscious perception has been under research recently (Crick & Koch, 1995; Ffytche & Zeki, 2011). According to an influential model (Lamme & Roelfsema, 2000; Lamme, 2004), basic stimulus features are extracted during the feedforward sweep from V1 to higher areas. Feedforward processing is assumed to be sufficient for unconscious processes, whereas recurrent (feedback) processing between higher cortical areas and V1 is necessary for visual awareness (Lamme, 2004). Unconscious visual processing may also rely on geniculo-extrastriate or extrageniculate retinotectal visual pathways which bypass V1 (Goodale & Milner, 1992; Ptito et al., 1999; Ro et al., 2004; Boyer et al., 2005; Ro, 2008). This would explain the residual visual abilities of V1-damaged patients with blindsight (Stoerig & Cowey, 1997; Ptito et al., 1999) who do not consciously perceive simple stimulus features in the blind field, but are still able to perform forced-choice discriminations concerning these features better than expected by chance.

Unconscious visual processing can be studied with metacontrast-masked response priming (Vorberg et al., 2003; Breitmeyer et al., 2004a) in which the visibility of a prime is masked by the target but the prime still influences the speed and accuracy of the responses. Sack et al. (2009) found that transcranial magnetic stimulation (TMS) 60–100 ms after prime onset over early visual cortex disrupted supraliminal (conscious) response priming of arrow figures. We studied the mechanisms of unconscious response priming and conscious perception in normal brain by applying TMS over early visual areas (V1/V2) or lateral occipital cortex (LO), an intermediate area along the ventral stream that is activated during the processing of the visual shape of objects (Grill-Spector et al., 2000; Grill-Spector, 2009) and is causally involved in priming of shapes (Silvanto et al., 2010). If unconscious visual processing relies on the geniculo-striate pathway (Lamme, 2004), TMS over V1/V2 should suppress masked priming. A feedforward mechanism would be implicated if the contribution of V1/V2 is critical before but not after the onset of LO activity. On the other hand, if V1/V2 continues to be critical after the onset of LO activity, support for recurrent processing is obtained. In contrast, if unconscious processing is mainly mediated by the geniculo-extrastriate or extrageniculate retinotectal visual pathway (Ro et al., 2004; Boyer et al., 2005; Ro, 2008), disruption of V1/V2 activity with TMS should not interfere with unconscious priming.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Participants

The participants were 15 healthy, right-handed, undergraduate or graduate students from the University of Turku (20–30 years old, six males) with normal or corrected-to-normal vision. Eight of them took part in the early visual cortex stimulation and seven in the lateral occipital cortex stimulation condition. The stimulation site was manipulated between participants to reduce the number of testing sessions, possible training or fatigue effects, and TMS pulses for each observer. The participants were blind to the purpose of the study, with the exception of H.R. who was blind to the timing of the TMS pulses. The study was conducted in accordance with the Declaration of Helsinki and with the understanding and written consent of each subject. It was accepted by the ethical committees of the University of Turku and the Hospital District of Helsinki and Uusimaa.

TMS

Single biphasic TMS pulses were administrated with a Nexstim (Helsinki, Finland) eXimiaTM stimulator and Nexstim biphasic 70-mm figure-of-eight coil. Earplugs were used to attenuate the sound of the TMS pulse-induced noise and a chin rest was used to obtain a stable head position. The coil was fixed on a holder and the coil plane was positioned tangentially on the head. In the LO stimulation condition, the TMS pulses were directed on the LO target site by using a magnetic resonance imaging (MRI)-guided eXimia Navigated Brain Stimulation (NBS) system (Nexstim), which registers continuously the relationship between brain and the TMS coil with a spatial resolution of 2 mm. The coordinates of the LO target site were derived on the basis of the functional MRI (fMRI) procedure (see below) and the retinotopic V1/V2 target site was determined with a functional procedure (see below).

Localization of the stimulation areas

The fMRI measurements for each participant in the LO stimulation condition were carried out with a 3-T MRI scanner (SignaTM HDxt; General Electric Inc.) with a phased array eight-channel coil. In the scanner, the visual stimuli were presented to a semitransparent screen with a three-micromirror data projector (Christie X3TM; Christie Digital Systems Ltd.) using presentationTM software (Neurobehavioral Systems Inc.). The major imaging parameters were: TR 1.8 s, TE 30 ms, FA 60°, FOV 20 cm, matrix 64 × 64 and slice thickness 3 mm. Twenty-nine slices were acquired in interleaved order.

Retinotopic LO representations were determined with the aid of 50 achromatic photographs of objects (1.3° or 3.1° in diameter), which were first contrasted with fixation alone. Four fMRI runs, each of 4 min, comprised blocks at nine different locations of the visual field. The low-level retinotopic areas (V1, V2, V3) and the visual motion sensitive area V5 were used as functional landmarks. Here the localization of low-level retinotopic areas was based on 24-region multifocal fMRI (Vanni et al., 2005). Four runs, each of 4 min, comprised 32 miniblocks of 7.3 s duration. During each miniblock, a subset of the 24 regions was stimulated. V5 was localized with one run of low-contrast expanding and contracting rings (24°) vs. rest. Standard pre-processing with slice-time and motion correction were followed by estimation of general linear model with the SPM8 MatlabTM toolbox. Consistent with earlier TMS studies of LO (Silvanto et al., 2010; Koivisto et al., 2011a), the LO in right hemisphere was selected for TMS target area. The LO in the right hemisphere was localized on the basis of the activation elicited by the object pictures presented at the visual field location corresponding to the stimulus location in the main experiments (lower left visual field at 2.1° eccentricity) in such way that the activation did not overlap with those elicited by the multifocal and motion stimuli, but was located posterior from V5, about halfway between V5 and low-level retinotopic areas. The coordinates of the approximate centre of the LO areas were extracted by using SPM and used as the TMS stimulation target sites. Figure 1A demonstrates the LO target area in the lateral occipital cortex (LO) with the borders of the low-level retinotopic visual areas and the location of V5 in reconstructed and inflated cortical surface (Dale et al., 1999) of a representative participant.

image

Figure 1.  (A) The lateral occipital cortex TMS target area in the right hemisphere of one of the participants. The TMS target area in the lateral occipital cortex (LO) is shown together with the borders of the low-level retinotopic visual areas and the location of the visual motion sensitive visual area V5 for a representative participant. The participant’s reconstructed and inflated cortical surface of the right hemisphere is viewed from the back. (B) Priming task and conscious recognition task. In the priming task, the participants responded as fast and accurately as possible to the left–right orientation of the target (which also served as a mask). The masked prime visibility task was physically identical to the priming task, but the participant’s task was to recognize the left–right orientation of the masked prime and to rate their aware perception of it. In the conscious recognition task, the participants were asked to recognize the unmasked prime and to rate their aware perception of it. In all the tasks, the stimuli were presented randomly to the lower left visual field (contralateral field relative to the position of the TMS coil) or to the upper right visual field (ipsilateral field).

Download figure to PowerPoint

In the V1/V2 condition the target site was determined with a functional method (Pascual-Leone & Walsh, 2001; Silvanto et al., 2005; Sack et al., 2009) in which phosphenes, TMS-induced flashes of light, were used to locate the stimulation site. The inclusion criterion in the V1/V2 condition was that each participant had to see phosphenes and that their position in the visual field (VF) depended retinotopically on the location of the TMS coil. By making use of the retinotopic organization of the phosphenes, the procedure aimed to position the TMS coil in such location that the induced phosphenes overlaid the spatial position where the critical stimuli in the experiment were presented. An array of four rectangles (1.26° × 0.38°), with a fixation point in the centre of the screen, was continuously displayed on the CRT monitor. The centre of each rectangle was positioned 2.1° from the fixation, either in the lower left visual field (VF), upper left VF, lower right VF or upper right VF. During the localization, the participants sat with their eyes closed in a dimly lit room, they imagined that they were fixating on the fixation point and kept the array of rectangles in mind, and reported verbally after each TMS pulse whether they saw phosphenes or not. The starting point of stimulation was 2 cm dorsal from the inion. The stimulation began with an intensity of 40% maximum stimulator output. This was gradually increased in steps of 5% until phosphenes were reported. When the participant saw a phosphene, they were asked to describe exactly where in the VF the phosphene appeared, and whether it overlapped with any of the rectangles in the array that they were instructed to keep in mind. Between trials, they were allowed to open their eyes and refresh their image of the array. The coil was systematically moved vertically or laterally according to the participant’s reports to find a region from which a clear phosphene touched the lower left VF rectangle (i.e. the position of the critical stimuli in the main experiment) but not the upper right VF rectangle (i.e. the location where noncritical stimuli were presented). The mean coil position eliciting phosphenes in the lower left field was 2.5 cm (SD = 0.5) dorsal and 1.6 cm (SD = 0.5) to the right from the inion. Recent studies modelling the TMS-induced electric field in brain suggest that from this coil position V1 and V2 (and V3 to a lesser extent) will be stimulated (Thielscher et al., 2010; Koivisto et al., 2011b; Salminen-Vaparanta et al., 2012).

During the actual experiment, the TMS intensity was set to 65% of the maximum output of the stimulator. This intensity did not produce eye blinks, muscle twitches or other uncomfortable sensations. The direction of current in the second phase of the TMS pulse was from lateral to medial.

Stimuli

The stimuli were presented with e-prime software on a 19-inch CRT monitor (75 Hz, 1024 × 768 pixels) from a viewing distance of 150 cm. The stimuli were black arrows (0.4 cd/m2) pointing either to the left or to the right, presented on a white background (51 cd/m2) (Fig. 1B). The prime was a small arrow (1.26° × 0.38°) and the target, which also served as mask for the prime, was a larger arrow (1.9° × 0.7°). The inner centre of the target was white so that the outer contour of the prime touched the inner contour of the target (i.e. mask). The stimuli were centred 2.1° away from the fixation cross and presented randomly to the left lower VF or to the upper right VF in all tasks. Because the right hemisphere was magnetically stimulated, the ipsilateral right VF stimuli served as control stimuli. Due to the representation of the right VF in the early visual areas of the left hemisphere, the processing of ipsilateral stimuli was not expected to be influenced by TMS in the V1/V2 condition (Sack et al., 2009; Koivisto et al., 2011b).

Experimental procedure

Each participant was tested in two sessions, separated by about 1 week. The experiment consisted of three types of tasks: conscious recognition task, masked prime visibility task and priming task. In the first session, the participants performed the conscious recognition task (four TMS blocks) and half of the priming task (four TMS blocks), with the order of these tasks counterbalanced across the participants. In the second session, each participant performed first the remaining priming blocks (four TMS blocks) and then the masked prime visibility task (four TMS blocks). Baseline performance without TMS (no TMS) in each task was measured by separate blocks that were performed between the TMS blocks.

Priming task

In the priming task (Fig. 1B), each trial began with the presentation of the fixation point for 550 ms. This was followed by the prime for 13 ms in either the lower left VF or the upper right VF. After an interstimulus interval of 53 ms, the target (i.e. mask) was presented for 160 ms. Thus, the prime–target stimulus-onset asynchrony (SOA) was 67 ms. The prime and target appeared always on the same location (in half of the trials in the lower left VF and in half in the upper right VF). The orientation of the prime and target was the same (congruent) in half of the trials and different (incongruent) in the other half. The participants were asked to keep their eyes fixated on the fixation point and to respond as quickly and accurately as possible to the orientation of the target. Responses were given with the left or right index finger by pressing a button in a response pad. According to pilot testing, the selected 67-ms prime–target asynchrony should result in robust priming (faster and more accurate responses to congruent targets than to incongruent ones) without subjective awareness of the prime’s direction.

The TMS condition included eight stimulus blocks, with 48 stimuli in each block. A single TMS pulse was applied randomly 30, 60, 90, 120, 150 or 180 ms after the onset of the prime. For each participant, there were 16 congruent and 16 incongruent trials in the lower left VF (contralateral VF relative to TMS coil) at each SOA and 16 congruent and 16 incongruent trials in the upper right VF (ipsilateral VF) at each SOA. Each of the six baseline (no TMS) blocks included 24 trials, with six congruent and six incongruent trials in each VF, so that the total number of trials was 36 in each congruency/VF condition.

Conscious recognition task

Each trial began with the presentation of the fixation point for 550 ms, followed by the prime stimulus for 13 ms. The prime was presented in half of the trials in the lower left VF and in the other half in the upper right VF. The participants made a forced-choice discrimination response (left or right) to the orientation of the stimulus (i.e. prime) and then reported their subjective awareness of it by pressing one of four buttons (0 = ‘I did not see the prime at all’; 1 = ‘I saw something but I did not have any trace of the orientation of the prime’; 2 = ‘I saw the orientation but not clearly’; 3 = ‘I saw the orientation rather clearly’). It was stressed that ratings 0 and 1 meant that the participant did not have any idea of the orientation of the stimulus. To make sure that the participants followed the scale correctly, they were asked to explain the meaning of each button after performing 16 practice trials. In the forced-choice task, accuracy was stressed but the participants were asked to press the corresponding button immediately when they knew the answer. Four TMS blocks (48 trials in each) were run with the same SOAs as in the other tasks: a single TMS pulse was applied randomly 30, 60, 90, 120, 150 or 180 ms after the onset of the prime (a total of 16 trials in each SOA/VF condition). Three baseline (no TMS) blocks (24 trials in each) were run between the TMS blocks (a total of 36 trials in each VF condition).

Masked prime visibility task

In this task, visual stimulation was physically identical to the priming task (Fig. 1B). It was conducted to control for the visibility of the masked primes and to estimate the possible effects of TMS on the visibility of the primes. This task was always performed as the last task in the end of the second testing session. The participants were informed about the existence of the prime and that this task would be run in order to test whether or not they could see the prime. In each trial, the participants made a forced-choice discrimination response concerning the orientation of the prime (left or right) and after that they rated their subjective awareness with the same scale that was used in the conscious recognition task (see above). Only accuracy was stressed in the instructions of the forced-choice task. Four TMS blocks (48 trials in each) were run with the same SOAs as in the other tasks (a total of 16 trials in each SOA/VF condition). Three baseline (no TMS) blocks (24 trials in each) were run between the TMS blocks (a total of 36 trials in each VF condition).

Data analyses

Response times (RTs) were measured from the onset of the target in the priming task and from the onset of the prime in the conscious recognition task. In priming tasks, in particular, speed–accuracy trade-off may result in strong priming effects in accuracy (fast and accurate responses to congruent targets and fast but incorrect responses to incongruent targets), which attenuates the priming effects in RTs because the fast but erroneous incongruent trials do not contribute by decreasing RTs to incongruent targets. To take such trade-offs into account, the RT data were analysed by dividing each subject’s RTs (correct and incorrect trials) by the accuracy in each condition. This procedure is well established and has often been used in behavioural (Townsend & Ashby, 1983) and TMS (e.g. Cattaneo et al., 2009) studies.

The accuracy and subjective awareness data were combined into a single variable in the conscious recognition task: a stimulus was scored as consciously recognized if its orientation was correctly discriminated and the subjective rating of awareness indicated that the participant was aware of its orientation (2 = ‘I saw the orientation unclearly’ or 3 = ‘I saw the orientation rather clearly’). The results of the conscious recognition task and the masked prime visibility task were first analysed with Stimulation Site (2: V1/V2 vs. LO) × VF (2: contra- vs. ipsilateral) × TMS (7: no TMS and the SOAs of 30, 60, 90, 120, 150 and 180 ms) anovas. Significant interactions were followed by separate anovas on each Stimulation Site or VF condition. Significant main effects of TMS were further analysed with post hoc tests (Fisher’s procedure) by comparing the results at each SOA with those in the no TMS condition. In the anovas, we report Huynh–Feldt-corrected P-values always when the degrees of freedom are higher than 1, i.e. when TMS is included as a factor.

In the main analyses of the priming task, TMS × Congruency × VF anovas were conducted on the data. Based on the masked prime visibility (control) condition we restricted the main analyses of priming to those SOAs where TMS did not affect processing of the masks (which served also as the targets).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We report first the results from the conscious recognition task and masked prime visibility task as they are important in the interpretation of the results from the priming task.

Conscious recognition task

Proportion of consciously recognized unmasked primes

A Stimulation Site (2) × VF (2) × TMS (7) anova on the proportion of consciously recognized primes (Fig. 2) showed that the main effects for Site and TMS as well as all the interactions were statistically significant (F ≥ 9.40, P ≤ 0.009). This suggests that the effect of TMS was different depending on the stimulation site and VF.

image

Figure 2.  Conscious recognition of unmasked prime. The proportion of consciously recognized primes (correct recognitions with subjective rating of awareness indicating that the participant saw the orientation) and reaction times in response to primes in the contra- and ipsilateral visual field (relative to the position of the TMS coil) when no TMS was applied or when TMS was applied 30, 60, 90, 120, 150 or 180 ms after the onset of the prime. Error bars in Figs 2–5 are standard errors of the mean.

Download figure to PowerPoint

In the V1/V2 condition, a VF × TMS interaction (F6,42 = 13.82, P < 0.001) suggests that the effects of TMS on awareness depended on whether the stimuli were displayed in the contralateral or ipsilateral VF. Further one-way anovas separately on each VF showed a main effect for TMS in the contralateral VF (F6,42 = 15.75, P < 0.001) but not in the ipsilateral VF (F6,42 = 1.08). Post-hoc tests indicated that in the contralateral field conscious recognition was impaired in relation to no TMS baseline at all SOAs (all P ≤ 0.022). As Fig. 2 shows, there is one large dip in the contralateral field which begins at an SOA of 30 ms, the bottom of the dip occurs at 60 and 90 ms SOAs, and this is followed by a gradual improvement as SOA increases. Analysis of LO condition did not reveal any statistically significant effects, suggesting that TMS on LO did not impair conscious recognition.

Response times

The Stimulation Site (2) × VF (2) × TMS (7) anova on RTs showed that all the main effects and interactions were statistically significant (F ≥ 7.66, P ≤ 0.003). In the V1/V2 condition, a VF × TMS interaction (F6,42 = 11.39, P < 0.001) indicates a strong retinotopic influence of TMS on RTs. Further analyses showed that TMS had an effect on RTs in the contralateral VF (F6,42 = 11.25, P < 0.001) but not in the ipsilateral VF (F6,42 = 2.41, P = 0.143). In the contralateral VF, RTs were delayed at all the SOAs relative to no TMS (P ≤ 0.009).

In the LO condition, TMS had less clear retinotopic effects than in the V1/V2 condition. The main effect for TMS (F6,36 = 5.43, P < 0.001) indicates that responses were delayed by TMS relative to no TMS at SOAs of 90 ms (P = 0.018), 120 ms (P = 0.023) and 150 ms (P = 0.011). The VF × SOA interaction (F6,36 = 2.40, P = 0.047) shows that TMS on LO had a retinotopic influence on RTs, but this was restricted to the longest SOA. Although separate analyses of VFs indicated that the main effects of TMS were significant in both VFs (contralateral: F6,36 = 3.66, P = 0.006; ipsilateral: F6,36 = 4.00, P = 0.004), the retinotopic influence of TMS was due to the linear increase of RTs, which was present only in responses to stimuli in the contralateral field (linear trend: P = 0.009). Unlike in the ipsilateral VF, in the contralateral VF TMS delayed RTs also at the longest (180 ms) SOA relative to no TMS (P = 0.027).

In summary, TMS applied on V1/V2 had a clear retinotopic suppressive effect on the measures of conscious recognition, with the bottom of the dip at 60 and 90 ms after the onset of the stimulus, followed by gradual improvement at longer SOAs. TMS over LO only had an effect on RTs, delaying them from the SOA of 90 ms onwards. Stimulation of LO produced less clearly lateralized effects than stimulation of V1/V2, which may reflect weaker retinotopic organization at higher levels of hierarchy (Grill-Spector, 2009). The pattern that the critical (although weak) contribution of V1/V2 was observed both before and after the contribution of LO had started suggests that conscious perception relied first on feedforward activation from V1 to ventral stream, followed by a later period during which early visual areas engaged in recurrent interaction with higher areas.

Masked prime visibility

Subjective awareness

On average, the ratings of subjective awareness (scale 0–3) were below 1, indicating that the participants were unconscious of the orientation of the prime (1 = ‘I saw something but I did not have any trace of the orientation of the prime) (Fig. 3). The Stimulation Site (2) × VF (2) × TMS (7) anova on the ratings of subjective awareness in the masked prime visibility task revealed a main effect for TMS (F6,78 = 9.89, P < 0.001) which was modified by TMS × Site (F6,78 = 5.02, P = 0.001) and TMS × Site × VF (F6,78 = 3.20, P = 0.016) interactions.

image

Figure 3.  Masked prime visibility task. (Top) The subjective rating of awareness (scale 0–3) and (bottom) recognition accuracy in the contra- and ipsilateral visual field (relative to the position of the TMS coil) when no TMS was applied or when TMS was applied 30, 60, 90, 120, 150 or 180 ms after the onset of the prime.

Download figure to PowerPoint

In the V1/V2 condition, the effect of TMS (F6,42 = 10.26, P < 0.001) was retinotopic, as indicated by the VF × TMS interaction, F6,42 = 3.73, P = 0.013. TMS influenced subjective awareness of masked primes in the contralateral VF (F6,42 = 8.06, P < 0.001) by reducing it at the SOAs of 60 ms (P = 0.022) and 90 ms (P = 0.019) relative to no TMS and by increasing it at the SOA of 180 ms above the level observed in no TMS baseline (P = 0.043). TMS over V1/V2 influenced subjective awareness also in response to masked primes in the ipsilateral VF (F6,42 = 2.54, P = 0.043). This effect is due to the decrease of awareness at the shortest (30 ms) SOA (P = 0.009) relative to no TMS. In the ipsilateral VF, TMS did not improve awareness above the no TMS baseline level at any of the SOAs. Stimulation of LO did not have any statistically significant effects on subjective visibility of masked primes.

Thus, as can be expected on the basis of the findings in the conscious recognition task, TMS on V1/V2 impaired visibility of masked primes presented to the contralateral VF in the time windows (60 and 90 ms SOAs) where it also had the strongest effects on conscious perception of the unmasked prime. In later time windows, TMS increased the visibility of masked primes. This effect must be an indirect one in which TMS influenced processing of the target (i.e. mask) by reducing its masking effectiveness and hence indirectly enhanced the visibility of the prime (see also Ro et al., 2003). Figure 3 reveals that this effect seems to begin even at an SOA of 120 ms where the visibility of masked primes in the contralateral VF increases nonsignificantly above that of primes in the ipsilateral VF. It is important to keep in mind that the SOAs here refer to the interval between the onsets of prime and TMS. The asynchrony between the onsets of the prime and target was 67 ms. Thus, at the 120-ms prime–TMS SOA, the TMS pulse is delivered 53 ms after the onset of the target; the results of the conscious recognition task indicated that at this latency range TMS over V1/V2 impaired conscious recognition of the prime.

Accuracy

Stimulation Site (2) × VF (2) × TMS (7) anova on accuracy (Fig. 3) in the masked prime visibility task showed only a Site × VF interaction (F1,13 = 4.81, P = 0.047). This result was due to the fact that in the V1/V2 condition performance was better in the ipsilateral than in the contralateral VF (F1,7 = 13.90, P = 0.007), while there was no differences between VFs in the LO condition. In the contralateral field of V1/V2 condition, forced-choice performance did not exceed the 50% chance level in any TMS conditions (one-sample t-tests, P > 0.11); in the ipsilateral field, forced-choice accuracy was higher than expected by chance in the no TMS baseline (P = 0.014) and at the SOAs of 30 ms (P = 0.005) and 180 ms (P = 0.03). In the LO condition, accuracy never exceeded the chance level (P > 0.05). Note that the subjective visibility rating must be considered as the primary measure of awareness because it more directly assesses the conscious experience of observers and because forced-choice accuracy in the absence of reported awareness may exceed the chance level under some circumstances (Boyer et al., 2005; Koivisto et al., 2010, 2011b).

Priming

Analyses of the conscious recognition of unmasked primes and masked prime visibility task suggest that TMS over V1/V2 at the SOAs of 30, 60 and 90 ms did not indirectly influence the processing of primes via affecting the processing of the target (i.e. mask). Therefore, the primarily statistical analyses of the V1/V2 condition focused on these SOAs. In the LO condition, all six SOAs were included as there was no evidence that TMS over LO would influence processing of primes indirectly via affecting the processing of the targets. Priming in RT measures and accuracy was defined as the difference between responses to incongruent and congruent trials. Accuracy data replicated the RT results and are available online as Supporting Information (Fig. S1).

The 2 × 2 × 4 anova on RTs in the V1/V2 condition (Fig. 4) shows a main effect of congruency (F1,7 = 21.45, P = 0.002), indicating that priming occurred. RTs were longer in response to targets in the contralateral VF as compared with those in the ipsilateral VF (F1,7 = 9.63, P = 0.017). VF × Congruency interaction shows (F1,7 = 27.44, P = 0.001) that there was larger priming in the ipsi- than contralateral VF. Most importantly, TMS influenced priming effects differently in the VFs (VF × Congruency × TMS: F3,21 = 3.64, P = 0.035). Separate analysis of the data from the contralateral VFs showed a Congruency × TMS interaction (F3,21 = 5.79, P = 0.005), indicating that TMS to V1/V2 reduced priming. Post hoc comparisons indicated that priming effects were reduced by TMS relative to no TMS condition at SOAs of 30 ms (P = 0.006), 60 ms (P = 0.024) and 90 ms (P = 0.026). Further analyses comparing RTs to congruent and incongruent targets showed that the priming effect in the contralateral VF was significant in the no TMS condition (t7 = 4.28, P = 0.004), but not at SOAs of 30 ms (t7 = 1.07) and 60 ms (t7 = 1.46). At an SOA of 90 ms, the priming effect was significant (t7 = 3.15, P = 0.016), although it was smaller than that in the no TMS condition. In the ipsilateral VF, congruency did not interact with TMS (F3,21 = 2.27, P = 0.133), suggesting that TMS to V1/V2 did interfere with priming only in the contralateral VF relative to the position of the TMS coil.

image

Figure 4.  Response times in the priming task to targets presented in the contralateral and ipsilateral visual field (relative to the position of the TMS coil) when no TMS was applied or when TMS was applied 30, 60, 90, 120, 150 or 180 ms after the onset of the prime. In congruent trials the prime and target had the same left–right orientation, whereas in incongruent trials it was different. The priming effect is the difference in response times between congruent and incongruent trials. Asterisks indicate the latencies at which TMS impaired priming relative to no TMS baseline. The data points on the left side of the dotted vertical line (upper-left panel) represent data points in which TMS to early visual cortex did not impair processing of targets which served also as masks.

Download figure to PowerPoint

Analysis of the longer SOAs of 120, 150 and 180 ms (at which TMS over V1/V2 directly affected the processing of the target) revealed a reliable effect for congruency (i.e. priming) (F1,7 = 2.27, P = 0.009), which did not interact with TMS or VF. It is important to note that there is no indication (Fig. 4, upper-left panel) that TMS would interfere with priming at these long SOAs. It seems to slow down specifically the RTs to incongruent targets in the contralateral field while the responses to congruent targets are less affected. This pattern is expected because the incongruent condition is more demanding and hence more vulnerable to interference.

The 2 × 2 × 7 anova on the LO condition revealed a highly significant main effect of congruency (F1,6 = 125.11, P < 0.001), but no interactions. Thus, it appeared that TMS applied to LO did not affect priming. However, as each participant was tested in two separate sessions, we observed after the first sessions that all the participants in the LO condition showed TMS-induced suppression of priming in the contralateral field at an SOA of 120 ms (62 ms priming effect in the no TMS condition, 30 ms at an SOA of 120 ms; TMS × Congruency: F1,6 = 16.42, P = 0.007), while there was no corresponding suppression in the ipsilateral VF. This unexpected pattern suggests that the role of LO in priming may change as a function of training. This hypothesis deserves further testing.

To perform a stronger and independent test of whether the contribution of LO was critical for priming only in the beginning of the priming task, we analysed separately the RTs in the beginning of the task (the first two TMS and no TMS blocks of the first session) and in the end of the task (the last two TMS and no TMS blocks of the second session) (Fig. 5). Thus, the trials in the middle of the task were excluded from the analyses. The Period (2: beginning vs. end) × VF (2) × Congruency (2) × TMS (7) anova on RTs in the LO stimulation condition showed a significant four-way interaction (F6,36 = 3.39, P = 0.015), confirming that TMS had a retinotopic effect on priming only at the beginning of the task. At the beginning of the priming task, TMS suppressed priming in relation to no TMS baseline in the contralateral VF at SOAs of 90 ms (F1,6 = 10.50, P = 0.018) and 120 ms (F1,6 = 12.50, P = 0.012), and marginally significantly at an SOA of 150 ms (F1,6 = 5.87, P = 0.053). In the ipsilateral VF, the priming effects did not differ significantly from those in the no TMS condition at any of the SOAs. At the end of the task, TMS did not significantly suppress priming either in the contra- or the ipsilateral field at any of the SOAs. Thus, TMS to LO interfered with priming only at the beginning of the task. Analysis of V1/V2 stimulation condition did not reveal this kind of phenomenon.

image

Figure 5.  Response times to targets in the contralateral and ipsilateral visual field (relative to the position of the TMS coil) at the beginning and at the end of the priming task in the lateral occipital cortex stimulation condition. The priming effect is the difference in response times between congruent and incongruent trials. Asterisks indicate the latencies at which TMS impaired priming relative to no TMS baseline.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The present TMS study examined whether the contribution of early visual cortex (V1/V2) is necessary only for conscious perception or also for unconscious visual processing as measured with metacontrast-masked priming. The role of early visual areas was critical for masked priming at 30–90 ms after stimulus onset (Fig. 4) and for conscious perception 30–180 ms after stimulus onset (Fig. 2). These effects were retinotopic, occurring for stimuli presented to the contralateral visual field relative to the position of the TMS coil. Therefore they cannot be due to nonspecific effects of TMS.

The critical activity period of LO in conscious recognition speed began about 90–120 ms after stimulus onset (Fig. 2); in this time window LO also played a role in masked priming at the beginning of the priming task (Fig. 5). In general, TMS on LO produced less robust and lateralized effects than TMS on V1/V2. This is consistent with current brain imaging data that show weaker retinotopy in intermediate (e.g. LO) and high areas along the ventral stream than in early visual areas (Grill-Spector, 2009). The masked prime visibility task in which the observers tried to recognize the primes confirmed that the orientation of the masked primes remained at unconscious level (Fig. 3).

Conscious recognition

In the conscious recognition task (Fig. 2), the critical activity period of V1/V2 occurred 30–180 ms after stimulus onset, with the largest TMS suppression occurring at SOAs of 60–90 ms, which correspond to the latency of the classical occipital dip (Amassian et al., 1989; Corthout et al., 1999; Kammer et al., 2005; Sack et al., 2009; Koivisto et al., 2011b; Railo & Koivisto, 2012; Salminen-Vaparanta et al., 2012). The critical activity period of LO, reflected in RTs, began about 90–120 ms after stimulus onset, i.e. after the peak contribution of V1/V2. The activation latency of LO fits well with the estimates that all the extrastriate visual areas in the ventral pathway are activated 100–120 ms after stimulus onset (Lamme & Roelfsema, 2000; Boehler et al., 2008; Liu et al., 2009). Importantly, the contribution of V1/V2 overlapped and even extended beyond the onset of the LO time window, suggesting that the role of V1/V2 in conscious perception was not restricted to feedforward processing. This timing fits with the estimates that recurrent processing in V1 begins rapidly after information has been sent to extrastriate areas, around 100–120 ms after stimulus onset (Boehler et al., 2008; Koivisto et al., 2010; Lamme & Roelfsema, 2000; Pascual-Leone & Walsh, 2001; Silvanto et al., 2005; Supèr et al., 2001).

Thus, our results for conscious recognition are consistent with the models assuming that recurrent processing between higher areas and early visual cortex plays a role in shaping visual awareness (Lamme & Roelfsema, 2000; Hochstein & Ahissar, 2002; Lamme, 2004). The present type of experiment cannot, however, directly prove that the early visual cortex engages in recurrent interaction specifically with LO. The results only show that during the late V1/V2 time window the feedforward sweep of processing has reached higher extrastriate areas and therefore the late activity probably serves other functions than feeding information forward. This account is also consistent with the possibility that posterior parietal areas and prefrontal areas exert attentional modulation on ventral activity (Bullier, 2001; Bar, 2003; Chambers et al., 2004; Koivisto & Silvanto, 2012). Finally, the present suppressive effects at late V1/V2 latencies were relatively weak, raising the possibility that the role of recurrent feedback is to amplify or enrich conscious visual perception in a modulatory manner (Macknik & Martinez-Conde, 2007) rather than to serve as a mechanism which allows the representations of visual objects to access consciousness.

However, we cannot rule out the possible existence of a late TMS dip, reflecting recurrent V1/V2 activity after 180 ms. Such a period seems unlikely because previous TMS studies have not found any visual suppression for visual perception of simple forms or features after 180 ms; such late occipital suppression effects have been restricted to processing of more complex stimuli such as feature conjunctions (Juan & Walsh, 2003; Koivisto & Silvanto, 2012), photographs of objects in their natural settings (Camprodon et al., 2010; Koivisto et al., 2011a) and figures requiring segregation from ground (Heinen et al., 2005).

Priming

A major question that was addressed here was whether unconscious visual processing of shape (as measured with metacontrast-masked priming) depends on activity in the early visual cortex or whether it proceeds via pathways that bypass them. The results of the priming task (Fig. 4) were clear-cut: TMS applied to early visual cortex reduced priming, suggesting that unconscious processing depended on the functioning of the geniculo-striate pathway. Moreover, unconscious priming was completely eliminated as early as during the 30-ms SOA, although TMS did not yet strongly reduce conscious recognition in this time-window. This means that during the 30-ms SOA TMS suppressed visual signals that were essential for priming, but not necessary for conscious perception. These could be transient magnocellular signals, which have been suggested to be spared by metacontrast masking (Breitmeyer et al., 2004b; Railo & Koivisto, 2012). Thus, the neural activation in extrastriate areas that is produced by input from pathways bypassing V1 may support only simple unconscious cognitive functions such as detection of the mere appearance or localization of the stimulus (Christensen et al., 2008; Koivisto et al., 2011b; Railo & Koivisto, 2012) or detection of simple features (Alexander & Cowey, 2010) but not sufficiently detailed processing of the shape that was needed for priming in the present study.

Stimulation of LO had a weak, but consistent, effect on priming at the beginning of the task but not at the end (Fig. 5). The effect of LO stimulation on priming occurred in the same time window (about 90–120 ms after prime onset) where the stimulation also started to influence the speed of conscious recognition. Involvement of the ventral stream (which is dominated by parvocellular neurons) in unconscious priming is supported from a recent psychophysical study by Tapia & Breitmeyer (2011). They observed that metacontrast masked priming of arrow shapes was governed more strongly by contrast-response characteristics of parvocellular neurons than by contrast-response characteristics of magnocellular neurons, suggesting that unconscious priming relies on feedforward activation of primarily the parvocellular channels along the ventral pathway.

The role of LO in priming was very limited. LO contributed to priming only at the beginning of the priming task and TMS only reduced the priming effect to about half that in the baseline without TMS, suggesting that areas other than LO play more critical roles in unconscious priming. Thus, the present results are consistent with the role of dorsal stream (from V1 to parietal cortex) in unconscious processing (Goodale & Milner, 1992). In posterior parietal cortex, intraparietal sulcus has been associated with relatively high-level ability to represent object form (Konen & Kastner, 2008) and therefore it is a strong candidate for a neural substrate of unconscious response priming. Its potential causal role remains to be tested in further experiments.

The finding that masked priming was well preserved in the LO stimulation at the end of the priming task also implies that the contribution of ventral stream and LO is necessary only when participants in the task are not yet well trained. It might be necessary for processing the form of the objects for a limited time period until automatic stimulus–response mappings (Neumann, 1990) or action triggers (Kiesel et al., 2007) are formed. The action triggers are created in response to consciously visible targets but will be triggered also by other similar items, such as the invisible prime. Thus, the creation of action triggers is intentional and probably consciousness-mediated (Kunde et al., 2003) and therefore likely to rely on ‘conscious’ ventral stream. Our data suggest that the cerebral organization of masked priming becomes independent of ventral stream activity only after a relatively long exposure to the task. An alternative explanation, that LO would become desensitized to the influence of TMS as a result of the accumulating number of TMS pulses, is not likely as visual cortex has not been found to be vulnerable to such effects (Murd et al., 2010). It can be argued also that because LO constitutes a relatively large area in brain, TMS did not hit the critical location in LO. However, the TMS hotspots in LO were defined and stimulated on the basis of fMRI and navigated brain stimulation as precisely as currently possible.

It is important to note that the stimuli were simple arrow figures with two possible orientations and presented repeatedly during the experiment. They may have been too simple stimuli for detecting the contribution of LO on the accuracy of conscious perception (LO contributed only to the speed of conscious recognition). In addition, TMS target site was manipulated between participants. We do not believe that the between-participants design was a problem for making inferences about the critical time windows of the stimulated areas, because there was strong consistency across participants in the time windows within both stimulation site conditions: the LO activity period in conscious recognition and masked priming started after the peak of the V1/V2 activity had occurred at 60–90 ms. Thus, the temporal order of the activity periods of V1/V2 and LO could be reliably concluded in the present experiment. A within-participants design with large number of trials and testing sessions for each observer (e.g. manipulating the target site between successive stimulus blocks) might not have been sensitive enough to detect the effects of LO stimulation in the early parts of the priming task when it was novel for each observer.

We conclude that both conscious and unconscious processing of simple shapes rely on feedforward activity via V1 to higher areas. The feedforward sweep may be sufficient for unconscious processing to occur, but as influential models suggest (Lamme & Roelfsema, 2000; Bullier, 2001; Hochstein & Ahissar, 2002; Lamme, 2004, 2010), full-blown conscious perception depends also on recurrent feedback to early visual areas. Whether the recurrent feedback mechanism is directly related to the mechanism of visual consciousness (Lamme, 2010) or whether it only amplifies or enriches the conscious visual experience in a modulatory manner (Macknik & Martinez-Conde, 2007) remains to be studied in further experiments.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This work was supported by the Academy of Finland (grant numbers 125175 to M.K., 124698 to L.H., 124623 to A.R.) and the Finnish Cultural Foundation to L.H.

Abbreviations
fMRI

functional magnetic resonance imaging

LO

lateral occipital cortex

RT

response time

SOA

stimulus onset-asynchrony

TMS

transcranial magnetic stimulation

VF

visual field

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Fig. S1. Accuracy in the priming task in response to targets presented in the contralateral and ipsilateral visual field (relative to the position of the TMS coil) when no TMS was applied or when TMS was applied 30, 60, 90, 120, 150 or 180 ms after the onset of the prime.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset by Wiley-Blackwell. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
EJN_7973_sm_FigS1.doc204KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.