Divergence of fMRI and neural signals in V1 during perceptual suppression in the awake monkey

Abstract

The role of primary visual cortex (V1) in determining the contents of perception is controversial. Human functional magnetic resonance imaging (fMRI) studies of perceptual suppression have revealed a robust drop in V1 activity when a stimulus is subjectively invisible. In contrast, monkey single-unit recordings have failed to demonstrate such perception-locked changes in V1. To investigate the basis of this discrepancy, we measured both the blood oxygen level-dependent (BOLD) response and several electrophysiological signals in two behaving monkeys. We found that all signals were in good agreement during conventional stimulus presentation, showing strong visual modulation to presentation and removal of a stimulus. During perceptual suppression, however, only the BOLD response and the low-frequency local field potential (LFP) power showed decreases, whereas the spiking and high-frequency LFP power were unaffected. These results demonstrate that the coupling between the BOLD and electrophysiological signals in V1 is context dependent, with a marked dissociation occurring during perceptual suppression.

Figure 1

Generalized flash suppression paradigm for block design experiment. A. Stimulus conditions. The parameters of the salient target and the random dot surround were adjusted to create five different conditions, with individual trials lasting six seconds (one-minute blocks consisted of up to 9 such trials). In the generalized flash suppression (INV) condition, the appearance of surrounding dots to both eyes consistently induced the monocular target to disappear (the time period during which the target stimulus disappears is indicated with a thick orange line). In the physical removal (OFF) condition, the target was physically extinguished upon appearance of the dots, mimicking perceptual suppression. In the temporal reversal (VIS_TR) and binocular (VIS) conditions, the stimuli were adjusted by reversing the order of target (red disk) and surround (moving random dots), or by presenting the target to both eyes, respectively, to ensure that the target did not disappear. Finally, in the fixation (FIX) condition (not depicted), the screen remained blank as the monkey fixated a small cross throughout the trial. (Ton = target on, Toff = target off, Son = surround on, Soff = surround off). B. The two modes of block design used for the fMRI experiments. In the sequential block design, the five stimulus conditions alternated in a fixed temporal order, keeping the number of stimulus conditions per run constant. In the randomized block design, each condition was presented in pseudorandomized order, each preceded by a blank screen fixation condition. C. Psychophysical responses during the different conditions. The probability of perceptual suppression is shown based on 10 psychophysical test sessions (5 from each animal), and are plotted as the mean and standard error of the mean (s.e.m.) over these sessions.

Figure 2

Modulation of BOLD responses during perceptual suppression in two monkeys. A. Single-session examples of target-specific activation shown on axial slices (anterior is up, posterior down; lateral is right and medial is left). Colors represent the thresholded t-score map corresponding to the statistical comparison between four repetitions of 30s target presentation and four interleaved 30s blocks of a blank screen (see Methods for the parameters used for the anatomical and functional MR scans). B. Mean BOLD responses over all sessions in the study for both monkeys. Time course computed for those voxels showing significant decreases in activity during the OFF period, when the target stimulus was physically removed (see Methods for details). Mean + s.e.m. over 8 and 16 sessions, respectively.

Figure 3

Divergence of V1 single unit activity and fMRI BOLD response during perceptual suppression. A. Single session examples of V1 BOLD responses and single neuron firing rate inside and outside the target representation during the fMRI block design experiment. The colored region on the dorsal view of one hemisphere (anterior is up, medial is right) corresponds to the region of V1 activated by the target stimulus (data shown as t scores for a representative localizer experiment - see Methods and Suppl. Fig. 1). In each of the panels, activity levels for the five different stimulus conditions are shown in the sequential block paradigm (mean of 4 repetitions). Each trace represents the continuous activity level throughout 5 minutes of alternating 60s stimulation blocks consisting of up to 9 individual trials (vertical lines indicate the beginning and end of each stimulus block). Note that inside the target representation, but not outside, the BOLD and spiking activity drops in the OFF condition, when the target is physically removed. The spiking and fMRI signals are in close correspondence except for the GFS (INV) condition inside the target region. During this period, the BOLD signal shows perceptual modulation whereas the spiking activity reflects the unchanged physical. Data from monkey CB35. Each plot is mean and s.e.m. from four cycles of testing within one session. B. Population average across both monkeys and all experiments. The left panel depicts the mean BOLD response for all 24 scan session with time collapsed from beginning to end of each 60s block condition. Each bar represents the activation level within the ROI as a function of condition. Note the drop in BOLD during the INV condition where the target stimulus became invisible. The panel on the right represents the mean firing rate of all 172 recorded targetselective neurons, expressed as %-change to baseline activity (as assessed during the FIX condition). The two control conditions in which the target stimulus remained visible (VIS_TR and VIS) resulted in statistically indistinguishable firing rates from the invisible (INV) condition (multiple t-tests, error bars represent s.e.m. between imaging sessions and neurons, accordingly).

Figure 4

Spectral analysis of local field potential signals obtained during suppression and control trials. A. Time-frequency plots during the INV, VIS and OFF condition, featuring activity changes in the suppression period (following surround onset). Each panel depicts the average spectrogram for an entire trial period, with magnitude changes relative to the 500ms period preceding the surround onset, when the target alone was present in each condition. Population data is shown over all channels over all recording sessions (successfully completed trials only). All relevant stimulus events are marked with dashed lines (T = target, S = surround). Note that while all conditions showed a drop in low frequency power following the surround onset, the drop for the INV condition during the suppression period was larger than the corresponding drop for the VIS condition, but closely resembled physical removal (OFF). B. Statistical timefrequency analysis of perceptual suppression vs. physical removal (for the time period indicated by a dashed square in A). The upper plot (comparing the INV and VIS conditions) shows a decrease during perceptual suppression that is limited to the low frequencies, while the lower plot (comparing the OFF and VIS conditions) shows a large, broadband decrease when the target is physically removed. Note that these two conditions are nearly identical perceptually. T values are indicated by grey scale values shown in inset. C. Average t-score as a function of frequency for the entire period of perceptual suppression (and physical removal, accordingly). The threshold for statistical significance (p < 0.001) is indicated by a dashed line.

Figure 5

Population average of band limited power (BLP) and spiking time courses for different experimental conditions. A. Grand average of low frequency (5–30 Hz) BLP over time as a function of experimental condition (all channels recorded during all sessions with both monkeys). All relevant events are indicated with dashed lines (T = target; S = surround). The time period of interest that we focused our analysis on (during which target is perceptually suppressed during the INV condition) is indicated by an orange bar on top of the panel. B. Grand average of high frequency (30–80 Hz) BLP, same conventions as in A. C. Spiking density function of all units recorded during the neurophysiological experiments. In all cases, data are convolved with 50 ms stdev Gaussian kernel. Error bars are s.e.m. T=target, S=surround.

Figure 6

Summary of perceptual modulation in the BOLD and each of the electrophysiological signals (as computed from the raw data of both monkeys shown in Fig. 3B and Fig. 5). The Suppression Index corresponds to the percent of signal drop during the invisible condition (INV) compared to the physical removal condition (OFF), both relative to visible (VIS).

Figure 7

Schematic illustration of main results. From top to the bottom, each line represents the state of presentation (stimulus either on or off), the reported percept of the subject (stimulus visible or invisible), as well as the various measures of neuronal activity in primary visual cortex (high or low activity). The left column represents the case where the target is physically removed from the screen (OFF condition). As shown in Fig. 2–Fig. 5, all measures of neuronal activity, including the fMRI BOLD response, show a decline in signal when the stimulus is both physically removed and perceptually disappears. The right hand column represents our finding for the case of perceptual suppression of the target stimulus caused by GFS (invisible condition, INV). Under these circumstances, the percept is dissociated from the physical stimulus in the form of perceptual suppression. Under these conditions, the spiking activity (and high frequency LFP) maintain their activity, reflecting the continually present stimulus, while the fMRI response and (to a lesser extent) the low frequency LFP reflect the perceptual disappearance.