SUPPLEMENTARY FIGURE 1 fmri responses to target and surround

Transcription

1 SUPPLEMENTARY FIGURE 1 fmri responses to target and surround 1deg anterior lateral medial posterior t score anterior lateral medial posterior Target response zone Surround response zone ge ge % Signal Chan - % Signal Chan Seconds Seconds 35 TARGET ONLY + SURROUND ONLY + FIXATION +

2 Supplementary Figure 1. Definition of the region of interest (ROI) for fmri data analysis. Example data from a localizer experiment. Each session started with a 5-15 minute scan that consisted of a pseudorandomized block design with three conditions: a blank screen baseline, a presentation of the target stimulus in isolation and a condition during which the GFS-inducing surround stimulus (a random dot movie with a gap in the target region) was shown. Upper panel: Position of the target stimulus relative to the point of fixation in the visual field and the resulting statistical parametric map for target stimulus versus blank screen presentation, superimposed on a 3D brain model derived from a structural scan of one monkey (CB35). Lower panel: Average BOLD time courses for all voxels within and outside of the ROI. While the target stimulus elicited strong deflections of the BOLD response inside the ROI, there was increased activity throughout the surround presentations outside of the ROI only.

3 SUPPLEMENTARY FIGURE Details of fmri modulation A D EXAMPLE TIME COURSE VIS/VIS TR EXAMPLE VOXEL FIX VIS TR VIS OFF INV FIX VIS TR VIS OFF INV 5 OFF INV % Change B % Ch hange C % Cha ange ALL VOXELS IN ROI E Change % FIX VIS TR VIS OFF INV FIX VIS TR VIS OFF INV FIX VIS TR VIS OFF INV 1 3 FIX VIS TR VIS OFF INV ALL SESSIONS n=4 1 3 FIX VIS TR VIS OFF INV n=4 F Change % % Ch hange SESSION AVERAGE SESSION AVERAGE FIX VIS TR VIS OFF INV n =5 n =5 scan CB35.A As1/7

4 Supplementary Figure. Detailed fmri results and pseudorandomized control experiment. A. BOLD signal change as a function of experimental condition for an example voxel during the standard block design experiment. Left panel: Average BOLD time course for a single voxel inside the ROI (mean over 4 block repetitions). Right panel: Mean and s.e.m. BOLD response for each experimental condition for 4 block repetitions (same data as in left panel, but collapsed over time). B. Average BOLD signal change for all voxels inside the ROI (see Suppl. Fig. 1) of the same scan as in A. C. Grand average BOLD response as a function of perceptual condition for both monkeys and all scanning sessions (see also Fig. ). Error bars represent the s.e.m. between sessions. D. Example time fmri response during a pseudorandomized block paradigm (monkey CB35; average of all voxels within ROI). No preprocessing other than co-registration has been applied to the data. Each block (as indicated by colored bars below) was preceded by an equally long period of blank screen presentation. Note the reliable pattern of activation evoked by each block presentation and the reduced magnitude during the OFF condition indicated in blue. E. Average time course (over all blocks) for five sessions with the pseudorandomized block design experiment. Note that even though the data is plotted continuously, each block was actually preceded by a block of fixation on a blank screen of similar length. F. Same data as in E, but averaged over block (+- sem across sessions).

5 SUPPLEMENTARY FIGURE 3 Block and trial analysis A Power (µv) LF LFP POWER (5-3Hz) HF LFP POWER (3-9Hz) SPIKING 8 n = 4 n = Power (µv) Imp./sec 1 n = B n = 4 n = 4 n = 7 %-change 4 %-change 1 Imp./sec FIX VIS TR VIS OFF INV 1 FIX VIS TR VIS OFF INV FIX VIS TR VIS OFF INV C T on S on T off S off n = change %

6 Supplementary Figure 3. Detailed neurophysiology results and explanation of analysis A. Time course for an entire experimental run for each of the different neuronal signals. Leftmost panels: Band-pass filtered LFP power in the low frequency range as a function of time. Plots represent the average of all 4 simultaneously recorded contacts (sampling all layers of V1) of one multicontact electrode during one example session. Activity traces represent the time course of the entire experiment, consisting of 4 block repetitions of the structure shown in Fig. 1A. Center panels: Same data representation as in leftmost channels, but for the band-pass filtered LFP power in the high frequency range as a function of time. Rightmost panels: Average spiking activity of seven simultaneously measured units (across different laminae) during one example session. B. Mean activity after averaging all four block repetitions (compare to the BOLD response in Fig.,3 and Suppl. Fig. ). Each time period corresponding to one of the experimental conditions is delineated with dashed lines. Note that all three signals exhibit a drop during the OFF condition, but only the lower frequency LFP power is similarly low during the GFS condition. Lower row: Single trial data. Each block consists of up to 1 repetitions (depending on the animal's performance) of the single trial structure outlined in Fig.1A. Neuronal data was analyzed on a trial-by-trial basis in order to temporally align (trigger) the data and compute visually evoked potentials. Each of the main events is indicated with a dashed line and an arrow (T = target; S = surround stimulus).

7 SUPPLEMENTARY FIGURE 4 Perceptual report A LFP Power (5-3 Hz) B LFP Power (3-9 Hz) LFP POWER (µv) Suppression NO Suppression n=95 n= Time following surround onset (ms) Time following surround onset (ms) monkey reports invisible monkey reports visible MODULAT TION INDEX (t-sc core) p <.1-4 p <.1 n=95 n= Time following surround onset (ms) Time following surround onset (ms)

8 Supplementary Figure 4. LFP power as a function of target stimulus visibility (perceptual report paradigm). In this testing condition the physical stimulus was identical but the probability of target disappearance for each trial was adjusted to be roughly.5. A. Upper panel: Low frequency power (5-3Hz) following onset of the surround stimulus. The average power in this frequency band was lower during trials where the monkey reported the target stimulus to disappear (orange) than in trials where the monkey reported to see the stimulus (black). Lower pane; T-score as a function of time for the difference in low frequency LFP power between trials of target visibility and perceptual suppression. The dashed line indicates a significance at the level of p <.1. B. High frequency (3-9Hz) LFP power and t-score for the same trials as in A. (all data was collected in 4 sessions with monkey CB35). Thick orange bar corresponds to period in which perceptual suppression is expected.

9 SUPPLEMENTARY FIGURE 5 Absolute power A B LF LFP HF LFP POWER SPECTRAL DE ENSITY 1-6 Rest Task u.) LF FP POWER (a n=1594 VIS OFF INV FREQUENCY (Hz) FREQUENCY (Hz)

10 Supplementary Figure 5. A. Power Spectral Density of V1 local field potentials during the task and a similarly long period where the monkey is resting in a dark room (one session, monkey CB35). Abscissa is frequency in Hz and ordinate is power units on a log scale. Note the difference in high frequency power between the active state and resting state that coincides with the power bands chosen for our LFP analysis (indicated in blue and orange). B. LFP power during two seconds following surround onset as a function of experimental condition. The OFF condition revealed the largest drop in power during this time period. During perceptual suppression (INV), power decreases were restricted to the low frequencies (see Fig. 4 for the significance of these power differences, as well as Fig. 5 for the complete temporal profile).

11 SUPPLEMENTARY FIGURE 6 Eye Position A THROUGHOUT MR SCANS: Monkey CB35 THROUGHOUT MR SCANS: Monkey 98X9 % TIME n = 9 sessions % TIME n = 16 sessions HORIZONTAL DISPLACEMENT (dva) HORIZONTAL DISPLACEMENT (dva) % TIME % TIME 15 1 n = 9 n = 16 sessions 5 sessions VERTICAL DISPLACEMENT (dva) VERTICAL DISPLACEMENT (dva) B THROUGHOUT RECORDINGS: Monkey CB35 % TIME 6 4 n = 17 sessions HORIZONTAL DISPLACEMENT (dva) THROUGHOUT RECORDINGS: Monkey 98X9 % TIME n = 14 sessions HORIZONTAL DISPLACEMENT (dva) % TIM ME 6 4 VERTICAL DISPLACEMENT (dva) n = 17 sessions % TIME n = 14 sessions VERTICAL DISPLACEMENT (dva)

12 Supplementary Figure 6. Comparison of eye positions during electrophysiology and fmri sessions. Each histogram represents time spent at a certain location as a function of visual space ( being the point of fixation). Data was computed for the entire time of the experiment (note that this included a substantial amount of time during which the monkeys were not even supposed to fixate) and pooled across all sessions. A. Monkeys' eye displacement throughout the fmri sessions. Each column depicts data for one of the two monkeys we tested, with horizontal displacement on top. B. Monkeys' eye movements throughout the neurophysiological test sessions. Same data representation as in A. dva = degrees visual angle.

13 SUPPLEMENTARY FIGURE 7 Refined ROI Analysis mean time course of innermost voxels ( monkeys, 4 sessions) mean time course of outermost voxels ( monkeys, 4 sessions) FIX VIS TR VIS OFF INV FIX VIS TR VIS OFF INV Percent BOLD Signal Ch hange Seconds Percent BOLD Signal Ch hange Seconds Percent BO OLD Signal Cha nge 1-1 FIX VIS TR VIS OFF INV Percent BO OLD Signal Cha nge 1-1 FIX VIS TR VIS OFF INV

14 Supplementary Figure 7. Sub-region analysis of the ROI. Plotted here is the conditionspecific BOLD response for both monkeys (see Fig. B, 3A for comparison, as well as Suppl. Fig. 3B for further explanation). This analysis was designed to rule out the possibility of contamination by lateral spread of (ie. negative) BOLD responses induced by the onset of the surround stimulus. Specifically, we re-analyzed our data averaging the mean BOLD response of 3 randomly chosen voxels at the center of the ROI, and comparing the result to the mean response of 3 randomly chosen voxels at the outer margin of the ROI. There was no significant difference between the datasets (multiple t-tests). The result shows that the grand average shown in Fig. B is representative of the activity profile throughout the entire ROI.

15 SUPPLEMENTARY FIGURE 8 Refined Eye Movement Analysis Monkey CB35 Monkey 98X9 % TIME n = 14 sessions 1 5 n = 17 sessions HORIZONTAL DISPLACEMENT (dva) % TIME VERTICAL DISPLACEMENT (dva)

16 Supplementary Figure 8. Refined eye movement analysis. In order to rule out that any residual eye movements were affecting the neuronal responses, we re-analyzed the eye movement data collected from both monkeys during the neurophysiology sessions. Plotted here are the eye displacements during the stimulus/surround presentation epoch. Note that this time period corresponds to that used for the neurophysiology analysis, whereas the histograms plotted in Suppl. Fig. 6 are representing the (entire minute) time period that went into the block design/fmri analysis.

17 SUPPLEMENTARY FIGURE 9 Effect of Eye Movements on Neuronal Responses A Ti Trials with smaller eye movements ( monkeys, all sessions) Power (µv) 3 5 Low LFP (5-3 Hz) High LFP (3-9Hz) SPIKING n=1594 n=1594 n=17 Power (µv) Time [s] Time [s] Time [s] ike Rate (imp/s sec) Sp B VIS OFF INV VIS TR Trials with larger eye movements (monkeys, all sessions) Low LFP (5-3 Hz) High LFP (3-9Hz) SPIKING Power (µv) 3 5 n=1594 n=1594 n=17 Power (µv) Time [s] Time [s] Time [s] Spike Rate (imp/sec c) 18 14

18 Supplementary Figure 9. Direct measure of the effect of residual fixation eye movements on neuronal responses. We divided our neurophysiological dataset in two sets with equalized trial numbers. A. The first set consisted of all trials in which the standard deviation of the monkeys' horizontal eye displacement during stimulus presentation (see Suppl. Fig. 8) did not exceed the median standard deviation of all trials combined ("trials with smaller eye movements"). We then computed the grand average of each neuronal signals and plotted it as a function of time and conditions (see Fig. 5 for direct comparison). B. The second dataset consisted of the rest of the trials ("trials with larger eye movements). We applied the same analysis as in A. and plot the result as a function of time and experimental conditions. Despite the loss of statistical power, each of the split datasets showed the same main effects as the whole dataset combined (ie. Fig. 5) and there was no qualitative difference between the time courses of each signal between the two eye movement dependent datasets.

19 Supplementary Methods Visual Stimulation. All stimuli were generated using OpenGL-based custom written software (ESS/STIM, copyright Dr. D. Sheinberg, Brown University). Stimuli were generated on a PC (Kontron, Munich, Germany) equipped with a Quadro FX 3 graphics board (NVIDIA, Santa Clara, CA) using custom written software on Microsoft Windows XP (Microsoft Co., Redmond, WA) and projected on a screen via a MR-compatible TFT projector mounted outside the scanner (Avotec, Stuart, FL). The image projected on the screen 7 cm away from the mirror covered 19 x 14 cm. During neurophysiological recordings, stimuli were presented on either a single 18 TFT monitor placed in front of the animals (NEC MultiSync LCD 186NX with a 14x768 resolution) providing a field of view of 47.3 x 37.5 degree, or two 7 TFT monitors (XGen MV71, 14x768 resolution) mounted to the side of the recording booth and a mirror stereoscope with a field of view of 87.7 x 49.7 degree. For MRI experiments, visual stimuli were projected onto a screen mounted to the top of the animals chair, perpendicular to their line of sight. Dichoptic stimulation inside the scanner was achieved by fitting commercially available (red/green) anaglyph glasses (Rainbow Symphony Inc., Reseda, CA) to the monkeys faces. Ambient light in testing rooms was minimized throughout all experiments in order to achieve strong visual stimulation and reduce distraction to the animals. We divided the recording sessions into sessions with a similar kind of binocular stimulation as used inside the scanner (7 out of 41), and sessions where we used a Wheatstone-type mirror stereoscope to achieve separate stimulation of the two eyes. No systematic differences were found in the neuronal data between those sessions. Eye Tracking. The animal s eye movements inside the scanner were monitored and recorded using an infrared light sensitive MR-compatible camera (MRC Systems, Heidelberg, Germany) in combination with optical eye tracker software (SMI, Berlin Germany). The camera was mounted together with two DC-powered infrared light emitting diodes behind two layers of IR-permissible half-mirrors (Edmund Optics, Barrington, NJ) in front of the monkeys eyes, providing a high contrast, high resolution image of one of their eyes.

20 We divided the neuronal recording sessions into sessions with similar image-based eye tracking (Eye Link II), and sessions where we monitored the monkey s gaze using implanted scleral search coils (DNI, Newark, DE). No difference was found between the datasets of these sessions. During all MR scans (both the localizer scans and the main experiment), the monkeys had to fixate a central spot on the screen in a 1-4 dva window for up to 1 seconds in order to receive reward. During initial training and physiology sessions in which a scleral search coil was used, the monkey was required to hold fixation in a window of.8 dva radius. See Suppl. Figs. 6-9 for all eye movement data.