Figures and data
Experimental paradigm.
a.) Participants took part in a total of 8 fMRI imaging runs on two consecutive days where they performed 40 vibrotactile detection tasks per run. Imaging runs were preceded by the acquisition of functional localizers on Day 1 and followed by the acquisition of structural MRI scans on Day 2. b.) Vibrotactile stimulation intensities of each imaging run were sampled around the vibrotactile detection threshold of the respective finger, determined by an interleaved staircase prior to the imaging run. Shown are the Day 1 results of an example participant. c.) Task structure of a vibrotactile detection trial: Each trial within an imaging run started with either a ring finger cue (square on the left), thumb cue (square on the right) or a non-informative cue (square in the middle of screen), followed by vibrotactile stimulation of either the ring finger (80% likely after ring finger cue; 50% likely after non-informative cue) or thumb (80% likely after thumb cue; 50% likely after non-informative cue) of the participant’s left hand. During the response period participants indicated where they had perceived the stimulation (‘Thumb’, ‘Ring finger,’ or ‘None’ if no stimulation was detected). Illustrated is an example congruent trial (ring finger cue followed by ring finger stimulation; dashed squares show possible alternatives). ‘+’: Fixation cross; ITI: Intertrial interval.
Effect of prior information on behavioral detection performance and BOLD signal.
a.) Behavioral detection accuracies are enhanced for congruent (70.2±4.81%; mean±SEM) as compared to incongruent (56.02±6.7%) and non-informative trials (61.05±6.8%; one-way repeated measures ANOVA for Experimental condition: F2,48 = 87.839, p<0.001, η2 = 0.416). b.) Group-level GLM contrasts reveal distinct clusters of voxels in contralateral S1 for vibrotactile stimulation of ring finger versus thumb. Ring finger > Thumb: peak t-value = 5.182 (MNI coordinates: 42, -24, 60; cluster of 73 voxels); Thumb > Ring finger: peak t-value = 8.451 (MNI coordinates: 54, -14, 50; cluster of 128 voxels); p<0.05, familywise error (FWE) corrected. c.) The mean BOLD signal in contralateral S1 is dampened for congruent (0.796±0.368) as compared to incongruent (3.887±0.539) and non-informative trials (2.029±0.415; mean±SEM in a.u.: arbitrary units; one-way repeated measures ANOVA for Experimental condition: F2,48 = 44.466, p<0.001, η2 = 0.252). Pairwise t-tests with post-hoc Bonferroni correction: *** p<0.001, **** p<0.0001. Single-participant-results are visualized by gray lines (N=25 throughout all Figures).
Multivariate decoding of vibrotactile ring finger versus thumb stimulation.
a.) Contralateral S1 ROI-decoding accuracies are enhanced for congruent (63.067±1.311%, p<0.001) as compared to incongruent (49.872±1.003%, p=0.4) and non-informative trials (56.51±1.419%, p<0.001; mean±SEM; p-value calculation against empirical chance values from shuffled data). Decoding accuracies for ipsilateral S1 were not significantly greater than chance (p>0.4). Repeated-measures two-way ANOVA: Interaction Experimental condition x ROI, F2,96 = 17.847, p<0.0001, η2 = 0.294. Pairwise t-tests with post-hoc Bonferroni correction: **** p<0.0001. Single-participant-results are visualized by gray lines. b.) Improvements of behavioral detection accuracies (x-axis) and ROI-based decoding accuracies (y-axis) for congruent trials are strongly correlated for contralateral S1 (Spearman’s r, **** p<0.001, robust regression slope: 0.594; shaded area with 95% CI) but not for ipsilateral S1 (Spearman’s r, p=0.575, robust regression slope: -0.272 with shaded area: 95% CI). c.) Group-level searchlight maps reveal voxels throughout S1 and adjacent M1 with significantly higher decoding accuracies for the Congruent > Incongruent (left) comparison. The Congruent > Non-informative comparison (right), revealed two distinct clusters proximal to univariate ring finger and thumb ROIs (one-sided t-tests with p<0.01 and FWE-correction at the cluster-level with p<0.01). d.) Contralateral S1 searchlight decoding accuracies assigned to a voxel (y-axis) as a function of its Euclidean distance to the peak BOLD voxel (x-axis; MNI coordinates of peak BOLD voxel obtained by a functional localizer) reveal a strong correlation for congruent trials (Spearman’s r: **** p<0.001, * p<0.05, shaded area: 95% CI around robust regression line). Results averaged across ring finger and thumb, for a detailed discussion see Materials and methods.
Somatotopic modulation of the BOLD signal during the Cue-stimulation interval (CSI).
a.) The period between finger cue display and vibrotactile stimulation (CSI: ‘Cue-stimulation interval’; Figure 1C) was categorized as ‘informative’ when following a ring finger cue (square on left of screen) or a thumb cue (square on right of screen). Non-informative cue-stimulation intervals (square in the middle of screen) served as a control and were divided into two groups according to the following vibrotactile stimulation (ring finger vs. thumb). b.) After display of a finger cue, the BOLD signal was greater within the respective ROI (ring finger vs. thumb ROI) than within the ROI associated with the non-cued finger. No such modulation of BOLD activity could be observed after display of a non-informative finger cue. Ring finger and thumb ROIs were extracted for each participant from independent functional localizers (Materials and methods). A 2×2×2 way repeated measures ANOVA with CSI type x ROI x Cue condition revealed significant differences between the BOLD signal associated with ring finger vs. thumb cue for the informative CSI (p<0.01, t>4.98) but not for the non-informative CSI (p>0.258, t>0.0508). Pairwise t-tests with post-hoc Bonferroni correction: * p<0.01, ** *p<0.0001.
Decoding of ring finger versus thumb cue during the Cue-stimulation interval (CSI).
a.) Accuracies of ring finger vs. thumb cue decoding are greater than chance for contralateral S1 during the informative but not during the non-informative CSI (57.327±2.8%, mean±SEM; p<0.0001 vs. 49.024±5.501, p=0.72). Decoding accuracies are not greater than chance for either condition for ipsilateral S1 (49.451±4.712 vs. 50.122±2.514%; p>0.62). Two-way repeated measures ANOVA: Interaction CSI type x ROI, F1,24 = 63.26, p<0.0001, η2 = 0.423. b.) Results from decoders trained on the CSI (ring finger vs. thumb cue decoding) and tested on vibrotactile ring finger vs. thumb stimulation trials and vice versa. Decoding accuracies were significantly greater than chance for the informative but not for the non-informative CSI for contralateral S1 (accuracies averaged over both train-test schemes: 56.921±3.754% vs. 50.191±1.207%, p<0.0001), but not for ipsilateral S1 (48.934±4.813%, p>0.7 vs. 49.325±3.204%, p>0.6). Two-way repeated measures ANOVA: Interaction CSI type x ROI, F1,24 = 35.385, p<0.0001, η2 = 0.308. Pairwise t-tests with post-hoc Bonferroni correction: **** p<0.0001. Single-participant-results are visualized by gray lines. c.) Improvement of behavioral detection accuracies (x-axis; from non-informative CSI to congruent stimulation trials) and ROI-based decoding accuracies (y-axis) for congruent trials are correlated for contralateral S1 (Spearman’s r, * p=0.039, robust regression slope: 0.539 with shaded area: 95% CI) but not for ipsilateral S1 (Spearman’s r, p=0.728, robust regression slope: -0.138 with shaded area: 95% CI). d.) Representational dissimilarity matrices visualizing the average cross-validated Mahalanobis distances (Walther et al., 2016) between ring finger (RF) and thumb (Th) stimulation (black labels) and ring finger and thumb cues during the CSI (green labels) for the contralateral (left) and the ipsilateral S1 (right).
Psychometric curves for the vibrotactile detection task.
Psychometric curves for vibrotactile stimulus detection (% correct) during a total of 320 trials over 8 imaging runs as a function of vibrotactile stimulation intensity (Volt). The psychophysical curve of the congruent condition is shifted to the left as compared to the psychophysical curves of the incongruent and the non-informative experimental conditions, indicative of a lower vibrotactile detection threshold for the former. Curves represent logistic fits to mean values across all participants (N=25), error bands around psychophysical curves represent the 95% bootstrap confidence intervals.
Average vibrotactile stimulation intensities for experimental condition and finger.
The ipsilateral S1 BOLD level is not modulated by prior information.
The mean BOLD signal in ipsilateral S1 is not significantly different for congruent (3.229±0.249) vs. incongruent (3.355±0.33) and non-informative trials (3.743±0.261; a.u.: arbitrary units; one-way repeated measures ANOVA for Experimental condition: F2,48 = 1.789, p=0.71). Single-participant-results are visualized by gray lines (N=25).
ROI-based decoding of vibrotactile stimulation of ring finger vs. thumb for secondary somatosensory and primary motor cortex
a. Left and center: Ipsilateral S2 ROI-decoding accuracies are not significantly different from chance level for any of the experimental conditions (Ipsilateral S2; congruent: 49.928±5.703%, incongruent: 49.726±6.392%, non-informative: 50.372±5.991% Contralateral S2; congruent: 49.527±5.382%, incongruent: 49.448±4.976%, non-informative: 48.89±5.832%, mean±SEM, p>0.5). Right: Decoding accuracies for contralateral M1 were significantly greater than chance for congruent (54.605±5.78%) and non-informative trials (52.368±6.11%), but not incongruent trials (49.368±4.928%). Repeated-measures two-way ANOVA: Interaction Experimental condition x ROI, F2,96 = 9.452, p<0.0001, η2 = 0.311. Pairwise t-tests with post-hoc Bonferroni correction: * p<0.05; **** p<0.0001. Single-participant-results are visualized by gray lines. b. Improvements of behavioral detection accuracies (x-axis) and ROI-based decoding accuracies (y-axis) are not correlated for contralateral (Spearman’s r, p=0.181, robust regression slope: 0.594 with shaded area: 95% CI) or ipsilateral M1 (Spearman’s r, p=0.71, robust regression slope: -0.272 with shaded area: 95% CI).
Peak t-values associated with vibrotactile stimulation of ring finger versus thumb.
