The present results include behavioral (17 sessions, in a total of 33 runs) and/or neurophysiological data (30 runs) from 17 subjects (9/17=52.94% female; with 28.9±7.28 years old; min: 18 years old, max: 40 years old). The behavioral performance in the task was 76.55%±14.32 and the response latency was 1.045±0.266 seconds. Response latency was not correlated to performance (Rs=0.1277, P=0.5091, Q=0.8265). In the 17 sessions (each subject was schedule to perform two runs), a total of 33 runs were performed. Two subjects presented a performance below 50% in the first run and their analysis was not included in the results presented, except when specifically indicated. In four subjects, only data from one run is presented. Three of these runs were not analyzed due to the low quality of the recorded signals (N=3). The fourth run was from a subject who performed below 50% and, in addition, which had to be removed from the study due to excessive stress generated by the task. For the fourteen subjects that completed both runs, and overall improvement was found between the first and the second run (first run: 69.81±14.31% correct; second run: 85.36+-9.5%; Paired samples t test: t=5.315; df=13, P=0.0001) (Fig. 1 d).

Analysis of power in delta (0.5-4.5Hz), theta (4.5- 8.5Hz), alpha (8.5-13.5Hz), beta (13.5-30.5Hz), and low gamma (30.0-45Hz) bands indicated that the periods of discrimination (1500ms) (i.e., stimulus delivery) and response (1500ms) (i.e., choosing and pressing one button) were characterized by fundamentally different patterns of activity throughout the network of electrodes recorded (Fig. 1 d). Table 1 presents the statistical results for the comparison between the discrimination and response periods for each frequency band in each electrode. An overall reduction in power was present, frequently in more than one frequency band. The exception to this were electrodes F3, C4, and TP10; where no differences were found between the two periods analyzed for any frequency band. For channels with significant changes between the two periods, the theta frequency band was the most frequently affected (10/13=76.9% channels), followed by alpha (6/13=46.2%) and beta frequency bands (5/13=38.5%). Lastly, delta (4/13=30.8%) and low gamma bands (3/13=23.1%) were less often affected. These results indicated that the discrimination and the response periods were associated, with a bilateral asymmetrical distribution of changes in power, occurring most commonly in theta, but also in other frequency bands in a network of electrodes recording from frontal, temporal, parietal, and occipital regions.

To test our second hypothesis, we compared the power in the alpha frequency band in the C3 electrode with tactile discrimination performance.* *No significant correlation was found between the* *performance and the power in the alpha frequency band for the discrimination period. (Rs=0.007383, P=0.7035, Q=0.8338, n.s.).

To analyze changes occurring simultaneously in multiple frequency bands across the scalp, an analysis of state maps composed by ratios of frequency bands was performed [25] (Fig. 2 a-h). It has been previously shown that this analysis accurately captures global cortical dynamics and/or associated behaviors when multiple frequencies are involved [25,27]. As depicted in Fig. 2 panel a; ratio 1 (used to analyze changes occurring in higher frequency bands), was calculated as (0.5-20Hz)/(0.5-45Hz) and ratio2 (used to analyze changes occurring in lower frequency bands) was calculated as (0.5-4.5Hz)/(0.5-9Hz). In Fig. 2 panel b, an example of the state maps for channel Fp2 during the discrimination (Dis) and response (Resp) periods is presented. The arrow tail corresponds to the ratio1 and ratio 2 coordinates during the discrimination period, while the arrowhead corresponds to the ratio 1 and ratio 2 coordinates during the response period. The displacement of each vector reflects the neural dynamics occurring as the subject moved from a state associated with the discrimination to the neural state corresponding to response.

Analysis of these ratios revealed that the discrimination and response periods were characterized by different ratios of frequencies. In other words, the dynamics of change between the two behaviors corresponded to an overall increase in ratio 1 and ratio 2 throughout a network involving multiple electrodes (Table 2). Namely, significant differences were found for ratio 1 in all electrodes with the exception for Fp1, Fz, C4, Pz, and Tp10 (with Fp1, C4 and Tp10 presenting non-significant P values below 0.10). Meanwhile, for ratio 2, significant differences were found for all electrodes, except for Fp1, F3, Cz, C4, O2, and Tp10 (with Fp1 and Cz presenting P values below 0.10). A visual summary of these findings is depicted in Fig. 2 C, where the network of electrodes associated with significant differences in either or both ratios during the discrimination and response periods is presented.

Having determined that the discrimination and response periods were characterized by different ratios of higher and lower frequencies, we then analyzed if these differences could be relevant for the width discrimination task. More specifically, we asked if the variability observed in ratio1 and ratio 2 values during the discrimination period reflected the tactile performance in the task (note, for example, the large variability in the arrow tails depicted in Fig. 2 b). For this, ratio 1 and ratio 2 values were compared to the performances in each session using Spearman’s Rho correlation (i.e., the value of the session was compared to the overall session ratio 1 and ratio 2 values). As presented in Fig. 2 (d), ratio 1 (left panel) and ratio 2 (right panel), significantly correlated to subjects’ performance in a bilateral asymmetrical network across the scalp. For ratio 1, significant Spearman’s Rho correlations were present in electrodes Fp2 (Rs=0.5545, P=0.0018, Q=0.0192), F4 (Rs=0.5285, P=0.0023, Q=0.02048); T4 ( Rs=0.5007, P=0.0057, Q=0.026057), P3 (Rs=0.5796, P=0.001, Q=0.032), P4 (Rs=0.5647, P=0.0014, Q=0.0224), and O2 (Rs=0.489, P=0.0071, Q=0.0284). Meanwhile, for ratio 2 significant Spearman’s Rho correlations were present for electrodes P3 (Rs=0.5416, P=0.0024, Q=0.0192) and P4 (Rs=0.5094, P=0.0048, Q=0.0256). These results indicated that the state maps composed by ratios of higher and lower frequencies allowed not only describing the dynamics associated with the two different periods of the task, but also, that these ratios reflected the overall performance in the task.

To determine if these correlations would still be present when task difficulty was changed, three subjects (that had already been previously tested in this task) performed one session with two runs in an easier version of the tactile task while EEG recordings were performed. For this the difference between the Wide and the Narrow stimuli was increased by 4.0 mm. Under such conditions the behavioral performance was of 96.67±4.08%. Examples of the state maps associated with these runs are shown in Fig. 2 panels g-h (ratio 1 Fp2 and ratio 2 P3, respectively) where the red filled circles represent these additional runs. Comparison of the ratios of frequencies and the performance, when these six runs were pooled with the remaining sessions and analyzed, still generated significant correlations for channels Fp2 (Fig. 2 panel g), F4 and P4 in ratio 1 (Fp2: Rs=0.4671, P=0.0047, Q=0.0376; F4: Rs=0.4422, P=0.0073, Q=0.0292; P4: Rs=0.4232, P=0.0113; Q=0.0226), but not for T4, P3, and O2 (T4: Rs=0.1202, P=0.4915, Q=0.4915, n.s.; P3: Rs=0.1711, P=0.3257, Q=0.3722, n.s.; O2: Rs=0.3139, P=0.0663; Q=0.0884, n.s.). Meanwhile, a significant correlation between the performance and ratio 2 in electrode P3 was maintained (P3: Rs=0.4457, P=0.0073, Q=0.0292), but not for electrode P4 (P4: Rs=0.3334, P=0.00504, Q=0.08064, n.s.). These results suggested that changing the degree of difficulty did not disrupt the overall correlation between neural activity and performance for channels Fp2, F4, P3 and P4.

As we have observed an improvement in performance between the first and the second runs that took place in each session (Fig. 1 c), and as we have observed significant correlations between the ratios of higher and lower frequencies in an asymmetrical network involving frontal, temporal, parietal, and occipital electrodes, we then asked if the dynamics of ratios in this network would reflect the dynamics of improvements in performance occurring between the first and the second run for the same subject. In other words, we asked to which extent these neurophysiological measures could reflect within-subjects differences in behavior. For this, we reanalyzed the data from the subsample of subjects that completed both runs (N=14 subjects in 28 runs; the subject that presented a performance below 50% in the first run and then improved in the second run was included in this analysis). First, we calculated the changes occurring in ratio 1 and ratio 2 in the same subject. In Fig. 3 a and b, the changes occurring in ratio 1 and ratio 2 of electrodes P4 (panel a) and O2 (panel b), between the first and the second runs of the session are presented. Then, we analyzed both ratios simultaneously, to determine if a clear pattern could be identified. In Fig. 3 panels c and d, both coordinates are arranged to form the arrow tail (Ratio 1: X-axis, and Ratio 2: Y-axis in the first run) and the arrowhead (Ratio 1: X-axis, and Ratio 2: Y-axis in the second run) of a vector. While no clear pattern could be observed in the P4 electrode (Fig.3 c), an overall shift towards the right lower quadrant could be identified in O2 (Fig.3 d). Although this suggested that changes occurring in ratios could be related to the differences between the first and the second run, no clear pattern could be identified. Then, the difference between neural activity in the two runs was calculated and plotted against the difference in performance. This allowed comparing the physiological and behavioral evolution throughout the session. In Fig. 3, panels e and f, show the difference in ratio 1for the two runs simultaneously with the difference in performance. Even though no clear correlation could be observed between these two variables, visual inspection suggested that an overall increase in the variance of ratio 1 (i.e., increase in the variance of values in the coordinates axis) was present, as the difference in performance increased (i.e., abscissa values increased). To be able to explain this variation, we then tested if the symmetrical pattern observed in O2 (Fig.3 f) converted to its absolute values (which now reflected “a change in performance” instead of “an improvement in performance”) could explain the large variability in the data. As presented in Fig. 3 panels g and h, a significant correlation was found between the dynamics of ratio 1 for channels P4 (P4 Ratio 1: Rs=0.7209, P=0.0036, Q=0.0252; note that this result is also significant if the same analysis is performed without subject 14) and O2 (O2 Ratio 1: Rs=0.6452, P=0.0127, Q=0.04445), but not for Fp2 (Ratio 1: Rs= 0.1424, P=0.6273, Q=0.73185, n.s.), F4 (F4 Ratio 1: Rs=-0.02225, P=0.9398, Q=0.9398, n.s.), T4 (T4 Ratio 2: Rs=0.3471, P=0.2241, Q=0.5229, n.s.), P3 (P3 Ratio 2: Rs= -0.1891, P=0.5173, n.s.) or P4 (P4 Ratio 2: Rs=0.2937, P=0.3082, Q=53935, n.s.).

To further test if these dynamics were independent of the level of difficulty in the task, the six runs from the sessions of the small sample of three subjects presented in Fig. 2 g and h, was combined with the remaining subjects and the pooled data was reanalyzed solely for channels P4 and O2. As presented in Fig. 3, panels g and h, the inclusion of these additional runs still presented a significant correlation for P4 (Rs All=0.6395, P=0.0057) and for O2 (Rs All=0.5541, P=0.0210). These results indicated that the dynamics of ratio 1 for electrodes P4 and O2 encoded the changes in performance in the behavioral task between the two different runs of the same session.