Neural representations of the individual items and their global interaction.
Figure 1B and 1C showed the rERP time courses for the two levels of information, respectively, the individual stimuli and the global interaction, in an antero-posterior gradient from prefrontal to occipital electrodes. To quantify the overall activity at each time point, we calculated the global field power (GFP) of the rERPs across all valid electrodes [40]. The GFP exhibited three peaks after the circle array onset in both predictors. These peaks were obvious for the individual items (Peak 1: 0.076 s; Peak 2: 0.142 s; Peak 3: 0.288 s) and their global interaction (Peak 1: 0.086 s; Peak 2: 0.144 s; Peak 3: 0.3 s). Therefore, the brain generated independent neural responses to the individual items as well as to their global interaction.
To compare the rERPs of the individual items and their global interaction, we restricted the analysis to six electrodes (P5, P7, PO7, P6, P8, and PO8) from the occipital region. Figure 1D and 1E showed the averaged rERPs across the six electrodes to the individual items and their global interactions, respectively. We then extracted the response latencies of the rERPs in the selected electrodes to the individual items and their global interaction. The response latencies for the two predictors were extracted from the first rERP peaks (time range: 0.06–0.12 s) in each participant (Fig. 1F). The paired samples t-test showed that the latency of the global interaction (Mean ± SD: 0.088 ± 0.011 s) was significantly later than that of the individual items (Mean ± SD: 0.078 ± 0.008 s; t(464) = -25.11, P < 0.001, Cohen’s d = -1.17), confirming that the global interaction was processed at a higher level of the information processing hierarchy than the individual items [43].
The response to the global interaction predicts ensemble perception.
We first examined whether the neural response to individual items and the global interaction were related to the ensemble perception. A cumulative Gaussian function was fitted to the proportion of responses for ‘‘probe is larger’’ against the ratio of the probe size to the mean size. The sigma value of the fitted function indicated the perceptual sensitivity to the difference between the probe size and the mean size. For the rERP response to individual items and global interaction, we extracted its 3 peak amplitudes (averaged over 10 ms around the peak time) at the 6 selected electrodes in the occipital area. Pearson correlations were calculated between the fitted sigma value and peak rERP values for individual items and the global interaction. As revealed at the top of Fig. 2, we found that the three response peaks to the individual items were not significantly correlated with the perceptual sensitivity (Ps > 0.125). However, the perceptual sensitivity was significantly and positively correlated with the first peak (bottom left of Fig. 2; r = 0.20, p < 0.001) but not the other two peaks of the global interaction (Ps > 0.072). Therefore, the results validated that ensemble perception is reflected in the early neural response to the global interaction, which replicated the findings in our previous study [18].
The correlation between ADHD symptoms and the responses of the individual items.
We then examined whether the amplitude of the rERP responses to individual items correlated with the ADHD symptoms. We extracted the 3 peak amplitudes as above and calculated the correlation between the amplitude of each peak with the total score and the score for each subcomponent of ADHD, respectively. As shown in the left column of Fig. 3, the amplitude of the first peak was not correlated with any of the ADHD symptom scores (Ps > 0.423). In contrast, as shown in the middle column of Fig. 3, the amplitude of the second rERP peak (latency: 0.142 s) was significantly and positively correlated with the ADHD total score (r = 0.34, P < 0.001) and the ADHD subcomponent score of inattention (r = 0.43, P < 0.001) and hyperactivity/impulsivity (r = 0.10, P = 0.036). Specifically, larger peak amplitudes (more negative rERP values) were accompanied by lower ADHD total score, lower inattention score, and lower hyperactivity/impulsivity score. Critically, when comparing with hyperactivity/impulsivity, the significance level of the correlation between the second rERP peak and inattention was significantly larger (z = 6.15, P < 0.001, one-tailed), suggesting that the second peak mainly reflected the inattention symptom in ADHD. Finally, as shown in the right column of Fig. 3, the amplitude of the third peak (latency: 0.288 s) did not correlate with any of the measured ADHD symptoms (Ps > 0.165). These results demonstrated that subjects with a higher level of ADHD symptoms, inattention, in particular, showed reduced neural processing of the individual visual stimuli.
The correlation between ADHD symptoms and the responses of the global interaction.
Next, we examined whether the amplitude of rERP responses to global interaction correlated with the level of ADHD symptoms. We used the same procedure to extract the peak values of the rERP response to global interaction and calculated correlations of the 3 peak amplitudes with ADHD total score, inattention, and hyperactivity/impulsivity, respectively. As shown in the left column of Fig. 4, the amplitude of the first peak was significantly and positively correlated with the score of ADHD (r = 0.26, P < 0.001) and inattention (r = 0.36, P < 0.001), but not with hyperactivity/impulsivity (r = 0.03, P = 0.483). Larger peak amplitudes (negative values) are accompanied by lower ADHD total scores and lower inattention levels. Furthermore, the amplitude of the first peak was significantly more correlated with the inattention symptom than with the hyperactivity/impulsivity symptom (z = 5.91, P < 0.001, one-tailed). However, as shown in the middle and right columns of Fig. 4, the amplitudes of the second and third peaks did not correlate with any of the ADHD symptoms (Ps > 0.119). These results directly demonstrated that participants with higher levels of ADHD symptoms showed reduced neural processing of the global information in ensemble perception.
The response latencies of the rERPs do not predict ADHD symptoms.
Finally, we examined whether the latencies of the first peaks (shown in Fig. 1F) in response to individual items and the global interaction were related to the levels of ADHD symptoms in adults. As revealed in Fig. 5, we found that the response latency to individual items was not correlated with any of the measured ADHD symptoms (Ps > 0.385). The same pattern was observed for the correlation between the response latency to global interaction and the levels of the measured ADHD symptoms (Ps > 0.281). Therefore, the neural response time for the individual items and global interaction does not predict ADHD symptom levels.