Increased brain signal variability accompanies lower behavioral variability in development

Abstract

As the brain matures, its responses become optimized. Behavioral measures show this through improved accuracy and decreased trial-to-trial variability. The question remains whether the supporting brain dynamics show a similar decrease in variability. We examined the relation between variability in single trial evoked electrical activity of the brain (measured with EEG) and performance of a face memory task in children (8-15 y) and young adults (20-33 y). Behaviorally, children showed slower, more variable response times (RT), and less accurate recognition than adults. However, brain signal variability increased with age, and showed strong negative correlations with intrasubject RT variability and positive correlations with accuracy. Thus, maturation appears to lead to a brain with greater functional variability, which is indicative of enhanced neural complexity. This variability may reflect a broader repertoire of metastable brain states and more fluid transitions among them that enable optimum responses. Our results suggest that the moment-to-moment variability in brain activity may be a critical index of the cognitive capacity of the brain.

Figure 1. Behavior and EEG data by age.

(A) Behavioral results for mean RT, coefficient of variation (cvRT) and accuracy (percent correct responses). Error bars indicate group standard errors. B) Waveforms for group average ERPs across all electrodes, together with corresponding P100 latencies (marked by vertical dashed line) and P100 scalp maps. C) Group average results for spectral power distribution (SPD) during baseline interval for electrode O2. Error bars indicate group standard errors. Similar pattern was present at all channels. With maturation, decreases were observed in lower frequencies (<10 Hz) combined with increase in higher frequencies (>10 Hz).

Figure 2. Within-subject brain variability measures across age groups.

A) shows group mean results across the scalp for pre- and post-stimulus PCA dimensionality estimate of trial-to-trial variability (top two rows) and MSE area under the curve (bottom row). Scalp maps of group means were obtained by interpolating values from single channel group mean values. Gradual increase in all three brain variability measures across age groups is evident. PLS statistical analysis of each measure detected a pattern of linear increase, which was significant (p<<0 for all three measures) and was stably expressed (absolute bootstrap ratio >3.5) for most channels. B) shows group means for MSE estimates across temporal scales for channel O2, together with corresponding standard errors. Similar entropy curves were obtained for all channels and showed maturation-related increase in entropy at all scales. Given consistent age-related differences at all time scales, the area under the MSE curve was taken as a summary measure of maturational changes in entropy (e.g., (A), bottom row).

Figure 3. Within-subject brain variability in relation to behavior and chronological age.

A) Correlations are arranged in a table where rows represent three within-subject brain variability measures (pre- and post- stimulus PCA and MSE) and columns represent behavioral measures (cvRT, accuracy and meanRT) and age. Each entry in the table shows a scalp map resulting from interpolated values from single channel correlations between given brain variability measure and given behavior or age. Unstable correlations (where the 95% CI included 0) were set to 0. Colormap corresponds to [−0.7 0.7] range of correlation values. Along with each scalp map of correlations, there is an inset showing a scatter plot representing subject measures for a single channel (electrode O2). Subjects are grouped by color according to age group membership. Estimated value of correlation ( r ) along with the associated 95% CI is given on top. B) Results of the statistical analysis of the observed correlations. For each brain measure, the partial least squares (PLS ) analysis detected one significant pattern of correlations (p = 0). The bar graph plots the global correlation of brain variability and each behavior measure or age (+/− bootstrap estimated standard error). As can be seen, for all three brain measures PLS detected similar global patterns of simultaneous negative correlation with cvRT, positive correlation with accuracy and age, and no stable correlation with mean RT. For each of the three within-subject brain variability measures, the bootstrap analysis of PLS confirmed that the correlation patterns were robustly expressed across most channels (data not shown).

Figure 4. Relationship between within-subject brain variability and behavior after adjusting for age differences across individuals (age was regressed out from all measures).

Color scale and legends are the same as for Figure 3. A) Correlations are arranged in a table where rows represent three within-subject brain variability measures (pre- and post- stimulus PCA and MSE) and columns represent three behavioral measures (cvRT, accuracy and mean RT). Each entry in the table shows a scalp map resulting from interpolated values from single channel correlations between given brain variability measure and given behavior or age. Unstable correlations (where the 95% CI included 0) were set to 0. Colormap corresponds to [−0.7 0.7] range of correlation values. Along with each scalp map of correlations, there is an inset showing a scatter plot representing subject measures for a single channel (electrode O2). Subjects are grouped by color according to age group membership. Estimated value of correlation (r) along with the associated 95% CI is given on top. B) Statistical analysis of brain-behavior correlations adjusted for age. PLS detected one significant pattern of correlations for pre- and post-stimulus PCA dimensionality estimation (p = 0.003, and p = 0.011), and one pattern for MSE that did not exceed conventional statistical thresholds (p = 0.131). Bootstrap analysis of the two significant patterns for PCA showed, however, that the patterns were unstable for most channels (data not shown).