Neural measures of subsequent memory reflect endogenous variability in cognitive function

Abstract

Human cognition exhibits a striking degree of variability: Sometimes we rapidly forge new associations whereas at other times new information simply does not stick. Correlations between neural activity during encoding and subsequent retrieval performance have implicated such "subsequent memory effects" (SMEs) as important for understanding the neural basis of memory formation. Uncontrolled variability in external factors that also predict memory performance, however, confounds the interpretation of these effects. By controlling for a comprehensive set of external variables, we investigated the extent to which neural correlates of successful memory encoding reflect variability in endogenous brain states. We show that external variables that reliably predict memory performance have relatively small effects on electroencephalographic (EEG) correlates of successful memory encoding. Instead, the brain activity that is diagnostic of successful encoding primarily reflects fluctuations in endogenous neural activity. These findings link neural activity during learning to endogenous states that drive variability in human cognition. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Figure 1.

(a) Illustration of an individual trial in our experiment consisting of a study list followed by a distractor task, and a free recall test. There were 24 of these trials in each experimental session and each study list consisted of 24 items. See methods for details. (b) z-transformed power around the presentation of study words during the beginning and end of one participant’s (ID: 374) 4th study list in the 16th experimental session. The study words are indicated at the top of each sub-panel with bold italic font indicating subsequent recall. (c) Average power for subsequently unrecalled (left) and subsequently recalled (right) words during study across all lists from all participants (we averaged all data within participants and calculated the shown t-values across participants). All of our analyses were based on neural activity between 0.3 and 1.6 s following study word onset (indicated with vertical black lines) and the average power across this time interval is also illustrated. For this visualization, we aggregated EEG activity across 28 superior electrodes (see methods for details).

Figure 2.

Mean probability of recall as a function of serial position across all participants (top row) and associated neural activity (averaged between 0.3 and 1.6 s after the onset of study items) for all, subsequently recalled, and subsequently not-recalled trials respectively (we averaged all data within participants and calculated the shown t-values across participants). Error bars indicate 95% confidence intervals. For this visualization, we aggregated EEG activity across 28 superior electrodes (see methods for details).

Figure 3.

(a) Distribution of uncorrected item-level SMEs (“item”) across all participants and of corresponding corrected SMEs accounting for all factors or all but the indicated factor respectively (a ¬ prefix signifies that the indicated factor was omitted). Overlaid boxplots indicate the quartiles of the distribution with a notch showing the bootstrapped 95% CI around the median. Whiskers extend to 1.5× the inter-quartile range. (b) Mean correlations between power at different frequencies (aggregated across 28 superior electrodes) and the respective (residuals of) item-level recall performance across all participants (lined up with the corresponding SMEs in Panel a). The black horizontal lines indicate zero. Error regions indicate 95% CIs.

Figure 4.

(a) Distribution of uncorrected list-level SMEs (“list”) across all participants and of corresponding corrected SMEs accounting for all factors or only the indicated ones (here “list #” refers to the joint effects of both list number and average recallability of words in each list). Boxplots are as in Figure 3. (b) Mean correlations between power at different frequencies (aggregated across 28 superior electrodes) and the respective (residuals of) list-level recall performance across all participants (lined up with the corresponding SMEs in Panel a). The black horizontal lines indicate zero. Error regions indicate 95% CIs.

Figure 5.

Distribution of uncorrected list-level SMEs (“list”) across all participants for synthesized lists made up from randomly selected items within a session (see methods for details) and of corresponding corrected SMEs accounting for all factors or only the indicated ones (here “list #” refers to the joint effects of both list number and average recallability of words in each list). Boxplots are as in Figures 3 and 4.