Theta rhythmicity governs human behavior and hippocampal signals during memory-dependent tasks

The hippocampus plays a crucial role in encoding and retrieving information, a process believed to be orchestrated by theta rhythm oscillations. This study investigates whether theta oscillations influence behavioral responses during memory tasks, leading to rhythmic patterns in behavior. Memory formation involves cortical processing sending information to the hippocampus, creating an associative index. Retrieval, conversely, involves cues triggering the reinstatement of memory in associated cortical regions. Both encoding and retrieval are linked to oscillatory patterns in hippocampal local field potentials (LFPs). While rodent LFPs show 4-8 Hz theta oscillations, human intracranial recordings show a broader low-frequency band, often peaking between 1 and 5 Hz during memory tasks. Existing research shows that encoding of later-remembered items involves higher theta power than later-forgotten items, and phase-amplitude coupling between theta and gamma oscillations enhances successful encoding. During retrieval, theta power increases in cortical areas involved in reinstatement, and synchronization between these areas and the hippocampus increases at theta frequencies. Recall signals in the hippocampus precede cortical reinstatement by about one theta cycle, suggesting communication within theta "windows." Theoretical work suggests that locking to opposing theta phases minimizes interference between new and reactivated information. Synaptic strengthening is more likely near the theta trough, while synaptic depression is more pronounced at the peak. Rodent studies indicate that new information transmission from cortex to hippocampus primarily occurs around the theta trough, while retrieval-related hippocampal spiking activity is observed around the peak. Intracranial recordings from epilepsy patients show similar dynamics, with entorhinal cortex and hippocampus theta phase synchronization during encoding, and hippocampus locking to the subiculum during retrieval. Optogenetic suppression of task-irrelevant theta oscillation phases improves performance, demonstrating the theta phase's functional link to memory. The consistent locking of encoding and retrieval to the theta rhythm predicts rhythmic behavior, potentially observable in behavioral markers dependent on long-term memory. This study tests for such rhythmicity in memory tasks by analyzing button presses indicating the timing of associative memory encoding and recall. The hypothesis is that the theta rhythm's presumed clocking of neural memory processes results in observable oscillatory behavioral modulation.

Previous research has established a strong link between hippocampal theta oscillations and memory processes. Studies in rodents have shown that theta rhythms influence the timing and efficacy of synaptic plasticity, with long-term potentiation (LTP) favored during the trough of the theta cycle and long-term depression (LTD) during the peak. These findings suggest a functional role for theta oscillations in regulating the flow of information within the hippocampal network and between the hippocampus and neocortex. In humans, studies using electroencephalography (EEG) and intracranial EEG have demonstrated that theta oscillations are modulated during memory encoding and retrieval. Higher theta power during encoding is associated with better subsequent memory performance. Furthermore, phase-amplitude coupling between theta and faster frequency bands, such as gamma, has been shown to be crucial for successful memory encoding. The phase of theta oscillations also seems to play a role, with studies suggesting that different phases might be associated with encoding and retrieval processes. However, the direct link between these neural oscillations and overt behavioral manifestations during memory tasks has not been thoroughly explored. While research on attentional scanning suggests oscillatory activity can manifest in behavioral performance, reflecting periodic switches in attention, there has been a lack of research specifically focusing on memory-dependent behavioral oscillations. This study addresses this gap by directly investigating the presence and characteristics of behavioral oscillations during memory tasks.

This study employed a multi-faceted approach involving behavioral experiments, EEG recordings, and intracranial EEG (iEEG) recordings from epilepsy patients. The behavioral experiments involved 226 healthy participants performing associative memory tasks. These tasks consisted of encoding, distractor, and retrieval phases. During encoding, participants associated cues (verbs or images) with objects, pressing a button to mark the moment of association. During retrieval, cues were presented, and participants indicated the moment they recalled the associated object. Two groups of participants had slightly different retrieval phases: Group 1 pressed a button upon recall, followed by answering catch questions; Group 2 answered catch questions before the cue, pressing a button upon being able to answer. A control group of 95 participants performed visual tasks using the same stimuli and questions, but without the memory component. The button press timings provided data to analyze for oscillatory patterns. An Oscillation score (O-score) method was used to identify and quantify oscillations in the behavioral response times (RTs). This involved outlier removal, autocorrelation histogram (ACH) computation, smoothing, Fourier transform, and identification of the dominant frequency. To account for potential biases due to response structure and limited data points, a trend curve was fit to the RTs, and 500 randomized datasets were created. The O-score was then Z-transformed against these random O-scores to assess significance. The phase-locking of correct versus incorrect trials was analyzed by establishing a continuous reference trace representing the oscillation, performing Hilbert transform to determine instantaneous phases, and comparing phase distributions using V-tests and permutation tests. Intracranial EEG recordings were obtained from 10 epilepsy patients performing a similar memory task. Hippocampal LFPs were analyzed using wavelet transform, pairwise phase consistency (PPC) to quantify phase similarity across trials, and cluster-based permutation tests to assess statistically significant changes in PPC from baseline. Finally, phase differences between encoding and retrieval were assessed at peak PPC and in event-related potentials (ERPs).

The study found significant oscillations in button presses during encoding and retrieval phases of the memory tasks, primarily in the slow theta frequency range (1-5 Hz). O-scores were significantly higher for memory-dependent task phases compared to memory-independent phases (linear mixed-effects model: coefficient = 0.28, 95% CI: 0.19–0.36; t(556) = 6.55; p < 0.001). The proportion of participants with significant O-scores was substantially higher for memory-dependent tasks (74.7% for Encoding, 69.6% for Retrieval, 76.4% for Catch-with-retrieval) than for memory-independent tasks (64.3% for Catch-after-retrieval and 37.9% for Visual). Incorrect trials did not show locking to the oscillation of correct trials. Intracranial EEG recordings showed increased phase locking of hippocampal LFPs in the slow theta range during encoding and retrieval for correct, but not incorrect, trials. The PPC increase was significantly stronger for correct than incorrect trials. Furthermore, encoding and retrieval trials showed maximal phase alignment at opposite phases of the theta rhythm, supporting theoretical suggestions of encoding and retrieval occurring at different phases. This phase opposition was demonstrated both at peak PPC and in ERPs.

The findings demonstrate a clear link between hippocampal theta oscillations and behavioral responses during memory tasks. The rhythmic modulation of button presses during memory encoding and retrieval suggests that theta oscillations act as a "clock" for these cognitive processes. The absence of oscillations in memory-independent tasks and the lack of phase-locking in incorrect trials highlight the specificity of this rhythmic influence to successful memory formation and retrieval. The iEEG data provide strong neurophysiological support for the behavioral observations, showing increased phase consistency in hippocampal LFPs during successful memory processes. The phase opposition between encoding and retrieval aligns with theoretical models proposing that separating these processes in theta phases minimizes interference. These results contribute significantly to our understanding of the neural mechanisms underlying human memory, demonstrating that theta-rhythmic hippocampal activity is not only a neural correlate but also a determinant of behavior. Future research should explore the specific cortical areas involved in this hippocampal-cortical interaction and the precise mechanisms that translate hippocampal theta oscillations into rhythmic behavioral responses.

This study provides compelling evidence that hippocampal theta oscillations influence both neural activity and overt behavior during human memory tasks. The rhythmic patterns observed in behavioral responses during memory encoding and retrieval, coupled with the increased phase locking in hippocampal LFPs, highlight the importance of theta rhythmicity in memory processing. The findings confirm theoretical predictions regarding the functional role of theta phase in separating encoding and retrieval processes. Further research can build on these findings to unravel the precise mechanisms that transform hippocampal theta activity into rhythmic behavioral patterns and to investigate the wider implications of phase-coding in cognition.

The study relies on subjective button presses to mark memory events, which may be influenced by factors beyond the targeted memory processes. The sample size for intracranial EEG recordings was relatively small, limiting the generalizability of these findings. While the O-score method is robust, its ability to detect oscillations in sparse data may be limited, potentially underestimating the true extent of rhythmic activity. The study primarily focused on associative memory; whether these findings generalize to other types of memory remains to be explored.