Abstract
Neuronal processes underlying the formation of new associations in the human brain are not yet well understood. Here human participants, implanted with depth electrodes in the brain, learned arbitrary associations between images presented in an ordered, predictable sequence. During learning we recorded from medial temporal lobe (MTL) neurons that responded to at least one of the pictures in the sequence (the preferred stimulus). We report that as a result of learning, single MTL neurons show asymmetric shifts in activity and start firing earlier in the sequence in anticipation of their preferred stimulus. These effects appear relatively early in learning, after only 11 exposures to the stimulus sequence. The anticipatory neuronal responses emerge while the subjects became faster in reporting the next item in the sequence. These results demonstrate flexible representations that could support learning of new associations between stimuli in a sequence, in single neurons in the human MTL.
Introduction
The ability to predict future events based on past experience is a fundamental aspect of human cognition. The Medial Temporal Lobe (MTL), including the hippocampus and surrounding cortical areas, is known to play a pivotal role in associative memory. However, observing the cellular-level changes that occur during the formation of these associations in humans is challenging. This study utilizes rare single-unit recordings from epilepsy patients to investigate how individual neurons adapt as participants learn to predict the next item in a visual sequence.
Methods
The study involved eight patients with drug-resistant epilepsy, implanted with depth electrodes for clinical purposes. Researchers first identified "preferred stimuli" for individual neurons—images that elicited a strong, selective firing response. Participants then viewed these images in a fixed, predictable sequence repeated 60 times. To monitor learning, 20% of trials were "test trials" where participants had to identify the upcoming image in the sequence.
Results
Behavioral learning was rapid, with participants achieving high accuracy in predicting the next image by the 8th repetition. This was accompanied by a significant decrease in reaction times as the sequence became more familiar.
Electrophysiological analysis focused on 42 MTL neurons that showed selectivity for images in the sequence. As learning progressed, these neurons exhibited an "asymmetric shift" in their firing patterns. Instead of merely reacting to the preferred stimulus, they began firing in anticipation, starting approximately 297 ms before the image appeared. This anticipatory activity became statistically significant after 11 exposures. Notably, the strength of this neuronal anticipation correlated with the speed of the participants' behavioral responses ($r^2 = 0.18$).
Discussion
These results provide direct evidence of predictive coding in individual human MTL neurons. The rapid emergence of anticipatory firing suggests high levels of plasticity in the MTL during the initial stages of associative learning. The finding that these responses are context-dependent—disappearing when images are shown in random order—highlights the role of the MTL in representing sequential relationships. This mechanism of "firing into the future" likely supports our ability to navigate and predict a structured environment.
While the findings are robust, the study is limited by a small sample size inherent to invasive human electrophysiology. Additionally, the transient nature of the observed neuronal shifts suggests that the MTL may primarily facilitate the initial acquisition of associations, with long-term storage potentially occurring in other cortical regions.
Summary updated