Behavioral evidence for pattern separation in human episodic memory

Ewa Zotow, James A. Bisby, Neil BurgessLearning 2020 · DOI

TL;DR

This paper presents a novel behavioral paradigm to investigate "pattern separation," a proposed mechanism in the hippocampus that makes similar memories more distinct. They found that while recalling elements within the same event is highly dependent (pattern completion), recalling elements from similar, overlapping events shows less dependency than expected, suggesting these memories are actively separated to reduce interference.

What Problem Does This Paper Address?

Episodic memory allows us to recall specific events with all their associated details. A key question in cognitive neuroscience is how the brain manages to store and retrieve these complex memories, especially when they are similar to each other. The hippocampus is thought to play a crucial role, with two proposed mechanisms: "pattern completion," which allows us to retrieve a whole memory from a partial cue, and "pattern separation," which makes similar memories more distinct during encoding to prevent interference. While pattern completion is well-supported by behavioral and neural evidence, direct behavioral evidence for pattern separation in humans, particularly for multi-element episodic memories, has been elusive. This paper aims to provide such evidence by developing a task that can disentangle these two processes.

Methods

The researchers developed a novel experimental paradigm involving two phases: study and test.

  • Study Phase: Participants were presented with "events," each composed of three elements (e.g., a person, a location, an object). These events were either unique ("nonoverlapping") or shared one element with another event ("overlapping"). Participants were instructed to vividly imagine these events.
  • Test Phase: Participants were tested on their ability to recall associations between elements. They used a "dependency measure" to quantify how the success of retrieving one association from an event related to the success of retrieving other associations. This measure was compared against two models:
  • Independent Model: Assumes retrieval of associations is completely independent.
  • Dependent Model: Accounts for general performance levels and potential shared factors influencing retrieval.

The core idea is that:

  • Pattern Completion within a single event should lead to high dependency between element retrievals.
  • Pattern Separation between similar (overlapping) events should lead to lowered dependency (or "anti-dependency") between element retrievals from those events, compared to unrelated events.

They conducted two experiments, with Experiment 1 focusing on behavioral measures and Experiment 2 replicating the behavioral findings while also preparing for fMRI data collection (though the fMRI results are not reported here).

Datasets

Not discussed in the paper. The stimuli were custom-generated for the experiment, consisting of lists of famous people, locations, and objects.

Key Insights and Ideas

  • Dependency as a Window into Memory Processes: The paper ingeniously uses a statistical measure of "dependency" between retrieval successes to infer underlying cognitive processes. High dependency suggests a unified retrieval process (pattern completion), while low or negative dependency suggests separation.
  • Pattern Completion is Robust: As expected, retrieving elements from the same event showed high dependency, confirming that multi-element episodic memories are indeed retrieved holistically, consistent with pattern completion.
  • Pattern Separation is Real (and Behavioral): The most significant finding is that retrieval success between overlapping events (events sharing a common element) was less dependent than between unrelated events. This "anti-dependency" is the key behavioral signature of pattern separation, suggesting these similar memories were actively made more distinct during encoding.
  • Specificity of Pattern Separation: The reduced dependency was specific to matched overlapping pairs (where the shared element was explicitly linked). Unrelated events, even if they were of the "overlapping type" (meaning they could have overlapped with other events in the study), did not show this reduced dependency when paired randomly. This highlights that pattern separation acts on specific similar memories, not just general event types.

Key Results

  • Within-Event Dependency: Retrievals of associations from within the same event were highly dependent, consistent with pattern completion. This was observed in both experiments.
  • Across-Event Dependency (Unrelated Events): Retrievals between unrelated events were independent, as predicted by the independent model.
  • Across-Event Dependency (Overlapping Events): Retrievals between overlapping events showed lower dependency than predicted by the independent model. In fact, it was "negatively dependent" (anti-dependent), meaning success in retrieving an element from one event was associated with less success in retrieving an element from its overlapping counterpart, compared to what would be expected by chance. This effect was statistically significant in both experiments.
  • Experiment 1: Accuracy was around 66.4%. Dependency analysis showed significant differences between data and models, with within-event dependency being higher than predicted by the independent model. Crucially, across-event dependency for matched overlapping events was lower than for unrelated events.
  • Experiment 2: Accuracy was lower at 54.85%, but the pattern of dependency results replicated Experiment 1.

Limitations and Future Work

  • Interpretation of "Anti-dependency": While the authors argue for pattern separation, they acknowledge that retrieval-induced forgetting could be an alternative explanation for reduced dependency. However, post-hoc analyses on specific trial types did not fully support this, and the pattern separation explanation remains stronger.
  • fMRI Data Not Reported: Experiment 2 was designed to collect fMRI data, but these results are not presented in this paper, leaving the neural correlates of this behavioral finding to be explored elsewhere.
  • Generalizability: The study used specific types of stimuli (famous people, locations, objects) and a particular experimental design. Future work could explore the generalizability of these findings to other types of episodic memories and contexts.
  • Future Directions: The authors suggest their paradigm offers a more direct behavioral measure of pattern separation than previous tasks, allowing for further investigation into its role in memory and its potential disruption in various clinical conditions. They also highlight the complementary roles of pattern separation and completion in supporting episodic memory.

Summary updated

Abstract

An essential feature of episodic memory is the ability to recall the multiple elements relating to one event from the multitude of elements relating to other, potentially similar events. Hippocampal pattern separation is thought to play a fundamental role in this process, by orthogonalizing the representations of overlapping events during encoding, to reduce interference between them during the process of pattern completion by which one or other is recalled. We introduce a new paradigm to test the hypothesis that similar memories, but not unrelated memories, are actively separated at encoding. Participants memorized events which were either unique or shared a common element with another event (paired "overlapping" events). We used a measure of dependency, originally devised to measure pattern completion, to quantify how much the probability of successfully retrieving associations from one event depends on successful retrieval of associations from the same event, an unrelated event or the overlapping event. In two experiments, we saw that within event retrievals were highly dependent, indicating pattern completion; retrievals from unrelated events were independent; and retrievals from overlapping events were antidependent (i.e., less than independent), indicating pattern separation. This suggests that representations of similar (overlapping) memories are actively separated, resulting in lowered dependency of retrieval performance between them, as would be predicted by the pattern separation account.
BibTeX
@inproceedings{zotow2020behavioral,
  author = {Ewa Zotow and James A. Bisby and Neil Burgess},
  title = {{Behavioral evidence for pattern separation in human episodic memory}},
  booktitle = {Learning},
  year = {2020},
  doi = {10.1101/lm.051821.120},
}
Figure