An official website of the United States government
Many daily activities and psychophysical experiments involve keeping multiple items in working memory. When items take continuous values (e.g., orientation, contrast, length, loudness) they must be stored in a continuous structure of appropriate dimensions. We investigate how this structure is represented in neural circuits by training recurrent networks to report two previously shown stimulus orientations. We find the activity manifold for the two orientations resembles a Clifford torus. Although a Clifford and standard torus (the surface of a donut) are topologically equivalent, they have important functional differences. A Clifford torus treats the two orientations equally and keeps them in orthogonal subspaces, as demanded by the task, whereas a standard torus does not. We find and characterize the connectivity patterns that support the Clifford torus. Moreover, in addition to attractors that store information via persistent activity, our networks also use a dynamic code where units change their tuning to prevent new sensory input from overwriting the previously stored one. We argue that such dynamic codes are generally required whenever multiple inputs enter a memory system via shared connections. Finally, we apply our framework to a human psychophysics experiment in which subjects reported two remembered orientations. By varying the training conditions of the RNNs, we test and support the hypothesis that human behavior is a product of both neural noise and reliance on the more stable and behaviorally relevant memory of the ordinal relationship between the two orientations. This suggests that suitable inductive biases in RNNs are important for uncovering how the human brain implements working memory. Together, these results offer an understanding of the neural computations underlying a class of visual decoding tasks, bridging the scales from human behavior to synaptic connectivity.
more »
« less
Learning efficient representations of environmental priors in working memory
Experience shapes our expectations and helps us learn the structure of the environment. Inference models render such learning as a gradual refinement of the observer’s estimate of the environmental prior. For instance, when retaining an estimate of an object’s features in working memory, learned priors may bias the estimate in the direction of common feature values. Humans display such biases when retaining color estimates on short time intervals. We propose that these systematic biases emerge from modulation of synaptic connectivity in a neural circuit based on the experienced stimulus history, shaping the persistent and collective neural activity that encodes the stimulus estimate. Resulting neural activity attractors are aligned to common stimulus values. Using recently published human response data from a delayed-estimation task in which stimuli (colors) were drawn from a heterogeneous distribution that did not necessarily correspond with reported population biases, we confirm that most subjects’ response distributions are better described by experience-dependent learning models than by models with fixed biases. This work suggests systematic limitations in working memory reflect efficient representations of inferred environmental structure, providing new insights into how humans integrate environmental knowledge into their cognitive strategies.
more »
« less
- Award ID(s):
- 2207700
- PAR ID:
- 10545506
- Editor(s):
- Serre, Thomas
- Publisher / Repository:
- PLoS Computational Biology
- Date Published:
- Journal Name:
- PLOS Computational Biology
- Volume:
- 19
- Issue:
- 11
- ISSN:
- 1553-7358
- Page Range / eLocation ID:
- e1011622
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Abstract Almost all models of visual working memory—the cognitive system that holds visual information in an active state—assume it has a fixed capacity: Some models propose a limit of three to four objects, where others propose there is a fixed pool of resources for each basic visual feature. Recent findings, however, suggest that memory performance is improved for real-world objects. What supports these increases in capacity? Here, we test whether the meaningfulness of a stimulus alone influences working memory capacity while controlling for visual complexity and directly assessing the active component of working memory using EEG. Participants remembered ambiguous stimuli that could either be perceived as a face or as meaningless shapes. Participants had higher performance and increased neural delay activity when the memory display consisted of more meaningful stimuli. Critically, by asking participants whether they perceived the stimuli as a face or not, we also show that these increases in visual working memory capacity and recruitment of additional neural resources are because of the subjective perception of the stimulus and thus cannot be driven by physical properties of the stimulus. Broadly, this suggests that the capacity for active storage in visual working memory is not fixed but that more meaningful stimuli recruit additional working memory resources, allowing them to be better remembered.more » « less
-
Abstract Persistent delay-period activity in prefrontal cortex (PFC) has long been regarded as a neural signature of working memory (WM). Electrophysiological investigations in macaque PFC have provided much insight into WM mechanisms; however, a barrier to understanding is the fact that a portion of PFC lies buried within the principal sulcus in this species and is inaccessible for laminar electrophysiology or optical imaging. The relatively lissencephalic cortex of the New World common marmoset (Callithrix jacchus) circumvents such limitations. It remains unknown, however, whether marmoset PFC neurons exhibit persistent activity. Here, we addressed this gap by conducting wireless electrophysiological recordings in PFC of marmosets performing a delayed-match-to-location task on a home cage-based touchscreen system. As in macaques, marmoset PFC neurons exhibited sample-, delay-, and response-related activity that was directionally tuned and linked to correct task performance. Models constructed from population activity consistently and accurately predicted stimulus location throughout the delay period, supporting a framework of delay activity in which mnemonic representations are relatively stable in time. Taken together, our findings support the existence of common neural mechanisms underlying WM performance in PFC of macaques and marmosets and thus validate the marmoset as a suitable model animal for investigating the microcircuitry underlying WM.more » « less
-
Abstract In reinforcement learning (RL) experiments, participants learn to make rewarding choices in response to different stimuli; RL models use outcomes to estimate stimulus–response values that change incrementally. RL models consider any response type indiscriminately, ranging from more concretely defined motor choices (pressing a key with the index finger), to more general choices that can be executed in a number of ways (selecting dinner at the restaurant). However, does the learning process vary as a function of the choice type? In Experiment 1, we show that it does: Participants were slower and less accurate in learning correct choices of a general format compared with learning more concrete motor actions. Using computational modeling, we show that two mechanisms contribute to this. First, there was evidence of irrelevant credit assignment: The values of motor actions interfered with the values of other choice dimensions, resulting in more incorrect choices when the correct response was not defined by a single motor action; second, information integration for relevant general choices was slower. In Experiment 2, we replicated and further extended the findings from Experiment 1 by showing that slowed learning was attributable to weaker working memory use, rather than slowed RL. In both experiments, we ruled out the explanation that the difference in performance between two condition types was driven by difficulty/different levels of complexity. We conclude that defining a more abstract choice space used by multiple learning systems for credit assignment recruits executive resources, limiting how much such processes then contribute to fast learning.more » « less
-
van den Berg, Ronald (Ed.)Categorical judgments can systematically bias the perceptual interpretation of stimulus features. However, it remained unclear whether categorical judgments directly modify working memory representations or, alternatively, generate these biases via an inference process down-stream from working memory. To address this question we ran two novel psychophysical experiments in which human subjects had to reverse their categorical judgments about a stimulus feature, if incorrect, before providing an estimate of the feature. If categorical judgments indeed directly altered sensory representations in working memory, subjects’ estimates should reflect some aspects of their initial (incorrect) categorical judgment in those trials. We found no traces of the initial categorical judgment. Rather, subjects seemed to be able to flexibly switch their categorical judgment if needed and use the correct corresponding categorical prior to properly perform feature inference. A cross-validated model comparison also revealed that feedback may lead to selective memory recall such that only memory samples that are consistent with the categorical judgment are accepted for the inference process. Our results suggest that categorical judgments do not modify sensory information in working memory but rather act as top-down expectations in the subsequent sensory recall and inference process.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1371/journal.pcbi.1011622
