Skip to content

When you choose to publish with PLOS, your research makes an impact. Make your work accessible to all, without restrictions, and accelerate scientific discovery with options like preprints and published peer review that make your work more Open.

PLOS BLOGS The Official PLOS Blog

How Does the Brain Learn which Memory Retrieval Strategies are Most Effective? By Joe Moran

By Joe Moran

The profoundly unfortunate side effects of the operation to remove HM’s hippocampi in 1953 laid bare that we humans have multiple memory systems. After his operation, HM couldn’t learn new facts, recall the people he had seen, or the places he had been. Through this work, we learned that the hippocampus serves us by laying down episodic memories—memories for events that have a time and a place. This ability is part of declarative memory; we are able to say out loud the information these memories store.

By contrast, we use a different memory system to learn skills, known as procedural knowledge. This memory system is mediated by the nigrostriatal brain pathway, which encompasses parts of the midbrain and the ventral striatum, and mostly expresses the neurotransmitter dopamine. After HM’s operation, he learned to navigate a simple maze, among other skills, demonstrating that his procedural memory system was intact. Conversely, patients with Huntington’s disease who have a damaged nigrostriatal pathway have difficulty learning these skills, but no problems with declarative memory. This ‘double dissociation’ reveals that the declarative and procedural memory systems are independent, but gives us no clues as to how these memory systems might interact. In particular, several regions in the frontal lobes are involved in strategic retrieval, which uses control processes to guide how we search for information in declarative memory. Just how are these strategies acquired, and is there an influencing role of the mute nigrostriatal procedural learning system in their expression?

Recent research from David Badre’s lab at Brown University aimed to answer this question. The team used an elegant design that provided a strategy with which participants could organize their retrieval efforts (see Figure 1). Participants performed a simple lexical decision task (LDT; “Is this letter string a word?”), with a twist. Participants saw a visual object ‘prime’ (e.g., a car) that had been associated via a rule with a different semantic category (e.g., animals). Once the prime had disappeared, a short or long delay preceded the probe letter string. 75% of the time, the rule was true, and the target word was indeed an animal. 25% of the time, the rule was false, and the target word was in fact a car. So the researchers crossed how long participants had available to formulate a retrieval plan (“I saw a car, so that means the letter string is probably an animal name”), with whether or not the resulting probe matched the primed category. The authors found that guided rule-based retrieval can make us faster to identify a word from a primed category, rather than a word from the category usually associated with that item, but only when participants had long enough to implement the rule. So far, so good. What is really interesting about these findings however, is how the brain implemented the rule.

Figure 1. Schematic of the lexical decision task (LDT).
Figure 1. Schematic of the lexical decision task (LDT).

When the authors compared trials with successful versus unsuccessful strategy use, they found activation in a number of regions, including the ventral striatum (see Figure 2). The ventral striatum was more active when target letter strings conformed to the arbitrary stimulus association rule than when they did not. This suggests that the ventral striatum is serving as a behavioral reinforcer, with its higher activation on expected trials signaling that the strategy was used effectively, and that the cognitive control systems should continue to implement the strategy in future trials. Two further analyses narrowed down the contribution of the ventral striatum in this task.

Figure 2. Location of the ventral striatum.
Figure 2. Location of the ventral striatum.

 The authors modeled individual participant responses across the study using the drift diffusion model framework and compared these responses to activity in the ventral striatum. The model framework independently estimates a participant’s drift rate towards a word/non-word decision, and the amount of their non-decision time. The drift rate is the strength of retrieved evidence that the probe letter string is a word, and the non-decision time allows for processes like word encoding. This analysis showed that ventral striatal activation was related to how much easier the decision was when participants followed the retrieval rule, and was not related to basic word encoding.

The second follow-up analysis divided the experiment into two halves, and found that participants’ drift rates for expected trials increased from the first to the second half of the experimental session; relying on the rule through the course of the experiment accelerated participants’ response times. Most interestingly, those participants who learned the rule best—and thus whose performance was facilitated most—were most likely to be those participants who had greater striatal activation during the first half of the experiment when the rule was being learned. Thus, the ventral striatum served as a learning signal to apply the rule. Its activation was associated not with global LDT facilitation, but rather with the degree to which evidence accumulated in favor of the primed category, an effect only observable if participants reliably followed the decision rule.

This paper begins uncovering how our fantastically adaptive memory system hits upon the correct retrieval strategy to flexibly solve the payoff between obtaining a useful piece of information from memory, and the amount of time required to operate on our memory stores to get the information. Such a process requires the flexible use of retrieval rules, so that (a) we do not get bogged down long enough to be eaten while waiting for our memory retrieval machinery to give us a course of action, and that (b) the actual outcomes of that rapid process are indeed relevant and useful. Badre and colleagues here have demonstrated that this adaptive process of choosing amongst possible retrieval solutions – the guidance of memory retrieval by cognitive control – is at least partly under control of the supposedly independent procedural memory system.

There are a couple of study limitations that should be considered. First, the small sample size of 19 participants, endemic in cognitive neuroscience, limits the study’s ability to find anything but large effects. Smaller, but reliable effects are likely to have been obscured, while the effects observed need to be replicated in larger samples to ensure their reliability. Second, the authors provided a rule to guide retrieval, but this is more like an explicit strategy game than real life. An open question is whether these findings generalize to situations where participants must unearth the association between the prime and the target for themselves—is this learning process also under striatal control? Until such data are in, we can’t be sure whether the mechanisms revealed here generalize.

Overall, this is an excellent addition to the literature, for at least three reasons. First, it reveals one way in which the cortical mechanisms involved in controlled, declarative memory retrieval are guided and influenced by the subcortical mechanisms responsible for associative, procedural learning. Second, it highlights how the interaction between these memory systems complicates the dichotomy of explicit vs. implicit, or declarative vs. nondeclarative, memory, to show that in intact brains, these systems are in fact in regular cooperation. And third, this work goes beyond the standard approach in memory research to provide a clever demonstration of the dopaminergic system’s role in how we acquire and use memory retrieval strategies. Such papers form a growing corpus of brain imaging investigations that can inform our understanding at both the psychological and neurological levels, a feat that repeated observers have argued may be impossible.

All views expressed in this post are those of the author, and not necessarily of PLOS.


Badre, D., Lebrecht, S., Pagliaccio, D., Long, N. M., and Scimeca, J. M. (2014). Ventral striatum and the evaluation of memory retrieval strategies. Journal of Cognitive Neuroscience, 26, 1928-1948.

Dobbins, I. G., Foley, H., Schacter, D. L., & Wagner, A. D. (2002). Executive control during episodic retrieval: multiple prefrontal processes subserve source memory. Neuron35(5), 989-996.

Gabrieli, J. D., Fleischman, D. A., Keane, M. M., Reminger, S. L., & Morrell, F. (1995). Double dissociation between memory systems underlying explicit and implicit memory in the human brain. Psychological Science6(2), 76-82.

Martone, M., Butters, N., Payne, M., Becker, J. T., & Sax, D. S. (1984). Dissociations between skill learning and verbal recognition in amnesia and dementia. Archives of Neurology41(9), 965.

Milner, B., Corkin, S., & Teuber, H. L. (1968). Further analysis of the hippocampal amnesic syndrome: 14-year follow-up study of HM.Neuropsychologia6(3), 215-234.

Ratcliff, R., Gomez, P., & McKoon, G. (2004). A diffusion model account of the lexical decision task. Psychological review111(1), 159.

Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of neurology, neurosurgery, and psychiatry20(1), 11.

Schultz, W., Tremblay, L., & Hollerman, J. R. (2000). Reward processing in primate orbitofrontal cortex and basal ganglia. Cerebral Cortex10(3), 272-283.

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving and W. Donaldson (Eds.), Organization of Memory (pp. 381-402). New York: Academic Press.

Moran_hdShot3Joe Moran is a cognitive scientist with the US Army Natick Soldier Research, Development, and Engineering Center, where he studies decision making, creativity, and social influences. @jtneuro

Back to top