Science Briefs

Extinction and the Erasure of Memories

A common finding from the animal laboratory is that even when the learned behavior is eliminated, the behavior often returns with time, a result that points to an extremely important distinction between memory storage and memory expression.

By Matt Lattal, PhD

The idea that specific memories can be erased has been central to plot lines of many bad soap operas, and to both good and bad science fiction. It is an idea that never seems to get old, but years of research on memory in animals have shown that once a memory has formed, it is extremely difficult to eliminate. Because invasive memories are a component of many psychiatric disorders, behavioral and pharmacological interventions are often designed to weaken, if not eliminate the impact of those memories on behavior. A common finding from the animal laboratory, however, is that even when the learned behavior is eliminated, the behavior often returns with time, a result that points to an extremely important distinction between memory storage and memory expression. This difference between what the organism knows and what the organism tells us is perhaps most apparent in research and theory on experimental extinction.In extinction, organisms learn that a previously established relation between two events has been severed. The classic example of this comes from Pavlov's (1927) work with dogs in which he paired a neutral stimulus (the conditioned stimulus (CS), such as the ticking of a metronome) with a biologically relevant stimulus (the unconditioned stimulus (US), such as meat powder). As a result of learning this relation, conditioned responding (salivation in Pavlov's experiments) emerged to the CS. Pavlov found that when the CS was then presented without the US, responding attenuated and eventually ceased (Figure 1). The finding that conditioned responding extinguished was not very surprising or even particularly interesting. The interesting finding came from the next phase of the experiments, in which the dog was presented with the CS again following a moderate delay. Pavlov showed that the extinguished response had spontaneously recovered after a delay (Test in Figure 1). This spontaneous recovery effect and other related findings documented since - such as contextual renewal, reinstatement of responding by the US, and Pavlovian-to-instrumental transfer - are convincing demonstrations that the original learning is preserved in extinction, despite the absence of responding. In the years that have followed, extinction and spontaneous recovery have been documented in many species (e.g., worms, mice, rats, humans) and behavioral preparations (e.g., fear conditioning, drug conditioning, operant choice, spatial learning). Many theories have been offered to explain both the decremental process that occurs during extinction and the spontaneous recovery that occurs with time (reviewed in Delamater, 2004). Although much remains unknown, it is quite clear that extinction involves learning processes that suppress but do not erase the original memory.

Understanding the mechanisms of extinction and spontaneous recovery is important for developing a broad knowledge of basic learning processes. Extinction also is one of the great success stories of extrapolations of basic findings to applied settings, where it is used as part of behavioral interventions for many psychiatric disorders. Recent studies of the cellular and molecular mechanisms of memory have identified receptors and signaling molecules that may be involved in extinction, suggesting possible targets for pharmacological interventions designed to enhance extinction (see Davis, Myers, Chhatwal, & Ressler, 2006). These are exciting developments, but many questions remain unanswered.

What is enhanced in enhanced extinction effects?

Because there are many learning processes involved in extinction, there are different ways that a given behavioral or pharmacological manipulation could enhance extinction. Enhancements could occur because some aspect of the extinction memory is strengthened. For example, many theories suggest that extinction is a form of inhibitory learning. The development of this inhibition could be enhanced through a variety of mechanisms (e.g., Brandon, Vogel, & Wagner, 2003; Morris & Bouton, 2007; Rescorla, 1993). The challenge is determining where that inhibition occurs--for example, between CS and US systems or between CS and response systems--and how that inhibition is controlled--for example, by the context of extinction or by the passage of time.

Enhancements in extinction are most commonly attributed to some new, perhaps inhibitory association that develops during extinction, but extinction also could be enhanced through weakening of some aspect of the original memory. This may include weakening of the CS-US association, weakening of the CS representation, or weakening of the US representation. Most theories that appeal to a weakening of some aspect of the original memory emphasize that this is a temporary depression and not a permanent erasure. What is important to recognize is that there are a variety of ways that extinction can be enhanced, and that these processes may act alone or in combination. At a theoretical level, this means that when a manipulation enhances extinction, it may be oversimplified to say that the manipulation enhances the "extinction memory."

At the behavioral level, we can be much more concrete about what it means to enhance extinction. A manipulation that enhances extinction is likely to enhance either the rate of response loss during extinction or the persistence of this response loss across time. Effects on rate of extinction are informative because they potentially speak to how quickly the organism learns the new contingencies during extinction. This is important from a clinical perspective because with fewer treatment sessions, it is less likely that a patient will drop out of a program before reaching the clinical goal.

The persistence of extinction across time and environmental situations is of even greater clinical importance than the rate of extinction. An ideal intervention would eliminate spontaneous recovery to prevent the recurrence of the original behavior. Again, from a clinical perspective, a weakening of spontaneous recovery means that when conditioned cues are encountered long after the completion of the treatment program, the behavior does not come back.

Thus, even if we do not have a handle on the psychological process that is affected during extinction, we can assess the efficacy of an extinction treatment by examining the development and persistence of response loss. It is possible that a manipulation may enhance the rate of extinction without weakening spontaneous recovery. It also is possible that a manipulation may not affect the rate of extinction, but may weaken (or enhance) spontaneous recovery. A complete picture of the effects of a given manipulation on extinction therefore needs to include effects on both the development and persistence of extinction.

Pharmacological enhancements of extinction: Effects on rate and persistence

Research on the neurobiology of memory has discovered many receptor and second messenger systems, as well as nuclear targets, that appear to be critical for initial memory formation and consolidation. Although the study of extinction has revealed some different cellular and molecular mechanisms, there are many common pathways that appear to be involved in initial memory formation and extinction (e.g., Lattal, Radulovic, & Lukowiak, 2006). A natural strategy for enhancing extinction would therefore be to pair behavioral extinction with the delivery of a pharmacological agent that accelerates the function of these pathways, with the hope that the effects of an extinction treatment will be enhanced when coupled with the activation of appropriate cellular and molecular targets.

Such findings have been reported with D-Cycloserine (DCS), a partial NMDA receptor agonist. Many studies have demonstrated a critical role for NMDA receptors in initial learning and in extinction. Several laboratories have demonstrated that systemic administration of DCS during extinction may enhance the development of extinction and may also weaken the amount of spontaneous recovery that occurs with time (reviewed in Davis, Ressler, Rothbaum, & Richardson, 2006). Other studies show that these effects may not always be robust (Guastella, Dadds, Lovibond, Mitchell, & Richardson, 2007) and the enhanced extinction effects may not persist across different physical contexts (Woods & Bouton, 2006). Together, these DCS studies provide a good model for the type of careful, systematic analyses that are going to be needed for any demonstrations of pharmacological enhancements of extinction.

Altering the action of other receptors and intra-cellular signaling cascades has been shown in many studies to enhance the development of extinction (e.g., Berlau & McGaugh; Morris & Bouton, 2007). A recent collaborative study between my lab and Marcelo Wood's lab at the University of California, Irvine, found that fear extinction in mice can be enhanced by drugs that are thought to enhance gene transcription (Lattal, Barrett, & Wood, 2007). This is an exciting finding as we have seen this effect with two different drugs and two different injection regions (systemic or directly into the hippocampus). However, this excitement must be tempered by the caveat that we do not know how persistent these effects are across time. This is an important caveat, as other manipulations that may enhance extinction in the near-term (e.g., one day) may not enhance extinction in the long-term (e.g., many days to weeks). Such a dissociation was reported by Isiegas, Park, Kandel, Abel, & Lattal (2006), in which genetic inhibition of protein kinase A, a second messenger involved in memory, enhanced the development of extinction, but did not weaken spontaneous recovery.

Pharmacological enhancements of extinction are complicated by the many demonstrations that the behavioral expression of extinction is often tied to the context in which it occurs. These contexts can be external (such as physical cues) or internal (such as drug states). For example, Lattal (in press) found that ethanol administered prior to a fear extinction trial can decrease fear expression and enhance the development of extinction, as well as the persistence of extinction when testing occurred in the presence of ethanol. However, when testing occurred in the absence of ethanol, fear again was high, suggesting that ethanol served as an internal cue for the operation of extinction contingencies (see also Bouton, Kenney, & Rosengard, 1990; Briggs & Riccio, 2007; Cunningham, 1979). Together, all of these findings suggest that enhancements in extinction need to be examined closely. Many genetic and pharmacological manipulations appear to facilitate the development of extinction, but these effects do not always persist across time, contexts, and different internal states.

The reconsolidation hypothesis: Is it possible to erase a memory?

Because so many studies have found that the suppressive effects of extinction can be reversed, much of the discussion about enhancing extinction takes for granted that extinction manipulations will affect the rate or persistence of extinction without necessarily affecting the original memory. But are there ways to enhance extinction by actually erasing the original memory? If there were, we would not need to worry about spontaneous recovery and associated phenomena because there would be no memory to support the behavior. A great deal of recent excitement has centered on the idea that the simple act of retrieving a memory moves that memory out of permanent neural storage and into a labile state, where it remains until it is reconsolidated into a fixed state (reviewed in Tronson & Taylor, 2007). While this memory is labile, it is vulnerable to behavioral or pharmacological disruption. The broad implication of this account is that if the right treatment is given during memory retrieval, it may be possible to actually eliminate a memory.

Consistent with this reconsolidation hypothesis, many studies have demonstrated that retrieved memories are vulnerable to disruption (e.g., Bernardi, Lattal, & Berger, 2006; Misanin, Miller, & Lewis, 1968; Nader, Schafe, & LeDoux, 2000). However, just like effects on extinction, these disruptive effects are often temporary, either reversing with time or through reminder treatments (e.g., Fischer, et al., 2004; Lattal & Abel, 2004; Prado-Alcala, et al., 2006). Findings like these suggest that the original memory is at least somewhat preserved and that the loss of behavior seen soon after pharmacological disruption does not necessarily reflect the loss of memory.

It is important to note that other studies have failed to show spontaneous recovery after reconsolidation-like behavioral deficits, instead demonstrating that the behavioral impairments are long lasting. These null findings suggest that reconsolidation deficits may be permanent, but it also is abundantly clear that the absence of spontaneous recovery can be caused by many factors besides loss of memory (e.g., Lattal & Abel, 2004; Pavlov, 1927; Rescorla, 2004). For example, the use of repeated test sessions or an insufficient recovery interval will weaken spontaneous recovery, as will increasing the amount of extinction. Indeed, the difference in spontaneous recovery seen by Pavlov in Figure 1 was taken as evidence for a difference in the depth of extinction and not for a difference in the state of the original association. Clearly, the reconsolidation idea has generated excitement, but much more work is needed to determine whether it is possible to permanently erase specific memories.

Memory and the problem of behavior

One of the most perplexing issues in the behavioral and neurobiological analysis of memory is determining how learning is expressed in behavior. Years of laboratory research have shown repeatedly that the absence of behavior does not necessarily reflect the absence of a memory. Indeed, it is clear that the answers we get about memory from animals depend to a large extent on how we ask the questions (e.g., Stout & Miller, 2007). Recent research on enhanced extinction and impaired reconsolidation effects is confirming the importance of assessing behavior under multiple conditions, such as testing after different retention intervals, in different contexts, and under different drug states. Even if we can never know whether a memory has been erased, this careful approach to behavior will give us a very clear picture of the conditions in which previously formed memories appear in behavior.


Berlau, D. J. and McGaugh, J. L. (2006). Enhancement of extinction memory consolidation: the role of the noradrenergic and GABAergic systems within the basolateral amygdala. Neurobiol Learn Mem 86, 123-32.

Bernardi, R. E., Lattal, K. M. and Berger, S. P. (2006). Postretrieval propranolol disrupts a cocaine conditioned place preference. Neuroreport 17, 1443-1447.

Bouton, M. E., Kenney, F. A. and Rosengard, C. (1990). State-dependent fear extinction with two benzodiazepine tranquilizers. Behav Neurosci 104, 44-55.

Brandon, S. E., Vogel, E. H. and Wagner, A. R. (2003). Stimulus representation in SOP: I. Theoretical rationalization and some implications. Behav Processes 62, 5-25.

Briggs, J. F. and Riccio, D. C. (2007). Retrograde amnesia for extinction: similarities with amnesia for original acquisition memories. Learn Behav 35, 131-40.

Cunningham, C. L. (1979). Alcohol as a cue for extinction: State dependency produced by conditioned inhibition. Anim Learn Behav 7, 45-52.

Davis, M., Myers, K. M., Chhatwal, J. and Ressler, K. J. (2006). Pharmacological treatments that facilitate extinction of fear: relevance to psychotherapy. NeuroRx 3, 82-96.

Davis, M., Ressler, K., Rothbaum, B. O. and Richardson, R. (2006). Effects of D-cycloserine on extinction: translation from preclinical to clinical work. Biol Psychiatry 60, 369-75.

Delamater, A. R. (2004). Experimental extinction in Pavlovian conditioning: behavioural and neuroscience perspectives. Q J Exp Psychol B 57, 97-132.

Fischer, A., Sananbenesi, F., Schrick, C., Spiess, J. and Radulovic, J. (2004). Distinct roles of hippocampal de novo protein synthesis and actin rearrangement in extinction of contextual fear. J Neurosci 24, 1962-6.

Guastella, A. J., Dadds, M. R., Lovibond, P. F., Mitchell, P. and Richardson, R. (2007). A randomized controlled trial of the effect of D-cycloserine on exposure therapy for spider fear. J Psychiatr Res 41, 466-71.

Isiegas, C., Park, A., Kandel, E. R., Abel, T. and Lattal, K. M. (2006). Transgenic inhibition of neuronal protein kinase A activity facilitates fear extinction. J Neurosci 26, 12700-7.

Lattal, K. M. (in press). Effects of ethanol on the encoding, consolidation, and expression of extinction following contextual fear conditioning. Behav Neurosci.

Lattal, K. M. and Abel, T. (2004). Behavioral impairments caused by injections of the protein synthesis inhibitor anisomycin after contextual retrieval reverse with time. Proc Natl Acad Sci U S A 101, 4667-72.

Lattal, K. M., Barrett, R. M. and Wood, M. A. (2007). Systemic or intrahippocampal delivery of histone deacetylase inhibitors facilitates fear extinction. Behav Neurosci 121, 1125-31.

Lattal, K. M., Radulovic, J. and Lukowiak, K. (2006). Extinction: does it or doesn't it? The requirement of altered gene activity and new protein synthesis. Biol Psychiatry 60, 344-51.

Misanin, J. R., Miller, R. R. and Lewis, D. J. (1968). Retrograde amnesia produced by electroconvulsive shock after reactivation of a consolidated memory trace. Science 160, 554-5.

Morris, R. W. and Bouton, M. E. (2007). The effect of yohimbine on the extinction of conditioned fear: a role for context. Behav Neurosci 121, 501-14.

Nader, K., Schafe, G. E. and LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature 406, 722-726.

Pavlov, I. P. (1927). Conditioned reflexes, an investigation of the physiological activity of the cerebral cortex. London, Oxford University Press.

Prado-Alcala, R. A., Diaz Del Guante, M. A., Garin-Aguilar, M. E., Diaz-Trujillo, A., Quirarte, G. L. and McGaugh, J. L. (2006). Amygdala or hippocampus inactivation after retrieval induces temporary memory deficit. Neurobiol Learn Mem.

Rescorla, R. A. (1993). Inhibitory associations between S and R in extinction. Animal Learning & Behavior 21, 327-336.

Rescorla, R. A. (2004). Spontaneous recovery. Learn Mem 11, 501-9.
Stout, S. C. and Miller, R. R. (2007). Sometimes-competing retrieval (SOCR): a formalization of the comparator hypothesis. Psychol Rev 114, 759-83.

Tronson, N. C. and Taylor, J. R. (2007). Molecular mechanisms of memory reconsolidation. Nat Rev Neurosci 8, 262-75.

Woods, A. M. and Bouton, M. E. (2006). D-cycloserine facilitates extinction but does not eliminate renewal of the conditioned emotional response. Behav Neurosci 120, 1159-62.

About the Author

Matt Lattal received his BA in Psychology from the University of California, San Diego, in 1993 and his PhD in Psychology from the University of Pennsylvania in 1998. His graduate work focused on theories of learning with Robert A. Rescorla. His post-doctoral work focused on molecular mechanisms of learning and memory with Ted Abel in the Department of Biology at the University of Pennsylvania. He has been an Assistant Professor in the Department of Behavioral Neuroscience at Oregon Health & Science University in Portland, Oregon, since 2005.