A new PNAS paper
casts doubt on an influential theory of memory. The reconsolidation hypothesis holds that when a memory is recalled, its molecular trace in the brain becomes plastic, meaning that the memory has to be consolidated or ‘saved’ all over again in order for it to persist. In other words, remembering makes a memory vulnerable to being modified or erased. Reconsolidation has generated lots of research interest and even speculation that blocking reconsolidation could be used as a tool to ‘wipe’ memories. Much of the evidence for reconsolidation comes from experiments on animals, but there's also some human evidence - in particular, a much-cited 2003 Nature article by Walker et al. However, in the new paper, UCL researchers Tom Hardwicke and colleagues say that they were unable to replicate Walker et al.'s landmark findings. Here's the key data in Hardwicke et al.'s paper:
Volunteers were trained on a simple task: pressing some keys in a particular sequence as fast as possible, over and over again. The graph above shows that participants gradually got better at performing the sequence during the 'training' phase of the experiment. On the second day, participants were 'reminded' of the sequence by typing it out a few times, and were then trained to perform a different pattern. According to reconsolidation theory, learning this second pattern should have disrupted the memory of the first one, acting as 'interference' to block reconsolidation of the first sequence. This is what Walker et al. found: they showed that on the third, 'test' day, the participants' ability to perform the first sequence was impaired. However, Hardwicke et al. couldn't confirm this. They ran the paradigm seven times, with various modifications, but in no case did they observe the predicted effect. The authors conclude that "the considerable theoretical weight attributed to the original study [Walker et al.] is unwarranted" i.e. that theories of memory based on the experimental results of Walker et al., may need a rethink. Hardwicke et al. don't address the question of why Walker et al. got positive results and they didn't, however. In my view, this is another piece of evidence that challenges the reconsolidation model. Last year I blogged about an experiment in rats that cast doubt on much of the animal evidence for reconsolidation. Now the human side of the equation is also in trouble. I spoke to Tom Hardwicke, who says he was surprised by the results because he and his colleagues had originally expected to replicate Walker et al.'s data. I asked him if, in his opinion, there is any good evidence for reconsolidation. He said
Not replicating a single experiment does not necessarily signal the death of a theory of course, but one might say a significant crack has appeared in its foundations. [Walker et al.] was one of the first and most convincing demonstrations of reconsolidation in humans, and critically one of the few studies that actually observed performance impairments (consistent with much of the non-human animal literature). There are numerous other human reconsolidation studies, but the observed effects are often qualitatively different and/or the experiments are missing vital control groups... In addition, many other initially promising findings have also proved unreliable in subsequent replication attempts by other groups (see discussion)... Perhaps there is signal amongst the noise in this field - I’m not sure - but I certainly think we need a greater focus on close replications, transparent reporting of null findings, and the use of pre-registration to clearly delineate when studies are exploratory and when they are confirmatory. I think these are important steps to ensure that we are robustly testing the theory - maybe we’ll find that the cracks deepen, or maybe we’ll find that the foundations are solid after all - either way, we’ll be making progress.
Hardwicke TE, Taqi M, & Shanks DR (2016). Postretrieval new learning does not reliably induce human memory updating via reconsolidation. Proceedings of the National Academy of Sciences of the United States of America, 113 (19), 5206-11 PMID: 27114514
