Dissociating memory accessibility and precision in forgetting

Forgetting is an everyday experience for all of us yet little is known about what happens to memories as they are forgotten. Do our memories gradually lose detail, or are they ‘erased’ and ‘overwritten’?

Like Comment
Read the paper

When attempting to recall an event from memory, there is a chance that the information we are looking for is simply out of reach; inaccessible to us in the present moment. If this happens, we may be able to recall the information sometime later, or it may have been lost for good. Such losses in memory ‘accessibility’ are well documented and have been studied since the advent of experimental psychology. However, if a memory remains accessible, does the precision of the information that it represents change for any reason? Do our memories become ‘noisier’ or less reliable and is this affected by the type of information our memories encode?

We started exploring these questions in a few lab-based pilot experiments. This allowed us to formulate an analysis pipeline and establish a set of hypotheses to test in a large-scale experiment. Discussion between Aidan (the senior author) and myself then turned to what the large-scale experiment should look like.

It was a Friday afternoon, and following a meeting, Aidan and I went to lunch - Friday was Fish and Chips Day! Over lunch, we discussed potential plans for the new experiment. By the time we arrived back at the lab, we had decided to collect data online, using a web-based task, and to submit the study as a registered report. We chose to collect data online as we needed a fairly large sample to properly test our hypotheses. Additionally, as part of a push to widen our use of open science practices, Aidan was keen to run the project as a registered report.

I was fully on board with this plan, but I knew that the project would be a challenge. I had no experience of conducting online research experiments (particularly one this complex), nor had I ever submitted a registered report. However, given the pilot data we already had, I was confident that things would work out. Simulating the final dataset ahead of time assured us that the experimental design was sound. Thankfully, Blake (the second author) was on hand to provide theoretical guidance and to challenge some of the initial analysis ideas I proposed.

Finalising the protocol for the registered report took some time but this process ensured the study was the most rigorous scientific project that I have undertaken to date. The peer review process was extremely helpful in this respect. We did have to make some substantial changes which included running another (online) pilot and switching from using frequentist to Bayesian inference. However, the whole process was well worth it. Once the protocol was accepted, it took 8 months to collect the main dataset, run all the pre-registered and exploratory analyses, and produce the final report.

So, what did we find? Our results showed that forgetting principally involves losses in accessibility with little to no changes in memory precision. However, when memories can be grouped together and ‘summarised’ by a general pattern, we saw an apparent boost in the level of accessibility with a corresponding loss in precision.

Sam C Berens

Research Fellow In Cognitive Neuroscience, University of Sussex