Why can one kind of memory fail while another remains?

For: Independent learners beginning a systematic study of memory, with an optional mathematical continuation.

A useful first question is not simply whether someone has a “good memory”. Ask what they are being asked to do: retain a short sequence, recount an event, recognise something encountered before, or carry out a practised skill. The introductory MIT sessions organise memory around such distinctions. Their second session uses amnesia and the case of H. M. to ask how different abilities can be affected differently. That makes it a productive starting point for investigating memory systems, rather than beginning with a list of brain-region names. MIT’s session overview sets out this purpose.

Keep three questions separate as you work: what information entered the system, what was retained, and what could be accessed later? MIT’s memory discussion introduces these as encoding, storage and retrieval. The distinction gives you a way to analyse a task. A wrong answer at the end does not, by itself, tell you which part failed. Before proposing an explanation, specify the information presented, the delay, the competing information and the response required. This is a framework for asking better questions, not a method for diagnosing an individual.

The first memory session provides vocabulary for retention and interference. Read its overview before the starting lecture if these ideas are new. Then compare the patient-based discussion with the everyday and social questions in Yale’s memory lecture. The Yale overview includes suggestibility and testimony. Those topics require care: this 2007 teaching record should help you recognise a research question, not settle the accuracy of a particular recollection or a present-day legal dispute.

A second comparison concerns levels of explanation. Describing performance on a memory task, relating performance to brain evidence and constructing a network that stores patterns are different achievements. The MIT recurrent-network lecture and its Hopfield-network continuation offer an advanced route into the third. For each model, identify what counts as a stored item, what counts as successful retrieval and which assumptions create its limits. A successful model behaviour is a reason to examine the proposal; it is not, on its own, evidence that the same mechanism explains every human memory task.

Use the discussion exercise to turn these distinctions into predictions. State what you expect before reading the course’s explanation, then identify an alternative account and an extra comparison that could separate them. This prevents a plausible story from standing in for a discriminating test. Keep the lifespan lecture as a separate extension: changes across adulthood introduce another set of questions about tasks, participants and comparisons. By the end of this route, aim to explain why two memory tasks can require different evidence, and what additional observation would strengthen your preferred account.

Next step

After the two MIT memory sessions, complete the discussion exercise. Choose either adult development for a lifespan comparison or the neural-computation course before attempting recurrent networks.

Limits of this route

The MIT psychology course is from 2011 and Yale psychology from 2007. Their source pages and selected teaching text were inspected, not the full videos. The MIT Memory II answer key says “three” explicit types while naming two; the guide does not repeat that error. Historical discussions of amnesia and testimony are not current clinical or legal guidance.