Special Coverage

Degree of Learning and Linear Forgetting (117)

Jerry Fisher and Gabriel Radvansky (University of Notre Dame)

Summary by Taylor Curley, Digital Content Associate Editor

This recap is part of a special series of session summaries from the Psychonomic Society's 61st Annual Meeting. To read the rest of the series, click here.

Forgetting Isn't Always Curvilinear

One of the most recognizable curves in cognitive psychology is Ebbinghaus’s forgetting curve, or when the proportion of information remembered from a task decreases in a curvilinear fashion over time. The forgetting curve was instrumental in early investigations of mnemonics and spaced practice and continues to be an important fixture in memory research.

But the rate of forgetting is not always curvilinear. In many cases, the data are better fit using a linear function. Jerry Fisher and Gabriel Radvansky performed a series of three experiments to examine why such differences in forgetting rates may occur.

Their main hypothesis rests on the Retrieval Accuracy from Fragmented Traces (RAFT) framework of memory. It proposes that the decay rates of memory traces are reliant on the contents of those memory traces. 

For memory traces with different quantities of constituent components, for example, those with fewer components would display decay in the form of Ebbinghaus’s curve. And traces with more components would display decay functions over time that are more linear. The best way to increase the number of components in a memory trace, the authors argue, is through better learning.

In three experiments, Fisher and Radvansky manipulated the number of components in memory traces by varying the types of learning. For example, in Experiment 1, participants either studied sentences once (low degree of learning), studied sentences once and then took a practice test (higher degree of learning), or engaged in two study/test trials for each sentence (even higher degree of learning).

The results support the RAFT theory: With better learning, leading to greater number of components in each memory trace, forgetting was best fit with a linear function. With less learning, forgetting was best fit with a curvilinear function. The results also suggest that engaging in effective learning strategies lead to reduced.