Recency is sufficient for reconciling categorisation and memory: Commentary on Devraj et al. (2024).

Area of Science:

• Cognitive Psychology
• Memory Research
• Machine Learning Models

Background:

• Existing research presents conflicting findings on memory accessibility over time versus categorization performance.
• Devraj et al. proposed that forgetting adapts to task demands, favoring prototype use in classification.
• This study re-examines data to challenge the strategy-shifting hypothesis.

Purpose of the Study:

• To reconcile conflicting findings between memory accessibility and categorization performance.
• To investigate the role of forgetting in classification tasks.
• To evaluate exemplar-based memory models against strategy-shifting models.

Main Methods:

• Re-analysis of existing experimental data from Devraj et al.
• Systematic manipulation of stimulus testing delays to isolate forgetting effects.
• Comparison of exemplar classification models with and without a forgetting function.
• Evaluation of model fit to predict performance and strategy shifts.

Main Results:

• The observed performance patterns can be explained by exemplar forgetting in both experimental and control conditions.
• Increased forgetting effects in later trials, due to longer delays, reversed performance growth.
• A forgetting function improved exemplar model fit, predicting observed patterns a priori.
• Exemplar forgetting provided equivalent or better model fit than strategy-shifting models.

Conclusions:

• Power-law memory decay does not create a conflict between categorization and memory findings.
• Increased forgetting across longer retention intervals explains performance changes, not a shift to prototype use.
• Exemplar-forgetting models are more parsimonious and better explain the data than strategy-shifting models.