Interference and the representation of order within associations

Area of Science:

• Cognitive Psychology
• Human Memory
• Computational Neuroscience

Background:

• Recall accuracy in paired-associate learning can depend on memory for within-pair item order.
• Existing models of pair memory differ in their assumptions about the storage of order information, ranging from no explicit storage to perfect order memory.

Purpose of the Study:

• To investigate the interplay between associative memory and within-pair order memory under conditions of interference.
• To test the predictions of different computational models of pair memory.

Main Methods:

• Examined forward and backward recall probes of word pairs using a double-function list paradigm.
• Introduced order-dependent associative interference to assess its impact on recall accuracy and correlations between forward and backward probes.
• Analyzed test-retest correlations to evaluate the reliability of memory encoding and retrieval.

Main Results:

• Associative interference reduced the correlation between forward and backward recall accuracy, challenging matrix and concatenation-based models.
• Participants demonstrated some ability to overcome interference related to within-pair order ambiguity, challenging certain convolution-based models.
• Interference decreased test-retest correlations, indicating unreliable order encoding that is not captured by current models.

Conclusions:

• Within-pair order memory is neither perfect nor entirely absent, suggesting a fallible encoding mechanism in verbal association tasks.
• Existing models need revision to incorporate unreliable order encoding, with the specific mechanism potentially varying by model architecture.
• The findings necessitate new assumptions about the unreliability of order encoding, distinct from retrieval processes, to accurately model human associative memory.