Forgetting curves : Implications for connectionist models
(2002) In Cognitive Psychology 45(1). p.95-152- Abstract
Forgetting in long-term memory, as measured in a recall or a recognition test, is faster for items encoded more recently than for items encoded earlier. Data on forgetting curves fit a power function well. In contrast, many connectionist models predict either exponential decay or completely flat forgetting curves. This paper suggests a connectionist model to account for power-function forgetting curves by using bounded weights and by generating the learning rates from a monotonically decreasing function. The bounded weights introduce exponential forgetting in each weight and a power-function forgetting results when weights with different learning rates are averaged. It is argued that these assumptions are biologically reasonable.... (More)
Forgetting in long-term memory, as measured in a recall or a recognition test, is faster for items encoded more recently than for items encoded earlier. Data on forgetting curves fit a power function well. In contrast, many connectionist models predict either exponential decay or completely flat forgetting curves. This paper suggests a connectionist model to account for power-function forgetting curves by using bounded weights and by generating the learning rates from a monotonically decreasing function. The bounded weights introduce exponential forgetting in each weight and a power-function forgetting results when weights with different learning rates are averaged. It is argued that these assumptions are biologically reasonable. Therefore power-function forgetting curves are a property that may be expected from biological networks. The model has an analytic solution, which is a good approximation of a power function displaced one lag in time. This function fits better than any of the 105 suggested two-parameter forgetting-curve functions when tested on the most precise recognition memory data set collected by Rubin, Hinton, and Wenzel (1999). Unlike the power-function normally used, the suggested function is defined at lag zero. Several functions for generating learning rates with a finite integral yield power-function forgetting curves; however, the type of function influences the rate of forgetting. It is shown that power-function forgetting curves cannot be accounted for by variability in performance between subjects because it requires a distribution of performance that is not found in empirical data. An extension of the model accounts for intersecting forgetting curves found in massed and spaced repetitions. The model can also be extended to account for a faster forgetting rate in item recognition (IR) compared to associative recognition in short but not long retention intervals.
(Less)
- author
- Sikström, Sverker LU
- organization
- publishing date
- 2002-08
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Connectionist models, Forgetting curves, Massed and spaced repetitions, Power law
- in
- Cognitive Psychology
- volume
- 45
- issue
- 1
- pages
- 58 pages
- publisher
- Elsevier
- external identifiers
-
- pmid:12127503
- scopus:0036674345
- ISSN
- 0010-0285
- DOI
- 10.1016/S0010-0285(02)00012-9
- language
- English
- LU publication?
- yes
- additional info
- Funding Information: This research was supported by a grant from NSERC (APA 146) to Ben Murdock and by a post-doctoral grant from STINT to the author. I thank Ben Murdock for the inspiring discussion and for a careful reading of the manuscript. I also thank Hans Markowitsch, James McClelland, John Wixted, Richard Anderson, Tim Busey, Geoffrey Loftus, and the anonymous reviewers for comments on an earlier draft of the manuscript.
- id
- 40ddbb24-44aa-40bc-a349-e0b7443bfe2c
- date added to LUP
- 2021-11-04 14:27:21
- date last changed
- 2024-03-08 22:03:04
@article{40ddbb24-44aa-40bc-a349-e0b7443bfe2c, abstract = {{<p>Forgetting in long-term memory, as measured in a recall or a recognition test, is faster for items encoded more recently than for items encoded earlier. Data on forgetting curves fit a power function well. In contrast, many connectionist models predict either exponential decay or completely flat forgetting curves. This paper suggests a connectionist model to account for power-function forgetting curves by using bounded weights and by generating the learning rates from a monotonically decreasing function. The bounded weights introduce exponential forgetting in each weight and a power-function forgetting results when weights with different learning rates are averaged. It is argued that these assumptions are biologically reasonable. Therefore power-function forgetting curves are a property that may be expected from biological networks. The model has an analytic solution, which is a good approximation of a power function displaced one lag in time. This function fits better than any of the 105 suggested two-parameter forgetting-curve functions when tested on the most precise recognition memory data set collected by Rubin, Hinton, and Wenzel (1999). Unlike the power-function normally used, the suggested function is defined at lag zero. Several functions for generating learning rates with a finite integral yield power-function forgetting curves; however, the type of function influences the rate of forgetting. It is shown that power-function forgetting curves cannot be accounted for by variability in performance between subjects because it requires a distribution of performance that is not found in empirical data. An extension of the model accounts for intersecting forgetting curves found in massed and spaced repetitions. The model can also be extended to account for a faster forgetting rate in item recognition (IR) compared to associative recognition in short but not long retention intervals.</p>}}, author = {{Sikström, Sverker}}, issn = {{0010-0285}}, keywords = {{Connectionist models; Forgetting curves; Massed and spaced repetitions; Power law}}, language = {{eng}}, number = {{1}}, pages = {{95--152}}, publisher = {{Elsevier}}, series = {{Cognitive Psychology}}, title = {{Forgetting curves : Implications for connectionist models}}, url = {{http://dx.doi.org/10.1016/S0010-0285(02)00012-9}}, doi = {{10.1016/S0010-0285(02)00012-9}}, volume = {{45}}, year = {{2002}}, }