Learning Versus Performance: An Integrative Review
Nicholas C. Soderstrom and Robert A. Bjork | Perspectives on Psychological Science 2015
In a Nutshell 🥜
Soderstrom and Bjork1 draw the distinction between learning and performance and point out that working towards achieving one may go against working towards achieving another. The paper begins with a review of early work on the learning-performance distinction, namely, latent learning, overlearning, and fatigue:
Latent learning. Learning occurs even in the absence of any reinforcement behavior. Reinforcement is necessary to reveal learning but not to induce learning.
Overlearning. Continued practice of a task, even beyond the achievement of some criterion of mastery, enhances retention and accelerates the rate of relearning.
Fatigue. Despite fatigue limiting noticeable gains in short-term performance, learning can still occur.
The paper then discusses key three concepts that show how learning and performance can be divergent objectives:
Distribution of practice. In massing practice, students study the same thing repeatedly in one block. In distributing practice (or spacing), students separate study sessions with unrelated activities or interleave study sessions for different tasks. Distributing practice impairs short-term performance growth. However, it improves long-term learning. In addition, students will be able to transfer learned skills more easily to a new response pattern.
Variability of practice. Varying the conditions of practice can have negative effects on performance during acquisition but foster long-term learning and generalization capability. For example, if one wishes to shoot more accurate free throws, one should not only practice from the foul line only but also from various other positions.
Retrieval practice. Retrieval practice by testing not only passively assesses what is stored in memory but also modifies them to be more recallable in the future, enhancing long-term retention compared to reading them over and over. However, in measuring performance immediately after acquisition, rereading performs better, leading students to prefer rereading. This occurs even in cases when there is no corrective feedback. Failed retrieval attempts also enhance learning. However, in measuring performance immediately after acquisition, rereading performs better.
The paper then shows that students generally favor learning activities that enhance immediate performance during acquisition, over those that are beneficial to long-term learning. For example, students prefer massing practice, low variability of practice, and rereading over distributing practice, high variability of practice, and retrieval practice.
Finally, the paper concludes with contemporary theories that seek the explain the distinction between learning and performance:
Theory of disuse2. Storage strength (learning) is expressed as a negatively accelerated function of current retrieval strength. Situations that reduce current retrieval strength (e.g., distributing practice) yield greater storage strength.
Schema theory3. Variability of practice enables one to become more familiar with the general underlying program of the skill needed to successfully perform the task.
Reloading hypothesis4. Distributing practice enables one to reload (or reproduce) the skill during acquisition, promoting learning.
Desirable difficulties framework5. Difficult manipulations can be desirable because they encourage active encoding processes that support better long-term retention and transfer.
Some Thoughts 💭
The concepts illustrating how learning and performance can be divergent objectives makes me wonder about its implications for machine learning. While we are optimizing for performance during machine learning training, are we also somehow working against learning in terms of generalization and transfer?
The distinction between long-term learning and performance during acquisition also makes me reflect on my own learning practices.
Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2), 176-199.
Bjork, R. A., & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. From learning processes to cognitive processes: Essays in honor of William K. Estes, 2, 35-67.
Schmidt, R. A. (1975). A schema theory of discrete motor skill learning. Psychological review, 82(4), 225.
Lee, T. D., & Magill, R. A. (1985). Can forgetting facilitate skill acquisition?. In Advances in psychology (Vol. 27, pp. 3-22). North-Holland.
McDaniel, M. A., & Butler, A. C. (2011). A contextual framework for understanding when difficulties are desirable. Successful remembering and successful forgetting: A festschrift in honor of Robert A. Bjork, 175-198.