2014 Theses Doctoral

Learning by making errors: When and why errors help memory, and the metacognitive illusion that errors are hurtful for learning

Huelser, Barbie

This body of work begins to investigate the following three overarching questions on errors and learning. First, when are errors helpful for memory? Second, why are errors beneficial in certain circumstances? Third, are learners aware of when errors are advantageous for learning?
These questions cover two unique dimensions of learning by making errors, both from a memory and a metacognitive point of view. From a memory perspective, it might seem surprising that making an error compared to simply studying (no mistakes) could be beneficial for memory. We began our investigations with a replication and extension of previous work on the error generation effect: When does making errors enhance correct retention above studying? By investigating boundary conditions, this helps inform theory of the mechanism responsible for the error generation effect. We found that error generation only enhanced retention for related materials but not for unrelated word-pairs, and therefore, confirm that the error generation benefit is more than simply due to the act of generation.
However, what is the role of the error: Does it serve as a semantic mediator linking directly to the semantically related target, or can the error serve as an episodic link, bridging to the original learning episode, even if it is not directly linked to the target? If a learner remembers her error, does this help or hurt memory for the correct answer? By using materials that enabled errors that were either congruent (related) or incongruent (unrelated) to the correct answer, we found generating errors during learning led to benefits of memory, both when the error was congruent and incongruent to the target. Furthermore, when one could recall her error at test, correct answer memory was higher than when one could not recall her original error. These findings suggest that just a semantic explanation for the error generation is likely insufficient, and point to the importance of episodic recollection at retrieval for error generation to aid memory above study alone.
Lastly, we investigated this errorful learning methodology from a metacognitive perspective. Even when errors were beneficial for learning, we found that learners were unaware of the memorial advantage. We sought to ensure this underconfidence was not merely a function of poor performance accuracy or source monitoring. We were also interested in exploring if this bias was stable, or if one could correctly update her metacognitive knowledge simply by making item-level judgments. These initial projects open the doors for exciting research investigating individual differences on learning from generating errors, from both a memory and metacognitive perspective.