D&M Chapter 5, textual notes
Judgments of Learning
Do all monitoring judgments tap the same information? (FOK, JOL, EOL) [Answer: Nope]
**JOL is the most studied recently**
When comparing all 3 monitoring judgements, research found that these factors are positively but only weakly correlated
Variables Influencing JOL Accuracy
JOL accuracy influenced by:
- number of trials
- timing of those judgments
- ...others, not covered here
Historical background
Arbuckle & Cuddy had college students do a simple JOL on word pairs, with reasonable accuracy. Groninger had similar results. Everyone seems to be testing college students, very convenient :). Did not tend to assess degree of accuracy. Nelson introduces gamma correlation to measure relative accuracy in 1984 and this becomes the standard. 0 correlation == chance. Correlations come in ~.30 for their relative accuracy. Opportunity here for improvement. Increased trials impacts calibration and relative differently, whereas timing improves both.
Number of Trials and Underconfidence-With-Practice Effect
When doing multiple trials, people's relative accuracy improves (they more consistently know themselves relative to the trial in front of them), they underestimated their calibration (actual accuracy) -- apparently not realizing how much the multiple trials would help them.
I wonder how this applies to non-college students and non-Americans -- might people from cultures that value reflective behavior or memorization perform differently?
Global judgments -- your sense of how you'll do overall, as opposed to the item-by-item type of JOL. Your assessment tends to follow your previous performance rather than increase when you have more practices.
Timing of Judgments and Delayed-JOL Effect
Nelson & Dunlosky found that immediate JOL has low-to-moderate relative accuracy, but a short delay boosts rel. accuracy. Essentially we are more consistent in our self-judgments if we wait a moment.
Note on Impact of Drugs:
Drugs seem to impact learning but not JOL -- depressants decreasing learning, stimulants improving it slightly.
One experiment to explain delayed JOL found that if you give people the pair, you throw off even a delayed JOL. So the metamemory 'if I know if now, I'll know it later' is removed. But if you give them half, they have reasonable rel accuracy. (prompting "undermines the students' ability to attempt retrieval of the response")
Overall assumption is that you're making a retrieval attempt as part of your JOL -- either metacognitive monitoring is real, or else there's a self-fulfilling prophesy at work here. Research debate -- do delayed JOLs change your memory or does the elapsed time actually improve your judgment process?
Monitoring-Dual-Memories Hypothesis
- Nelson and Dunlosky
- since JOL is attempt to retrieve, short-term memory interferes with an immediate JOL; you don't know if it's in your long-term store
- gap before JOL means you're just looking at long-term memory -- which is what JOL is meant to assess
Self-Fulfilling Prophesy Hypothesis
- Practicing retrieval improves assessment of retrieval (seems like you should be able to test for this)
- A delayed JOL acts on long-term memory, so it's a more accurate time to JOL.
Probably a combination of both -- but MDM is challenged by the fact that accuracy does not vary with delay length. And the underlying assumption of a retrieval attempt doesn't seem borne out by timing -- a negative JOL is sometimes very fast (lack of recognition of the prompt). Also, there seem to be may kinds of JOL.
Hypotheses about JOLs
Ease-of-processing Hypothesis -- Begg 1989 proposes that JOLs are based on how easy it is to processing an item immediately prior to making the judgment. A heuristic/rule. Given a pair, you might look at it and decide on a mnemonic -- a good one, and you'll feel confident. Nothing comes to mind? Not confident. Then again, easy-processing may still lead to difficult-to-remember (names seem like an example of this). Explored this via common and rare words -- but rare words were actually remembered better. To more fully assess processing ease, Hertzog et al had participants use imagery (press a button when you've got an image). The time taken to come up with an image was used as the ease of processing. Quick imagery generation does indeed seem to predict JOL.
Retrieval-Fluency Hypothesis -- also heuristic. Benjamin and Bjork 1996 described. More fluent when: retrieval is quick and when retrieval includes more details. (related to FOK accessibility hypothesis) Fluency demonstrated by first asking a relatively easy trivia question, then ask them to predict if they'll remember their response (JOL). JOLs were highest when they answered the trivia question quickly -- BUT -- actual recall is greater when retrieval is slower! People may have thought they'd get the cues on the final test, and therefore paid less attention. Fluency heuristic misled them. JOL/recall correlation was zero.
Cue-Utilization Approach for JOL
Ease of processing hypothesis doesn't explain underconfidence-with-practice effect. Accuracy comes from a variety of cues both in the brain and in the study conditions -- and use those to infer memory. Accuracy comes from correlation of cues and the test.
As expected, item relatedness is influential on JOL -- salt-pepper gives you more confidence than dog-chair. Study trials influence JOL, but not as strongly as they influence actual performance (UWP). Relative accuracy is influenced by test trials and delayed JOL.
Koriat proposed a taxonomy of cues and their influence.
Koriat's Cue Taxonomy
- intrinsic cues -- things about the items themselves, such as relatedness
- extrinsic cues -- using imagery or other "encoding operations" actual mnemonics and loci also a trick here?
- mnemonic cues -- internal indicators based on subjective experience ease of processing, previous recall
Overall theory is that individual JOLs are differentially sensitive to cues. JOL more susceptible to intrinsic than extrinsic, but performance is impacted by intrinsic and extrinsic as well.
Overall -- people seem relatively unaware of the impacts of extrinsic cueing
So why are we bad at metacognition, including both metamemory and metacomprehension?
Still studying . . . re-reading and summarizing helps, but we don't know why
What's the function of JOL?
- monitoring regulates behavior
- FOK is retrieval strategy
- JOL is studying control -- validated using self-paced study and multiple tests
Theory of Self-Regulated Study
You likely establish a 'norm of study' for the subject -- how much do you care and how well do you want to do? Usually people follow their norm of study in how they allocate their time. If told to emphasize accuracy ("until you're absolutely sure"), students had a higher norm of study -- if told to emphasize speed ("only as long as you need").
Discrepancy-Reduction Model
Why do people study 'difficult to learn' materials longer than 'easy to learn' ones?
DR Hypothesis says it's a measure of the difference between your monitoring of learning and your norm of study.
But there seems to be more to it than this. When cramming, students are likely to study easier material. Given varying times to study, test subjects tended to focus on easier (less time given) or harder (more time given) topics. This is called 'STEM' Effect -- the "Shift-To-Easier-Materials" effect. Seems to be adaptive decision making conducted during study time.
Region-of-Proximal-Learning Hypothesis
Metcalf argues that discrepancy reduction isn't enough -- would cause students to spend too much time on items they might be unable to learn. He offers Region-Of-Proximal-Learning (RPL) hypothesis -- that study time is allocated to material in the region that is "just beyond the grasp of the learner and that is most amenable to learning."
