What we know about our own memory (Cue utilisation view…
- What we know about our own memory
procedure for examining metamemory= Judgment of Learning (JOL) - p estimates how likely it is/how confident they are that they will remember that item on a later test
- Koriat et al. (2004) p's studied 60 paired associates (table-apple). Immediately after, p's rater from 0-100% how likely they'd be able to recall the associate given the cue on a following test. Overall accuracy on a subsequent cued recall test was 53% and the mean JOL was 52%, indicating extremely accurate tuning of JOLs to the true level of recall.
- also found that items p's tended to recall worst were the ones they previously gave the lowest JOLs to and those they tended to recall best they previously gave the highest JOLs to.
- Consistent with other studies, the data do however show some evidence of underconfidence at low memory levels and overconfidence at high levels.
- BUT, insensitive to the effects of long retention interval on memory performance --> further expt. involved 2 additional groups asked how likely they'd remember a day later and a week later. Recall much poorer in day group and even worse in week group, yet mean JOLs were no lower than immediate group
Metamemory innacuracy: Potts and Shanks (2014)
- p's shown foreign language word pairs: some simply read, others cue word shown and p's had to guess what it might mean before being shown correct answer. P's also made JOL of how likely they would later recall the correct answer to the cue words
- P's rated themselves considerably less likely to remember the generated than read words (worried own intuitive but incorrect responses would intrude into their memory at test?)
- generated words were actually sig. better recalled at test than read ones = inaccurate meta memory judgments
The spacing effect
“Spaced” repetitions are far more beneficial in general than “massed” ones, yet people seem unaware of this benefit.
- Son (2004) p's studied definitions of unusual words (for 1s) and made JOL. Decided whether to study pair again immediately for 3s (massed condition), or study later (spaced), or didn't need to see again.
-given test where needed to give definition for the word
-lower the JOL for a definition, more likely p wants to see again immediately, and higher JOL the more they'd want to see it later
-for all final recall was better after spaced than massed trials = p's could have done much better on the final test by spacing definition yet don't acknowledge this benefit
The testing effect
People don't seem to appreciate testing (retrieval practice) compared to further study:
- Roediger and Karpicke (2006) p's study a passage containing around 30 central points. Encoding took place across four 5-min periods comprising either 4 consecutive study periods (SSSS), 3 study periods and one written recall test (SSST), or 1 study period and 3 tests (STTT).
-p's gave JOL for recalling a week later: highest for SSSS, then SSST, and lowest for STTT
-recall after 1w was highest in STTT group showing benefits of testing on memory = metamemory judgments of learning do not accurately reflect beneficial effect of retrieval practice
- The underconfidence with practice effect (UWP) refers to the finding that people's JOL shift from overconfidence to underconfidence on and after a first study-test trial (Koriat et al. 2002)
- Finn and Metcalfe (2007, 2008) proposed that people show UWP because they use their memory of the most recent test performance as a cue to make subsequent JOL, failing to account for engagement in more study (i.e. the Memory for Past Test (MPT) heuristic)
= another example of how people underestimate the benefits of external factors e.g. studying and practice on memory
- Different for children: Finn & Metcalfe (2014)
In contrast to adults, 3rd and 5th graders' JOL (of definitions, tested through cued-recall task) showed persistent overconfidence on and after a first study-test trial. 2nd expt tested ability to remember their prior test performance. Children's prior performance discriminations were accurate for items that they answered correctly on the prior trial, but were overconfident for items they had answered incorrectly =their continued overconfidence is a result of faulty memory, rather than a failure to use the MPT heuristic --> could be that they didn't use the MPT heuristic at all and this is a more advanced strategy that appears by young adulthood
Errors and mistakes --> typically viewed as something to avoid in learning process, in case they are learned/they document our inadequacies, but making errors is often essential to efficient learning
- Introducing desirable difficulties (e.g. interleaving, testing oneself, varying the conditions of learning) tend to increase errors in acquisition but tend to enhance long-term retention (Lee, 2012)
- Eliminating errors often eliminate learning- when retrieval of info to be learned is made so easy that success is ensured (e.g. through recency, cue support) tends to eliminate its benefits/reduce learning (Rawson and Kintsch 2005)
Major benefit on memory is effort: evidence that people are not only inaccurate about the benefits of effort, but also remember effort inaccurately, showing bias in how effortful something was
- Finn (2010) p's learned Spanish-English word pairs. Short list=30 very difficult pair. Long list=30 equally difficult followed by 15 less difficult ones
-every 3 trials p's rated discomfort from 0-10 --> discomfort was quite high and dropped on 15 easier items of long list.
-subsequent test, recall higher for words on short compared to long list
-p's asked which list had been easier and which they'd prefer to repeat- majority rated long list
=easier items at end of list made it appear less painful retrospectively, when it contained more discomfort --> p's thought something was easier to learn when it was actually harder
BUT, use of ELER heuristic may depend on individual differences: i.e. people's theories about the nature of human intelligence
- Research on naive theories of intelligence 1) people who believe that intelligence is a fixed entity tend to attribute their academic performance to innate ability more often than to effort, 2) whereas people who believe that intelligence can be developed incrementally attribute their performance to effort as much as they attribute it to ability. (Molden & Dweck, 2006)
Does this relate to JOL? YES
- Miele et al. (2011) p's who viewed intelligence as fixed, and who tended to interpret effortful encoding as indicating that they had reached the limits of their ability, used the ELER heuristic to make judgments of learning.
-p's who viewed intelligence as malleable, and who tended to interpret effortful encoding as indicating greater engagement in learning, did not use the ELER heuristic and at times predicted greater memory for items that they found more effortful to learn.
The simpler the information or the quicker the learning the more confident people are about their ability to recall it in the future (i.e., the higher their JOLs)- Koriat et al. (2006)
- led to the suggestion that people uniformly use an easily learned = easily remembered (ELER) heuristic to interpret experiences of encoding fluency during learning (e.g., Koriat, 2008)
= ignore the benefit of effort on memory
Evidence to suggest metamemory biases can be overcome to some extent:
- if JOLs are made after a delay, rather than immediately after studying an item they are more accurate --> consistent with cue utilization theory- item-judgment delay is likely to reduce impact of intrinsic features of the item and boost reliance on extrinsic/mnemonic properties
- some studies (e.g., Tauber & Rhodes, 2010; Castel, 2008) have found that accuracy improved when p's performed multiple study-test sessions in which they were able to judge for themselves the accuracy of their metamemory judgments and use this feedback to make more accurate subsequent judgements.