Episodic Memory Flashcards
What does the Dual‐coding hypothesis state? What’s some evidence for it?
Highly imageable words are easy to learn because they can be encoded both visually and verbally.
Imagery: forming mental images helps & combining objects in the image helps (Bower & Winzenz, 1970).
Carmichael et al. (1932). Verbal labels of ambiguous pictures influence the memory of the picture.
Bower et al. (1975). Subsequent recall was greatly enhanced when a set of droodles was accompanied by their titles.
Bower et al. (1969). The “minerals” conceptual hierarchy. Presenting words from each category together aids recall. Recall is much higher than when the same words were presented in scrambled order.
Explain what are Type I and Type II levels of processing according to Craik and Lockhart (1972).
Type (I) Maintenance rehearsal (just rehearsal): shallow processing
Type (II) Elaborative rehearsal: Include associations with information (especially semantic) that has been stored before: deep processing –> Stronger encoding & more retrieval cues
Rehearsal of information in STS keeps the information in STS and increases probability of transfer to LTS. But maintenance rehearsal usually did not result in good LTM performance. Elaborative rehearsal (deep semantic processing) is better than maintenance rehearsal (rote repetition).
Name three tasks adeguate to explore three LEVELS OF PROCESSING when enconding information in episodic memory.
> Shallow Graphemic Task: Decide whether each word is in upper‐ or lower‐case letters.
Deeper Intermediate Phonemic Task: Decide whether each word rhymes with a target word.
Deepest Semantic Task: Decide whether each word fits a sentence containing a blank.
Craik & Tulving (1975). Effects of type of encoding task on subsequent word recognition. Deep semantic encoding produced the best overall memory, especially for ‘Yes’ responses.
What does the TRANSFER APPROPRIATE PROCESSING hypothesis state? Describe some evidence.
Performance is improved if there is a relationship between the processing during encoding and the processing during retrieval (e.g. Fisher & Craik, 1977).
Morris et al. (1977)
Rhyme encoding condition: First present sentence: “___ rhymes with legal.” > then present target: “EAGLE” > resp Y/N
Semantic encoding condition: First present sentence: “The ___ had a silver engine.” > then present target: “TRAIN” > resp Y/N
Following the encoding task there was a surprise recognition task: either standard (e.g. “Was EAGLE one of the targets?) or rhyme recognition (e.g. “Does REGAL rhyme with one of the targets?)
Interaction between encoding task and recognition task: better performance when tasks at encoding and recognition match.
ROSE ET AL. (2010) - SIMILARITIES AND DIFFERENCES BETWEEN WM & LTM
Mazuryk and Lockhart (1974): Immediate and delayed recall of 5-item lists under 4 different processing conditions: ❶ Silent rehearsal ❷ Overt rehearsal ❸ Rhyme generation ❹ Generating verbal associates.
What were the results and how could those be interpreted?
Better immediate recall in the rehearsal conditions (overt and silent) & better delayed recall in the associate condition. Different effects of processing task on immediate vs delayed recall task could reflect a difference WM vs LTM. Or rather (alternative suggestion by Rose et al., 2010) nature of retrieval differs in WM (immediate memory) and LTM (delayed memory), different types of retrieval processes in immediate relative to delayed memory tasks.
ROSE ET AL. (2010) - SIMILARITIES AND DIFFERENCES BETWEEN WM & LTM
Describe the Levels-of-processing span task used in the first experiment.
Participants are presented with a list target words followed by two processing words. Conditions (task is to determine): Same color? (shallow) Rhyme? (interm.) Semantic. related? (deep)
> Immediate serial recall = STM
> Surprise delayed recognition after all blocks and a time-consuming arithmetic task (for delay) = LTM
ROSE ET AL. (2010) - SIMILARITIES AND DIFFERENCES BETWEEN WM & LTM
Exp 2. Sum up the changes relative to Exp 1.
– new (more time-consuming) shallow processing task: match number of vowels.
– list length either 4 (typical WM span) or 8 (supraspan).
– Between subjects manipulation of immediate recall (only half the subjects receive the immediate recall task in addition to the later delayed recognition task). Rationale: retrieval of items from long-term (or secondary) memory is known to benefit later memory, but retrieval of WM items is not.
ROSE ET AL. (2010) - SIMILARITIES AND DIFFERENCES BETWEEN WM & LTM
What were the results form exp 2? What do they imply?
- still no LOP effect in the immediate serial recall task.–> immediate testing doesn’t involve LTM
- Benefit of immediate testing on delayed recognition: suggests immediate testing involves LTM.
- Greater forgetting (from immediate to delayed testing) of short lists relative to long lists: suggests long-lists benefit more from retrieval practice. Consistent with the idea that long lists involve LTM to a greater extent –> immediate testing DOES involve LTM
Describe the early chaining and primacy models of serial recall form episodic memory.
Chaining Models (Ebbinghaus): A > B > C > D > E
Primacy Model (Page and Norris, 1998): Items are activated in memory upon presentation (activation based). The earlier the item the greater the activation (first items prioritized). Activation decays as a function of time. At recall items are retrieved in order of their activation level. Errors commonly occur when items have similar activation levels (especially in the middle of the list)
Between presentations subjects rehearse to prevent decay (similar to phonological loop). When there are too many items to rehearse during the interval decay of activation sets in.
Make an example and describe a temporal ratio model for serial recall from episodic memory.
Temporal Ratio Models (Brown et al., 2007): Temporal distinctiveness model
Episodic memories are located along a dimension representing temporal distance from the observer. The retrievability of an item is inversely related to its summed confusability with other items in memory. The confusability of items along a temporal dimension is given by the ratio of the temporal distances of those items at the time of recall (logarithmic scale of time). Thus, recent events will typically be less confusable with neighbouring events.
Describe an example of non-temporal interference model of serial recall and provide some evidence for it.
Interference Models (Oberauer & Lewandowsky, 2008): Non-temporal approach to forgetting Forgetting occurs through interference arising from the encoding of additional information. The encoding strength of each item is a function of its novelty or dissimilarity compared with what has already been memorized.
To test their model they compared performance in three serial recall tasks:
Standard serial recall of letter strings H-J-M-Q-R-V
Serial recall with a intervening distractor word: H-super-J-super-M-super-Q-super-R-super-V-super
Serial recall with an intervening triple-distractor word: H-super super super-J-super super super-M..
Subjects read aloud all letters and words during encoding. No difference between one (1E) or three (3E) distractors, because interference only comes from the first ‘super’ after each letter, repetitions are not novel they do not interfere. The same effects are found for interfering stimuli during retrieval.
All temporal models (primacy model; temporal ratio models) have problems accounting for this finding.
Describe Polyn, Norman and Kahana (2009) CONTEXT MAINTENANCE AND RETRIEVAL MODEL (CMR)
Items are associated with the context (= pattern of activity in the cognitive system) during stimulus presentation.
o TEMPORAL context (contiguity effect or temporal clustering effect)
o SOURCE context (the source clustering effect)
Items also have pre-existing semantic associations (semantic clustering effect)
Retrieving an item from memory also retrieves the associated contextual information (source and temporal context). The maintained context is gradually updated with each retrieval (i.e. previous context representation is not completely overwritten).
All items compete with each other for retrieval.
Describe Polyn, Norman and Kahana (2009) CONTEXT MAINTENANCE AND RETRIEVAL MODEL (CMR) again, this time using the classical lamp methafore.
Context as a set of spotlights: each lamp can illuminate a different subset of memories. The temporal lamp always illuminates a set of traces that where stored nearby in time (the light becoming more diffuse for more distant items). The source lamp illuminates memories that were associated with similar source characteristics. The context retrieved upon successful recall of an item may swing each lamp to illuminate a different set of items.
The active context functions as a cue to retrieve certain item features. This then updates the current context and the updated context allows the retrieval of other item features associated with this context.
How does CMR2 model (Lohnas, Polyn and Kahana 2015) explain recency and primacy effects of free recall from episodic memory?
The most recently presented items have the strongest representation in the current context (which is used as a retrieval cue in free recall) –> typical recency effect of free recall is the result. [without a STM buffer!]
The first items of the list have the strongest association to the context –> there is also a primacy effect. [no effect of rehearsal!]
The final item is the most likely item to be recalled first.
After recall of any given item the neighbouring items are most likely to be recalled next, because each the context is retrieved together with an item and the neighbouring items have a similar context. [no direct item-to-item associations: items are indirectly associated through a similar context!]
Upon presentation of an item the context representation is updated. Each item’s representation in the current context gradually decays. When an item is retrieved during recall the current context is updated to include this item’s context. The current context now no longer corresponds to the recency-weighted activation of each item.
After presentation of a second list, the items from the second list have a recency-weighted level of activation in the current context. Items from the first list also have some representation in the current context (especially if they were retrieved after presentation of the first list).
Between the lists there is a more rapid change in context (allowing the model to separate items from different lists).
