Retrieval & Recognition Memory Flashcards
What is the frontal lobe role in recall and in recognition forms of retrieval?
Frontal cortex critical for recall tasks, but not for recognition tasks: because recall tasks require a strategic, controlled search of memory.
Mangels et al. (1996): effects of frontal lobe damage on free recall, cued recall but not on recognition tests of remote memories.
Factors determining retrieval success: try to retrieve as many as you can (ehehe)
- Attention to cues: a useful cue might be present, but if it is not attended it is not used for retrieval
- Relevance of cue: a good cue was encoded together with the target.
- Cue-target strength: a good cue is strongly associated with the target (compare associations in semantic memory)
- Number of cues: the more the better, more likely sufficient activation will spread to the target.
- Target strength: how well was the target encoded/stored in memory?
- Retrieval strategy: which cues are you going to focus on; frontal cortex involvement
- Retrieval mode: you must be actively engaged in memory search.
What is ENCODING SPECIFICITY? Provide some evidence for it.
The relation between context at encoding and context at retrieval also affects memory performance. We learn information together with its context.
Godden & Baddeley (1975) “diving experiment”: Words learned and tested in the same environment are better recalled than those items for which the environmental context varied between study and test.
Grant et al. (1998) “studying” experiment with a noise and quite condition: replicated the result.
State‐Dependent Learning: learning is associated with a particular internal state. Better memory if person’s mood at encoding matches mood during retrieval.
Eich & Metcalfe (1989) “mood” experiment with a sad and happy condition.
Encoding specificity may also involve the effect of cues at encoding and retrieval. Explain how and provide evidence.
Memory performance at retrieval depends on the similarity between cues used at retrieval and the way in which the stimulus was processed.
Tulving and Thomson (1973) Stage 1: learning of weak associates: SPIDER ‐ BIRD > test: SPIDER ‐ ?
Stage 2: free associate to EAGLE > generated resp: WINGS BIRD MAJESTIC … > recognition task: which word was in the first list? = 24% correctly recognized
Stage 3: paired associates: SPIDER ‐ ? = 63% correct
Subjects were better at recall than recognition!
When associated with SPIDER, the word BIRD reminds the subject of medium‐sized birds that eat spiders (e.g. crows). After generating BIRD in response to EAGLE this will not easily be recognized as an ‘old’ word, because they were imagining different kinds of birds.
The conversion threshold proposed by Tulving (1985) is a line that indicates when an item will be retrieved. What does it depend on?
It depends on the strength of the episodic trace and how much semantic cue information subjects are given.
• (c,x) words subjects could recall without a cue;
• (b,y) words that required a category cue to be recalled.
• (a,z) words that required a category cue and a first-letter cue.
Moving along the conversion threshold to the upper left of the graph, subjects required more semantic cue information to compensate for the weaker episodic trace.
Tulving demonstrated that the more semantic cue information was required for correct recall, the lower the percentage of “Remember” responses of correctly recalled words.
Describe the DRM paradigm [Deese‐Roediger‐McDermott]. What does it help to study?
The DRM is designed to elicit FALSE MEMORIES in recognition memory.
12‐word lists consisting of 12 primary associates of a critical non-presented word. e.g. critical word (lure): needle; List: thread, pin, eye, sharp, point, pricked, thimble, haystack, pain, hurt, injection.
Subjects regularly have false memories of semantically related lures, both in free recall (Deese, 1959) and recognition task (Roediger and McDermott, 1995). Subjects often give “Remember” responses to critical lures.
In a DRM paradigm, why do false memories occur for words that are strongly related to the words on the list?
– Roediger and McDermott (1995): error in reality‐monitoring: perhaps the words on the list made subjects think of the related word and later they mistook their thought of the word for seeing the word on the list.
– Compare: source monitoring error: subjects remember/recognize the word, but make a source misattribution (think the source of the memory was the external list instead of an internal thought)
In a SIGNAL DETECTION THEORY account for Recognition Memory performace: what are d’, beta, and C?
Recognition memory strength is normally distributed so it can be converted to z-score (number of SD from the mean).
The discriminability d’ is a measure for the subjects ability to discriminate targets from lures. The greater d’, the more sensitive (better able to distinguish targets from lures) the observer.
β is the subject’s criterion for deciding whether an item is a target. β < 1 is a liberal observer; β > 1 is a conservative observer.
C (alternative measure for bias) < 0 is liberal; C > 0 is conservative.
Tulving initiated the dual-process theory tradition in recognition memory by distinguishing two states of awareness associated with the conscious experience of memory. Which ones?
Retrieval from episodic memory (awareness of a past event as autobiographical in nature) & retrieval from semantic memory (awareness of previously acquired knowledge without any autobiographical component)
What are the basic assumptions of the DUAL‐PROCES SIGNAL DETECTION MODEL – Yonelinas (1994)?
“Know” responses are based on familiarity (continuous signal-detection process, ahistorical memory-strength) and “Remember” responses are based on recollection (categorical process, it either occurs or it does not, historical information)
*Explains the almost universal experience of encountering the familiar face that you cannot specifically remember having met before: Mandler’s (1980) classic “butcher-on-the-bus”.
What do UNIDIMENSIONAL MODELS of recognition memory state?
recognition memory is based on a single signal of “memory strength” or “evidence” that the word was on the list. Unidimensional models simply assume that “remember” responses reflect a stronger memory signal than “know” responses. Concept of memory strength (high confidence, high accuracy, fast reaction times).
Targets and lures on a recognition memory test have normally distributed memory strengths (targets: greater mean and variance than lures). If the strength of memory associated with an item exceeds a decision criterion: old.
What is the problem with unidimentional models in terms of zROC slopes predictions?
ROC (Receiver Operating Characteristic) curve shows the hit rate as a function of the false alarm rate for different criteria of the subject. ROC curve closer to the upper left corner: higher d’. After transforming the ROC curves into z-scores, a slope can be calculated for the linear zROC curves: indicate the relationship between the old and new distributions.
Slope < 1: greater variance for the old items.
Slope > 1: greater variance for the new items.
The uni-dimensional model predicts similar zROC slopes for the R-K task (old-know-remember) compared to traditional Recognition Memory tasks (old-new). Rotello et al (2004): different mean zROC slopes provides evidence against the one-dimensional recognition model. R and K judgments share more than one confidence rating (5, as signal-detection anticipates, but also 6) so: the slope of zROC derived from confidence ratings differs from the one derived from R/K judgments.
Describe the Rotello et al. (2004) SUM-DIFFERENCE THEORY OF REMEMBERING AND KNOWING (STREAK).
– old-new decisions are made on the basis of a weighted sum of the specific strength (recollection) and the global strength (familiarity): greater sum greater confidence.
– R-K decision is made on the basis of a weighted difference between specific and global strength: greater the difference, more likely remember response.
Describe Wixted & Stretch (2004) VARIABLE CRITERION UNIDIMENSIONAL MODEL. How does it account for different zROC slopes?
A comparison between R/K and standard recognition memory tasks assumes subjects use constant criteria. But the differing z-ROC slopes could be due to subjects varying the criterion for R-K responses so these results do not refute one-dimensional models: can account for overlap in confidence ratings between R and K judgements (i.e. both may occur with confidence ratings of 4, 5 and 6, but with differing frequencies).
The proportion of the target distribution that falls to the right of the remember criterion is much greater than the one of the lure distribution. Explains why remember false alarms have higher confidence than know false alarms. Remember hit and false alarm rates correlation is strongly positive. Es. manipulation that induces a more liberal criterion (leftward shift) results in both a higher remember hit and remember false alarm rate.
Reaction times are greater for K judgements (closer to old/new decision criterion: point of maximal subjective uncertainty).
WIXTED & MICKES (2010) – CONTINUOUS DUAL‐PROCES MODEL (CDP)
Describe the model in relation to STREAK alternative theory.
Both familiarity and recollection are continuous signal-detection processes; each has its own “noise” distribution (for lures) and its own “signal” distribution (for targets). “Old” responses are based on the sum of the familiarity and recollection signals (=STREAK), for both targets and lures. The confidence criteria are placed along the aggregated memory-strength axis.
R responses are based on recollection and K responses on familiarity (≠ STREAK). The remember criterion is placed on the recollection axis, the know criterion on the familiarity axis. Participants first interrogate the recollection dimension, if there was enough (≠standard dual-process theories) recollection = remember (the strength of recollection signal exceeds the remember criterion) remember hit rate correlates with remember false alarm rate.
Know judgment implies that the amount of familiarity exceeds a criterion on the familiarity dimension, but it does not imply the absence of recollection.
