Presentation on theme: "PSY 368 Human Memory Memory Recognition cont.. Experiment 2 Signal Detection (Download details from BB) Like last time, find 3 participants You’ll need."— Presentation transcript:
Experiment 2 Signal Detection (Download details from BB) Like last time, find 3 participants You’ll need index cards for the words (write one word per card) Read instructions to participants, the IV is manipulated with different instructions for each condition, so make sure that you read the correct instructions. You’ll need to print out 3 copies of the “memory test” (1 for each participant) Fill out the datasheet and bring it to class on Monday (March 5 th, date in assignment is old Fall date). I will compile data for whole class and bring it on Wednesday March 7 th Reports will be due the Wednesday after Spring Break (March 21 st ),
Experiment 1 results Overall (N = 18) Immediate 7.6 items Delayed 6.0 Distraction 3.5 General report comments Don’t identify your participants Stick to APA style as much as you can Include your datasheet Primacy effect in all three conditions Recency effect strongest in immediate condition
Two classes of theories Single process theories - retrieval is one process regardless of task Tagging Model (Yntema & Trask, 1963) Strength Theory (Wickelgren & Norman, 1966) Dual process theories - two processes needed for retrieval - can be task dependent Generate-recognize model (G-R) e.g., Anderson & Bower (1972)’s HAM Remember/Know processes model (R/K) How does Recognition work?
Generate-recognize model (G-R) Recall is made up of two processes First, generate a set of plausible candidates for recall (G eneration stage ) Second, confirm whether each word is worthy of being recalled (R ecognition stage – not the same as the recognition test) Recognition is made up of only one process Because the experimenter provides a candidate, recognition does not need the generation stage Dual-process theories
(Tulving, 1985; Gardiner, 1988) Relatively recent change in recognition methodology When you recognize something, do you: Specifically remember (linked to Episodic memory) Conscious recollection of the information’s occurrence at study Just somehow know (linked to Semantic memory ) Knowing that it was on the list, but not having the conscious recollection, just a “feeling of knowing” Remember versus Know Process Model
Dual-process theories Tulving (1985) Present subjects with 27 c ategory - member pairs (FRUIT– pear) Recall tests: Free recall test Cued recall test (category) FRUIT Cued recall test (category + first letter of target) FRUIT- p Results The proportion of remember judgments decreased over the three kinds of tests Remember versus Know Process Model Prob(remember) 0.88 0.75 0.48
Remember/Know processes Make R/K judgment for “ Old ” items Remember = consciously recollect details of the item’s presentation Know = sure an item was presented, but can’t recall any of the details of presentation Dual-process theories Picture superiority effect R: P > W K: W > P Generation effect R: G > R K: R = G Word frequency effect R: L > H K: H = L Gardiner et al (1990, 1991, 1993) Remember versus Know Process Model R/K differ by:
Remember Versus Know Remember judgments are influenced by conceptual and attentional factors Know judgments are based on a procedural memory system This is similar to a distinction between explicit and implicit memory (more on this next week) Remember versus Know Process Model Gardiner et al (1990, 1991, 1993) gives an explanation:
Techniques used to distinguish dual processes Signal Detection Theory A technique for separating discrimination (“true” detection) from response bias Process Dissociation (next week) A technique for separating intentional (effortful) retrieval processes from incidental (automatic) retrieval processes May want to go back and review pages 111-114
Signal Detection Theory Signal Detection Theory : A model for explaining recognition memory Based on auditory perception experiments: Typical Task: Ask participants to detect a faint tone (signal) presented against a background of noise The tone’s loudness against the background noise is manipulated Easy-to-Detect Signal Hard-to-Detect Signal Background Noise Volume
Signal Detection Theory Brief History In World War II radar waves were used to detect enemy aircraft. The soldiers had to determine if the little spots of light are enemies, or simple noise (I.e. birds). There was no clearly defined criteria for making these kinds of decisions. Hit False alarm Miss Correct reject yes no SIGNAL: Are the spots on the screen enemies? DECISION: Should you scramble the jets? yesno Consequences: If an enemy went undetected, people could be killed. If noise was interpreted as an enemy, time and money would be lost and people would be put in harm’s way
Signal Detection Theory Response bias is based on a participant ’ s preference for a particular outcome. Preferences are based on costs & rewards Hit False alarm Miss Correct reject yes no SIGNAL: Are the spots on the screen enemies? DECISION: Should you scramble the jets? yesno For example, People will die because I failed to detect enemy, that is a very high cost. If congress yells at me for spending money, that is not a very high cost.
Signal Detection Theory Hit False alarm Miss Correct reject yes no SIGNAL: Are the spots on the screen enemies? DECISION: Should you scramble the jets? yesno High Criterion : less hits but also less false alarms Low criterion : more hits but also more false alarms Criterion level (C or β ) is set based on outcome preferences. Criterion level: The intensity at which a signal will be reported as being present ( Not the intensity at which it is perceived).
Signal Detection Theory Criterion level (C or β ) is set based on outcome preferences. Criterion level: The intensity at which a signal will be reported as being present ( Not the intensity at which it is perceived). High Criterion : less hits but also less false alarms Low criterion : more hits but also more false alarms stimulus intensity probability Noise Signal (enemy) Call for jets No alert - Criterion +
Signal Detection Theory d’ (“Dee-prime”) = Discriminability The difference between the means If d ’ is low, then this means there is low discriminability. The noise and stimulus are highly overlapping. d ’ = 0: pure chance If d ’ is high, then this means there is high discriminability. d ’ = 1: moderate performance d ’ = 4.65: “ optimal ” (corresponds to hit rate=0.99, false alarm rate=0.01) stimulus intensity probability Noise Signal (enemy) stimulus intensity probability Noise Signal (enemy) Low d’ high d’
Signal Detection Theory Recognition accuracy depends on : Whether a signal (noise/target memory) was actually presented The participant’s response Thus, there are four possible outcomes: Hits Correctly reporting the presence of the signal Correct Rejections Correctly reporting the absence of the signal False Alarms Incorrectly reporting presence of the signal when it did not occur Misses Failing to report the presence of the signal when it occurred CORRECT INCORRECT
Signal Detection Theory Assumptions: Memory traces have strength values (i.e. activation levels) Activation levels dictate how “ familiar ” a stimulus feels Traces vary in terms of their familiarity, based on: Attention paid to the stimulus during encoding The number of repetitions Familiarity values for “old” and “new” items are each normally distributed On average, “new” items are less familiar than “old” items However, some distractors are quite familiar because they appear often in other contexts or are similar to “old” items Thus, there can be overlap between the distributions Items that surpass a threshold (i.e. response criterion ) of familiarity are judged “old” 18
Signal Detection Theory Everything more familiar than (to the right of) the response criterion ( beta or β ) will be judged “old” A centrally placed β is unbiased Everything less familiar (i.e. to the left of β ) will be judged “new.” Hits (in green) Misses (in red) Above, the same distribution with the focus on the lure distribution to highlight: Correct rejections (in green) False alarms (in red) D prime (d’) represents: The distance between the distributions The participant’s ability to discriminate the two distributions 19
Signal Detection Theory A more liberal guesser will: Have a response criterion shifted to the left Accept more targets as “old” (i.e. hits) Accept more lures as “old” (i.e. false alarms) A more conservative guesser will: Shift β to the right Have fewer hits Have fewer false alarms Thus, the overlap in the distribution leads to: Trade offs between hits and false alarms Depends on the placement of the response criterion 20
Signal Detection Theory Calculating d’ and C (or β) Discriminability (d’): Step 1) Look up the z-score for the average Hit and False Alarm rates. Step 2) Apply the formula d ’ = z HIT – z FA, where z FA is the z-score for FAs and z HIT is the z-score for Hits. Criteria C (or β) : Take the negative of the average of z HIT and z FA. This is the criterion value C. Remember that positive C values indicate a conservative response bias, while negative C values indicate a liberal response bias. We will go over this in class again next week when we have our data for Experiment 2 http://memory.psych.mun.ca/models/dprime/ 21
Evidence for special ability: (1)Prosopagnosia The inability to recognize previously seen faces, with relative sparing of other perceptual, cognitive and memory functions. Intact ability to identify people using nonfacial features (voice) Due to brain injury (typically to the right temporal lobe) Broad Subtypes: 1. Apperceptive - failure to generate a sufficiently accurate percept to allow a successful match to stores of previously seen faces. 2. Associative - accurate percept, but failure to match because of loss of facial memory stores or disconnection from them. Face Recognition
Evidence for special ability: (2) Newborn preferences Studies done by Fantz (1961, 1963) - had kids look at three kinds of figures Morton and Johnson (1991) report that new-born babies will preferentially view faces Face Recognition
Yin (1969) found that whilst people are generally better at recognising upright faces than they are other objects. They are worse for inverted faces than they are for other inverted objects. Evidence for special ability: (3) Face inversion effect This suggests that the processing underlying normal face recognition is different from those underlying object recognition. Face Recognition
28 Why does the ‘Thatcher illusion’ occur? Bartlett and Searcy (1993) conducted experiments to measure face ‘grotesqueness’. Their results supported the “configural processing hypothesis” i.e. We have a difficulty in understanding the configuration of features when faces are inverted. We aren’t aware of the odd configuration of elements within the inverted Thatcher image.
Evidence for special ability: (4) Pop-out effect for faces (Herschler & Hochstein, 2005) Face Recognition Find the human face in the display as fast as you can. Ready?
Face Recognition Find the human face in the display as fast as you can. Ready?
Evidence for special ability: (4) Pop-out effect for faces (Herschler & Hochstein, 2005) Face Recognition Now find the animal face. Ready?
(1) Recognition is an explicit memory test. (2) Single- and dual-process theories of recognition (3) Single-process can ’ t account for differences across recall and recognition (4) G-R theory can ’ t account for items that are recalled, but not recognized (5) Face recognition seems to be a special ability Summary
The Mirror Effect Observed when “The type of stimulus that is accurately recognized as old when old is also accurately recognized as new when new. The type that is poorly recognized as old when old is also poorly recognized as new when new.” (Glanzer & Adams, 1985, p.8) Pervasive in recognition tests High/low word frequency and hit/false alarm rates, presentation rate, age of subject,...
The Mirror Effect - Example The Mirror Effect and the Word Frequency Effect Word Frequency HighLow Hits27.8431.00 False Alarms10.207.63 Source: Human Memory, p. 214
The Mirror Effect Significance: It eliminates all theories of recognition based on a unidimensional conception of strength or familiarity (single process models) May be able to be explained by dual process models Explanations for the mirror effect are still being formed
Dual-process theories Process Dissociation Procedure (Jacoby, 1991) Task : Participants study two sets of items in different contexts Two different recognition tests follow: Inclusion Condition : Say “yes” if they recognize an item from either context Correct recognition = Recollection + Familiarity Exclusion Condition : Say “yes” only if they recognize an item from one of the two contexts Familiarity = False alarms in exclusion condition Recollection = Inclusion’s correct recognition minus Familiarity Dissociating Recollection and Familiarity