Psych 240 Exam 2

How thinking involves working memory

Being better at cognitive tasks requires a strong working memory (Raven’s + mental arithmetic)

Raven’s progressive matrices

Storing info from memory and using it to determine missing shape from matrix is a strong predictor of other cognitive task performance

Mental arithmetic

Doing an addition problem in your head involves storing numbers for retrieval (storage buffer), knowledge about what to do next (executive processes) while incorporating long term skills

Individual differences in WM

Reading span (reading sentences and recalling the final words) gives a measure of working memory span

Aging (WM)

Raven’s scores worsen with age and worsens with lower WM span and young and old people with the same WM score have the same Raven’s score (correlation), so Raven’s result may be caused by WM span (causation)

Interference (WM) (syllogism)

Random number generation (repeatedly stating random numbers aloud) while doing syllogistic reasoning (answering a question like all A’s are B’s, all B’s are C’s, so are all A’s C’s?) shows interference, meaning reasoning uses WM

Reading comprehension + WM

People with high WM spans have better reading comprehension, they are able to store more information for immediate retrieval after reading

Evidence for WM/LTM distinction

Double dissociations between amnesia types and serial position curve

Anterograde amnesia/WM disorder

Anterograde amnesia only affects LTM but WM is fine, impaired WM only affects WM, not LTM = neuroimaging double dissociation

Serial position curve

Testing recognition of a list of words leads to a primacy (LTM) and recency (WM) effect, people remember the earlier and most recent words better

• Delayed test: primacy effect holds, recency effect is eliminated = no longer in WM

• Speed of test: primacy effect increases (slower speed) or decreases (faster speed), recency effect doesn’t change = the faster, the less gets into LTM

Baddeley's 3-part model

phonological loop, visuospatial sketchpad, central executive

Phonological loop

Auditory inputs go into buffer (verbal store), non-auditory inputs are converted through subvocal rehearsal in the articulatory loop and then enter the buffer, inputs in buffer are refreshed through rehearsal

Phonological loop behavioral evidence (4)

• Phonological similarity

• Articulatory suppression

• Irrelevant speech effect

• Phonological storage capacity

Phonological similarity

Similar sounds impair recall

also called acoustic confusion, you store sound not meaning or visual features (acoustic coding) so sound similarity decreases accurate recall of words whereas meaning or visual features don’t matter

Articulatory suppression

Saying “the” disrupts recall/rehearsal of other words

Formation of phonological code can be prevented by rehearsing distracting material during WM task, eliminating the phonological similarity effect (ex. saying “the” over and over while trying to remember words eliminates the phonological similarity effect and interferes with remembering any words at all)

Irrelevant speech effect

Hearing audio at the same time as remembering a visual list disrupts memory because it disrupts rehearsal

Phonological storage capacity (5)

Subjects can remember about as many words as they can say in 2 seconds

• Word length effect: you can remember more shorter words 

• Speed of speech: you can remember words for people that talk faster

• Speed of pronunciation: you can remember more words pronounced quickly (bishop vs. harpoon)

• Speed of language: you can remember more words in languages where they’re said faster (counting in Mandarin/English)

• Chunks: letters are easier to remember if they are related and make meaningful chunks

Phonological loop PET evidence

The 2 back task (recognizing the letter two letters ago) requires rehearsal in phonological loop, activating Broca’s area (language, left hemisphere) and not right frontal, whereas the search task (recognizing if its a certain letter) does not require rehearsal or activate Broca’s area → double dissociation when compared to dot/circle visuospatial PET scan (between right and left hemispheres)

Visuospatial sketchpad

Visual imagery and spatial processing enters directly from visual perception/visual memory and can be treated like a percept (scanned/zoomed/etc)

Visuospatial sketchpad behavioral evidence (Brooks task)

Double dissociation between visuospatial task and response and phonological task and vocal response, they interfere with each other but not the opposite modality

• Visuospatial task (letter-scanning task): is the corner of the F on the top or bottom in a mental image? → faster at responding vocally (phonological) than pointing to yes/no (visuospatial)

• Phonological task (sentence task): remember a sentence and say which words are nouns → faster at pointing response (visuospatial) than saying which (phonological)

Visuospatial sketchpad PET evidence

Visuospatial WM condition (remembering if dot was in same location as circle) activates right hemisphere and not left compared to control condition (if the dot is currently in the same location as the circle) → double dissociation when compared to 2 back task/search task phonological PET scan (between right and left hemispheres)

Visuospatial vs. phonological brain hemispheres

Visuospatial = right, phonological = left

Central executive

Attentional supervisor, planning, monitoring

• Central executive evidence: frontal lobe syndrome causes distractibility, and perseveration (fail to stop inappropriate behavior, also called utilization behavior)

• Brain area: prefrontal cortex

Double dissociation logic (using lesion example)

• A lesion in Area A affects System A (WM) which impairs Task 1 (WM) and does NOT impair Task 2 (LTM)

• A lesion in Area B affects System B (LTM) which impairs Task 2 (LTM) and does NOT impair Task 1 (WM)

Modal model (three-stores model)

Sensory memory is under a second, short-term memory and long-term memory follow

• The longer an item is in STM, the more likely it makes it to LTM

• Problems: the longer time in STM != LT learning (anterograde amnesia + WM deficit double dissociation)

Levels of processing (depth of processing) model

Learning increases as it is processed at deeper levels (physical/appearance → acoustic/sound → semantic/meaning) solution to problem above

Explicit memory

Conscious recollection (declarative knowledge), recall and recognition 

Implicit memory

Implicit is unconscious change (includes procedural knowledge, priming, and facilitated processing after repeated exposure) 

Amnesia

Psychogenic is forgetting due to psychological trauma and organic is forgetting due to physical damage

• Anterograde: learning new explicit info in the future is impaired

• Retrograde: retrieving info prior to trauma is impaired, more recent is more impaired while childhood memory can be intact (temporally-graded)

Patient H.M.

Had part of brain removed that included the hippocampus (part of brain where damage causes amnesia) before they knew the importance of it 

Spared implicit memory (3)

People with amnesia are often fine at implicit tasks, while explicit memory is clearly impaired

• Mirror reading: practice makes you better at reading mirrored words, amnesics performed the same as controls for non-repeated words after practice (implicit), but controls improved faster at repeated words because they recognized them (explicit) 

• Tower of Hanoi: disk moves decreased just as fast for amnesics as controls (implicit) even though they didn’t remember doing it before (explicit)

• Priming (word fragment completion): participants see a list of words and then get their fragments to complete, seen words are completed faster than new words = priming → amnesics showed normal priming (completed old words faster) (implicit performance) even though they didn’t know the word was on the prior list (explicit performance) 

PET study evidence of dissociation between explicit/implicit (word stem completion)

Double dissociation between explicit and implicit task suggest the two tasks rely on different systems

• Explicit task: say the word from the list you remember that began with the given word stem → hippocampus and frontal lobe activity increased

• Implicit task: say the first word that comes to mind that completes the word stem → posterior visual area activity decreased

Dissociation between immediate (working) memory and long-term memory (HM)

• WM: good digit span, can remember something consistently rehearsed like the time someone left his room

• LTM: can’t learn stories, patterns, or new vocab

Dissociation between remote and anterograde memory (HM)

• Remote: can recall life from age 16 and under (classmates, seizures, famous people), knows he has memory problems and that his surgery went wrong

• Anterograde: doesn’t know his age, anyone he’s met since surgery, doesn’t know current events and confuses them even though he spends all his time watching TV (Elvis/JFK assassination)

Dissociation between implicit memory and explicit memory (HM)

• Implicit: has memory that doesn’t require conscious recollection, got better at puzzles/games without knowing he’d done them before, broke his ankle and knew how to use a walker but not why, improved on mental rotation task

• Explicit: couldn’t traditionally recognize or recall, often said he didn’t know what he was supposed to do in a task he’d done before but then guessed correctly (implicit)

No dissociation between verbal and non-verbal memory (HM)

Lost temporal lobe structure for both

No dissociation between semantic and episodic memory (HM)

Context-free generic knowledge and autobiographical records are both explicit and he has neither

Behavioral study evidence of dissociation explicit/implicit

Single dissociations between explicit and implicit memory are shown in modality of presentation (affects on implicit) and depth of processing (affects only explicit), suggesting a double dissociation between explicit and implicit memory

• Modality of presentation: getting visual or auditory word list and then completing a task only affects ability to complete reading (priming) task (implicit) and not recognition task (explicit) 

  • Explicit task: complete a yes/no recognition task for visually and auditorally presented words, no difference in performance based on modality

  • Implicit task: read words that were presented either visually or auditorally as fast as you can, performance improves for words that were visually presented (primed)

• Depth of processing: varying depth of processing of words on a list and then completing a task only affects ability to complete a recall task (explicit) and not a fragment completion/priming task (implicit)

  • Explicit task: recall as many words as you can on a list after encoding the words in memory by either rating how pleasant they are (semantic/deep processing) or thinking of a word that rhymes (acoustic/medium processing), depth of processing improves performance

  • Implicit task: complete fragments/word stems of words on list (priming), priming effect is constant and performance does not improve with deeper processing

Taxonomy of long-term memory

Semantic vs. episodic

Remembering meaning vs. remembering personal episodes → both are within explicit memory (knowing that, declarative knowledge)

Categorization

We classify things because the world is clumpy to know how to treat similar things, and it allows inferences about members of a class

Categorization by pigeons

Pigeons are trained to peck one of four keys depending on a stimulus, were accurate in categorizing never-before-seen images suggesting they were using categories

Physical similarity-based vs. conceptual knowledge/reasoning-based categorization

Categories are not always similarity based, hawks and flamingos are birds but hawks are more physically similar to bats, children as young as 4 will correctly say hawk feeds mashed-up food to its young as opposed to milk knowing they’re not mammals (like bat) and are birds (like flamingo)

Classical view of categorization

Defining properties are necessary and sufficient for categorization

• Problem: what defines categories like “game”? Some categories don’t have necessary and sufficient conditions that are enough to form the category 

Modern probabilistic view of categorization

Categories have characteristic properties and fuzzy boundaries, similarity to other members of the category determine if something is likely to belong to it

Typicality evidence of categorization (what does it show up in) (3)

Exemplars with more characteristic properties are rated as more typical of a category, could be a measure of similarity (they correlate)

• Ratings: apples and bananas are commonly rated highly typical while olive and figs are rated less typical, people generally agree on these ratings

• Sentence verification: people are faster to verify sentences that contain more typical exemplars than less typical (robin is a bird = fast, chicken is a bird = slow)

• Hedges: people are more likely to say “technically” (hedging) when something is less typical (whale is technically a mammal = common, cow is technically a mammal = uncommon)

Exemplar vs. prototype theory

Exemplar is when multiple examples of a category are stored in memory and we categorize new ones based on similarity to old ones, prototype is when one prototype is stored in memory as an ideal example and categorization is based on similarity to prototype

Geometric approach to categorization

Concepts lie in a geometric space, ratings of similarity correlate to where they are geometrically placed

3 metric axioms + violations: Rules that must be true about spatial relations according to geometric approach

• Minimality: the dissimilarity (distance) between a concept and itself must be 0 → familiar concepts are rated as more similar to themselves than unfamiliar (apple-apple more similar than pomegranate-pomegranate)

• Symmetry: the dissimilarity (distance) between two concepts must be the same regardless of order → an unfamiliar category is judged more similar to a familiar one than vice versa (similarity of apple to plum is less than similarity of plum to apple)

• Triangle Inequality: if concept A is similar to concept B and concept B is similar to concept C, then concepts A and C must be reasonably similar → some concepts are similar for different reasons so their mutuals aren’t similar (Jamaica is similar to Cuba but not China, which Cuba is similar to)

Feature-based measure (vs. geometric)

People give similarity ratings inconsistent with geometric space but feature-based similarity doesn’t require metric axioms because it involves comparing two lists of discrete features to one another

Tversky's feature comparison (contrast) model

Similarity(X, Y) = a(shared) - b(unique to X) - c(unique to Y) 

• How violations are accounted for:

  • Minimality: familiar things have more features so a(shared) is higher 

  • Symmetry: b and c weights can differ, so the order of the comparison makes a difference 

  • Triangle inequality: two concepts can have similar feature lists to a third but their feature lists don’t share anything

Coding experience by category (Smith)

Reduces perceptual load

Reduces storage space

Simplifies reasoning process

Taxonomy where medium levels are most likely to code experience

Misattribution brain area

Frontal lobe: damage in this area is associated with source memory errors and false recognition → frontal regions are involved in strategic monitoring of medial temporal lobe structures

Categories allow inferences (Smith)

Categorization allows for inductive inferences about categories, everything that is true in a higher category is true below unless otherwise stated

Category level and distinctness/similarity (Smith)

At basic level, members resemble each other and are distinct, at superordinate level, members may not resemble each other as much, and at subordinate level, members may not be as distinct

Teachable language comprehender (Original TLC - Collins and Quillian)

Model with hierarchical structure with superordinate, subordinate, and basic levels that organize categories for storage in the mind

Feature storage TLC (highest node)

Features are stored at the highest node they apply to, are automatically applied to lower levels

Sentence verification TLC (distance effects) + problems (3)

Can answer true/false questions about any concept

Distance effect: the more hierarchy links you must go through to access the concept, the longer it takes (more links = more time)

• Problems:

  • Time to verify category membership doesn’t change based on the number of inferences needed (how subordinate it is) but this predicted it would

  • Reverse distance: people are faster at verifying some sentences that contain more links (ex. faster at verifying a dog is an animal than a mammal)

  • Typicality: doesn’t explain the typicality effect of being faster to answer if a robin is a bird than a chicken (distance for these would be assumed the same)

  • Basic-level effects: we recognize something first as a chair rather than a piece of furniture or a specific type of chair, this doesn’t explain that

Revised TLC model (spreading activation) (structure + intersection search)

One node is activated when a person sees, hears, or thinks about a concept and activation spreads to adjacent nodes

• Structure: not hierarchical, links vary in strength/length (shorter = stronger, longer = weaker)

• Intersection search: when activation intersects it allows for sentence verification (ex. is a robin a bird → robin and bird activated + spread, when they intersect the person decides if it’s true)

Spreading activation accounts for (3)

• Reverse distance: dog is more close to animal than mammal so activation spreads faster

• Typicality: less typical examples are located farther away from category so activation takes longer

• Priming: verifying if both words are words is faster when they are part of the same semantic category (verifying if doctor is word after nurse is faster than verifying if beans is a word after nurse)

Privileged categories in spreading activation (McClelland)

Info is accessed directly, not by spreading activation in other categories, allowing for faster verification

Category prototypes in spreading activation (McClelland)

Stored representations are used to compare similarity between instances when activation spreads

Semantic dementia and language development as evidence for hierarchical structure (Table 1.1 McClelland)

Deterioration of semantic knowledge first causes loss of info about low-level/specific categories, children/infants are more sensitive to high-level categories first

Basic level supporting findings (Table 1.2 McClelland) (5 findings)

Maximizes informative + distinctness, balance between distinct and similar

• Faster to verify

• Adults prefer to use them

• Easier to discriminate between

• Show largest priming effects

• Children learn first

Expertise effects on basic level (Table 1.3 McClelland)

Experts differentiate more so basic-level advantages can shift lower, can verify faster, the nature of their expertise can cause differences in which concepts are faster

Evidence that conceptual information is stored in prototypes (Table 1.4 McClelland)

Typical instances are verified faster, children and people with dementia overuse familiar (prototype) names when they don’t know, likely because they are restricted to knowing typical instances because of similarity to stored prototype

Remembering verbatim vs. gist information

Long term memory involves remembering gist/meaning, not syntactic/verbatim info 

In-class demonstration of sentence memory

The more similar ideas from original sentences in new/old sentences, the more likely people are to say they’d heard the sentence before

Semantic vs. syntactic information

People are good at recognizing differences between syntax immediately after, but are only good with semantic info after a delay

Sachs (Galileo paragraph) study

Sachs (Galileo paragraph) study

Participants hear identical sentences with either a semantic, syntactic, or word order difference and are tested immediately or after a delay. They are most accurate at verifying syntactic change immediately and semantic change only after a delay.

Central vs. peripheral information (3 supporting findings)

There’s a clear distinction between memory based on how important the info is to the person

• Rating importance: most important = best memory

• Children extract central info implicitly, even though they don’t know what is most important when asked

• Traffic/weather on radio are usually not memorable unless it matters to you

Prior knowledge facilitating comprehension and retrieval (2 examples)

 Laundry and balloon examples show how context helps memory later

• Laundry: given an ambiguous paragraph about laundry, no title and title after = less ideas recalled vs. given title before = more ideas recalled

• Balloons: shown an ambiguous paragraph that only makes sense with a picture, being shown good image before = better recall vs. seeing it the text once, seeing the text twice, being shown bad image, or being shown good image after = worse recall

Prior knowledge hindering comprehension and retrieval (1 example)

War of the Ghosts was Native American story told to British students, students distorted info by neglecting what didn’t fit their expectations and distorting facts to fit, this increased over time (ex. canoe to boat, seal to fish)

Schema (3 types)

General conceptual knowledge used in understanding, meaningfully organizes concepts, tells us what to expect and what to infer

• Event schema (script) and evidence: a script tells us what to expect from a certain situation (restaurant)

  • People agree on things in scripts (would write down what happens when you enter a restaurant), recall events in a certain order, read things faster knowing schema/context, and can recall script items not in story

• Scene schema and evidence: people remembering things consistent with the schema (desk and chair in office), don’t remember things not related to it (rug and bulletin), and often have a false memory for things in schema but not in office (books)

• Story (narrative) schema: follows an episode format (meet characters, learn problem, see resolution), there are rules for what to expect and good authors often violate these

Transience (2 types)

Decreasing accessibility of info over time

• Gradual (long-term): long-term memories not retrieved dissipate

• Rapid (short-term): dissipates within seconds, within working memory

Absent-mindedness (3 properties)

Forgetting due to insufficient attention or superficial processing (affects encoding/attempted retrieval)

• Encoding: lapse of attention, often happens with automatic actions (forgetting car keys)

• Depth of processing: shallow encoding makes memory worse over time than deeper encoding

• Change blindness: people fail to detect changes like an actor changing in a scene

Blocking (3 properties)

Temporarily cannot retrieve, inaccessible info

• Tip-of-the-tongue: know its there and can’t produce it, could be due to interference of similar but incorrect items, can know partial info, often resolves itself quickly

• Old age: blocking increases with old age, especially with remembering names

• Nonretrieved items inhibited by retrieving related items: retrieval cues related to a previously studied word can block retrieval of a target word

Misattribution (4 properties)

Memory is present but misattributed to a time, place, or person

• Source confusion: recalling that you learned a fact from one source that you actually learned from a friend 

• Cryptomnesia: think a real thing was imagined, absence of subjective remembering

• False fame effect: people think previously studied names are famous people because they studied them before in the experiment

• Falsely recognition effect: claim to remember something that didn’t happen, people recognize words falsely because they may have been semantically related to words they did see

Roediger & McDermott experiment (misattribution - false recognition effect)

Studied 15 associates to one theme word, participants false alarmed 65-80% of nonpresented words and had as much confidence in these as actually presented ones

Suggestibility

Tendency to incorporate info provided by others into one’s own recollections

Inference

When getting info, we infer automatically based on prior knowledge 

Logical inference + example

Inference using reasoning

• Spatial relations example: box on chair, box to right of tree → infer chair is also to right of the tree, logical because it's consistent with spatial relations

Pragmatic inferences + example

Rely on context of situation, go beyond reasoning based on what was said

• Reading comprehension example: John was pounding a nail vs. looking for a nail to fix a bird house → if someone reads he was pounding a nail they infer it was with a hammer and will verify that sentence with higher confidence than the actual sentence

Advertising (assertion vs. implication)

Take advantage of people making pragmatic inferences and remembering implications as assertions

• Assertion commercial: taking this med will get you through the winter without a cold → could cause a lawsuit

• Implication commercial: get through cold season, take this med as directed → people accept this as an assertion (make pragmatic inference)

• Hedges: people ignore that they are uncertain when presented in ads (pills MAY help, VIRTUALLY spotless)

• Unclear comparisons: last longer, are healthier → avoid lawsuits and people make pragmatic inferences

Inferences during encoding

Giving context affects what’s being encoded and later recall (washing clothes context or balloon context helped make inferences)

Inferences during storage

Memory changes during storage, more over time, to fit schema better (ex. War of Ghosts) and forgetting is possible

Inferences during retrieval + example

Manipulating details about a story people read in past causes them to make false inferences

• Helen Keller experiment: read story about Carol Harris, 1.5 weeks later are told it was about Helen Keller and 50% of these falsely remember hearing that she was deaf and blind vs. 5% not deceived

Misinformation experiments (Loftus) (2)

Eyewitness memory for complex events can be distorted, because they may forget things or remember things that didn’t happen

• Hit vs. smashed: control group asked how fast cars were going when they hit, misled group asked how fast cars were going when they smashed → 14 vs. 32% wrongly say there was broken glass 

• Yield sign vs. stop sign: control group asked if the car passed the other at the stop sign, misled group is asked if the car passed the other at the yield sign → 73 vs. 41% correctly said it was a stop sign, most misled group was unaware of the deception

Overwriting hypothesis

Idea that maybe original info is replaced by new info when someone is misled (yield replaces stop sign in memory)

Hammer experiment (overwriting hypothesis)

Everyone is shown a man carrying a hammer, proves that hammer memory is still there even (not replaced/overwritten) when a group is originally misled 

• Group 1 is told the man carried a tool, asked “hammer or screwdriver” → 72% correct

• Group 2 is misled the man carried a screwdriver, asked “hammer or screwdriver” → 63% correct

• Group 3 is misled the man carried a screwdriver, asked “hammer or wrench” → 75% correct

Misinformation acceptance

When given new information about a memory, people often accept it as if it were true (due to recency effect where newer info is often stronger) but it is not replaced, instead they experience source confusion when they aren’t clear which is real

Interviewing techniques (cognitive interview) (5)

• Reinstate conditions: exploits encoding specificity, conditions present to context when encoded

• Tell story without interruption: leading questions can bias original memory like smashed cars example

• Ask questions in reverse order: earlier changes are tied to later ones, biasing answers to questions asked in chronological order 

• Sequential lineups: successive presentations are better than simultaneous, forcing yes/no decisions eliminates relative judgment (when person picks innocent suspect relatively closest to perp)

• Sequential questioning: multiple short interviews are better than one long one, as eyewitnesses create a bias as soon as they claim one thing within an interview to make everything else line up

Hypnosis

Not effective, causes people to generate lots of info even if not accurate, Loftus’s misinformation effects increase

Problems illustrated by Brewster case

Woman did not confirm suspect at least four times and only did because she was previously exposed to him multiple times since the event, detectives’ bad strategies interfered with leading questions and a poor understanding of memory deterioration 

Planting false childhood memories (lost-in-the-mall study)

Participants asked to remember childhood events written by family members, ¼ is false but made realistic with details from family → people recall 25% of the details of false events and 68% of true (high rate of recall for something made up)

Effects of imagining fictitious events

Participants rated the likelihood of 40 events happening and were asked to imagine some, ~20% who imagined had increased likelihood scores on retest

General impairment vs. suspect-bias variables in eyewitness identification (+ example)

Suspect bias variables can account for why the eyewitness specifically selected one innocent person out of all possibilities, while general impairment cannot

• If suspect was other race it doesn’t explain why they picked 1/7 people that were

• Only prior exposure, disguise, overheard descriptions, etc. can explain why they picked 1 person falsely over the rest

Memory taxonomy