PSY 246: Introduction (see notes) Tools of discovery in Cognitive Pysch: Electrical Electroencephalography (EEG/ERP) Single cell (neuron) recordings Electromagnetic Transcranial Magnetic Stimulation (TMS) Magnetoencephalography (MEG) Magnetic Magnetic resonance imaging (MRI) Magnetic/haemodynamic Functional magnetic resonance imaging (fMRI) Temporal resolution: accuracy when one can measure when an event occurs in the brain Spatial resolution ?accuracy when one can measure where an event occurs in the brain Invasiveness: Whether equipment is internal/external, ethical issue to do with this EEG: Or Electroencephalography Thinking and feeling: electrical activity in the brain, measures electrical activity in the brain. Electrodes on the scalp. (ERP?s recorded, event-related potentials) (see example on slide) Talks about differences in processing (N170, negative, 170ms after stimulus) Same thing many times and then an average is given. ERP: timing (latency) and amplitude of the peaks, polarity of the peaks is of little consequences as this is caused by how the cereal cortex is folded. Learning changes electrical activity in your brain Electrode location does not equal brain location (likely a number of different sources contribute to it) Pros/Cons: Takes many time and then average them together, hard to know where/what is generating a response, however the ms accuracy is quite high Very non-invasive can be used on children. Single-Cell recording Measures the responsiveness of single neuron?single, electrodes into the brain , record action potentials in the neuron (refresher on action potential) Spikes (AP per second) measured, used as a measure of activation. Can reveal stimulus selectively (what it reacts most to.) Pros/Cons Need to be inserted, typically only used with animals, however recordings placed on brain in surgery can produce the same effect Pros Information about the activity at the level of a single cell (rather than a population) Excellent spatial resolution Excellent temporal resolution Cons Highly invasive (can typically only be used in animal studies) No insight into the activation of the system Very time-consuming (often need to record from many cells to make sense of data) Electrically stimulating the brain (See video) create virtual lesions, important info to guide surgery (what disrupts difference processes) Transcranial Magnetic Stimulation (TMS) Virtual Lesions Weak electric current. Electric stimulation takes region ?off-line? temporarily Causes neurons to fire, disrupts activity in neurons making them unable to respond as they typically would Asked to perform task while activation is disrupted Researches can then infer the role of a given region in the task. Advantages over real lesions: Temporary, localized, no time for reorganization of function, same subject can be control and lesion. Pros/Cons: Pros: Good temporal resolution Experimenter controlled lesions Cons: Only sites just below the scalp can be targeted Stimulating one area could affect areas connected to it What questions asked through TMS? Is a particular brain area involved in a specific task? At what point in time is area involved in a particular task? Magnetoencephalography (MEG) Measures magnetic signals as opposed to electrical ones generated by the brain, very weak generated by electric currents. (associated magnetic field) See slide Pros/Cons: Pros: Better spatial resolution then EEG (2-3mm) Excellent temporal resolution (ms) Cons: Expensive Technically complex to maintain (cool and magnetically shielded) Limited availability Magnetic resonance imaging (MRI) Important advances in medicine in 20th century Nobel Prize in 2003 Exploiters magnetic properties of tissue Body full of hydrogen patterns, positive charge, tiny bar magnets At rest randomly arranged MRI focus patterns to align Radio frequency pulse knocks them out of alignment Emit a signal as they return back to alignment Can provide very high spatial resolution (3D) of brains structure after a few minutes Pros/Cons: Pros High spatial resolution (less then 1 mm) Non-invasive (many times) Cons: Can?t be used with metal in their bodies Expensive Can?t tell you about the function of the brain What sort of questions can we ask with MRI? How does the structure of the brain differ across individuals? What structures in the brain are impacted by different diseases? What parts of the brain are damaged in persons suffering brain damage? Structural differences in the brain: Schizophrenia and ventricular enlargement. (See slide) Functional magnetic resonance imaging (fMRI) About brain function and structure: Combines structural/functional MRI techniques Neurons require oxygen and glucose Detects oxygenation level in local cerebral blood flow. (magnetic properties of haemoglobin) Active brain regions, higher ratio of oxygenated to deoxygenated blood They each have different magnetic properties Measuring? Voxel (see slide for sample) FMRI? Not neural activity directly, measures metabolic changes correlated with neural activity, relies on haemodynamic properties Anatomical images, then functional processes on top of the images (not at the same time See slide for example. Subtraction method, subtract one from the other to see where they differ Subjects perspective: safety questionnaire, small confined place, motionless for 60 mins, strange noises (usually button presses) Pros/Cons: Pros: High spatial resolution in terms of both structure and function (<1mm voxels/pixels) Non-invasive Cons: Poorer temporal resolution then EEG (typically seconds instead of milliseconds) Can?t be used with person with metal in their body Expensive What questions can we ask with FMRI? What part of the brain is activated by given task? Does training/learning change which parts of the brain are activated while performing a task? Do different diseases impact the functional use of different brain regions? What are the neural correlates of individual differences in performance? What?s the best tool? Depends on the question you are asking Spatial resolution, temporal, time restrictions? humans or animals? How do they compare? (See slide) Summary: Cognitive neuroscientists use numerous tools to explore the relationship between the brain and behaviour Different tools have different strengths and weaknesses A given research question is often addressed using multiple tools in order to take advantage of the strengths of each tool. Rest of intro: See notes. Lecture 2: Short Term Memory: Used constantly Act on, then leaves or then retain the information Characteristics: Short term memory tasks: Serial presentation of short lists Fixed rate of usually 1 item/second or 1 item/0.5 sec Many different types of stimuli: words, letters...etc. (verbal) Presentation is either auditory or visual Immediate Memory Span: Number of items recalled in correct order 7 (+ or ? 2) Limit: Span is longest accurately recalled list (digit span sub-set of IQ tests) (testing, usually 2 items then 3...etc. stops when recall errors occur) Modality Effect: Recall is better after auditory then visual. Effects of Delaying recall of short list: Unfilled delay: Usually rehearse sub vocally (over and over) so still remember Filled delay: Distracter task given during delay period then drops rapidly with filled short delays of 10-30 seconds (demanding?less demanding tasks) Associations between STM deficit and other cognitive skills: Reduced memory span often found for Children with difficulty learning to read Children with difficulty in acquiring spoken vocabulary (specific language impairment Acquired dyslexia readers( e.g. deep dyslexia) acquired as result from injury/damage to brain Also some types of acquired aphasic patients (language disorders) Effects of a redundant: Prefix/Suffix (adverse effect) Prefix effect: read all of list items reduced by a redundant item at the beginning of the list : reduced recovery for whole list Suffix effect: Recall of final auditory list items reduced by a redundant item at the end of the list N.B. Suffix must be categorized as linguistic. Only ?human? sound impairs recall. So does lip-read silent suffix (rather than other sounds i.e. music/animal) In auditory. Coding in STM: Conrad (1964) Visual letter lists (not related to words or anything) Examined incorrectly recalled items Errors seemed to be acoustic confusions i.e. resembled errors found in listening to letters spoken in noisy background. Memory errors: T -> D, S -> F, B -> P similar errors to reading/acoustic (articulated similarity) Conclusion: Short-term memory uses an acoustic/phonological code. i.e. Sounds like Variables affecting STM: 1: Phonological Similarity effects: Adults recall fewer phonologically similar items then dissimilar items (even visually) letters words, pictures However recall for children under 5-6 years and congenitally deaf people?unaffected by this effect. (Discussed later in lecture) Only for people born deaf Conclusion: only hearing adults/older children use phonological coding in STM 2: Word Length: How long each word is within a list Adults fewer long then short words remembered. As many words said in 1.5-2 seconds So speech rate positive correlation to memory span Young children <6 years: recall unaffected but recall in generally low Model of Working Memory: Baddeley and Hitch (1974) Central Executive: Integrates info, planes/controls behaviour Articulatory/Phonological Loop Actively retains speech?based information Separated into Phonological Short-term Store which retains phonological information a Verbal Rehearsal Process - used to register and refresh information in PSTS Visual-Spatial Sketchpad: Visual info, sets up and maintains visual information (See slide explaining all of this) Rehearsal processes Speech Code for info to remember Effect of Irrelevant Concurrent Articulation on STM By uttering irrelevant repetitive speech while during list presentation Recall reduced However with visual lists, phonological similarity doesn?t reduce recall With visual/auditory lists words length does not affect recall if articulation occurs during presentation and recall N.B. * Baddeley et al. (1975) found a word length effect with auditory lists and articulation only if recall was in silence Conclusion: Concurrent articulation disrupts phonological coding and rehearsal. Effects of Irrelevant Sounds on STM: Sounds presented during STM task: 1. White noise 2. Tones 3. Speech 4. Foreign speech 5. Music - instrumental vs singing Speech (including foreign) and singing reduce recall, while the others do not Speech has obligatory access to phonological store (interprets few lines coming in) Filter out speech? Can occur if speech has taken place for > then 20 minutes, recall may improve STM Recall by Deaf People: Depends on ability levels, degrees of deafness and oral skills Acoustic similarity impairs hearing recall. (doesn?t affect deaf children) Visual similarity impairs deaf recall. (doesn?t affect hearing children) Most deaf people don?t use STM phonological coding, may rehearse with sign language Some deaf skilled speakers rehearse orally or using an articulatory code, some use finger specify English/ASL (American sign language) deaf people impaired on STM takes requiring recall order. (see picture on slide) Developmental Studies of STM: Memory span increases with age Children < 6years do not rehearse, can be taught but abandon Memory worse for visually similar then dissimilar pictures at ages <6 yrs Phonological similarity/word length impair recall ages >6yrs Children < 6rs don?t encode items phonologically/rehearse may relay on visual STM. Developmental Changes in digit span (See graphs on slide) Digit Span in other languages: Word length of number names affects size of digit span Welsh names longer than English, Chinese shorter Digit spans in Welsh Smaller, Chinese larger than English bilingual N.B. Problem with Welsh study - digit names were less familiar in Welsh - memory span for familiar words is larger. Silent Test of Phonology: Do picture names rhyme? (just pictures not words) Involves silent name retrieval and comparison Rhyme judgements by Speechless People: Anarthric (can?t speak) dysarthric (only some sounds) both from birth. They compared picture pairs and judged whether names rhymed. They could judge rhymes as well as controls who had speech Therefore speechless people can generate/compose phonological features of names STM in Congenitally Speechless People: Anarthric/dysarthric: Children saw picture lists either short or long words or phonologically similar. Recall worse for pictures with long/similar names suggesting use for phonological coding/rehearsal. The STM phonological code is abstract and available to speechless people Articulatory Loop and Vocabulary Acquisition Children with specific language repeating spoken non-words 8 years olds worse than normal 4 years olds Nonword repetition scores (4-5 years) predict: Vocab levels 1-2 years later, learning new names for toys Second Language Learning: Memory span and non-word repetition scores predict English (second language) word learning by Italian adults and Finnish children. The Visuo-Spaitial Sketchpad Tasks requiring visualization Concurrent visual tasks impair visual memory, spatial tasks impair spatial memory Concurrent articulation does not affect visual memory Visual STM Tasks (see the slide) Read block positions/patterns. Visual interference reduced visual memory, spatial interference reduced spatial memory. See other arrays on slide: Decide whether or not if the test array is the same or not Theoretical Development: STM recall uses language processing/processing system: Phonological Input buffer: Used to perceived speech (brief memory) (represent a sequence of sounds) Phonological Output Buffer: Used to speak (number stages?mental?retrieve words?then speech) Phonological similarity effects arise in the Input Buffer. Word length effects arise from capacity limits of the Output Buffer More elaborate theory: Extended Baddely model: Now includes Episodic Buffer Retain info from sentences, senses, more complex?all connected to LTM When given list of words to remember-make contact with information in the language system in the LTM. Lecture 3: Attention. Meanings: Alertness, concentration, selectively, control Focussed attention: Auditory selective attention: Dichotic listening task: How do we follow a conversation at a party: Two listening tasks, two messages, subjects told to listen to and shadow one message, to repeat later. Demo: (example given in class) subject noticed her own name , and random words Cherry?s finding: Later asked about unattended message, did not notice it was foreign speech or reversed But noticed it was a pure tone or a male/female voice Concluded that only physical characteristics processed in unattended ear. Moray?s (1959) cocktail party phenomenon: Subjects did not notice the same word 35 times But did notice own name mentions Incompatible with idea that only physical characteristics processed in unattended ear Bottleneck Models of Attention: Similarities: Structure of memory the same: Input?Sensory register?STM?LTM Sensory register: large capacity (fleeting) STM store (limited in capacity, but hold for as long as possible) All models assume transfer of info from sensory register to STM. They differ to where they regard the bottleneck to be and the nature of it. Broadbent?s filter model: Assumption: (See images in textbook) Stimuli gain accesses in parallel?sensory register Input filtered on the basis of physical characteristics (like ear location) Selective filter prevents overloading of the STM store (limited) One of the inputs allowed through a filter on basis of physical characteristics Inputs remaining in the buffer after filter undergo later (semantic) processing Evaluation: Consistent with Cherry?s findings Inconsistent with Morey?s Cocktail party phenomenon. Treisman?s attenuation Model: Assumptions: Instead of an all-or-none filter, attenuator turns down the processing of unattended info The thresholds of all context-appropriate stimuli are lower Partially processed stimuli sometimes exceed the threshold of conscious awareness, leading to ?breakthroughs? Evaluation: Own name processed is unattended channel, but because it has a low threshold (high salience) meaningful context also reduces threshold (see slide) When heard words worked into context of shadowed ear, were heard in unattended ear. Deutsch and Deutsch?s (1963) late selection model: Assumptions: Information is analysed fully, without attenuation Argued that attenuator is redundant; only idea of different thresholds necessary Bottleneck is late, at selection for action (response e.g. cannot shadow two messages. Evaluation: Can explain what the attenuation model can explain, so which better? Treisman and Riley?s (1969) experiment Two messages to two areas Shadowed one message, while there was a tapping response to words in either messages Predictions: Attenuation: Attenuated processing of non-shadowed message?worse in this message Late selection model: Full processing of both messages: Target detection should be equal in both Target detection 87% in shadowed while 8% in non-shadowed, consistent with attenuated Deutsch and Deutsch argued that shadowed message was more important because it had two responses and was therefore higher salience Johnston and Heinz?s ?flexible bottleneck view? Proposed location of bottleneck is flexible (early or late) Unattended message not always processed fully Johnston/Wilson (1980) List of words dichotically. Detect a target=semantic category in either ear. ?non-target? word presented cocindentelly with target ( same time) Focused attention condition: knew which ear targets at Divided attention condition: did not know where the targets at Critical targets: ambiguous meaning Meaning interpretation biased by non-target Either appropriate, neutral or inappropriate (basis against target) If meaning is processed in unattended message, meaning of non-target would affect target attention. Experiment: Results see graph on the slide) Under focussed, no different Under divided, appropriate has an effect Supports claim bottleneck flexible Meaning of non-targets affected target detection in divided, but not focused attention condition Processing of meaning of non-target when attention was divided but not when focused on the other ear. Evaluation: More steps of processing (physical?semantic) greater demands on capacity Selection occurs as early in processing as possible to minimise demands on capacity. Divided Attention: Multi-tasking: Driving and mobile phones (Strayer and Johnston 2001) Conversing on hands-free or handheld, phones impairs driving Stimulated-Driving task: Tracking moving target with a joystick Occasionally target flashed red or green?button if light goes red Single task or dual task: Phone convo: discussion about political events with another person in a different room Radio group (control) radio broadcast of their choosing. (See slide for data) Phone conversation worse than passenger talking next to you Suggest cannot dual task Practice: With practice, can perform other task simultaneously, such as driving and convo Performance, is said to become automatic But automatic processes are unavoidable and inflexible Automaticy: Characteristics: Automatic process (vs. Controlled processes) Are fast Required less attentional capacity ( other tasks simultaneously) Unavailable to consciousness Unavoidable (ballistic) Inflexible (once established difficult to modify) Shiffrin and Schneider?s experiment: Target detection task, memorise 1-4 targets (memory set) shown display conditioning 1-4 items. Respond as quickly as possible is display contained from memory set (examples on slide) Consistent mapping: Targets and distracters do not overlap from trial to trial Memory set: numbers only distractor: labels only Varied mapping: Target on one trial may be a distractor in the next trail Memory set numbers Distractions: letters and numbers Results (200 trials) Extensive practice with consistent mapping (RT does not increase with set size, so parallel search) Varied mapping (RT increases with set-size=serial search) Practice with CM leads to automatic processes (not regaining attentional capacity) Other characteristics of automatic processes: Inflexible: difficult to modify Extensive practice with CM: sonsonents B-L as targets and Q-Z as distractors Followed by reversal mapping of distractors and target sets?increase in RT worse than baseline Unavoidable: Intially searched for targets anywhere, but then told to search directly iun only one part (ignore others) Less able to ignore part of the display after CM then after VM (difficult to suppress) Learning example (see slide) Stroop inference effect: (Name colour of the word) Demonstrates unavoidable aspect of word reading Well practiced and automatic?unconscious How does practice make performance more automatic? Automaticy associated with gradual reduction in the use of attentional processes Shiffrin & Schneider: Extended practice with consistent mapping resulted in parallel search pattern But how does automaticity develop? Logan?s instance theory: Automaticity is memory retrieval Instance representation: each encounter with a stimulus is encoded, stored, and retrieved separately as a memory episode With practice performance becomes automatic when based on a single-step-direct-access retrieval of solution from memory (rather than rules or an algorithm which is time consuming.) Lassline and Logan (1993) Numerosity judgement (6-11 elements) Initially involves counting algorithms but with automatic Automaticity indexed by an absence of display set size effect (near zero slope between 6-11) Provided evidence for the instance theory by showing that a change in some attribute of the display increased the set size effect to original. Examples (see slide) Changing identity of element doesn?t matter, but the orientation does Instance theory: Automatic processes: ?Are fast because they require only the retrieval of solutions from memory ?Make few demands on attentional resources because retrieval of over-learned information is effortless Lecture 4: Episodic Memory: Original multistorey model of memory (1960) See slide) Different stores for different information Sensory Stores: decay Attention Short-Term: Displacement Rehearsal Long-Term: Interference Different types of memory still accepted, each can be impaired 1: Each store does not operate in a separate uniform function They have many functions (working memory consisting of many components interacts with LTM Long term memory, consists of many components as well: Procedural Memory: Actions, perceptual and motor skills Declarative: factual information Semantic: General knowledge about the world, concepts, language...etc. Episodic: Dated recollections of events (includes autobiographical) tied to temporal/spatital context. Reference to self What?s in semantic memory learned through Episodic memory (where you learned the information) To the extent the memory you received eventually becomes separate and you don?t remember where you learn it. 2: Information in STM transferred to the LTM via passive rehearsal Doesn?t work, STM makes contact with LTM. Instead now an emphasis on memory processes, not structure. (What to do to make memory more durable) Memory Processes: Encoding: transforming input into suitable formats, Held in durable way Storage: Maintaining info Retrieval: Getting info out of storage. Encoding: Maintenance (rote) rehearsal is not effective. Then what type of encoding is? Levels of Processing framework: Stimulus processed via a continuum of progressively deeper processing visual -> phonological -> semantic Retention is a function of the depth of processing (i.e. the final process is better) Experiment: Surprise memory text, asked questions about words to be presented.. Study phase (orienting task) Visual ?? Is the word in capital letters? ?? Does the word contain a letter e? Phonemic ?? Does the word rhyme with train? Semantic ?? Is it a type of animal? ?? Does it fit the sentence ?He dropped the precious ____?? ?? Test phase (memory test) Different questions biasing you to different parts of the word: Then a test phrase (study phase just answer) Main finding is that memory retention is best for Semantic. Very robust finding Spread of Processing: Elaboration Amount of processing of a different kind (within a level) also important Complexity of sentence in semantic processing, makes recall better, makes it more distinctive/unique. Retrieval: Testing Memory Varieties of memory tests: Free recall: e.g., ?tell me all the words I just showed you? ??Cued recall: e.g., ?tell me a word I showed you that starts with my-? ??Recognition: e.g., ?Was ?mystery? in the list I showed you?? ??Implicit retrieval: e.g., ?Tell me the first word that comes to mind that starts with my-? Memory performance depends on how it is tested Encoding Specificity Principle: Memory performance depends on the match between context cue between encoding/retrieval Thomson and Tulving: 1970 Strongly associated over weakly associated cue-target pairs. At test cued with same or different cues for recall of target Recall better when output cues matched input cues (See slide) Recognition failure of recallable words: Failing Should be eaiser: Tulving & Thomson (1973) 1. Ss learned a list of weakly associated cue-target pairs (e.g.,black-ENGINE) 2. Ss were then given strong associates of target words and asked to generate associates (e.g., steam-?) 3. Ss were asked to circle target words that were presented earlier 4. Ss were asked to recall target words given cues from Stage 1 Recognition in Stage 3 worse than recall in Stage 4. Fixed the idea of Encoding Specificity Principle (context of what you learned) Variation by Muter: 1978: Ss presented with surnames and asked to circle those they recognized as ?a person who was famous before 1950? e.g., DOYLE, FERGUSON, THOMAS ? 29% correct ?? Ss then asked to recall surnames from description + first names e.g., ?Author of the Sherlock Holmes series: Sir Arthur Conan _____?; Welsh poet: Dylan _____ ? 42% correct Better in recall then recognition, where you learned it. Context is also significant Interpretation: Tulving & Thomson interpreted it within the encoding specificity principle ??due to mismatch between encoding context (weakly associated cues) and test context (strong associates) Episodic memory is about events bound to context?match between encoding and test context important But some contexts effect different memory tests differently Different types of Context: Intrinsic context: direct impact on the meaning of a to-be-remembered item: Extrinsic context: Does not change the meaning of a to-be-remembered item Environmental Context: Change of environmental context: affects recall but not recognition Learn a list of words, encoding/text context: On land or 20 feet under water. Change of environmental context, affects recalls but not recognition (See slide) Dry is better overall Mood-State dependent memory: People should remember events more when mood at encoding matches mood at test: Ss shown a map ??Learned oral directions to landmarks (e.g., pub, library) in happy mood or sad mood ??Tested for recall of directions in happy or sad mood Free recall, cued recall (given outline of map) Mood match affected free recall but not cued recall Different types of Context: Intrinsic context (e.g., strawberry-JAM vs. traffic-JAM): ??Mismatch affects recall and recognition (consistent with the encoding specificity principle) ?? Extrinsic context (e.g., environment, mood) ??Mismatch affects recall but not recognition (or cued recall) Recall vs. Recognition: Recall involves self-generating retrieval cues: Matching extrinsic context may help generation of retrieval cues (use that to get info you need) Developmental implication: young children and elderly perform worse in free recall rather then cued or recognition Associated with prefrontal cortex (implication for children as witnesses) Not finished developing or declined. Harder to supply cues, maybe leading witnesses Inter personality exercise in DID (multiple personality) Nissen, et al. (1988, Brain & Cognition) ??Case study of one subject ?? Alice: 39 years old, nurses? assistant ?? Charles: 45 years old, aggressive heavy drinker ?? Bonnie, 36 years old, social, and others ?. ??Between-personality memory was poor with free recall ?was good when tested for recognition (of faces) Dominement personality (remembered more) Not one simple model Implicit Retrieval: Do not make explicit reference to study episode Example: Word fragment completion test ?? Complete this word fragment: _ys_e_y Word stem completion test ?? Write the first word that comes to mind that starts with STR- (strip, strap, stripe, ..) Multiple to only one solution, but no reference to study episode Repetition priming: Refers to facilitation as a result of prior exposure Better able to complete fragment when used in study Performances on implicit/explicit tests disassociated (normal vs. Amnesia) amnesia perform as well as healthy on implicit Example: Study phase: learn a list of words (e.g., tobaggan, mystery) ?? Test phase: implicit test vs explicit test Amnesia?s worse then controls on all conventional memory tests, but not word connections. Amnesia Impairments of memory: Causes of organic amnesia ?? Korsakoff?s disease: (chronic alcohol abuse) ?? Encephalitis e.g., Clive Wearing (The man with 30 second memory) http://www.youtube.com/watch?v=WmzU47i2xgw&feature=related ?? Temporal lobe removal ?? Stroke ?? Closed head injury: Most common cause of amnesia Most cases have other cognitive impairments (e.g., attention) (See Oliver Sacks (1985) ?The Lost Mariner? in ?The man who mistook his wife for a hat? Still believed 1945, and believed he was 19 (Showed him a mirror then forgot later) Symptoms: Anterograde amnesia: inability to retain new facts (after events) Retrograde amnesia: Inability to remember old facts Usually intact STM while paying attention Can learn new skills (new facts can?t learn) Show recognition priming in implicit memory?can always show? No Explanation of Amnesia: Explicit vs. Implicit memory: Amnesic patients have impaired explicit memory but intact implicit memory?? This is a description, not an explanation Amnesics perform poorly on some implicit memory tests (those that require memory binding) 1995: Subjects heard words spoken by different speakers: Implicit memory test: Word identification: Study phase: Subjects heard words spoken by different speakers ?? Subjects heard muffled words and identified them (same voice, re-paired, new) See slide, Recognition priming for new and re-paired Amensics lacked the ability to bind voices with specific studied words Not showing benefit of the same voice Binding of item and context: fMRI study Familiarity vs. Recollection: Familiarity: memory based on assessment of strength Recollection: ? memory accompanied by retrieval of context (correct source judgment; ?REMEMBER ?response) Reviewed functional MRI (magnetic resonance imaging) studies of familiarity vs. recollection 1. Item information (perirhinal cortex) 2. Context information (parahippocampal cortex) - 3. Binding of item and context (hippocampus) Different parts of the brain involved in recollection and familiarity Huppert & Piercy (1976) Korsakoffs and controls were shown pictures on Day 1 and different pictures on Day 2 On Day 2, 10 minutes after picture presentation, Ss were asked to respond only to pictures shown on Day 2 Cannot distqniush as well, distinguishing between different contexts, see different in graphs When respond to all pictures they perform the same: (see slide) Found Korsakoff patients were able to distinguish between never-seen pictures and seen pictures, but not between pictures seen on different days Korsakoffs are responding on the basis of familiarity of pictures?: failure to bind context with item Amnesiacs: Repetition priming Usually provide cues that guide performance (do not require distinguishing between contexts) Explicit memory tests required retrieval of contextual information bound at the episodes Requirement of the tasks Lecture 5: Semantic Memory: Semantic memory (In long term memory?declarative memory) Episodic memory: time, place, (autobiographical) Semantic memory: general knowledge, shared across populace (culture) Semantic memory (also called conceptual knowledge) is the aspect of human memory that corresponds to general knowledge of objects, word meanings, facts and people, without connection to any particular time or place. Knowing that you are Szechuan scallops at the Peking Restaurant in Cambridge last Thursday evening is episodic, not semantic, memory. Knowing that Szechuan refers to a province of China that food from this region tends to be spicy and that scallops are sea-creatures that live in brittle bivalve shells are all forms of conceptual knowledge. Memory for episodic events is not only specific to times and places; it is also largely specific to an individual. Conceptual knowledge on the other hand, is mostly shared across individuals in a given culture, although its precise scope depends on the individual?s experience. Patterson, Nestor & Rogers (2007), where do you know what you know? The representation of semantic knowledge in the human brain. Nature Review Neuroscience, 8 (December), 976-988 Impairments in Semantic memory?can?t comprehend meanings of words or pictures or express ideas (Semantic dementia impairments) Different from amnesia Experiments on Semantic Memory: Sentence verification?true or not, reaction time is measured Found lead to development of theories of Semantic memory Subject (like canary) Predicate (what it describes) Either set inclusion (part of a set) Or Property attribute (subject has something) Network Models: Semantic memory, network of concepts, interconnected or linked Concepts represented by nodes Relationships between concepts are represented by links (is a, has a ) see slide (basic models only two links) Hierarchal Network Model Collins & Quillian (1969) Concepts in a hierarchy Cognitive economy: property attribute stored only once at highest/general level (see slide) RT is a function of levels (evidence/date see slide) levels to transverse in hierarchy Only shows true, consistent with idea, higher levels longer RT. (with prosperities longer then supersets) Need to trace more levels/links therefore longer, consistent with idea Problems: Challenge to cognitive economy: Conrad (1972) RT data better explained in terms of frequency of co-occurrence of concept/property Manipulated subject-property frequency (associative strength: determined from norms) E.g. level 2 sentences (both of them) Hierarchal, should take as long for both Not the case. RT functions of co-occurrence (associative strength) only thing that determines it (how frequently property access to the mind) Challenge hierarchal More Problems: RTs did not always mirror hierarchical relationship e.g., A dog is an animal < A dog is a mammal ?(longer) dog is an animal should be longer Within-category typicality effects e.g., A canary is a bird < An ostrich is a bird (longer)should be same Faster negative judgments for closer concepts e.g., A canary is a salmon < A canary is an ostrich (longer) canary is a salmon should be longer. (no easy way to explain this one Negative judgements take shorter time here, shouldn?t according to model: Spreading Attention Model: no hierarchy Explains lack of hierarchal effect (no organisation) Links vary in associative strength/accessibility (longer/shorter links) Explains Conrad?s finding (typicality) Activation of a concept spreads to other concepts linked to it (priming) See slide ) Some nodes slightly activated by interconnected nodes Explains Semantic priming effect: Meyer & Schvaneveldt, 1976) Lexical decision task: Is ?truck? a word? car-truck < flower-truck Faster when preceded by something similar Effect used to study organisation of semantic memory May be automatic, study patients with impairments in semantic. Such as schizophrenic (showing effect is intact, so is memory) Feature Comparison Model (not network) Network models assume knowledge is prestored Feature comparison models, assumes knowledge is computed Concepts: Not represented as networks Rather composed of semantic features Multidimensional Scaling: (See slide) Rate how similar pair of concepts are Data represented in semantic space and underlying dimensions (features) are extracted. Part of data, infer dimensions people use (interpret what the dimension might be) i.e. What do the axis become? These are interpretations, how you interpret one theory. Idea concepts just points on the space (more than two dimensions) Defining vs. Characteristic factors/dimensions Defining features: essential features (i.e has wings, beaks..) Characteristic features: less important features (birds can fly..) Linguistic hedges: Such as technically, loosely speaking (see slide) Technically: subject has defining but not characteristic Loosely: subject has characteristic but not defining Two stage decision process (slide) Initially: all features of subject and predicate to determine overall similarity (if high/low..quick descion) If intermediate: Compare features of subject and predicate to those of characteristic Then longer then the first This explains: Typicality effect (on positive decisions): A canary is a bird < An ostrich is a bird Similarity effect (on negative decisions) A bat is a fruit < A bat is a bird; A canary is a salmon < A canary is an ostrich Problems Clear definition of defining and characteristic features is lacking (slide) Revisit issue later in further lectures Sentence Verification: Important data to develop theories of semantic theories Typicality effect (on positive decisions): Similarity effect (on negative decisions) Limitations of sentence verification data Reflect structure of semantic memory of the task process? Stimuli are words? Same as objects in general? Neuropsychological studies (wider range) Studies of brain-impaired patients show: Selective impairments of living things vs. Non-living can occur Selective impairments of living things more common than the other. Case studies (See slides) More category-specific impairment Preserved knowledge of body-parts together with impairment of living things Preserved knowledge of non-living things together with impaired knowledge of musical instruments How to explain? Selective Impairment of living things: JBR 23rd old electronics undergraduate: intact language - fluent speech Verbal and performance IQ normal Visual identification impaired (e.g., famous faces); picture-word matching impaired (Densely amnesic) Variety of tasks to tap semantic memory, naming/identifying pictures or give definition of spoken word...etc. Identification of picture: Animals/plants-6% inanimate objects 90%, but impairment with food category Description of objects given its name living things: poor Parrot - Don?t know Snail - an insect animal Eel - not well Inanimate objects: reasonably well preserved Tent - temporary outhouse, living home Briefcase - small case used by students to carry papers Compass - tools for telling directions you are going Definition of words from different categories (12 most common in each category) Expected is for norm: Better than expected: Clothing (e.g., shirt, trousers), furniture, kitchen, utensil Worse than expected: animals, insect, fish, flower, fruit, vegetables also precious stones, musical instrument categories impaired Organisation of Concepts with brain: Perceptual functional theory: category-specific impairments reflect different types of property Distributed-plus-hub-theory: There is a hub for each concept or object in addition to distributed modality-specific information Perceptual functional theory: Category-specific impairments reflect different types of properties that distinguish between category members Living things are distinguished from each other on the basis of perceptual(visual)properties (e.g., lion, tiger, leopard -> plain, striped, spotted) Non-living things are distinguished from each other on the basis of functional properties (e.g., chalk, crayon, pencil -> involve different writing surface) Explains more common impairments with living things in terms of visual properties being more frequent Visual prosperities being more frequent: Evaluation: Support for it, seem to go together Problems Patients do not necessarily show impairments of one type of knowledge (some prosperities neither sensory or functional but conceptual, patients also problems with these) Distributed-plus-hub-theory Hub for each concept or object in addition to distributed modality specific information Predicts occurrence of item-specific and modality-independent deficit of semantic memory Semantic dementia? (front-temporal dementia) patients show this type of impairment e.g., Mr. M (story) Do have episodic memory Loose specific output of concept (generalise to concepts) SD patients choose green celery and green pumpkin (common colour for vegetables in general...) Gradual detoriation: starts with specific then goes down Delayed copy drawing: SD patients draw general properties (see slide, duck with 4 legs, camel without humps) What is common to concept Evaluation: Hub for each concept or object in addition to distributed modality-specific info Why hubs? Provide efficient way of integrating our knowledge of any given concept Easier to detect semantic similarities across concepts differing greatly in their modality-specific attributed (e.g. scallops and prawns) Concept hubs are stored in the anterior temporal lobes (ATL) Degeneration of ATL is ?invariably evident? in semantic dementia (SD) patients Disrupting anterior-causes this Lecture 6: Word Recognition and Recall (Skilled Reading) Skilled Reading: Average adult person (12 years of school) Reading rate of 250 words a minute, (not only read, more difficult read slowly) Words in less then 100 ms (identify) less then 7/8 letter (responding takes long) Read aloud words 500?600 ms Word meaning 600-900 ms 10% + of adults in Aus U.K. U.S.A> below grade 4 level, functionally illiterate (develomentary deslexia) Writing Systems: Logographic: Characters represent whole words, morphemes. Egyptian modern Chinese...etc. Syllabary: Symbols represents syllabic units, Japanese kana: Hiragana & Katakana (words not in Chinese, second one different unitary system, foreign words) Alphabetic: Graeco-Roman alphabetic script represents phonemes (simple speech sounds) more than 1 letter can be used (see slide) Alphabetic Script: Represent phonemes with one or more letters Grapheme: letter or letter cluster represented simple phoneme (ck=k) Some languages: Represent a phoneme in more than one way such as ee/ea ph/f...etc. Have exception words which disobey spelling-sound rules (complicated for learning new language) English has Regular words obeying the rules and Irregular/exception words which do not Eye Movement/Fixation: Eye tracker: measuring the movements and when stops still, allow precise measurement, location/duration( either scenes, faces..etc) Finding out cog processes when they see the info Sentences are shown on the screen one at a time and are read silently. Reading Text: Fixations: stationary, periods of 250 ms (average) info take in (only here) Saccades: rapid/ballistic forward left?right movements lasting 25-50 ms, average is 7-8 characters Regression: backward movements right?left usually 2-5 character (average) End of line movements: regression to next line Movements blur across retina, very jerky but to not notice Info only during fixation, these process reveal immediate ongoing cog processes (because only properties) window on the mind. (See example) Longer words have more fixation (syllables and unfamiliar words) Perceptual Span (Size of) see slide) Characters are seen in details, 4 to left and 6 to right ( different in other societies, left to right or up and down which are different) Fuzzy up to 9 characters (where to move the eye Gaze-contingent display change: text changes when eye moves. Does the reader notice the change? No Spend longer fixating changed word? Yes, so some information has been in prior to fixation on word See slides for example. More fixation on uncommon or unpredictable words (even if common) So taking in meaning, spend longer...not consciously accessible (slow down) Fixation time for words depends on how common the words is and how predictable it is in the context (See slide) Sentence Wrap up time: Extra long fixation used to integrate current sentence and text meaning Recovery of Ambiguity: If interpretation needs revision, very long fixation then normal eye movements Or regression to ambiguous region Or re-reading of whole sentence with long fixation and short saccades (look to see how reader comprehends) looks like constantly updating while reading Inferences while reading: Draw inference which reduce later fixation time (slide) Cognitive Process: Eye movement studies challenged idea that comprehension occurs at end of a phrase or sentence (off-line hypothesis) Fixation times vary with the type of text. Number of fixations increases on rare and unpredictable words. Syntactic (grammatical) analysis and interpretation occur during fixations. The simplest interpretation occurs with eye movements disrupted (by regressions and short saccades) if revision is needed. Conclusion Sentence processing occurs ?one-line? word by word as each is fixated Evidence does not support ?off-line? hypothesis delay. Speed Reading: (Slide) If speed up better? Also purpose of reading tasks (different reading for different tasks) Comprehension better? Good reader faster than average, correlation not causation Normal vs. Speed Reading (slide) Fixated 50% fewer words with shorter fixations. Did not slow down at difficult pages, just more words fixations. They comprehended/recalled less than normal readers (same as skimmers ) but recalled gist better then skimmers (Isn?t surprising involved in course) Background knowledge also increases speed Dual Route Cascaded Processing Model: (DRC) see slide Visual feature uni?letter unit (then 3 routes) Computer program system use same processes as humans This model can?t read longer than 8 letters (possible more) Orthographic (spelling) phonological (word pronunciations/sounds) Some words 1 pronunciation together If recognise then to this system then to spoken phonemes of English. May go through meaning to read word if judgement needed Speak then (Contains exicitory or inhibitory, depending on meaning or form of a word) Forward/backward connection, further stimulation before Visual feature units - separate visual feature segments making up each letter Letter units - detectors for individual letters, recognize up to 8 letters at once. ? Orthographic lexicon - store of >8000 words ? Phonological lexicon - store of ~8000 word pronunciations. ? Grapheme-phoneme conversion - application of stored letter-sound rules - ph -> /f/, ee ->/i/. ? Phoneme units - spoken phonemes of English Connections are excitatory ---> or inhibitory ---feedforward or feedback. If don?t recognise word?Grapheme-phoneme conversion?application of stored letter-sound rules, Stored rules to read aloud words or non-words Provided follow some rules Dual route theories, Between 3 routes 1: a Lexical route (via semantics): look up written word in orthographic lexicon, access meaning, then stored word pronunciation Letter feature Recognition -> Letter Recognition -> Orthographic Lexicon -> Semantic system -> Phonological Lexicon-> Speech output buffer 2: a Non-lexical route: use spelling-sound rules to obtain a pronunciation Letter feature Recognition -> Letter Recognition ->Grapheme-phoneme translation -> Speech output buffer 3: Lexical route bypassing semantics: look up word, access stored word pronunciation Letter feature Recognition -> Letter Recognition -> Orthographic Lexicon -> Phonological Lexicon -> Speech output buffer Can use rules to process non-words (by Grapheme-phoneme) 3rd route usually for young children, or unseen words (reading without comprehension) Lexical Decision and Speeded Reading Aloud Lexical decision: Whether letter string is a word (50% words/non-words) 600-700 ms Speeded Reading Aloud: Words/nonwords one at a time. Vocalization latency?time between word onset and speech 500-600ms (around there) Stroop Task (see slide and tut) Automatic processing Interference effects with other words (slides and tut) Also Picture word interference (slide) Variables affecting Word Recognition: Lexical decisions and reading aloud RT depend on: Word Frequency faster RT for high frequency (common) words Regularity of spelling-to-sound correspondence faster reading aloud of regular (meat) than of exception words (steak, bread) DRC: Common words have lower thresholds in the orthographic lexicon Regular words have one pronunciation with both routes, exception words have different pronunciations but the lexical route ?wins ?competition over the non-lexical one Priming of Word Recongition: Effect of recognition of a word on a word following it. May speed up or slow down Effect of recognition of a word on a later word Repetition priming- Prime and target are the same Effect is facilitation, can be long lasting (up to 2 days) or brief (with brief primes). Semantic priming Prime and target are semantically related (bread - butter, nurse-doctor) Effect is facilitation, short-lived, a few sec. Masked priming A visual pattern (mask) precedes a brief prime (50 ms) followed by a long duration target (500+ ms). Ss are unaware of the prime but it can facilitate (speed) RT to targets Lecture 7: Cognitive neuropsychology (part 1) Cognitive disorders caused by brain damage in adults with previously normal cognition Advances knowledge of perception, memory and language Patients? problems and how to develop theories related to them Aim: explain performance through damage?referring to model of normal processes and to extend/test processing models (either discard or redfine theories to be able to explain disorders) Study: Simple case studies in 1890s and 1900s, descriptive, not really theories Group studies 1960s to 1990s, groups of patients with same syndromes, compared to control (try to understand all, not that straight forward) varied a lot, no one reason, which data right then? 1980s: detailed single case studies with experimental manipulation of tasks If theoretical basis, draw conclusions (try to) performance via tasks Memory, perception, language manipulated If data on one person, need enough for stats, when groups don?t need as many items Questions asked (an example of how to regulate theories, not even well-known facts...) Does visual object recognition involve a single common visual recognition system? If this theory is correct, a person with impaired object recognition should be impaired at face recognition and reading. The theory is falsified if some people can recognize objects and read are but impaired at face recognition. This condition has been observed ? termed prosopagnosia Others can read and recognize faces but not common objects ? visual agnosia So, the theory above is incorrect. Assumptions: Modularity: Module is domain-specific?if only specific class of stimuli, may be made up of sub-modules (I.e. face reading) Anatomical modularity: Cognitive modules not necessarily represented in small regions of the brains (distributed functioning) Uniform across ?normal? individuals: Cognitive psychology and neuropsychology assume this Subtractivity: Impaired cognitive system, same architecture as intact, but components deleted or impaired Associations and Disassociations Single associations: If performance on Task 1-4, are associated tasks may use separate processes served by adjunct anatomical structures (not connected but close) Single disassociation: If can do 1, but not 2, may suggest different processes Double disassociation: X can do 1 but not 2, Y can do 2 but not 1, Implication that tasks are different cognitive processes. Set of associations: Gerstmann?s syndrome Acalculia (inability to do arithmetic), finger agnosia (inability to distinguish among fingers), right-left disorientation and dysgraphia (inability to spell and write) Not related cognitively, damage to left-parietal cortex, where separate systems may be located (so if damage cortex, likely to damage a number of areas) can?t say separately, very different. Reading disorders: Developmental dyslexia Children of normal intelligence and regular schooling?but fail to read Distinguished from poor readers of low intelligence (born with or genes...etc.) Acquired dyslexia?s: Skilled readers that have dyslexia through brain damage Theoretical based work: Marshall & Newcombe (1973) Work has influenced research on deep dyslexia?leading to international researchers and a book (1978) even across writing systems and languages. (See book) and journal. Investigation of normal?to help research?to give theoires Diagnosing Cognitive Disorders: Cognitive neuropsychology has led to theoretically based Cognitive assessment tests ? Such as Language processing in Aphasisa BORB visual perception and recognition (see slide) Assessing reading skils: Letter knowledge: Recognition - Are these the same letter? A-a, A-b Naming - What is the name of this letter? A, a, B, . . .? Sounds - What sound does this letter make? A, K, B, . . . Word knowledge: Recognition - Lexical decision -Is it a real word? e.g., dog, deg, chair, thaip Reading aloud - Can you read aloud YACHT ? Is accuracy equal for exception and regular words?? ?Semantic decision - Is it a four-footed animal? e.g., horse, table, snake Letter-Sound-rules: Nonwords: Reading aloud - Can you read aloud SLINT, VIB ? Does it sound like an animal name? FOCKS, BRANE, FACKS Letter-by letter reading, Alexia can?t read, but can write. Letter reading precceds word reading Word length and exposure duration affect reading accuracy, espically for similar letters Writing is unimpaired Letter naming may be impaired Number sparing 1361 vs. Nine Impaired reading of handwriting and degraded print Associated symptoms: Colour agnosia (not colour blindness, can no longer identify colour with name) Right hemianopia (blindness in right visual sphere Write something down then come back and can?t read it Also many associated symptoms, for anatomical reasons Many cases: Acquired Phonological Dyslexia: Crucial Symptoms: Very poor nonword-reading Good word reading No semantic errors (some derivational/inflectional, such as suffix/prefix) Associated symptoms: Visual errors in word reading Impaired spelling/writing Aphasis can be absent Both acquired and developmental Part to model, the graphic-phoneme correspondence rules, seem not work (see example ) Surface dyslexia: Severely impaired irregular word reading Regularization errors such as bear?beer Homophone confusions sale/sail Phonological spelling: familiar?fermillyer Moderately accurate nonword reading Stress assignment errors, letter omissions, letter order errors: Stress on wrong part of word, ordering the letter wrong Both acquired and developmental: Reliant on letter-sound rules (lead to errors) Some variations, can be explained as above (only phoneme correspondence rules) Deep dyslexia: Semantic errors in oral word reading boy to girl Very impaired function word reading Nonword reading impossible Visual errors Derivational/inflectional errors Imageability /concreteness effect on reading?very limited word reading ability Very reduced immediate STM memory span Associated: Aphasia usual, typically Brocas, agrammatism Usually right hemiplegia Don?t know made error, lots of semantic errors Bad function reading (small words, but if not function has a go) Concret words are better, abstract impossible Not always present (the associated) usually right paralysis (so left cortex affected) Rasmussens Encephalitis and Hemispherectomy: (begins in one hemisphere?spreads, have to get ride of diseased hemisphere) 1. N.I. - illness from age 13, LH surgery at age 15 Severely anomic, agrammatic speech but used function words Lexical decision 72-89% correct, nonword reading 0% Poor word reading with semantic errors arm -> finger, pigeon -> cockatoo Visual-semantic errors: mud -> muck, swallow -> swan 2. H.P. - illness from age 10, RH surgery at age 15 Lexical decision 92-97% correct, nonword reading 80% Normal word reading Left-Right Hemispherectomy: Right hemisphere language: If removed left also able to swear, if left is removed. If abstract words, classified as non-words, description of deep dyslexia (why symptoms probs bad left hemisphere for deep) limited vocabulary, mainly common concrete words limited syntax -> agrammaticspeech recognizes letters and numbers limited oral word reading but distinguishes words and (unfamilar) nonwords ?no grapheme-phoneme conversion ability Left Hemisphere: (right hemisphere worse for spatial areas) normal speech, normal reading skills spatial abilities severely impaired after right hemispherectomy Think about the DRC model. Right=grapheme-phoneme correspondence skills, Left the others: lexial and semantic Explanations: 1: letter-by ?letter: Partly impaired feature > letter detectors, impaired parallel access from letter detectors > orthographic lexicon, so letter by letter identification used to access orthographic lexicon: bad feature detectors 2: Phonological dyslexia: abolished letter sound rule system (grapheme phoneme conversion), lexical route (mostly intact) 3: Surface dyslexia: impaired orthographic lexicon (or access to it), excessive reliance on letter sound rules (gpc graphemephoneme) conversion 4: Deep dyslexia: multiple impairments no gpc, impaired semantic access to phonological lexicon, or possibly right hemisphere Symptoms vs. Syndromes: The description of syndromes lead to focussing on cognitive mechanisms Now more emphasis on describing disorders with reference to symptoms and their mechanisms. Thus there can be differences among people with the same syndrome E.g. Some patients with letter-by-letter reading can have some covert semantic access (patient thought ? church? had something to do with religion and some have access to orthological lexicon (can perform lexical decision.) Lecture 8: Cognitive neuropsychology (part 2) DVD: Memory All molecules in brain, replaced..but still memory Can?t remember things happening (amnesic) what happened? Classic amnesic: virus infection, stroke, car injury...etc. Can remember the past (used to have a good memory) Can?t live alone, disabling condition...how to see own life Can still remember walking...no new memories, damages Respond to cues, skills in cerebellum...automating Conscious elsewhere, consciousness essence of self Amnesia roles of this patient (encphilius) hippocampus, frontal lobes?repeating, highly emotional...blinkered moment. Goes in perfectly well, can?t keep it Almost same thing, doesn?t acknowledge previous dairy, every repeating moment Love is raw...encoded elsewhere in the brain... Dealing with lucid person, no knowledge of own life, way frustrated (relate to past, present, future) can recall music part of memory, but can?t active other things...has skill still, confusion afterwards, because has control while doing music Hippocampus: Conscious memories: inside temporal lobe, establishing relationships, putting things into context. Providing clarity, and a record of events...like a diary As old: Senile dementia, Aletizmiers disease: long degenerative disease (uses tricks to get around memory loss) loss of personality?nucleus bassamius, (which bits are stored) interfere with chemicals, eventually dies...doesn?t know which to remember. Can affect certain aspects (some better, preserved then others) Memory Disorders: Auditor verbal Short term memory disorder Long term memory disorders Amnesic syndrome Semantic memory disorder Impaired Auditory-Verbal STM: Impaired Auditory-Verbal STM: Defining Symptoms: Abnormally small digit span, but normal LTM recall (abnormal for age/intelligence) Poor immediate repetition of lists of words or letters or longer sentences, better for short sentences Not as common, KF had a digit span of 2, and similar and word spans, deep dyslexia, also bad STM (motorbike accent) 2,3 better when visual Visual lists are better recalled then auditory?with normal STM recall auditory lists better No effect of phonological similarity on list recall Don?t seem to convert into speech coding (some other problems) Associated Symptoms: IN free recall with slow list presentation, there is normal primacy, but a very reduced or absent recency effect. Serial position function for normal memory?immediate vs. Delayed recall Slow down, good for first (can be normal speech wise...also can be affected) Speech perception and production can be normal Sentence comprehension deficits with more complex types of syntax can occur. HB: report bilinguial Austrian WAIS IQ 122 (after stroke) so smarter beforehand LH parieto-temproal stroke at age 60 STM impaired-digit span=3 and span low on other types of items Better with visual then auditory stimuli No effect of: Phonological similarity on word recall Word length 1 vs. 3 syllables for auditory lists Word length effect was found for visual lists- worse recall of long words?a visual length effect. (reffering to longer visual stimuli) 2: The Amnesic Syndrome: More severe condition, can?t learn new things, have good long term...memory from before...(spouses death, nope after amnesia) Symptoms Inability to learn new tasks, names, faces, facts, places, Inability to remember anything for more than a few minutes Cannot remember: having lunch, events on TV before an ad, being told of the death of relative Repeat the same remarks and stories in conversation, and cannot live independently Many causes: (symptoms more important) NB: digit span and STM list recall are normal: Surgical excision of hippocampus and amygdala (to relieve sever epilspsy or tumor removal before they knew what this dead) Brain damage by accicdent Herpes Simplex Enc ephalitis (very rare, if diagnosed very quickly, stop symptoms) Korsakoff Syndrome: severe chronic alcoholism and thiamine deficiency causing brain damage?memory and other cognitive defects?more younger cases these days Amnesic Memory Performance: Retrograde amnesia: less dramatic in brief concession Loss of memory for events prior to brain damage (weeks to several years) Antrograde amnesia: inability to learn and remember after brain damage, Normal Receny effect but no Primary effect in list recall Preserved early Cognitive and Motor Skills Piano skills kept, or second language, understanding technical information Residual learning capability for motor skills Purist rotor, keep indicator on moving object, can get better at these tasks, not as fast thought...so do have procedural memory No conscious memory of previous experiences with the tasks Can do mazes etc if same next day, do better Tasks which amnesic Ss can learn have the following features: 1. Task demands can be immediately understood. 2. Responses are in the S existing repertoire. 3. The goal of the task can be optained without reference to a specific past even Theoretical Accounts: LTM System Deficit Episodic memory impaired, intact Semantic memory (knowledge base) Residual learning suggests some preservation of episodic memory. Procedural memory cannot be inspected consciously, includes Cognitive and motor skills e.g., we cannot describe how we ride a bicycle Declarative memory can be inspected and reported e.g., memory for episodes and facts ? Declarative memory impaired, intact Procedural memory Problem: amnesiacs can recall some facts and the declarative procedural distinction is not clear cut. Relatively preserved, when originally learned, very conscious of it...once learned can perform, not clear cut (concussion may recover lost memories) Semantic Memory Disorders: All can go the either way Patients have: normal STM (shortterm working memory) and normal Episodic LTM* can recall past events, learn lists but comprehension deficit for single words, objects, pictures of objects, etc. * Herpes simplex encephalitis patients have amnesic syndrome and semantic deficits Semantic Dementia: progressive loss of semantic functions: Symptoms: Severe anomia for concepts, objects, people , very impaired spoken and written single word comprehension Impoverished general knowledge Normal syntax and phonology Normal perceptual skills and nonverbal problem solving abilities relatively normal autobiographical and day to day memory Surface dyslexic reading disorder Particular case where loosing different things: perhaps become encoded in Semantic memory, all modalities Cause: progressive atrophy of the anterior temporal lobes usually worse in the left, Age of onset from 40 yrs on Can?t read aloud irregular words are conscious of this happening, know can?t do it or other things, progressively getting worse (not known cure) Semantic Memory Disorders: More specific: 1: Category-Specific: Different patients presented with different things Abstract vs Concrete Words patient could define ABSTRACT but not CONCRETE words e.g., vocation, pact, arbiter,.... vshay, needle, poster,... Animate vs Inanimate concepts e.g., Impairment on Objects, Colour , Body Parts, and Animals; Herpes simplex encephalities impaired on Living Things and Foods (e.g., fruit and vegetables) 2: Input-modality specific: pictures not spoken: Semantic task can be performed with pictures but not with spoken words, or vice versa e.g., S can report lots of properities of dolphins shown a picture, but is unsure if it is a fish or bird 3: Attribute specific: Either visual or spoken: Perceptual vs non perceptual atributes Perceptual atribute impairment: e.g., S can respond: lion? to ?Which animal is king of the jungle?? but not ?zebra? to ?Which animal is like a black and white striped horse?? Can?t do descriptive, can do perceptual, Disorder of Visual Object Recognition: How info represented: AC 67 yr old clerk at NSW Rail, had cardiovascular and cerebrovascular disease then a stroke and lesions in the left middle cerebral artery (and smaller earlier lesions). Reading and writing almost completely abolished, He did not know that ?A? and ?a? were the same letter, but knew ?A? and ?A? were the same and ?A? and ?E? different (this requires just pattern matching). Could copy letters, so could see normally. Had word finding problems and impaired word comprehension. Done without knowing letters Very interesting, tested on Animal legs, only chance, not specific to animals or to only legs at all Perceptual: couldn?t do is it round or to colour questions, suggests lost of Perceptual, but non-percetual such as is it Australian, or dangerous...very good. But couldn?t tell you what is was Preserved perceptual knowledge, only lots of visual This pattern of responses indicates that object knowledge is not represented in a single store. It suggests separate stores for visual, auditory, and other forms of sensory and non sensory information i.e., separate stores of knowledge in people with intact cognitive processes. Leads to a model of Semantic representation (see slide) Simpler version, complicated includes a lot more Can link to other levels (lemma level: word knowledge) see reading from before, if even more generally link patterns Conclusions: Memory deficits fall into distinct groups: STM (short term working memory) deficit Episodic LTM impairment Semantic memory deficits. As for acquired dyslexia, memory deficits are understood with reference to theories of normal memory Provide some support for multiple memory systems theories Studying memory disorders: contributes to theoretical development leads to differing approaches to remediation and specific memory aid Lecture 9: Visual Attention Spatial visual attention Visual info in large array. Such as Conjunction search: Two or more features How do we allocate attention/locate objects in complex arrays/scenes Treisman?Visual search task for experiment: Ss search for a target among distractors. Set Size = total n(items) in the display p(Target) = 50% Ss judge whether target is Present or Absent. Usual R measure = Correct Decision RT. How is decision time (RT) affected by the number of items searched? How large set size is varied. (reaction time is measure) Correct diescion RT is measure Targets easy to find or difficult (see slide) Some call attention to themself, others blend in or more difficult Conjunctions: Becomes more different A red circle ?pops out? if distractors are all blue, or are all squares Search time for a red circle (conjunction of features) among red squares and blue circles varies with set size. Treisman & Gelade (1980) found RT increased as a linear function of set size for conjunction targets argued that conjunction searches are serial and self-terminating Self-terminating (done with non-trials, trials and no-trials take longer?yes quicker, because when found target terminate) See slide: by Search process indicated by slope of RT and set size function. Single feature not longer. RT increases with more set items and with conjunction What features are basic features? Colour Orientation Curvature Vernier offset: Whether lines meet together when you line them up in field of vision (depends on how similar they are) Size Motion Shape (? not all do) Depth Gloss Easy to see when they are the only features, lead to Feature Integration theory: Basic features in parallel by pre-attentive processing?gazing at the scene Pre-attentive features ?free-floating? attention to bind them together (not bound till focus) This operates serially, so conjunction requires serial and self-terminating Feature of attention can lead to illusory conjunctions: These can occur with brief masked displays and attention focussed on the ends of the display Ss report the numbers first and then the coloured shapes (slide) Responses are mostly correct but Ss sometimes report ?red circle?, or ?blue square? If subjects told they made a mistake?believed they were correct (only attention elsewhere and display brief) Search Asymmetry: Target unique feature pops out from distracters. If reversed search becomes serial as now difficult (see slide) Search takes longer as the distracters are increased. Relationship important. With unique feature easy to find (see slide) Some problems for Feature Integration theory: Search is not always clearly parallel or serial Wolfe suggests using the terms: efficient vs inefficient search (for parallel/serial). So, a common terminology can be used by different theorists Basic features? Colour is a basic feature only if colour set is small (<~9) and colours are dissimilar Prototype colours are easier to find among non-prototype distractors - red target in magenta distractors. (see slide) So Wolfe?s Guided Search Theory: Attention needed to identify targets Preattentive basic features, e.g., colour can be used to guide search For feature search (e.g., ?red?) attention is directed to the item immediately For ?red circle? T, search is directed to red items Slow (serial) search involves repeated selection until the target is found at random RT by set size function: Target: Feature, Conjunction or Spatial Configuration Some conjunction targets are easier to find than others Targets defined by a spatial configuration and other features (e.g., colour, occlusion) can be very hard to find Efficient to inefficient, see slide for examples Cost of missing something, can be disastrous Detecting distressed or drowning swimmers in pools Distress, treading water or normal swimming Computer vision scientists and engineers are developing visual surveillance systems for use at public pools to assist lifeguards (see slide Lifeguards attend to behaviour of swimmers on the surface of a pool. Unconscious swimmers at the bottom of a pool can be missed in crowded pools. Do find those on bottom of pool Special problem: Rare target incidence: Wolfe et al. (2006-) studied weapon detection in X-rays by airline security staff. Rare targets are often missed. How can detection rates be improved? In real life the target are quiet rare One of the ways to help is to exposure the person to a target/weapon before this shift to prime them Also problem for other things, such as anagrams, craking...etc. (not problem in most life cases) Visual Attention Problem: Looking at particular target and miss something clearly present They were attending to the heads-up instrument display. Can close attention to one part of a display prevent seeing another part? Missed something very obivious Intentional Blindness: Precise when bit attending properly, Mack and Rock: Ss do not expect or look for the object of interest They make a difficult judgement - decide which was the longer arm of a 200 ms cross changing on each trial On the 3rd or 4th trial an extra stimulus is shown Ss are asked if they saw anything other than the cross If they say ?yes? they are asked to describe or select it in a recognition test Ss are tested only once with the extra object 5000 were tested (over 7 yrs)! Only do it once, trying to find conditions that cause this (see slide) 60-80% failed to detect the diamond shape at fixation All Ss could detect it in Control experiments Did it with heaps of manipulations (and still found connections) even though 100% saw it in the control Here background circles were green and one changed from green to red on the critical trial 44% of Ss did not see the change 100% saw it in the full attention condition Mack & Rock concluded that attention can be actively inhibited and there is no conscious perception without attention. Inattentional blindness for own name? Mack & Rock (1998)showed Ss their own name on critical trials. Some Ss were tested with other names or common words, House or Time. % Ss who saw ?something?: Own name = 88 Other name = 65 Time/House = 50 Some blindness for own name, Miss words but not own name, saw something, so salient easy to remember, less affected by this effect Attentional Capture: Although we miss things in front of us, irrelevant items can attract attention. Abrupt irrelevant changes in a visual scene can capture attention, e.g., a new object in an unoccupied area a new colour a sudden motion Visual attention, eye movement in Parkinson?s difference between them and normal Deijen (2006) But also alone with other non-clinical groups 1. Ss see grey ?8?s in circles 2. 5 circles change to red with letters . one changes to ?C? or reverse ?C? 3. Ss decide which ?C? it is 4. On distractor trials an extra circle. appears in a new position This distractor disrupts ?C? detection - slower less accurate responses. See slides This new shape...whatever distracts detection of (C) even though in irrelevant location Also Parkinson?s patients were more slowed down When Does capture occur? Attention widely spread across visual field (inattention is focussed) Can?t cross inattention blindness and attentional capturing In the ?C? detection task the ?C? can occur in any of the 6 locations and attention is not focussed on a small area. Temporally distributed attention: The attentional blink: Task: Ss search for 2 targets in a sequence of 10-20+ items shown at rates of ~10 items/sec Possible Targets - report 2 red letters (T1 and T2) in list of black ones or - report 2 letters in list of digits. T2 detection is impaired if T2 occurs soon after T1 (~ 500+ ms) but then recovers. The transient loss of detection is the 'attentional blink'. Sequences can be digits, letters, words or patterns. If closer together, better at reporting both (not real blink) Chun?s metaphor: Fishing in a messy stream Boots, tyres, fish, rubbish float by. The fisherman sees a fish and dips net in to catch it. It takes time to remove the fish and put the net back in. So, a 2nd fish will be missed if it comes soon after the 1st. If 2 fishes come very close together both may be caught in the net at once. If T1 and T2 close (yes) if far away then also yes (see slide for example) most stimuli like this ) For adjacent T1 and T2 - Lag 1 - there may be no ?blink? known as ?Lag 1 sparing? Blink occurs when at least 1 distractor follows both T1 and T2 Theories of attentional blink: Gating - early selection model -Raymond, Shapiro & Arnell (1992): T1 is detected preattentively triggers an attentional response to assist target identification. This closes a ?gate? regulating flow of visual information, so T2 is not processed Bottleneck model - 2-stage model Chun & Potter, 1995 (this one is better) Stage 1 - all stimuli are identified but Stage 2 is needed for report Stage 2 - registration in STM - slow serial process - T2 is lost because Stage 2 processing is busy with T1. Problem for Gating Model: Meaning of distracters and unreported T2 has been accessed: Word in moving to second target e.g. chair, cat, dog (T2) cat primes dog so remember that one petter, early stage identify these. Failures of attention : Summary: Failure of Attention Task Rare targets missed Visual search Inattentional blindness Spatially focussed attention Attentional capture Spatially distributed attention Attentional blink Temporally distributed attention To previous lecture: Semantic Memory, see slide for big model of speech and written word recognition. Input may be different but all of them converge on Semantic System DRC ?reading components, in larger model STM?may involve impairment of phonological input and or output buffers?Sequence errors...etc. Lecture 10: Comprehending and Remembering: Immediate memory for a single sentence: (Better then unrelated material) Memory span usually small (sentence only straight away, power of after brief delay 8th of a second per word, remember quite well (Straight away) Verbatim immediate recall of a single sentence greatly exceeds memory span for unrelated words. People can recall a spoken sentence of 15-21 words in length with close to 100% accuracy Butterworth, Campbell & Howard (1986) Visually presented sentences ? 1 word at a time ? at rates of 8-10 words/sec are also well recalled Potter (1999) However, verbatim memory for sentences and passages of prose is poor if tested after a brief delay Early Research: Cambridge study, Male students, area quite old Asked Cambridge students to read and later recall an unusual rather disconnected story with a supernatural element - The War of the Ghosts. (Indian war story) Recall was brief, information distorted, had intrusions based on Ss own experiences, even worse after delay. Was this due to the unfamiliar content? No, because recall was similar for passages about familiar topics ? Cricket match. Conclusion: memory was an active, constructive process based on scanty memory traces, not a literal record. Memory for Wording/Meaning: Consistent with Conclusions When sentence repeated, asked if heard before...even though active listening, their understanding for what is meant is good, wording is not good Memory after Listening to Prose: History, science & folk tales Recognition memory after delay Delay period: 0 80 syllables (~27 sec) 160 " (~46 sec) Recognition of Altered Sentences - people accurately detected changes in: Meaning at all delays Wording at 0 delay only. Memory for gist is good, verbatim memory is brief. Constructive Processes: Clear sum to understand it (if tired to memorise better) Then whole sentence is intergrated, remember topic to topic, also catch sentences with flase information, perhaps happens because statements contain visual image, but also happened with abstract info so Memory constructive Ss heard several statements about different topics randomly mixed. e.g., The ants were in the kitchen. The boulder rolled down the hill. The sweet jelly was on the table. Whole Topic: The ants in the kitchen ate the sweet jelly which was on the table. False recognitions occurred to longer sentences with more ideas. Thus, Ss integrated meanings from separate statements. Similar results occurred for abstract and concrete ideas e.g., Whole Topic: The intense desire to be successful can determine all personal actions. Inferences: are drawn during reading or listening -become part of the memory representation. Types of Inferences Lexical: inferences about word meaning for ambiguous words. (example such as actress, when actually operating theatre) Temporal and Spatial/Bridging: when and where events occur. (2nd statement to location of first statement) Extrapolative/Elaborative : extra events from world knowledge added to link two sentences (extrapolating what has happened or will happened) Inferences in Memory: With an inference or not (not picture) Thought sentences were the same, not consistent with context/meaning stored Remembering based on situation/meaning Ss heard sentences with or without a potential inference. Potential Inference (PI) Three turtles rested on a floating log and a fish swam beneath them. No Inference (NI) Three turtles rested beside a floating log and a fish swam beneath them Recognition Distractor Three turtles rested on/beside a floating log and a fish swam beneath it .Ss who heard the PI version accepted the distractor, Ss who heard NI did not. Provision of Context: confusing passage unless has context If picture as well, easy to understand, if without harder Perspective then, information related to perspective remembered: Thematic title or Picture e.g., Washing the dishes Picture of serenade Context before passage enhances comprehension and memory Context after passage increases comprehension ratings, but NOT recall. Asking Ss to take the perspective of a burglar or estate agent (selling a house) affects recall of information about a house. Role of Schema: Schema : is a body of knowledge about classes of objects, common events, or people. Story Schema: is a set of expectancies about the structure of stories. Story Grammar: is an analysis of story structure: the components and how they are linked. (Rumelhart, Mandler) Mandler?sStory Grammar represents narrative stories, folk tales, TV serials, fairy tales. has setting, episodes, outcome. How stories likely to be structured: Ideal form of a story All fashion similar structures, a developed structured, and basic overview (see slide) What happens when messing around with grammar of the story? Thought children less affected, actually more affected Once upon a time...activates the story grammar Stories structured according to the grammar are well recalled. e.g., folk tales, fairy stories, fables, TV serials for children Disruptions to structure reduce recall, especially for children. Interleaved episodes are separated spontaneously in recall by adults and children. From an early age, people store knowledge about the likely form of stories and use it to interpret and recall new stories. Scripts: Describe typical situation?s Classified in order: are stored stereotyped knowledge about common events, include action plans for likely sequences of events. Normative Data on Scripts e.g., Going to a Restaurant Most people (>73%) include: SIT DOWN LOOK AT MENU ORDER EAT FOOD PAY BILL LEAVE in the order above. Memory for stories are based on scripts: indicates people have this stored knowledge Omitted script events may be inserted in recall, or falsely recognized. Incorrectly ordered events may be re-ordered in recall. Motive or Character Schema : Beliefs about motives or character may act as a schema to influence memory. The "Soap Opera" effect (Owens et al, 1979) Recall of mundane events in the life of Nancy was influenced by prior information on motives and feelings attributed to her. Ss recalled more about the ?pregnant? Nancy?s day. False inferences implied by the ?motive schema? increased in recall a day later. Conclusions to draw from these studies: Memory for prose is an active and constructive process. Inferences are drawn during reading or listening and later during recall of text. Verbatim storage is relatively short-lived. Knowledge structures, e.g., schema, scripts, story grammars, are automatically invoked in encoding and retrieval. Recall intrusions and false recognition consistent with knowledge structures occur in memory tests, especially after delay. Comprehending Pictures and Scenes: Objects in scenes: Support- do not float about, are supported by surfaces. Interposition- occlude objects behind. Probability- occur in some settings and not others. Position- occupy specific positions. Size- have a narrow range of familiar sizes What if come across differences Identification of objects in Scenes: Beforehand, having appreciate contrext...etc. Intact and Jumbled Scenes were shown for 50 and 100 msec. Pictures were colour photos of streets, kitchens, offices, etc. Task: Is there a fire hydrant in the picture? Ss were slower and less accurate at identifying objects in Jumbled scenes. (see slide) Relational Violations seen earlier: (arrow pointing where) Ss saw an object name, then a 150 msec scene then a location cue in a mask. The target objects violated from 0 - 3 relations Violations reduced detection accuracy and slowed RTs Detection was worse with more violations See slide Seeing pictures but remembering Scenes: Half the time changed the picture Also done with pro artists (always more background, more of scene then in the photograph) Study Task: Remember photos of objects in a background. Memory Tests: Recall task: draw pictures Recognition task: Are pictures the same, closer, or further away than originals? Recall: Ss drew more background than in originals Recognition: Ss chose wider angle pictures Both tasks showed BOUNDARY EXTENSION See slides LTM for Scenes and Objects: Photo, (not ,literal memory) Some scenes several days ago, chance would be 50% surprising because of Change blindness Shepard (1969), Standing (1973) Found very high levels of recognition memory for scenes shown for 5 sec each Standing found 90% recognition for 10,000 scenes Recognition task: 2 alternative forced choice Distractors dissimilar to old items Recently Brady et al. (2008) found 87-93% recognition for pictures of 2500 objects shown for 3 sec each However, surprising memory lapses are also observed Recognition memory with different types of distractors: Unrelated Exemplar Same item, different state % Correct recognition scores See slide for results Memory failures: McConkieet al. (1979) CaSe changes not noticed when reversed between saccades , i.e., (See slide Even the experimenters did not see change, thought equipment was defective! O'Regan?s(1992) view: there is no internal representation of the world, the world is continuously 'out there?, serves as an 'outside memory'. (bit extreme) No info during saccade reading/eye cotingent during change Failure to detect Scene changes across Saccades Change blindness: Grimes (1996) Ss were shown 10 sec. scenes which changed during a saccade. Instructions: Study scenes for a memory test and press a key if picture changes. Detection of changes = 33% Some Changes % Misses A 25% size increase of a building 100 2 men change different style hats 100 33% of 30 puffins disappear 92 A pink swimsuit changes to green 58 Two cowboys change heads 50 Disneyland castle rotates from L to R 25 Methods that cause this: 1. Flicker task - scenes and blank fields alternate 2. Scene moves - forcing eye movement 3. Scene interruption by visual mask field 4. Mudsplash? appears on scene then disappears Change blindness is less likely to occur for: an explicitly attended object a previously described object objects/scenes familiar to an expert Memory for Familiar object: Not good as would expect it to be: How well do people remember coins? Can you draw both sides of a $2 coin? Recall Only 1/20 Ss recalled 8 features of a penny. Mean recall = ~ 3 features. Nickerson & Adams (1979) Recognition Only 15/36 Ss chose the correct one from 15 drawings. British studies had similar results for coins & stamps Processing Info in Scenes: As with stories stored mental representation, need to be promoting memory: Top-down processes(higher order knowledge about scenes - schema) influence object recognition. Each glimpse (fixation) is interpreted with reference to an abstract mental representation of the environment. Memory for Objects and Scenes Early scene recognition memory and a recent object memory study indicate good long-term memory for pictures of objects and scenes. But change detection studies show poor memory for detail. Little detail is retained for some familiar objects - coins, stamps. What causes the discrepant findings? Focussed attention, expertise and verbal description promote memory for objects and detail in scenes. Lecture 11: Concept formation and Reasoning: Categories useful? Why do we categorise the world? Class of stimuli that are treated in a equivalent manner: ?Categorizing serves to cut down the diversity of objects and events that must be dealt with uniquely by an organism of limited capacity Bruner, Goodnow & Austin, 1956 Generalise from on experience to another (not each stimulus as unique event) Natural Categories: (See slide for what categories are not) What occur in natural and everyday language, different to artificial categories, contrast between theses Rosch?s work suggests natural categories have a correlational structure, artificial do not. Category learning study: Artificial category study. 81 4 dimensions (see slide) Concept universe: 4 features (3 x 3 x 3 x 3 = 81) Shape (cross, circle, square) Colour (white, black, grey) Number of shapes (1, 2, 3) Number of borders (1, 2, 3) Must try and figure out category experimenter has in mind (certain features) try to define the real Feedback as in or not, learn slowly (contrasts to natural categories) Concept defined by a rule Ss task is to work out the rule E is using e.g., ?Black cards with circles? :Colour: black Shape: circle Features may be combined arbitrarily Natural: Some features go together, others do not (correlational) Characteristics of Natural: fuzzy sets, family resemblance and internal structure Fuzzy Sets: Boundaries are ill-defined: different from artificial: Asked Ss to decide category membership for each exemplar-category pair (e.g., chair-furniture, cucumber-furniture, bookend-furniture) twice For intermediate exemplars (bookend-furniture), both between-subject agreement and within-subject consistency were low On the border: many disagreed, even same person inconsistent Family Resemblance: No simple attribute is shared by all members of a category, but each member has at least one attribute in common with others: Wittgenstein (1953): ?game? has no attributes that are shared by all members of the category Involves boards (chess, monopoly) but ball games? Have winners and losers but Solitaire? Played for amusement but sometimes not? Natural categories have family resemblance: Typicality: Vary in how typical a member of category is Rosch& Mervis(1975): Ss rated how good a member of the category an exemplar was (1-7 scale): typicality is not uniform (See slide) Different to artificial: Which are equally typical; Typicality/family resemblance different Family how many features of a category a member has, typicality how typical a member is Rosch& Mervis(1978): Ss listed attributes possessed by each exemplar of a category (e.g., bird) robin: flies, sings, lays eggs, is small, .. Derived the family resemblance score: sum of the weighted scores of attributes Greater weight given to attribute shared with other exemplars High correlation between rated typicality and family resemblance score, (see slide) lead to Prototype: Member of a category with highest family resemblance score? prototype Different to artificial, all typical resemblance. Natural: internal structure Internal structure: Typicality gradient Some members of a category are better exemplar than other members (e.g., apple is a more typical fruit than a coconut) Better exemplars share more attributes in common with other exemplars A category is centred around a prototype Basic Level categories (Within hierarchy, most fundamental) Hierarchal model (see slide) This is the most inclusive level which attributes are common to many members of a category and few in other categories. Most informative level: Natural language, use in language: communication is effective e.g., ?The chair broke?; not ?A piece of furniture broke? Learned first by children ? Expressed economically Has a short name e.g., ?chair?, not ?arm chair? ? In sign language also, expressed in simple gestures (see slides) Concept formation: How people categorize? Similarity based views: Prototype view: Abstracted prototype in memory: each new instance is compared to this and then decide if in category Exemplar view: Comparing to individual instances: basis on similarity to amount of instances, prototype to a single person. is the new neighbour like people in your neighbourhood? Unlikely to have a stored prototype for ?people in my neighbourhood? Instead, compare to individual instances Exemplar view and prototype view make similar predictions (due to internal structure of natural categories) ? Faster classification of typical exemplar can be explained by either the prototype or exemplar view typical exemplar is more similar to a prototype, OR typical exemplar is more similar to many instances Theory Based view: Categorization is based on a theory developed from instances: Rips (1989): ?A circular object with a diameter of 6 cm: Is it more likely to be a pizza or a 20-cent coin? ?Ss classified it as a pizza more often Feature value is more similar to a coin (2.7 cm) than a pizza (25 cm) -> against similarity-based categorization Theory (explanation): pizzas vary in size; 20 cent coin doesnt Consistent with effects of context: ?Which is it more similar to? Grey cloud: white cloud vs. Black cloud Grey hair: white hair vs. black hair Based on theory about rain vs. aging Not similarity alone, theories predictive of future in this case, can?t be just typicality, not on surface, more scientific. Reasoning: Generating a conclusion that is not stored in LTM, from premises Are we rational? (no) reasoning ability may be a product of schooling and western culture rather than an innate ability. (see slide, with tribal elder) Deductive Reasoning: (logic) Syllogistic reasoning: Conclusion not the point of true/false...but if inference is logical, as long as conclusion follows premises: Syllogism consists of two premises followed by a conclusion Inference is valid if conclusion follows necessarily from the premises given e.g., If this rock is a garnet, it is a semi-precious stone. This rock is a garnet. Therefore, it is a semi-precious stone. How do people decide this inference is valid? Logical rules: Know and apply (deductive reasoning) Failure of reasoning due to failure of comprehending the premises. Valid inferences from the conditional (If p then q) Modus ponens (given p, q) Modus tollens (given Not q, Not p) Modus ponens rule: If p then q: (conditional): p: (antecedent) Therefore q: (consequent) e.g., If this rock is a garnet, then it is a semi-precious stone (If p then q) The rock is a garnet (p) Therefore, this rock is a semi-precious stone (q) The syllogism is valid because it follows the modus ponens rule Modus tollens rule If p then q Not q Therefore not p e.g., If the rock is a garnet, it is a semi-precious stone (If p then q) The rock is not a semi-precious stone (not q) Therefore, this rock is not a garnet (not p) The syllogism is valid because it follows the modus tollens rule Wason?s selection task (see slide) do as efficiently as possible, logical rules that follow Have to use the rules above, a lot of people don?t do so at all. Not P/Q rule not correct Most people get the E, but do not get the ?7? (less then 10%) SO not logical? Different version Exactly the same form, but about alcohol, really easy to do. Content makes it much easier. The content of propositions affects our reasoning performance Pragmatic rules: (not logical, but rather domain-specific to specific situations) ? Rules we use may be domain-specific (apply to specific situations) Permission rule: driving, drinking, voting If an action is to be taken, then a precondition must be satisfied Pragmatic rules are activated by content of reasoning problem In the Wason selection task, permission rule was not activated by description (vowels and numbers) Evidence for this: Same form as before (Given context or not) Envelope problem If the letter is sealed, then it has a 5-d stamp on it Hong Kong Ss are familiar with the ?sealed envelope? scenario Cholera problem If the form says ?Entering? then the other side lists cholera among the list of diseases Most cases given context were much better (but if knew it before hand didn?t need it) Inductive Reasoning: Probalistic reasoning: John majored in accountancy John now works in an accountancy firm Therefore, John is an accountant Not deductively valid but inductively strong Rules: Base-rate rule The probability of something being a member of a class is greater the more class members there are Conjunction rule The probability of a proposition cannot be less than the probability of the proposition conjoined with another proposition (More of 1 then 2, making it smaller, similarity to prototype) But doesn?t seem to be the rules we apply here, we give into Heuristics or rules of thumbs, which are quick answers but not always correct: Tversky& Kahneman In inductive reasoning, people apply heuristics and ignore probability rules Representativeness heuristic Availability heuristic Representatives heuristic: estimating probability on the basis of similarity to prototype: (see slide) Jane is medium height, wears nice but comfortable clothes, is single and likes to read 18th century English poetry and plays. Is Jane more likely to be a lecturer in English literature or a shop assistant? ?Lecturer in English lit? ignoring the base rate rule (there are more shop assistants than lecturers in English Lit) Is Jane more likely to be a bank teller or a bank teller who is a member of a book club? ?Bank teller and member of book club? ignoring the conjunction rule Arability Heuristic: How easy it is to generate instances: Example Are there more words in English beginning with k or with k in third position? Which is a more common cause of death? Accidents vs stroke Homicide vs diabetes Cancer vs heart disease
Want to see the other 47 page(s) in PSY 246.docx?JOIN TODAY FOR FREE!