Lecture 1: Short Term and Working Memory Outline • What is memory? • The Short Term / Long Term distinction • Baddeley’s model of Working Memory • Evidence for Baddeley’s model • The episodic buffer By the end of the lecture you should be able to appreciate: • The importance of memory in our everyday lives • Why a distinction is made between short and long term memory • Evidence for the 3 original components of Baddeley’s model of Working Memory • Why the original model was revised Clive Wearing Without our memories, what is left? QuickTime™ and a Sorenson Video decompressor are needed to see this picture. “Its exactly like death.” History of the STS/LTS distinction 1890 - William James draws distinction between primary and secondary memory. PRIMARY MEMORY SECONDARY MEMORY Reward portion of present Genuine past space of time Linked to conscious experience Unconscious - permanent Retrieval is effortless Retrieval is effortful The development of computers in the 1960s provided another analogy which split memory into two: CPU/RAM = STS, STORAGE = LTS Modal model of memory developed by Atkinson & Shiffrin (1968) The modal model of memory Sensory Rehearsal Store Short Long Sensory Term Term Store Store Store Transfer Sensory Store Displacement (Forgetting) But what is the evidence for separate STS / LTS? Evidence for STS / LTS distinction Converging evidence appeared to support the STS / LTS distinction as proposed by the modal model: • Capacity differences - STS = limited / LTS = unlimited • Encoding differences - STS = phonological / LTS = semantic • Serial Position Curves - STS = Recency / LTS = Primacy + Asymp / • Forgetting - STS = trace decay / LTS = interference • NP Evidence - HM = intact STS, impaired LTS KF = intact LTS, impaired STS BUT - psychology is never simple... Evidence for STS / LTS distinction • Encoding differences - How do we comprehend text / learn language / remember faces? • SPCs - Recency effects after 20sec distraction following each item (Tzeng, 1973). Long term recency (Baddeley & Hitch, 1977) constant ratio rule (t / T) (Glenberg et al, 1980). • Forgetting - Interference effects in STS (e.g. Release from Proactive Interference - RPI) • NP Evidence - Why is KF able to encode information in LTS if the STS is a critical bottleneck? The modal model provided the first systematic attempt to account for the structures and processes which comprise the memory system But by the end of the 1960s there were several well established findings that it was unable to account for. Working Memory “Courtney Young… was the undisputed Security Service crossword king. He always claimed it was too easy to do the Times crossword with a pencil. He claimed to do it in his head instead. For a year I watched him do this, until finally I could resist the temptation no longer. I challenged him, whereupon he immediately wrote in each answer, without hesitation.” Peter Wright, in Spy Catcher. (Quote used at beginning of chapter on Working Memory in Al Parkin’s text book) Baddeley and Hitch’s model of Working Memory Articulatory loop Central Executive Background By late 1960s - STS research “laboratory bound” Modal model unable to account for important data: 1. Relationship between type of encoding and type of store 2. Why STM patients had normal LTM 3. Dual task data Baddeley and Hitch - “What is the STS for?” Assumed to be used for learning, reasoning and comprehension - but little or no evidence. Baddeley and Hitch developed dual task paradigms :- Ss perform a primary task whilst simultaneously performing a secondary task which is presumed to take up STS capacity. Overt rehearsal of secondary task ensures that subjects are not simply switching between tasks. Dual Task Paradigms Ss remember (and overtly rehearse) sequences of 0-8 digits At the same time subjects perform a simple reasoning task A precedes B: AB (TRUE) B is not preceded by A: AB (FALSE) Increase in reasoning time is significant, but not large (35%). More importantly: No effect on errors Dual Task Paradigms Primary task: learn a list of words Secondary tasks: 1) copying pairs of digits 2) rehearsing 3 digit sequence 3) rehearsing 6 digit sequence Condition 3 has a small effect on primary task performance, but it does not effect the recency component of the SPC. Evidence for the Phonological Loop Dual task results imply system responsible for digit span cannot be the same as system responsible for learning / reasoning. Baddeley and Hitch (1977) - performance on verbal span tasks involved a speech-based system. The phonological loop (AKA articulatory loop) comprises two components: Phonological Store: holds small amount of speech based information Articulatory Control Process: Based on inner speech Phonological Auditory Visual Store Presentation Presentation Phonological Similarity Effect Phonological similarity effect: Recall of characters or words is impaired if they are phonologically similar. (Conrad, 1964; Baddeley, 1966) PVCGE is harder to recall than XRFYZ This effect can be explained because items in phonological store are based on phonological codes. PGCVE all have similar phonological codes. Similar codes are harder to discriminate. Hence recall is worse Word Length Effect Phonological similarity effect: Span for short words is greater than span for long words. Is this an effect of syllables or spoken duration? E.g. is syllable a “Unit of storage”? Word Length Effect Spoken duration appears to be crucial: Memory spans are greater for words like “Bishop” and “Wicket” than for “Harpoon” and “Labile” (Baddeley et al, 1975). Language Articulation Rate Digit Span Chinese 265ms/digit 9.9 English 321ms/digit 6.6 Welsh 385ms/digit 5.8 (Hoosain & Salili, 1988; Ellis & Hennelly, 1980) Memory span and articulation rate are highly correlated in all age groups - our span increases as we are able to articulate more rapidly. Overt or covert articulation serves to maintain items in the phonological store by refreshing their fading traces. The faster it can run, the longer the memory span. Unattended Speech Effect Performance on span tasks is impaired if items are accompanied by other verbal material: Colle & Welsh (1976) - immediate recall of digits is impaired if accompanied by sound of someone reading German. Explanation - unattended phonological material can gain access to the phonological store. Salame & Baddeley, (1987) - Spoken digits - “one”, “two” impair digit span to the same degree as similar phonemes like “tun, woo” Both impair span more than non-similar words “happy, tipple”. Explanation - code is phonemic, not semantic. This suggests that listening to music with vocals may impair your comprehension of complex texts. Patient Data Learning to read Modal model cannot explain data from STM patients such as KF - LTM should also be impaired (STS is bottleneck) Data can be explained if it is assumed that these patients have impaired phonological stores. To what extent does rehearsal depend on “inner speech”? Dysarthria: Inability to use speech musculature due to brain damage. Dysarthric patients have normal digit spans, and show normal phonological similarity effects: Articulatory control processes do NOT depend on peripheral speech musculature - we can run a motor program centrally. SO: Phonological loop preferred to articulatory loop What is the phonological loop for? 1. Learning to read: Children with impaired reading ability have reduced memory spans and have difficulties in tasks which require the manipulation of phonological information (e.g. given Stop, reply Top). 2. Language comprehension: STM patients such as TB have some difficulty in comprehending verbose or complex sentences e.g. “The boys pick the apples” = OK; “The two boys pick the green apples from the tree” = Impaired 3. Vocab acquisition There is a strong correlation between non-word repetition (which strongly taxes the phonological loop) and vocabulary size (Gathercole & Baddeley, 1989) Imagery Imagery ignored by the behaviourists, - studied again in the 1960s. “Imagery debate” - Kosslyn and Pylyshyn Are processes involved in imagery same as those involved in visual perception? Imagery = Analogue or Propositional ? Debate proved rather barren - but generated interesting phenomena. Neuropsychological and neuroimaging data suggests neural substrates involved in vision also involved in imagery. The visuospatial sketch-pad Baddeley - listening to American football disrupts driving VSSP - A workspace in which an image can be stored and manipulated to guide behaviour. 3 4 Brooks Matrix Task (1967) Subjects learn sequence of 1 2 5 sentences: Spatial: “In the next square to the 6 right put a 2” Non-spatial: “In the next square to the quick put a 2”1 Subjects remembered 8 spatial vs 6 non-spatial. Spatial instructions better when presented auditorily Non-spatial instructions better when presented visually The visuospatial sketch-pad Baddeley et al, 1975 - Ss perform Brooks matrix task with and without concurrent distractor - pursuit rotor. Tracking disrupts the spatial task, but not its verbal equivalent: Sketchpad relies on spatial coding What is the sketchpad for? Not as well studied as the articulatory loop Geographical orientation: - learning our way around our environment. Planning and performing spatial tasks Hatano & Osawa (1983) -Japanese abacus experts memory for numbers is disrupted by concurrent spatial but not verbal task. What is the sketchpad for? (Eye Movement Desensitisation Reprocessing) From the website: What is EMDR “an innovative method of psychotherapy…The focus of EMDR treatment is the resolution of emotional distress arising from difficult childhood experiences, or the recovery from the effects of critical incidents, such as automobile accidents, assault, natural disasters, and combat trauma [PTSD]. Other problems treated with EMDR are phobias, panic attacks, distress in children, and substance abuse. Another innovative focus of EMDR is performance enhancement: which aims to improve the functioning of people at work, in sports, and in performing arts” EMDR EMDR is a “Therapy” used by Psychotherapists, Psychiatrists, Clinical Psychologists etc primarily for the treatment of PTSD. Involves following the finger of therapist whilst imagining “negative information identified with problem” From the website: HOW DOES EMDR WORK? “It is not clear how EMDR works because neuroscience researchers are still exploring how the brain works... However, there is evidence for an innate information processing system that exists as part of human thinking processes….” “EMDR appears to produce a direct effect on the way the brain processes upsetting material. Researchers have suggested that the eye movements trigger a neurophysiological mechanism that activates an "accelerated information processing system." Accelerated information processing is a phrase used in EMDR to describe the rapid working through, „metabolizing‟, of upsetting experiences.” Stickgold R (2002) EMDR: A putative neurobiological mechanism of action JOURNAL OF CLINICAL PSYCHOLOGY 58 (1): 61-75 …EMDR induces a neurobiological state, similar to that of REM sleep, which is optimally configured to support the cortical integration of traumatic memories into general semantic networks. We suggest that this integration can then lead to a reduction in the strength of hippocampally mediated episodic memories of the traumatic event as well as the memories' associated, amygdala-dependent, negative affect. EMDR Experimental Psychology to the rescue… McNally RJ (1999) EMDR and mesmerism: A comparative historical analysis J ANXIETY DISORD 13 (1-2): 225-236 JAN-APR 1999 Most disabling symptom of PTSD is recurring, intrusive images of precipitating trauma. These images are presumably instantiated in VSSP Effective volitional eye tracking involves the storage and manipulation of both visual and spatial information IN OTHER WORDS IT COMPETES FOR VSSP RESOURCES Therefore EMDR is simply another “desensitisation” procedure - of the type traditionally used by behavioural therapists. BUT… Christman SD, Propper RE, Dion A. (2004) Increased interhemispheric interaction is associated with decreased false memories in a verbal converging semantic associates paradigm. Brain Cogn. 56:313-9. The Central Executive Most complex and least understood component of WM model “In some ways the central executive functions more like an attentional system than a memory store” - Baddeley Model suggests CE coordinates the activity of the two slave systems Other potential roles for the CE include coordinating retrieval strategies, and selective attention Baddeley suggests that a model of action control developed by Norman and Shallice (1980,1986) may serve as a model of the central executive Norman and Shallices (1986) model of attentional control Supervisory Attentional System Trigger Perceptual Effector Data System System Base Schema Control Inhibition Units Contention Scheduling Evidence for the CE Dual task performance and DAT: Pursuit rotor and digit span tasks were adjusted so that individual performance was identical in DAT and matched controls Combining the two tasks caused greater costs in the DAT patients than the controls - CE impaired in DAT Random number generation: Ss required to generate random sequences of letters make more repetitions and stereotyped responses the faster the task Dominant schema (ABC, ITV etc) must be constantly inhibited by the SAS and novel schema activated. Evidence for the CE Neuropsychological evidence: - dysexcutive syndrome Original Norman and Shallice model developed to account for behaviour of patients with frontal lobe lesions. Perseveration - patients have lost ability to interrupt ongoing schemas Catatonia - patients can remain motionless and speechless for hours - unable to initate schemas. Distractibility - schemas easily “captured” by external/internal stimuli Utilisation behaviour - Lhermite (1983) Utilisation Behaviour Utilisation behaviour: The tendency to grasp common objects when presented, and perform the function commonly associated with the object E.g Lhermite (1983) Problems for original WM model 1. Articulatory suppression: According to the model, AS should prevent registration of visual material (which must be recoded phonologically) In fact, span only drops slightly (Baddeley et al, 1994) 2. Neuropsychological data: STM patients, with digit spans of 2 or less, have visual spans of about 4 (Baddeley et al, 1997). 3. Chunking: If stimuli comprise a meaningful sentence, span is considerably increased (e.g. info in LTM is used to chunk) Problems for original WM model 4. Rehearsal Not all rehearsal can be subvocal: How are items in VSS rehearsed? What about children? 5. The role of consciousness CE originally proposed to assist in binding - our ability to integrate information about location, colour, size, smell, feel etc of objects. How could it do this without a multimodal short term store? Baddeley (2000) suggests the above problems can be solved by an Episodic Buffer. A Revised WM Model Central Executive Visual Episodic Language Semantics LTM The Episodic Buffer “A limited capacity temporary storage system that is capable of integrating information from a variety of sources” • Controlled by the CE • Feeds information into and retrieves information from LTS • Uses a common “multidimensional” code The Episodic Buffer makes the link between Working Memory and LTM more explicit BUT: Are VSS and AL still necessary? Summary • Original model was able to account for a considerable body of data that the modal concept of an STS could not explain. • The concept of WM has proven to be enormously influential, and is used by Neuroscientists, Neuropsychologists, and AI researchers. • Problems with the original model led to the addition of the Episodic Buffer • The central executive component remains under-specified, and controversial.