Short Term Memory

Modal model of memory-Atkinson and Shiffrin, 1968

• Defined structural components of memory and how they interact with each other

• cognitive control processes

  • Brief time which info extends beyond sensory memory-buffer

  • Info can then be kept indefinitely through maintenance rehearsal

    • Through elaborate rehearsal you can transfer the knowledge into long term memory at which point you can retrieve it back into short term memory at different times

Capacity

measured by span

depends on remembering item and order

Chunking-definition

• Typical memory measured as 7 plus or minus 2 chunks

• Depends on knowledge, understanding and ability to recognize patterns

  • Allows us to overcome 7 item limit

Chunking-Ericsson, Chase and Faloon, 1980

nine items to about 80 with daily practice

Simon and Chase, 1973-Chunking

Experts recalling more chess pieces placements than novices but only when notable (a play of sorts)

Coding-Conrad, 1964

• Listening task-letters embedded in white noise

• Repeat letter name as heard them

• Identify kinds of errors made when difficulty hearing speech sounds in noise

• if short-term memory depends on an acoustic code, when people have trouble remembering, mistakes should likewise be based on similar sounds

recall confusion matrix

Errors seemed to resemble listening errors, even though letters presented visually and reported from memory (B as D, P as T)

Decay and Interference (Brown/Peterson & Peterson)

Performance declined sharply as recall interval increased

In the absence of rehearsal, memory trace decays in under 30s

Decay or Interference (Keppel & Underwood, 1962)

Trial 1 showed little change in performance over 18 seconds, performance declined on subsequent trials

• In the absence of rehearsal, memory trace is lost due to interference

Time versus Interference (Waugh & Norman, 1965)

Performance dropped most sharply as number of interfering items increased

Presenting items slowly (more time for decay) had less pronounced effect

• Interference is more important than decay

Corroboration (Wickens, Dalezman and Eggemeier, 1976)

• Proactive interference can build up over time

  • Previous experience alters ability to learn now

  • Interference is released with a change in meaning

• The greater the change in category, the better performance on trial 4

• STM recall determined not by how much time has passed but by how much interference has built-up or been released

Retrieval search strategies

Ascertain nature of search by looking at reaction time to determine whether reaction item is present in the set

• parallel search

• serial self-terminating search

• serial exhaustive search

Parallel search

• Time to search items currently being held in short term memory-would be unaffected by memory set size

• All items searched simultaneously 

• Should be as fast to respond to target present trial as target absent trial-instantaneously know if target is or is not present

Serial self-terminating search

• Each item must be searched one at a time but once it is found the search is halted

• Longer to search larger set sizes

• Implies a search should on average take half as long for a target present trial than a target absent trial

Serial Exhaustive

• Search to end of list every time even if already found target

  • Done so sequentially

• Same amount of time on average regardless of whether or not target is present

Test STM search strategy-Sternberg, 1966

• Measured reaction time to indication of target presence of absence 

• On average reaction time increased with set size-serial search

• Slope was same regarding of probe absence/availability-exhaustive

• Conclusion serial exhaustive search

Free Recall, Serial order effects

Primacy, Recency

Primacy effect

Better memory for first list items

Less likely with difficult-to-name items

Recency effect

Better memory for last list items

Primacy due to LTM encoding, Rundus, 1971

Probability of recall varies with number of rehearsals in the primacy portion of the curve but not in the recency portion

Primacy due to LTM encoding

Primacy effect corroboration, Glanzer and Cunitz, 1966

If increase time available for rehearsal, primacy should increase

Recency should not

• confirmed

Recency due to read out from STM (Glanzer and Cunitz, 1966)

The longer the distractor duration, the smaller the recency effect

Recency due to read-out from STM

Recency-Suffix effect

Recency effects smaller following a speech suffix than following a nonspeech suffix

Why

• Recency due to STM

• STM uses speech based code

• Other speech interferes with this encoding

Can recency of rehearsal explain both primacy and recency effects? (Tan & Ward, 2000)

Nominal serial position

• Performance best for items at the beginning (primacy) and end (recency) of the list

Functional serial position

• Performance best for items rehearsed most recently

No need to postulate two different memory systems

Long-Term recency effect-Baddeley and Hitch, 1977)

Recency occurs over months/years and cannot be due to read-out from STM

Recency effects reflect a retrieval strategy

• Temporal distance between an item and its nearest competitor determines which items are most available for recall

• Recent items more discriminable

• Not specific to STM

Serial Order effects-Free recall

• Allowed to recall items in any order they want but they look at such to analyze the serial position of that item at study

• Free recall has a u shaped serial order function

Models of Serial Order

Chaining models, Ordinal models, Positional models

Chaining models

• Order stored as item-to-item associations (a chain)

• Items recalled in sequence

• Recall cannot continue if one is not recalled

  • Contradicted by the recency effect

Ordinal models-types

Perturbation model, inhibition model

Ordinal model-definition

Order conveyed by where one item occurs along a dimension (e.g., control unit hierarchy, activation) relative to others

Perturbation model

• Information is chunked in a hierarchy of control units

• Order is conveyed by item-to-control unit associations

• Lower level mis-orderings more likely than higher level mis-orderings

Inhibition model

• each studied item receives activation

  • the first receives the most, the next less

• Items are recalled in order of strength

  • Once recalled, the item is inhibited

    • This allows the next strongest to be selected

• Can’t account for recency effects

Positional models-definition

• Order determined by associating each item with position

  • Position represented as slots or relative to changing context

Positional models-types

Slot-based, Context models

Slot-based models

• Memory has slots at each position

• Information is stored slots as it is encoded

• Recall involves moving through the slots

• Item and order information are stored together (if one is forgotten, both are)

  • Little evidence to support this

Context models

• Context shifts over time

• Events closer in time have similar contexts

• Context can guide serial order

• Mis-orderings occur because contexts are similar