Working memory

• The phonological similarity effect is a phenomenon where lists of items that sound similar when spoken, such as "B, T, P, V, G, C", are much harder to recall than lists of nonconfusable items, such as "R X H Z L Q".
• The phonological similarity effect is not that subjects forget the items; rather they tend to get the order wrong, as demonstrated by Wickelgren in 1965.
• The phonological similarity effect is found in congenitally deaf children, especially those who are good speakers, suggesting the short-term store is more articulatory than acoustic.
• The phonological similarity effect is also found in patients who can't speak, suggesting a role for central speech planning rather than articulation itself.
• Page, Cumming, Madge & Norris (2007) showed that the "memory" errors may indeed be speech errors, i.e., to do with speech planning.
• There is little difference in short-term serial recall between lists of semantically similar words and their controls, such as "LONG, HUGE, WIDE, TALL, BIG" and "OLD, LATE, THIN, WET, FOUL".
• Semantically similar words are actually easier to recall after a long delay than unrelated words, suggesting that long-term memory is more meaning-based, and short-term memory is more sound-based.
• Concurrent articulation involves asking subjects to say repeatedly a word or phrase so as to occupy their speech apparatus while they perform, for example, serial recall of visually presented materials.
• When concurrent articulation is used during serial recall of visually presented verbal materials, there is no longer a phonological similarity effect, supporting the assumption that the loop is speech-based.
• Concurrent articulation does not get rid of the PSE for auditorily presented materials, even when concurrent articulation continues throughout both input and recall, suggesting that auditorily presented material access the speech loop directly without needing to be converted by a speech-based mechanism.
• The structure of the phonological loop includes a speech loop, a phonological store, and a phonological threshold.
• The word-length effect, as demonstrated by Baddeley, Thomson & Buchanan (1975), is that lists of long words are harder to remember than lists of shorter words.
• Subsequent experiments with slow-to-pronounce words and fast-to-pronounce words suggested a large component of the word-length effect is due to pronunciation time rather than word-length per se, which is also suggestive of a speech-based system.
• People can remember about as many things as they can say quickly in two seconds.
• The word-length effect results have been challenged by Caplan, Rochon & Waters (1994), Service (1998), and notably by Lovatt, Avons & Masterson (2000, 2002).
• The word-length effect also explains the otherwise puzzling result that, for example, Welsh speakers have a lower digit span than English speakers, as the Welsh digits just take longer to say.
• The word-length effect is primarily attributed to the fact that long words take longer to rehearse and recall due to the time it takes for the store to decay.
• If you stop rehearsal altogether using Concurrent Articulation (CA) and control the output time, then even though auditory items access the store, they will not show a word-length effect.
• Serial recall from phonological short-term memory is disrupted by irrelevant sound played either during presentation or in between presentation and recall.
• The effect of irrelevant sound is most significant when the sound comprises at least one identifiable source that changes state.
• In a particularly elegant experiment, Jones & Macken (1995) showed that the same sequence of syllables would either give an effect that depended on how they were grouped spatially.
• If the syllables came from one location, the subjects heard a changing stream “ABCABC…” and there was a large effect on recall.
• If the syllables came from three different locations, subjects heard three nonchanging streams “AAA”, “BBB” and “CCC”, and the effect on recall was reduced.
• Speech-like irrelevant stimuli, even if they're in a language completely foreign to the subject, are most effective - as a result, the effect is often referred to as the irrelevant speech effect.
• The irrelevant sound effect is not just a general attentional effect, it seems to be specific to disruption of the phonological store, as is evidenced by the fact that CA removes the ISE for visually presented material (Salame & Baddeley, 1982; Hanley, 1997) but not for auditorily presented material, which has direct access.
• Auditory presentation: concurrent articulation = material still gets into phonological store so phonological similarity effect is observed (this suggests that auditory items have direct access to the loop), irrelevant sound effect is also observed, but word-length effect is not observed since there is no rehearsal
• Concurrent articulation = blocks articulatory control process (can't say stuff to yourself)
• Visual presentation: concurrent articulation = material does not make it to the phonological store so phonological similarity effect is not observed, irrelevant sound effect and word-length effect is also not observed since there is no rehearsal (no PSE, ISE & WLE) material is probably recalled from a visual store that is not speech based
• Neuropsychological evidence suggests that the phonological loop is primarily a device which assists in the learning of vocabulary, particularly word-forms.
• The phonological loop seems to be intimately connected to the learning of phonological word forms (i.e., the sequence of sounds of which vocabulary items are comprised).
• The Hebb effect in immediate serial recall seems to be based on something like word learning (lexicalization)
• In a study of short-term memory patients, P.V, it was found that she could remember only 40% of lists with 3 nonconfusable items and none of the confusable lists of the same length when the input was auditory.
• With visual presentation, P.V's span was about 4 which didn't depend on similarity at all.
• P.V showed no word-length effect, though she had normal skill in fast articulation.
• P.V tried to use her articulatory loop even though it only has a capacity of about 2 items.
• With visual input, there is no incentive for P.V to convert to a speech-based code, so she relies on a visual store with better capacity.
• P.V's comprehension of speech was remarkably good, only breaking down for long verbose sentences which had had two words exchange positions.
• P.V's spatial short-term memory, verbal and spatial long-term memory, and general intelligence were all normal.
• P.V had almost complete inability to learn new vocabulary after the onset of her stroke, even though she could continue to recognize words that she had learned prior to her stroke.
• The last observation, that patients with an impaired phonological store have trouble learning new vocabulary, gives a hint as to the answer to the question raised in the title of the next section.