Sensory Memory

advertisement
Modal Model
Sensory Memory
Lesson 2
External Response
What is Sensory Memory?
• Memory system that retains purely
sensory information for very brief
amounts of time (milliseconds – seconds).
• Does it have a function (???)
What is Sensory Memory?
• Two different aspects (Coltheart, 1980; Loftus & Irwin, 1998)
– Stimulus persistence: phenomenological,
can be measured either subjectively (e.g., continuity ratings)
or objectively (e.g., temporal integration).
• basis: RESIDUAL NEURAL ACTIVITY
– temporary “buffer” on which attention can operate.
– integration of sensory information over time.
– Information persistence: precategorical/unanalyzed,
can be measured objectively (e.g., partial report).
• basis: “TRUE” MEMORY (e.g., STM).
What is Sensory Memory?
• Modality specific:
– Visual = Iconic
– Auditory = Echoic
– Olfactory ?
– Tactile ?
– Taste ?
1
What is Sensory Memory?
Issues:
• Capacity
• Duration
• Representation
• One store or more?
Iconic Memory
Iconic Memory
• Span of apprehension (e.g., Averbach, 1963)
– 5 or 6 dots accurately
perceived/counted within 600 ms
• George Sperling (1960)
– Partial vs. whole report
(600 ms exposure, OR
150 ms exposure + “persistence”)
F
2
T
9
3
Y
7
P
N
5
X
8
Iconic Memory
• George Sperling (1960)
– Partial vs. whole report
Iconic Memory
Information
persists about
500 ms
Stimulus persistence
• Haber & Standing (1969)
– Flashing circle perceived as continuous, ISI < 300 ms.
• Afterimages
% recalled in cued row
multiplied by total
number of displayed
items
– Persistence should be of complementary color
(but it’s not; e.g., Banks & Barber, 1977).
• Masking (e.g., Turvey, 1973) http://www.scholarpedia.org/article/Visual_masking
Performance based
on iconic memory
Performance based
on STM
– Brightness masking (retinal – limited to same eye)
– Pattern masking (central – not limited to same eye)
• Implication: Some of iconic memory is post-retinal.
SOA = stimulus onset asynchrony
(between visual display and row cue)
Problems / Complications
• Not precategorical
– Merikle (1980): partial report of digits vs. letters
• Output interference: recall of early items may interfere with
memory of later items
– No advantage of partial report over full report with
regard to accuracy of first reported item (Dick, 1971).
Problems / Complications
• Stimulus vs. Information persistence
– Inverse duration effect:
• Stimulus persistence decreases with increased stimulus
exposure time (Bowen et al., 1974; Di Lollow, 1980)
• Information persistence (partial report) is unaffected by
stimulus exposure time (Yeomans & Irwin, 1985)
– Tendency to make acoustic errors in partial report
– Mewhort & Leppmann (1985)
• Letter identity information stable over time
• Location information decreases over time
2
persistence
(150 ms)
exposure
(100 ms)
• Errors
stimulus
onset
stimulus
offset
neural
offset
neural
onset
(variable)
(constant)
Problems / Complications
• Stimulus vs. Information persistence
• Stimulus vs. Information persistence
– Inverse duration effect:
– Dissociations between measures (Loftus & Irwin, 1998)
• Stimulus persistence decreases with increased stimulus
exposure time (Bowen et al., 1974; Di Lollow, 1980)
• Information persistence (partial report) is unaffected by
stimulus exposure time (Yeomans & Irwin, 1985)
exposure
(200 ms)
Problems / Complications
persistence
(50 ms)
stimulus
onset
stimulus
offset
neural
offset
neural
onset
(variable)
(constant)
• Co-temporality ratings (subjective, measures stimulus persistence)
• Temporal integration (objective, measures stimulus persistence)
• Partial report (objective, measures information persistence)
– Elegant design: Stimuli and conditions held constant while only
type of task (type of measure) varies.
– Exps. 1 and 2:
•
•
•
•
5 x 5 matrix, 12 dots – ISI – 12 dots
Co-temporality rating: 2 distinct events or one (on 4-point scale)?
Temporal integration: which cell is empty (% accuracy)?
Very high correlation (Spearman = .98) between measures across
ISI and exposure manipulations.
Problems / Complications
• Stimulus vs. Information persistence
• SUMMARY
– Exps. 3 - 7:
• 2 x 5 array of letters followed by visual probe (bar) above/below
a particular array location.
• Co-temporality rating: 2 distinct events or one (on 4-point
scale)?
• Partial report: what was the letter?
• Lower correlations (Spearman = .63-.81) between measures
across ISI and exposure manipulations.
– ISI affected both measures similarly
– Exposure duration affected the measures differently
(inverse duration effect for rating but not for partial report)
Echoic Memory
• Auditory partial report
Darwin et al. (1972)
– 3 lists of 3 digits
– 3 auditory “channels”
• Results:
– longer persistence
– smaller advantage
– methodological
problems
3
Iconic Memory
Dissociated effects of stimulus exposure duration
(+ similar results using masking)
+ categorical effects/errors in partial report
imply two different types of storage:
• Residual neural activity (retinal, unanalyzed, very brief) that
accounts for stimulus persistence. “RESIDUAL SENSATION”
• Short term memory (central, categorical, longer lasting) that
accounts for information persistence. “TRUE MEMORY”
Echoic Memory
Information
persists 3-4
seconds
• Stimulus Persistence (100 – 300 ms)
e.g., Efron (1970)
– Tones presented from 30 to 100 ms.
– Subjective duration 130 ms regardless of actual duration.
RESIDUAL NEURAL PROCESSING
• Information Persistence (longer lasting).
– Modality effect
– Suffix effect
SHORT TERM MEMORY
Echoic Memory
• Modality Effect –
In immediate serial recall, a recency effect is observed
for auditory presentation but not for visual presentation.
Echoic Memory
• Suffix Effect –
Recency effect for auditory presentation is eliminated
by an irrelevant suffix item that interferes with
short-term auditory storage.
Conrad & Hull (1968)
• read silently vs. aloud
Crowder (1972)
• auditory presentation
• suffix: buzzer or speech
Results:
• speech suffix eliminates
recency effect, but buzzer
suffix does not.
Echoic Memory
• Precategorical Acoustic Store (PAS) –
Theory of echoic memory (Crowder & Morton, 1972)
Assumptions:
Echoic Memory
• Precategorical Acoustic Store (PAS)
Support:
– Stores unanalyzed acoustic information for approx. 2 sec.
– New items interfere with existing items to the extent that they are
acoustically similar.
– Steady-state (e.g., vowel) retained longer than dynamically changing
(e.g., consonant) information.
Explanations:
– Modality effect: Recency effect observed for auditory items because
terminal items* can be retrieved from PAS.
– Suffix effect: Suffix item interferes with recent list items in PAS to the
extent that they are acoustically similar.
i.e., information about terminal items (see Crowder, 1978a). Homophone list:
WRITE-RIGHT-RIGHT-RITE-WRITE-RITE-WRITE <== No recency effect
Echoic Memory
•
Precategorical Acoustic Store (PAS)
Not precategorical – suffix effect is context-dependent.
• musical sound vs. human speech sound (Ayres et al., 1979)
• animal sound vs. human speech sound (Neath et al., 1993)
2.
Not specifically acoustic – modality and suffix effects observed
with non-acoustic stimuli.
• lip reading (Nairne & Walters, 1983; Spoehr & Corin, 1978)
• mouthing words (Campbell & Dodd, 1980)
• touch (Nairne & McNabb, 1985; Watkins & Watkins, 1974)
• ASL
3.
(Shand & Klima, 1981)
Modality and suffix effects when PAS should not be available.
• 60-item old/new recognition (Conway & Gathercole, 1987)
• articulatory suppression (Suprenant et al., 2000)
4
OTHER THEORIES
• Changing-State Hypothesis (Campbell & Dodd, 1980)
Problems:
1.
Echoic Memory
– Modality/Suffix effects will be observed with dynamic
(e.g., auditory) but not static (e.g. visual) stimuli.
– However:
• Modality/suffix effects are observed with (static?) tactile stimuli.
• Numbers presented dynamically (exposed gradually) do not yield
modality/suffix effects (Crowder, 1986).
• Primary Linguistic Code Hypothesis (Shand & Klima, 1981)
– Speech (or ASL) is a primary linguistic code; other types of input
must be recoded before storage in primary memory (STM).
– However:
• Modality/suffix effects are observed with tactile stimuli.
• Tasks that force recoding do not eliminate effects (Manning et al. (1990).
Echoic Memory
• Multi-component View (Cowen et al., 2002)
Serial position effects in standard serial recall may stem
from a combination of 3 sources:
• serial position (as assumed by theories of echoic memory)
• output interference (best recall for earlier items)
• response set size (best recall for later items)
To deconfound these factors, compared various
procedures involving partial or full recall (see next slide).
Echoic Memory
• SUMMARY
As with iconic memory, the data cannot
be explained by a single, modality-specific
sensory memory.
Instead:
1) Modality-specific sensory memory can account
for stimulus persistence (up to ~250 ms)
2) Information persistence can best be explained by
general theories of “immediate” memory (STM)
5
Multi-Component View
Echoic Memory
RESULTS:
1) Input position:
relatively small primacy
effect and somewhat
larger recency effect.
2) Output interference:
Greater for visual than
for auditory input
(yielding modality effect
in terminal positions)
3) Response set size had
no effect.
Ayres et al. (1979)
• Goal: To test the validity of the PAS model as an
explanation of the suffix effect.
• Rationale: If suffix effect stems from interference
with a precategorical acoustic store, then
categorization of the suffix should not matter.
• General Design: Same physical stimulus
(trumpet note – “wa”) used as suffix in two
contexts:
– All other suffixes were musical notes
– All other suffixes were spoken syllables
Ayres et al. (1979)
Exp. 1
Results:
Ayres et al. (1979)
Exp. 2
Control experiment designed to
discount the possibility that suffix
effect induced by musical note
suffix in the context of a majority
of speech suffixes (25% vs. 75%)
was due to its being “unexpected.”
Here, a different musical note
(bowed violin string) was
substituted for the trumpet note
(wa) in the context of other the
other (75%) speech suffixes.
If being unexpected caused the
suffix effect for wa, then a suffix
effect for the bow note should also
be observed here. BUT IT IS NOT.
Ayres et al. (1979)
Main Conclusions
• Interference between suffix and list items
occurs in a memory store that is postcategorical.
• Primary candidate for such storage is
STM.
1
Download