Psyc30017 - Visual Memory

Front 1

Sperling 1960

Back 1

Iconic Memory
-showed people arrays for 50ms, had them report.
-could only ever recall four or five, regardless of the number of items displayed.

-In second experiment, varied duration of a 3x3 array.
-had no effect on correctly reported letters

-In third experiment, used partial reporting
-could estimate memory capacity (num rows * num correctly recalled in one row)
-this estimate, T, was fairly high.
-observers were forgetting much of the display as they wrote it down.

in 4th experiment, repeated exp. #3 using different delays between stimulus offset and time of the tone.
-partial report advantage mostly gone after .25s, gone by 1s.
-Iconic memory lasts less than 0.25-1 second.

Front 2

Evidence for Mechanism of Iconic Memory.

Back 2

-Seems as if the neural activity generated by a visual stimulus takes time to decay.
-During its decay, it can be inspected.
-Very Fragile (eye blinks, movements, and bright backgrounds following stimulus can break it).

Front 3

Who studied VSTM?

Back 3

Luck & Vogel 1997

Front 4

VSTM (Visual Short Term Memory)

Back 4

-lasts for several seconds
-highly capacity limited
-robust (not destroyed by eye movements)
-requires 20-50ms per item to form.

Front 5

Luck & Vogel, 1997

Back 5

-A sampe array of shapes presented
-50% of the time, one item was altered
-observers indicated wether or not a change occurred.
-Performance fell after about 4 items in array.
-adding verbal load did not decrease VSTM capacity. (verbal memory separate from VSTM)

In Experiment 2, compared 100ms stimulus duration to 500ms stimulus duration. no difference.

Front 6

How is Memory Capacity Measured? Complications?

Back 6

-Estimating capacity is complicated by subject guessing
-formula used that allows authors to take guessing into account so as to generate an unbiased estimate of memory capacity.

Front 7

Estimating Capacity

Back 7

-Know the proportion of trials where subject reports a change(H), the total number of n objects.
-deduce g by looking at false alarms.
-then can accurately deduce m, memory capacity.

Front 8

Possible Confounds (Luck & Vogel, 1997

Back 8

-As set size increases, the number of comparisons also increase.
-Suppose it wasn't a memory limitation, but errors in comparisons. Performance would still decrease with set size.
-Solution: cue just one object and ask if this object had changed.

Front 9

(Luck & Vogel, 1997) Experiment 2

Back 9

-O’s asked only if cued item had changed
-total number of comparisons did not vary with set size.
-No difference in performance found. indicating performance reflects memory errors not comparison errors.

Front 10

Luck & Vogel 1997 Summary Experiment 1 & 2

Back 10

- Experiments 1 & 2 showed that memory capacity was approximately 4 items
• However, the objects varied only along one feature dimension (color)
• Is memory capacity object-based or feature-based?

Front 11

Luck & Vogel 1997 Experiment 3

Back 11

-3 conditions: color changes, orientation changes, color or orientation changes.
-In condition 3, observers had to memorize double the number of features. PERFORMANCE DID NOT DECLINE.
**VSTM difficulty is determined by number of objects to be memorized, not
number of features

Front 12

Luck & Vogel 1997 Experiment 4

Back 12

Same as #3, but more conditions.
-conjunction condition: any feature could change.
-no change in performance.

Front 13

Potential Concern with Luck & Vogel's results

Back 13

-Is still possible that memory capacity is feature-based.
-maybe observers have independent memory capacity for each type of feature (eg. 4 colours, 4 orientations, etc.)

Front 14

Luck & Vogel 1997 Experiment 5

Back 14

Same as experiment 1, but each item has two colour features.
-If memory were feature based, then performance should have been worse in the conjunction condition
-THIS WAS NOT THE CASE.

Front 15

Luck & Vogel overall Summary

Back 15

- what this study really showed is that you are more efficient at remembering features when they are bound together on the same object. (eg. 8 features on 4 objects is better than 8 features on 8 objects).
- they are implying that you can remember an infinite number of features.
- In the next study we show that this statement is too extreme.

Front 16

Alvarez & Cavanagh (2004)

Back 16

-Luck & Vogel used very simple shapes.
-Would results hold in more complex shapes?

Front 17

What Is Object Complexity?

Back 17

• Is the bar simpler than the cube?
• One way to define complexity is by the length of the description required to define the object
• The shorter the description, the simpler the object
• Because it is quicker to describe the red bar, this must be the simpler object

PROBLEM: Too imprecise.
-Alvarez & Cavanagh (2004) used visual search efficiency to measure object complexity.

Front 18

Rational behind Alvarez & Cavanagh (2004)

Back 18

• In a serial (i.e. conjunction) visual search we search for a target by attending to each item in turn and comparing the attended item to our memory of the target to determine whether it is the target.
• Thus, the rate of visual search is a measure of object complexity.
• Search rate is defined as the slope of the visual search function.

Front 19

Alvarez & Cavanagh (2004) Experimental Design

Back 19

-First target presented
-Then blank interval (900ms)
-Then the search array
-Subject reported wether or not the target was present as quickly as possible.
-Response time vs set size of search array is the visual search function
-Search rates varied from 10ms to 130 ms/item

Front 20

Now that they have a way to determine complexity, can measure memory capacity....

Back 20

-Array of object shown for 500ms, blank interval of 900ms, then a test array.
-O's asked: was there a change?
-Only a change on 50% of trials.
-Set sizes between 1 -15
-Evidence that memory capacity varies with object complexity
-"Complex" stimuli of Luck & vogel were not complex enough.

• Take home message: you cannot remember an infinite number of features.
• However, Luck & Vogel’s other claim still stands: there is an object advantage, it is easier to remember features when they belong to the same object than when they belong to different objects.

Front 21

Todd & Marois (2004)

Back 21

-performed an fMRI experiment to determine which brain areas are involved in VSTM.
-activity in the posterior parietal cortex is tightly correlated with the number of objects stored in VSTM.
-more object stored, the higher the activity.

-O's shown an auditory stimulus to rehearse / object array / retention interval / probe disk(Y/N) / auditory probe(Y/N).

Front 22

Todd & Marois (2004) Iconic memory experiment

Back 22

(unfortunate name, not relevant. nothing to do with memory, or iconic memory)
-same experiment, without recall
-just said yes or no to wether or not there was an item in the very centre of the sample array.
-Allows us to determine what fraction of the BOLD fMRI response in the VSTM experiment was due to the stimulus and not the memory requirement.

Front 23

Bilateral Intraparietal / Intraoccipital Sulcus

Back 23

-activity correlates with number of items stored in VSTM
-activity does not correlate with the number of items in the display. (only 4 items can be stored in VSTM)

Front 24

Ventral-Occipital Cortex

Back 24

-involved in response to stimulus, not memory. (when there is no visual stimulus, there is no activity)
-correlated with complexity of stimulus and decision process.

Front 25

Anterior Cingulate Cortex

Back 25

-signal not driven by memory because same activity in "iconic memory experiment" as in VSTM
-active when observer makes a response.
-correlated with just the decision process.

Front 26

Brady et al. (2008) Design

Back 26

Visual Long Term Memory:
-presented with 2500 real-world objects for 3s each.
-informed to remember all the details for a test
-During trial, O's repeats were occasionally shown and O's told to response by pressing a key.
• When finished, 30 minute test (300 images)
• In testing session, O’s shown pairs of images and had to indicate which they had seen before
• Test pairs could be “Novel”, “Exemplar” (categorical) or “State” (minor differences)

Front 27

Brady et al. (2008) Results

Back 27

-Memory Performance very good , even in state conditions! (93%/88%/87%)
-Tells us LTM is huge and highly detailed - much more than previously thought.
-Why so much different than Luck & Vogel? Maybe difficult to notice changes in orientation?
-VLTM not sensitive to rotations. This is why Luck & Vogel participants couldn't use VLTM

Front 28

Konkle et al. (2010) Overview

Back 28

Why is VLTM highly sensitive to state changes but not to changes in orientation?
-Konkle et al. explains this by making a distinction between conceptual distinctiveness and perceptual diststinctiveness

Front 29

Konkle et al. (2010) Design

Back 29

-2800 images, repeat detection task (about 5 hours)
-Observers presented with up to 16 different examples of the same category of object (ie. 16 different apples). UNLIKE BRADY.

Front 30

Konkle et al. (2010) Results

Back 30

-More exemplars = decreased percent correct.
-This slope is the 'interference slope'.
-More interference when exemplars were conceptually similar. perceptual similarity didn't matter much.
CONCLUSION: Humans can remember a vast number of conceptually distinct images. Seems to be specific to visual memory - does not hold for auditory.

Front 31

Cohen et al. (2009) Experiment 1

Back 31

-Experiment 1:
-Participants listened to 64 5-second sound clips.
-after those, another 64 sound clips (half new half old)
-had to indicate which ones were new.
-overall, poor performance (78% correct, chance is 50)
-Whereas in the brady study, 2500 images were presented and performance was 93% correct.

Front 32

Cohen et al. (2009) Experiment 2 Design

Back 32

-same as experiment 1, but better sound quality and 5 conditions:
-sound = only heard sound
-sound+picture = in training phase heard sound and saw picture. In test phase just heard sound.
-sound+name = in training phase heard sound and saw name. In test phase just heard sound.
-name=just saw name.
-picture = just saw picture (similar to Brady)

Front 33

Cohen et al. (2009) Experiment 2 Results

Back 33

-Sound had worst result; picture had best result.

Front 34

Cohen et al. (2009) Experiment 3

Back 34

Two Conditions:
-Language: 90 unique speech clips (7-15s)
-Music: 90 popular music clips (5-15s)

-Language much higher than music.

-REVIEW THIS>