On the Failure to Detect Changes in Scenes Across

Brief Interruptions

Professor: Liu

Student: Ruby

Motive & purpose

• Motive– The authors think that the past papers didn’t

have a same detailed representation.

• Purpose– To test how the lack of consistency comes

about.

ReferencesYear Authors Result

197519961979

Bridgeman, Hendry, & Stark

GrimesMcConkie & Zola

The observers find it hard to discover changes in pix made during a saccade.

19921997

1997,1996

Hayhoe, & WhiteheadHenderson

Irwin

It can be carried across a saccade are a few characteristics of a few previously fixated items.

1997 Simons & Levin

In the test, observers are poor at detecting change when the displays are separated by an interstimulus interval (ISI) of more than 70-100msec and the presentation time range from 250msec to 2sec.

ReferencesYear Authors Result

1997 Rensink

Coherence theory which is the early level structures simple doesn’t have much coherence in the absence of focused attention.

1997 Rensink

Some people think the representation is too fast, they have to descript what they discover when the incoming stimuli at the same retinal location.

1992Kahneman, Treisman,

& Gibbs

The focused attention is required to provide the coherence which the early level structures into large scale objects and allows these objects to keep an identity over time.

General method• Pix were displayed for 240msed and blanks for 80

msec.• The blank fields being medium gray.• 48 color pix.• 27wide and 18 high.• Change the presence , color, or location.• Central interests (CIs) were objects or areas menti

oned by most observers.• Marginal interests (MIs) were objects or areas me

ntioned by none.

General method• Ten observers in each test.

• Observers were required to press the keyboard when they discover the change then say what the different in these two pix.

• Using computer presented the pix and controlled and collected the data.

• 6 trials for practice.

The procedure of the change blindness

Experiment 1 - Extended preview

• Giving observers an 8 sec preview for the original image before the test.

• They were asked to remember as much of the image as possible in order to improve the ability to notice change.

Result for Experiment 1 • There is no significant in

detection time (all Ps > .2).• No effect of extended

preview, F(1,10) = 0.6, P > .4.

• The change blindness is not caused by non-enough time to construct a serial pix, or by some flicker-induced interference with its consolidation in memory.

Experiment 2 – Different blank durations

• Change the blank field duration of 40msec, 160msec, and 320msec.

• Results here were combined with those obtained for the standard durations of 80msec.

Result of Experiment 2

Result of Experiment 2

• No significant interactions between duration and interest,F(3,30) = 2.8, p > .05.

• No significant interactions between duration and change type, F(6,60) = 0.87, p > .5.

• No significant interactions between all three together, F(6,60) = 0.98, p > .4.

• However, a significant effect of duration was detected, F(3,30) = 3.66, p < .03.

Result of Experiment 2

• Blanks of 320msec are slower than for 80msec blanks, this slowdown was significant only for marginal interests; CI: p > .1;MI: p < .05.

• This slowdown perhaps because some kind of memory are reduce.

Experiment 3 – Different blank colors

• Three different colors of the blank field:

black, white, and red.

• The red field was isoluminant with the standard gray.

Result of Experiment 3 • No interactions between

luminance and interest, F(2,20) = 0.55, p > .5.

• No interactions between luminance and change type, F(4,36) = 0.44, p > .7.

• No interactions between all three factors, F(4,36) = 0.047, p > .9.

• No overall effect of luminance was found, F(2,20) = 1.03, p > .3.

Result of Experiment 3

• Performance was not significantly affected for changes in central interests (all ps > .2).

• Color changes in marginal interests were significantly more difficult to detect (p < .05),

Experiment 4 – Reduced coverage of the interruption

• Using a brief, localized non-color “splats” that did not cover the areas of the image that were changed.

• Splats were rectangles having a structure of black and white checks, each check being approximately 20 arc minutes in size.

Result of Experiment 4

• There was no effect of change type, F(2,18) = 2.70, p > .05

• There was an interaction between change type and interest, F(2,18) = 9.83, p < .002.

• No significant difference between CIs and MIs (p > .6).

• There was a strong effect of interest, F(1,9) = 154.06, p < 10-6, with detection slower for MIs

Discussion

• 8 sec should to be enough to discover the change more easily.

• If change blindness were caused by disruption, performance might worse (or improve) with increasing duration.

Discussion

• A detailed representation does exist at each moment that our eyes are open, we are easily able to see that part of the scene being fixated.

• Although we have representations that large amounts of detailed information, and representations that provide coherence over time and space, we do not have representations to do both.