Example: seeing a bird that is singing in a tree or miss a road sign in plain sight Cell phone use while driving reduces attention and memory for billboards despite that fact that people still fixated on the signs (Strayer, Drews, and Johnston, 2003).
Luck and Vogel (1997) used a sequential comparison change detection task. They found that performance dropped with set sizes larger than four. Verbal loads and cues did not impact performance.
Change detection is only good for the saccade target (Bridgeman, Hendry, and Stark, 1975; Grimes, 1996; McConkie and Zola, 1979) Hollingworth, Richard, and Luck (2008) found gaze corrections occur unconsciously and require VSTM. Target circle expanded in size to inform participants they should focus attention on that item.
The flicker paradigm eliminates both of these. If they are to the cause then the detection should be easy in the flicker paradigm. If attention is to blame, then the flicker caused by blank fields would interfere with the local motion signals preventing attention from being drawn to that area. People would fail to see large changes even with plenty of viewing time and when changes aren’t synchronized to saccades. They argue that attention is the key factor in both types of change blindness.
Gray screen with a white rectangle in the center. A word appeared in the rectangle in the cue condition (3000 ms) (word cues were only used in experiment 3) 1000 ms gray field Flicker sequence for 60 sec or until a response was made Pictures were displayed for 240 ms Gray screen was displayed for 80 sec A = original image A` = modified image Image order was A, A, A`, A`, …
Ten participants were used in each experiment Five additional participants were used to determine what was central interest (CI) and marginal interest (MI)
Experiments used 48 color images of real world scenes 27 deg. W x 18 deg. H One change (color, location, or presence v. absence) was made to an object or area Changes were also divided by degree of interest (CI v. MI) CI had 3+/5 people mention, MI had 0/5 people mention. Changes in intensity and color were similar for CI and MI The area of change in MI was on average 22 sq. deg. larger than in CI
Participants had to push a key when a change is observed, then give a verbal description of the change. Participants were told of the types of changes and had six practice trials (two ex. of each type). Images occurred in random order. DV was the average number of alterations (proportional to RT) before the change was detected. Only responses that included the correct type of change and the object or area being changed were counted as correct. Low error rate of 1.2% in all the experiments.
MI: change detection took an average of 17.1 alterations (10.9 s). Some images required over 80 alterations (50 s) for a change that appeared obvious once noticed. CI: much faster with an average of 7.3 alterations (4.7 s). MI took significantly longer than CI (p <.001 for presence vs. absence, p <.05 for color, p <.001 for location) despite the fact the MI changes were 20 % larger in area on average. The blanks were then removed to see if the changes were easy to see. Detection only took an of average 1.4 alterations (0.9 s). There was no difference between MI and CI or in any different type of change.
Because it takes about 400 ms to process and consolidate an image in memory (Potter, 1976), the 240 ms display may not be sufficient to form a memory. The blanks between the two original and two modified images were removed. New pattern is A, A`, A, A`, … Image A is shown for 560 ms, then blank for 80 ms, then image A` is shown for 560 ms. RTs for MI and CI for all three change types were NOT significantly different from the RTs in experiment 1. The dashed lines represent the results from experiment 1.
The flicker paradigm may reduce visibility making the image difficult to see. To test for this, a verbal cue (one or two words) was placed in a white rectangle for three seconds at the beginning of each trial Partially valid: half valid and half invalid Completely valid: always valid Valid cues always caused MI and CI detection to speed up. Partially valid (p <.001 for MI; p <.03 for CI) Completely valid (p <.001 for MI and CI) No difference between MI and CI for completely valid. Poor visibility did not cause change blindness in the flicker paradigm.
They propose: Changes in an image can only be perceived when attention is focused on the object that changes. Visual memory contents are overwritten by subsequent stimuli without focused attention. The overwritten contents cannot be used to make a comparison. Attention appears to allow structures to be entered in a relatively durable store (VSTM) which allows comparisons to be made. It may be that the faster performance in CI compared to MI was due to attracting attention to the high interest areas. If correct, this points to a connection between four different areas of vision: Eye movements Visual attention Visual memory Scene perception
The ISI of 80 ms is well within the 300 ms limit of iconic memory. This implies that attended items enter VSTM and can be transformed when a change is made but unattended items are replaced by new items without awareness. There is a strong connection between attention and scene perception. Valid cues helped detection: Creates a stronger representation that can allow for change detection. Invalid cues did not help: Increased attentional scanning alone does not increase performance. Because attention can only be allocated to a few items at one time (Pashler, 1987; Pylyshyn & Storm, 1988; Wolfe et al., 1989), it is likely that only a few changes can be perceived at one time By measuring the detection rate for different elements in a scene it may be possible to determine the order in which different elements in a scene are attended and thus perceived. The relationship between RT and level of interest opens up the possibility that the flicker paradigm could be adapted to determine what nonverbal observers (e.g., infants and animals) find interesting in a scene.