You are on page 1of 3

VISUAL SEARCH : HOW DO WE FIND WHAT WE ARE LOOKING FOR?

A task familiar to everyone is searching through a page in the phone book for a
particular number, e.g. having located the correct surname and not being sure of
the initial, we search down the suburbs, keeping the relevant one in mind. We can
sometimes search down the numbers themselves, knowing the target number
begins, say, with 90. Another example of visual search would be scanning a list of
index words looking for particular topics.

How do we perform such tasks? Do we just use visual information in these


tasks, e.g. looking for visual features such as curved lines? How do we know
when we have reached the item we want? Can we look for more than one target
at a time just as easily as for one? The whole problem concerns the skill of visual
search and pattern identification.

Neisser (1967), discussed some of the answers proposed to these questions.


He referred to earlier experiments in which subjects scan down lists of letters for
a target letter at an unpredictable position. The average time taken to find a letter
varies as a function of the context: for example, it takes longer to find a `K' among
straight-line letters than among those with curved features like `S' and `C'.
Neisser also found that after considerable practice at the task, subjects were just
as fast locating any of ten target letters as they were locating only one.

Following these experiments with single-letter targets, Neisser and Beller


(1965) reported two experiments in which the targets were words. They compared
the search time required to locate three types of targets: firstly, a familiar word,
e.g. `Monday'; secondly, one of a familiar group of words, e.g. one of the
American States; and thirdly, a target defined in terms of its meaning alone, e.g.
`an animal' or `a person's first name'. Neisser and Beller suggested that the first
two tasks involve `stimulus examination' or processing of the physical properties
of the stimuli, while the third also involves `memory examination' or reference to
stored semantic information.

This workshop is based on Experiment 1 of Neisser and Beller and examines


the hypothesis that scanning will be faster when the target can be distinguished
by stimulus examination alone than when its definition requires memory
examination.

GENERAL PROCEDURE
You will need to decide first what sort of targets you will use to test the
hypothesis - you can follow Neisser and Beller's experiment, or, better still,
devise your own stimuli. Each target will be embedded in a list and there will
have to be several lists for each target type. Neisser and Beller discuss how
they compiled their lists using a computer, but several student experimenters
can collaborate in preparing this workshop and share the task of devising the
word lists so that they are drawn up reasonably quickly.
Lists of 50 words are a good length, i.e. 49 extra words plus the target word.
Each list can be typed on a separate sheet, vertically down the page, with one
word on each line. The pages can then be collated into booklet form. The target
items are randomly positioned in each list.

Each subject is tested individually, so, depending on the time available in the
laboratory session, you may need either to collect data beforehand and simply
demonstrate the task in the session or have subjects working in pairs in class,
taking turns to be experimenter and subject. (This would mean that you would
need two sets of the lists: simply use the same 49 words of each list and slot in
different targets.)

Chronoscopes or stopwatches are needed for timing the search of each list. On
any trial, the experimenter tells the subject the target and then starts timing when
the subject begins to search, stopping the timer when the target is located. You
will need to prepare record sheets for experimenters to note down the time taken
to search each list. It is suggested that you have 8 lists for each target type.
Leave the first two trials as practice and score only trials 3 to 8 for each target
type.
This might be an example of a record sheet:

Set 1 Target a.

List Time Position of Time diff. Position Time per


target diff. item
3 3s 11 5s 10 5/10=0.5s
4 8s 21 etc

The procedure for calculating the average search time per item is as follows.
For each target type, compare lists 3 and 4, 4 and 5, 5 and 6, 6 and 7, 7 and 8.
Calculate the time difference between the two searches and the position
difference, i.e. the number of lines separating the targets. To calculate the time
per item, divide the time difference by the position difference. This may seem
complicated but you will find it straightforward once you have your data. (Note
that it is theoretically possible to get negative times per item: these can be
ignored and omitted from final averages calculations.) For each subject, find the
median search time per item for each target type. Then finally calculate the group
average for each target type by finding the mean of these subject medians.

Is scanning time slower when memory examination is involved, as suggested


by the hypothesis? When you look at Neisser and Beller's results, remember that
they were examining the effect of practice on visual search performance over a six-
week period. You can compare your data only with the initial trials of their study (p.
353).

Useful information could be gained by asking the subjects how they did the task.
What role do sub-vocalisations play in their performance?
ALTERNATIVES
Although Neisser (1967) presented a model of visual search which emphasised the
visual processes involved, subsequent research has shown that this is not always
applicable: for example, acoustic factors can be involved, as demonstrated in the
everyday example of repeating the target word to yourself as you search. Rabbitt
(1978), in a thorough analysis of visual search, discussed studies which show this
effect (pp. 107-9). For an alternative experiment, you could adopt the method of one
of the studies he mentioned there, e.g. ask subjects to search through lists for
targets varying in their acoustic similarity to the background stimuli. If you find an
effect of acoustic factors on visual search, you could consider the implications for
Neisser's theory, noting Rabbitt's point that practice in searching may remove the
effect.

Other data inconsistent with Neisser's model were obtained by Brand (1971). She
reported that for numeral targets, search through lists of other numerals was slower
than search through letter lists. You could make up lists of letters and digits to test
for this effect and refer again to Rabbitt (p. 100 ff.) for a discussion of this
categorisation effect.

REFERENCES
Neisser, U. Cognitive Psychology. New York: Appleton-Century-Crofts, 1967
(especially ch. 3).
Neisser, U., and Beller, H. K. Searching through word lists. British Journal of
Psychology, 1965, 56, 349-58.

REFERENCES FOR ALTERNATIVES


Brand, 7. Classification without identification in visual search. Quarterly Journal of
Experimental Psychology, 1971, 23, 178-86.
Rabbitt, P. Sorting, categorization, and visual search. In E. C. Carterette and M. P.
Friedman (eds). Handbook of Perception (vol. 9): Perceptual Processing. New
York: Academic Press, 1978, 85-134.

This workshop is from :


Bennet, A., Hausfeld, S., Reeve, R. A., & Smith, J. (1982). Workshops in cognitive
processes. Sydney: New South Wales University Press.

You might also like