johnston & wilson - perceptual processing of nontargets in an attention task, artykuły psychologiczne
[ Pobierz całość w formacie PDF ]
Memory & Cognition
1980, Vol. 8 (4), 372-377
Perceptual processing of nontargets
in an attention task
WILLIAM A. JOHNSTON and JOLYNN WILSON
University of Utah, Salt Lake City, Utah 84112
The capability of nontargets to qualitatively influence the semantic processing of coincident
targets was investigated in three experiments. Subjects were aurally presented a series of word
pairs and attempted to detect homonymic instances of a predesignated category (e.g., animals).
The nontarget with which a target (e.g., ANT) was paired was appropriate (e.g., CRAWLING),
inappropriate (e.g., UNCLE), or neutral (e.g., STRAW). Experiments 1 and 2 established that
detection of targets can be facilitated by appropriate nontargets and inhibited by inappropriate
ones. Thus, nontargets can influence the way in which targets are semantically represented.
Experiment 3 showed that this effect is eliminated when subjects are precued as to the ear of
entry of targets. Thus, precuing appears to curtail the perceptual processing of nontargets. The
data run counter to theories that claim that focused attention does not entail the perceptual
suppression of nontargets.
Attention is generally considered focused to the
extent that a target stimulus receives more processing
than concurrent nontarget stimuli. One class of theory
localizes this difference in processing at perception:
Target stimuli undergo more perceptual processing than
nontarget stimuli because of the perceptual suppression
of the latter stimuli. Theories of this class include the
early-selection theories of Broadbent (1958, 1971)and
Treisman (1964, 1969). Another class of attention
theory conceives of the difference in processing between
targets and nontargets as being nonperceptual in nature:
Target stimuli are accorded priority by a special attention
mechanism that operates on the irrepressible buildup of
perceptual data for all concurrent stimuli. Theories of
this class include the late-selection theories of Deutsch
and Deutsch (1963) and Norman (1968).
The trend in recent years has been away from early-
selection theories and toward late-selection theories
(e.g., LaBerge, 1975; Shiffrin & Schneider, 1977).
1
The decline in popularity of early-selection theories is
based, in part, on empirical indications that nontargets
in attention tasks are not suppressed but, rather, undergo
deep, semantic analysis (e.g., Corteen & Wood, 1972;
Lewis, 1970; MacKay, 1973; Shiffrin & Schneider, 1977,
Experiment 4). The present research is most closely
related to the Lewis (1970) and MacKay (1973) studies.
Lewis (1970) asked subjects to shadow (repeat)
words from one of two dichotically presented lists and
to ignore words from the other list. The principal
finding was that shadowing latency was longer when
coincident target (shadowed) and nontarget ("ignored")
words were semantically related than when they were
unrelated. Since the response to a target word was
affected by the meaning of the coincident nontarget
word, Lewis concluded that the nontarget words must
have undergone semantic analysis. MacKay (1973)
asked subjects to shadow an ambiguous message (e.g.,
"They threw stones at the bank yesterday.") that was
presented to one ear and to ignore any words presented
to the other ear. The principal finding, as inferred from
subsequent recognition tests, was that the interpretation
of the shadowed message was biased in the direction of
disambiguating words (e.g., "river" or "money") that
were presented to the other ear. Since the interpretation
of the target message was affected by the semantic
properties of nontarget words, MacKay concluded that
the nontarget words must have been processed to a
semantic level. In general, this entire line of research
reduces the plausibility of early-selection theories
by showing that attention to targets does not necessarily
prevent the semantic analysis of nontargets.
The present studies extend this line of research.
Subjects listened to a series of word pairs and attempted
to detect instances of a category such as "articles of
clothing." Experiments 1 and 2 sought to determine
whether the detectability of a homonymic target word
such as "socks" can be either enhanced by a coincident
nontarget word such as "smelly" (appropriate nontarget)
or hindered by one such as "punches" (inappropriate
nontarget). These studies were intended to show that
the semantic properties of nontargets can affect the
detectability of targets by controlling how they are
represented perceptually. Experiment 3 attempted to
discriminate between early- and late-selection theories
empirically by determining whether a manipulation that
increases the overall detectability of targets also reduces
the effects on target detectability of appropriate and
This entire research effort, from its inception, has profited
from the ideas, comments, and suggestions of Larry Jacoby.
We are grateful for his interest and assistance in this research
project. Requests for reprints should be addressed to William A.
Johnston, Department of Psychology, University of Utah,
Salt Lake City, Utah 84112.
Copyright 1980 Psychonomic Society, Inc.
372
0090-502X/80/040372-06500.85/0
NONTARGET PROCESSING 373
inappropriate nontargets. The reasoning behind Experi-
ment 3 will be developed after Experiments 1 and 2 are
presented.
test confirmed the pattern shown in Table 1. Relative to
neutral nontargets, inappropriate nontargets reduced the
likelihood of detecting targets but appropriate nontargets
had no effect on target detections. Although the inhibi-
tory effect of inappropriate nontargets was small in
absolute terms, it was highly consistent across subjects.
The inhibitory effect of inappropriate nontargets
is consistent with the findings of Lewis (1970) and
MacKay (1973) in indicating that nontargets in an
attention task undergo semantic analysis. However, a
symmetrical facilitative effect of appropriate nontargets
was not obtained. One possible reason for the lack of a
facilitative effect of appropriate nontargets is that they
were often mistaken for targets. Indeed, 6.5% of appro-
priate nontargets were actually given as responses
in lieu of the targets with which they were paired.
This target-nontarget confusion factor was successfully
reduced in Experiment 2 by changing from binaural
presentation of word pairs to dichotic presentation.
EXPERIMENT 1
Method
Subjects.
The subjects were 36 male and female students at
the University of Utah. They were recruited from introductory
psychology classes and received credit toward a higher grade for
their participation.
Procedures.
Subjects performed a target-detection task on a
succession of nine different lists of word pairs. A list consisted
of 81 word pairs, 9 of which contained a target word (i.e., an
instance of the target category). The word pairs were presented
at the rate of one every 1.5 sec. The interlist interval was 15 sec.
A target category was announced 10 sec in advance of each list.
The nine target words in a list were homonyms with at least two
distinct meanings, only one of which fell within the domain of
the target category. Targets were presented at the rate of one in
every successive block of nine pairs. Placement of targets
in
blocks was random, with the restriction that there be at least
four word pairs between any two targets.
The manipulation of interest was the nature of the nontarget
word with which a target word was paired. An appropriate
nontarget was related to the target meaning of the target word,
an inappropriate nontarget was related to a different meaning of
the target word, and a neutral nontarget was not related to the
target word. All three kinds of nontargets were represented, in
random order, in each successive set of three targets in a list.
Moreover, across subjects, all three kinds of nontargets were
represented for each target word. The full set of target and non-
target words used in the experiment is provided in the Appendix.
Nine different orders of both lists and target words within lists
were represented across subjects.
The subjects served individually in a sound-deadened cubicle.
The prerecorded lists were played to subjects from a tape recorder
through an amplifier and over stereophonic earphones. Presenta-
tion of the word pairs was binaural, rather than dichotic, in that
both words in a pair were played to both of a subject’s ears.
Subjects were instructed to repeat aloud every target word that
they detected; their responses were tape-recorded and later
scored for accuracy. The task may be considered a divided-
attention one, in that subjects had to process both members of
each pair in order to detect targets.
Results and Discussion
Across the nine lists, each subject received 27 targets
for each type of nontarget. The unit of analysis was the
number of targets detected. These units, pooled across
subjects and converted into percentages, are summarized
in the first row of Table 1. Type of nontarget attained
statistical significance in an analysis of variance of these
data [F(2,70)=4.05, MSe = .48].2 A Newman-Keuls
EXPERIMENT 2
Method
Experiment 2 was a precise replication of Experiment 1
except that presentation of word pairs was dichotic (one word to
each ear) rather than binaural and the number of subjects was
45 rather than 36. The target words in a list alternated randomly
between the two ears. Thus, divided attention was still required.
Results and Discussion
The data are summarized in the second row of
Table 1. The results deviated from those of Experi-
ment 1 in two ways: The overall rate of target detection
was increased from 41% to 54%, and both a facilitative
effect of appropriate nontargets and an inhibitory effect
of inappropriate nontargets were obtained. Type of non-
target attained significance in an analysis of variance of
these data [F(2,88) = 7.09, MSe
= .54]. Newman-Keuls
tests confirmed that all three types of nontargets dif-
fered from each other in terms of accuracy of target
detection. Again, the differences were small in magnitude
but highly consistent across subjects.
Experiment 2 extends the findings of Experiment 1
by indicating that the semantic analysis of nontargets
can enhance, as well as impede, the detection of targets.
The emergence of a facilitative effect of appropriate
nontargets was accompanied by a reduction in the rate
Table 1
Percent of Targets Detected in Experiments 1-3 as a Function of Type of Nontarget
Type of Nontarget
Experi-
Attention
Appro-
Inappro-
ment
N
Condition
priate
Neutral
pria te
1
2
3
3
35
Divided, Binaural
42
43
>
39
45
Divided, Dichotic
57
54
>
51
27
Divided, Dichotic
67
55
>
47
27
Focused, Dichotic
74
72
=
72
Note The > and = signs indicate whether or not two percentages were statistically different (p <. 05}.
374 JOHNSTON AND WILSON
at which these nontargets were "detected" in lieu of
targets: 2.6% in Experiment 2 compared with 6.5% in
Experiment 1. As a composite, the data of Experiments
1 and 2 show that the perceptual representations of
target stimuli can be affected qualitatively by the non-
targets with which they are paired. Experiment 3 was
designed to assess the implications of this phenomenon
for theories of attention.
four target categories for which that effect was most pronounced
in Experiment 2 were used in Experiment 3. These categories
were "body parts," "animals," "edibles," and "clothing." In
addition, each list was increased in length from 81 to 324 word
pairs. Targets still occurred at the rate of 1 in every nine word
pairs, but 27 of the targets in each list were new "filler" targets
that were unrelated to the nontargets with which they were
paired. The remaining nine targets in each list were the ones
used in Experiments 1 and 2 and with respect to which type of
nontarget was manipulated.
The lists were identically constructed for the focused-attention
condition, except that all targets ~vere presented to the left ear.
The left ear was chosen for target presentation so that appropri-
ate and inappropriate nontargets would enter the right ear and
be represented predominantly in the left hemisphere. This
procedure was expected to increase the opportunity for an
effect of type of nontarget in the focused-attention condition
and, thus, for the confirmation of late-selection theory. Subjects
in each condition were instructed as to the actual distribution
of targets between the two ears and were encouraged ~o dis-
tribute thek attention accotdirtgly.
EXPERIMENT 3
The third experiment attempted to discriminate
empirically between early- and late-selection theories
of attention by orthogonally manipulating type of
nontarget and precuing of targets. Ninio and Kahneman
(1974) examined the effect of precuing on detectability
of targets in a dichotic-listening task. In a precuing or
focused-attention condition, targets (animal names) were
confined to just one ear and subjects were so informed.
In a divided-attention condition, targets were randomly
distributed between the two ears, and subjects were so
informed. Speed and accuracy of target detection were
higher in the focused-attention condition than in the
divided-attention condition.
Early- and late-selection theories provide different
interpretations of the precuing effect. Early-selection
theories would argue that precuing enhances the detect-
ability of targets by reducing the perceptt~al processing
of nontargets. Late-selec~tion theories would argue that
precuing enhances the selection precision or automaticity
of the attention mechanism but has no effect on the
amount of perceptual processing of nontargets. An
empirical decision between these two classes of theory
could be rendered by examining the effect of precuing
on the amount of perceptual processing of nontargets.
Amount of perceptual processing of nontargets was
measured in terms of the magnitude of the effects of
appropriate and inappropriate nontargets. The logic
behind this technique was as follows: If the component
effects of type of nontarget, like the Lewis (1970) and
MacKay (1973) effects, reflect the perceptual processing
of nontargets, then variations in the magnitude of these
effects may reflect variations in the amount of perceptual
processing of nontargets. Within the context of this
logic, then, the two classes of attention theory make
different predictions. According to early-selection
theory, precuing should reduce the effects of appropri-
ate and inappropriate nontargets. According to late-
selection theory, precuing should not alter these effects.
of these data: precuing [F(1,52)
= 30.77, MSe
= 5.20],
type of nontarget [F(2,52)= 13.89, MSe
= 1.74], and
their interaction [F(2,52)
= 8.I0, MSe = 1.89]. The
precuing effect replicates the finding of Ninio and
Kahneman (1974): Target-detection accuracy was
higher under focused attention than under divided
attention. The effect of type of nontarget was the same
as that found in Experiment 2: Relative to neutral
nontargets, accuracy of target detection was enhanced
by appropriate nontargets and inhibited by inappropri-
ate nontargets.
The interaction is of principal interest. A Newman-
Keuls test revealed that the effect of type of nontarget
was restricted to the divided-attention condition. More-
over, the effect in that condition was larger, in absolute
terms, than it was in either of the first two experinaents.
Evidently, the list-selection procedure succeeded in
accentuating the effect of type of non’target. Nonethe-
less, this large effect was eliminated by precuing subiects
as to the ear of entry of targets.
Despite the fact that the average rate of target detec-
tion was well below 100% in the focused-attention
condition, one might argue that an effect of type of
nontarget in that condition was prevented by a perfor-
mance ceiling. This possibility was examined by breaking
both the focused-attention and divided-attenrion groups
into bottom, middle, and top thirds in terms of the
overall rate of target detection. The data for each
subgroup are summarized in Table 2. One-way analyses
of variance revealed that the effect of type of nontarget
was significant in none of the focused-attentior~ sub-
groups [F(2,16)
= 1.87, MSe =2.57, for the bottom
one-third; F = 1.26, MSe
= 3.82, for lhe middle one-
Method
A 2 by 3 design was used in which type of nontarget was
manipulated under both focused-attention and divided-attention
conditions. Each of the latter conditions was administered to a
group of 27 students.
The divided-attention condition was, in essence, a replication
of Experiment 2. That is, list presentation was dichotic and
targets were randomly distributed between the two ears. How-
ever, in an effort to increase the effect of type of target, only the
third; F
= 1.79, MSe
= .58, for the top one-third] but
was significant in all of the divided-attention subgroups
Results and Discussion
The results are summarized in the third (divided
attention) and fourth (focused attention) rows of
Table 1. All three experimental sources of variance
attained statistical significance in an analysis of variance
NONTARGET PROCESSING 375
Table 2
Effect of Type of Nontarget Under Focused and
Divided Attention
in Experiment 3 as a Function of Overall Performance
Level of Subjects (Bottom, Middle, and Top)
Type of Nontarget
(1979) found that the Lewis effect is reduced as targets
and nontargets are made more discriminable physically.
Finally, Underwood (1976) examined a visual version of
the Lewis effect and found it to be reduced when sub-
jects are precued as to the spatial position of the target
in the stimulus display) If the effect of type of nontar-
get reflects the perceptual processing of nontargets, then
the present and other recent data indicate that the
amount of this processing is variable. Perceptual process-
ing of nontargets appears to be curtailed, or suppressed,
as targets are rendered more detectable as a function of
such variables as practice, physical discriminability, and
precuing.
Proponents of late-selection theory might argue that
the observed variations in the effects of nontargets on
responses to targets reflect variations in amount of
nonperceptual, rather than perceptual, processing of
nontargets. Indeed, variability in the selection precision
of the attention mechanism has been explicitly built
into the theories of LaBerge (1975) and Shiffrin and
Schneider (1977). Under some conditions, the "attention
director" has to search through all of the perceptual data
for that portion that represents the target. Under other
conditions, the attention director "automatically detects"
the target representation. Perhaps responses to targets
are affected by the semantic (and other) properties
of nontargets only when conditions force the attention
mechanism to process nontargets in its "controlled
search" for targets.
As convincing as this nonperceptual line of reasoning
might be, it is inconsistent with the empirical attack
that late-selection theorists have launched against early-
selection theory. Specifically, the case against early-
selection theory has rested primarily on the argument
that the findings of Lewis (1970), MacKay (1973),
and others reflect the full perceptual processing of
nontargets. Late-selection theorists appear to be left
with the following choice: Either interpret Lewis-like
effects in terms of the perceptual processing of non-
targets and admit that this processing is suppressible,
or interpret these effects in terms of the nonperceptual
processing of nontargets and dismiss their empirical
case against early-selection theories. In either event,
early-selection theories of attention must be considered
to remain a viable alternative to late-selection theories.
Performance
Appro-
Inappro-
Mean
Level
priate
Neutral
priate
Level
Divided Attention Condition
Bottom
57
42
36 45
Middle
68 54 47 56
Top
76 68 58 67
Focused Attention Condition
Bottom
63
60
59 61
Middle
71 70 74 72
Top
88 86 82 85
Note-Bottom = lower third (N = 9) of subjects in terms of
overall detection accuracy; middle = middle third; top = upper
third.
[F=15.51, MSe=I.02, for the bottom one-third;
F=6.75, MSe=2.08, for the middle one-third; F=7.12,
MSe
= 1.41, for the top one-third].
It is important to note in Table 2 that the mean level
of target-detection accuracy was higher in the top
subgroup of divided-attention subjects than in the
bottom subgroup of focused-attention subjects. None-
theless, the effect of type of nontarget was reliable
only in the case of the former subgroup. Hence the
important interaction between precuing and type of
nontarget cannot be attributed to a mensural ceiling
under focused attention. This interaction is consistent
with early-selection theory but difficult to reconcile
with late-selection theory.
GENERAL DISCUSSION
The present data extend prior findings in two ways.
First, they establish further that nontarget stimuli can
be processed to a semantic level. As in the Lewis (1970)
and MacKay (1973) studies, processing of target stimuli
was affected by the semantic properties of nontargets.
The present results extend the boundaries of the prior
effects of type of nontarget by demonstrating that
the semantic properties of nontargets can affect the
accuracy, as well as latency, of target detection and can
enhance, as well as inhibit, target detection. Thus, non-
targets can have a qualitative effect on the perceptual
processing of the targets with which they are paired.
The second way in which the present research
extends prior f’mdings is by indicating that the amount
of semantic processing of nontargets is not invariant.
The component effects of type of nontarget disappeared
when targets were made more detectable by the precuing
manipulation. Similar findings have been obtained in
other recent studies. Treisman, Squire, and Green (1974)
found that the Lewis (1970) effect dissipates over the
course of a dichotic-listening trial. Johnston and Heinz
REFERENCE NOTE
1. Dawson, M. E., & Schell, A. M.
Electrodermal responses to
attended and nonattended significant stimuli.
Paper presented at
the meeting of the Society for Psychophysiological Research,
Cincinnati, October 1979.
REFERENCES
BaOADBF, r~T, D. E.
Perception and communication.
London:
Pergamon Press, 1958.
BaOADn~:NT, D. E.
Decision and stress.
New York: Academic
Press, 1971.
376
JOHNSTON AND WILSON
CORTEEN, R. S., & WOOD, B.
Automatic responses to shock
asso-
ciated
words in an unattended channel.
Journal of Experimental
Ps.vcholouy,
1972, 94, 308-313.
D~:t~TSC~t. J. A., & DEUTSCn, D. Attention: Some theoretical
considerations.
Psychological Review,
1963, "/0, 80-90.
,hntNS’rON, W. A., & HEINZ, S. P. Depth of nontarget processing
in an attention task.
Journal of Experimental Psychology:
Human Perception and Performance,
1979, $, 168-175.
LAB~:R~;r:, D. Acquisition of automatic processing in perceptual
and associative learning. In P. M. A. Rabbitt & S. Dornic
(Eds.},
Attention and performance V.
New York: Academic
Press, 1975.
L~:w~s. J. L. Semantic processing of unattended messages using
dichotic listening.
Journal of Experimental Ps.vcholog.v,
1970,
~5,225-228.
MACKAY, D. G. Aspects of the theory of comprehension, memory,
and attention.
Quarterly Journal of Experimental Psychology,
1973, 25, 22-40.
NINIO, A., & KAttNEMAN, D. Reaction time in focused and divided
attention.
Journal qf Experimental Psychology,
1974, 103,
394-399.
NO~MAh, D. A. Toward a theory of memory and attention.
Psycholo.~ical Review,
1968, 7~, 522-536.
S~t~vw~, R. M., & Sc~Ir~WID~m W. Controlled and automatic
human information processing: II. Perceptual learning, auto-
marie attending, and a general theory.
Ps~vchol¢~ical Review,
1977, ~4, 127-190.
T~:ISr~,,,N, A. M. Verbal cries, language and meaning in selective
attention.
American Journal ~f PEvchology,
1964, 77, 206-218.
TR~.:ISMAN, A. M. Strategies and models of ~electixe attention.
PEw’ho!o~ical Review,
1969, "/~, 282-299.
T~:Isr~A~’, A. M., S~t:I~r:, R., & O~:r:N, J. Semantic processing
in dichotic listening? A replication..’¢Iemot~v
& Co.~nition,
1974,
2,641-646.
UNm:rtwoor), G. Semantic interference from unattended primed
words.
BritZsh Journal ~fP.s~vchology,
1976, 67, 327-338.
NOTES
l. The term "late selection" does not accurately describe
these recent theories because they allow for the attention
director to operate in parallel with, as well as after, perceptual
processing of targets and nontargets. However, nontargets are
assumed to run their full perceptual course in either case.
2. The level of significance was set at p < .05 in all of the
statistical analyses referred to in this report.
3. Recent data (e.g., Dawson & Schell, Note 1) also call
into question other empirical indicants of the uniformly deep
perceptual processing of nontargets (e.g., Corteen & Wood,
1972).
Appendix
Target Categories, Homonyms, and the Nontargets of each Type (Appropriate, Inappropriate, and Neutral) Used in Experiments 1-3
Type of Nontarget
Target
Homonymic
Category
Target
Appropriate
Neutral
Inappropriate
Body Parts
arms
sleeves
plant
weapons
head
bald
time
leader
calf
muscular
stapler
cow
foot
shoe
cigar
inch
waist
belt
fountain
squander
chest
hairy
ruler
treasure
hair
comb
sun
rabbit
neck
choke
travel
kiss
nose
kleenex
arrest
understands
Household Items
sink
wash
green
titanic
hamper
dirty clothes
tortoise
irnpede
range
electric
contract
ranch
bureau
bedroom
sale
federal
chest
storage
cord
breast
cabinet
medicine
jacket
government
geiger
counter
kitchen
nest
vanity
make-up
degree
selfish
iron
steam
fish
ore
Child’s Toys
train
choo-choo
robin
teach
blocks
wooden
hem
barricades
banquet
ball
bouncing
smoke
scrabble
words
hour
quarrel
jacks
recess
family
flat tires
price
tag
touch
river
statue
marble
round
link
bottom
top
spinning
paper
swing
playground
chimney
sway
Animals
duck
quacking
movie
dodge
ant
crawling
straw
uncle
swallow
nesting
spoon
drink
turkey
gobble
ash
india
fly
buzz
wax
pilot
badger
caged
violin
harass
bear
hibernate
luck
naked
steer
roundup
jam
guide
deer
antler
amber
sweetheart
[ Pobierz całość w formacie PDF ]