Journal
of
Experimental Psychology:
Human
Perception
and
Performance
1989.
Vol.
15. No. 2,
315-330
Copyright 1989
by the
American Psychological Association, Inc.
0096-1523/89/S00.75
Reflexive
and
Voluntary Orienting
of
Visual Attention:
Time Course
of
Activation
and
Resistance
to
Interruption
Hermann
J.
Muller
Department
of
Psychology,
Birkbeck
College
University
of
London, England
Patrick
M. A.
Rabbitt
Age
and
Cognitive Performance Research Centre
University
of
Manchester, England
To
study
the
mechanisms underlying covert orienting
of
attention
in
visual
space, subjects were
given
advance cues indicating
the
probable locations
of
targets that
they
had to
discriminate
and
localize. Direct peripheral cues (brightening
of one of
four
boxes
in
peripheral vision)
and
symbolic
central cues
(an
arrow
at the
fixation
point indicating
a
probable peripheral box) were
compared. Peripheral
and
central cues
are
believed
to
activate
different
reflexive
and
voluntary
modes
of
orienting (Jonides,
1981;
Posner, 1980). Experiment
1
showed that
the
time courses
of
facilitation
and
inhibition
from
peripheral
and
central cues were characteristic
and
different.
Experiment
2
showed that voluntary orienting
in
response
to
symbolic central cues
is
interrupted
by
reflexive
orienting
to
random peripheral
flashes.
Experiment
3
showed that irrelevant periph-
eral
flashes
also compete with relevant peripheral cues.
The
amount
of
interference varied
systematically
with
the
interval between
the
onset
of the
relevant
cue and of the
distracting
flash
(cue-flash
onset asynchrony)
and
with
the
cuing condition. Taken together, these
effects
support
a
model
for
spatial attention with distinct
but
interacting
reflexive
and
voluntary orienting
mechanisms.
In
daily
life
persons move their eyes
so as to
foveate
parts
of
the
visual
field to
which they wish
to
attend.
However,
it
is
well
established
that
even without making
eye
movements,
persons
can
choose
to
attend
to
input
at
particular extrafoveal
locations (e.g., Colegate,
Hoffman,
&
Eriksen,
1973;
Eriksen
&
Colegate,
1971;
Eriksen
&
Hoffman,
1972; Posner, 1980;
Posner, Snyder,
&
Davidson, 1980). Posner
has
termed this
covert
visual orienting.
Posner
(1980)
proposed that there
are two
modes
of
control
over
covert visual orienting:
The first
mode
is
exogenous,
involving
reflexive
orienting
in
response
to
salient stimuli
in
the
visual
field,
such
as a
peripheral
flash
(peripheral cue).
The
second mode
is
endogenous, involving voluntary orient-
ing
in
response
to
symbolic indicators, such
as a
central arrow
(central
cue).
Posner's
two
modes
of
control over spatial
attention
are a
distinction between
two
ways
in
which
one
common
orienting
mechanism,
which brings
limited-capacity
attention
processes
to
bear
on
stimuli
at
selected spatial
regions,
can be
activated.'
Henceforth these
two
modes
of
orienting
are
referred
to as
reflexive
and
voluntary
(in
analogy
with
the
guidance
of
saccadic
eye
movements).
Jonides
(1981)
extended this distinction
by
proposing that
reflexive
and
voluntary orienting
differ
in
their automaticity.
This
research
was
supported
by
Medical Research Council grant
G84/03193N
to
Patrick
M. A.
Rabbitt
and by
Deutsche
Forschungs-
gemeinschaft
postdoctoral
fellowship
Mu773/l-l
to
Hermann
J.
Muller.
We
thank
J.
Duncan,
G. W.
Humphreys,
W.
Epstein,
and
three
anonymous reviewers
for
their extremely
helpful
comments
on
earlier versions
of
this article. Many
thanks
are
also
extended
to M.
Shepherd
for
providing
his
EMDISP
system
and to J. M.
Findlay
for
his
help
with
eye
movement monitoring.
Correspondence concerning this article should
be
addressed
to H.
J.
Muller,
Department
of
Psychology, Birkbeck College,
University
of
London,
Malet
Street, London,
WC1E
7HX, England.
Reflexive
orienting, triggered
by
peripheral cues,
is
automatic;
voluntary
orienting,
initiated
by
central
cues,
is
controlled.
Jonides
(1981)
found
that unlike controlled orienting
in re-
sponse
to
central cues, automatic orienting
to
peripheral cues
is
not
affected
by a
secondary memory task (Experiment
1)
or by the
relative
frequency
with which
different
kinds
of cue
are
given (Experiment
3) and
cannot
be
voluntarily sup-
pressed (Experiment
2).
Furthermore,
if the
magnitude
of
priming
is
computed
by
summing benefits
of
correct cuing
and
costs
of
incorrect cuing, peripheral cues produce
a
greater
effect
than central cues.
Similar
to
Posner
(1980),
Jonides
(1981)
assumed that there
is
one
mechanism responsible
for
movements
of the
mind's
eye
(i.e., orienting)
but
that this mechanism
may be
guided
either
by
voluntary
or by
automatic control.
If the
distinction
between automatic
and
controlled orienting rests only
on the
ways
in
which
attentional
orienting
is
initiated,
the
results
of
two
of
Jonides's
(1981)
experiments
are
unsurprising. Central
symbolic
cues, such
as
arrows, must
be
decoded before
the
spatial location that they designate
can be
determined.
Pe-
ripheral
cues convey spatial information directly because they
occur
at the
locations
at
which
the
subsequent target signals
are
likely
to be
presented. Thus,
it is not
surprising that
a
secondary
memory
load
(Experiment
1)
or the
relative
unex-
pectedness (within
a
block
of
trials)
of
central cues,
by de-
manding additional processing resources (see Navon, 1984),
may
interfere with
the
interpretation
of
symbolic central cues
more than that
of
direct peripheral cues.
1
Spatial orienting means allocation
of a
limited-capacity attention
system
to
selected
spatial
locations.
Thus,
determining
the
locations
to
which
spatial cues
refer
must occur prior
to
selection
and
orienting
itself
(see
Duncan's,
1980b,
notion
of a
selection schedule). Allocation
of
limited-capacity attention
is
accomplished through
an
orienting
mechanism
that
is
initiated
by the
output
of the
anteceding
cue
processing.
315
316
HERMANN
J.
MULLER
AND
PATRICK
M.
A.
RABBITT
Jonides's
(1981)
Experiment
2 is
more interesting.
It was
designed
to
test whether observers could suppress orienting
in
response
to
peripheral
and to
central cues. Jonides
(1981)
found
that when subjects were instructed
to
attend
to the
cue,
reaction times (RTs) were faster
to
targets
at
cued rather than
uncued
locations, both with peripheral
and
with central cues.
However,
when subjects were told
to
ignore
the
cue,
an
advantage
for
cued over uncued locations occurred only
with
peripheral cues. Jonides
(1981)
concluded that
it is
possible
to
suppress orienting
in
response
to
central cues
but not to
peripheral cues.
However,
on its own
this
finding
does
not
establish that
orienting
in
response
to
symbolic central cues cannot
be
automatic.
It can be
argued that attentional orienting, once
triggered,
proceeds
automatically
but
that
processes
that
an-
tecede orienting, such
as
decoding
of cue
directionality
or
validity,
can be
voluntarily suppressed. Nevertheless, Jon-
ides's
(1981)
Experiment
2 did
demonstrate that peripheral
cues,
by
contrast, cannot
be
prevented
from
activating
the
orienting
mechanism.
In
summary,
the
available evidence suggests that direction-
ality
of
direct peripheral
and
symbolic central cues
is
deter-
mined through separate processes,
but the
information
de-
rived
by
these processes
is fed
into
the
same orienting mech-
anism.
Processing
of
central
cues
can be
suppressed,
so
initiation
of the
orienting mechanism
is not
obligatory.
Miiller
and
Findlay
(in
press) replaced
the
assumption
of a
common orienting mechanism underlying
reflexive
and
vol-
untary
orienting
by the
proposal that there
are two
separate
mechanisms
that
come
into
play
at
different
times
after
cue
onset: Peripheral cues trigger both
a
fast-acting
reflexive
(au-
tomatic)
and a
slower-acting voluntary (controlled) orienting
mechanism.
The
rapid automatic mechanism
has a
powerful
but
transitory response that
fades
out
100-300
ms
after
cue
onset.
The
controlled mechanism
has a
longer
rise
time
and
is
less
effective
but
shows
a
longer persistence.
The
reflexive
mechanism
is
triggered
by
immediate physical properties
of
the cue
(e.g.,
the
abruptness
of
onset;
see
Jonides
&
Yantis,
1988;
Yantis
&
Jonides, 1984)
and is
little
affected
by cue
validity, that
is, the
probability with which
the cue
predicts
the
target location.
The
voluntary mechanism, however,
re-
quires development
of a
spatial expectancy (set)
on the
basis
of
the
probabilistic information provided
by the
cue. Central
cues initiate only
the
voluntary orienting mechanism.
Miiller
and
Findlay's
(in
press) suggestion
was
based
on
differences
in the
time course
of
facilitatory
(i.e., benefits
for
cued locations)
and
inhibitory (i.e., costs
for
uncued locations)
effects
produced
by
peripheral
and
central
cues.
2
They com-
pared relatively large ranges
of
intervals between
cue and
target onsets (stimulus onset asynchronies; SOAs), that
is, 50
to 700 ms.
With peripheral cues
the
peak facilitation
for
cued
locations occurred within
150
ms
after
cue
onset,
was
followed
by
a
decline, between
150-
and
300-ms SOAs,
and
sustained
facilitation
at the
lower
level;
inhibition
for
uncued locations
was
strongest
at
short SOAs
and
then showed
a
marked
reduction
within
300 ms
after
cue
onset. With central cues
the
facilitation
for
cued locations built
up
more gradually,
requiring
300 ms to
reach optimum,
and was
then maintained
at
optimum level; inhibition
for
uncued locations showed
some initial increase
and
later decrease. With SOAs
of
less
than
300 ms, the sum of
costs
and
benefits
was
greater
for
peripheral than
for
central cues.
At
longer SOAs peripheral
and
central cues
had the
same
effects.
The finding
that peripheral cues produce stronger
facilita-
tory
and
inhibitory
effects
at
short rather than
at
long SOAs
(see
also Jonides,
1981)
is
crucial
to
Miiller
and
Findlay's
(in
press) argument
for two
separate orienting mechanisms. Note
that although this
is a
quantitative
effect,
it
suggests that
separate mechanisms
are
involved. Were there only
one
ori-
enting
mechanism, that
is,
were allocation
of the
limited-
capacity
attention system triggered
on
registration (direct
peripheral cues)
or
derivation (symbolic central cues)
of one
common stimulus, then
the
action
of the
system ought
to
depend
in an
all-or-none
fashion
on the
presence
or
absence
of
this
stimulus.
3
Miiller
and
Findlay
(in
press) interpreted
the
stronger
effect
of
peripheral cues
at
short rather than
at
long SOAs
as
evidence that
reflexive
orienting, occurring rapidly
after
cue
onset,
is
characterized
by
greater
resistance
to
interruption
than voluntary orienting, which comes into play only later.
Strong
effects
of
peripheral cues were also
found
with single-
element displays
in
which targets (presented
for a
limited
exposure duration) appeared
on
their
own
(i.e.,
in
which
nontarget
locations
remained
empty).
When
a
single target occurred
at an
uncued location,
the
luminance
increment associated with
its
onset ought
to
have
attracted attention
by the
same (automatic) mechanism
as the
preceding brightening
from
the
peripheral
cue at the
cued
location.
However,
the
strong
inhibitory
effect
at
short
cue-
target SOAs suggests that when
the
automatic orienting mech-
anism
is
engaged
by the
peripheral cue,
it
cannot
be
inter-
rupted
by the
onset
of a
target
at an
uncued location.
The
decrease
in the
inhibitory
effect
at
longer SOAs
can be ex-
plained
if
after
some delay
automatic
orienting
fades
out and
is
replaced
by the
controlled mechanism. Voluntary orienting
to the
cued location
can be
interrupted
by an
automatic
orienting
response
to a
competing target, reducing
any
inhi-
bition
from
an
incorrect cue.
If
reflexive
and
voluntary ori-
enting
involve separate mechanisms that, nevertheless,
ad-
dress
a
common limited-capacity attention system,
it is not
immediately clear
why
voluntary orienting ought
to be
inter-
rupted
by a
stimulus triggering
a
competing
reflexive
orienting
response. However, this
can
easily
be
explained
by
assuming
that
the
reflexive
mechanism modulates
the
voluntary
mech-
2
Differences
in
time courses (i.e.,
rise
times) between peripheral
and
central cuing presumably
reflect
differences
in the
time required
to
process
the
cues (see
Eriksen
&
Colegate,
1971;
Eriksen
&
Hoffman,
1972)
and as
such
are not
surprising;
in
contrast,
differences
in
magnitudes
of
cuing
effects
are.
3
This presupposes that
the
orienting mechanism
is
released
by its
proper (suprathreshold) trigger stimulus, that
is,
that quantitative
aspects
of the
trigger
stimulus
have only
little
effect
on the
response
of
the
mechanism. With regard
to
peripheral cuing, this assumption
is
supported
by
Jonides
and
Yantis's (1988)
finding
that whether
or
not
reflexive
orienting occurs depends
on the
presence
or
absence
of
an
abrupt stimulus onset rather than
on
quantitative
stimulus
aspects
such
as
differences
in
luminance
or
hue.
REFLEXIVE
AND
VOLUNTARY ORIENTING
317
anism
through
an
inhibitory connection.
To the
extent that
interruptability
constitutes
a
criterion
for the
automaticity
of
a
given mechanism (e.g., Jonides, 1981;
LaBerge,
1981;
Schneider.
&
Shiffrin,
1977;
Shiffrin
&
Schneider,
1977),
reflexive
and
voluntary orienting
can be
characterized
as
automatic
and
controlled, respectively.
Whereas
Miiller
and
Findlay's
(in
press)
two-mechanism
model seems
a
plausible extension
of
Posner's
(1980)
and
Jonides's
(1981)
proposals,
their time course data
do not
however
rule
out
alternative interpretations.
One
effect,
the
decline
in
facilitation
at
long SOAs that occurs with peripheral
but
not
with central cues,
may
reflect
Posner
and
Cohen's
(1984)
inhibition
effect
(see also Maylor,
1985;
Posner,
Rafal,
Choate,
&
Vaughan, 1985) rather than
the
transition
from
reflexive
to
voluntary orienting. According
to
Posner
and
Cohen (1984),
"some
part
of the
pathway
from
the
cued
location
is
reduced
in
efficiency
by the
[peripheral] cuing"
(p.
537), thus
favoring
the
sampling
of
areas
of the
visual
field
at
which
there
was no
previous change
in
light energy.
Posner
and
Cohen's
inhibition
effect
has
only been
found
with spa-
tially
uninformative cues
but not
when targets were more
likely
to
occur
at
cued than
at
uncued locations.
Posner
and
Cohen concluded
that
inhibition
may be
overlaid
by
facilita-
tion
produced
by
active orienting
to the
cued
location.
Be-
cause central cuing does
not
involve
a
change
in
light energy
at
the
indicated
location,
no
inhibitory consequences
for
this
location
are to be
expected. This
is
consistent with Miiller
and
Findlay's
(in
press) result with
central
cues,
for
which
facilitation
(cued locations)
is
maintained
at
optimum level
at
long SOAs.
A
second
finding
that
may be at
variance with Miiller
and
Findlay's
(in
press) two-mechanism model
is
that central cues,
too, showed evidence
of
stronger
inhibition
at
short
rather
than
at
long SOAs. This
may
indicate that early orienting
to
central
cues
is as
resistant
to
interruption
as is
early orienting
to
peripheral cues.
These alternative accounts cannot
be
refuted
from
Miiller
and
Findlay's
(in
press)
data,
because direct comparisons
between
their peripheral-
and
central-cuing conditions
are
limited.
Either
these
conditions were presented
one
after
the
other, with
different
subjects
and
different
threshold exposure
times
(Experiments
1 and 2), or
crucial
cue-target
SOAs were
not
tested (Experiment
4).
Experiment
1 of our
study
was
designed
to
establish
the
time course
of
reflexive
and
volun-
tary
orienting mechanisms
by
directly comparing
peripheral-
and
central-cuing conditions. This comparison
was
expected
to
show whether
reflexive
orienting
is
more
effective
and
less
interruptable
than voluntary orienting
and
whether Posner
and
Cohen's
(1984)
inhibition
effect
can be
suppressed
by
voluntary
orienting.
If
reflexive
orienting
is
more
effective
and
less interruptable than controlled orienting, optimum
facilitation
produced
by
peripheral cues
at
short SOAs ought
to be
greater than
facilitation
that
can be
sustained
with
peripheral
or
central cues
at
long SOAs,
and
inhibition
at
short
SOAs ought
to be
stronger with peripheral than
with
central
cues. Furthermore,
if
Posner
and
Cohen's
inhibition
effect
is
suppressed
by
voluntary orienting,
facilitation
ought
not to
differ
between peripheral
and
central cuing
at
long
SOAs.
Experiments
2 and 3
were designed
to
test directly
the
interruptability
of the two
orienting mechanisms hypothe-
sized
to
underly
the
time course data
of
Experiment
1.
Inter-
ruptability
was
indexed
by the
degree
to
which
orienting
in
response
to
cues
was
interfered with
by
task-irrelevant (i.e.,
uninformative)
peripheral
flashes.
Unlike Jonides
(1981,
Ex-
periment
2),
Experiments
2 and 3 did not
measure
resistance
to
suppression
but
rather resistance
to
competition, that
is,
interruptability
of
orienting
after
it was
initiated.
If
reflexive
orienting
is
indeed automatic,
it
ought
to be
highly resistant
to the
competition
of
peripheral
flashes at
uncued locations.
If
voluntary orienting
is
controlled,
it
ought
to be
less resistant
to
competition.
Experiment
1
Method
Apparatus
and
materials.
The
stimuli were presented
on a
Tek-
tronix
608 X-Y
display
with
P-31
phosphor.
The
cathode-ray tube
(CRT)
was
controlled
by a
LSI-11/23
computer through
a
CED-502
interface;
the
display system used
was
EMDISP
(Shepherd,
1984).
The
laboratory
was
dimly illuminated. Stimulus luminance
was
0.343
cd/
m
2
,
and
screen background luminance
was
0.034
cd/m
2
.
Subjects
viewed
the CRT
from
a
distance
of 50 cm
with
their heads resting
on
a
chin rest.
Display
and
timing. Figure
1
illustrates
the
sequence
of
frames
presented
on a
given trial. Frame
1
displayed
a
central
fixation
dot
and
four
peripheral
boxes
marking
the
possible
target
locations.
After
500
ms
the fixation dot was
replaced
by a
central
box
containing
a T
in
one of the
four
orthogonal orientations (Frame
2),
which
was
presented
for
1,500
ms;
this
T was the
comparison stimulus
for the
trial.
Then
the fixation dot
reappeared (Frame
3), and
1,000
ms
later
a
spatial
cue
indicating
the
likely target location
was
presented.
The
cue
was
either
a
50-ms
brightening
of the
outline
of one of the
peripheral boxes (peripheral cue; Frame
4a) or a
50-ms arrow indi-
cator
in the
center (central cue; Frame 4b).
In
addition, baseline trials
were
given
on
which
either
all
four
boxes were brightened
simulta-
neously
(peripheral
cuing)
or a
cross
appeared
in the
center (central
cuing);
these events indicated that
the
target
was
equally likely
to
appear
at all
four
locations.
After
variable
cue-target
SOAs
the
target
stimulus,
a T in the
same
or a
different
(orthogonal) orientation
as
the
comparison
T, was
presented
in one of the
four
boxes
for a
limited
exposure duration;
the
three
nontarget
locations contained distractor
crosses
of the
same size
and
luminance
as the
target.
On
valid trials
(Frame
5a),
the
target appeared
at the
cued location.
On
invalid trials
(Frame 5b),
the
target occurred
at one of the
uncued locations. Target
eccentricity
was
4.40°,
and its
size
was
0.25".
In
Frame
6, the
target
and
distractors were terminated
by
contour masks (squares composed
of
0.25°
lines).
Task.
The
subjects
had to
give three responses
on a
hand-held
keypad:
(1)
a
same-different response
to
indicate whether
the
orien-
tation
of the
target
T
(Frame
6) was the
same
as or
different
from
that
of the
comparison
T
(Frame
2); (2) a
position response
to
indicate
in
which
of the
four
boxes
the
target
T had
appeared;
and (3)
another
position
response
to
indicate
which
position
had
been cued
(0 on
baseline
trials). This third response ensured that
a
trial
was
accepted
only
when
decoding
of cue
direction
was
successful.
4
If the
decoding
was
not
successful,
the
trial
was
rejected
and
rerun later
in the
block.
4
A
correct cued position response
to
accept
a
trial
was
required
because Miiller
(1984)
observed that
if
central arrows
are
presented
for
brief times
only,
their
direction
may be
confused
(i.e.,
confusions
between
cued
and
diagonally opposite locations),
which
may
lead
to
orienting
to an
uncued
location
and
thus reduced
priming
effects.
318
HERMANN
J.
MULLER
AND
PATRICK
M. A.
RABBITT
HL
U
L
•
D
E
1
1
1
Tl
'
1
1
1
1
L
1 1
1
Figure
L
Sequence
of
frames
presented
on a
trial.
Design
and
procedure. There were three variables:
(a)
peripheral
or
central cuing,
(b)
valid
or
invalid trial,
and (c)
cue-target
SOA
(100, 175, 275, 400, 550,
or 725
ms).
The
experiment
was
divided
into
two
4-hr sessions, each consisting
of two
blocks
of
peripheral-
and two
blocks
of
central-cuing trials (360 trials
per
block);
the
order
of
blocks
was
counterbalanced across subjects
and
sessions. Central
and
peripheral cues were presented
in
separate blocks, rather than
in
random order within
the
same block, because expectedness
of
central
indicators
affects
cue
processing (Jonides, 1981).
Cue
validity
and
SOA
conditions were presented
in
random order within
a
block
of
trials.
The
four
orientations
of
target
Ts
were presented with equal
frequency.
In
half
the
trials
the
targets were
the
same
as
comparison
Ts; in the
other half they were
different.
Targets appeared with equal
frequency
at
each
of the
four
peripheral locations. Cues were valid
on
half
the
trials
and
invalid
on the
other
half.
The
cued
position
was
three times
as
likely
to
contain
the
target
as any one of the
three
uncued locations.
The
number
of
trials totaled 2,880,
that
is,
1,440
peripheral-cuing
and
1,440
central-cuing trials
(6
SOAs
x [96
Valid
Trials
+ 96
Invalid Trials
+ 48
Baseline
Trials]).
Determination
of
the
threshold exposure durations.
At the
begin-
ning
of
each session, target exposure times were determined individ-
ually
for
each subject
and
separately
for
each cuing condition (with
the
order
of
conditions counterbalanced across sessions
and
subjects).
The
estimation procedure used
was a
modified Probability Estimation
by
Sequential Testing (PEST) adaptive staircase (Findlay, 1978),
which
aimed
at a
threshold
level
of 75% on
baseline trials. (Threshold
trials
presented
either
a
central
cross
or
four
simultaneous
peripheral
flashes,
followed
by the
target
at
SOAs randomly chosen
from
the set
of
SOAs
in
Experiment
1.)
This procedure
was
based
on
jointly
correct
same-different
and
position responses; that
is, an
error
was
counted
if one or
both responses were incorrect.
The
target exposure
times introduced
in
peripheral-
and in
central-cuing blocks were
the
means
of the 75%
thresholds estimated
for the two
cuing conditions.
Peripheral-
and
central-cuing baseline trials (with
fixed
exposure
duration)
in
Experiment
1
served
to
provide
a
check
on the
validity
and
stability
of the
threshold estimates.
Instructions.
The
subjects were asked
to
respond
as
accurately
as
possible. They were
fully
briefed
on all
probability contingencies
described
earlier. They were
instructed
to
attend
to the
cued
location
but to
maintain
fixation
on the
central dot.
Subjects.
Four subjects participated
in
Experiment
1.
Their ages
ranged
from
20 to 30; all had
normal vision. Each
was
paid £5.0
(about $8.75)
for
each 4-hr session. Three subjects
had
taken part
in
the
study
by
Muller
and
Findlay
(in
press).
One
received
a
similar
amount
of
practice
in
four
preexperimental
sessions.
The
subjects'
exposure times were
as
follows (means
of two
sessions): 49.5, 44.0,
48.0,
and
46.5
ms.
Eye-movement control. Because
eye
position
was not
monitored
in
Experiment
1,
effects
of
attention
on
accuracy might
be
con-
founded
with
effects
of
retinal eccentricity (i.e.,
saccades
to
cued
locations could
be
executed
at
longer
cue-target
SOAs).
To
rule
out
this possibility, Experiment
1 was
repeated
for all 4
subjects with
the
monitoring
of eye
movements. Design
and
procedure were
the
same
as
before, except
that
only
two
SOAs
(175
and 725 ms)
were presented,
and
baseline trials were
not
included. Horizontal movement compo-
nents were monitored
by
using
a
modification
of
Findlay's
(1974)
limbus-tracking technique
in
which
a fiber-optic
Y-guide
is
used both
to
illuminate
the
iris-sclera
boundary
and to
detect reflected light
(see Findlay,
1981,
for
further details). Output
from
the
eye-move-
ment
recorder
was
sampled
every
10
ms,
starting
at cue
onset
and
ending
150
ms
after
target onset. Samples were analyzed
at the end
of
the
trial.
If
a
saccade
was
detected (velocity criterion:
two
successive
sample differences, both
in the
same direction, exceeding approxi-
mately
30°/s),
the
trial
was
rejected
and
rerun later
in the
block.
5
This
eye-movement control leaves
the
possibility
of
slow
fixation
drifts
(e.g.,
Kowler
&
Steinman, 1979a,
I979b).
Muller
and
Findlay (1987),
investigating
eye fixation in a
spatial-cuing task
with
the
same displays
as
in
this study, found that such
drifts
do
indeed occur. However,
only
about
60% of
drifts
were
in the
cued, that
is,
attended, direction,
and
drift
size
was too
small (75% were smaller than 0.375°,
and
95%,
smaller than 0.625°)
to
have
a
noticeable
effect
on
accuracy. Drifts
in
excess
of
about
1° are
required
to
affect
accuracy.
Analysis.
The
main performance measure
was the
joint proba-
bility
of a
correct position (CP)
and a
correct same-different (CSD)
response, p(CP, CSD). Note that
any
bias inherent
in one
measure
can
call into question whether
effects
observed
are
attentional
in
nature (see Duncan, 1980a).
In
particular, guesses based
on the
knowledge
that targets
are
more likely
to
occur
at
cued than
at
uncued
locations (i.e.,
a
bias
in the
localization task)
can
inflate
accuracy
for
valid
trials
at the
expense
of
invalid trials.
One
possible solution
is to
analyze
the
discrimination data
on
their own. However, detailed
analyses
of
errors
in
this task
(Muller
&
Rabbitt, 1988) revealed that
if
target localization
fails,
the
same-different response
is a
guess; that
is,
the
conditional probability
of a
correct
same-different response
given
an
incorrect position (IP) response, p(CSD
/
IP),
is
close
to .5.
That
is,
discrimination accuracy
with
localization
failures
adds
little
5
Saccade probabilities averaged .009, .013,
and
.015
for
Experi-
ments
1, 2, and 3,
respectively,
and did not
differ
between experi-
mental
conditions.
REFLEXIVE
AND
VOLUNTARY ORIENTING
319
information
to
that
provided
by
discrimination accuracy with suc-
cessful
localization, that
is,
p(CP, CSD). Thus, analysis
of
p(CP,
CSD)
can be
justified
if
p(CSD
/ IP) is
close
to
chance
and
inde-
pendent
of
experimental
variables.
6
Results
and
Discussion
Figure
2
presents mean values
ofp(CP,
CSD)
as a
function
of
cue-target
SOA, separately
for
central
and
peripheral cuing,
PV A
CVA
PB
•
CB
O
PI
•
ci
a
.9-
n
tr>
.8
u
o_
<-"
.7
.6-
.5-,
I I
100
175
275
400 550
SOfl
(msec)
725
.9-
a
in
.8-
u
.7-
a.
.6-
.5-
\
\
100 175
1 1
1
275
400
550
SOR
(msec
)
1
725
Figure
2. The
probability
of
correct
position-correct
same-different
responses, p(CP, CSD),
as a
function
of
cue-target
stimulus onset
asynchrony
(SOA) plotted separately
for
peripheral-cuing
(P)
valid
(V),
baseline (B),
and
invalid
(I)
trials
and for
central-cuing
(C) V, B,
and
I
trials. (Top,
data
from
experimental trials; bottom, data
from
eye-movement
control trials.)
valid
and
invalid trials,
and
baseline trials. Figure
2
also
presents
the
corresponding
data
for the
eye-movement control
condition.
The
values
ofp(CP,
CSD) were
arcsin-transformed
(Winer,
1971)
and
evaluated
in a
three-way analysis
of
vari-
ance
(ANOVA)
with main terms
for
central
or
peripheral cuing,
valid
or
invalid trial,
and SOA
duration.
All
comparisons
between
experimental conditions
presented
are
based
on the
Tukey
method.
The
three-way interaction
was
significant, F(5,
15)
=
40.79,
p
<
.001,
because
at
100-
and
175-ms
SO
As,
costs
plus benefits
were
greater
for
peripheral than
for
central cuing.
SOA
func-
tions
for
peripheral
and
central cuing
and for
valid
and
invalid
trials
are
characterized
as
follows:
Peripheral
cuing. Valid trials (cued locations) showed
a
fast
rise in
accuracy, peaking
at
175-ms
SO
As;
thereafter,
accuracy declined toward
400-ms
SOAs
and
then remained
constant
(p <
.001
for all
comparisons between SOAs shorter
and
longer than
275
ms).
Invalid trials (uncued locations)
showed
an
improvement between
100-
and
400-ms
SOAs
and
then
remained invariant
at the
higher level
(p <
.001
for all
comparisons between 100-
to
175-ms
SOAs
and the
longer
SOAs).
Central
cuing. Valid trials showed
a
gradual buildup
in
accuracy, particularly marked between
100-
and
275-ms
SOAs
(p
<
.001
for all
comparisons between SOAs shorter
and
longer
than
175
ms). Invalid trials showed some
decrease
between
100-
and
175-ms
SOAs
(p <
.05), followed
by a
tendency
to
increase towards 400-ms SOAs.
Valid
trials. Peripheral cues showed higher accuracy
at
100-
and
175-ms
SOAs than central cues
at the
same
(p <
.001)
and at
longer SOAs
(p <
.01).
Invalid
trials. Peripheral cues showed lower accuracy
at
100-
and
175-ms
SOAs than
did
central cues
(p <
.005).
Valid
and
invalid trials.
There
are no
reliable
differences
between
peripheral
and
central cuing
at
SOAs longer than
175
ms.
Figure
2
shows that
the
eye-movement control condition
is
characterized
by the
same pattern (Peripheral
vs.
Central
Cuing
X
Valid
vs.
Invalid Trials
x SOA
Interaction).
That
is,
this
pattern cannot
be
attributed
to
strategic
eye
movements.
These results
are
consistent with
the
two-mechanism model
of
spatial orienting:
The
fast-acting mechanism triggered
by
peripheral cues produces
a
powerful
facilitatory
effect
for
cued
locations
(100-
to
175-ms
SOAs). This mechanism
is
effective
only
for a
short time,
and as it
fades
out, facilitation
for
cued
locations declines. However, within
275-400
ms
after
cue
onset,
a
second mechanism comes into
effect
that enables
attention
(i.e., facilitation)
for the
cued position
to be
sus-
tained, even though
at a
lower level. Central cues initiate only
this second, slower-acting mechanism.
6
Analyses
of
variance
of
arcsin probability
of
correct
same-differ-
ent
response given incorrect position response, p(CSD/IP),
for Ex-
periments
1,2,
and 3 did not
reveal
any
significant
effects.
The
overall
values
of
p(CSD/IP)—.533,
.535,
and
.528,
respectively—differed
only
little
from
chance, thus
justifying
the use of
probability
of
correct
position-correct
same-different
responses, p(CP, CSD),
as a
perform-
ance measure.