RSVP課題における初頭効果とヴィジランスの低下(第23回大会 優秀発表賞抄録)

The Japanese Psychonomic Society

NII-Electronic Library Service

The JapanesePsychonomic Society

TheltpanesEJourrtalofRs),chonom･icSc,ience

2005,Vol,24,No.1,1I7-1l8

Summary

ofAwarded

PresentationIP25

The

transitions

of

time-dependentperformance

in

a

task

AtsunoriARiGA

and

Kazuhiko

7Vze

Universit.v

of

7bk.vo*YOKOSAWA

For prolonged cognitive tasks the eMciency ofan observer. as a

funetion

of test_period,isoften

illustratedas a

downward-sLoping

curve, This deteriorationiscalled thevigilancedecrement and

is

due

toashift

in

attention

during

atask. We havc investigatedthetemporal

features

of attention

by

using an RSVP task.We focused on

transient

_performance

in

a trial,

in

addjtion toprolongecl

perfurmance throughout the experiment,

Throughout

_the experiment a vigilance decrement

occurred.

During

a trial

however,

the dctectionof a targetat the beginning of a sequence was

dramatically

low

and recovered as itappcared later.This result has been shown neither

in

the vigilance studies, which _predicthigh _performance after the task onset, nor

in

the RSVP studies, which _prcdict high and fairdetectionof only targetinthe scquence. This result wou]d refiect a

gradual modulation of temporal attention toa rapid sequence.

Key

words: preparation cost, vigilance

decrement,

RSVP

Vigi!ance

studies

investigate

the ability of

observ-ers tomaintain their focusof awareness and remain alcrt tostimuli for_{prolonged periods}of time

_(Davies

&

Parasuraman,

1982).

These

studies

have

found

that thecthciency of observers

is

highest after the onset of a task and declinesover time. This iscalled

the vigi]ance decremenL However, _previous studies

have rarely ±ocused on transientchanges

in

perfor-mance.

Instead

they

have

compared perforrnance at several periods of time,_generallyat the beginning of

an episode and then

during

some

time

later.

Indccd

it

is

known that preparation for a task isonly

com-pleted byobservers after the onset of thctask.This suggests that_performance

is

hindered

just

after the onset of

the

task,

due

tothecost of this_preparation

(Rogers

&

Monsell,

1995).

The

present study investigated _the temporal

pro-filesof attention

during

a task.

By

focusing

on the

preparation cost the state of attention during this

briefperiod of the trialcould be understood. But

focusing only on

the

vigilance

decrement

would not

represent _the overall state of attention inthe

experi-ment.

In

our study, we used a rapid serial visual

presentation

(RSVP)

task inwhich stimuli were

pre-sented serially and rapidly at the same spatial

loca-tion on a

display.

The

observers were required to

detect

a target defined by a certain feature. By

manipulating when _thetargetappenred

in

thc RSVP stream we codld observe the bricftransitionsof

at-tention

during

the task _performance. We _predicted

thata vigilance decrement wou]d occur throughout the experiment: theeMciency of an observer would

decline

as the

trials

continued,

In

particular during a

trial,because of the preparation cost, ebservers

would often miss a

target

inserted

into

the

beginning

(or

preparation period)of an RSVP stream,

*Department _{of Psychology, The University of}

To-kyo,

7-3-]

Hongo,

Bunkyo-ku,

Tokyo

113-O033

Method

Twelve naive students participated

in

the

experi-ment,

After

thepresentation of a

fixation

_point

{500

ms), an

RSVP

stream of upper-case lettersbegan

(SOA=100

rns, ISI=80ms). In_the

RSVP

stream 20

different

letters

were sequentially presented fora

totalof 2500ms. The task oi each ob$erver was to

report a white _target

Ietter

which was presented

among tight-blue

distracter

lettersby _pressing the

corresponding

key

on a keyboard after the finishof the

RSVP

stream. The _positionof

the

target

in

the

stream was vamed between the second and

nine-teenth frames. A response was recorded

if

it

oc-curred within the response _perjod of 2500 ms after

t,hetermination of the sequence. Following the

re-sponse period

the

fixation

point reappeared and the

next triat

began,

regardless of whether an observer

made a response or did not, Each trial

therefore

lastedfor5000 ms. The experiment consisted of 432

The Japanese Psychonomic Society

NII-Electronic Library Service

The JapanesePsychonomic Society

118

TheJapanese

Journal

of

Psychonomic

Science Vo].24,

No.

1

trials

(4

sessions of

108

trials,

for

a totalof 9

min-utes). The firstand second sessions of thetrialswere

conducted sequentially as one

block,

and

the

third

and fourth trialsessions were cenducted as another.

The

duration

of each

block

was therefore18minutes.

Because

the observers were allowed to

have

a

5-minute break

between

each

block

the entire experi-ment took

41

minutes,

There

were 2 experimental factors

<Time

Frame

and Target Part).The Time Frame had

2

conditions

(the

fir$t-half

and

last-half

conditions),

in

which the

firstand thirdsessions

(the

first

9minutes of a block) were the

first-half

condition, and

the

second and

fourth

sessions

(the

last9 minutes of a

block)

were

thelast-halfcondition. The

Target

Part had 6

_parts

(the

first

tosixth parts), which were based on the primary target positions. We integrated every

3

prirnary

target

positions

in

a sequentially order as

one _part.

That

is,

we treatedthe primary second to

fourth

target

_position

as the first_part,the fifthto scvcnth as the second _part,and so on

(a

totalof

6

partswerc seO,

Results

and

Discussion

ThD mean _percentages of correct detection of the

target,foreach condition of

Target

Part and Time

Frame,

are shown inTable 1. A two-way ANOVA

with theTime Frame _Cthe

first-half

and

last-half

con-ditions)and

Target

Part

Cthe

first

to sixth _parts)

revealed significant main effects of the Time Frame

[F(1,

11)=6,55,

_P<O.05]

and of the Target Part

_[F(5,

55)-32.59, p<O.OOII. A significant interactic)n

be-tween thefactorswas not revea)ed

[F<5,

₅₅₎₌

1.13,

n.

s.1.

A

Ryan's

tnultiplc eomparison analysis, which

collapsed the Time Frame facter,revealed that the correct detectionrates inthe firstand second partsof

the target _positions were significantly lower than those

in

any other _part

[ts(55}

>3.33,

P<O.05:.

The

present study demonstrated the changes of time-dependent eMciency inthe experimenta] tasks.

Bccause

thecorrect

detection

rates were lower

in

the

last-halicondition than inthefirst-halfcondition, a

Table 1

The

mean correct

detection

rate

(%)

Time frame

Target

part

lst 2nd 3rd 4th 5th

6th

Ftrst-half Last-ha]f

535066587270787281787474

vigilance

decrement

occurred throughout the experi-ment. Inthe prcsent experiment, the vigilance

decre-mcnt occurred within 9minutes, which isalmost

consistent with other vigilance studies,

On

theother

hand,

we

found

a preparation co$t

in

a single trial:in both the first-and last-half

condi-tions

the

correct

detection

of a

target

in

the earlier

positions was dramatically low, and gradually

in-creased as thetargetappeared later.This result

has

not been reported in _previous vigilance studies,

which have focusecl on the changes

during

pro-longed _performance

in

an experiment. This resu]t

is

also a new

finding

forRSVP tasksbecause the earlier

RSVP studies have _predicted a fairand

high

detec-tion

of only one target,whenever

it

appears. We

assumed that thcreduction indetection of a target thatwas early inasequence wou]d reflect a gradual

modulation of temporal attention to a rapid

se-quence. At the

beginning

of an RSVP stream thc

temporal attention of an observer may have

diMcul-tyin adjusting to a rapid scquence and

detecting

a

target.Then, the observer might

begin

tomodu]ate

their

temporal attention

in

order to accurately

ex-tracta targetfrom the tempora]ly congested stream.

The _preparation cost obscrved in

the

_present study

would refiect thismodu]ating process of attention to

the t.ask.

References

Davies,D. R,

&

Parasuraman,

R,

I982 The

ogy of vigilance.

London:

Academic Press.

Rogers,

R,

D.

Monsell,

S.

1995 The costs of a

dictable

switch between simple cognitive tasks.

,loorrnal

of

ExPerimental Ilsychotog),:

Geneizzt,

124,