Content uploaded by Laura A Brannon
Author content
All content in this area was uploaded by Laura A Brannon on Jul 01, 2014
Content may be subject to copyright.
Psychonomic Bulletin & Review
/999,6
a:
/48-/56
Timeless demonstrations
of
Parkinson's first law
LAURA A. BRANNON
University
of
Oklahoma, Norman, Oklahoma
PAULJ.HERSHBERGER
Veterans
Affairs
Medical Center, Dayton, Ohio
and
TIMOTHY C. BROCK
Ohio State University, Columbus, Ohio
"Workexpands so as to fill the time available for its completion," Parkinson's law, is an explanation
classic that has survived without an artifact-free demonstration at the individual level. To evaluate
Parkinson's law, undergraduate subjects expected to judge four sets of photos of faces with reference
to a subjective criterion. The experimental subjects, who were told that the fourth set was canceled be-
fore they began work on the third set, dallied on the third set; that is, as compared with controls, they
prolonged
work
The cancellation---dalliance effect was reobtained in two exact replications. It was ob-
tained again in a fourth study, a conceptual replication wherein subjects processed negatively toned
phrases against an objective criterion. The generalizability of the effect and explanations for it are dis-
cussed,
The temporal elasticity
of
a set of activities, the degree to
which the activities take more or less time as a function
of
the circumstances under which they are carried out, is the
central idea
of
Parkinson's notorious first law
....
It seems
to us an important idea, one that deserves more study than
it has thus far received. (McGrath
& Kelly, 1986, p. 124)
"Work expands so as to fill the time available for its
completion" became an explanation classic soon after its
promulgation as a "law" by Parkinson (1957). Fourdecades
later, the familiar aphorism continues to be evoked to ac-
count for a plethora
of
human and organizational ineffi-
ciencies. However, although Parkinson's formulation has
become a
household
expression, its popularity never
elicited a convincing empirical demonstration.
Parkinson's First Law: Whence and Whither
When Parkinson's (1957) mathematical formula for
his "law" is examined, its satirical character is apparent;
although tongue-in-cheek, it expressed sincere revulsion
at metastasizing bureaucracies.' Parkinson later exam-
ined additional data and concluded that he should be re-
garded "as a true prophet" (Parkinson, 1980, p. 23) be-
cause "work does expand so as to fill the time available"
(ibid. p, 35, Parkinson's italics). In 1980, as in his initial
1957 statement, the evidence adduced was increases in
aggregate numbers
of
staffmembers rather than a direct
The
comments
of
Joe
McGrath, Janice Kelley,
and
Craig
1. Russell
on earlier drafts are
acknowledged
with gratitude.
Our
indebtedness to
several
anonymous
reviewers for
their
comments
on an earlier draft is
also acknowledged.
Correspondence
concerning
this article should be
addressed to L. A. Brannon,
Department
of
Psychology, University Ok-
lahoma,
Norman,
OK
73019
(e-mail:
lbrannon@ou.edu).
study
of
the persons who were the presumed perpetrators
of
work expansion.
Mixed Endorsement by Scholars
Parkinson's law is widely taught in a variety
of
disci-
plines that deal with behavior in work settings. While the
quality
of
the evidence for the "law" appeared to be be-
yond question in most sociology textbooks (Giddens,
1989, p. 277; Kornblum, 1988, pp. 362-363; Macionis,
1991, pp. 187-188; Thio, 1991, pp. 87-88), in other text-
books (e.g., Conklin, 1984, p. 163) there is acknowledg-
ment
of
failures to confirm (e.g., Reimann, 1979). Sim-
ilarly, in many organizational behavior and management
texts, the validity
ofthe
"law" is accepted, and it plays an
important explanatory role (Carvell, 1980, p. 419; Hersey
& Blanchard, 1988,p.128; Koontz & Weirich, 1990, p. 12;
Luthans, 1981, pp.
446-447;
Miner, 1980, pp. 170-177;
and Robbins, 1990, pp. 162-167). However, neither ac-
cepting nor doubting scholars have been able to rely on
experimental tests
of
individuals, susceptibility to excess
time effects.
Aronson's Experimental Explorations
"Beyond Parkinson's Law"
In quest
of
compelling evidence for Parkinson's law,
the early results
of
Aronson and Gerard (1966), Aronson
and
Landy
(1967), and Landy, McCue, and Aronson
(1969) were examined. Unfortunately, the Aronson stud-
ies were inconclusive because, rather than demonstrating
the excess time effect, it was simply taken for granted
and, also, because time was artifactually emphasized for
the research subjects. In both the 1966 and 1967 experi-
ments, Aronson allowed research subjects too much time
Copyright 1999 Psychonomic Society, Inc.
148
to perform a task. Subsequently, when presented with a
similar task and allowed to work at their own pace, sub-
jects who initially were allowed excess time required
more time to complete the subsequent task. The subject
was seated at a table with a timer and asked to prepare a
2-min talk (with materials supplied by the experimenter);
then an interruption occurred, the critical manipulation
of
excess time. A secretary "barged in," asked the ex-
perimenter to help another experimenter, and announced
that it would take 15 min (or 5 min). The experimenter
"gave in resignedly, 'Only 15 (5) minutes, eh?'" When
the experimenter returned, the subject was asked to work
on preparing another speech (Aronson
& Gerard, 1966),
and time on this second task was the principal dependent
variable.
The data did show that subjects who were allowed
more time for the first task took more time on the second
task. However, the phenomenon
of
excess time on the
first
task was not examined and the procedure was re-
plete with time cues: the subjects faced the timer on the
table, and the interruption
of
the experimenter was at-
tended by emphatic time
stipulations-for
example, "only
fifteen minutes; only five minutes."
The 1967 (Aronson
& Landy) replication employed a
different task and included a further manipulation, with
the second task being similar to or different from the first
task. However, when the initial task was explained to the
subjects, the interruption manipulation had the same
prominent time-cuing features. Subjects who were al-
lowed 5 min (and knew they were allowed just 5 min) uti-
lized an average
of288
sec, while subjects who were al-
lowed 15 min (and knew they had 15 min) spent an
average
of640
sec [F = 66.94,p < .001; effect size was
calculated as
d = 2.02; Johnson, 1993].
Unfortunately, because of the aforementioned perva-
sive time cuing, this possible demonstration of the Parkin-
son effect could not be convincing. Importantly, Aronson
and his co-workers acknowledged this ambiguity: "dif-
ferential time norms may have been conveyed by E as to
what constituted an appropriate time to spend on a given
task" (Landy et al., 1969, pp. 236-237). In their final ex-
periment, Aronson and his coexperimenters nevertheless
continued the time cuing and even increased it. An in-
terruption with time announcements was used again.
Furthermore, the subject was confronted by a confeder-
ate who sat very close to the subject (allegedly because
all the other cubicles were in use) and repeatedly asked
the subject to hurry.
Of
course, subjects told to hurry did
take less time than subjects in a control condition in
which they were not badgered to hurry. In sum, the Aron-
son studies may have gone "beyond" Parkinson's law but
they did not demonstrate the law itself.2
In this first stringent test of Parkinson's prediction, we
sought to eliminate time demands and time cues while
probing for dawdling on their current work by subjects
who had less future work to do than they expected. Our
first three studies used a photo-ranking procedure, sim-
PARKINSON'S LAW 149
ilar to that used by Aronson and Landy (1967; Landy
et al., 1969); it required subjects to look at photographs
and rank-order the people depicted in terms of various
criteria, such as
subjective impression
of
intelligence.
Use
of
a similar task was considered critical to the cu-
mulativeness of the present investigations with Aron-
son's earlier findings. Our final experiment, Study 4,
was a conceptual replication in which subjects processed
negatively toned phrases against an
objective criterion.
THREE
STUDIES
OF
TIME
SPENT
JUDGING
SETS
OF
PHOTOS
Method
Overview
Subjects individually followed a procedure in which they ex-
pected four trials but actually completed only three. A trial con-
sisted
of
judging photos
of
faces with reference to a criterion. The
experimental subjects were told about the "cancellation"
of
the
fourth trial just before they began work on the third, the test, trial;
they thus had "extra" time. In contrast, the control subjects expected
a fourth trial as they worked on the third trial, the test trial. The pri-
mary dependent measure was the change in time taken to complete
the third (test) trial as compared with the time taken to complete
the second (baseline) trial. Oral instructions were used in Study I,
whereas written instructions were used in Studies 2 and 3.
Subjects
The subjects were undergraduate students, male and female,
from introductory psychology classes at a Midwestern university,
who participated in the study in partial fulfillment of a course re-
quirement:
N = 22 (Study I), N = 47 (Study 2), and N = 109
(Study 3). Importantly, all subjects from this pool expected to spend
up to an hour in each required research participation.
Materials
The stimulus materials consisted of four sets
of
10 pictures, 20
of
males and 20
of
females. The portrait pictures were taken from
newspaper wedding pictures so that all persons would be dressed in
formal attire. This source for the pictures also meant that educa-
tional and occupational information about the persons depicted was
plausibly obtainable from the accompanying announcements. The
pictures were trimmed so that only faces
of
the brides and
of
the
grooms were visible.
Procedure
Individually tested subjects sat facing a wall, so their field
of
vi-
sion excluded both the experimenter and the timekeeper. The time-
keeper had no interaction
of
any kind at any time with the subject.
The subjects were given instructions which indicated that the ex-
periment was attempting to determine whether or not it was possi-
ble to accurately estimate the educational and occupational levels
of
persons by looking at pictures
of
the persons' faces. The instruc-
tions further indicated that the experiment would consist
of
two
equal parts and that each part would involve ranking two sets
of
pic-
tures. The text of the instructions was as follows:
This experiment is part
of
an exploratory study in which we are trying
to determine whether it is possible to guess the education and occupa-
tional level
of
persons simply from looking at their pictures. This is a
skill that could be very important in police work. for persons in service
occupations. and for fundraisers. Our procedure is quite straightfor-
ward. The experiment has two parts. In Part I
of
the experiment. you
will rank pictures
of
10 men according to your best guess
of
their edu-
cation and occupational level and then 10 women based on the same cri-
150 BRANNON, HERSHBERGER, AND BROCK
teria. Part II
of
the experiment is identical to Part I, except that you will
be using different pictures.
Rank the pictures from lowest education/occupational level to the
highest. For example, someone whom you believe has a high school
diploma but no further formal education would be ranked before some-
one whom you believe has completed graduate school. In the same way,
an assembly line worker would be ranked before a high level corporate
executive.
At this point the timekeeper began surreptitious timing. The tim-
ing was stopped when the subject indicated that he or she was fin-
ished with the set
of
pictures. During this practice/warm-up trial,
questions were allowed from the subjects and were answered by the
experimenter. The practice trial enabled standard administration
of
the next trial, the baseline trial.
Trial
cancellation
manipulation.
Control-condition subjects
completed the two picture sets (the practice/warm-up and the base-
line trials) in Part I, and were then told that Part I was now com-
pleted and that the procedure in Part II would be exactly the same.
The complete text
of
the instructions was: "You have now com-
pleted Part I
of
the experiment. The identical procedure will be fol-
lowed in Part II
of
the experiment."
After completing the first set
of
pictures in Part II (the test trial),
the control-condition subjects were told that there would not be an-
other set to rank (no fourth trial). Thus, the control-condition sub-
jects completed all three trials with no basis for suspecting that the
fourth trial would not occur. An additional control group
(N
= 14)
in Study 3 was told at the outset that there would be only three trials.
Experimental-condition subjects also completed the two picture
sets in Part I (the practice/warm-up and the baseline trials), but
these subjects were then told that one
of
the sets
of
pictures for
Part II of the experiment had been canceled. Therefore, they would
have only one picture set to rank in Part II. The complete text
of
the
instructions was:
"I'm
unable to find the pictures
of
the women for
Part II of the experiment. Instead
of
ranking both men and women
in Part II
of
the experiment, you'll only rank 10 men. Here are the
10 pictures, and again, let me know when you are finished."
Finally, the subjects were fully debriefed: the actual purpose was
disclosed, Parkinson's law was defined and illustrated, and the can-
cellation manipulation was explained.
Dependent
measures.
The timekeeper surreptitiously scored,
by stopwatch, the beginning and end
of
the subject's use
of
the pho-
tos. The principal dependent variable was the change in the amount
of time spent on the test trial (the picture ranking in Part II) as com-
pared with the baseline trial (the second picture ranking in Part I).
Probing
task
perceptions
in
Studies
2
and
3. After ranking the
one set
of
pictures in Part II (the test trial), the subjects were also
asked to answer five posttask questions:
I. Did you find the sets
of
pictures to be
of
equal interest" lf no, which
more interesting"
2. Did you find the sets
of
pictures to be
of
equal difficulty"
(fno,
which
more difficult"
3. Estimate the actual time you spent on each
of
the sets, in minutes and
seconds
4. Did you feel you were being evaluated in any way beyond what was
explained at the beginning
of
the experiment"
Experimental subjects were further asked about the thoughts that
went through their minds when they were informed that the exper-
imenter had misplaced one of the sets
of
pictures for Part II; they
were also asked, "How did this affect the amount
of
time you spent
on the third set')"
Results
All subjects readily completed their photo-rankings
according to the instructions. No subject's data have
been omitted for any reason. The principal results for the
three studies are presented in the top, middle, and bottom
panels
of
Figure 1. The left side
of
each panel shows
mean speed in seconds for the baseline and test trials for
the control condition (no cancellation,
N =
11
in Study 1;
N = 22 in Study 2; and Ns = 46 and 14 in Study 3); the
right side
of
each panel shows the same data for the ex-
perimental conditions (fourth trial canceled,
N = 11 in
Study 1;
N = 25 in Study 2; N = 49 in Study 3). Gender
of
subject did not enter into any reliable main or inter-
action effects for the three studies. Gender is therefore
not further discussed.' The principal dependent variable
was the change in the amount
of
time spent on the test
trial (the picture ranking in Part
11)
compared with the
baseline trial (the second picture ranking in Part I). The
propensity
of
a subject to be quick or slow on the test trial
was examined by a 2 (condition: experimental vs. con-
trol)
X 2 (trial: baseline vs. test) analysis of variance on
the number
of
seconds required to complete the task,
with repeated measures on the second factor. There was
a significant condition
X trial interaction
[F(l,20)
=
5.30,p
< .04, in Study 1
andF(l,45)
=
5.87,p
< .02, in
Study 2]. Study 3 included a second control condition
(expect three trials). A similar 3 (condition: experimen-
tal, control 1, control 2)
X 2 (trial: baseline vs. test) analy-
sis with repeated measures on trial yielded a borderline
significant condition
X trial interaction
[F(2,
106) =
2.98, p = .055].
As shown in Figure 1, the control subjects worked
through the sets of pictures more quickly on the test trial
than on the baseline trial. The times were also faster for
the baseline trial than for the practice/warm-up trial (not
shown). In contrast, while the experimental subjects also
worked faster on the baseline trial than on the practice
trial, these subjects tended to work more slowly on the
test trial. In comparison with the control subjects, the ex-
perimental subjects dallied! over the first set
of
pictures
in Part II: work appeared to expand to fill the time avail-
able for its completion, as Parkinson proposed.
The parallel pattern for the second Study 3 control,
expect three trials (bottom panel
of
Figure 1), showed
that the cancellation-dalliance effect did not depend on
comparison with another group (the primary control)
that was also expecting a fourth trial. The comparison
between the expect-three-trials group (control 2) and the
experimental group also appeared to rule out the hy-
pothesis that the latter worked relatively more slowly
than the control group in order to prolong their enjoy-
ment
of
the task. If that had been the case, the expect-
three-trials subjects might also have worked more slowly
to "savor" the final set
of
photos. An additional manip-
ulation
of
expected trials is reported in Study 4 below.
Effect sizes,
ds (Johnson, 1993) for Studies I, 2, and
3 were .94, .70, and .46, respectively. These medium-to-
high effect sizes were obtained with no explicit time
cuing. Although robust, they were lower than the 2.02
calculated for the Aronson and Landy (1967) time-cued
demonstration
of
the effect. However, our data and the
PARKINSON'S
LAW 151
OJ
Baseline Trial
• Test Trial
Experimental
Group
CJ
Baseline Trial
• Test Trial
Experimental
Group
Experimertal
Group
Expect Three
Trials Group
Control
Group
Control
Group
Control
Group
170
~
160
!
i
150
E
0
140
0
s
-!
130
c
0
J 120
c
:I
110
100
170
160
~
•
l-
I
150
Go
E
0 140
0
.s
-! 130
c
0
~
II)
120
c
i
110
100
170
160
~
1lI
I-
1
150
I
Go I
E
140 "
0
0
I
s
III
I
~
130 1
0
!
120
c
•
II
~
110·
100
Figure
I.
Effects
of
trial
cancellation
on
work
speed
in
seconds
in
Study
1,
top
panel;
Study
2,
middle
panel;
and
Study
3,
bottom
panel.
152 BRANNON, HERSHBERGER, AND BROCK
Aronson data, taken together, suggest that the Parkinson
effect is not only quite real but that it may be robust in
real-life work settings because such settings do com-
monly include time cues and time-related accessories.
Task Perceptions (Studies 2 and 3)
In addition to using written instructions, Studies 2 and
3 probed subjects' perceptions with virtually the same
results for both studies. Subjects' estimates
of
the time
spent on each set
of
photos were entered into an index,
namely, the difference between actual time and estimated
time divided by the actual time taken. Whether verbal es-
timates were affected by condition was determined by 2
(condition: experimental vs. control) X 2 (trial: baseline
vs. test) analyses
of
variance, with repeated measures on
trial. The interaction Fs for these
analyses
for
both
Study 2 and Study 3 were below unity. When alternative
indices were calculated, such as the simple difference
between actual and estimated time and the ratio
of
each
person's verbal estimate to his or her actual time, main
and interaction effects for condition
and
trial again
yielded Fs < unity. Hence, there was no evidence that
cancellation
of
the fourth trial caused experimental sub-
jects to somehow arrive at more accurate recollection
of
their work durations on the test trial.
To further test subjects' sensitivity to dallying, t tests
were conducted comparing the verbal estimates
of
third-
trial duration made by subjects in the experimental and
control conditions. In both Study 2 and Study 3, these
differences yielded ps > .20. An apparent tendency for
experimental subjects to report longer durations fell well
short
of
conventional criteria for reliability.
There were no differences between experimental and
control subjects in their perceptions
of
the interest and
difficulty
of
the sets
of
pictures. Fewer than 5%
of
the
subjects commented about being evaluated, and the only
comment made by more than I subject was to surmise
that the experiment dealt with comparing how males and
females made such judgments.
Finally, subjects in the experimental conditions were
reminded that the experimenter had misplaced pictures
for Part II and were asked,
"How
did this affect the
amount
of
time you spent on the third set?" Most sub-
jects had no comment. The most important observation
of
these comments was that none
of
the subjects men-
tioned "tried harder," "tried to be more accurate," "spent
more time on the pictures," or the like.
STUDY
4
Conceptual Replication With Negative Stimuli
and Objective Measurement
of
Performance
A fourth experiment addressed a number
of
questions
that were not resolved by the first three studies. Was the
Parkinson effect limited to seemingly neutral stimuli?
Would people prolong work that was inherently unpleas-
ant?
Could
the effect be
observed
for tasks in which
work quality had an objective criterion? Was the effect
proportional to the amount
of
further work that had been
canceled? The first three studies deliberately used a task,
ranking photos, that was quite similar to that used origi-
nally by Aronson (e.g., Aronson & Landy, 1967). How-
ever, that task did not allow for observation
of
dawdling
in an activity that was somewhat unpleasant, and it did
not allow for objective measurement
of
performance or
for assessment
of
the possible impact
of
the magnitude
of
the canceled activity on amount
of
dalliance.
If
dal-
liance was related to the magnitude
of
the cancellation,
the condition X trial interactions
of
Studies I, 2, and 3
might be qualified by a higher order interaction with
number
of
canceled trials. Study 4 therefore utilized an
objectively scorable and somewhat unpleasant task, and
amount
of
cancellation was manipulated. Finally, al-
though
estimated
durations were not affected by the
treatments in Studies I, 2, and 3, these estimates might
have been insensitive measures
of
actual motivation to
try harder. Consequently, we included new measures
of
subjects'
impressions
of
the
task
and
of
their effort
(Bryan & Locke, 1967).
Method
Overview
As in Studies
1,2,
and 3, subjects individually followed a proce-
dure in which they expected 4 (or 12) trials but actually completed
only 3. A trial consisted of counting the number of letters in a set
of25
phrases (see materials below). Cancellation instructions were
read to the subjects by the experimenter, but other instructions were
printed.
Subjects
The subjects were 86 undergraduate students, male and female,
from introductory psychology classes at a Midwestern university,
who participated in the study in partial fulfillment
of
a course re-
quirement. As in Studies 1,2, and 3, all subjects from this pool ex-
pected to spend up to an hour in each required research participa-
tion.
Materials
The stimulus materials consisted
of
three sets
of25
negative two-
word
phrases-one
set for each trial. For example, in the test trial,
some
of
the phrases were "suffocating air," "poisoned water,"
"gloomy
day,"
" menacing
dogs,"
"parking
penalty,"
"vicious gos-
sip," "criminal threat," "foul odors," and so forth. The word pairs
were printed with variable spaces between them above a scale with
the numbers from 8 to 28. For example, for "DISTRESSED CHIL-
DREN" the correct letter-counting score was 18.
Procedure
Individually tested subjects, who were unaware that they were
being timed, sat facing a wall in a cubicle so that their field
of
vi-
sion excluded both the experimenter and the timekeeper. The sur-
reptitious timekeeper had no interaction
of
any kind at any time
with the subject.
The subjects were given instructions which indicated that the ex-
periment was attempting to determine how well reading and count-
ing could be done when they were done simultaneously. The text of
the instructions was as follows with the expect-12-trials wording in
brackets:
Reading and Enumeration Study:
The experiment today deals with two human processes, reading text and
counting objects, and how well these basic processes can be done when
they are done together. These basic processes, reading and counting,
are essential to many kinds of work and often occur together in every-
day life. The experiment has two parts. In the first section
of
Part I of
the experiment, you will read and count the letters in a set
of25
phrases
as well as you can. In the second section
of
Part I you will again read
and count the letters in another set
of25
phrases. The two [ten] sections
of Part II
of
the experiment are identical to Part I, except that you will
be using different sets of phrases. Therefore, after Part I, there will be
two [ten] more sets of phrases with 25 phrases in each set. In this study,
read a PHRASE and then circle the number that represents the number
of letters in the entire PHRASE.
A worked example followed.
The counting task was rendered nontrivial by varying the num-
ber of spaces between the words in the pair: it was not possible for
a subject to correctly count by simply learning the counts that cor-
responded to the positions of the last letter in the second word.
At this point, the subject began work, and the work was timed.
The timing was stopped when the subject indicated that he or she
was finished with the first set (practice/warm-up)
of
phrases (word
pairs). During the warm-up set, questions from subjects were al-
lowed and answered by the experimenter. Then the subject was
given the second set (baseline)
of
phrases to score and timing
resumed.
TrialcanceUation manipulation
and
debriefing. With respect
to the third (test) set, the cancellation procedures for the succeed-
ing set(s) and the debriefing for control and experimental subjects
were similar to those
of
Studies I, 2, and 3, except that subjects
were told that the experimenter had just learned that the remaining
sets had been canceled (rather than misplaced) and the number
of
remaining sets was 10 (rather than I) in the expect-12-trials condi-
tion. The subjects' total work time on the baseline set
of
phrases
averaged 199 sec
(SD = 30 sec). Therefore, even subjects in the
expect-12-trials condition could complete the task in approximately
40 min (much less than the 60 min allowed for a unit
of
research
participation). No subject would have felt "hurried" (by the nature
of the task or the number
of
anticipated trials) to finish.
Objective measures of performance
and
concentration/ef-
fort:
Letter
counting
and
surprise
recall. As an objective
measure
of
performance, we scored correct letter counting. In ad-
dition, in order to assess subject concentration/effort, we measured
subjects' ability to recall one
of
the words from the test-trial word
pairs when either the first or the second word was presented. Pre-
sumably, greater concentration/effort during the letter-counting
task would result in greater recall of the word pairs on a surprise re-
call test.
Task perceptions. Study 4 subjects used a 5-point Likert scale
to rate how pleasant and how interesting it was to read the phrases
in each set. A self-assessment
of
work speed was adapted verbatim
from Bryan and Locke (1967, p. 266):
During the third set, the last set, I was trying to:
Work faster than on the first two sets
__
Work at about the same pace as on the first two sets
Work a little slower than on the first two sets
__
Work as fast as possible
_~Work
quickly, but not as fast as possible
__
Work with as little effort as possible
Other.
Finally,we included Likert items to measure the extent to which the
subjects were trying "harder to circle the correct number ofletters
in the last set
of
words" and the extent to which they were "hurry-
ing to complete the experiment."
PARKINSON'S LAW 153
Results
All subjects were given as much time as needed to
count characters in the three sets
of
word pairs; thus, the
subjects achieved perfect counting scores on the third set.
No subject data were deleted for any reason. Sex of a sub-
ject did not enter into any reliable main or interaction ef-
fects in Study 4 and gender is not further discussed.
Subjects found reading the phrases to be moderately
unpleasant. The Likert-scale points were labeled 1
=
very unpleasant, 2 = unpleasant, 3 = neither pleasant
nor unpleasant, 4
= pleasant, and 5 = very pleasant.
Overall means
(N = 86) for the three word sets were 2.7,
2.5, and 2.4, respectively. The pleasantness ratings were
unaffected by condition (experimental vs. control) or ex-
pected trials (4 vs. 12) or the interaction
of
these factors.
Similar Likert ratings
of
"interesting" were slightly
below the midpoint, 2.9, for the three sets
of
phrases and
were unaffected by the independent variables. Therefore,
whatever dalliance was observed was not attributable to
different levels
of
task enjoyment or
of
task interest.
The principal dependent variable was the change in
the amount
of
time spent on the test trial (counting the
number
of
characters in the third set
of
phrases) com-
pared with the baseline trial (counting the second set
of
phrases). The graphs
of
Figure 2 show the mean speeds
in seconds for the baseline and test trials for control
(N
=
21) and experimental
(N
= 24) groups under expect-4
trials (leftmost) and for the control
(N = 18) and exper-
imental
(N = 23) groups under expect-12 trials (right-
most). The propensity
of
a subject to be quick or slow on
the test trial was examined by a 2 (condition: experi-
mental vs. control)
X 2 (expected trials: expect 4 vs. ex-
pect 12)
X 2 (trial: baseline vs. test) analysis
of
variance
on the number
of
seconds required to complete the task,
with repeated measures on the third factor. There was a
significant condition
X trial interaction
[F(l,82)
=
75.69, P < .0001]; this interaction reproduced the same
pattern that was observed in Studies
1,2,
and 3. The ef-
fect size (Johnson, 1993) was d
= 1.87, a high outcome
comparable with the d
= 2.02 calculated (above) for
Aronson's data (Aronson
& Landy, 1967).
In addition, there was a condition
X expected trials X
trial interaction [F(I,82) = 10.02, p < .002, d = .68].
Figure 2 shows that the higher order interaction was at-
tributable to higher dalliance in response to greater can-
cellation. The experimental subjects' mean work dura-
tion on the test trial under expect-12 trials, 233, was
reliably higher than their mean duration under expect-4
trials, 211
[t(45) = 2.2, p < .03, d = .63]. Thus, dal-
liance here appeared to be proportional to the magnitude
of
the canceled work.
Objective
Measures
of
Performance
and
ConcentrationlEffort:
Letter
Counting
and
Surprise
Recall
As noted previously, given the relative simplicity
of
the letter-counting task, all subjects, regardless
of
con-
dition (experimental vs. control), correctly counted the
154 BRANNON, HERSHBERGER, AND BROCK
250
.a::
240
II)
~
230
S 220
.!! 210
E200
e 190
o 180
.s
170
~
160
~
150
~
140
tn 130
c
Z 120
:::E
110
100
Illll
BaselineTrial
• Test Trial
Expect 12 Trials
Control Experimental
Group Group
Expect 4 Trials
Control
Group
Experimental
Group
Figure
2. Effects of
trial
cancellation on
work
speed in seconds in Study 4 (expect 4 trials, left panels;
expect 12 trials,
right
panels).
number
of
letters in each set. Dalliance in the experi-
mental condition was not associated with better perfor-
mance on the counting task.
Recall
of
word pairs was difficult for the subjects, in
part because they did not expect a recall test. When the
third set of words was readministered with one word in
each set missing, overall correct mean recall
of
the miss-
ing words was 5.1 (obtained range
= 0 to 14; perfect re-
call
= 25). Between-subjects analyses showed that recall
performance was not affected by condition (experimen-
tal vs. control) or by expected trials (4 vs. 12) or by their
interaction. Moreover, recall and third-trial performance
time were only slightly correlated
(r
= .13, n.s.), and the
within-condition correlations in the experimental and
control groups did not differ from each other. The ob-
tained range
of
scores (0 to 14) indicated that the ab-
sence of effects
of
the independent variables was not at-
tributable to a restriction
of
range. Dalliance was not
associated with greater concentration/effort.
Task
Perceptions
To measure subjects' assessment
of
their own motiva-
tion, we adapted a question from Bryan and Locke
(1967, p. 266). With respect to the test trial, 69% of the
control subjects and 72%
of
the experimental subjects
used the response alternatives "work faster than on the
first two sets" and "work at about the same pace as on the
first two sets." Chi-square analyses for these frequencies
and frequencies for all response alternatives (see Proce-
dure above) yielded no reliable differences between the
control and experimental conditions. With respect to the
test trial question, only 2 experimental subjects and 2
control subjects said "worked a little slower than on the
first two sets." To the extent that this question was sen-
sitive to subjects' motivation, the results indicated that
experimental subjects were not aware
of
their actual
slowing down relative to the control subjects.
We also asked subjects to what extent they were "try-
ing harder to circle the correct number of letters on the
last set
of
words, the third set, than on the previous set
of
words." The 5-point Likert item was anchored with 1 =
"I tried much harder on the third set than the second set"
and 5
= "I tried much harder on the second set than on
the third set." The means for the control group were,
under expect-4 trials,
M = 2.6 and, under expect-12 tri-
als,
M = 3.0; the corresponding means for the experi-
mental group were 2.5 and 3.0. The only reliable effect
was that subjects under expect-4 trials reported trying
harder on the last set
[F(I,82)
=
5.54,p
< .02]. Condi-
tion itself (experimental vs. control) did not affect self-
report
of
trying harder on the third than on the second
set. Once again, cancellation, the Parkinson treatment,
did not influence reported motivation. The final Likert
item, "To what extent were you hurrying to complete the
experiment," had five anchors: "A
lot,"
"Somewhat," "A
little," "Slightly," and "Not at all." The overall mean fell
between "somewhat" and "a little," but there were no ef-
fects due to treatments. In sum, there was no tendency
for experimental subjects, as compared with controls, to
report trying harder on the third set or to report hurrying
less to complete the experiment.
GENERAL
DISCUSSION
The present studies have consistently demonstrated
Parkinson's "law" under conditions free of the explicit
time-cuing and time demands that confounded previous
research by Aronson and his co-workers (Aronson &
Gerard, 1966; Aronson & Landy, 1967; Landy et aI.,
1969) and by Bryan and Locke (1966). While the Aron-
son studies appeared to demonstrate that taking excess
time instigated further taking
of
excess time, the basic
Parkinson phenomenon was never unambiguously demon-
strated. In contrast to subjects who were externally paced
by investigator-provided cues, the present subjects paced
themselves: in this context, an unexpected cancellation
instigated dalliance. Furthermore, the three-way inter-
action
of
Study 4, condition X expected trials X trial,
suggested that magnitude
of
dalliance might be affected
by the magnitude
of
the canceled work. In other words,
the more unexpected time people have, the more they
dally.
The present experiments comprised two simple replic-
able Parkinson-effect paradigms, one with seemingly
neutral stimuli and a subjective task and the other with
moderately unpleasant stimuli and an objective task. In
both paradigms, next-task cancellation increased work
time on the focal task even though all subjects knew that
research participation required up to an hour. Expansion
of
time by the experimental subjects in Study 4 did not
result in better work (accuracy in counting characters in
word pairs was unaffected) or more concentration/effort
(there was no better recall
of
the word pairs on a surprise
recall test).
To make our findings cumulative with Aronson's (e.g.,
Aronson & Landy, 1967), we used a similar task in Stud-
ies I, 2, and 3 in which performance could not be evalu-
ated objectively. But, even
if
subjects who dallied on the
task did not actually do a better job, is it possible that
doing better work became their purpose? In Study 4, dal-
liance did not improve task performance, but was better
performance an aim nonetheless?
The postexperimental questionnaires had several items
which might have revealed that experimental subjects, in
contrast to the controls, were trying harder, aiming more
for accuracy, and so forth. As shown in Studies 1, 2, and
3, on none
of
these scaled and open-ended questions did
the experimental subjects differ from the controls. Fur-
thermore, since there was ample time left over, it is doubt-
ful that control subjects were hurrying beyond what was
reasonable in the situation in order to
"fit
in" extra trials.
In Study 3, control subjects accelerated their work from
baseline to test trial in parallel with a second control group,
which had expected three trials from the outset (bottom
panel
of
Figure 1).
New probes in Study 4 disclosed no differential effort
or perception
of
effort by subjects who experienced next
trial cancellation as compared with those who completed
Trial 3 without knowing it was the last trial. Experimen-
tal and control subjects appeared to have the same task
goals. Thus, the longer durations on the test trial for the
experimental subjects cannot be readily attributed to ei-
PARKINSON'S LAW 155
ther conscious "hurrying" by control subjects or to con-
scious "lingering" or extra effort by experimental sub-
jects. The Parkinson effect at the individual level appears
to occur insidiously and without obvious benefit (Study 4)
to work quality.
Although virtually all laboratory experiments
carry
with them implicit time norms, each
of
the three work
sets in both
of
the present paradigms was easily accom-
plished in a few minutes by the present subjects. Parkin-
sonian work expansion may be generalizable to a wide
variety
of
judgment, preference, classification, and mon-
itoring tasks. Whenever anticipated work on the next
task is canceled or, more generally, excess time arises,
dalliance by workers on their present task could amount
to substantial and costly inefficiency.
REFERENCES
ARONSON,
E., &
GERARD,
E. (1966). Beyond Parkinson's law:
The
ef-
fect
of
excess time on subsequent performance. Journal
of
Personal-
ity
& Social Psychology, 3, 336-339.
ARONSON,
E., &
LANDY,
D. (1967). Further steps beyond Parkinson's
law: A replication and extension
of
the excess time effect. Journal
of
Experimental Social Psychology, 3, 274-285.
BRYAN,
J.
E,
&
LOCKE,
E. A. (1967). Parkinson's law as a goal-setting
phenomenon.
Organizational Behavior & Human Performance, 2,
258-275.
CARVELL,
E J. (1980). Human relations in business. New York:
Mac-
millan.
CONKLIN,
J. E. (1984). Sociology: An introduction. New York:
Macmil-
lan.
GIDDENS,
A. (1989). Sociology. Cambridge, U.K.: Polity Press.
HERSEY,
P., &
BLANCHARD,
K. H. (1988). Management
of
organiza-
tional behavior-Utilizing human resources.
Englewood Cliffs, NJ:
Prentice-Hall.
JOHNSON,
B. T. (1993). DSTAT 1.10: Software for the meta-analytic re-
view
of
research literatures
(Version
1.10)
[Computer
software].
Hillsdale, NJ: Erlbaum.
KOONTZ,
H., &
WEIRICH,
H. (1990). Essentials
of
management. New
York: McGraw-Hill.
KORNBLUM,
W.(1988). Sociology in a changing world. New York: Holt.
LANDY,
D., MCCUE, K., &
ARONSON,
E. (1969). Beyond Parkinson's
law:
Ill. The effect
of
protractive and contractive distractions on the
wasting
of
time on subsequent tasks. Journal
of
Applied Psychology,
53,236-239.
LUTHANS,
E (1981). Organizational behavior. New York: McGraw-Hill.
MACIONIS,
J. (1991). Sociology. Englewood Cliffs, NJ: Prentice-Hall.
MCGRATH,
J. E., &
KELLY,
J. R. (1986). Time and human interaction:
Towarda social psychology
of
time. New York: Guilford.
MINER,J.
B. (1980). Theories
of
organizational behavior. Hinsdale, IL:
Dryden.
ORPEN,
C,
&
RIESE,
E. (1973). Beyond Parkinson's law: A failure to
replicate.
Journal
of
Social Psychology,
89,151-152.
PARKINSON,
C. N. (1957). Parkinson slaw. Cambridge, MA: Riverside
Press.
PARKINSON,
C. N. (1980). Parkinson: The law. Boston:
Houghton
Mif-
flin.
PETERS,
L. H., O'CONNOR, E. J.,
POOYAN,
A., &
QUICK,
J. C. (1984).
The relationship between time pressure and performance: A field test
of
Parkinson's law. Journal
of
Occupational Behaviour, 5, 293-299.
REIMANN,
B. C. (1979). Parkinson revisited: A
component
analysis
of
the use
of
staff
specialists in manufacturing organizations. Human
Relations,
32, 625-641.
ROBBINS,
S. P. (1990). Organization theory: Structure, design and ap-
plications.
Englewood Cliffs, NJ: Prentice-Hall.
156 BRANNON, HERSHBERGER, AND BROCK
THIO,
A, (1991), Sociology' A
brief
introduction, New York: Harper-
Collins,
NOTES
I, "In any public administrative department not actually at war, the
staff increase may be expected to follow this formula: x
= (2km+ 1)/n,
where
k is the number of staff seeking promotion through the appoint-
ment of subordinates; I represents the difference between the ages of
appointment and retirement;
m is the number of manhours devoted to
answering minutes within the department; and
n is the number
of
ef-
fective units being administered, x will be the number of new staff re-
quired each year" (Parkinson, 1957, p. 12),
2, Time cuing affected other Parkinson studies as welL Bryan and
Locke (1967) obtained an effect similar to Aronson's, Subjects who
were given excess time took longer to complete a multitrials task than
did subjects who were given minimum time, As in the work by Aron-
son, the instructions were replete with time cues and with the experi-
menters
announcements of temporal expectations, Orpen and Riese
(1973) reported a conceptual replication that likewise included time
cues, Peters, O'Connor, Pooyan, and Quick (1984) reported an am-
biguous correlational study in which employees were asked about task
times and the adequacy of those times,
3, In the absence of main and interaction effects of gender, a con-
founding of subject gender with the gender of people in
photos-for
example, during the test trial, male subjects saw photos of own-gender
faces and females did nor-s-could not account for the overall pattern of
Parkinson effects,
4, "Dalliance" has the primary meaning of "a trifling away of time;
dawdling" (unabridged Random House Dictionary
of
the English Lan-
guage), In the present context, however, no pejorative connotation is at-
tached to "dalliance" other than taking more time than necessary,
(Manuscript received July 7, 1997;
revision accepted for publication April 1, 1998,)
