-
This article was downloaded by: [Western Michigan University]On:
10 January 2012, At: 11:35Publisher: RoutledgeInforma Ltd
Registered in England and Wales Registered Number:
1072954Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH,UK
Journal of OrganizationalBehavior ManagementPublication details,
including instructions forauthors and subscription
information:http://www.tandfonline.com/loi/worg20
Behavior-Based Safety andWorking AloneRyan Olson a & John
Austin aa Western Michigan University, USA
Available online: 12 Oct 2008
To cite this article: Ryan Olson & John Austin (2001):
Behavior-Based Safety andWorking Alone, Journal of Organizational
Behavior Management, 21:3, 5-43
To link to this article:
http://dx.doi.org/10.1300/J075v21n03_02
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use:
http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private
study purposes.Any substantial or systematic reproduction,
redistribution, reselling, loan,sub-licensing, systematic supply,
or distribution in any form to anyone isexpressly forbidden.
The publisher does not give any warranty express or implied or
make anyrepresentation that the contents will be complete or
accurate or up todate. The accuracy of any instructions, formulae,
and drug doses should beindependently verified with primary
sources. The publisher shall not be liablefor any loss, actions,
claims, proceedings, demand, or costs or damageswhatsoever or
howsoever caused arising directly or indirectly in connectionwith
or arising out of the use of this material.
http://www.tandfonline.com/loi/worg20http://dx.doi.org/10.1300/J075v21n03_02http://www.tandfonline.com/page/terms-and-conditionshttp://www.tandfonline.com/page/terms-and-conditions
-
EXPERIMENT
Behavior-Based Safety and Working Alone:The Effects
of a Self-Monitoring Packageon the Safe Performance
of Bus Operators
Ryan OlsonJohn Austin
ABSTRACT. Experimental evaluations of Behavior-Based Safety
(BBS)processes applied with lone workers are scarce. Clinical and
organiza-tional researchers alike have studied the effectiveness of
self-monitoringas a performance improvement strategy, but further
work is needed to de-termine the power of such interventions for
improving safe behavior and
Ryan Olson and John Austin are affiliated with Western Michigan
University.Address correspondence to Ryan Olson, 3308 Miami Avenue,
Kalamazoo, MI
49048 (E-mail: [email protected]).The authors would like to
thank Adam VanAssche and Lisa Olson for implement-
ing critical aspects of the intervention.They would also like to
recognize Alicia Alvero and Scott Traynor for their helpful
input regarding the design of the study.
Journal of Organizational Behavior Management, Vol. 21(3) 2001
2001 by The Haworth Press, Inc. All rights reserved. 5
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
to explore the best practices for using such processes with lone
workers.
In the current study, four male bus operators (20.5 years
average experi-
ence) self-monitored their safe performance and received
feedback
based on self-monitoring data. Dispatch supervisors used radio
commu-
nication to prompt participants to complete self-monitoring
forms and also
conducted special observations of participants to measure target
perfor-
mances. Both operators and supervisors were unaware of
experimental
observers who measured the performance of each participant by
riding on
busses as passengers. A multiple baseline design across
performances was
used to assess the effects of the intervention on four
performance targets.
The intervention resulted in a 12.3% increase in safe
performance for the
group, with individual increases in performance ranging from 2%
to 41%
for specific target performances. The results are discussed in
terms of the
value of BBS processes for employees who work alone and the
research
needed to determine the components of self-monitoring processes
that are
most critical for generating improvements in safe performance.
[Article copiesavailable for a fee from The Haworth Document
Delivery Service: 1-800-HAWORTH. E-mail ad-dress: Website: © 2001
by The Haworth Press, Inc. All rights reserved.]
KEYWORDS. Self-monitoring, behavior-based safety, safe
driving,lone workers, bus transit safety, bus operator
performance
Over the past 20 years behavioral research in the field of
Behav-ior-Based Safety (BBS) has grown steadily. Some of the first
conceptualarticles discussing the application of behavior analysis
technology to im-prove occupational safety were published in the
late 1970’s (e.g., Smith,Cohen, H., Cohen, A., & Cleavland,
1978). The first experimental applica-tions of behavioral
technology applied to occupational safety occurred dur-ing the same
time period (Komaki, Barwick, & Scott, 1978; Smith, Anger,
&Uslan, 1978; Sulzer-Azaroff, 1978). The central foundation of
all BBS re-search since these early applications has been the
identification and mea-surement of safe and at-risk behaviors and
conditions, and the use ofbehavioral technology to increase the
frequency of those safe behaviorsand conditions. The body of
research has demonstrated the effectiveness ofmany different
intervention packages designed to achieve these effects.
Studies have evaluated experimentally the effectiveness of
training(Cohen & Jensen, 1984; Komaki, Heinzmann, & Lawson,
1980; Reber &Wallin, 1984; Reddell, Congleton, Huchingson,
& Montgomery, 1992),
6 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
goal setting and/or prompts (Austin, Alvero, & Olson, 1998;
Berry,Geller, Calef, R. S., & Calef, R. A, 1992; Engerman,
Austin, & Bailey,1997; Fellner & Sulzer-Azaroff, 1986;
Ludwig & Geller, 1991, 1997;Phillips, Sutherland, & Makin,
1994: Reber & Wallin, 1984; Reber,Wallin, & Chhokar, 1990;
Saarela, 1989), verbal and graphic feedback(Alavosius &
Sulzer-Azaroff, 1986, 1990; Babcock, Sulzer-Azaroff,
&Sanderson, 1992; Chhokar & Wallin, 1984; DeVries,
Burnette, &Redmon, 1991; Fellner & Sulzer-Azaroff, 1984;
Komaki, Heinzmann, &Lawson, 1980; Nasanen & Saari, 1987;
Phillips, Sutherland, & Makin,1994; Sulzer-Azaroff & de
Santamaria, 1980), contingent incentives andreinforcement (Austin,
Kessler, Riccobono, & Bailey, 1996; Fox, Hopkins, &Anger,
1987; Komaki, Barwick, & Scott, 1978; McAfee & Winn,
1989;Petersen, 1984), and self-monitoring procedures (McCann &
Sulzer-Azaroff, 1996) at increasing safe behaviors and conditions.
For a recentlypublished, more thorough, review of BBS in
manufacturing settings, seeGrindle, Dickinson, and Boettcher
(2000). For a review of the impact ofBBS on injury rates, see
Sulzer-Azaroff and Austin (2000).
Studies by Geller and colleagues have clear relevance when
discuss-ing driving safety. For example, Ludwig and Geller (2000)
described aseries of seven studies designed to improve the safe
driving of pizza de-liverers. The interventions they evaluated
included public and private feed-back, corporate policy changes,
commitment card strategies, participativeand assigned goal setting,
competition and rewards, and involving thedeliverers as community
intervention agents. Ludwig and Geller (2000)reviewed these seven
studies in terms of the multiple intervention level(MIL) hierarchy.
The MIL is characterized by a continuum of interven-tion
intrusiveness and cost, where the least intrusive and most
inexpen-sive interventions tend to reach the most people and the
most intrusiveand most costly interventions tend to impact the
fewest people. TheMIL is not unlike other discussions of treatment
intrusiveness in ap-plied behavior analysis (e.g., Meinhold &
Mulick, 1990), but the MILspecifically applies these concepts to
organizational behavior. Ludwigand Geller (2000) recommended that
least intrusive interventions be ap-plied to create large-scale
change and that those individuals who remainunaffected by
non-intrusive interventions should be exposed to succes-sively more
intrusive interventions.
Although discussed in theory by the MIL, experimental evaluation
ofself-monitoring procedures to improve safe behavior is a
relatively newdevelopment. The field of BBS is growing and reports
of successful com-mercial applications with lone workers have begun
to surface (e.g.,Krause, 1997; Pettinger, Click, & Geller,
2000). The research base exam-
Experiment 7
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
ining the best practices for improving the safe performance of
lone work-ers is small, however, self-monitoring has been widely
used in othercontexts as a behavior change technique.
SELF-MONITORING
Richman, Riordan, Reiss, Pyles, and Bailey (1988) conducted
astudy that demonstrated the power and utility of self-monitoring
proce-dures for improving organizational performance. Richman et
al. (1988)used in-service training, self-monitoring, and
self-monitoring plusfeedback to improve the on-schedule and on-task
performance of staffat a residential setting for persons with
mental disabilities. A multiplebaseline design across groups was
used to assess the effects of the dif-ferent intervention phases.
Three months prior to the study, participantswere informed that a
special project was going to take place wherestaff/client
interactions would be observed. Under this guise, experi-mental
observers collected data for on-schedule and on-task behaviorfor
the duration of the study. After baseline data were collected,
anin-service training session was held to review job
responsibilities thatincluded topics related to on-schedule and
on-task performance. Duringthe self-monitoring phase, staff members
carried individual schedulecards during the workday and
self-recorded the extent to which theywere on-schedule and on-task
during the shift. These self-recorded datawere handed in at the end
of each shift. For the self-monitoring plusfeedback component,
supervisors provided periodic on-the-spot feed-back regarding
target performances while the self-monitoring proce-dure continued
as before. Two houses participated in the study and werelabeled A
and B. For house A, on-schedule behavior averaged 50%,50%, 80%, and
94% across baseline, in-service, self-monitoring,
andself-monitoring plus feedback conditions respectively. For house
B,on-schedule behavior averaged 39%, 39%, 75%, and 81% across
base-line, in-service, self-monitoring, and self-monitoring plus
feedbackconditions respectively. For on-task behavior, baseline for
both housescombined was 28%. In-service increased the on-task
performance ofstaff in house A to 36% but did not affect the
performance of staff inhouse B. For house A, on-task behavior
averaged 72% and 88% forself-monitoring and self-monitoring plus
feedback respectively. Forhouse B, on-task behavior averaged 77%
and 80% for self-monitoringand self-monitoring plus feedback
respectively.
8 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Self-monitoring, as part of intervention packages, has also been
usedto improve academic performance (Dean, Malott, & Fulton,
1983;DiGangi, Maag, & Rutherford, 1991; Kneedler &
Hallahan, 1981; Lan,1996; Stecker, Whinnery, & Fuchs, 1996), to
improve the performanceof teachers (Browder, Liberty, Heller, &
D’Huyvetters, 1986), to im-prove the performance of athletes
(Kessler, 1985; Srikameswaren,1992; Whelan, Mahoney, & Meyers,
1991), to increase interactions be-tween staff and patients at an
institution (Burgio, Whitman, & Reid,1983), and to help
individuals stop smoking and reduce their caloric in-take (Moinat
& Snortum, 1976). Only some of the research listed abovewas
conducted with adults and targeted “workplace” performance.However,
self-monitoring procedures are potentially relevant across abroader
scope of organizational behavior for people who work aloneand for
people who work in groups. This broader scope includes supportfor
improving the quality, quantity, or timeliness of the performance
ofsalespeople or consultants working outside of the home office
with cli-ents. People working in teams could use self-monitoring
procedures inconcert with peer feedback to track progress on
long-term projects or totarget specific “team relevant” skills. In
relation to the topic of the cur-rent paper, self-monitoring
procedures could complement or substan-tially improve the current
performance management strategies utilizedto support and improve
the performance of people operating any num-ber of different
vehicles in the general transportation and product deliv-ery
industries. For example, the first author has been
exploringsupplementary performance measurement systems for student
pilotsduring the early phases of flight training. An especially
risky phase offlight training involves the first series of solo
flights without an instruc-tor on board. As part of the exploratory
research mentioned above, volun-teer students have been
self-monitoring aspects of landing performance ondual (with an
instructor) and on solo (without an instructor) flights.
Thisproduces data that would otherwise not be available and could
poten-tially improve learning and performance, thereby reducing
risk. Stu-dents have reported that the procedure has enhanced the
learningprocess. If these self-monitoring procedures were used in
combinationwith instructors rating the same performances, such
systems couldprompt feedback and coaching for specific critical
performances. Downthe road when these students become professional
pilots working in thecockpits of planes for major airlines,
self-monitoring procedures andpeer feedback and discussion could be
used to target critical crew re-source management skills. With
these potential applications in mind,we can predict that
self-monitoring procedures will probably prove
Experiment 9
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
valuable across a wide variety of contexts. However, as
self-monitoringresearch and practices expand in organizational
settings, we should alsopredict limits to the effectiveness and
relevance of self-monitoring pro-cedures in organizational
settings.
Self-monitoring procedures have contributed to performance
im-provement across many settings and represent a set of methods
that maybe especially relevant for improving learning and
performance in work-place environments. However, the question of
which components ofself-monitoring procedures are most critical for
generating behaviorchange is still being explored. For example, the
extent to whichself-monitoring data need to be reliable is not
clear. Some research sug-gests that self-monitoring procedures
produce performance improve-ment even when the self-recorded data
are not accurate (Hayes &Nelson, 1983; McCann &
Sulzer-Azaroff, 1996). However, whenself-monitoring data are more
reliable, effects seem to be enhanced(Baskett, 1985; Kanfer, 1970;
McCann & Sulzer-Azaroff, 1996). Itwould be useful to know
whether training participants to reliablyself-monitor is a worthy
investment. An additional consideration re-lated to the
effectiveness of self-monitoring procedures is identifyingthe
behavioral functions of the stimuli generated by such
procedures.Some of the potential behavioral functions of stimuli
produced byself-monitoring processes include: (a) an antecedent
function (i.e., in-formational or task clarification), (b) a
consequence function (condi-tioned reinforcement or punishment),
(c) a rule generating function(i.e., contingency specifying,
function-altering stimuli are evoked), and(d) a conditioned
establishing operation function.
When a participant is asked to record aspects of his or her
behavior,looking at the form, and filling it out may clarify
performance expecta-tions or prompt the most appropriate
performance. Based upon subse-quent observations of behavior with
respect to such informationalstimuli, we would say that
participants “know” or “do not know” thesafe manner in which to
behave (Skinner, 1953). If self-monitoringfunctions primarily as
information or a prompt it would make sense toask participants to
self-monitor at the beginning of the workday or justprior to
opportunities to perform. Antecedents without consequencesare
likely to have only temporary effects due to habituation or
extinc-tion (Daniels, 1989). Therefore, it would also make sense to
ensure thatpositive consequences were correlated with the
antecedent process andthat the self-monitoring procedure was
periodically changed.
Aspects of self-monitoring processes may also function as
conse-quences. Scoring oneself high or low may function as analogs
to rein-
10 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
forcement or punishment for the desired performance depending
uponthe quality of the most recent relevant performance (Malott,
R., Malott,M., & Trojan, E., 1999). Thinking of the potential
consequence func-tion of self-monitoring procedures may also
explain in part why com-pliance with self-monitoring processes is
normally less than perfect.Because performance varies, scoring
aspects of one’s own performancemight sometimes reinforce and
sometimes punish filling out a self-moni-toring form.
Due to the fact that management systems utilize numerous
perfor-mance management strategies, filling out self-monitoring
forms mayalso cause participants to generate rules related to those
strategies. If anorganization regularly uses aversive consequences
to discourage unsafepractices, filling out a self-monitoring form
might evoke rules such as,“If I improve this performance, I can
avoid punishment from my super-visor (because the performances on
this form are what he/she caresabout right now).” Schlinger (1993)
proposed that a rule such as thisone might produce behavioral
effects because it specifies contingenciesand alters the function
of stimuli in the immediate environment. For ex-ample, the rule
above specifies a new contingency (i.e., my supervisorwill punish
me if I don’t improve these behaviors on the form) andmight alter
the function of stimuli in the immediate environment (a pre-viously
ineffective stop sign now evokes behavior that results in a
com-plete stop).
Another way of accounting for the effects of verbal behavior
describ-ing contingencies is the concept of the conditioned
establishing opera-tion (CEO). An establishing operation is a
stimulus or procedure thathas at least two effects; it (1)
momentarily alters the effectiveness of areinforcer or punisher,
and (2) momentarily alters the frequency of be-havior that has been
correlated with the consequence whose effective-ness has been
altered (Michael, 1993). Michael has delineated threetypes of CEOs
with specific characteristics, but discussing these typesis beyond
the scope of this paper. In most cases, CEOs alter the
effec-tiveness of conditioned reinforcers or punishers, and given
the fact thatmost organizational performance is maintained and
shaped by suchconsequences, we should consider the CEO concept a
potentially im-portant motivational variable. When approaching a
relevant opportu-nity to perform, a rule statement related to the
performance beingself-monitored might be evoked. The covert verbal
behavior, or perhapsthe stimulus that evoked the covert behavior,
may then function as aCEO that alters the effectiveness of salient
consequences. For example,a bus driver may perform rolling stops at
stop signs because the brakes
Experiment 11
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
squeal less than when he/she performs a complete stop.
Participating ina self-monitoring procedure that targeted complete
stopping mightcause the sight of a stop sign and/or evoked rule
statements to functionas a CEO that momentarily alters the value of
the squealing sound, mak-ing it less aversive (weakening motivation
to escape or avoid the squeal-ing). Alternatively, CEOs could
momentarily establish the squealingsound as an effective
reinforcer, thereby evoking behavior (firm footpressure on the
brakes) that had produced that consequence in the past.
It is likely that performance improvement generated by
self-monitoringprocedures is caused by a complex set of
contingencies and behavioralmechanisms. Considering these
mechanisms and explanatory conceptsmay guide future research and
help discover the most effective practices.With self-monitoring
research in BBS being scarce, the field may requiremore studies
that demonstrate the effectiveness of self-monitoring proce-dures
to improve safe performance before technical questions can be
ad-dressed. Below we review two applications of self-monitoring
proceduresto improve safe performance that informed the design of
the current study.
BBS APPLICATIONS OF SELF-MONITORING PROCEDURES
Preventing Cumulative Trauma Disorders
McCann and Sulzer-Azaroff (1996) used a behavioral approach
toprevent cumulative trauma disorders with employees who spent
muchof each workday typing in an office setting. Part of the
interventionpackage required typists to self-monitor performance
along particularbehavioral dimensions. Participants were divided
into two groupswhere one group monitored hand and wrist position
and the other moni-tored posture. Each participant was exposed to
conditions in the follow-ing sequence: (a) baseline, (b) training
and self-monitoring, and(c) feedback, goal setting, and
reinforcement. During training, partici-pants were taught
discriminations between safe and at-risk performanceand were
required to pass a discrimination test with a score above
80%correct. Self-monitoring procedures required participants to
estimatethe percentage of time they performed target behaviors
safely. Duringthe final phase of intervention participants met
prior to each session andwere given both graphic and verbal
feedback based on levels of safetyobserved by experimenters on the
previous days. The graphic feedbackwas in the form of
transparencies that, when laid over the participants’self-monitored
data, revealed the accuracy of participants’ reports. Ex-
12 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
perimenters guided participants as they set goals to ensure that
goalswere not set higher than the highest data point from the
previous ses-sion. And finally, praise was provided for progress
and attainment ofgoals.
The study produced consistent improvements in safe
performanceacross all participants with moderate to high
improvements during thetraining and self-monitoring phase, and very
high improvements duringthe feedback, goal setting, and
reinforcement phase. Posture ultimatelyimproved to near perfect
levels for all participants in the posture group.Hand and wrist
position improved to levels clearly above baseline forall
participants in the hand/wrist position group.
Participants were not initially given information about the
accuracyof their self-estimations of safe performance. Without
accuracy infor-mation participants achieved acceptable levels of
agreement betweenself-monitored data for posture and experimenter
data for posture.However, self-monitoring data for hand and wrist
position did not agreewith experimenter data at this stage.
Researchers postulated that the“gross motor” nature of the
movements involved with posture made thebehavior easier to
self-monitor than the “fine motor” hand and wrist po-sition
movements, which resulted in the different agreement levels
be-tween posture and hand/wrist position. The goal setting,
feedback, andreinforcement phase increased the agreement between
self-monitoringdata and experimental data for hand and wrist
position. The reinforce-ment component (verbal praise) was
contingent upon performance im-provement and accurate
self-estimations of performance. The researchersreported that high
agreement between typists and experimenters was as-sociated with
enhanced performance improvement of safe hand and
wristposition.
Improving the Safe Performance of Bus Operators
Krause (1997) reported a Behavioral Science Technology, Inc.
(BST)consultation effort with a public transportation system where
self-moni-toring procedures were utilized. Thirty drivers and
several supervisorsparticipated in the project. Interviews with
drivers were utilized to de-velop a checklist that contained 34
performances. Drivers estimatedtheir safe performance on these 34
targets once or twice daily and plot-ted their own data on graphs.
Every two weeks a supervisor rode witheach driver and collected
data using the same checklist.
When the intervention was initially implemented drivers
reportedhigh percent safe scores that did not agree with
supervisors’ scores of
Experiment 13
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
driver performance. Supervisors discussed these discrepancies
withdrivers and plotted the self-monitoring data and supervisor
data to-gether on feedback graphs. Over a period of 20 weeks,
supervisor datatrended upward and driver data began to trend
downward slightly to al-most match supervisor data. Agreement
between employees and supervi-sors appeared to take place over time
and Krause (1997) reported a 66%decline in injuries and accidents
in the organization over the 20-weektime period. However, the
project did not employ an experimental designand did not include
any formal assessment of the reliability of either su-pervisor or
driver data. Therefore, the degree to which driver’s
behavioractually changed because of the intervention could not be
evaluated.
In order to evaluate experimentally the degree to which
self-monitor-ing procedures can improve the safe performance of
lone workers, a dem-onstration study similar in design to the
McCann and Sulzer-Azaroff(1996) study is needed. The current study
was an attempt to synthesizeaspects of Krause (1997) with McCann
and Sulzer-Azaroff (1996) andexperimentally evaluate the
effectiveness of self-monitoring proceduresfor improving the safe
performance of bus operators.
METHOD
Participants and Setting
A public transportation system serving two midwestern cities
with acombined estimated population of 160,000 was the sponsoring
organi-zation for the study. In addition to operating and
maintaining 17 busroutes, the organization operated rail and other
public transportationsystems. Within the bus system an operations
supervisor managed theperformance of seven dispatch supervisors,
who in turn supervised 65bus and other vehicle operators. A
university campus route serving acampus of approximately 26,000
students was the location for the studywhere two to eight busses
operated from 7 a.m. to 12 midnight on week-days. The bus route
consisted of two directional patterns, each lastingabout 30
minutes, and served all major campus locations includingon-campus
housing.
Four experienced drivers who worked a 10-hour shift (about 6:30
a.m.to about 4:30 p.m.) were selected by the operations supervisor
to partic-ipate in the study (male, ages approximately 40-50;
average experience20.5 years, range: 19-23 years). Organizational
leadership was inter-
14 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
ested in this shift because of its duration and the busy
pedestrian andtraffic conditions of the university campus.
Prior to the study, the transit system used five methods to
motivatesafe driving after drivers were initially trained upon
hire. These meth-ods were (a) a $25 bonus for all drivers who
worked an entire quarterwithout having a preventable collision, (b)
a 7 step progressive disci-pline program for moving violations and
preventable collisions, (c) hir-ing private investigators to
monitor drivers receiving serious complaints,(d) yearly safety
awards at a banquet, and (e) bi-monthly general perfor-mance
evaluations by dispatch supervisors. The operations supervisor
re-ported that this general management strategy had produced a
plateau in totalcollisions per year that had remained relatively
stable over the past five years.
Dependent Variables
Dependent variables were identified through an assessment that
in-cluded a review of one year of collision reports from the
organization’srecords. Passenger and pedestrian injury reports were
also reviewed butwere so infrequent that no patterns could be
discerned from them. Giventhe high pedestrian traffic conditions of
the campus bus route, it islikely that risk for these kinds of
events was high when compared toother routes within the transit
system. The degree to which acceptableIOA could be achieved was the
final consideration for the selectionof dependent variables.
Performances observed were divided intothree categories: (a)
loading/unloading passengers, (b) bus in motion,and (c) complete
stop. Bus in motion performances related to corneringsafely and
maintaining adequate following distance were excludedfrom the study
due to ceiling effects.
Loading/unloading performances included bus stopping position,
re-maining motionless for two seconds after an unload/load
instance, andmirror checking. The assessment discovered that 20% of
preventablecollisions had occurred at loading zones and another 12%
of prevent-able collisions had occurred at parking lots or
driveways. Checking mir-rors was identified as a behavior that may
have helped prevent 56% ofthe collisions reviewed in the
assessment. The route involved in thestudy passed through six major
campus parking lots. Moreover, manyloading zones were located near
parking lot exits and various otherthroughways.
Bus stopping position was defined as “bus doors must remain
shutuntil the bus is completely stopped, and the bus should be
positioned sono cars can pass on the right.” Observers scored this
performance by
Experiment 15
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
watching the front doors of the bus as it slowed. If the bus was
still mov-ing when the doors separated, or if a car could pass on
the right, the per-formance was scored at-risk. Two seconds
motionless was defined as“the bus should remain motionless for at
least two seconds after the lastloading/unloading passenger either
steps behind the yellow line on thebus, steps off the bus to the
right, or steps clear of the front left corner ofthe bus.”
Observers were instructed to count “one-thousand one,one-thousand
two,” to themselves to measure this performance andused a
wristwatch to periodically calibrate the pace of their counting.
Ifthe observer was able to reach “two” before the bus moved the
perfor-mance was scored correct while any movement before the
observerreached “two” was scored as at-risk. Mirror checking was
defined as“the driver should visually check both side mirrors after
loading/un-loading passengers as the bus pulls out of a loading
zone.” Observerswere instructed to mark this performance as correct
if both mirrors werechecked before or as the bus started moving.
Checking mirrors after theback of the bus cleared the original
load/unload location was scoredat-risk. From the driver’s right
hand side of the bus in the second row offorward facing seats his
eyes were visible in the center mirror and headmovement could be
viewed. If a driver looked in the general directionof either mirror
it was assumed he checked that mirror.
Complete termination of forward motion at stop signals is a
legal re-quirement and was considered an important safe performance
for thecampus route. Drivers making complete stops have a better
opportunityto scan traffic and pedestrian conditions at busy
intersections. Therewere over 20 stop signals during each 30-minute
loop regardless of thedirection the bus was traveling. Rolling
stops and jumping a traffic sig-nal were scored as at-risk. The
observation technique that achieved reli-ability for complete stops
involved picking out an outside object like apole and watching it
as the bus slowed. If the outside object stood still inthe
observer’s field of vision the performance was scored as safe.
A percent safe score for each dependent measure was calculated
bycounting the number of correct scores and dividing that number by
thetotal number of observations for that dependent measure, and
then mul-tiplying by 100. An overall percent safe score for each
observation ses-sion was also calculated in a similar fashion.
Observers and Observation Procedures
The first author and two undergraduate research assistants
worked asexperimental data collectors over the course of the study.
Undergradu-
16 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
ate research assistants were selected on the basis of good
performancein an organizational psychology class, interest, and
availability. Allthree researchers worked for research credits at
Western Michigan Uni-versity. Observers sat at the driver’s right
hand side of the bus in the sec-ond row of forward facing seats
about 10 feet from the driver’s chair.Bus schedules were marked
with color codes to locate participants andeach driver was
identified by their color code throughout the study
forconfidentiality reasons. As an additional measure of
confidentiality colorcodes were changed to participant numbers for
this paper. Each driverwas generally observed at least once each
day for at least 30 minutes (i.e.,one directional loop of the
route). However, observers were required tomonitor at least 10
instances of loading/unloading of passengers per ses-sion,
resulting in some sessions longer than 30 minutes. On average,
therewere 10 or more load/unload instances and over 20 stops
observed eachsession.
Once or twice each week all four participants were observed by
twoobservers to assess IOA, which was calculated by dividing the
numberof agreements by the number of agreements + disagreements,
and thenmultiplying by 100. During reliability sessions, the first
author was theprimary observer. To protect the independence of
observations, the ob-server sitting on the right hand seat next to
the window used a three-ringbinder with the left cover held upright
to block the visibility of the datasheet. The observer sitting on
the left hand seat covered his/her datasheet with his/her right arm
and hands (all observers were right handed).
Methods to Minimize Driver Reactivity to Experimental
Observers
Participating drivers were not informed of experimental
observers un-til a post-experiment debriefing.1 However, it was odd
for passengers toride an entire loop of the route without arriving
at a destination and driv-ers occasionally asked questions.
Observers, who were earning researchcredits for participation, were
instructed to answer such questions by say-ing “I’m collecting a
survey for a class.” Surveys on bus ridership hadtaken place
recently and this proved to be an effective strategy. To
furtherreduce the possibility of untoward interactions with drivers
observerswere instructed to wear headphones when collecting data by
themselves.
Independent Variables
Project Kick-Off Meeting. After an initial baseline phase
interven-tion began with an hour and a half meeting at the transit
station hub that
Experiment 17
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
consisted of an introduction to BBS and the rationale for
piloting such aprocess at the transit system, an introduction to
and rationale for aself-monitoring process, and finally a
description of the details of run-ning the project. The meeting was
conducted by a doctoral student (male,age 26) not involved in data
collection and by the operations supervisor.The student was
introduced as an external safety consultant without men-tioning his
ties to the university. Participants were informed that their
in-put about the project would be solicited at a post-project lunch
and thatorganizational leadership had signed an agreement that
information ob-tained during the project could not be used for
disciplinary purposes.Drivers were also told that one or two
additional meetings with the con-sultant would be scheduled over
the next few weeks. Immediately afterthe kick-off meeting, the
student consultant and the operations supervisormet with dispatch
supervisors to introduce them to the project. Supervi-sors were not
informed of the presence of experimental observers.
Self-Monitoring. Three different self-monitoring forms were
usedover the course of the study and were introduced during
meetings at thetransit hub (see Experimental Design section below).
Drivers used theseforms twice each day during their 10-hour shift
to estimate the percent-age of time they performed each of the
target performances safely.Blank squares were provided on the form
for writing estimations, whichis one strategy suggested by research
to avoid shaping respondent an-swers (Schwarz, 1999). At the
drivers’ request the locked drop box forself-monitoring forms was
located in the drivers’ lounge at the transitsystem hub. Drivers
were also told that they would be prompted twice aday by their
dispatch supervisors via radio when it was time to
self-moni-tor.
Feedback. The first author generated daily color-coded
individualand group graphs based on self-monitoring data from the
previous day.A research assistant posted a new set of graphs each
evening between 8and 9 p.m. in the drivers’ lounge near the drop
box and collected com-pleted self-monitoring forms. Each driver was
asked to initial the groupgraph at the conclusion of each shift to
demonstrate that the feedbackhad been viewed.
Supervisor Prompts and Observations. Dispatch supervisors were
in-structed to prompt participating drivers via radio twice each
day to usethe self-monitoring forms and record the date and time of
their promptson a chart posted in the dispatch office. In addition
to deliveringprompts, supervisors conducted special observations of
drivers using adata sheet (identical in format to experimental data
sheets) limited to theperformances currently being self-monitored
by drivers. Experimental
18 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
observers arranged to measure performance concurrently with
supervi-sor observations. On these occasions, experimental
observers boardedthe bus prior to the supervisor visit and left the
bus one or two stops afterthe supervisor left the bus. This
procedure was added to the design ofthe study as a type of probe,
where performance changes generated bythe presence of a supervisor
could be measured and compared to datacollected on the same day
without supervisor presence. To create thiscomparison, each driver
was observed for an additional session on thesame day either before
or after the supervisor probe was completed.
Independent Variable Integrity. Three measures of independent
vari-able integrity were calculated. Percentage of compliance with
theself-monitoring procedure was calculated by counting the actual
num-ber of self-monitoring forms completed by each driver, dividing
thatnumber by the expected number of completed self-monitoring
formsfor each driver (two per day), and then multiplying by 100.
Percentageof compliance with feedback procedures was calculated by
counting thenumber of days each driver signed the feedback graph,
dividing that fig-ure by the number of days the driver was expected
to sign the feedbackform, and then multiplying by 100. And finally,
the percentage of super-visor compliance with delivering prompts
was calculated by countingthe number of prompts recorded on the
supervisor form, dividing thatfigure by the number of prompts that
were expected to be given, andthen multiplying by 100.
Experimental Design
A multiple baseline design across performances was used to
assessthe effects of the intervention. Intervention began after a
baseline of 9 to11 sessions for each individual driver (group
baseline sessions totaled13 because individual driver baseline
sessions were obtained across dif-ferent days). Intervention was
first implemented for complete stop per-formance and lasted for
eight workdays while baseline conditionscontinued for the remaining
dependent variables. Phase two added theperformance of remaining
motionless for two seconds after loading/un-loading passengers and
lasted for five workdays while baseline condi-tions continued for
the remaining dependent variables. The third andfinal phase of
intervention introduced checking mirrors and bus stop-ping
position. After five more working days using this final form,
theroute stopped running for the semester and the study was
concluded.
Experiment 19
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
RESULTS
Group Performance
Group performance was calculated by creating a group percent
safescore for each day of the experiment for each dependent
measure. Forpractical purposes, average improvement percentages for
each depend-ent measure were then summed and divided by four to
obtain a simpleoverall improvement percentage. Using this method,
the group im-proved safe driving by an average of 12.3% over
baseline conditions.The dependent variable realizing the largest
improvement for the groupwas complete stop, which improved by an
average of 21.2% (range:14%-41%). Two seconds motionless after
loading/unloading passen-gers improved by an average of 11.8%
(range: 3%-19%), mirror checkimproved by an average of 10% (range:
3%-15%), and bus stopping po-sition improved by an average of 6.2%
(range: 2%-12%). Figure 1 rep-resents the grouped data (i.e.,
averaged across the four drivers for eachday of data collection)
for each of the four dependent variables in themultiple baseline
design. As can be seen during visual inspection ofgroup data,
calculations of average “improvement” may not indicateclear
effects, especially with regard to the last phase of the
intervention.For example, visual inspection does not indicate any
clear effect for busstopping position (patterns in intervention
data closely resemble pat-terns in baseline data). For mirror
check, an up trend in the data suggestsan effect, but more data
would be required to draw this conclusion.
Individual Performance
The results of individual participants are presented in order of
largestto smallest overall improvement. The word “improvement” is
usedthroughout the discussion of individual performance, although
it shouldbe understood that small average increases for specific
dependent mea-sures do not necessarily indicate clear effects of
the intervention. Forexample, average improvement percentages for
phase 3 of the interven-tion are based on only 3 to 4 data points
and should therefore be inter-preted conservatively. Overall
improvement percentages for eachparticipant are necessarily
influenced by this characteristic of the exper-iment, and should
also be interpreted conservatively. Several alterna-tives for
computing overall improvement were considered, but forpractical
purposes, the same method used to calculate overall group
im-provement was applied to individual participants. Percentages
related
20 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Experiment 21
COMPLETE STOP
86.5%
68%46.8%
Perc
entS
afe
100
80
60
40
20
01 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Days
LOAD/UNLOAD 2 SECONDS MOTIONLESS
60%
45.3% 57.1%
Perc
entS
afe
100
80
60
40
20
01 3 5 7 9 11 13 15 17 19 21
Days
23 25 27 29
LOAD/UNLOAD MIRROR CHECK99%
68.3%58.3%Perc
entS
afe
100
80
60
40
20
01 3 5 7 9 11 13 15 17
Days
19 21 23 25 27 29
LOAD/UNLOAD STOPPING POSITION
72.8% 79%
95.5%
Perc
entS
afe
100
80
60
40
20
01 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Days
FIGURE 1. Group results in multiple baseline design format.
Closed circle datapoints are experimenter data averaged for each
day of the experiment andopen circle data points are
self-monitoring data averaged for each day of theexperiment.
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
to individual performance have been rounded to the nearest whole
num-ber to make them easier to read.
Participant 1 improved by an average of 14% over baseline
levels.The largest average improvement was for two seconds
motionless witha 19% improvement (baseline, 43% safe; intervention,
62% safe). Busstopping position improved 8% (baseline, 70% safe;
intervention, 78%safe), mirror check improved 15% (baseline, 73%
safe; intervention,88% safe), and stopping improved 14% (baseline,
63% safe; interven-tion, 77% safe). A supervisor probe on the first
day of phase two of theintervention created systematic effects on
the performance of partici-pant 1. Complete stop and two seconds
motionless, which were beingself-monitored, improved to over 20%
above the levels measured on thesame day without supervisor
presence. Mirror check and bus stoppingposition, which were still
under baseline conditions, did not change inthe presence of the
supervisor. For a graphic display of these data seeFigure 2.
Participant 2 improved by an average of 13% over baseline
condi-tions. His largest average improvement was for complete stop
with a41% improvement (baseline, 51% safe; intervention, 92% safe).
Thisimprovement stands out as the most clear and dramatic effect of
the in-tervention procedures. Bus stopping position improved 3%
(baseline,49% safe; intervention, 52% safe), two seconds motionless
improved3% (baseline, 28% safe; intervention, 31% safe), and mirror
check alsoimproved 3% (baseline, 38% safe; intervention, 41% safe).
For agraphic display of these data see Figure 3.
Participant 3 improved by an average of 12% over baseline
condi-tions. His largest average improvement was for mirror check
with a15% improvement (baseline, 65% safe; intervention, 80% safe).
Busstopping position improved 12% (baseline, 81% safe;
intervention,93% safe), two seconds motionless improved 12%
(baseline, 47% safe;intervention, 59% safe), and complete stop
improved 9% (baseline,38% safe; intervention, 47% safe). For a
graphic display of these datasee Figure 4.
Participant 4 improved by an average of 10% over baseline
condi-tions. His largest average improvement was for complete stop
with a19% improvement (baseline, 38% safe; intervention, 57% safe).
Busstopping position improved 2% (baseline, 94% safe; intervention,
96%safe), two seconds motionless improved 5% (baseline, 66% safe;
inter-vention, 71% safe), and mirror check improved 15% (baseline,
58%safe; intervention, 73% safe). For a graphic display of these
data seeFigure 5.
22 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Experiment 23
COMPLETE STOP
77%63%
100
89
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
24
Sessions
Perc
entS
afe
100
80
60
40
20
0
Perc
entS
afe
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
24
Sessions
LOAD/UNLOAD 2 SECONDS MOTIONLESS 100
83
62%43%
Sessions
242322212019181716151413121110987654321
Perc
entS
afe
0
20
40
60
80
100 LOAD/UNLOAD MIRROR CHECK
88%73%
242322212019181716151413121110987654321
Sessions
0
20
40
60
80
100
Perc
entS
afe
LOAD/UNLOAD STOPPING POSITION
70% 78%
FIGURE 2. Participant 1 results in multiple baseline design
format. Closed cir-cle data points are experimenter data, open
circle data points are self-monitor-ing data, closed triangles are
experimenter data during supervisor probes, andopen triangles are
supervisor data.
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
24 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
1 2 3 4 5 6 7 8 9 10 11 12 13
Sessions
14 15 16 17 18 19 20 21 22 23 24 25
100
Perc
entS
afe
80
60
40
20
0
92%
51%
100
100
COMPLETE STOP
1 2
Sessions
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
25
Perc
entS
afe
0
20
40
60
80
100
LOAD/UNLOAD 2 SECONDS MOTIONLESS
28% 31%
57
55
2524232221201918171615
Sessions
1413121110987654321
0
20
40
60
80
Perc
entS
afe
100
38%93
64
41%
LOAD/UNLOAD MIRROR CHECK
100
80
60
40
20
0
Perc
entS
afe
Sessions
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
25
49% 52%
100
91
LOAD/UNLOAD STOPPING POSITION
FIGURE 3. Participant 2 results in multiple baseline design
format. Closed cir-cle data points are experimenter data, open
circle data points are self-monitor-ing data, closed triangles are
experimenter data during supervisor probes, andopen triangles are
supervisor data.
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Experiment 25
1
Perc
entS
afe
2
100
3
80
4 5
60
6 7 8
40
9 10 11
Sessions
20
0
12
COMPLETE STOP
7667
38% 47%
13 14 15 16 17 18 19 20 21 22 23 24
Perc
entS
afe
20
0
40
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sessions
16 17 18 19 20 21 22 23 24
LOAD/UNLOAD 2 SECONDS MOTIONLESS
86
70
47% 59%
60
80
100
1 2 3 4 5 6 7
0
20
40
60
Perc
entS
afe
8Sessions
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
65% 80%
100
86
LOAD/UNLOAD MIRROR CHECK
81% 93%
100
75
Perc
entS
afe
0
1 2 3 4 5 6
20
40
7Sessions
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
60
80
100
80
100
LOAD/UNLOAD STOPPING POSITION
FIGURE 4. Participant 3 results in multiple baseline design
format. Closed cir-cle data points are experimenter data, open
circle data points are self-monitor-ing data, closed triangles are
experimenter data during supervisor probes, andopen triangles are
supervisor data.
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
26 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Perc
entS
afe
100
80
60
40
20
0
COMPLETE STOP100
80
1
38%
57%
2 3
Sessions
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
26
26252423222120
Sessions
191817161514131211109876
0
20
40
60
Perc
entS
afe
54321
80
100
66%
71%
LOAD/UNLOAD 2 SECONDS MOTIONLESS
2625242322212019181716151413
Sessions
121110987654321
Perc
entS
afe
0
20
40
60
80
100
58%73%
LOAD/UNLOAD MIRROR CHECK
Perc
entS
afe
0
20
40
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Sessions
17 18 19 20 21 22 23 24 25 26
80
100
94%
96%
LOAD/UNLOAD STOPPING POSITION
FIGURE 5. Participant 4 results in multiple baseline design
format. Closed cir-cle data points are experimenter data, open
circle data points are self-monitor-ing data, closed triangles are
experimenter data during supervisor probes, andopen triangles are
supervisor data.
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Results of Self-Monitoring Estimations
It should be noted that drivers estimated their performance for
an en-tire day with two self-observations, and experimenters only
sampledtheir behavior between 30 minutes to 60 minutes each day.
Therefore,the comparison between experimenter and self-monitoring
data is notan exact comparison. Drivers’ self-estimations are
plotted as open cir-cles on individual results figures.
For participant 1, the average of percent safe estimations
across allintervention phases was 72% (complete stop, 79%; two
seconds mo-tionless, 67%; mirror check, 94%; bus stopping position,
81%). His ac-tual overall percent safe score, as calculated from
experimental observa-tions, was 73% (complete stop, 77%; two
seconds motionless, 62%;mirror check, 88%; bus stopping position,
78%). The largest discrep-ancy between his self-monitoring data and
experimenter data occurredfor mirror check, with a difference of
6%. The smallest discrepancy oc-curred for complete stop, with a
difference of 2%.
For participant 2, the average of percent safe estimations
across allintervention phases was 98% (complete stop, 99%; two
seconds mo-tionless, 100%; mirror check, 100%; bus stopping
position, 81%). Hisactual overall percent safe score as calculated
from experimental obser-vations was 53% (complete stop, 92%; two
seconds motionless, 31%;mirror check, 41%; bus stopping position,
52%). The largest discrepancybetween his self-monitoring data and
experimenter data occurred for twoseconds motionless, where the
difference was 69%. The smallest discrep-ancy occurred for complete
stop, where the difference was 7%.
For participant 3, the average of percent safe estimations
across allintervention phases was 78% (complete stop, 85%; two
seconds mo-tionless, 99.9%; mirror check, 100%; bus stopping
position, 100%). Hisactual overall percent safe score as calculated
from experimental obser-vations was 65% (complete stop, 47%; two
seconds motionless, 59%;mirror check, 80%; bus stopping position,
93%). The largest discrep-ancy between his self-monitoring data and
experimenter data occurredfor two seconds motionless, where the
difference was 40.9%. Thesmallest discrepancy occurred for bus
stopping position, with a differ-ence of 7%.
For participant 4, the average of percent safe estimations
across allintervention phases was 74% (complete stop, 82%; two
seconds mo-tionless, 18%; mirror check, 100%; bus stopping
position, 100%). Hisactual overall percent safe score as calculated
from experimental obser-vations was 71% (complete stop, 57%; two
seconds motionless, 71%;
Experiment 27
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
mirror check, 73%; bus stopping position, 96%). The largest
discrep-ancy between his self-monitoring data and experimenter data
occurredfor two seconds motionless, where the difference was 53%.
The smallestdiscrepancy was for bus stopping position, where the
difference was 4%.
Independent Variable Integrity
Group compliance with the rule to fill out two estimations of
safeperformance each day was 76.5%. During phases one, two, and
three ofthe intervention, compliance was 91.5%, 72.5%, and 60.5%
respec-tively. Group compliance with the rule to sign the feedback
graph at theend of each shift was 58.8%. During phases one, two,
and three of theintervention, compliance was 43.3%, 52%, and 85.5%
respectively.Drivers received 68.3% of the supervisor prompts via
radio that wereplanned. During phases one, two, and three of the
intervention, supervi-sor compliance with the prompting procedure
was 66%, 81.5%, and57.5% respectively. Individual participants
received at least one prompton 88.3% of the days during the
project, and received two daily promptson 48.3% of the days during
the project. Independent variable integrityfor individual
participants is summarized in Table 1.
Reliability
A total of 99 experimental observations of driver performance
tookplace over the course of the study. Two independent observers
collecteddata simultaneously for 30 sessions (30.3% of total
sessions). The aver-age agreement percentage was 89.8% (range:
70-100). IOA scores werecalculated for each dependent variable for
every IOA session. Agree-ment scores under 80 percent were limited
to 11 out of 120 total IOAcalculations. Table 2 shows ranges of IOA
scores for each dependentvariable over the course of the study.
Debriefing
At the conclusion of the study the participants met with the
opera-tions supervisor and student consultant for lunch and
debriefing. A sur-vey was administered to the drivers to
investigate issues related to thestudy and solicit their opinions
about the process, and afterwards, par-ticipants were informed
about experimental observers and were eachprovided with a coded
summary of self-monitoring results and the aver-age percent
improvement for each individual as observed by experi-
28 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
mental observers. Participant responses to this information were
positive.After discussing all questions that were raised during
debriefing the oper-ations supervisor left the room while consent
was obtained for the use ofdata.
Survey results showed that participants believed their
self-monitor-ing estimations were accurate to slightly high.
Participants also identi-fied which performances had actually
changed the most and which hadchanged the least over the course of
the study. Drivers were given op-portunities in the survey to
describe why they thought their perfor-mance had improved or stayed
the same. Comments on this topic wereinformative and will be
presented when relevant in the discussion sec-tion. Participants
rank-ordered aspects of the project from most to leastuseful in the
following order: (1) being able to share opinions about theproject,
(2) talking with co-workers about safety and aspects of theroute,
(3) meetings to discuss the project, (4) using
self-monitoringforms, (5) graphs of safe performance, (6) process
not attached to disci-
Experiment 29
TABLE 1. Independent Variable Integrity
Participant and Variable Phase One Phase Two Phase Three All
Phases
Participant 1
Self-Monitoring 100.0 100.0 33.0 82.0
Feedback 50.0 33.0 67.0 50.0
Supervisor Prompts 80.0 66.7 66.7 73.0
Overall IV Integrity 76.7 66.6 55.6 68.3
Participant 2
Self-Monitoring 100.0 69.0 67.0 83.0
Feedback 83.0 75.0 100.0 85.0
Supervisor Prompts 58.3 87.5 50.0 65.0
Overall IV Integrity 80.4 77.2 72.3 77.7
Participant 3
Self-Monitoring 83.0 63.0 67.0 73.0
Feedback 40.0 75.0 75.0 62.0
Supervisor Prompts 58.3 75.0 50.0 62.0
Overall IV Integrity 60.4 71.0 64.0 65.7
Participant 4
Self-Monitoring 83.0 38.0 75.0 68.0
Feedback 0.0 25.0 100.0 38.0
Supervisor Prompts 80.0 75.0 62.5 73.0
Overall IV Integrity 54.3 46.0 79.2 59.7
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
pline in any way, (7) supervisors observed the same behaviors we
did,and (8) more frequent contact from supervisors. All four
participantsrecommended extending the use of customized
self-monitoring pro-cesses to other parts of the organization for
both new and experienceddrivers. They also responded favorably to
having the union participatein choosing target behaviors.
DISCUSSION
The results of the study suggest that a self-monitoring package
canchange the safe performance of bus operators. Furthermore, the
studyrepresents a rare empirical evaluation of lone worker
performance.However, because of the small number of participants
and short dura-tion of the study, it cannot be concluded that
changes in safe behaviorled to an important decline in collisions.
All four participants were “col-lision free” for five weeks, but
the transit system as a whole had threeseparate months without
collisions in 1997.
The overall effects of the intervention were small to moderate
(12.3%overall average improvement; individual performance
improvement onspecific targets range: 2% to 41%). This may have
been due in part tothe lack of participant involvement in
activities such as performancetarget selection and design of the
process (i.e., “buy-in” activities). Thefact that participants were
aware of the short-term nature of the projectmay have also
contributed to this effect and, for some participants, mayhave been
the reason for low treatment compliance. Perhaps some didnot take
the procedures “seriously” because the process was presentedas
temporary rather than permanent. Only future research can
answerthese questions conclusively. However, the results of this
study do sug-
30 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
TABLE 2. Inter-Observer Agreement Percentages for Each Dependent
Vari-able
Dependent Variables Average % IOA Range % IOA Sessions <
80%
Bus Stopping Position 93.2 70.0-100 2
2 Seconds Motionless 90.0 73.0-100 0
Mirror Check 84.1 70.0-100 8
Complete Stop 91.8 77.2-100 1
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
gest that there are limits to the effectiveness of
self-monitoring inter-ventions (i.e., not all behaviors improved;
certain antecedent conditionsor additional variables may be
necessary to ensure that all behaviors im-prove). Indeed, it may be
very difficult to ensure that self-monitoringpackages produce
substantial behavior change.
The Cumulative Benefits of Small Effects
In the current study the intervention phases were relatively
short,with the entire intervention lasting only three weeks.
Without any op-portunity to significantly involve participants or
to allow participants tobecome familiar with the new process, a
12.3% improvement in overallsafe performance was achieved. It is
possible that larger effects wouldoccur under more supportive
circumstances. This fact notwithstanding,we should consider the
possible practical importance of the level of be-havior change
observed in the current study. Mawhinney (1999) notedthat it is
important to consider how small to moderate improvementswould
impact an organization over time. To investigate this issue wewill
consider the potential impact of the changes made by participant
3,whose overall average improvement of 12% was not clearly visible
to usin graphic form until after mean lines were added (see Figure
4). As sug-gested by Mawhinney (1999), we agree that “cumulatively
large benefitscan result from incrementally small intervention
effects” (p. 83).
On the campus route there were usually about 10 instances of
load-ing/unloading passengers every 30 minutes. During a ten-hour
shiftwith a regular flow of passengers, each driver could stop to
unload orload 200 times each day. During baseline conditions,
participant 3checked both side mirrors 65% of the time when
loading/unloading pas-sengers. This would represent 130 safe mirror
checks out of 200 oppor-tunities each day. During brief
intervention conditions, he checked bothside mirrors 80% of the
time. This would represent 160 safe mirrorchecks out of 200
opportunities each day. During one month perform-ing at baseline
levels participant 3 would achieve 2080 safe mirrorchecks out of
3200 opportunities whereas one month of interventionlevel
performance would achieve 2560 safe mirror checks out of 3200.So a
15% average improvement on checking mirrors could result in asmany
as 480 fewer at-risk load/unload instances each month. If the
re-maining 64 drivers working in the transit system were also
participatingand improved to similar levels (assuming similar
passenger rates), thetransit system could realize 31,200 fewer
at-risk behaviors each monthand 374,400 fewer at-risk behaviors
each year. Managers and practitio-
Experiment 31
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
ners applying BBS with lone workers should find moderate effect
sizespromising, especially when maintained over longer periods of
time.
Behavioral Functions of the Self-Monitoring Package
The effect size and variability produced in this study are
similar to ef-fects and variability generated with antecedent
interventions targetingsafety (e.g., Austin, Alvero, & Olson,
1998; Engerman, Austin, &Bailey, 1997; Ludwig & Geller,
1997; Streff, Kalsher, & Geller, 1993).Although the performance
targets of the above studies are not identicalto the targets in the
current study, all seem to involve a common safetydilemma where
immediate and probable consequences support riskyperformance, while
delayed or improbable consequences fail to supportsafe performance.
Given the similar behavioral underpinnings of perfor-mance targets,
it is interesting to note that the self-monitoring package
gen-erated effects similar in magnitude to purely antecedent
strategies. Incontrast to antecedent strategies, most safety
studies that use pro-grammed consequences have demonstrated much
larger changes in be-havior (e.g., Austin, Kessler, Riccobono,
& Bailey, 1996; Sulzer-Azaroff & de Santamaria, 1980).
These results, combined with the ef-fect size and variability
issues discussed above, leads us to believe thatour self-monitoring
procedure might have served an important anteced-ent function.
One may also argue, in line with Hayes and Nelson (1983), that
thewhole self-monitoring package (the instructions, the sheets,
theprompts, and posted feedback) made more effective the natural
conse-quences of the particular performances we measured. That is,
monitor-ing complete stops, for example, could have made the
potentialconsequences of behaving unsafely (e.g., colliding with
student pedes-trians or other vehicles) more salient. One driver’s
answer to a surveyquestion highlights this issue. When explaining
why some of his behav-ior did not change very much, participant 2
circled the statement “acci-dents/collisions just don’t happen
often enough to warrant any extraeffort to prevent them.”
Alternatively, participant 1 reported the follow-ing with regard to
the self-monitoring package, “It caused me to con-sider the effects
on others (students) of my errant behavior (rollingstops).”
Aversive outcomes like collisions, as horrific as they may be,tend
to be too improbable to motivate safe behavior. In addition, the
saf-est way of doing things often requires the person to endure
immediateaversive conditions (taking longer to complete a task,
wearing uncom-fortable personal protective equipment, etc.).
Reports such as those
32 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
from participant 1 suggest that the self-monitoring package, in
somecases, overcame the contingencies favoring risk-taking by
making theconsequences for unsafe acts more salient. From a
molecular analyticperspective, the self-monitoring package may have
generated CEOsthat altered the effectiveness of direct-acting
reinforcers or punishers(Malott, R., Malott, M., & Trojan, E.,
1999; Michael, 1993).
Discussion of Individual Performance
It was hoped that very consistent effects would be observed
across par-ticipants, or at least systematic improvements related
to the degree towhich participants complied with intervention
procedures. However,each participant’s largest improvement was not
necessarily the most ac-curately self-estimated performance. Among
individual participantsthere were very small to very large
improvements for specific target per-formances. Understanding these
individual differences in performancerequires a consideration of
the accuracy of each participant’s self-moni-toring estimations,
the extent of exposure to the independent variables(i.e.,
independent variable integrity) for each participant, the
self-reportdata obtained from each participant, and anecdotal
information obtainedby experimental observers. Because aggregate
data can obscure interven-tion effects and relationships between
variables, we chose to providemore detailed analyses of the data
for participants 1 and 4.
Participant 1 Performance
Participant 1 realized the greatest average improvement (14%)
andthe most consistent improvements of any participant. He was also
themost accurate self-estimator of safe performance. Upon visual
inspec-tion of his data, it is clear that his estimations closely
tracked his actualperformance (see Figure 1). The data from
participant 1 support thefindings of McCann and Sulzer-Azaroff
(1996), suggesting that thegreatest improvements in safe
performance occur when participants aremost accurate in their
self-estimations.
Overall independent variable integrity for participant 1 was
68.3%(Phase one, 76.7%; Phase two, 66.6%; Phase three, 55.6%). The
declinein integrity percentages was largely the result of decreased
participationin self-monitoring procedures (he was 33% compliant
during phasethree). This may explain the sharp drop in his
performance on bus stop-ping position during the last two days of
intervention (see Figure 1).
Experiment 33
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
Anecdotally, participant 1 appeared to be deliberate and
conscien-tious and seemed to take great pride in his profession. It
is possible thatcertain “personality characteristics” (i.e., verbal
responses to questionson valid psychometric instruments) could
predict the effectiveness of orcompliance with self-monitoring
procedures in some cases. Participant1 also responded very
systematically to supervisor observations, duringwhich his
performance on the variables being self-monitored was about20%
higher than his performance on the same day without
supervisorpresence. During supervisor observations, dependent
variables thatwere not being self-monitored remained at baseline
levels. This effectsuggests relatively low reactivity to
experimental observers as com-pared to reactivity to supervisor
presence.
Participant 4 Performance
Participant 4 achieved the smallest overall improvement for
thegroup (10%) and also had the lowest overall independent variable
integ-rity of all participants (59.7%). The clearest effects for
this participantoccurred during the first phase of intervention.
During baseline condi-tions he seemed to come to a complete stop
only when he was forced todo so by traffic conditions. His typical
pattern of performance was toroll slowly through stop signs. This
distinctive performance duringbaseline made behavior changes
observed on the first day of interven-tion very dramatic. The
deterioration of this improvement in perfor-mance was also
distinctive as it gradually returned to baseline levelsover the six
sessions of phase one (see Figure 5). Contributing to this ef-fect
may have been the fact that participant 4 did not ever sign the
feed-back graph during phase one of the intervention, suggesting he
did notview the graphs, which may have eliminated a consequence
componentfrom the self-monitoring package.
Another clear effect achieved during the first phase of the
study forthis participant occurred during the supervisor probe. He
scored 30%higher on complete stops when the supervisor was present
than he didwhen the same performance was measured on the same day
without thepresence of a supervisor. In addition, the baseline
dependent measuresall showed slightly lower performance with the
supervisor present thanthey did without the presence of the
supervisor, showing that the partici-pant was reactive only to the
performance being self-monitored.
At the onset of phase two, participant 4’s performance dropped
to 20%and 16% safe on 2 seconds motionless and complete stop
respectively. Atthat time we hypothesized that this might represent
counter-controlling
34 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
behavior in response to the intervention procedures (Ludwig
& Geller,1999; Skinner, 1953). However, this low performance
did not continuebeyond the first day of phase two of the
intervention.
We selected participant 4’s performance to analyze because it
demon-strates the fact that grouping data across behaviors within a
single partici-pant can conceal true effects; just as effects can
be concealed by groupingthe performance data of several individual
participants (as in group re-search). To clarify, we consider his
performance relative to the accuracyof his estimations in more
detail. Although his estimations were the leastaccurate and his
overall performance changed the least of all participants,the
accuracy of participant 4’s self-estimations did not seem to
systemati-cally vary with his performance improvement. The smallest
discrepancybetween his self-estimations and experimenter data
occurred for busstopping position with an average difference of 4%.
However, this per-formance improved by only 2% over baseline levels
(see Figure 5). Incontrast, his estimations differed from
experimenter data by 25% forcomplete stop, where he realized his
greatest improvement (19% overbaseline levels). Another large
discrepancy occurred for two seconds mo-tionless, which improved by
only 5%, where his self-estimations were53% lower than experimenter
data. If we were to look only at his overallbehavior change and his
overall estimations (i.e., data grouped across be-haviors), the
data would suggest a clear relationship between his estima-tions
and the resulting behavior change (i.e., he had the smallest
overalleffect size and reported the least accurate data),
especially in light of par-ticipant 1’s results discussed above
(i.e., who reported the most accuratedata and had the largest
overall effect size). However, as one can see fromthe more detailed
analysis that considers each behavior singly, the rela-tionship
between estimations and behavior change is far from clear. It
islogical that accuracy of self-estimations would affect the degree
to whichbehavior changes. However, the degree of agreement between
self-moni-toring and experimenter data in our study did not predict
the degree of im-provement for each participant on particular
performance targets.Accuracy is only estimated by assessing IOA
(Johnston & Pennypacker,1993). Therefore, with regard to the
importance of the accuracy ofself-monitoring data, we have only
scratched the surface of the topic.
FUTURE RESEARCH
There are many unanswered questions regarding applications
ofself-monitoring procedures to improve the performance of lone
work-
Experiment 35
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
ers. The most important question in terms of safety is the
effectivenessof these kinds of procedures at reducing accidents
and/or injuries.Krause (1997) reported a 66% reduction in accidents
and injuries in hisapplication with bus drivers, but methodological
issues prevent draw-ing firm conclusions about the degree of
behavior change generated bythe self-monitoring procedure. While we
were able to successfullybuild upon the consultation effort
reported by Krause (1997) by experi-mentally evaluating behavior
change, we were not able to answer ques-tions about accident/injury
reduction. In order for researchers to drawconclusions about
accident/injury reduction, it may be necessary topartner with
companies or consulting firms implementing behav-ior-based safety
processes with lone workers on a large scale. Academicparties could
ensure reliable assessment of behavior change and compa-nies or
consultation firms could ensure a large-scale implementationwith a
long duration that could impact accident/injury rates. An
addi-tional question that still remains unanswered is the extent to
which ac-curacy of self-monitoring influences its effectiveness
(McCann &Sulzer-Azaroff, 1996). Such research could begin by
training partici-pants to accuracy at the beginning of a study and
then assessing drift andconcurrent performance levels with
confederate observers or someother unobtrusive measurement system
over time. Some electronicforms of driving performance measurement
are now becoming avail-able that might be of use in such research.
The issues discussed aboveare excellent research questions,
however, both require that behaviorchange be produced reliably by a
self-monitoring procedure. It is hopedthat the successes and
failures of the current study will inform the devel-opment of
self-monitoring procedures that can reliably produce sub-stantial
changes in safe behavior.
Several methodological improvements are needed in order for
exten-sions and/or replications of this work to (a) more clearly
assess behaviorchange and (b) generate greater behavior change.
First, baseline and in-tervention phases of longer duration would
allow researchers to discernbehavioral effects more reliably. This
is especially relevant for the thirdphase of the intervention that
targeted mirror check and bus stoppingposition, which lasted for
only a few sessions. Even if methodologicalimprovements indicate no
effect for any one dependent measure, thisenables researchers to
draw stronger conclusions about the “bound-aries” of the
effectiveness of an intervention. In other words, research-ers
could then explore questions about why some performancesimproved
while others did not. Research of this kind would be costly interms
of labor and time, but would enable researchers to draw
stronger
36 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
conclusions about behavioral changes. Second, potentially more
pow-erful intervention variables should be tested. An obvious
choice wouldbe the addition of valuable consequences that were
dependent uponsome aspect of driver performance (e.g.,
participation or performancelevels as assessed by observers). In
addition to adding more powerfulconsequences to such interventions,
we feel that there are at least twoless costly or less intrusive
conditions that might add power to the inter-vention: these are (a)
participants selecting performance targets of per-sonal value or
interest and (b) evaluating an implementation that wasperceived as
permanent rather than temporary. We discuss these two is-sues, as
well as research that might reveal the behavioral functions
ofstimuli generated by self-monitoring procedures, below.
A key component missing from the current study was the absence
ofemployee participation in the design stages of the project and
other ac-tivities said to generate “buy in.” Both Krause (1997) and
McSween(1995) heavily promote employee participation in BBS
processes. As-pects of such employee involvement may function as
motivational vari-ables where the value of consequences related to
safety improvement isincreased and behaviors correlated with those
improvements are morefrequently evoked. When behavior is analyzed
on a molecular scale(i.e., behavior is analyzed in terms of its
immediate antecedents andconsequences), Michael’s (1993) taxonomy
of CEOs may be relevant.
Participant comments on the debriefing survey suggest that
employ-ees may make the greatest improvements when they “value” the
targetperformance. Technically, values can be defined as a set or
constella-tion of conditioned reinforcers (Malott, R. Malott, M.,
& Trojan, 1999).Participant 2 improved complete stops by 41%
and his survey com-ments regarding the self-monitoring process
emphasized this specifictarget performance. He wrote, “Complete
stops are important. A lot canhappen in a short amount of time at
an intersection. Really have to stopcompletely to see the whole
picture.” These results and self-report com-ments concur with
McSween (1995), who suggested that learning expe-riences prior to
the onset of a BBS initiative may be importantstrategically and
that individuals’ values should be incorporated intoperformance
improvement initiatives. Future research should examinemore closely
this potential relationship between employee “buy-in” ac-tivities
and the effectiveness of and compliance with treatment
proto-cols.
Whether self-monitoring procedures with lone workers tend to
func-tion primarily as antecedents, as does a prompt for safety
belt use, is aninteresting research question. Researchers have
suggested that self-moni-
Experiment 37
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
toring is effective as a result of an individual applying
consequences con-tingent upon his or her own behavior (Kanfer,
1970). Future researchcould explore this question by requiring the
self-monitoring to take placeeither just before (antecedent
function), or just after (consequence func-tion) a work shift
and/or by measuring the use of self-delivered conse-quences through
talk-aloud procedures. If self-monitoring tends tofunction
primarily as an antecedent, practitioners should emphasize the
ad-dition of programmed reinforcement to ensure prolonged
effectiveness.Future researchers should also consider the clinical
literature on self-moni-toring. For example, some clinical research
suggests that the power ofself-monitoring procedures is enhanced
when participants monitor the fre-quency of undesired, rather than
desired, performance (Kopp, 1988).
DISCUSSION AND CONCLUSION
Strengths of the current study include measures of independent
vari-able integrity, collection of self-report measures at the
conclusion of thestudy, and supervisor probes. In some cases IV
integrity measures pro-vided insight into unusual patterns in the
data. The survey instrumentgave participants a chance to express
their opinions about aspects of theprocess, and gave us a chance to
collect information about covert be-havior that may have impacted
their performance. Supervisor probes af-fected performance
systematically across all participants and mayrepresent one method
for assessing participants’ understanding of thetarget
performances. For some participants the probes demonstratedthat
they understood and were capable of performing the target
behav-iors at high percent safe levels. For participant 2, however,
the probeshowed that he may not have understood or discriminated
certain targetbehaviors. His performance improved when the
supervisor was present,but only to 55% and 64% for 2 seconds
motionless and mirror check, re-spectively. In general, the probes
demonstrated that supervisor presencewas a more powerful
intervention than self-monitoring, and that partici-pants improved
only the behaviors that supervisors were observing.
Weaknesses of the current study include the relatively short
durationof the intervention, the absence of meaningful outcome
measures (alsodue to the short duration), the small number of
participants, the lack ofemployee “buy-in,” and the small to
moderate effect size of the interven-tion. The study was cut short
because the particular bus route terminatedfor summer break, so we
could not determine whether performancechanges maintained,
improved, or deteriorated over time. As mentioned
38 JOURNAL OF ORGANIZATIONAL BEHAVIOR MANAGEMENT
Dow
nloa
ded
by [
Wes
tern
Mic
higa
n U
nive
rsity
] at
11:
35 1
0 Ja
nuar
y 20
12
-
previously, future researchers should consider methods that
provide moretime for the stabilization of performance under each
experimental condition.In addition to concerns about the ability to
assess behavior change, a reduc-tion in collisions or injuries is
not possible unless more participants are in-volved over long
periods of time. With regard t