JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR VOLUME 5, NUMBER 2 APRIL, 1962

SOME EFFECTS OF RESPONSE COST UPON

HUMAN OPERANT BEHAVIOR
HAROLD WEINER
SAINT ELIZABETHS HOSPITAL, DEPARTMENT OF HEALTH, EDUCATION,
AND WELFARE, WASHINGTON 20, D.C.1' 2

Three experiments are reported which investigated the effects of cost (point loss per response)
upon human-observer responses maintained by VI and Fl schedules of reinforcement (acquisi-
tion of points via critical-signal detections). (I) Cost attenuated VI response rates without sub-
stantially disturbing the constancy of responding, regardless of the presentation sequence of
the no-cost and cost conditions. (II) Fl scalloping appeared only under cost conditions. Under
no cost, a constant rate of responding (similar to VI performance) characterized inter-rein-
forcement intervals. Exposure to cost did not prevent the recovery of previously established no-
cost baselines. (III) FI irregularities, analogous to those commonly observed under Fl rein-
forcement schedules, may be produced by different temporal presentations of the no-cost and
cost conditions.
The results of all three experiments emphasize the importance of cost as a factor in the
maintenance of human behavior on schedules of positive reinforcement.

The human organism seldom produces posi- responses. Holland (1958) has shown that
tive reinforcement without some response relatively constant inter-reinforcement rates of
"cost"(e.g., physical, monetary). Relatively few observer responses are emitted when critical
studies of operant behavior, however, have signals (reinforcements) are programed on VI
systematically investigated response cost as schedules. The frequency of observer responses
a determinant of the human organism's was found to decrease as signal rate decreased.
adjustment to the contingencies of positive Unlike Holland's study, the present data were
reinforcement. obtained when observer responses were emit-
The purpose of the experiments reported ted under two conditions of schedule informa-
here was to assess the effects of response- tion and under explicitly programed response-
produced cost (point loss per response) upon cost contingencies in which reinforcement was
human operant behavior maintained by VI not only signal detections but also concomitant
and Fl schedules of positive reinforcement 100-point rewards per detection.
(acquisition of points via critical-signal detec-
tions). The concurrent reinforcers chosen Method
can be quantified precisely along qualitatively Subjects. The subjects were four male
comparable reinforcement dimensions, and humans, aged 18-21 who were paid $1.25 per
are particularly germane to the human hour for their services.
situation, Apparatus and Task. The Ss sat alone in
an experimental cubicle facing a display
which consisted of a series of 40 lamp baffles
EXPERIMENT I: RESPONSE COST arranged in three concentric circles behind a
AND VI PERFORMANCE frosted-glass screen. Each baffle contained a
This experiment was concerned with the red light bulb. Switching circuitry was used
effects of concurrent schedules of response cost to program presentation schedules of red
and VI reinforcement upon human observer (critical) signals and to randomize their
"The views expressed here are those of the author spatial location on the frosted-glass screen.
and do not necessarily reflect the opinion of Saint Eliza- The Ss' 1-hr task was to monitor the display
beths Hospital. for red signals. The red signals were non-
'This study was supported by Grant M-5600 from the
National Institute of Mental Health, U.S.P.H.S. The transient, i.e., they remained available until
data were collected at the American Institute for Re- detected. Observation of the display was made
search, Washington, D.C. contingent upon the pressing of an observa-
201
202 HAROLD WEINER
tion lever (with a force of 20 g through a sence of a yellow light; a 20-min VI 3 schedule
distance of 1 cm). By pressing this lever presented in the presence of a green light; a
(observer response), the Ss could observe the 20-min VI 9 schedule presented in the pre-
display for only 0.3 sec, even when the lever sence of a blue light. The Ss were exposed to
was held down. Thus, each successive "look" three cycles (i.e., three daily sessions) of this
at the display required the release and re- multiple schedule.
depression of the observation lever. The three-component multiple schedule
If scheduled, a red signal would appear on was then subdivided into a six-component
the display (0.3-sec duration) after a single multiple schedule by introducing either a
observer response. If no red signal was 10-min cost or a 10-min no-cost period, under
scheduled, the display remained blank after the control of the on-off white light, during
an observer response. each 20-min component. The Ss initially con-
When a red signal was detected, the Ss ditioned under the no-cost condition were
were required to "report" it by pressing exposed to the cost condition and vice versa.
another lever. A reinforcement bell sounded; The components of the six-component sched-
and a Veeder-Root digital counter, mounted ule for the Ss initially conditioned under no
in front of the Ss, registered 100 points each cost were presented in a fixed order: a 10-min,
time a red signal was detected and reported no-cost VI 1 schedule; a 10-min cost VI 1 sched-
correctly. This 100-point reinforcement pro- ule; a 10-min no-cost VI 3 schedule; a 10-min
vided feedback to the Ss on the day-to-day cost VI 3 schedule; a 10-min, no-cost VI 9
quality of their performance. schedule; a 10-min cost VI 9 schedule. The
Observer responses were recorded contin- presentation order of components for the Ss
uously on a Gerbrands cumulative recorder. initially conditioned under cost was also fixed
Procedure. The observer responses of each and was'identical except that the presentation
S were conditioned separately (1-hr sessions order of no cost and cost for each schedule
daily) under a 1-min variable-interval (VI 1), was reversed.
a 3-min variable-interval (VI 3) and a 9-min The Ss were conditioned for 3 hr on the
variable-interval (VI 9) presentation schedule six-component multiple schedule. The colored
of red signals, in that order. Subjects S15 and lights of the three-component multiple sched-
S12 were conditioned with schedule informa- ule were associated with the same VI schedules
tion under no-cost and cost conditions, re- under the six-component multiple schedule.
spectively. Subjects S34 and S48 were The on-off white light and the auditory-visual
conditioned without schedule information cues produced by the action of the counter
under no cost and cost, respectively. Under provided information on the differential cost-
the no-cost condition, the observer responses to-payoff contingencies.
of the Ss were emitted without loss of points Instructions. At the beginning of the first
(cost/payoff = 0/100). In the cost condition, hour of the conditioning (under the VI 1
the counter subtracted one point from the schedule), instructions were read to the
S's score for each observer response (cost/pay- two subjects (S15 and S12) given schedule
off= 1/100). information:
For all Ss, each of the VI schedules was
associated with a different stimulus light. The "This is a vigilance task. Your job it to
no-cost and cost conditions were also associ- detect red signals that will appear on the
ated with different environmental stimuli. A screen every once in awhile. You will not
white bulb was lit during the cost condition, be able to see these signals unless you press
and it was extinguished during the no-cost and release this lever (indicating the obser-
condition. vation lever). Holding down this lever will
After 3 hr of conditioning on each of the not enable you to see the scope contin-
three VI schedules, the schedules were pre- uously. Each time you wish to take a look
sented as a three-component multiple schedule at the scope to search for red signals, you
under either no-cost or cost conditions. The must re-press this lever and re-release it.
components of this schedule for both no cost When you detect a red signal, report it as
and cost were presented in a fixed order: a rapidly as you can by pressing this other
20-min VI 1 schedule presented in the pre- lever (indicating the report lever) and then
 j~ Ihu
SOME EFFECTS OF RESPONSE COST 203
releasing it. Press the report lever only after lower than their respective no-cost rates. The
you detect a red signal. Each time you response suppression produced by cost did
correctly detect and report a red signal, a not markedly disrupt the constancy of re-
bell will ring and the counter, which is sponding between reinforcements which char-
placed before you, will register 100 points. acterized no-cost performance. These effects
This is your reward for each red-signal are analogous to those occurring when re-
detection. Your task for the next hour is to sponse-produced shock punishment is added
detect as many red signals, that is, score to the food-rewarded VI performance of
as many points, as possible. pigeons (Azrin, 1960). Unlike Azrin's data,
You will notice that there are three however, no evidence could be found in the
differently colored light bulbs mounted present data to suggest any recovery from the
on the wall. Throughout this session, the effects of cost, either within or between ex-
yellow light will be lit. Whenever the perimental sessions. Furthermore, the re-
yellow light is on, it means that the red introduction of the no-cost condition did not
signals will appear 60 times during the produce any transitory enhancement of pre-
session. Their occurrence, however, will viously established cost baselines.
be unpredictable; that is, two successive red
signals can appear anywhere from less than
a second apart to as long as 2 min apart.
On the average, however, a red signal will
appear once a minute or 60 times during
the next hour." A

Additional brief instructions were given to

the "schedule information" subjects before
their first exposure to the VI 3 and VI 9 sched- Suhiec S15
ules. As in the instructions for the VI 1 sched-
ule, they were informed that the red signals NO COST NOv
CO1 9
would appear unpredictably, either on the
average of once in 3 min (for the VI 3 sched-
ule) or once in 9 min (for the VI 9 schedule).
They were also told that for both the VI 3
and VI 9 schedules, the inter-signal range
was less than a second to twice their respec-
tive mean inter-signal interval. They were
given no other instructions.
The two Ss who did not receive schedule
information were given only the first para- Fig. 1. Effects of cost, either with schedule informa-
graph of the instructions above. tion (S15) or without schedule information (S34), upon
several variable-interval performances conditioned un-
der no cost. Vertical marks on the cumulative-response
Results and Discussion curve indicate the occurrence of 100-point rewards.
Figures 1 and 2 present the response rates
during the third hour of conditioning under With or without schedule information, no-
the six-component multiple schedules. Figure cost and cost response rates decreased (except
1 shows the effects of cost upon VI 1, VI 3, and for the VI 9 component under cost in Fig. IB)
VI 9 performance, with and without schedule as the rate of VI reinforcement decreased, re-
information, previously maintained under no- gardless of the conditioning sequence of no
cost conditions. In Fig. 2, response rates were cost and cost. Holland (1958) had previously
conditioned, with and without schedule in- demonstrated this relationship under no-cost
formation, under cost before the introduction conditions.
of the no-cost condition. The response rates on the VI schedules
Inspection of these figures reveals that for which had been conditioned separately, under
each VI component, with and without sched- either no cost or cost, were maintained under
ule information, cost-response rates were the three-component multiple schedule. This
204 HAROLD WEINER
was not true for the six-component multiple even in the early phases of training. Skinner
schedule. The introduction of a 10-min cost and Morse (1958) have noted the same general
period within each no-cost VI component finding in a study of wheel-running behavior
(Fig. 1) consistently reduced the no-cost re- of rats under Fl reinforcement. Deceleration
sponse rates of the three-component multiple in running before reinforcement (i.e., S-
schedule. The introduction of a 10-min no-cost shaped inter-reinforcement curves) was re-
period within each cost VI component (Fig. 2) moved (producing typical Fl scalloping) when
wheel friction was increased.
The purpose of the present experiment was
a further analysis of response cost as a factor
determining the emission of "typical" Fl
performance.
Method
Subjects. The subjects were four male
humans, aged 17-21, who were paid $1.25 per
hour for their services.
Su
Sj
Apparatus and Task. The same apparatus
and task were used as in Experiment I.
Procedure. The observer responses of the
Ss were conditioned without cost for an hour
daily under an Fl 1 presentation schedule of
red signals (and 100-point rewards per de-
tection) until their response rates stabilized.
Following stabilization of response rates
under no-cost conditions, two 30-min periods
of cost were introduced, involving a loss of 1
&Nod point per response. This was followed by the
reintroduction of the no-cost condition for a
Fig. 2. Effects of no cost, either with schedule infor- 30-min period.
mation (S12) or without schedule information (S48),
upon several variable-interval performances conditioned A yellow light remained lit during each
under cost. Vertical marks on the cumulative-response experimental session. This light provided a
curve indicate the occurrence of 100-point rewards. discriminative stimulus (SD) for the Ss, i.e.,
responses were reinforced only in its presence.
slightly increased (except for the VI 1 com- In addition, the yellow light indicated that
ponent under cost in Fig. 2B) the cost re- the presentation of red signals had not been
sponse rates previously established in the changed either within a session or from
three-component schedule. session to session. Pretrials indicated that some
Ss appeared to suspect the change in the sched-
ule of reinforcement, so that they emitted
EXPERIMENT II: RESPONSE COST AND erratic exploratory behavior. Because rapid
"TYPICAL" Fl PERFORMANCE response stabilization was desired, the yellow
In a recent study, Azrin (1958, p. 184) re- light was added to suggest to the Ss that
ported that when human Ss were required to comparable conditions prevailed from day to
exert little physical effort (i.e., 15 g of force day.
through a distance of approximately 2 mm) in The no-cost and cost conditions were associ-
order to emit an observing (button-pressing) ated with the same SDS (i.e., the on-off white
operant, ". . the rate of response usually was
. light) as in Experiment I. These associations
maintained at several responses per second were maintained throughout the experiment
with little or no scallops during prolonged ex- to provide the Ss with information on cost
posure to a fixed-interval schedule of rein- contingencies.
forcement." When a heavier button was Instructions. All subjects received the same
substituted, reduced response rates and typi- instructions as the no-schedule information
cal Fl scalloping characterized performance, subjects (S34 and S48) in Experiment I.
 SOME EFFECTS OF RESPONSE COST 205
INO COS COST

I,
1
Results and Discussion
All Ss exhibited a reasonably stable Fl 1
performance under no-cost conditions by the
end of the fifteenth hour of conditioning.
A I

Table 1 reveals the negligible difference be-

tween the average no-cost response rates = t -i
h~~~~~~~~ =
=:
emitted by two of the four Ss during Sessions il alo

12-15. Comparable differences in response Subjoot 8O 10th hour of ooditioblng

rates during these sessions were obtained from
the other two Ss. A constant inter-reinforce- COST NO COST
ment rate of responding characterized the
performances of all Ss during these sessions.
Table 1
Average No-cost Response Rates in Responses B
per Minute Before the Introduction of Cost
§.1 /'A4}/ /7!j
1~~~~~~I
__________lhou
Conditioning
Session S 10 S 45
12 229.8 60.9 uum3O 17o riumrat_ofom4-aft
13 21%3 60.0 Fig. 3. Cumulative-response records of S1O showing
14 215.0 59.3 the effects of no cost and cost upon instrumental ob-
server responses conditioned by an Fl 1 schedule of re-
15 233.7 58.5 inforcement. Vertical marks on the cumulative-response
curve indicate the occurrence of 100-point rewards.
Figures 3 and 4 show the effects of cost upon
Fl 1 performance conditioned under no cost. rates during the sixteenth hour of condition-
These figures present cumulative-response ing. Low response rates and flat scalloping
curves for two of the four Ss tested under again characterized performance. The rein-
both no cost and cost during the sixteenth
and seventeenth hours of conditioning. The NO COST COST aI
two other Ss that were tested showed similar
effects.
Inspection of Fig. 3 and 4 reveals that for
all Ss the no-cost response rates and response A
patterns were similar to their respective rates
and patterns during Sessions 12-15. Without
cost, observer responses did not exhibit ________ _hUr
scalloping between reinforcements. The con- Subjeot 846 Sth hour of oonditlonig
stancy of responding between reinforcements
resembled variable-interval (VI) performance,
not Fl performance. 1- COST NO COST
The introduction of the cost contingency A
midway through the sixteenth hour of con-
ditioning markedly reduced response rates
and produced fairly immediate flat scalloping B
between reinforcements. The magnitude of
the cost effects appeared somewhat unrelated
to no-cost baselines of responding. Exposure ______ _I hour
to cost produced scalloping and extremely
low response rates regardless of established ubjoot 417t hour of oouditong
no-cost baselines. Fig. 4. Cumulative-response records of S45 showing
Response rates under cost during the the effects of no cost and cost upon instrumental ob-
server responses conditioned by an FI 1 schedule of re-
seventeenth hour of conditioning (Fig. 3B and inforcement. Vertical marks on the cumulative-response
4B) were similar in general to cost response curve indicate the occurrence of 100-point rewards.
206 HAROLD WEINER
troduction of the no-cost condition during the The present experiment attempted to dem-
seventeenth hour of conditioning rapidly in- onstrate that several different types of Fl
creased response rates for all Ss to their re- response patterns may be produced by a
spective sixteenth-hour no-cost levels and judicious scheduling of the no-cost and cost
removed the flat scalloping patterns emitted conditions.
under the cost condition. Apparently, ex-
posure to cost did not produce a permanent Method
modification in no-cost response rates and Two of the Ss tested in Experiment II were
response patterns. given additional Fl 1 conditioning sessions
Differences in response cost may possibly under different presentation sequences of the
account for some of the inconsistency between no-cost and cost conditions in order: alter-
the findings of Holland (1957, 1958) and Blair nating inter-reinforcement intervals (1-min)
(1958). Studies by Holland demonstrated that of no cost and cost; alternating 30-sec periods
the mere detection of critical signals by of cost and no cost between reinforcements in
human observers can be used as a positive which the cost condition always followed rein-
reinforcer for instrumental observer re- forcement; alternating 30-sec periods of no
sponses. The Fl presentation schedules of criti- cost and cost between reinforcements in which
cal signals produced observer-response scallops the no-cost condition always followed rein-
characteristic of the Fl lever-pressing and key- forcement; alternating 15-sec periods of no
pecking performances of infrahuman organ- cost and cost between reinforcements in
isms controlled by food reinforcement. which the no-cost condition always followed
Holland's results demonstrate: (a) the con- reinforcement; alternating 15-sec periods of
sistency of response patterns within and cost and no cost between reinforcements in
among human Ss under Fl reinforcement which the cost condition always followed
schedules, and (b) the remarkable similarity reinforcement.
in the behavior emitted by human and infra- After 2 hr of Fl 1 conditioning on each
human organisms when both are exposed to of these five response-cost schedules, the sched-
Fl reinforcement contingencies. However, the ules were presented as a seven-component
results of experiments by Blair indicate that multiple schedule (Ferster & Skinner, 1957).
this behavior shows wide individual differ- Each of the five response-cost schedules was
ences. In a similar task and under similar Fl presented for 10 min (in the same fixed order
schedules, Blair's subjects did not always show as the individual conditioning sessions), fol-
Fl scalloping. lowing a 10-min no-cost period and a 10-min
In the Blair experiments, a light source was cost period within a single experimental
attached to the subject's head. Critical signals session. Performances under all response-cost
could be seen only when the subject's head schedules were maintained by concurrent Fl 1
was properly positioned for direct scanning positive reinforcements (critical-signal de-
of the display (observer response). On the tections and 100-point rewards per detection).
other hand, the observer responses in the As in Experiment I and Experiment II, a lit
Holland studies involved both appropriate white bulb was associated with the cost con-
positioning of the head (like Blair's observer dition. The no-cost condition was in effect
response) and the pressing of a key. This when the white light was off. The Ss were
additional observer-response component may exposed to two cycles (i.e., two daily 70-min
have increased the "physical cost" of respond- sessions) of the seven-component multiple
ing and may have contributed to the develop- schedule.
ment of Fl scalloping. This hypothesis is con-
sistent with the findings of Azrin (1958) Results and Discussion
discussed previously. Figure 5 presents the second-cycle perfor-
mance under the seven-component multiple
EXPERIMENT III: RESPONSE-COST schedule for one of the Ss (S45) tested. Similar
CONTINGENCIES AND "DEVIANT" effects were obtained from the second S.
Fl PERFORMANCE Examples analogous to many of the commonly
In Experiment II, Fl scalloping was con- observed FI irregularities can be found within
tingent upon the existence of response cost. and among the various components of the
 SOME EFFECTS OF RESPONSE COST 207
multiple schedule. Subject S45's previously As in the no-cost response pattern, the flat
established (Experiment II) Fl 1 performances scalloping emitted under cost may be con-
under no-cost and cost conditions were rep- sidered an Fl irregularity when compared to
licated (components 1 and 2 in this record), the deep scalloping usually obtained from
even when cost contingencies were changed infrahuman organisms. Such deep scalloping
after each reinforcement (component 3). has been effected in component 4, by program-
The constancy of responding between re- ing alternately the cost and no-cost conditions,
inforcements under the no-cost condition in that order, every 30 sec.
(component 1) represented a failure of S45 The record shows instances of single and
to respond to the fixed temporal aspects of multiple bites or runs (components 4 and 7),
Fl reinforcement. This pattern of responding single and multiple knees (components 5 and
has been commonly observed under VI sched- 6), double scallops (e.g., at "a" between com-
ules of reinforcement, in which the occurrence ponents 3 and 4), and "running through" rein-
of individual reinforcements is unpredictable. forcements (e.g., at "b" between components 4
On the other hand, S45's cost performance and 5). Although triple scallops are not
(component 2) was a maximally efficient produced in this record, they might have
adjustment to the Fl 1 pattern of reinforce- rather easily been effected by varying the
ment. Positive reinforcements were procured sequence of no-cost and cost conditions.
with minimum response expenditure, in- Negative accelerations (failures to sustain the
dicating highly accurate temporal discrimi- terminal rate) also do not appear in the re-
nations of the inter-reinforcement interval. cor(l, but could probably have been produced
The flat scalloping which characterized cost by introducing intermittent cost between the
performance is similar to highly efficient DRL no-cost and cost sequence within a single
performance. In a DRL schedule, the organism inter-reinforcement interval.
is required to space his responses temporally There appears to be little question that
in order to obtain positive reinforcements. As characteristics of Fl performance may be con-
here, there is a "cost" (loss of positive rein- tingent upon response cost. Whether or not
forcement) in responding inappropriately to specific instances of irregular (either intra-
the temporal contingencies of positive rein- subject or intersubject) Fl patterns in fact
forcement under a DRL schedule. develop as a function of response-cost con-

0)

Fig. 5. Deviant FI 1 performance in a seven-component multiple schedule produced by different presentation

sequences of the no-cost and cost conditions. The temporal presentation schedules of the no-cost and cost condi-
tions are indicated in the parentheses to the right of each condition. The components of the multiple schedule are
numbered in the parentheses above the response-cost contingencies and their associated presentation schedules.
Vertical marks on the cumulative-response curve indicate the occurrence of 100-point rewards.
208 HAROLD WEINER
tingencies is problematical. The important Azrin, N. H. Effects of punishment intensity during
point is that response cost must be attended variable-interval reinforcement. J. exp. Anal. Behav.,
to (even when not explicitly programed) as 1960, 3, 123-142.
Blair, W. C. Measurement of observing in human
a possible determinant of operant behavior. monitoring. Science, 1958, 128, 255-256.
Attempts to account for Fl responding in Ferster, C. B., and Skinner, B. F. Schedules of rein-
terms of only the contingencies of positive . forcement. New York: Appleton-Century-Crofts, 1957.

reinforcement are parsimonious, but they may Holland, J. G. Technique for behavioral analysis for
human observing. Science, 1957, 125, 348-350.
be inadequate to provide a complete speci- Holland, J. G. Human vigilance. Science, 1958, 128,
fication of the controlling factors. 61-63.
Skinner, B. F., and Morse, W. H. Fixed-interval rein-
forcement of running in a wheel. J. exp. Anal.
REFERENCES Behav., 1958, 1, 371-379.
Azrin, N. H. Some effects of noise on human behavior. Received July 12,1961
J. exp. Anal. Behav., 1958, 1, 183-200.