M9 Project

zanuyekeh

ArticleM9.pdf

Home >Applied Sciences homework help >M9 Project

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

CONTINGENCY-SHAPED AND RULE-GOVERNED BEHAVIOR: INSTRUCTIONAL CONTROL

OF HUMAN LOSS AVOIDANCE MARK GALIZIO1

UNIVERSITY OF NORTH CAROLINA AT WILMINGTON

Instructions can override the influence of programmed schedules of reinforcement. Al- though this finding has been interpreted as a limitation of reinforcement schedule control in humans, an alternative approach considers instructional control, itself, as a phenomenon determined by subjects' reinforcement histories. This approach was supported in a series of experiments that studied instructional and schedule control when instructions either did or did not accord with the schedule of reinforcement. Experimenit I demonstrated that accurate instructions control discriminative performances on multiple avoidance schedules, and that such control persists in a novel discrimination. Experiments II and III showed that elimination of instruction-following occurs when inaccurate instructions cause sub- jects to contact a monetary loss contingency. Experiment IV demonstrated the reinforcing properties of accurate instructions. Skinner's view of rule-governed behavior is consistent with these findings, and can be extended to account for many aspects of instructional control of human operant behavior. Key words: instructions avoidance, stimulus control, monetary loss, observing be-

havior, rule-governed behavior, adult humans

Several experiments have demonstrated that instructions may facilitate the development of schedule control (Baron, Kaufman, and Stau- ber, 1969; Turner and Solomon, 1962; Wiener, 1962). For example, Ayllon and Azrin (1964), in an experiment with institutionalized sub- jects, were unable to obtain schedule control until they gave instructions about the desired response. The instructions were effective only when the response they specified was rein- forced. A different result was obtained by Kaufman,

Baron, and Kopp (1966) in a laboratory set-

'The experiments reported were part of a dissertation submitted to the University of Wisconsin-Milwaukee in partial fulfillment for the requirements of the doctoral degree. I am indebted to Alan Baron, chairman of my committee, for his support and advice. I would like to thank Vince Adesso, Robert T. Brown, Eugene Eisman, Catherine Galizio, Robert Lowman, and Jessica Wirth for their helpful comments. The research was supported in part by the Graduate School and the College of Let- ters and Science at the University of Wisconsin-Mil- waukee. Thanks are due Linda Gleixner, Alan Kolski, Cindy Krieger, Chris Scallon, Craig Wassenburg, and Dale Wilson, who assisted in data collection. Experi- ments I and II were presented at the Midwest As- sociation of Behavior Analysis Annual Convention, May, 1977, Chicago. Reprints may be obtained from Mark Galizio, Department of Psychology, University of North Carolina at Wilmington, Wilmington, North Carolina 28401.

ting. Subjects were exposed to a variable-inter- val (VI) schedule of monetary reinforcement and one group was given the accurate informa- tion that money would be delivered on a VI basis. Other groups were given inaccurate schedule instructions: either that a variable- ratio (VR) or a fixed-interval (FI) schedule would be in effect. The inaccurate instructions exerted substantial control during a 3-hr period. The effects of the schedule instructions were just what would be expected from sub- jects who had been exposed for many sessions to the specified schedules, in spite of the fact that the VI schedule was programmed for all subjects. In a subsequent experiment, Kauf- man et al. (1966) examined the effects of false schedule instructions when reinforcement was unavailable throughout the 3-hr period. Re- sponding weakened, but was not completely eliminated during the extinction session. Thus, Kaufman et al. found that instructional control was capable of overriding the contin- gencies, a conclusion confirmed in a similar experiment by Lippman and Meyer (1967).

Several features of the procedures used by Ayllon and Azrin make direct comparison with the Kaufman et al. and the Lippman and Meyer experiments difficult. These latter two studies were laboratory experiments conducted with college students; Ayllon and Azrin

1979, 31, 53-70 NUMBER I (JANUARY)

MARK GALIZIO

worked with institutionalized patients. Fur- ther, the instructions in the laboratory experi- ments always specified that reinforcement was dependent on responding (although some- times, in fact, it was not); instructions in the Ayllon and Azrin study did not directly specify a response-outcome dependency. Another, per- haps more important, difference was that the laboratory experiments involved only a single, brief session, and as Kaufman et al. pointed out, instructional control of responding may weaken, given sufficient exposure to the pro- grammed contingencies. The above findings demonstrating the im-

portance of instructional control have gen- erated a theoretical controversy. One approach is represented by accounts of instructional con- trol that emphasize constructs within the do- main of a behavioral analysis. Thus, several accounts treat instructions as a form of dis- criminative stimuli that are associated with characteristic response patterns (Schutte and Hopkins, 1970; Skinner, 1957). Other theorists have focused on the results of the Kaufman et al. research, and have interpreted these find- ings as showing a limitation of reinforcement control in humans. For example, Bandura (1971, 1974) considered instructional variables as vehicles of vicarious reinforcement, similar in principle to modelling effects, and capable of inducing expectancies that influence subse- quent behavior. Some theorists have gone to the extent of taking the instructional control literature as supporting the claim that operant conditioning has not been demonstrated in adult humans (Brewer, 1974). The nonbehavioral approaches to instruc-

tional control have stressed apparent weak- nesses in the discriminative stimulus account. For example, Dulany (1968) argued that to provide an explanation, discriminative stimuli must be "defined by a history in which a stimulus gains control of a response by selec- tive reinforcement of that response in its presence-not by any demonstration of stimu- lus control." Dulany's point is simply that the instructions-as-discriminative-stimulus hypoth- esis is ad hoc. The criticism can be answered in principle by demonstrating that differential reinforcement does influence instructional con- trol. Dulany notes further that "one can easily demonstrate that a subject will respond selec- tively to an instruction he has never heard be- fore, much less been selectively reinforced for

responding to" (pp. 365-366, note a). Dulany's second criticism, the issue of control by novel stimuli, is applicable only to a simplistic model of discrimination learning. A number of fre- quently studied paradigms have extended mod- els of discrimination formation. In particular, the matching-to-sample and oddity paradigms demonstrate that a novel stimulus can control behavior in a predictable, lawful way (Nevin, 1973). Accounts of these paradigms have been developed that do not require abandoning the principle of reinforcement. A fundamental as- pect of such accounts is the recognition that the controlling stimulus dimension can be represented as a rule or concept (Skinner, 1974; Urcuioli, 1977; Urcuioli and Nevin, 1975). This type of approach may be applied to an analysis of instructional control. For example, consider Skinner's (1974) dis-

tinction between contingency-shaped and rule- governed behavior. Skinner pointed out that rules can exert rapid control over behavior, and that a person following instructions may behave differently from a person who has been exposed to the contingencies described by the instructions. Skinner further suggested that following instructions, heeding warnings, and obeying laws, all examples of rule-governed behavior, are themselves influenced by their consequences. That is, adults have long histo- ries of conditioning, which presumably involve favorable consequences for following instruc- tions. Such an analysis leads to the hypothesis that instruction-following can be controlled by its consequences. This is a testable hypothesis: instruction-following should be influenced by reinforcement, subject to extinction, and should come under discriminative control.

This hypothesis can best be examined in the setting of the long-term operant experiment, where the effects of instructions can be assessed over many sessions. If inaccurate instructions were presented in such a setting, and these in- structions led to loss of reinforcement, then a general elimination of instruction-following should occur. Alternatively, if instructions from one source are accurate and from another source are inaccurate, a discrimination should be formed. Subjects should show instructional control with the former source, but only sched- ule control with the latter. Finally, since in- structions are viewed in the present analysis as a complex form of discriminative stimuli, sub- jects should respond to produce instructions,

INSTRUCTIONAL CONTROL

just as other discriminative stimuli are rein- forcing events. The present research is an at- tempt to examine these predictions using a free-operant avoidance baseline with monetary loss as the aversive event (Baron and Kauf- man, 1968).

EXPERIMENT I MULTIPLE SCHEDULES OF LOSS AVOIDANCE: THE EFFECTS OF ACCURATE INSTRUCTIONS

Before the questions involving inaccurate instructions can be assessed, the role of ac- curate instructions must be established. In Experiment I, performances generated by a simple multiple schedule were compared with performances when accurate schedule instruc- tions were added. Multiple schedules permit the simultaneous analysis of several different temporal parameters, with and without in- structions that specify those parameters.

METHOD Subjects Two male (CB and MB) and four female

(CH, DH, WH, and PR) students enrolled at the University of Wisconsin-Milwaukee par- ticipated. All were between the ages of 18 and 25 yr and had no more than introductory level exposure to psychology. They were recruited by a campus employment advertisement, which depicted the research as a part-time job in which payment depended on performance. Subjects were informed that average earnings were about $1.80 per 50-min session, but that they could earn as much as $2.00, or as little as nothing at all in a given session. All sub- jects signed contracts agreeing to remain in the experiment for a minimum of 75 sessions, or until dismissed, and participated in a simu- lated session before signing the contract. Once the contract was signed, the experiment began. Sessions were scheduled at a rate of eight to 12 per week, each 50 min in duration, with a maximum of four sessions per day.

Apparatus The subjects worked in a well-lighted,

sound-attenuated room, 1.8 m square. They sat facing a table holding a vertical panel contain- ing the lever manipulandum and an array of colored lights. Six lights were spaced evenly

across the top of the panel. The leftmost light was green, and indicated when the session was in progress. The rightmost light was red, and served as a signal of monetary loss. The onset of the red light was always accompanied by a tone (1000 Hz, 76 dB) presented through a speaker beneath the table to ensure that the losses were noted. The middle four lights were amber, and

served as the discriminative and instructional stimuli for the components of the multiple schedule. A cardboard label could be inserted above each light. The manipulandum was mounted on a shaft

protruding from the face of the panel just above the table. A 10-cm lever was attached to the shaft and covered with a rubber hand- grip. The lever rested in the vertical position, but could be rotated 900 in either direction. The avoidance response was a 450 clockwise lever turn, which required 23 N of force. Com- pletion of the response produced a 0.5-sec blue feedback light just to the right of the lever whenever a 450 turn was accomplished. Programming and recording equipment were located in an adjacent room, sound-insulated from the subject.

Procedure Preliminary orientation. Before the first ex-

perimental session, each subject read a written description of various aspects of the experi- mental situation. Subjects were informed that books and magazines, but not purses or packs, or writing materials could be brought into the work area. Further instructions described the functions of the session light, the loss light and tone, and the lever. They were told that each time the loss light came on, they would lose five cents from their earnings, which would otherwise amount to $2.00 during that session. Finally, they were informed that turning the lever to the right would postpone the onset of the loss light:

By turning the handle to the right, you can postpone the next presentation of the red light. Sometimes the response will postpone the red light for only 10 seconds, sometimes for longer, and there may be times when you do not need to turn the handle at all because no red lights are pro- grammed. Remember, all you have to do is to turn the handle to the right, and you

MARK GALIZIO

will postpone the red light for some period of time.

The four-ply baseline schedule. Throughout the experiment, the same baseline schedule was programmed for all subjects. The schedule was composed of three 12.5-min free-operant avoidance schedules, each with a different re- sponse-loss interval, and a fourth 12.5-min component in which no loss was programmed. On the three avoidance components, 1-sec dis- plays of the red loss light accompanied by the tone were scheduled every 10 sec (the loss-loss interval was always 10 sec). Each response post- poned the next loss for either 10, 30, or 60 sec, depending on the component. Each compo- nent was presented once per session in random order.

Discriminative and instructional stimuli. Four of the six subjects (CB, PR, DH, and CH) were first exposed to the four-ply schedule with no instructions. The signal lights were ar- ranged to provide a simple multiple schedule, and no information was added about the schedules signalled by any of the stimuli. The leftmost amber light was associated with the No-Loss component, the second light with the 10-sec response-loss interval, the third with the 30-sec interval, and the fourth with the 60-sec interval. The next phase of the experiment for Sub-

jects CB, PR, DH, and CH, involved the ad- dition of instructional labels to the multiple schedule. The schedules signalled by the four amber lights were shuffled so that the leftmost light signalled the 30-sec interval, the second light the 10-sec interval, the third the 60-sec interval, and the fourth, No Loss. Labelled cards were inserted above the four amber lights in this phase. The label for each light accu- rately described the temporal properties of that schedule. The label above the light signalling the 10-sec interval read: "10 SEC", the label above the 30-sec and 60-sec schedule lights read "30 SEC", and "60 SEC", respectively. The label above the No-Loss schedule light read "NO LOSS".

In the third phase, lights and components were shuffled and the instruction labels with- drawn. Since Subject DH showed good dis- crimination of the component schedules with- out instructions, the second exposure without instructions was eliminated in her case. The procedures for the remaining two sub-

jects, MB and WH, were the same, except that the first condition was with instructions added to the multiple schedule. Then, the stimuli were shuffled and instructions withdrawn. Sub- ject WH went through this sequence twice.

Stability criteria. To determine when stable performance was reached, stability indices were calculated for rate of responding in each com- ponent (Sidman, 1960). The criterion to be met before changing conditions was as follows: the difference between the rate for the last two and the immediately preceding two sessions had to be less than 15% of the mean rate for the four sessions. Before advancing to a new condition, stability was required in each of the four components of the multiple schedule, with a maximum of 15 sessions in any com- ponent. When responding approached zero in the No-Loss component, and when conditions were repeated, visual inspection of the data de- termined when conditions could be changed.

RESULTS Response rate in each of the components

across sessions is presented in Figure 1 for Subjects DH, PR, CB, and CH, the four sub- jects whose initial exposure to the four-ply multiple schedule was without instructions. The first panel of Figure 1 shows the subjects' initial performances, and reveals that all sub- jects avoided regularly in their initial session, although rates were somewhat inconsistent and fluctuated from session to session in the early phases of training. Rates tended to stabilize after four to six sessions, although Subject CH did not reach the stability criterion by the fif- teenth session. The terminal performances of all four subjects revealed highly successful avoidance. During the last four sessions, Sub- ject DH received two losses, Subject CH, three losses, Subject PR, two, and Subject CB, none. Without instructions, only one subject, DH,

discriminated among all four components of the multiple schedule. By her third session, she showed a relatively high rate in the 10-sec com- ponent, a near-zero rate in the No-Loss com- ponent, and intermediate, appropriate rates in the 30- and 60-sec components. The other three subjects showed little indication of discrimina- tive control. The terminal performance of Sub- ject CB shows completely undifferentiated re- sponding in the four components, Subjects CH and PR showed some evidence of stimulus con- trol, emitting higher rates in the 10-sec compo-

INSTRUCTIONAL CONTROL

NO INSTRUCT

i I i I I I I I

INSTRUCT

I A I I

- -

NO INSTRUCT

..L....L.~I I-A&II I -i iiIII I I

lOSEC 30 SEC,A 60 SEC* DH

*~~If IA NO LOSSA II

l iii ai AAAAAAAAAAA lip

I I 9 A a a In I 0 ,IeA .C a IYt.V I 1' - 4 6 a 10 2 4 6 8 10

CONSECUTIVE SESSIONS Fig. 1. Responses per minute across the four components for conditions of Experiment I. Shown are the four

subjects whose initial exposure was without instructions.

nent than in the other three, but rates in the 30-sec, 60-sec, and No-Loss components were undifferentiated.

Figure 1 also shows the effects of introduc- ing the instructional labels. In the case of the

three subjects who had not previously shown complete discriminative control, PR, CB, and CH, the addition of instructions had major effects. Discrimination among all four com- ponents was evident within the first instruction

12 10 8 6 4 2

_j L I ^ n Jo I

a: LLI *XL U)

LLU U) z

C)LL.

8 6 4

E 5.

I0 8 6 4 2

a I a a I a A *- . .

s * * * . s II

dAa4)b-l ---...

-1

MARK GALIZIO

OH GB

I /

10 SEC

30 SEC

60 SEC

NO LOSS

10 SEC

7 3OSEC/000",~ ~ %,~~ - . bU b-L

NO LOSS Fig. 2. Cumulative records from Experiment I for Subjects CH and CB. Performances with no instructions (NI)

are shown in the top records; the lower records show performances with instructions (I). Pen deflection indicates the delivery of monetary loss.

session. Rates fell in the No-Loss component to near zero, while rates in the 30-sec and 60- sec components became well differentiated and more consistent with the programmed re- sponse-loss interval. Although rates, in general, were reduced with instructions, the actual number of losses was as in the earlier phase: DH=3, CH=4, CB=1, and PR=I total losses, for the last four instruction sessions. Cumulative records for Subjects CH and

CB (Figure 2) further clarify the effects of in- structions. Rates without instructions were high, undifferentiated, and regular, but be- came differentiated with instructions. Differ-

ential rates remained regular, with pauses after responding in the 30-sec and 60-sec compo- nents indicating development of temporal con- trol. To examine the after-effects of instructions,

the three subjects who had not shown discrim- inative control without instructions were re- turned to the initial discrimination. Although the schedules paired with the lights had been shuffled, and the instructions withdrawn, Sub- jects CB and PR showed clear lasting effects of instructions. For CB, response rates (shown in Fig. 1) remained virtually unchanged. The stimulus shuffling was detected rapidly and

WOOOOOO 00-00

INSTRUCTIONAL CONTROL

there was a rapid adjustment to the new stimu- lus-schedule pairings. Subject CH, however, appeared unaffected by the prior exposure to the instructions. After instructions were with- drawn, her rates quickly returned to those ob- served in her initial exposure to the discrim- ination.

Subjects WH and MB provide an additional analysis of the after-effects of instructions, since they were trained first with instructions and then switched to the uninstructed condi- tion. Figure 3 presents rates of responding for these two subjects. Acquisition of the avoid- ance response was rapid, and instructional con- trol developed in the first session for both sub- jects. When the schedules were shuffled, and the

instructions withdrawn, Subject MB rapidly acquired the new discrimination. Subject WH, however, showed no sign of forming the new discrimination. Instructions were inserted again for Subject WH, and as Figure 3 shows, differentiated performance was quickly re-es- tablished. After four sessions with instructions, WH was returned to the multiple schedule without instructions, but the discriminative stimuli signalling the various schedules were not shuffled. Nevertheless, when the instruc- tions were withdrawn, the discrimination broke down. Apparently, WH's differential performances were under the exclusive control of the instructions and when they were with- drawn, the light-schedule correlation, itself, exerted no control.

INSTRUCT

A AA A AAA

2 4 6 8 10 12 14

NO INSTRUCT

-A A A A A A

10 SEC 0 30 SEC , 60 NO

SEC- LOSS A

2 4 6 8 10 2 4 2 4

CONSECUTIVE SESSIONS Fig. 3. Responses per minute across the four components for the conditions of Experiment I. Shown are the two

subjects whose initial exposure was with instructions.

.D z

C/) LU z 0 CllJ

16 14 12 10 8 6 4 2

10 8 6 4 2

- - - - - - - - - .-.-.AL " AL

MARK GALIZIO

DISCUSSION The results illustrate the control that in-

structions can exert over human operant be- havior, and extend the analysis of instructional control to the study of behavior generated by avoidance schedules and to the case where in- structions are varied within a single session. The preliminary instructions, which specified that responding affected the delivery of loss, were sufficient to induce regular avoidance al- most immediately for all subjects. Even with prolonged exposure to the contingencies, though, only one subject showed discrimina- tion among the four components of the multi- ple schedule without instructions. These find- ings agree with previous studies of human performances on multiple schedules (Baron et al., 1969) in showing poor schedule control without instructions and more appropriately differentiated rates when instructions are added. The present experiment further analyzed

the effects of instructions, since following the acquisition of instructional control, subjects were studied with a reshuffled multiple sched- ule without instructions. Three of the five sub- jects studied learned the discrimination. For these subjects, learning a discrimination with the aid of instructions facilitated learning a new discrimination without instructions. The mechanism of this facilitation is unclear. Per- haps instructions alerted subjects to the range of the programmed contingencies, or to the sig- nal functions of the stimulus lights. The pic- ture is complicated by the performance of Subjects CH and WH, who showed no lasting after-effects of instructions.

EXPERIMENT II THE ELIMINATION OF

INSTRUCTIONAL CONTROL In Experiment I, the instructions always ac-

corded with the schedule, and instructional control occurred. Experiment II assessed the effects of inaccurate instructions. Given the present procedures, at least two types of in- accurate instructions are possible. One type evokes behavior that leads to point loss, while another type evokes behavior that leads to no clear aversive consequence. In Experiment II, subjects were studied under conditions where behavior evoked by inaccurate instructions led

either to programmed loss (Contact), or to no aversive consequence (No Contact). In the No- Contact condition, the four-ply schedule of Experiment I was transformed to a No-Loss schedule in all four components, while leaving the discriminative and instructional stimuli intact. The instructions were inaccurate under these conditions, but the subject who con- tinued to follow them did not come into con- tact with the discrepancy. The results of the Kaufman et al. (1966) study suggest that con- trol should persist. The behavior should re- main identical to that observed when the point-loss contingencies with accurate instruc- tions were in effect. An alternative possibility is that with extended exposure to inaccurate instructions, behavior may eventually come under schedule control.

In the Contact condition, an avoidance schedule with a response-loss interval of 10 sec was programmed in all four components. Again, the instructional and discriminative stimuli were left intact. Thus, conditions were established in which instruction following led to losses. Under these conditions, instruction following was expected to decline. Finally, the No-Contact conditions were reinstated after subjects were exposed to the Contact condi- tions.

METHOD Subjects and Apparatus

Four of the subjects studied in Experiment 1, MB, CH, DH, and WH, participated. The apparatus and general protocol were un- changed.

Procedure The last condition of Experiment I was the

four-ply multiple schedule without instruc- tions. To recover instructional control, the first phase of Experiment II was a return to the four-ply schedule with accurate instructions. After at least two sessions with instructional control, subjects were studied under the No- Contact condition. The loss programmer was turned off, but the stimulus lights with their accompanying labels still signalled the multi- ple schedule (reshuffled from the last phase of Experiment I). Again, the component duration was 12.5 min, and the lights appeared once per session in random order. The stability criteria were the same as in Experiment I, except that a minimum of eight sessions was required to

INSTRUCTIONAL CONTROL

permit the observation of relatively long-term effects.

After stability was reached with the No-Con- tact condition, the Contact condition was in- troduced. The Contact condition involved the same sequence of lights and labels as the No- Contact condition, but each of the four com- ponents was an avoidance schedule with a

NO CONTACT (NO LOSS)

±L L L L|

%. - - AAAAAAAAAA I I I

v v v v. v AAAA

response-loss interval of 10 sec. Finally, sub- jects were returned to the No-Contact condi- tion, where all components were No Loss.

RESULTS The leftmost panel of Figure 4 shows per-

formances on the four-ply schedule with accu- rate instructions. The instructional control

CONTACT (10 SEC)

I lIlI I l l

NO CONTACT (NO LOSS)

i CH

aYa,ml

DH XEb.am

llam.....

I10 SEC °30 SEC A 60 SEC - NO LOSS A

I I I I

2 2 4 6 8 10 12 14 2 4 6 8 10 12 2 4 6 8

CONSECUTIVE SESSIONS Fig. 4. Responses per minute across the four components for the various conditions of Experiment II.

16 14

12 10 8 6 4 2

8 6 4

12 10 8 6 4 2

z LLI

cr Cl)UJ C)

z0n LL.)uJ

10 8 6 4 2

. ^~~~~

MARK GALIZIO

observed in Experiment I was rapidly rees- tablished. The second panel shows perform- ances under the No-Contact conditions, when no point-loss was programmed in any com- ponent. In spite of the withdrawal of the avoidance contingency, subjects remained un- der instructional control. Response rates in the four components were virtually identical to those in the previous phase. Throughout the eight sessions of No-Contact exposure for Sub- jects CH and DH, the 11 sessions for Subject WH, and the 14 sessions for Subject MB, there was no trend toward reduction in rates. Rather, differentiated rates and temporal con- trol were observed throughout. The third panel of Figure 4 shows per-

formances in the Contact condition, where the 10-sec response-loss interval was in effect in all components. There was an immediate breakdown of instructional control for all four subjects. Rates were equivalent across com- ponents as early as the first session, and stabil- ized at the levels previously associated only with the 10-sec response-loss interval. Thus, responding in the Contact condition was con- trolled by the programmed avoidance sched- ule, not by the instructions. Accompanying exposure to the Contact conditions was an in- crease in losses. During the initial sessions, loss rates were higher than at any other point in the experiment, but by the final four ses- sions, subjects were again avoiding well: CH= 5, DH=4, WH=O, and MB=13 total losses during the last four sessions. A key question in Experiment II was

whether the elimination of instructional con- trol would persist after the Contact condition was removed. As the fourth panel of Figure 4 shows, instruction-following did not reappear in any subject during the second exposure to the No-Contact condition. The rates of three of the four subjects declined, appropriately to the No-Loss schedule, while those of the fourth subject, WH, were high and undifferentiated characteristic of performance under the im- mediately preceding condition. Extinction was abrupt for Subjects MB and DH; the high rates seen during the Contact condition were reduced to near zero within two sessions. Rate reductions for Subject CH were more gradual, with rates dropping to about three responses per minute during the first three sessions and stabilizing at that level. The high rates gener- ated by the Contact condition persisted for

Subject WH, whose rates showed no sign of declining after nine sessions. In spite of Sub- ject WH's lack of adjustment to the contingen- cies, there was no evidence for reinstatement of instructional control; the undifferen- tiated rates observed here were in marked con- trast to her performance during the initial ex- posure to the No-Contact condition.

DISCUSSION The results showed the role of contact with

schedule-instruction discrepancies in weaken- ing instructional control. In the No-Contact condition, the four instruction labels were superimposed on a No-Loss schedule. Instruc- tion-following under these conditions led to unnecessary responding, but because no losses were received, instructional control was main- tained with no sign of weakening. However, in the Contact condition, when instruction-fol- lowing led to exposure to the loss contingency, instructional control was rapidly eliminated. The elimination of instruction-following per- sisted when the No-Contact condition was rein- stated. This last finding is particularly impor- tant, since it shows that subject reactions to the instructions were irreversibly altered after ex- posure to the Contact condition. Subjects now "disbelieve" the instructions, and the schedule assumes control of behavior. But contact with schedule-instruction discrepancies is necessary for the elimination of instruction-following, not simply the existence of such a discrepancy. Instruction-following is controlled by its con- sequences. Although control by the avoidance contin-

gencies developed rapidly in the Contact con- dition, it was not always maintained. Subject MB nearly ceased responding during Sessions 4 and 5, which is difficult to understand, either in terms of the instructions or the contingen- cies. The transitory nature of this reaction sug- gest that it may have represented an emotional response to the increased monetary loss during the early stages of the loss of instructional con- trol. Control by the contingencies was also evi-

dent when subjects were switched from the Contact condition back to the No-Contact con- dition. Three of the four subjects showed re- ductions in response rate and two subjects, MB and DH, stopped responding altogether within three sessions. Subject WH, however, showed no reduction in rates during this period. While

INSTRUCTIONAL CONTROL

her performance revealed no indication of in- structional control, her high and undifferen- tiated rate of responding showed a lack of ad- justment to the No-Loss schedule. Perhaps the notion of contact may be invoked here also, since in WH's terminal performance during the Contact condition no losses were en- countered. Such successful avoidance perform- ance may insulate the subject from the con- tingencies (Baron and Galizio, 1976), just as instruction-following in the No-Contact con- dition did, thus rendering the response highly persistent.

EXPERIMENT III STIMULUS CONTROL OF

INSTRUCTION FOLLOWING An important property of operant behavior

is that it can be brought under discriminative control with differential reinforcement pro- cedures. If instructional control is influenced by subjects' reinforcement histories, then a stimulus associated with the elimination of in- structional control should come to inhibit in- struction following. If instructions remain ac- curate in the presence of some other stimulus, so that instruction-following is reinforced, that stimulus should control a high probability of instruction-following. In Experiment III, these possibilities were tested by programming two novel stimuli in conjunction with the four in- structional stimuli. In the presence of the other stimulus, the instructions were inaccu- rate, first under the No-Contact condition, then under the Contact condition, and finally, again under the No-Contact condition.

METHOD Subjects and Apparatus Two male subjects, SS and BK, served. The

apparatus was the same, except that two new stimulus lights, purple and orange, were added to the panel just below the amber stimulus lights.

Procedure The four amber lights plus instructions were

presented concurrently with either of two novel stimuli. The presence of the orange light characterized the four different avoidance com- ponents of Experiment I with the instructions always accurate (Sa). In the presence of the purple light, the instructions were inaccurate

(Si), first with the No-Contact condition (i.e., no loss in any component), then with the Con- tact condition (i.e., Response-loss interval of 10 sec in all components), and finally, with the No-Contact condition again. Each stimulus combination and its associated component oc- curred once per session, for 6.25 min, in ran- dom order. Thus, in each session subjects were exposed to each instruction twice, once in the presence of Sa, and once in the presence of Si.

RESULTS Figure 5 shows performances of Subjects BK

and SS during Experiment III. For each condi- tion, rates in the presence of Sa are presented in the panel to the left of rates in the presence of Si. The first panel of Figure 5 shows perform-

ances when the four instructions were accurate during Sa and the No-Contact condition pre- vailed during Si. Both subjects showed in- structional control in the presence of both Sa and Si, with no reduction of rates in the Si components despite the No-Loss schedule. The middle panel of Figure 5 shows the

effects of introducing the Contact condition in the presence of Si. Instructional control was immediately eliminated in the presence of Si, but not in the presence of Sa. Both subjects rapidly increased responding to the same level in all Si components. There was an early break- down in the precision of instructional control for both subjects in the Sa components. For example, both subjects showed moderate rates of responding in the Sa-No-Loss component during the first session, although before the Contact condition was introduced, both had near-zero rates in this component. This "dis- inhibition" was transient, lasting no more than a few sessions. The terminal performances show clearly that avoidance responding was un- der instructional control in the presence of Sa, and controlled by the response-loss interval in the presence of Si. The outcome of the second exposure to the

No-Contact conditions in Si is shown in the right panel of Figure 5. Unlike the initial ex- posure to the No-Contact condition, both subjects quickly adjusted to the No-Loss sched- ule programmed in the Si components. Both stopped responding in the Si components within three sessions. Although instruction- following ceased in the Si components, it con- tinued in the Sa components.

MARK GALIZIO

NO CONTACT Sa Si

CONTACT Sa i Si

NO Sa

CONTACT I Si I I

'10 130 '60 'NO I

SEC o SEC,A SEC* LOSSA

6 2 4 6 2 4 6

SESSIONS Fig. 5. Responses per minute for Subjects BK and SS for the various conditions of Experiment III. For each con-

dition, responding in the Sa and Si components are presented separately.

DISCUSSION Experiment III showed that instruction-

following can be brought under discriminative control. In the first phase, where instructions remained accurate in Sa, but No-Contact in- accuracies were introduced in Si, both subjects maintained differentiated rates and persisted in responding in spite of the No-Loss schedule programmed in Si. When the 10-sec loss condi- tion was introduced in Si, instruction-follow- ing was eliminated in that component, but al- though temporarily disrupted, persisted in Sa, where instructions remained accurate. The stimulus control of instruction-following per- sisted when subjects were returned to the No- Contact condition in Si. Both subjects showed rapid extinction of avoidance responding in all Si components, in spite of the contradictory suggestion to respond provided by the instruc-

tions. Performances in the Sa components re- mained unaffected.

EXPERIMENT IV THE REINFORCING PROPERTIES OF SCHEDULE INSTRUCTIONS

Wyckoff (1969) showed the reinforcing prop- erties of simple discriminative stimuli by ex- amining concurrent schedules in pigeons, where one response was reinforced with food, and another was maintained by the onset of discriminative stimuli signalling when food was available. Baron and Galizio (1976), using a similar procedure, showed that an observing response was maintained by human subjects when it produced time-correlated stimuli on an avoidance or Fl baseline. Experiment IV adapted the Wyckoff procedure, where observ- ing converts a mixed schedule to a multiple

14 12 10 8 6

,~~~~~~~~~~~~~~

2 w 0l CU w U) z 0 a.

I I

I I 'I I I

I I I I I I

I I I I

I c I %. I

. A A A A I

INSTRUCTIONAL CONTROL

schedule, for use with humans, with instruc- tions rather than simple discriminative stimuli.

In Experiment IV, subjects were exposed to the same four-ply avoidance schedule as previ- ously, but the lights with instruction labels were not illuminated unless an observing re- sponse was emitted. The prediction was that observing would be maintained under such conditions. However, in a later phase, the Con- tact condition was combined with the response- dependent instructions. Since the instructions did not provide accurate information during this condition, observing was expected to weaken.

METHOD Subjects Two of the subjects from Experiment 1, PR

and CB, were studied in Experiment IV, and a third subject, a female student (GW), was added to the experiment.

Apparatus The apparatus was the same as before, ex-

cept that turning the lever to the left, which previously had no effect, was designated the observing response.

Procedure The four-ply schedule with added, accurate,

instructions used in Experiment I, was pro- grammed throughout most of Experiment IV. However, the signal lights were no longer re- sponse independent. Rather, a counter-clock- wise turn of the lever (450, 23 N of force) was required to produce the lights. During various phases of the experiment, the observing re- sponses produced the lights for 10, 20, or 30 sec. For Subjects CB and PR, a descending se- quence was used. Since Subject GW was naive to the situation, her behavior was stabilized with response-independent instructions before her exposure to the response-dependent in- structions. An ascending sequence of display durations was then used. Two-session stability criteria were used to determine length of ex- posure to the various conditions. In order to ensure that constant-component durations did not provide cues to the programmed com- ponent, variable-duration components were used in Experiment IV, with a mean duration of 5 min. When subjects had completed the sequence of display durations, the observing response was made ineffective; that is, the in-

struction lights were no longer displayed con- sequent on an observing response. Thus, sub- jects were exposed to a mixed schedule during the Observing Extinction phase. In order to recover observing behavior after

the Observing Extinction phase, subjects were returned to the response-dependent instruc- tions conditions with a 10-sec display duration. After stability was reached, the baseline sched- ule was altered so that a response-loss interval of 10-sec was programmed in all components, but instructions varied as before and were re- sponse-dependent with a 10-sec display dura- tion. This was the Contact condition of Ex- periments II and III, but the instructions were response-dependent, instead of response-inde- pendent.

Immediately after Experiment I for Sub- jects CB and PR and immediately after stabil- ity was reached with the response-dependent instructions for Subject GW, the following typewritten instructions were presented to the subjects:

Starting today, things will be somewhat different, the four amber lights will no longer automatically come on. However, if you want to see the amber lights, you can turn them on by turning the handle to the left. Turning the handle to the left will produce the amber lights for a period of time.

RESULTS Table 1 and Figure 6 summarize the results

of Experiment IV. In Table 1, mean responses per minute for the last two sessions of each condition are presented. Subject GW showed instructional control when instructions were response independent, and all three subjects showed instructional control with the intro- duction of the observing contingency, and through the various display durations. The basis for the instructional control was high and regular rates of observing shown by all three subjects. Figure 6 shows observing rates for the last two sessions at each display duration, and rates were maintained at all three durations, and were inversely related to the duration of the display. However, the actual time the dis- play was presented was less, in spite of the higher observing rates, in the 10-sec compo- nent. For Subject PR, the display was on 75% of the session during the 10-sec display condi-

MARK GALIZIO

C/)

C/) z 0

Cl)

3 h

10 SEC

*mPR

cCB

* IF

20 SEC 30 SEC EXT CONTACT

CONDITIONS Fig. 6. Observing responses per minute for the three subjects of Experiment IV. Data points represent mean per-

formance for the last two sessions of each condition.

tions, compared to 83% during the 20-sec and responding on the mixed schedule generated 30-sec display conditions. Similar results were by these conditions was less well differentiated, obtained for Subjects GW and CB, who pro- although, as Table 1 shows, all three subjects duced the display during, respectively, 33% tended to show higher rates in the 10-sec com- and 28% of the 10-sec sessions, 40% and 45% ponent. Subjects tended to receive more losses of the 20-sec sessions, and 45% and 45% of during these conditions: CB=7, PR=12, and the 30-sec sessions. As in the previous experi- GW=9, for the last two sessions of Avoidance ments, there were few losses: five for CB, 10 Extinction. for PR, and seven for GW, during the last All three subjects showed rapid recovery of two sessions of the 10-sec display condition. observing when the observing dependency was When the Observing Extinction conditions reinstated with a 10-sec display duration, and

were introduced, observing rates dropped to this was accompanied by a reinstatement of in- near zero in all three subjects, and avoidance structional control (although the reinstate-

4 r

INSTRUCTIONAL CONTROL

Table 1

Mean avoidance responses per minute for the last two sessions of the conditions of Experiment IV.

Subjects

Conditions GW CB PR

Response-independent 10 sec 7.6 - - instructions 30 sec 3.5 - -

60 sec 2.4 - - EXT 0.2 - -

Response-dependent 10 sec 8.9 9.8 10.1 instructions: 30 sec 3.4 4.2 2.9 10-sec display 60 sec 2.1 2.0 1.5

EXT 0.0 0.0 0.0 Observing extinction 10 sec 6.5 9.5 9.0

30sec 4.5 6.9 8.8 60 sec 4.3 7.3 3.7 EXT 3.4 8.6 4.4

Response-dependent 10 sec 7.5 9.6 11.7 inaccurate 30 sec 8.1 8.6 12.0 instructions: 60 sec 7.6 9.0 12.3 10-sec display EXT 7.9 7.6 12.1

ment was uninstructed). When the Contact condition was introduced, instruction-follow- ing was eliminated, and as Figure 6 shows, ob- serving responses were near zero in all three subjects during this condition.

DISCUSSION When accurate instructions depended on

an observing response, observing was readily acquired. The functional relationship between the display duration and the rate of observing supported the conclusion that the instructional stimuli were the source of reinforcement for the observing response. Further support was provided by the rapid extinction of observing when the response no longer produced the in- structions. Finally, the elimination of observ- ing when the instructions were no longer ac- curate shows that instructions were reinforcing only when avoidance behavior was under in- structional control. These results thus support the view of instructions as discriminative stimuli, in the sense that both instructions and simple discriminative stimuli can serve as reinforcers in the observing paradigm. Of ad- ditional interest was the rapid loss of instruc- tional control when the Contact condition was introduced, once again demonstrating that in- struction-following is controlled by its conse- quences. The source of the reinforcing properties of

discriminative stimuli has long been of theo- retical interest. One explanation, which has

found support in infra-human studies with ob- serving procedures, is that the reinforcing properties are acquired through direct classical conditioning. This explanation appears un- likely in the present context. Although sub- jects received slightly fewer losses with the in- structions, the differences were quite small. An alternative viewpoint is that discriminative stimuli are reinforcing because they permit more efficient behavior. Certainly in the pres- ent study, the addition of instructions resulted in fewer avoidance responses overall, but when the additional effort of the observing response is considered, especially with the 10-sec dis- play duration, total work for two of the three subjects actually increased with instructions. This increased effort is clearly seen when avoidance response rates obtained in the Ob- serving Extinction condition (GW=4.7, CB= 8.0, PR=6.5 responses per minute overaged across components) are compared with total responding (observing and avoidance) in the 10-sec display conditions (GW=6.9, CB=7.0, and PR=8.1 responses per minute averaged across components). Thus, the most viable hy- pothesis regarding the reinforcing properties of instructions is the information hypothesis (Hendry, 1969), which asserts that stimuli that reduce uncertainty about forthcoming events are reinforcing. The instructions provided in- formation about the schedule then in effect, and this may have been sufficient to establish them as reinforcers.

GENERAL DISCUSSION

Taken together, the four experiments sup- port the view of instructional control as in- volving rule-governed behavior, and illustrate mechanisms by which subjects' reinforcement histories influence subsequent rule-governed behaviors. The rapid induction of avoidance by the preliminary instructions and the initia- tion of discriminative control by the instruc- tional labels in Experiment I were consistent with Skinner's (1974) analysis of the properties of rule-governed behavior. Instructional con- trol was shown to be influenced by the conse- quences of following instructions (Experiment II). Given differential reinforcement, instruc- tion-following was brought under stimulus con- trol (Experiment III). Finally, instructions were shown to possess reinforcing properties, a

MARK GALIZIO

characteristic shared by simple discriminative stimuli (Experiment IV). The view of instructional control as rule-

governed behavior that is controlled by its con- sequences both explains the present findings, and accounts for much of the literature re- viewed earlier. Studies showing that instruc- tions facilitate the acquisition of schedule con- trol (cf. Baron, Kaufman, and Stauber, 1969; Scobie and Kaufman, 1969; Turner and Solo- mon, 1962) can be incorporated into the pres- ent framework by positing that instructional control is at high strength when subjects first enter a psychology experiment. This assump- tion is warranted by the present results, as well as by research on the social psychology of the psychology experiment (Orne, 1962). Thus, when instructions are presented that specify re- sponses relevant to the experiment, subjects execute those behaviors, and may give the ap- pearance of having come under schedule con- trol.

Consider now the case where inaccurate in- structions have been studied. These studies have yielded particularly awkward results from the perspective of the reinforcement theorist. In one of the experiments reported by Kauf- man et al. (1966), some subjects were instructed that monetary reinforcement was available on an Fl basis, and others were told that it was programmed on a VR schedule. Although the actual schedule of reinforcement was a VI, both groups behaved as specified by the in- structions: subjects given VR instructions re- sponded at high rates, while those given Fl instructions showed scalloping. However, in- structional inaccuracies in the Kaufman et al. experiment are most analogous to the No-Con- tact inaccuracies of the present research. On the schedules programmed in the Kaufman et al. study, responding at a high rate or showing a scalloped response pattern was frequently reinforced, so there was never contact with the schedule-instruction discrepancy. The present results showed that instruction control can be maintained indefinitely when such contact is absent. A similar analysis of the Lippman and Meyer (1967) study, where subjects were given VR instructions but reinforcement was sched- uled on an Fl basis, reveals that no contact with the schedule-instruction discrepancy need have occurred. Thus, the Kaufman et al. ex- periment, as well as the Lippman and Meyer study, can be understood in terms of the per-

sistence of instructional control in the absence of contact with schedule-instruction discrep- ancies. One difficult finding for the present model

is the third experiment by Kaufman et al., where subjects were exposed to a single session in which no monetary reinforcement was pro- grammed, but were given instructions that a VR schedule was in effect. These conditions resemble the Contact conditions of the present study, and as expected, there was some weaken- ing of instructional control in this phase of the Kaufman et al. study. However, the break- down of instructional control was neither as complete or as rapid as that observed in the present study. But on the avoidance baseline, subjects that persisted in following instructions experienced immediate loss, while on the ex- tinction schedule of Kaufman et al., although instruction-following did not lead to reinforce- ment, it was not punished. Perhaps with the less-aversive consequences of continued in- struction-following in the Kaufman et al. study, more than a single session would be re- quired to extinguish instructional control com- pletely. Indeed, Ayllon and Azrin (1964) found that instructional control was weakened greatly over sessions with a similar procedure. The present findings have considerable im-

plications for the analysis of human operant behavior. For the researcher studying the sim- ple operant behavior of adult humans, caution seems indicated. Since some verbal interaction between subject and experimenter is nearly un- avoidable, care is needed to ensure that the be- havior under examination is under schedule, and not instructional control. Mathews, Shim- off, Catania, and Sagvolden (1977) suggested that inadvertant instructional control may ac- count for many cases of poor schedule control in humans (Streifel, 1972; Schmitt, 1974; Wei- ner, 1970). Thus, in cases where schedule con- trol of simple responses is of interest, minimal instructions should be employed. The analysis of instructional control itself

warrants more study. The rapid acquisition of discriminative control with instructions, and its persistence when instructions were with- drawn, demonstrates the power of instruc- tional control. But such persistence of control does not always lead to more efficient behavior, as was forcefully demonstrated in the initial No-Contact exposures of the present research. Under these conditions, steady rates of avoid-

INSTRUCTIONAL CONTROL 69

ance responding were maintained for many ses- sions in spite of the uniform No-Loss contin- gency. Other instances where human behavior is apparently in discord with the prevailing contingencies may be explained by the per- sistence of instructional control (cf. Herrn- stein's, 1966, analysis of human superstitions). The finding that instruction-following may

be brought under external stimulus control may have particular implications for the analy- sis of social behavior. Consider the instructions presented in Experiment III as verbal state- ments or appeals originating from different sources. After the experience of the contact condition, the probability is low tIat com- munications from the inaccurate souiPce would influence other's behavior. On the other hand, the present subjects continued to follow in- structions in the presence of the stimulus as- sociated with accurate instructions. Bandura (1969) argued that communicator's influence depends on the likelihood that the source's be- havioral recommendations will lead to favor- able consequences. The present research sup- ports Bandura, and its methodology might be profitably used for further examination of such issues. The present analysis provided a behavioral

account of some of the phenomena associated with instructional control. The potency of in- structional control can be interpreted not as a limitation of reinforcement control of hu- man behavior (cf. Brewer, 1974), but rather as an instance of reinforcement history affect- ing rule-governed behavior. Many questions raised by this type of analysis remain to be answered. For example, the initial acquisition of instructional control can be addressed only through developmental research. The details of how reinforcement affects rule-governed be- havior, and the conditions necessary for con- trol by instructions and rules of other forms remain incomplete, but the present research shows the utility of an experimental analysis of these phenomena.

REFERENCES Ayllon, T. and Azrin, N. H. Reinforcement and in-

structions with mental patients. Journal of the Ex- perimental Analysis of Behavior, 1964, 7, 327-331.

Bandura, A. Principles of behavior modification. New York: Holt, Rinehart & Winston, 1969.

Bandura, A. Vicarious and self-reinforcement proc-

esses. In R. Glaser (Ed), The nature of reinforce- ment. New York: Academic Press, 1971. Pp. 228-278.

Bandura, A. Behavior theory and the models of man. American Psychologist, 1974, 29, 859-869.

Baron, A. and Galizio, M. Clock control of human per- formances on avoidance and fixed-interval schedules. Journal of the Experimental Analysis of Behavior, 1976, 26, 165-180.

Baron, A. and Kaufman, A. Facilitation and suppres- sion of human loss-avoidance by signalled, unavoid- able loss. Journal of the Experimental Analysis oj Behavior, 1968, 11, 177-185.

Baron, A., Kaufman, A., and Stauber, K. A. Effects of instructions and reinforcement feedback on human operant behavior maintained by fixed-ihterval rein- forcement. Journal of the Experimental Analysis of Behavior, 1969, 12, 701-712.

Brewer, W. F. There is no convincing evidence for operant or classical conditioning in adult humans. In W. B. Weimer and D. S. Palermo (Eds), Cogni- tion and the symbolic processes. Hillsdale, New Jersey: Erlbaum, 1974. Pp. 1-42.

Dulany, D. E. Awareness, rules, and propositional con- trol: A confrontation with S-R behavior theory. In T. R. Dixon and D. L. Horton (Eds), Verbal be- havior and general behavior theory. Englewood Cliffs, New Jersey: Prentice-Hall, 1968. Pp. 340-387.

Hendry, D. P. Introduction. In D. P. Hendry (Ed), Conditioned reinforcement, Homewood, Illinois: Dorsey, 1969. Pp. 1-33.

Herrnstein, R. J. Superstition: A corollary of the prin- ciples of operant conditioning. In W. K. Honig (Ed), Operant behavior: areas of research and application. New York: Appleton-Century-Crofts, 1966. Pp. 33- 51.

Kaufman, A., Baron, A., and Kopp, R. E. Some effects of instructions on human operant behavior. Psycho- nomic Monograph Supplements, 1966, 1, 243-250.

Lippman, L. G. and Meyer, M. E. Fixed-interval per- formance as related to instructions and to subjects' verbalizations of the contingency. Psychonomic Science, 1967, 8, 135-136.

Mathews, B. A., Shimoff, E., Catania, A. C., and Sag- volden, T. Uninstructed human responding: sensi- tivity to ratio and interval contingencies. Journal of the Experimental Analysis of Behavior, 1977, 27, 453-469.

Nevin, J. A. Stimulus control. In J. A. Nevin and G. S. Reynolds (Eds), The study of behavior: learning, motivation, emotion and instinct. Glenview, Illinois: Scott, Foresman, 1973. Pp. 115-154.

Orne, M. T. On the social psychology of the psychol- ogy experiment: With particular reference to de- mand characteristics and their implications. Ameri- can Psychologist, 1962, 17, 776-783.

Schmitt, D. R. Effects of reinforcement rate and rein- forcer magnitude on choice behavior of humans. Journal of the Experimental Analysis of Behavior, 1974, 21, 409-419.

Schutte, R. C. and Hopkins, B. L. The effects of teacher attention on following instructions in a Kindergarten class. Journal of Applied Behavior Analysis, 1970, 3, 117-122.

Scobie, S. R. and Kaufman, A. Intermittent punish- ment of human responding maintained by inter-

70 MARK GALIZIO

mittent reinforcement. Journal of the Experimental Analysis of Behavior, 1969, 12, 137-147.

Sidman, M. Tactics of scientific research. New York: Basic Books, 1960.

Skinner, B. F. Verbal behavior. New York: Appleton- Century-Crofts, 1957.

Skinner, B. F. About behaviorism. New York: Alfred A. Knopf, 1974.

Streifel, S. Timeout and concurrent fixed-ratio sched- ules with human subjects. Journal of the Experi- mental Analysis of Behavior, 1972, 17, 213-219.

Turner, L. H. and Solomon, R. L. Human traumatic avoidance learning. Psychological monographs, 1962, 76, N. 40 (Whole No. 559).

Urcuioli, P. J. Transfer of oddity-from-sample per- formance in pigeons. Journal of the Experimental Analysis of Behavior, 1977, 25, 195-202.

Urcuioli, P. J. and Nevin, J. A. Transfer of hue matching in the pigeon. Journal of the Experimental Analysis of Behavior, 1975, 24, 149-156.

Weiner, H. Some effects of response cost on human operant behavior. Journal of the Experimental Analysis of Behavior, 1962, 5, 201-208.

Weiner, H. Human behavioral persistence. Psycho- logical Record, 1970, 20, 445-456.

Wyckoff, L. B. Jr. The role of observing responses in discrimination learning. In D. P. Hendry (Ed), Con- ditioned reinforcement. Homewood, Illinois: Dorsey, 1969. Pp. 237-260.

Received 28 August 1977. (Final acceptance 6 August 1978.)