INTERMITTENT SCHEDULES OF REINFORCEMENT AND PUNISHMENT:
IMPLICATIONS FOR THE TREATMENT OF SEVERE BEHAVIOR DISORDERS
IN INDIVIDUALS WITH DEVELOPMENTAL DISABILITIES
DOROTHEA C. LERMAN
A DISSERTATION SUBMITTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I wish to thank those individuals who helped make these investigations possible.
Appreciation is extended to the members of my committee, Marc Branch, Timothy
Hackenberg, Shari Ellis, and Cecil Mercer, for their input and support during the
preparation of this document. I would like to extend special thanks to Dr. Brian A.
Iwata, my committee chair and major professor, who offered assistance and
encouragement throughout my education. His guidance, friendship, and support have
profoundly influenced my academic, professional, and personal development.
I also would like to thank my fellow graduate students and colleagues, Iser DeLeon,
Han Goh, SungWoo Kahng, Jodi Mazaleski, Bridget Shore, Richard Smith, and Sonya
Ulrich, who assisted in the development and implementation of this project. Finally, I
dedicate this dissertation to my parents, Harvey and Roberta Lerman, for their
unconditional love and support. They are quick to offer help and encouragement when
my confidence falters, yet they continue to respect my independence.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS.............................. ................. ii
LIST OF TABLES ................................................................ v
LIST OF FIGURES .............................................................. vi
ABSTRACT ....................................................................... vii
Assessment and Treatment of Severe Behavior Disorders................ 1
Implications of Research Findings on Extinction and Punishment
for Treatment Programs in Applied Settings........................ 6
Reinforcement and Punishment Schedules................................... 7
Intermittent Schedules of Reinforcement:
The Partial-Reinforcement-Extinction Effect............................. 8
Intermittent Schedules of Punishment....................................... 15
Statement of Purpose......................................................... 19
GENERAL METHOD ............................................................ 20
Subjects and Setting.......................................................... 20
Human Subjects Considerations ............................................. 21
Response Measurement and Reliability ..................................... 22
STUDY 1: FUNCTIONAL ANALYSIS OF PROBLEM BEHAVIOR....... 24
Procedures ....................................................................... 24
STUDY 2: THE PARTIAL-REINFORCEMENT-EXTINCTION EFFECT.. 35
Procedures....................................... ............................. 35
Results............................................. ............................ 37
Discussion........................................ .......................... 53
STUDY 3: INCREASING THE EFFECTIVENESS OF INTERMITTENT
PUNISHMENT VIA SCHEDULE FADING............................... 58
Procedures........................................ ............................ 58
Results............................................ ............................ 60
D iscussion........................................... .......................... 68
GENERAL DISCUSSION............................................................... 73
REFERENCES................................................ ..... ................. 77
BIOGRAPHICAL SKETCH..................................................... 88
LIST OF TABLES
1 Summary and interpretation of findings following
FR-1 and INT reinforcement schedules........................... 52
2 Proportion of responses that followed punishment delivery
within specified time periods (in seconds)......................... 67
LIST OF FIGURES
1 Functional analysis results for Brandon (top panel)
and Sue (bottom panel).............................. ........ 28
2 Functional analysis results for Harold....................... 30
3 Functional analysis results for Paul (top panel)
and Merry (bottom panel).......................................... 32
4 Rates of SIB for Brandon across sessions in Study 2........... 39
5 Rates of SIB for Sue across sessions in Study 2................ 42
6 Rates of SIB for Harold across sessions in Study 2............. 46
7 Proportion of baseline measure for each extinction session
in Study 2 for Brandon (top panel), Sue (middle panel), and
Harold (bottom panel)............................... ........... 50
8 Percentage of intervals of hand mouthing for Paul across
sessions in Study 3................................... ......... 62
9 Percentage of intervals of hand mouthing for Merry across
sessions in Study 3.................................... ........ 64
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
INTERMITTENT SCHEDULES OF REINFORCEMENT AND PUNISHMENT:
IMPLICATIONS FOR THE TREATMENT OF SEVERE BEHAVIOR DISORDERS
IN INDIVIDUALS WITH DEVELOPMENTAL DISABILITIES
Dorothea C. Lerman
Chairman: Brian A. Iwata
Major Department: Psychology
Results of numerous studies indicate that extinction and punishment are effective in
treating a variety of severe behavior disorders, including self-injury and aggression. In
the natural environment, however, the history of reinforcement that precedes treatment
with extinction and the consistency of punishment contingencies may not resemble those
in studies examining the utility of these treatment procedures. Staff generally do not
deliver reinforcement and punishment following every instance of the problem behavior,
and basic research findings suggest that these intermittent contingencies could have
important implications for treatment efficacy. The current series of studies examined the
effects of intermittent reinforcement and punishment schedules on the treatment of self-
injury and related behavior disorders. Subjects whose behavior was maintained by social
reinforcement were exposed to baseline conditions with continuous and intermittent
reinforcement schedules, and their performance during subsequent extinction was
compared. In a separate study, a procedure designed to increase the efficacy of
intermittent punishment was examined with subjects whose behavior was not maintained
by social reinforcement. Delivery of punishment was leaned gradually while attempting
to maintain low levels of self-injury. Data are presented indicating that these intermittent
contingencies can have complex effects on responding during treatment. Various
strategies for the use of intermittent schedules in the natural environment are discussed,
as are potential refinements and extensions of these investigations.
The current series of studies examined the effects of intermittent reinforcement and
punishment on the treatment of self-injury and related behavior disorders in individuals
with developmental disabilities. After identifying the reinforcement contingencies
maintaining problem behavior, the effects of continuous and intermittent reinforcement
schedules during subsequent extinction were compared for subjects whose behavior was
maintained by social reinforcement. In a separate study, the effects of intermittent
punishment schedules on treatment with timeout or contingent restraint were examined
for subjects whose problem behavior was not maintained by social reinforcement.
Assessment and Treatment of Severe Behavior Disorders
Research findings during the past 30 years have demonstrated that many problem
behaviors exhibited by individuals with developmental disabilities are learned responses
and, as such, are treated most effectively with procedures based on principles of operant
conditioning. Severe behavior disorders such as self-injury and aggression have been
treated with a variety of procedures, including extinction, differential reinforcement, and
punishment. In recent years, studies have focused increasingly on the use of functional
analyses to identify the specific sources of reinforcement that maintain problem behavior,
and research findings indicate that treatments based on the outcome of such assessments
are more effective in reducing problem behavior than those selected arbitrarily (e.g.,
Iwata, Pace, Dorsey et al., 1994; Iwata, Pace, Cowdery, & Miltenberger, 1994; Repp,
Felce, & Barton, 1988).
During the past ten years, numerous studies have concentrated on the development
and refinement of functional assessment methodologies (see Iwata, Vollmer, & Zarcone,
1990, and Mace, Lalli, & Lalli, 1991, for recent reviews). The experimental (functional)
analysis, which involves direct and systematic manipulation of potential maintaining
variables, has been used most often. Compared to other assessment methods, the
functional analysis is considered superior for identifying causal relationships, and the
utility of this approach has been well established with a variety of behavior disorders,
including self-injury (Day, Rea, Schussler, Larsen, & Johnson, 1988; Iwata, Pace,
Dorsey et al., 1994; Steege et al., 1989); stereotypy (Adams, Tallon, & Stangl, 1980;
Sturmey, Carlsen, Crisp, & Newton, 1988), aggression (Mace, Page, Ivancic, &
O'Brien, 1986; Slifer, Ivancic, Parrish, Page, & Burgio, 1986), and disruption (Carr &
Before conducting a functional analysis, putative variables may be identified via other
assessment methods, such as interviews or descriptive (correlational) analyses.
However, most studies on the functional analysis of behavior have employed a general
assessment approach developed by Iwata, Dorsey, Slifer, Bauman, and Richman (1982),
which tests several variables potentially relevant to the target behavior. The experimental
conditions of the functional analysis described by Iwata et al. (1982) are designed to
examine the relative influence of social-positive reinforcement (in the form of attention),
social-negative reinforcement (in the from of escape from instructions), and automatic
reinforcement on rates of problem behavior.
Research on the functional analysis of behavior has isolated a number of
contingencies that maintain severe behavior disorders. These reinforcing consequences
can be mediated through the actions of others (i.e., "social" consequences) or produced
directly by the behavior (i.e., "automatic" consequences). Contingencies that maintain
maladaptive behavior include access to attention (e.g., Mace et al., 1986), play materials
(e.g., Day et al., 1988), and response-produced stimulation (e.g., Rincover, Cook,
Peoples, & Packard, 1979), as well as escape from instructions (e.g., Iwata, Pace,
Kalsher, Cowdery, & Cataldo, 1990), social interaction (e.g., Taylor & Carr, 1992), and
ambient stimulation (e.g., Iwata, Pace, Dorsey et al., 1994).
Treatment programs based on results of functional analyses can take a variety of
forms (see Iwata, Vollmer, & Zarcone, 1990, and Mace et al., 1991, for reviews). For
example, practitioners can implement procedures to alter antecedent events that occasion
behavior problems (e.g., Weeks & Gaylord-Ross, 1981), eliminate access to maintaining
reinforcers through extinction (e.g., Iwata, Pace, Cowdery, et al., 1994), or provide
those reinforcers contingent on alternative behavior (e.g., Carr & Durand, 1985).
Particularly relevant to the current study is the fact that extinction often may be a crucial
component of effective treatment (cf. Fisher et al., 1993; Mazaleski, Iwata, Vollmer,
Zarcone, & Smith, 1993; Wacker et al., 1990; Zarcone, Iwata, Smith, Mazaleski, &
Extinction involves terminating the reinforcement contingency that maintains a
response, which results in a reduction in the behavior's occurrence over time. Several
procedural variations of extinction have been examined in the literature. The most
common, extinction of responding maintained by positive reinforcement (e.g., access to
attention or leisure materials), usually involves withholding the specific reinforcing
stimulus. For example, extinction might be implemented by failing to provide attention
following occurrences of the target response (e.g., France & Hudson, 1990). Extinction
of responding maintained by negative reinforcement (e.g., escape from or avoidance of
instructional activities) usually involves continued presentation of the aversive stimulus
despite occurrences of the behavior, a procedure often called "escape extinction." For
example, extinction might be implemented by continuing the instructional activity
(i.e., failing to provide escape) contingent on occurrences of the target behavior (e.g.,
Iwata, Pace et al., 1990).
Applied researchers also have developed extinction techniques for behavior
maintained by nonsocial (automatic) reinforcement. In some cases, behavior appears to
produce its own maintaining consequences directly, which can serve as either positive
reinforcement (e.g., "pleasurable" tactile or auditory stimulation) or negative
reinforcement (e.g., attenuation of "painful" or aversive stimulation). Extinction of
behavior maintained by automatic reinforcement, often called "sensory extinction,"
includes a variety of techniques designed to attenuate or eliminate the hypothesized
reinforcing stimuli that follow occurrences of the target behavior. Procedural variations
of sensory extinction depend on the type stimuli (e.g., auditory, visual, tactile)
maintaining the response. For example, self-injurious behavior (SIB), such as head
hitting, arm biting, and hand mouthing, sometimes appears to be maintained by
response-produced tactile stimulation. Sensory extinction has been implemented in these
cases by placing equipment (e.g., helmets, gloves) on the individual to block or attenuate
the stimulation (e.g., Dorsey, Iwata, Reid, & Davis, 1982; Luiselli, 1988).
Results of numerous studies have shown that extinction can produce robust treatment
effects when the procedure is implemented as the sole intervention (e.g., Iwata, Pace,
Cowdery et al., 1994) or in combination with alternative procedures, such as differential
reinforcement of other behavior or DRO (Vollmer, Iwata, Zarcone, Smith, & Mazaleski,
1993) and differential reinforcement of alternative behavior or DRA (Lalli, Browder,
Mace, & Brown, 1993). In fact, the efficacy of such procedures has led some authors to
suggest that the current functional analysis technology has eliminated the need for
restrictive interventions such as punishment (e.g., Donnellan & LaVigna, 1990).
In a recent paper addressing this issue, however, Vollmer and Iwata (1993) discussed
several conditions under which treatment based on the results of functional analyses
would not necessarily preclude the use of punishment. One condition particularly
relevant to the current experiments involves the treatment of behavior maintained by
automatic reinforcement. In this situation, treatment options may be somewhat limited.
Unlike extinction of behavior maintained by social reinforcement, sensory extinction is
particularly complex because it can be difficult to identify or withhold the specific
reinforcing stimuli produced by the behavior. When these reinforcers cannot be
eliminated, the efficacy of other (nonaversive) interventions, such as DRO and DRA, will
depend on the identification of alternative sources of stimulation that successfully
compete with, or substitute for, the maintaining reinforcer.
Compared to reinforcement-based treatments, punishment generally may be more
successful in "overriding" the variables that maintain problem behavior (Paisey, Whitney,
& Hislop, 1990; Vollmer & Iwata, 1993). Punishment procedures, including timeout
from positive reinforcement and the delivery of aversive stimulation (e.g., electric shock,
loud noise, water mist, certain tastes and smells), have effectively reduced a variety of
severe behavior disorders maintained by unknown sources of reinforcement (see Axelrod
& Apsche, 1983, Matson & DiLorenzo, 1984, and Vollmer, 1994, for reviews). Such
procedures may in fact be the "most reliable treatments in the absence of a conclusive
functional analysis" (Vollmer, 1994, p. 201).
Results of studies on extinction and punishment suggest that these procedures would
be effective in treating problem behavior in applied settings such as schools and
residential facilities. However, as discussed in the next section, the efficacy of these
procedures may depend on the extent to which certain conditions operating in the natural
environment are similar to those found in the research setting.
Implications of Research Findings on Extinction and Punishment
for Treatment Programs in Applied Settings
It is generally assumed that the robust treatment effects demonstrated in research on
extinction and punishment can be replicated in applied settings when teachers or
caregivers attempt to reduce problem behavior. In many cases, however, treatment
efficacy may be compromised because the manner in which some contingencies are
implemented in the experimental setting is significantly different from those in the natural
In clinical research on extinction, it is common practice to expose subjects to a
baseline period during which every occurrence of the target behavior is followed by
reinforcement (the consequence that will be later withheld during extinction). In the
natural environment, however, reinforcement rarely follows every instance of the
response. For example, staff do not typically attend to each occurrence of inappropriate
behavior, or they may be unwilling to allow escape from certain demands (e.g., "Get
dressed."). In a similar manner, most studies on punishment involve consistent delivery
of consequences even though staff often fail to implement punishment contingencies
following every instance of the inappropriate behavior.
Thus, in the natural environment, the history of reinforcement preceding treatment
with extinction or the consistency with which punishment is implemented may not
resemble that found in studies examining the utility of these treatment procedures.
Consequences often are delivered intermittently in the natural environment, and results of
basic research suggest that these patterns of contingencies (called "intermittent schedules
of reinforcement or punishment") may influence treatment efficacy. The next section
describes these schedules of reinforcement and punishment in greater detail, thus setting a
framework for a review of studies on intermittent schedules, as well as for the current
series of studies.
Reinforcement and Punishment Schedules
Schedules of reinforcement and punishment are rules specifying the probability that a
specific occurrence of a response will be followed by reinforcement or punishment.
Under a continuous schedule, every occurrence of the response produces the reinforcer
or punisher. By contrast, intermittent (INT) schedules are those in which only some
instances of the response are followed by the consequence. A variety of INT schedules
of reinforcement and punishment have been examined in basic studies (e.g., Azrin, Holz
& Hake, 1963; Ferster & Skinner, 1957; Filby & Appel, 1966).
For example, delivery of consequences can be based on the number of responses that
occur (called "ratio schedules") or on the passage of time (usually since the last reinforced
or punished response, called "interval schedules"). The ratio or interval requirement for
delivery of a reinforcer/punisher can remain constant (i.e., "fixed") or it can vary around
some average value (i.e., "variable"). Thus, consequences may be delivered following a
constant number of responses ("fixed ratio" or FR), after a constant amount of time has
elapsed ("fixed interval" or FI), following a variable number of responses ("variable
ratio" or VR), or after a variable amount of time has elapsed ("variable interval" or VI).
When behavior maintained by reinforcement is exposed to punishment, different
schedules can be specified for the two contingencies. For example, a behavior
maintained on an FI schedule of reinforcement may be punished on a VR schedule.
Problem behavior can be exposed to one or more of these INT schedules in the
natural environment (although variable schedules are probably more common than fixed
schedules). For example, staff may withhold consequences until the individual's
behavior becomes "too bad to ignore" or "bad enough to warrant intervention." In this
situation, patterns of reinforcement or punishment delivery may approximate ratio
schedules. On the other hand, some caregivers may reliably provide consequences for an
individual's target behavior, but they do not observe the individual continuously
throughout the day. In this case, responding may produce a consequence only after a
period of time has elapsed (i.e., the interval during which the individual is unobserved), a
pattern that may approximate interval schedules.
Although these examples describe INT schedules of social consequences, similar
patterns could be in effect when the problem behavior produces its own reinforcing or
punishing stimuli (i.e., automatic consequences). In some cases, for example, the
occurrence of a certain number of responses may be necessary to produce the
consequence (i.e., the effects of a response may be cumulative). Compared to schedules
of social consequences, however, patterns of automatic reinforcement or punishment are
extremely difficult to identify and manipulate. Thus, studies examining INT schedules of
reinforcement and punishment have focused primarily on social consequences. Results
of this research and their implications for the treatment of problem behavior in applied
settings are discussed in the next two sections.
Intermittent Schedules of Reinforcement:
The Partial-Reinforcement-Extinction Effect1
Results of numerous basic studies with humans and nonhumans indicate that
exposure to INT, or partial, schedules of reinforcement can increase resistance to
extinction, a phenomenon that has been termed the "partial-reinforcement-extinction
effect" (see Kimble, 1961, and Mackintosh, 1974, for reviews). Resistance to extinction
(i.e., the extent to which responding persists in the absence of reinforcement) has been
measured by calculating response rate (e.g., Cowen & Walters, 1963), number of
responses (e.g., Bijou, 1958; Hearst, 1961), or amount of time to meet a prespecified
extinction criterion such as no responses for 5 min. (e.g., Perin, 1942). Using one or
1This review is limited to studies examining the effects of intermittent reinforcement on free-operant
responses. Results of these studies may be more relevant to the current experiment than those of studies
examining discrete-trial procedures because problem behaviors usually are free-operant responses (i.e.,
they can be displayed at any time).
more of these measures, basic researchers have demonstrated the partial-reinforcement-
extinction effect (PREE) with a variety of subjects, responses, and reinforcement
As a result, the PREE often is considered "one of the fundamental rules governing the
application of learning principles to practical problems" (Pittenger & Pavlik, 1988, p. 2).
In particular, results of studies on the PREE suggest that problem behaviors may be
difficult to treat with extinction if they have been maintained on INT rather than
continuous (FR-1) schedules. Some authors have even suggested that, due to potential
difficulties generated by the PREE, extinction should not be used as treatment for severe
behavior disorders (e.g., LaVigna & Donnellan, 1986). Thus, it is somewhat surprising
that no applied studies have examined the clinical significance of the PREE with problem
behavior, and that only a few studies have investigated the effects of INT schedules on
other types of responses.
Kazdin and Polster (1973), who reinforced the social interactions of two men
diagnosed with mild retardation during three daily break periods at a sheltered workshop,
compared the effects of two reinforcement schedules on response maintenance during
extinction. Reinforcement conditions were alternated with extinction conditions within a
reversal design. Initially, both subjects received tokens immediately following each
break period ("continuous reinforcement") for conversing with peers. They were then
exposed to extinction for three weeks, and the social interactions of both subjects
decreased to near zero levels by the second week. Following extinction, one subject
received tokens on the continuous schedule for conversing with peers, while the other
subject received tokens after either one or two of the three break periods ("intermittent
reinforcement"). Both subjects then were exposed to extinction for five weeks. The
subject who had received tokens on the continuous schedule exhibited few social
interactions by the second week of extinction, whereas the subject who had received
tokens on the INT schedule showed no reduction in behavior across the five weeks of
extinction. Although these results provide one of the few demonstrations of the PREE in
applied research, the effect of INT reinforcement may have been partially a function of
reinforcement delay, another variable that was included in the procedure (i.e., the
subjects received reinforcement after the break period rather than immediately following
each interaction). When combined with INT schedules, reinforcement delay can enhance
the PREE (cf. Peterson, 1956). In addition, results of this between-subject comparison
may simply reflect different extinction rates for the two subjects.
Results of a study by Koegel and Rincover (1977) also suggested that INT schedules
can facilitate behavioral maintenance. In the study's first experiment, the appropriate
behavior of two children diagnosed with autism generalized to settings unassociated with
the treatment contingencies but failed to maintain in these settings in the absence of
reinforcement. Consequently, in a separate experiment, the authors manipulated the
reinforcement schedule in the training setting to examine its effect on response
maintenance in the nontraining setting with four children diagnosed with autism. Results
showed that a relatively thin reinforcement schedule, in which every fifth instance of
appropriate behavior in the training setting received reinforcement (i.e., FR 5), was
associated with continued responding in the nontraining setting with no apparent
decrements for up to 500 trials, whereas FR-1 or FR-2 schedules were associated with
fairly rapid decreases in behavior in the nontraining setting. However, three of the
children were exposed to just one reinforcement schedule (FR 1, FR 2, or FR 5); thus,
results of this between-subject comparison may simply reflect different extinction rates
for the three subjects. Although the fourth child was exposed to two reinforcement
schedules (FR 2 and FR 5), each was paired with a different response, and the results
could reflect different extinction rates for the two types of behavior.
Finally, Baer, Blount, Detrich, and Stokes (1987) investigated the effects of INT
schedules on the maintenance of correspondence between verbal and nonverbal nutritious
snack choices in a day-care setting. Maintenance of correspondence following both FR-1
and INT schedules was compared for one of the three subjects within a reversal design.
(The other two subjects were exposed to extinction after INT only.) First, the subject
received reinforcement (e.g., hugs, stickers) for verbalizing healthy food choices prior to
the daily snack period, a procedure that resulted in little change in the amount of
nutritious items actually selected during snack time. During the next phase, the subject
received reinforcement only if these verbalizations matched the items chosen during snack
time (reinforcement of correspondence), and results demonstrated a large increase in the
number of healthy food items selected. When the experimenters reversed to the initial
condition (reinforcement of verbalizations), the amount of nutritious snack choices
gradually decreased to near zero levels. Reinforcement for verbal/nonverbal
correspondence then was reinstated, and the schedule was gradually thinned from 100%
to 33% (i.e., reinforcement was delivered on 33% of the days) prior to the second
maintenance phase, during which the subject was asked to verbalize food choices but
received no reinforcement for either verbal or nonverbal choices. Following INT, the
subject's appropriate snack choice behavior maintained for 17 experimental sessions
conducted across a 7-week period.
However, two factors other than the PREE may be responsible for the findings of
Baer et al. (1987). First, a larger number of reinforcers was delivered prior to the second
maintenance phase (i.e., when reinforcement was reinstated and gradually leaned),
possibly enhancing resistance to extinction. Second, procedures implemented during the
two maintenance phases were different and could account for the varying levels of
resistance. During the first maintenance phase (following the FR-1 schedule),
reinforcement was delivered prior to snack time for correct verbalizations, whereas
during the second maintenance phase (following the INT schedule), no reinforcement
was delivered for either verbal or nonverbal snack choices. In the conditions immediately
preceding both maintenance phases, however, reinforcement was delivered after the
snack period for verbal/nonverbal correspondence. As a result, the first maintenance
phase contained a salient stimulus (reinforcer delivery for pre-snack verbalizations) that
was absent from both the reinforcement and second maintenance conditions.
Accordingly, the subject's behavior may have maintained for a longer period of time
during the second maintenance phase because the transition from reinforcement of
correspondence to extinction was less obvious (i.e., more difficult to discriminate) than
the transition from reinforcement to the first maintenance phase. In a similar study, Baer,
Williams, Osnes, and Stokes (1984) obtained maintenance of verbal/nonverbal
correspondence by simply delaying the reinforcement for verbalizations, and the authors
concluded that delivery of the reinforcer immediately following verbalizations functioned
to signal the termination of reinforcement for correspondence (i.e., extinction).
Although results of these studies replicate those of basic research demonstrating that
INT schedules can increase resistance to extinction, each contains limitations that prevent
clear interpretation of the findings in terms of this variable. Thus, further studies should
investigate the clinical significance of the PREE, particularly with inappropriate behavior.
Additional applied research on the PREE also should be conducted because the
relationship between reinforcement intermittency and resistance to extinction is actually
somewhat complex. A brief discussion of several complications that have emerged in the
basic literature and their relevance to applied research on the PREE is in order.
Most basic studies on the PREE, using a between-subject design, exposed separate
groups of subjects to different reinforcement schedules and, after averaging the responses
of individual subjects within each group, compared the performance of the different
groups during extinction. However, this design may not be practical in applied research
because the high-degree of intersubject variability common among humans (who
generally have varied and extensive reinforcement histories prior to the study) would
require the use of large subject pools. Further, results of between-group comparisons
may not be directly relevant to the behavior of individuals (Sidman, 1960). For these
reasons, applied studies on the PREE must use experimental designs that permit within-
subject comparisons of responding during extinction.
In the basic laboratory, however, many attempts to replicate the effect using a within-
rather than a between-subject design have failed (e.g., Adams, Nemeth, & Pavlik, 1982;
Cohen, Riley, & Weigle, 1993; Pavlik, Carlton, Lehr, & Hendrickson, 1967; Warren &
Brown, 1943). In some of these studies, subjects were exposed to different
reinforcement schedules alternated with extinction in a reversal design. More commonly,
subjects were exposed to rapidly alternating reinforcement conditions, each paired with a
distinct stimulus. For this design (i.e., the multielement or multiple-schedule design),
rapid alternation of the stimuli continued during extinction, and responding in the
presence of the different stimuli was compared to determine the effects of the different
reinforcement schedules. Although most studies using the reversal design have obtained
the PREE (e.g., Cohen et al., 1993; Wertheim & Singer, 1964), studies using the
multielement or multiple-schedule design often have reported a "reversed PREE" (i.e.,
greater resistance following FR 1 than INT; e.g., Adams et al., 1982; Flora & Pavlik,
1990; Mellgren & Elsmore, 1991) in addition to the conventional PREE (e.g., Hearst,
1961; Pavlik & Flora, 1993).
In several papers, Nevin (1974, 1988, 1992) suggested that the usual finding for
both within- and between-subject comparisons is the reversed PREE, particularly if the
data are transformed to adjust for differences in response rates associated with different
schedules of reinforcement. Rate of responding under INT schedules is generally much
higher than responding under FR-1 schedules. As such, Nevin has argued that
traditional measures of resistance, including response rate, number of responses, and
time to meet an extinction criterion, should not be compared following baselines with
INT and FR-1 schedules because the terminal acquisition performances are not equitable.
Instead, rate of decrease in responding (i.e., slopes of extinction curves) should be
examined, and data on the PREE should be expressed as a proportion of the response rate
during baseline or during the initial extinction sessionss.
Nevin (1988) used these measures of resistance to reanalyze data from several
previous studies on the PREE and found that responding after the FR-1 baseline was
consistently more resistant to extinction than responding after the INT baseline. He
suggested that rate of reinforcement rather than reinforcement intermittency actually
determines resistance to extinction. Thus, exposure to an FR-1 schedule, which often is
associated with a higher rate of reinforcement than INT schedules, should generally
increase resistance to extinction in the natural environment. Nevertheless, it also could be
argued that the traditional measures of resistance, such as number of responses and time
to reach an extinction criterion, are more relevant to applied problems than the proportion
of baseline (or initial extinction session) measure (Pavlik & Flora, 1993). For example,
practitioners treating dangerous behavior generally would be most concerned about the
total number of responses exhibited during the course of extinction.
Applied studies on the PREE must include consideration of the most appropriate
measures of resistance and experimental designs. As noted above, applied studies
probably should use within-subject designs, such as the reversal and multielement
designs. Nevertheless, each design contains potential limitations, which might be
attenuated with certain refinements in methodology.
Although the reversal design permits a direct comparison of responding during
extinction following exposure to each type of reinforcement schedule, results may be
confounded by sequence effects. The reversal design necessarily exposes the subject to a
history of reinforcement and extinction, which can influence the outcome in two ways.
First, repeated exposure to reinforcement might alter responding during subsequent
extinction phases. For example, resistance to extinction might increase as the subject is
exposed to an increasing number of reinforcers (cf. Perin, 1942). Although order effects
(i.e., history with specific conditions that influence the outcome) could be identified by
varying the order of reinforcement conditions across subjects, previous exposure to
reinforcement per se cannot be eliminated with this design. Second, repeated exposure to
extinction might alter responding during subsequent extinction conditions. In some
cases, for example, resistance to extinction will decline across successive extinction
phases (cf. Clark & Taylor, 1960).
Although the multielement, or multiple-schedule, design minimizes sequence
effects, it may be limited by interaction effects across conditions during either
reinforcement or extinction phases, a problem that has been encountered in basic studies
on extinction (e.g., Amsel, Rashotte, & Mackinnon, 1966; Rashotte, Ross, & Amsel,
1968). For example, conditions presented during one component of the multielement
design could influence responding in a different component, obscuring any differences in
the effects of INT vs FR-1 schedules. However, interaction effects across conditions of
the multielement design may be less likely to occur if the reinforcement schedules are
associated with highly salient stimuli (e.g., different therapists, responses, settings, or
times of day). In addition, sequence effects could be minimized in the reversal design by
keeping conditions as brief as possible.
Unlike the PREE, the effects of intermittent punishment have been examined in a
number of applied studies. Results of this research and implications for the use of
punishment in applied settings are discussed in the next section.
Intermittent Schedules of Punishment
Basic studies with both humans and nonhumans have examined the effects of a
variety of INT punishment schedules on behavior concurrently exposed to a schedule of
reinforcement Results generally indicate that amount of response suppression depends
on such factors as the type of punishment schedule, intensity of the punishing stimulus,
and the particular schedule of reinforcement maintaining the target response (e.g., Azrin,
1956; Bradshaw, Szabadi, & Bevan, 1979; Scobie & Kaufman, 1969; Zimmerman &
Ferster, 1963). In a review of this literature, however, Azrin and Holz (1966) concluded
that punishment should be delivered on an FR-1 schedule to be most effective.
Nevertheless, results of applied studies have demonstrated that some INT punishment
schedules are as effective as FR 1, particularly if the response has already been reduced
to a low frequency. Clark, Rowbury, Baer, and Baer (1973) conducted one of the first
systematic investigations of INT punishment schedules after successfully treating an 8-
yr-old girl's disruptive behavior with an FR-1 schedule of isolation timeout. Using a
reversal design, they examined three different VR schedules (VR 3, 4, and 8) and a
schedule that specified delivery of timeout for any response that followed the previous
one by less than 10 min (i.e., differential punishment of high rates [DPH]). Results
indicated that when the punishment schedule was no leaner than about VR 4, treatment
was nearly as effective as it was when timeout was delivered on an FR-1 schedule.
Results of subsequent studies examining INT schedules of isolation timeout (e.g.,
Calhoun & Lima, 1977; Calhoun & Matherne, 1975; Jackson & Calhoun, 1977) and
studies examining other punishing stimuli, such as electric shock (Kircher, Pear, &
Martin, 1971), lemon juice (Cipani, Brendlinger, McDowell, & Usher, 1991), verbal
reprimands (Acker & O'Leary, 1988), and leg slaps (Romanczyk, 1977), generally have
been consistent with the findings of Clark et al., suggesting that caregivers must use
fairly short schedules of punishment -- but not necessarily an FR-1 schedule -- to
effectively reduce problem behavior. For example, Romanczyk (1977) found that both
FR-1 and VR-5 schedules had similar suppressive effects on two subjects' stereotypic
Although most studies on INT punishment did not involve attempts to identify the
sources of reinforcement maintaining the target responses, delivery of punishment is
almost always confounded with the termination of reinforcement in applied research
(Iwata, Pace, Cowdery et al., 1994). Thus, studies examining INT schedules may have
combined punishment with extinction, a procedure that probably would increase the
efficacy of INT punishment (cf. Azrin & Holz, 1961). In the Clark et al. study, for
example, the subject's disruptive behavior may have been maintained by attention from
peers or caregivers, and treatment with INT punishment may have included the cessation
of contingent attention following all instances of disruption (both punished and
unpunished responses). If so, certain INT schedules (e.g., VR 3) may have been
effective because they were combined with extinction. On the other hand, leaner
schedules (e.g., VR 8) combined with extinction may not have been powerful enough to
produce significant decrements in responding (i.e., to hasten the extinction process).
Thus, the utility of punishment as treatment for severe behavior disorders maintained
by unknown or uncontrolled sources of reinforcement may be limited unless every
instance of the target response is followed by punishment. However, treatment programs
that require such close, constant monitoring of behavior may be difficult to implement in
settings with low staff-to-client ratios such as schools and residential facilities. It would
be beneficial if caregivers could use INT schedules of punishment yet still obtain
significant treatment effects. Although several authors have suggested that punishment
schedules might be gradually "thinned" to increase the efficacy of INT punishment (e.g.,
Cipani et al., 1991; Kazdin, 1994), only one study has examined such a procedure.
Barton, Brulle, and Repp (1987) implemented a DPH schedule of timeout to decrease
the aggression of three children in a classroom. The initial DPH schedule was
determined for each subject by calculating the mean baseline interresponse time (IRT) of
aggressive behavior. On the first day of treatment, the interval used for the DPH
schedule was equal to the mean baseline IRT, and this interval was adjusted daily
throughout treatment to reflect the mean IRT from the previous day. The DPH schedule
was implemented as follows: If the mean IRT was 2 min, the first target response that
occurred during every 2-min interval did not produce timeout. All subsequent responses
that occurred prior to the end of the interval were followed by the timeout procedure
(contingent observation for two subjects and exclusion timeout for the third subject). The
interval did not reset following each timeout; at the start of every consecutive 2-min
interval, the first instance of aggression was not followed by timeout. Results
demonstrated that treatment produced significant reductions in aggression for all subjects
as the DPH interval was lengthened across several weeks.
This adjusting DPH schedule had several advantages. Because the schedule was
based on the subject's current response rate, a high proportion of aggressive responses
probably was followed by timeout. In addition, this time-based schedule may have been
easier for the teachers to implement than other INT schedules, such as VR schedules (cf.
Clark et al., 1974). However, results of this study are limited in several respects. First,
all instances of aggression were followed by a verbal reprimand during treatment, a
procedure that may have established the reprimand as a conditioned punisher and
increased the efficacy of the DPH timeout schedule. In actuality, an FR-1 schedule may
have been in effect throughout the study because all responses appeared to be followed
by some type of punisher. Second, the study did not determine if gradual adjustment, or
leaning, of the INT schedule was necessary to obtain significant reductions in
aggression. That is, the terminal DPH schedule implemented for each subject may have
been effective at the outset of treatment. Third, the punishment procedure may have been
confounded with extinction. Finally, the utility of the DPH schedule was somewhat
limited because, like FR 1, it required continual monitoring of behavior.
Additional research should determine if INT punishment schedules can be gradually
"thinned" after behavior has been reduced by an FR-1 schedule so that initially ineffective
schedules can maintain low levels of problem behavior. The INT schedule should be
relatively easy to implement and, if possible, allow discontinuous monitoring of
behavior. For example, FI schedules, which have not yet been examined in applied
research on INT punishment, seem particularly well-suited for this type of procedure.
Monitoring the passage of time is probably less effortful than counting responses. In
addition, interval schedules may be more effective than ratio schedules during the course
of a fading procedure because nearly every response will be followed by punishment as
long as response rates remain low (i.e., the behavior has a lengthy IRT).
Statement of Purpose
The current series of studies was designed to examine the effects of INT
reinforcement and punishment on the efficacy of treatment (i.e., extinction or
punishment) for severe behavior disorders. After identifying the variables maintaining
problem behavior through functional analysis (Study 1), the effects of FR-1 versus INT
schedules of reinforcement during subsequent extinction were examined for subjects
whose target behaviors were maintained by social consequences (Study 2). To
investigate the PREE, the study examined two different within-subject designs and
several measures of resistance based on results of basic studies in this area. For subjects
whose behaviors were not maintained by social reinforcement, the effects of INT
punishment schedules on treatment with timeout or contingent restraint were examined
(Study 3). After identifying an FI schedule that failed to produce significant reductions in
behavior, an FR-1 schedule was gradually leaned in an attempt to maintain low levels of
problem behavior under the initially ineffective FI schedule. The next section contains a
description of general methods employed throughout the current series of studies.
Subjects and Setting
Five individuals diagnosed with profound mental retardation participated. All
subjects lived in a public residential facility for individuals with developmental disabilities
and were referred to a specialized program for the assessment and treatment of self-
injurious behavior (SIB) based on histories of severe and/or chronic SIB.
Brandon, a 32-year-old man, was referred for treatment due to an extensive history of
head hitting that had resulted in a cauliflower left ear. On occasion, Brandon would also
exhibit aggressive and disruptive behaviors. He displayed no expressive language but
was able to respond to a few simple requests. He could walk independently and had no
visual or auditory impairment. Throughout the course of this study, Brandon received
prescribed medication (chlorpromazine) for problem behavior, but no medication changes
were implemented until the completion of the experiment. Brandon participated in
Studies 1 and 2.
Sue was a 24-year-old woman whose SIB consisted of head hitting and hand biting.
She did not display expressive verbal skills but was able to respond to some simple
directions. She could walk independently and had no visual or auditory impairment Sue
received medication to control seizures during the course of this study. She participated
in Studies 1 and 2.
Harold was a 39-year-old man who had a variety of severe behavior disorders,
including SIB (head and body hitting, hand biting), aggression, and disruption. He had
some expressive verbal skills, although his speech was difficult to understand, and he
could respond to simple requests. Harold could walk with assistance although he was
confined to a wheelchair. He was blind due to cataracts but had no auditory impairment.
Harold received medication to control seizures during the course of this experiment. He
participated in Studies 1 and 2.
Paul was a 39-year-old man who had a long history of chronic hand mouthing, which
had resulted in some tissue damage. Paul displayed no expressive language but was
responsive to simple instructions. He could walk independently and had no visual or
auditory impairment. Paul participated in Studies 1 and 3.
Merry was a 31-year-old woman who engaged in chronic hand mouthing that had
produced extensive tissue damage. Merry displayed no expressive or receptive language
and was confined to a wheelchair. She had no apparent visual or auditory impairment
She received medication to control seizures during the course of the experiment. Merry
participated in Studies 1 and 3.
All sessions were conducted in therapy rooms of a day program located on the
grounds of the residential facility. Rooms contained tables and chairs, as well as
materials necessary for conducting certain conditions (see Procedures sections of the
various studies). At least one observer was present during all sessions.
Human Subjects Considerations
Human subjects approval for these studies was obtained from the University
Institutional Review Board and the facility where the day-treatment program was located.
In order to assess and treat SIB (and related problem behaviors), subjects were permitted
to freely engage in SIB for brief periods of time. Although SIB was likely to produce
extensive physical damage within a session for only one subject (Brandon), several
safeguards were established to reduce the risk of injury to all subjects. First, subjects
was monitored daily by medical personnel at their homes, and a physician was available
(via telephone or beeper) during sessions at all times. Second, session termination
criteria were established by medical personnel for Brandon (this was deemed unnecessary
for the rest of the subjects). Third, a foam pad was readily available during all sessions
to block potentially severe self-injurious responses, if necessary. With these safeguards,
it was felt the degree of risk to which subjects were exposed was no greater than that
found in their natural environment.
Response Measurement and Reliability
Response definitions were developed on the basis of staff interviews and informal
observations of the subjects prior to the study. Self-injurious responses were defined as
follows: face/head or body hitting (Brandon, Sue, Harold) -- forceful contact of an open
or closed hand with any part of the face, head or other body part (e.g., leg, chest); hand
biting (Sue, Harold) -- closure of the upper and lower teeth on the flesh anywhere on the
hand or wrist; hand mouthing (Paul, Merry) -- contact of the tongue with any part of the
hand or wrist, or insertion of any part of the hand or wrist between lips without biting.
Aggression (Harold) was defined as hitting, kicking, or biting the therapist, and
disruption (Harold) was defined as throwing objects or tearing clothing. Data also were
collected on the following subject and experimenter responses: compliance with
instructions; appropriate interaction with play materials; experimenter delivery of
attention, instructions, or materials; and experimenter delivery of timeout or manual
Observers collected data using a hand-held computer (ASSISTANT, model A102) that
audibly signaled 10-s intervals. Observers were graduate and undergraduate students
who had previously demonstrated proficiency with this type of data collection by
attaining a 90% agreement criterion for three consecutive sessions. Data were collected
using frequency or partial-interval recording, and the data were calculated as responses
per minute (Brandon, Sue, and Harold) or percentage of 10-s intervals scored (Paul and
Interobserver agreement was assessed by having a second observer simultaneously
but independently record data during 32% of all sessions (29% of sessions in Study 1,
35% of sessions in Study 2, and 29% of sessions in Study 3). In comparing observers'
records, session time was divided into consecutive 10-s intervals, and agreement
percentages were calculated on an interval-by-interval basis. The smaller number of
responses in each interval was divided by the larger number of responses. These
fractions were then summed across all intervals and divided by the total number of
intervals in the session to get the percentage of agreement between the two observers.
Mean agreement scores for SIB, aggression, or disruption were 98% overall (range, 94%
to 99%). Mean agreement scores were 98% during Study 1 (range, 95% to 99%), 96%
during Study 2 (range, 94% to 98%), and 98% during study 3 (range, 97% to 99%).
STUDY 1: FUNCTIONAL ANALYSIS OF PROBLEM BEHAVIOR
All subjects participated in Study 1, which involved an experimental (functional)
analysis to identify the variables maintaining SIB and other target behaviors (i.e.,
aggression and disruption for Harold). Subjects were repeatedly exposed to four
conditions presented within a multielement design, based on procedures described by
Iwata et al. (1982). In addition, two subjects (Brandon and Sue) were exposed to a fifth
condition (see "Materials" below) based on information obtained from interviews with
staff on their residence and through informal observations. All sessions lasted 15 min,
and 2 to 3 sessions were conducted per day for each subject, usually 4 to 5 days per
In this condition, the subject was placed in a therapy room with a variety of leisure
materials available (e.g., books, games, musical instruments). At the beginning of each
session, the experimenter said, "I am here if you need me," and sat away from the
subject. Throughout the session, the experimenter did not attend to the subject.
Contingent on the target behavior(s), however, the experimenter provided attention in the
form of statements of concern and disapproval (e.g., "Stop. You'll hurt yourself.") and
physical contact (e.g., patting the subject's back). This condition was designed to test
the effects of positive reinforcement (in the form of attention) on the rate of problem
The subject was placed in a therapy room with a variety of training tasks available.
The experimenter presented learning trials to the subject every 30 s using a graduated
prompting procedure (i.e., verbal instruction, modeling, physical guidance). The
instructions included academic and self-care tasks to which the subjects were typically
exposed on their residences (e.g., fold the towel, comb your hair, walk over here).
Praise and pats on the back were delivered contingent on compliance without physical
guidance. Contingent on the occurrence of the target response, the experimenter
terminated the trial by removing the materials and turning away for 30 s. If the subject
was exhibiting the target response at the time of the next scheduled trial, the experimenter
delayed the trial until the subject had not exhibited problem behavior for 5 s. This
condition was designed to test the effects of negative reinforcement (in the form of escape
from instructions) on the rate of problem behavior.
In this condition, the subject was placed in a therapy room with no training or leisure
materials available. No one was present in the room except an observer, and no
interaction with the subject occurred. This condition, intended to simulate a "barren"
environment, was designed to test the effects of automatic or self-stimulatory
reinforcement on the rate of problem behavior.
This condition was designed to test the effects of positive reinforcement in the form
of access to a specific item on the rate of SIB. Prior to the start of the session, the subject
was permitted access to a preferred item (a game for Sue and shoes for Brandon). At the
start of the session, the therapist removed the item. Contingent on each occurrence of
SIB, the subject was provided access to the item for 30 s.
This condition, intended to simulate an "enriched" environment, was designed to
serve as a control for the other conditions. The subject was placed in a therapy room
with a variety of leisure materials available. The experimenter delivered praise, physical
contact, and materials every 30 s. No instructions were delivered, and all instances of
problem behavior were ignored.
Results of the subjects' functional analyses are shown in Figures 1, 2, and 3. Data
for Brandon and Sue are shown in Figure 1. Both subjects exhibited the highest rates of
SIB in the Materials condition. For Brandon (top panel), SIB ranged from 0 responses
per minute (rpm) to about 6 rpm in the Materials condition. In the other conditions, SIB
ranged from 0 rpm to about 1 rpm. Sue (bottom panel) exhibited SIB exclusively in the
Materials condition. These results indicated that both subject's SIB was differentially
sensitive to positive reinforcement in the form of access to a particular item.
Results of Harold's functional analysis are presented in Figure 2. All topographies of
SIB (top panel), aggression (middle panel), and disruption (bottom panel) occurred
exclusively in the Demand condition with the exception of one play session. These
findings suggest that Harold's problem behaviors were members of the same response
class, all differentially sensitive to negative reinforcement in the form of escape from
Results for Paul and Merry are shown in Figure 3. For Paul (top panel), the highest
levels of hand mouthing occurred in the Alone condition (M = 74.4%; range = 48% to
91%). In all other conditions, SIB generally ranged from 0% to 23%. For Merry
(bottom panel), the highest amounts of hand mouthing occurred in the Demand (M =
67.7%; range = 71% to 80%) and Alone (M = 25.7%; range = 5% to 51%) conditions.
In all other conditions, SIB ranged from 0% to 19%. Results for both subjects suggested
Figure 1. Results of the functional analysis for Brandon (top panel)
and Sue (bottom panel).
10 20 30
0. a a a
5 10 15
Figure 2. Results of the functional analysis for Harold. Rates of SIB are shown
in the top panel, rates of aggression are shown in the middle panel, and
rates of disruption are shown in the bottom panel.
4 a -
I 4 810 12-
2 4 6 8 10 12
Figure 3. Results of the functional analysis for Paul (top panel)
and Merry (bottom panel).
5 10 15 20
that SIB was maintained by automatic reinforcement In addition, the findings for Merry
suggested that her hand mouthing may have been sensitive to negative reinforcement in
the form of escape from instructions.
Results of these functional analyses replicate the findings of previous studies (e.g.,
Iwata et al., 1982, 1994; Derby et al., 1992), demonstrating the utility of the
experimental analysis for revealing the variables maintaining problem behavior. The
patterns of responding observed during the experimental conditions and the subsequent
data interpretations were consistent with previous research in this area. Results of the
functional analyses indicated that the subjects' SIB was multiply determined: two
subject's SIB was maintained by positive reinforcement in the form of access to materials
(Brandon and Sue), at least one subject's SIB was maintained by negative reinforcement
in the form of escape from instructions (Harold), and two subject's SIB was maintained
by automatic reinforcement (Paul and Merry).
Results also showed that the experimental analysis is useful for identifying the
function of multiple response topographies exhibited by a single individual. Results for
Harold indicated that all topographies of SIB, aggression, and disruption were members
of the same functional response class. Data on these three categories of maladaptive
behavior were analyzed separately because results of a recent study suggested that the
functions) of multiple topographies might be obscured when data for all behaviors are
aggregated (Derby et al., 1994). However, the clear differentiation in Harold's data
indicated that the negative reinforcement function would have been identified for all
behaviors even if the results had been plotted in an aggregate fashion.
Results for two subjects (Brandon and Sue) demonstrated that the functional analysis
is useful for identifying relatively idiosyncratic sources of reinforcement for SIB, such as
access to specific items. An additional ("Materials") condition, designed to examine the
effects of a putative reinforcer identified via interviews with staff and informal
observations of the subjects, was easily included in the general assessment protocol.
Findings for both subjects clearly demonstrated the relevance of these specific items to
maintenance of SIB and underscore the importance of using this type of background
information to construct the conditions of the functional analysis.
Relative to the findings for the other subjects, results of Merry's assessment were
somewhat equivocal. Merry exhibited moderate amounts of hand mouthing throughout
the alone condition, suggesting that her behavior was at least partly maintained by
automatic reinforcement. However, she engaged in even higher amounts of hand
mouthing during the demand condition, indicating her behavior was differentially
sensitive to negative reinforcement in addition to automatic reinforcement (i.e., that her
hand mouthing was multiply controlled). However, further manipulations would have
been necessary to investigate the source of these unclear results (see Smith, Iwata,
Vollmer, & Zarcone, 1992). Although Merry's hand mouthing might have been multiply
controlled, results of her assessment also could have been obscured by interaction effects
across conditions. For example, some automatically-maintained SIB might have
occurred during the Demand condition. On the other hand, escape-maintained SIB would
not be expected to occur in the Alone condition because the relevant establishing
operation for escape is absent from the alone sessions. Thus, additional manipulations to
test the negative reinforcement hypothesis would have been necessary for Merry.
Because Merry's chronic hand mouthing had produced extensive tissue damage and
increased her risk for infection, it seemed important to treat her automatically-maintained
SIB prior to conducting any further assessment manipulations.
Study 1 served as a screening device. Based on these results, individuals either
participated in Study 2 to examine the PREE or in Study 3 to examine INT schedules of
STUDY 2: THE PARTIAL-REINFORCEMENT-EXTINCTION EFFECT AND
IMPLICATIONS FOR THE TREATMENT OF PROBLEM BEHAVIOR
Study 2 examined the PREE by exposing subjects to baseline conditions with FR-1
and INT reinforcement schedules, then comparing their performance during extinction.
Brandon, Sue, and Harold participated in Study 2 because results of Study 1 indicated
that their problem behaviors were maintained by social consequences (positive
reinforcement for Brandon and Sue and negative reinforcement for Harold).
Two daily sessions were conducted for each subject, usually 4 days per week.
Brandon and Sue were exposed to baseline (reinforcement) and extinction conditions
alternated within a reversal design. For Harold, the PREE was examined using a
multielement design, in which each reinforcement condition was associated with a
specific therapist, setting, and time of day.
During baseline conditions, the therapist used either FR-1 or INT schedules to deliver
the maintaining reinforcer following occurrences of the target responsess. Subjects
received five reinforcers during each session. The number of reinforcers (rather than
session time) was held constant across baseline sessions because results of studies
indicate that amount of reinforcement can influence resistance to extinction (e.g., Perin,
1942). For Brandon and Sue, procedures implemented during baseline sessions were
identical to those implemented during the Materials condition of the functional analysis.
For Harold, procedures implemented during the Demand condition of the functional
analysis were modified during baseline in two ways: (a) the contingent 30-s escape from
instructions was increased to 1 min, and (b) instructions were delivered continuously
throughout the session rather than on a FT-30 s schedule.
FR-1 baseline. The maintaining reinforcer was delivered following each occurrence
of SIB (for all subjects), aggression (for Harold), or disruption (for Harold).
INT baseline. During this condition, the reinforcement schedule was gradually
"thinned" across sessions until responding was maintained under a predetermined INT
schedule. A number of considerations influenced the choice of the terminal INT schedule
for each subject, including the types of schedules used in previous applied studies on the
PREE and those that appeared to be operating in the natural environment based on
informal observations of the subjects prior to the study. For Brandon, Sue, and Harold,
the terminal INT schedules were VR 6 (range, 4 to 8 responses), FR 3, and VR 10
(range, 5 to 15 responses), respectively. The VR schedules were constructed by writing
numbers (i.e., the predetermined response requirement range) on individual slips of
paper. Prior to each session, the response requirement for each reinforcement delivery
was determined by randomly choosing five slips of paper from the box and adjusting the
fifth number as necessary to ensure that the correct average was obtained. Harold was
exposed to the FR-1 and INT conditions concurrently. His FR-1 baseline sessions were
conducted during the morning by one therapist in a room at the day program. His INT
sessions were conducted during the afternoon by a different therapist in his residence
During these sessions, reinforcement was no longer delivered following occurrences
of the target behaviors. For Brandon and Sue, the preferred item was removed at the
start of the session, and all SIB was ignored. For Harold, the graduated prompting
sequence simply continued while all instances of inappropriate behavior were ignored.
Unlike baseline sessions, however, instructions were delivered on a FT-30 s schedule to
ensure that equal numbers of instructions were delivered across all extinction sessions.
Session length was determined for each subject by calculating the average baseline
session length. Sessions lasted 10 min for Brandon and Harold and 5 min for Sue.
The results of Study 2 are shown in Figures 4, 5, 6, and 7, and are summarized in
Table 1. Several different measures of resistance were examined for each subject. Data
from all sessions were calculated as responses per minute by dividing the total number of
responses by the number of minutes of session time. The total number of responses and
sessions that occurred during each extinction phase also were calculated, and the slopes
of linear regression lines fitted to the data from each extinction phase (using the method
of least squares) were compared. Finally, response rates during extinction sessions were
expressed as proportions of the baseline rate (from the immediately preceding baseline
condition) by dividing the response rate for each extinction session by the average
baseline response rate, which was calculated from the last 5 sessions.
Figure 4 shows Brandon's response rates during all baseline and extinction sessions.
Brandon was first exposed to the FR-1 baseline. Although a slight descending trend is
seen during this phase, responding remained fairly stable across the last 8 sessions (M =
2.8 rpm during the last 5 sessions). With the introduction of extinction, responding
increased and became more variable, then decreased to zero levels. The extinction phase
was terminated when Brandon had not exhibited SIB for four consecutive sessions, and
this performance level (no SIB for four consecutive sessions) was established as the
termination criterion for the subsequent extinction phase. Brandon exhibited 255 self-
injurious responses across 16 sessions (M = 1.6 rpm) before the first extinction phase
Figure 4. Rates of SIB for Brandon across sessions in Study 2.
20 40 60
The therapist then attempted to reimplement the FR-1 baseline to recapture responding
before leaning the schedule during the INT baseline phase. However, Brandon
continued to exhibit no SIB for the next several days (data not shown), and as a result,
his behavior did not come into contact with the altered contingency. Brandon was then
placed in a different therapy room at the treatment center. SIB abruptly reappeared in the
new therapy room, and responding gradually increased as the reinforcement schedule
was leaned to VR 6 (M = 8.3 rpm during the last 5 sessions). During extinction, SIB
rapidly decreased to zero levels, and Brandon exhibited only 100 self-injurious responses
across 9 sessions (M = 1.1 rpm) before satisfying the termination criterion.
Rate of change in responding during extinction was examined by calculating the
slopes of linear regression lines fitted to the data from each extinction phase. Results
showed that rate of decrease in responding during Brandon's second exposure to
extinction (slope = 0.6) was greater than that during Brandon's first exposure to
extinction (slope = 0.2). Thus, Brandon's data showed a "reversed PREE" based on
several different measures of resistance, including response rate, total number of
responses, number of sessions to meet an extinction criterion, and rate of change in
responding (slopes). That is, resistance to extinction was greater following exposure to
an FR-1 reinforcement schedule than following exposure to an INT schedule.
Results for Sue are displayed in Figure 5. The figure shows responses per minute of
SIB across all baseline and extinction sessions. Sue was first exposed to the FR-1
baseline condition, during which rates of SIB were extremely stable (M = 1.7 rpm during
the last 5 sessions). With the introduction of extinction, responding initially increased
and then rapidly decreased to zero. To minimize potential sequence effects, the extinction
criterion selected for Sue (i.e., 2 sessions in a row with SIB at or below 0.5 rpm) was
Figure 5. Rates of SIB for Sue across sessions in Study 2.
more lenient than that selected for Brandon. Sue exhibited just 43 self-injurious
responses across 8 sessions (M = 1.1 rpm) before satisfying this criterion. During the
INT baseline phase, rate of SIB increased as the schedule was leaned from FR 2 to FR 3
(M= 5.6 rpm during the last 5 sessions). The reintroduction of extinction produced a
pattern of responding similar to that observed in the first extinction phase (i.e., SIB
initially increased and then decreased to zero). However, Sue exhibited 301 self-
injurious responses across 16 sessions (M = 3.8 rpm) before satisfying the extinction
criterion. Thus, Sue exhibited seven times more self-injurious responses and required
twice as many sessions to meet the termination criterion during the second extinction
phase (following the INT baseline) than during the first extinction phase (following the
FR-1 baseline). Based on these measures of resistance (response rate, number of
responses, and number of sessions to meet an extinction criterion), Sue's data appeared
to demonstrate a PREE.
Rate of change in responding also was examined for Sue by calculating the slopes of
linear regression lines fitted to the data from each extinction phase. Results showed that
rate of decrease in responding during the second exposure to extinction (slope = 0.4)
was somewhat greater than that during the first exposure to extinction (slope = 0.2).
These data were indicative of a small "reversed PREE."
The next phases were designed to replicate the previous conditions and to investigate
the advantages of switching from an INT to an FR-1 schedule prior to treatment with
extinction. During the first phase, the therapist reimplemented the INT schedule, which
was leaned to FR 3, and then changed the schedule to FR 1. Rates of SIB during the
INT baseline sessions were similar to those observed during the initial INT phase, and
responding abruptly decreased with the transition from the INT to the FR-1 schedule.
Results of these phases, which replicated those obtained in the first part of the study,
showed that SIB maintained at a much higher level under the INT schedule (M = 4.6
rpm) than under the FR-1 schedule (M = 1.8 rpm). The introduction of extinction again
resulted in an initial increase in SIB followed by a rapid decrease to low levels. Sue
exhibited 41 self-injurious responses across 6 sessions (M = 1.4 rpm) before meeting the
extinction criterion. During the final phases for Sue, the INT baseline and extinction
conditions were reimplemented to compare responding during extinction after INT
reinforcement to responding during extinction after a switch from an INT to an FR-1
schedule. As the figure shows, rates of SIB during the final exposure to the INT
baseline and extinction conditions were similar to those observed during Sue's previous
exposures to these conditions. Sue exhibited 127 self-injurious responses across 11
sessions (M = 2.3 rpm) before meeting the extinction criterion. Thus, Sue exhibited
about three times as many self-injurious responses and required nearly twice as many
sessions to meet the termination criterion during the last extinction phase (following the
INT baseline) than during the third extinction phase (following a switch from INT to FR-
1 baseline conditions). Rate of change in responding during these two extinction phases
was again examined for Sue. Results showed that rate of change during the final
exposure to extinction (slope = 0.4) was slightly lower than that during the first
exposure to extinction (slope = 0.5). Thus, results of these replication phases appeared
to indicate that resistance to extinction was greater following INT reinforcement than
following a switch from an INT to an FR-1 schedule. These findings suggested that
treatment with extinction might be improved by switching from an INT to an FR-1
schedule prior to extinction.
Data for Harold are shown in Figure 6. The figure displays rates of inappropriate
behavior (SIB, aggression, and disruption) during the reinforcement and extinction
sessions with each therapist. Results showed that responding was much higher during
the INT reinforcement sessions (M = 7.1 rpm for the last 5 sessions) than during the FR-
1 sessions (M = 1.5 rpm for the last 5 sessions). The therapists simultaneously switched
Figure 6. Rates of SIB for Harold across sessions in Study 2.
20 40 60 80
to extinction in their respective settings. The extinction criterion selected for Harold was
SIB at or below 0.5 rpm for 3 consecutive sessions with both therapists. That is,
extinction would continue with each therapist until this criterion was met during both
morning and afternoon sessions. Results for the extinction condition showed that
responding following the FR-1 baseline initially increased and became more variable
before gradually decreasing to low levels. Following the VR 10 baseline, responding
gradually decreased to low levels. The termination criterion was satisfied after each
therapist had implemented 15 sessions. Harold exhibited 360 responses (M = 2.4 rpm)
during extinction sessions with the FR-1 therapist and 419 responses (M = 2.8 rpm)
during extinction with the INT therapist. Rate of change in responding during extinction
also was examined for Harold by calculating the slopes of linear regression lines fitted to
the data from sessions with each therapist. Results showed that rate of change in
responding following INT reinforcement (slope = 0.3) was greater than following FR 1
(slope = .06). Thus, as measured by response rates, total number of responses, and
number of sessions to meet the extinction criterion, resistance to extinction following the
INT baseline was not significantly greater than that following the FR-1 baseline.
Further, the rate of change measure (slopes) was indicative of a "reversed PREE" for
The next phases were designed to replicate the previous baseline and extinction
conditions. It could be argued, for example, that rate of behavior was higher under the
VR-10 baseline than under the FR-1 baseline because INT reinforcement was associated
with a specific therapist, setting, or time of day. Thus, the therapist initially associated
with the FR-1 baseline implemented an INT baseline (in the morning at the day program).
Results showed that rates of problem behavior increased when the reinforcement
schedule was gradually leaned to FR 6 (M = 6.2 rpm during the last 5 sessions), and
responding abruptly decreased and maintained at a low level (M = 1.8 rpm) when the
therapist switched to an FR-1 schedule. These findings replicated those of previous
phases showing that response rates were much higher under INT reinforcement than
under FR 1. With the introduction of extinction, responding gradually decreased to near
zero levels and remained low across seven sessions.
Overall, these findings suggested that a clear "reversed PREE" occurred for one
subject (Brandon), a potential PREE occurred for another subject (Sue), and no PREE
occurred for a third subject (Harold). As noted above, however, some authors have
argued that data on the PREE should be transformed to adjust for baseline differences in
responding associated with different reinforcement schedules (e.g., Anderson, 1963;
Nevin, 1988). The terminal rate of problem behavior for all subjects was consistently
much higher under the INT schedule than under the FR-1 schedule. For Sue, this
difference might partially account for the higher response rates, number of responses,
and total number of sessions observed during extinction following the INT schedule (cf.
Nevin, 1988). For Harold, these differences might obscure a clear "reversed PREE."
Figure 7 shows the data for each extinction session expressed as a proportion of the
baseline response rate. Higher proportions indicate greater resistance to extinction.
Results for Brandon are displayed in the top panel. As the figure shows, the proportion
of baseline measure for the FR-l/extinction sessions (M = 0.6) was consistently higher
than that for the INT/extinction sessions (M = 0.1).
The middle panel shows the results for Sue. The left figure displays the data from
Sue's first exposure to the two reinforcement schedules and extinction. The proportion
of baseline measure for the INT/extinction sessions (M = 0.6) was not consistently
higher than that for the FR-1/extinction sessions (M = 0.6). The right figure, which
shows the data from Sue's second exposure to the two reinforcement schedules and
extinction, indicates that the proportion of baseline measure for the INT/extinction
sessions (M = 0.4) was significantly lower than that for the INT-FR-1/extinction
Figure 7. Proportion of baseline measure for each extinction session in Study 2
for Brandon (top panel), Sue (middle panel), and Harold (bottom
EXT from FR 1
--- EXT from INT
sessions (M = 0.7). Thus Sue's data demonstrated no PREE during her first exposure to
extinction and a "reversed" PREE during her second, using the measure of resistance
recommended by Nevin (1988).
The bottom panel displays the proportion of baseline measure for each extinction
session during Harold's first exposure to the two reinforcement conditions and
extinction. Results showed that the proportion of baseline measure for the FR-1/
extinction sessions (M = 1.6) was consistently higher than that for the INT/extinction
sessions (M = 0.4), indicating a "reversed PREE" for Harold.
Results for all subjects are summarized in Table 1. In addition, rates of reinforcement
delivered under the FR-1 and INT schedules are shown in the table. As discussed above,
Nevin (1974, 1979, 1988) has suggested that reinforcement rate (rather than
intermittency er se) can determine resistance to extinction. For all subjects, the average
reinforcement rate under each reinforcement schedule was calculated by dividing the total
number of reinforcers delivered by the total minutes of session time. Reinforcement rates
were then compared to the various data interpretations.
Results for Brandon are shown in the top panel. All measures of resistance appear to
demonstrate a "reversed PREE" for Brandon, and rates of reinforcement delivered under
the two baseline conditions are somewhat consistent with this interpretation. That is,
reinforcement rate under FR 1 (M = 1.2) was slightly higher than that under INT
reinforcement (M = 1.0)
Results for Sue are shown in the middle panel. For the initial demonstration (left
side), several measures of resistance suggest the occurrence of a PREE (i.e., response
rate, number of responses, number of sessions to meet the extinction criterion), whereas
other measures of resistance indicate either no PREE (proportion of baseline measure) or
a "reversed PREE" (rate of change in responding). These equivocal results are consistent
with the finding that reinforcement rates were equivalent under the FR-1 and INT
Summary of Data During Extinction and Interpretation of Findings
Following FR 1 and INT Reinforcement
(PREE = partial-reinforcement-extinction effect; RPREE = "reversed" partial-
reinforcement-extinction effect; NPREE = no partial-reinforcement-extinction effect)
Mean Resp Rate
# of Responses
# of Sessions
Mean Prop of BL
Rate of Rfmnt
Mean Resp Rate
# of Responses
# of Sessions
Mean Prop of BL
Rate of Rfmnt
Mean Resp Rate
# of Responses
# of Sessions
Mean Prop of BL
Rate of Rfmnt
schedules (M = 1.7). For the second demonstration (right side), data on response rate,
number of responses, number of sessions, and rate of change in responding support a
PREE interpretation of the findings, whereas the proportion of baseline measure indicates
a "reversed PREE." Again, these equivocal results are consistent with finding that
reinforcers were delivered at equal rates under the two reinforcement conditions (M =
Results for Harold are displayed in the bottom panel. Although several measures of
resistance (i.e., response rate, number of responses or sessions) indicate no significant
differences in responding following FR-1 and INT baselines, rate of response decrease
(slopes) and the proportion of baseline measure suggest the occurrence of a "reversed
PREE." Further, reinforcement rates (0.7 under FR 1 and 0.4 under INT) are consistent
with the "reversed PREE" interpretation of Harold's data.
This study investigated the clinical significance of the PREE with individuals who
engaged in severe behavior disorders. The study examined two different within-subject
experimental designs and several measures of resistance. Results suggested that problem
behaviors, such as SIB and aggression, may not be more difficult to treat with extinction
if they have been maintained on INT rather than FR-1 schedules and that texts on
application may have overemphasized the potential for treatment difficulties generated by
the PREE. When traditional measures of resistance (e.g., response rate, number of
responses or sessions) were examined, a PREE was obtained with one subject (Sue), a
"reversed PREE" was obtained with another subject (Brandon), and no PREE was
obtained with a third subject (Harold). Results further demonstrated a "reversed PREE"
for two subjects (Brandon and Harold) and no PREE for a third subject (Sue) when rate
of decrease in responding (slopes of extinction curves) and the proportion of baseline
measure were examined.
Results of this study also demonstrated that rate of inappropriate behavior can be
extremely sensitive to changes in reinforcement schedule. For all subjects, responding
under INT reinforcement was consistently higher than responding under FR 1. Most
likely, this resulted in the higher initial response rates during extinction following the INT
baseline (see results for Sue and Harold). These findings suggest that switching from an
INT to an FR-1 schedule prior to treatment with extinction might lower the baseline
response rate, as well as the total number of responses exhibited during extinction. Such
an effect would be particularly desirable when treating severe behavior disorders, such as
SIB. For example, Sue exhibited only 41 self-injurious responses during extinction
when the INT schedule was switched to FR 1, but she exhibited 127 responses when the
subsequent extinction phase followed an INT baseline. Although recommended by
various authors as a means to attenuate or eliminate the PREE (see Ducharme & Van
Houten, 1994), this treatment str:negyv is based solely on response rate as a function of
the baseline reinforcement schedule.
Nevin (1979, 1988, 1992) has suggested that reinforcement rate rather than the
particular reinforcement schedule can determine resistance to extinction, and results of
this study appear to support this hypothesis. For two subjects (Brandon and Harold), a
comparison of the reinforcement rates delivered under the INT and FR-1 schedules
demonstrated that the schedule associated with the highest rate of reinforcement was
associated with the greatest resistance to extinction (using the measures of resistance
recommended by Nevin [19881). Such a finding indicates that reinforcement rate also
should be examined (and perhaps altered) in the natural environment before treating
problem behaviors with extinction.
Although results of this study appear to have important implications for the use of
extinction in applied settings, the findings should be considered preliminary due to a
number of potential limitations. For example, use of within-subject designs may have
significantly influenced the results. First, sequence effects associated with the reversal
design may have been responsible for the "reversed PREE" and no PREE demonstrated
for Brandon and Sue, respectively. That is, their first exposure to extinction (following
FR 1) may have led to a reduction in resistance during their second exposure to extinction
(following INT). Results of some basic studies indicate that resistance can decrease
across repeated exposures to extinction (e.g., Bullock & Smith, 1953; Clark & Taylor,
1960). In this study, extinction phases were kept as brief as possible to minimize the
potential influence of sequence effects. However, data for Brandon, who had the most
stringent extinction criterion, appear to suggest the occurrence of such effects.
Second, interaction effects across conditions of the multielement design may have
been responsible for the absence of a PREE for Harold. Because several basic studies
have obtained the PREE when salient stimuli were associated with the different
components of a multiple schedule (e.g., Feider, 1973; Waters & Knott, 1970),
conditions for Harold were paired with specific therapists, settings, and times of day.
Nevertheless, Harold met the extinction criterion simultaneously with both therapists,
suggesting that interaction effects may have occurred during extinction. That is,
exposure to extinction in the morning at the day program (following FR 1) may have led
to less resistance to extinction during the afternoon sessions at Harold's residence
(following INT). Such an effect is not merely speculative; results of basic studies using
the multiple-schedule design indicate that interaction effects can occur during extinction,
obscuring the PREE (e.g., Amsel et al., 1966).
Other factors, such as the reinforcement schedules used during the INT baselines,
also may have decreased the possibility of obtaining a significant PREE in this study.
For example, leaner reinforcement schedules, lengthier baseline phases, or different
extinction (termination) criteria may have altered the findings. Nevertheless, the
parameters implemented in this study were similar to those used in studies that have
obtained the PREE.
Additional research on the clinical significance of the PREE with severe behavior
disorders seems warranted. Studies should examine the effects of reinforcement
schedule and rate of reinforcement on responding during extinction, as well as the
benefits of switching from an INT to an FR-1 schedule prior to treatment Although
several studies have attempted to examine the benefits of altering reinforcement schedules
while treating problem behavior (e.g., Foxx & McMorrow, 1983; Neisworth, Hunt,
Gallop, & Madle, 1985; Schmid, 1986), conclusions about the effects of switching from
INT to FR-1 schedules prior to extinction cannot be formed on the basis of their findings.
In these studies, the contingencies maintaining subjects' inappropriate behavior
(stereotypy) were not identified, and it was assumed that the behaviors were maintained
by INT schedules of automatic reinforcement. Because sources of automatic
reinforcement are difficult to manipulate, arbitrary reinforcers (e.g., food items) were
delivered following each occurrence of stereotypy (i.e., on an FR-1 schedule) and then
removed in an attempt to decrease the behavior. Results suggested that the procedure
produced short-term reductions in stereotypy for some of the subjects. However, these
studies demonstrated the effects of introducing and removing an arbitrary reinforcer on
behavior maintained by an unidentified reinforcer, not the effects of switching
reinforcement schedules prior to extinction.
Further studies on the PREE also should attempt to determine which measures of
resistance have the greatest relevance to applied problems. Possibly, all measures (e.g.,
response rate, slope of extinction curves) can be important, depending on the situation.
For example, results of additional research might indicate that INT reinforcement is
associated with more responses during extinction but faster decrements in responding
(i.e., steeper extinction curve slopes) than FR 1. In this case, the FR-1 baseline may be
more desirable than the INT baseline when treating severe behaviors disorders, such as
SIB, but less desirable than the INT baseline when treating other types of problem
behaviors, such as mild forms of stereotypy and disruption.
Texts and articles on application (e.g., LaVigna & Donnellan, 1986; Romanczyk,
Kistner, & Plienis, 1982) generally describe extinction as a relatively inefficient treatment
procedure that may be associated with a number of undesirable side effects, such as initial
increases in response frequency (i.e., "extinction bursts") and aggression (i.e.,
"extinction-induced aggression"). As a result, extinction is rarely recommended as a
singular intervention for severe behavior disorders. Nevertheless, robust treatment
effects were obtained in this study by simply terminating the contingency between
responding and reinforcement during brief (5- or 10-min) sessions. For all subjects,
target behaviors were reduced to low levels within 16 sessions (range = 6 to 16
sessions), and few problems were noted with the exception of response bursting during
the initial stages of treatment. However, all extinction bursts were relatively brief and
tended to follow FR-1 rather than INT reinforcement baselines. These results are
consistent with those of previous studies demonstrating the utility of extinction as
treatment for problem behavior (e.g., Carr, Newsom, & Binkoff, 1980; Forehand, 1973;
France & Hudson, 1990; Iwata et al., 1990; Iwata et al., 1994; Mazaleski et al., 1993;
Salend & Meddaugh, 1985).
The current investigation may serve as an impetus for additional studies in this area.
Basic research findings suggest that a variety of factors other than reinforcement
schedules can influence performance during extinction (see Mackintosh, 1974, for a
review). Further examination of these factors, including reinforcer magnitude,
reinforcement delay, and response effort, could lead to the development of a
comprehensive technology for the use of extinction in applied settings.
STUDY 3: INCREASING THE EFFECTIVENESS OF
INTERMITTENT PUNISHMENT VIA SCHEDULE FADING
Study 3 examined a procedure to increase the efficacy of intermittent punishment for
behavior maintained by unknown or uncontrolled sources of reinforcement. Delivery of
punishment was gradually leaned while attempting to maintain low levels of SIB. Paul
and Merry participated in Study 3 because results of Study 1 indicated that their hand
mouthing was not maintained by social consequences.
Three to four sessions were conducted daily for each subject, 4 to 5 days per week.
All sessions lasted 15 min. When the therapist delivered the punisher (20-s timeout for
Paul and 15-s manual restraint for Merry), session time was stopped, and these intervals
were not included in the total session time.
The effects of FR 1 punishment on SIB were first examined by exposing subjects to
baseline and treatment conditions within a reversal design. After subsequently
identifying an ineffective FI punishment schedule, the FR-1 schedule was reimplemented
and gradually "thinned" every few sessions as long as responding remained low.
Baseline. During these sessions, Paul had continuous access to a hand-held
massager. No one interacted with him.
Timeout (FR I). Paul had continuous access to the massager; however, the therapist
removed the massager for 20 s contingent on each occurrence of hand mouthing. If Paul
engaged in hand mouthing during the last 10 s of the timeout period, the interval was
extended until he had not engaged in hand mouthing for 10 s.
Timeout (FI 5 min). The timeout procedure was implemented as described in the
previous section. However, the therapist removed the massager for 20 s contingent on
the first occurrence of hand mouthing after 5 min had elapsed since the last timeout (or
the start of the session). If Paul was already engaging in hand mouthing at the end of the
5-min interval, the therapist immediately removed the massager for 20 s. Thus, a
maximum of three time-out periods was possible during these sessions.
Timeout (fading). During this phase, delivery of timeout was faded from FR 1 to FI
5 min. Starting with an initial INT schedule of FI 30 s, the interval was lengthened by 30
s every time levels of hand mouthing were equal to or below 10% of the intervals for four
For all conditions, Merry was removed from her wheelchair and placed in a regular
chair prior to the start of the session.
Baseline. These sessions were identical to the Alone sessions of Merry's functional
Contingent restraint (FR 1). No one interacted with Merry during these sessions;
however, contingent on each occurrence of hand mouthing, the therapist removed
Merry's hand from her mouth and held both arms to her sides for 15 s, using the
minimum force necessary to keep Merry's hands stationary near her lap.
Contingent restraint (FI 2 min). Contingent restraint was implemented as described
in the previous section. However, the therapist implemented the restraint procedure
contingent on the first occurrence of hand mouthing after 2 min had elapsed since the last
restraint delivery (or the start of the session). If Merry was already engaging in hand
mouthing at the end of the 2-min interval, the therapist immediately implemented the
restraint contingency. Thus, contingent restraint could be delivered a maximum of seven
times during each session.
Contingent restraint (fading). During this phase, delivery of contingent restraint was
faded from FR 1 to FI 2 min. The initial INT schedule was FI 15 s, and the interval was
lengthened by 15 s or 5 s (see results) when levels of hand mouthing were equal to or
below 10% of the intervals for five consecutive sessions.
Results of Study 3 are shown in Figures 8 and 9 and are summarized in Table 2.
Results for Paul are displayed in Figure 8. Paul engaged in moderate but variable levels
of hand mouthing during baseline (M = 33%). Treatment with timeout (FR 1) produced
an immediate reduction in the behavior to low levels (M = 4.2%). With the removal of
timeout in the next phase, Paul's hand mouthing rapidly increased and maintained at a
moderate level (M = 44.2%). These findings indicated that an FR-1 schedule of timeout
was effective in treating Paul's hand mouthing.
When an FI 5-min schedule was then implemented, hand mouthing decreased and
then increased again to baseline levels (M = 26%). The reintroduction of timeout (FR 1)
again produced a reduction in hand mouthing (M = 4.4%), and the behavior remained
low as the timeout schedule was gradually leaned to FI 5 min. Across the 57 sessions of
the schedule fade, hand mouthing occurred in an average of 3.4% of the intervals. Hand
mouthing also remained low when Paul was exposed to the final FI 5-min schedule of
timeout (M = 2.4%), suggesting that the fading procedure increased the efficacy of an
initially ineffective INT punishment schedule.
Results for Merry are shown in Figure 9. During baseline, levels of hand mouthing
were variable and moderate ( M = 46%). Treatment with contingent restraint (FR 1)
resulted in an immediate decrease in SIB to low levels (M = 6.6%). Hand mouthing then
Figure 8. Percentage of intervals of hand mouthing for Paul across sessions
in Study 3.
o o 0 0
I0 I I I I I I
20 40 60 80 100
Figure 9. Percentage of intervals of hand mouthing for Merry across sessions
in Study 3.
BL FR 1 BL FI 2' RESTRAINT (FADE)
- C' |C a-
20 40 60 80 100 120
increased and maintained at moderate levels with the return to baseline (M = 55.7%).
These findings indicated that contingent restraint (FR 1) was effective in treating Merry's
When contingent restraint (FI 2 min) was implemented, levels of hand mouthing were
similar to those in baseline (M = 58.7%), and the reintroduction of contingent restraint
(FR 1) again produced significant decreases in SIB (M = 7.9%). Hand mouthing
remained relatively low as the schedule was leaned to FI 45 s, under which responding
became more variable and increased to baseline levels. FR 1 was then implemented to re-
establish treatment effects before attempting to fade a second time. Hand mouthing
decreased to low levels under FR 1; however, behavior increased dramatically during the
first session of INT punishment (FI 15 s) and maintained at moderate levels for the next
few sessions. As a result, FR 1 was again introduced before leaning the schedule more
gradually (i.e., using 5-s increments). During the third attempt to fade the schedule,
levels of hand mouthing remained low until FI 25 s, when responding became more
variable and maintained at moderate levels across 7 sessions. These findings suggested
that contingent restraint would not effectively treat Merry's hand mouthing unless nearly
every response was followed by punishment. Instead of attempting to lean the schedule
any further, FR 1 was reimplemented and Merry's participation in the study was
terminated. Across the final 14 sessions with FR 1, levels of hand mouthing remained
low (M = 3.3%).
Results showed that FR-1 punishment was effective for both subjects. Paul's hand
mouthing remained unchanged throughout a lengthy fading procedure, whereas Merry's
hand mouthing increased when the punishment schedule was leaned slightly. A closer
examination of the subjects' response patterns during punishment could provide at least
one explanation for these discrepant outcomes. Under the FI schedules, occurrences of
hand mouthing were not punished when their interresponse times (IRTs) were relatively
short (i.e., shorter than the interval used in the FI schedule). If Merry's responses often
immediately followed delivery of punishment, FI punishment schedules (and hence the
schedule fade) probably would be ineffective. That is, levels of hand mouthing would
likely increase as the schedule interval was lengthened, and the delivery of manual
restraint eventually could function as a discriminative stimulus for the temporary
discontinuation of the punishment contingency.
Data on the relative frequency (or distribution) of various IRTs from selected
treatment sessions were compared for the two subjects. Specifically, amount of time that
elapsed since the previous punishment delivery (or the start of the session if no punisher
had been delivered) was calculated for each occurrence of hand mouthing. Data from the
last 5 sessions of FR 1 (immediately prior to the start of the schedule fade) and from the
last 5 treatment sessions were included in the analysis. These data are summarized in
Table 2. The table shows the proportion of responses that followed the previous
punishment delivery (or start of the session) by specific amounts of time for Paul (left
side) and Merry (right side). Results for the last 5 sessions of FR 1 (sessions 50-54 for
Paul and 33-37 for Merry) are displayed in the second and fourth columns. Results
generally showed that a large proportion of Merry's hand mouthing was distributed
among the short IRTs (i.e., 0 s to 2 min) compared to Paul's hand mouthing. In fact, a
relatively large proportion of Merry's responses (37%) occurred within the first 20 s of
the previous punishment delivery. By contrast, a small percentage of Paul's hand
mouthing (10%) had such short IRTs.
Results for the last 5 treatment sessions (sessions 101-105 for Paul and 117-121 for
Merry) are shown in the third and fifth columns. When compared to the findings from
the first 5 FR-1 sessions, these data show changes in the distribution of the IRTs
following exposure to the schedule fade. The distribution of Merry's IRTs shifted
somewhat toward the shorter values (0 s to 2 min). By contrast, Paul's distribution of
Proportion of Responses that Followed Punishment Delivery
Within Specified Time Periods (in seconds)
IRTs generally shifted toward the larger values. As a result, the majority of Merry's
hand mouthing (85%) occurred within 2 min of the previous punishment delivery,
whereas the majority of Paul's hand mouthing (54%) followed the previous timeout by
more than 5 min.
This study examined the efficacy of gradually fading delivery of punishment with two
individuals who engaged in chronic hand mouthing not maintained by social
consequences. Results for one subject (Paul) demonstrated that an FI schedule of
punishment could be leaned while maintaining low levels of responding. Findings for
Paul further suggested that the fading procedure increased the effectiveness of an INT
schedule that had previously failed to suppress hand mouthing to low levels. Results for
the other subject (Merry) indicated that any useful FI punishment schedule probably
would be ineffective, despite repeated attempts to gradually lean the schedule beyond FR
These conflicting outcomes were not anticipated prior to the fading procedure because
a continuous schedule of punishment was equally effective for both subjects. Results
showed that FR-1 punishment with either contingent timeout (Paul) or manual restraint
(Merry) successfully treated hand mouthing. However, a closer examination of these
data indicated that Merry's hand mouthing (when it did occur) often immediately
followed the delivery of punishment or the start of the session, a pattern of responding
that might jeopardize the effectiveness of FI punishment schedules. Thus, an analysis of
an individual's IRTs under FR-1 punishment may indicate if delivery of punishment can
be successfully leaned using FI schedules.
Results for Merry suggested that her behavior also became more sensitive to the FI
punishment contingencies with repeated attempts to lean the schedule. Initially, hand
mouthing remained low until the FI 45-s schedule. During the second attempt to fade
delivery of punishment, levels of hand mouthing escalated under the shortest FI schedule
(FI 15 s). When the fading procedure was then introduced more gradually (i.e., the
interval was lengthened in 5-s increments), the punishment schedule could not be leaned
beyond FI 20 s. The shift toward shorter IRTs under FR 1 at the conclusion of the
experiment (as shown in Table 2) also suggests that Merry's hand mouthing had become
more sensitive to the FI punishment contingencies. Under FI schedules, responses with
lengthy IRTs (i.e., those longer than the interval used in the FI schedule) are selectively
punished. As a result, the frequency of responses with relatively short IRTs will likely
increase (cf. Galbicka & Branch, 1981). Eventually, delivery of punishment also could
function as a discriminative stimulus for "punishment-free" periods, leading to a gradual
overall increase in responding under FI punishment Basic studies on FI punishment
schedules have shown that response rates often are highest immediately following the
delivery of punishment and decelerate across the schedule interval (e.g., Azrin, 1956).
During FI sessions with high levels of responding, Merry typically would hand mouth
continuously throughout the interval, but response patterns occasionally resembled those
obtained in basic studies.
On the other hand, Paul's hand mouthing remained low across the schedule fade,
and an analysis of response patterns during the final FI-5 min sessions showed that the
IRTs were still fairly lengthy. Findings indicated that about 50% of his responses
followed the previous punishment delivery by greater than 5 min. As a result, the FI-5
min schedule actually resembled an FR-2 schedule. Despite extended exposure to FI
punishment, Paul's behavior remained insensitive to the specific FI contingencies (i.e.,
he failed to learn that responses would not be punished if they occurred during a
prespecified period following each timeout). The processes) responsible for this
outcome are not clear. During the fading procedure, certain features of the treatment
setting (e.g., presence of the massager or therapist) may have been established as
powerful discriminative stimuli for punishment. As such, these stimuli would set the
occasion for low levels of hand mouthing, and Paul's behavior would rarely contact the
absence of the punishment contingency. In this situation, the long-term effectiveness of
INT punishment might depend on tightly controlled treatment conditions, which are not
typical in clinical settings.
A number of variables could have influenced the efficacy of INT punishment for
Merry. For example, a different type of punishing stimulus might have produced
lengthier IRTs under FR 1 or competed more successfully with the ongoing schedule of
reinforcement for hand mouthing. Cipani et al. (1991) compared the effects of
continuous and INT schedules of punishment using two different punishment procedures
and found that both FR-1 and VR-4 schedules of contingent lemon juice effectively
suppressed the stereotypic behaviors of an autistic child. On the other hand, VR 4 was
not as effective as FR 1 when overcorrection (contingent manual guidance) was
implemented during a different phase.
A different type of punishment schedule also may have increased the likelihood of a
successful schedule fade for Merry. For example, Azrin (1956) found that a VI schedule
of contingent electric shock produced lower levels of key pecking in pigeons than an FI
schedule. Under VI schedules, the delivery of punishment is somewhat unpredictable
and thus less likely to be established as a discriminative stimulus for the temporary
discontinuation of punishment For Merry, FI schedules may have been ineffective
because she often engaged in hand mouthing soon after the delivery of manual restraint
(i.e., within 20 s). Some of these responses would have been punished if VI schedules
had been implemented throughout the fading procedure.
Alternatively, FI punishment schedules may have been effective for Merry following
lengthier exposure to FR 1. The fading procedure was initiated when hand mouthing
was low for five consecutive sessions. However, continued exposure to FR 1 (e.g., for
20 to 30 sessions) may have altered the distribution of IRTs or increased the efficacy of
INT punishment in some other manner. In fact, the criterion for attempting to lean the
schedule could have been based on the pattern of IRTs rather than on overall level of
responding. For example, the FR-1 schedule could have been changed for Merry when
50% or more of the IRTs were greater than 2 min for 5 consecutive sessions. In
addition, all subsequent steps in the fading procedure could have been based on the
current distribution of IRTs, a procedure similar to that used in Barton et al. (1987).
Such a strategy would ensure that a large proportion of responses is punished even as the
schedule is gradually leaned.
Results for Merry also may be specific to behavior that continues to receive
reinforcement during treatment An extinction component would likely increase the
efficacy of INT punishment or, at the least, enhance the fading procedure (cf. Azrin &
Holz, 1961). Although no studies have compared the effects of INT punishment with
and without extinction, previous studies obtaining significant reduction of problem
behavior with INT punishment likely included an extinction component in the treatment
program (see Iwata, Pace, Cowdery et al., 1994, for a discussion of this issue). If
reinforcement for hand mouthing had been eliminated in some manner, the punishment
schedule may have been leaned successfully for both subjects. In fact, the initial INT
punishment schedules (FI 5 min for Paul and FI 2 min for Merry) might have been
effective prior to the fading procedure. However, the primary purpose of this study was
to examine the use of INT punishment when treatment with extinction is impractical (as in
the case of behavior maintained by automatic reinforcement; cf. Vollmer & Iwata, 1993).
Treatment procedures involving INT punishment schedules generally are unnecessary
when the response-reinforcer relationship can be terminated.
Although results for Paul indicated that punishment schedules can be leaned in the
absence of extinction, additional research should examine factors that might increase the
efficacy of INT punishment (e.g., different fading procedures, punishing stimuli, or
schedules). In addition, the utility of the schedule fade and its role in altering the
effectiveness of INT punishment should be established. For example, the FI 5-min
schedule might have reduced Paul's hand mouthing following lengthy exposure to FR-1
punishment (instead of the gradual fade) or following a more rapid fading process (e.g.,
using 2-min rather than 30-s increments). The generality of the findings for Paul also are
somewhat limited because the procedure was implemented under tightly controlled
conditions rather than in the natural environment. Results suggest that INT punishment
may be effective in applied settings if all caregivers systematically lean the schedule in all
relevant contexts (work, home, school, etc.). However, additional studies must examine
the utility of this relatively complex strategy. Long-term maintenance of treatment under
INT punishment also must be assessed in further research.
Caregivers often may reject the use of procedures such as timeout and contingent
manual restraint on the grounds that they are relatively intrusive and too time-consuming
to implement (O'Brien & Karsh, 1990). Nevertheless, punishment procedures may be
essential when the reinforcers maintaining behavior cannot be identified or controlled, or
when substitute reinforcers cannot be found. In this study, timeout was used
infrequently with Paul after the punishment schedule was leaned to FI-5 min, and the
schedule permitted discontinuous monitoring of behavior. Such an outcome has
important implications for the acceptability of punishment in applied settings.
The current series of experiments examined the effects of INT schedules of
reinforcement and punishment on the treatment of problem behavior. Results indicated
that these schedules can influence responding in important ways and suggested various
strategies for the use of intermittent contingencies in applied settings. These findings are
noteworthy because problem behavior often is exposed to INT schedules in the natural
environment. Caregivers and teachers rarely have the time or resources to respond to
each occurrence of behavior. As a result, most behavior disorders are maintained on INT
reinforcement schedules prior to treatment (with the possible exception of behaviors that
pose immediate danger to the individual or others), and behavior often may be exposed to
INT punishment schedules when the treatment program involves a punishment
Study 2 compared the effects of FR-1 and INT reinforcement schedules on rates of
problem behavior prior to treatment and on the course of responding during extinction.
Results demonstrated that INT schedules were associated with higher rates of problem
behavior during baseline conditions, a finding that replicates those of previous studies
examining the effects of ratio schedules on response rate (e.g., De Luca & Holborn,
1992; Lovaas, Freitag, Gold, & Kassorla, 1965; Schroeder, 1972; Stephens, Pear,
Wray, & Jackson, 1975). Responding also was higher during the initial stages of
extinction following baseline with INT reinforcement. Thus, the benefits of switching
from an INT to an FR-1 schedule prior to extinction were investigated with two subjects.
Results suggested that the total number of responses exhibited during extinction may be
reduced if caregivers simply implement an FR-1 baseline for a short period of time prior
to treatment Although caregivers may be somewhat reluctant to implement a strategy that
involves deliberate reinforcement of problem behavior, the advantages may be clear when
dangerous behaviors are targeted for reduction.
Nevertheless, continuous reinforcement was not associated with a significant
reduction in response persistence during extinction. In fact, robust treatment effects were
obtained for all subjects following baselines with either INT or FR-1 schedules.
Although this finding appears to contradict a well-established phenomenon (i.e., the
PREE), results are consistent with those of basic studies using within- rather than
between-subject designs (e.g., Adams et al., 1982; Flora & Pavlik, 1990). Factors
responsible for the conflicting outcomes among basic studies are still relatively unclear.
However, results of studies using within-subject designs may be particularly vulnerable
to confounding by sequence or interaction effects.
Research findings on the PREE also are equivocal because "resistance" has been
defined and measured in a variety of ways. Most studies measured resistance to
extinction by calculating response rates, total number of responses, or amount of time to
meet an extinction criterion. In general, these studies demonstrated greater resistance to
extinction following INT than FR-1 schedules. Others have attempted to adjust for the
differences in responding associated with the different baseline conditions by calculating
the rate of change in responding during extinction or the proportion of baseline response
rates (cf. Nevin, 1988). Results of these studies indicated that FR-1 schedules were
associated with greater resistance to extinction than INT schedules. The current study
provides further data showing that reinforcement schedules can produce very different
outcomes, depending on the particular measure used to reflect "resistance." These
findings suggest that the relationship between reinforcement schedules and responding
during extinction is somewhat complex and that many texts and articles on application
present an incomplete account of current research findings on the PREE.
Study 3 examined the effects of INT punishment schedules on rates of problem
behavior. Results showed that certain INT punishment schedules will not significantly
alter responding maintained by automatic reinforcement, but that these schedules can be
effective following a process that involves gradually "thinning" or leaning the schedule
parameters. Although the factors responsible for the efficacy of this procedure were not
determined, results may have hinged on the development of powerful discriminative
stimuli for punishment (e.g., presence of the therapist or massager, features of the
therapy room). Behavior that is under tight stimulus control probably will be somewhat
insensitive to the specific intermittent contingencies. In particular, a high proportion of
responses will be punished if behavior remains low under FI schedules (resulting in a
dense schedule of punishment). The role of the fading procedure in the development of
such discriminative stimuli is unknown; thus, this procedure may be just one of
numerous strategies for increasing the efficacy of INT punishment.
Further investigation of this procedure is warranted. The generality and reliability of
the findings for Paul should be established with other subjects, punishment procedures,
and INT schedules. More important, factors that might alter the outcome of schedule
fading, including the distribution of IRTs prior to and during the fading process, should
be directly manipulated. Although the utility of punishment as treatment for severe
behavior disorders has been firmly established in the literature (Axelrod & Apsche, 1983;
Matson & DiLorenzo, 1984), these findings likely depended on the consistent application
of punishment procedures. As a result, the robust treatment effects demonstrated in these
studies may be somewhat difficult to replicate in the natural environment. The intrusive
nature of punishment perse also makes it a relatively unpopular treatment option among
caregivers and clinicians. Thus, results of Study 3 have important implications for both
the efficacy and acceptability of punishment in applied settings.
The current investigation joins a surprisingly small number of applied studies
examining the effects of INT reinforcement or punishment schedules. Although basic
research findings indicate that INT schedules could influence the treatment of problem
behavior, the clinical significance of these effects and the development of related
treatment strategies should be established by conducting further studies with human
behavior in applied settings.
Acker, M. M., & O'Leary, S. G. (1988). Effects of consistent and inconsistent
feedback on inappropriate child behavior. Behavior Therapy, 19, 619-624.
Adams, J. F., Nemeth, R. V., & Pavlik, W. B. (1982). Between- and within-
subjects PRE with sucrose incentives. Bulletin of the Psychonomic Society, 20, 261-
Adams, G. L., Tallon, R. J., & Stangi, J. M. (1980). Environmental influences on
self-stimulatory behavior. American Journal of Mental Deficiency, 85, 171-175.
Amsel, A., Rashotte, M. E., & Mackinnon, J. R. (1966). Partial reinforcement
effects within subject and between subjects. Psychological Monographs: General and
Applied, 80, 1-39.
Anderson, N. H. (1963). Comparison of different populations: Resistance to
extinction and transfer. Psychological Review, 70, 162-179.
Axelrod, S., & Apsche, J. (1983). The effects of punishment on human behavior.
New York: Academic Press.
Azrin, N. H. (1956). Some effects of two intermittent schedules of immediate and
non-immediate punishment. Journal of Psychology, 42, 3-21.
Azrin, N. H., & Holz, W. C. (1961). Punishment during fixed-interval
reinforcement Journal of the Experimental Analysis of Behavior, 4, 343-347.
Azrin, N. H., & Holz, W. C. (1966). Punishment. In W. K. Honig (Ed.),
Operant behavior: Areas of research and application (pp. 380-447). New York:
Azrin, N. H., Holz, W. C., & Hake, D. (1963). Fixed-ratio punishment. Journal
of the Experimental Analysis of Behavior, 6, 141-148.
Baer, R. A., Blount, R. L., Detrich, R., & Stokes, T. F. (1987). Using intermittent
reinforcement to program maintenance of verbal/nonverbal correspondence. Journal of
Applied Behavior Analysis, 20, 179-184.
Baer, R. A., Williams, J. A., Osnes, P. G., & Stokes, T. F. (1984). Delayed
reinforcement as an indiscriminable contingency in verbal/nonverbal correspondence
training. Journal of Applied Behavior Analysis, 17, 429-440.
Barton, L. E., Brulle, A. R., & Repp, A. C. (1987). Effects of differential
scheduling of timeout to reduce maladaptive responding. Exceptional Children, 53., 351-
Bijou, S. W. (1958). Operant extinction after fixed-interval schedules with young
children. Journal of the Experimental Analysis of Behavior, 1, 25-29.
Bradshaw, C. M., Szabadi, E., & Bevan, P. (1978). Effect of variable-interval
punishment on the behavior of humans in variable-interval schedules of monetary
reinforcement. Journal of the Experimental Analysis of Behavior, 29, 161-166.
Bradshaw, C. M., Szabadi, E., & Bevan, P. (1979). The effect of punishment on
free-operant choice behavior in humans. Journal of the Experimental Analysis of
Behavior, 31, 71-81.
Bullock, D. H., & Smith, W. C. (1953). An effect of repeated conditioning-
extinction upon operant strength. Journal of Experimental Psychology, 46, 349-352.
Calhoun, K. S., & Lima, P. P. (1977). Effects of varying schedules of timeout on
high- and low-rate of behaviors. Journal of Behavior Therapy and Experimental
Psychiatry, 8, 189-194.
Calhoun, K. S., & Matherne, P. (1975). The effects of varying schedules of
timeout on aggressive behavior of a retarded girl. Journal of Behavior Therapy and
Experimental Psychiatry, 6, 139-144.
Carr, E. G., & Durand, V. M. (1985). Reducing behavior problems through
functional communication training. Journal of Applied Behavior Analysis, 18, 111-126.
Carr, E. G., Newsom, C. D., & Binkoff, J. A. (1980). Escape as a factor in the
aggressive behavior of two retarded children. Journal of Applied Behavior Analysis, 13,
Church, R. M., & Raymond, G. A. (1967). Influence of the schedule of positive
reinforcement on punished behavior. Journal of Comparative and Physiological
Psychology, 62, 329-332.
Cipani, E., Brendlinger, J., McDowell, L, & Usher, S. (1991). Continuous vs.
intermittent punishment: A case study. Journal of Developmental and Physical
Disabilities, 3, 147-156.
Clark, H., Rowbury, T., Baer, A., & Baer, D. (1973). Timeout as a punishing
stimulus in continuous and intermittent schedules. Journal of Applied Behavior
Analysis, 6, 443-455.
Clark, F. C., & Taylor, B. W. (1960). Effects of repeated extinction of an operant
on characteristics of extinction curves. Psychological Reports, 6, 226.
Cohen, S. L., Riley, D. S., & Weigle, P. A. (1993). Tests of behavior momentum
in simple and multiple schedules with rats and pigeons. Journal of the Experimental
Analysis of Behavior, 60, 255-291.
Cowen, P. A., & Walters, R. H. (1963). Studies of reinforcement of aggression: I.
Effects of scheduling. Child Development, 34, 543-551.
Day, R. M., Rea, J. A., Schussler, N. G., Larsen, S. E., & Johnson, W. L.
(1988). A functionally based approach to the treatment of self-injurious behavior.
Behavior Modification, 12, 565-589.
De Luca, R. V., & Holborn, S. W. (1992). Effects of a variable-ratio reinforcement
schedule with changing criteria on exercise in obese and nonobese boys. Journal of
Applied Behavior Analysis, 25, 671-679.
Derby, K. M., Wacker, D. P., Peck, S., Sasso, G., DeRaad, A., Berg, W., Asmus,
J., & Ulrich, S. (1994). Functional analysis of separate topographies of aberrant
behavior. Journal of Applied Behavior Analysis, 27, 267-278.
Derby, K. M., Wacker, D. P., Sasso, G., Steege, M., Northup, J., Cigrand, K., &
Asmus, J. (1992). Brief functional assessment techniques to evaluate aberrant behavior
in an outpatient setting: A summary of 79 cases. Journal of Applied Behavior Analysis,
de Villiers, P. (1967). Choice in concurrent schedules and a quantitative formulation
of the law of effect. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook of operant
behavior, (pp. 233-287). Englewood Cliffs, NJ: Prentice-Hall.
Donnellan, A. M., & LaVigna, G. W. (1992). Myths about punishment. In A. C.
Repp & N. N. Singh (Eds.), Perspectives on the use of nonaversive and aversive
interventions for persons with developmental disabilities (pp. 33-57). Sycamore, IL:
Sycamore Publishing Co.
Dorsey, M. F., Iwata, B. A., Reid, D. H., & Davis, P. A. (1982). Protective
equipment: Continuous and contingent application in the treatment of self-injurious
behavior. Journal of Applied Behavior Analysis, 15, 217-230.
Ducharme, J. M., & Van Houten, R. (1994). Operant extinction in the treatment of
severe maladaptive behavior. Behavior Modification, 18, 139-170.
Feider, A. (1973). Within subjects partial reinforcement extinction effects for a bar
pressing task. Canadian Journal of Psychology, 27, 356-366.
Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. New York:
Filby, Y., & Appel, J. B. (1966). Variable-interval punishment during variable-
interval reinforcement. Journal of the Experimental Analysis of Behavior, 9, 521-527.
Fisher, W., Piazza, C., Cataldo, M., Harrell, R., Jefferson, G., & Conner, R.
(1993). Functional communication training with and without extinction and punishment.
Journal of Applied Behavior Analysis, 26, 23-36.
Flora, S. R., & Pavlik, W. B. (1990). Conventional and reversed partial
reinforcement effects in human operant responding. Bulletin of the Psychonomic
Society, 28, 429-432.
Forehand, R. (1973). Teacher recording of deviant behavior: A stimulus for
behavior change. Journal of Behavior Therapy and Experimental Psychiatry, 4, 39-40.
Foxx, R. M., & McMorrow, M. J. (1983). The effects of continuous and fixed
ratio schedules of external consequences on the performance and extinction of human
stereotyped behavior. Behaviour Analysis Letters, 3, 371-379.
France, K. G., & Hudson, S. M. (1990). Behavior management of infant sleep
disturbance. Journal of Applied Behavior Analysis, 23, 91-98.
Galbicka, G., & Branch, M. N. (1981). Selective punishment of interresponse
times. Journal of the Experimental Analysis of Behavior, 35, 311-322.
Hearst, E. (1961). Resistance to extinction functions in the single organism. Journal
of the Experimental Analysis of Behavior, 4, 133-144.
Holz, W. C. (1968). Punishment and rate of positive reinforcement. Journal of the
Experimental Analysis of Behavior, 11. 285-292.
Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S.
(1982). Toward a functional analysis of self-injury. Analysis and Intervention in
Developmental Disabilities, 2, 1-20.
Iwata, B. A., Pace, G. M., Cowdery, G. E., & Miltenberger, R. G. (1994). What
makes extinction work: An analysis of procedural form and function. Journal of Applied
Behavior Analysis, 27, 131-144.
Iwata, B.A., Pace, G.M., Dorsey, M. F., Zarcone, J. R., Vollmer, T. R., Smith, R.
G., Rodgers, T. A., Lerman, D. C., Shore, B. A., Mazaleski, J. L., Goh, H.,
Cowdery, G. E., Kalsher, M.J., McCosh, K. C., & Willis, K. D. (1994). The
functions of self-injurious behavior: An experimental-epidemiological analysis. Journal
of Applied Behavior Analysis, 27, 215-240.
Iwata, B.A., Pace, G.M., Kalsher, M.J., Cowdery, G.E., & Cataldo, M.F. (1990).
Experimental analysis and extinction of self-injurious escape behavior. Journal of
Applied Behavior Analysis, 23, 11-27.
Iwata, B. A., Vollmer, T. R., & Zarcone, J. R. (1990). The experimental
(functional) ihalysis of behavior disorders: Methodology, applications, and limitations.
In A. C. Repp & N. N. Singh (Eds.), Perspectives on the use of nonaversive and
aversive interventions for persons with developmental disabilities (pp. 301-330).
Sycamore, IL: Sycamore Publishing Co.
Jackson, J. L., & Calhoun, K. S. (1977). Effects of two variable-ratio schedules of
timeout: Changes in target and non-target behaviors. Journal of Behavior Therapy and
Experimental Psychiatry, 8, 195-199.
Kazdin, A. E. (1994). Behavior modification in applied settings (5th ed.). Pacific
Grove, CA: Brooks/Cole Publishing Co.
Kazdin, A. E., & Polster, R. (1973). Intermittent token reinforcement and response
maintenance in extinction. Behavior Therapy, 4, 386-391.
Kimble, G. A. (1961). Hilgard and Marquis' conditioning and learning (2nd ed.).
New York: Appleton-Century-Crofts, Inc.
Kircher, A. S., Pear, J. J., & Martin, G. L. (1971). Shock as punishment in a
picture-naming task with retarded children. Journal of Applied Behavior Analysis, 4,
Koegel, R. L., & Rincover, A. (1977). Research on the difference between
generalization and maintenance in extra-therapy responding. Journal of Applied Behavior
Analysis, 10, 1-12.
Lalli, J. S., Browder, D. M., Mace, F. C., & Brown, D. K. (1993). Teacher use of
descriptive analysis data to implement interventions to decrease students' problem
behaviors. Journal of Applied Behavior Analysis, 26, 227-238.
LaVigna, G. W., & Donnellan, A. M. (1986). Alternatives to punishment: Solving
behavior problems with non-aversive strategies. New York: Irvington Publishers, Inc.
Lovaas, O. I., Freitag, G., Gold, V. J., & Kassorla, I. C. (1965). Experimental
studies in childhood schizophrenia: Analysis of self-destructive behavior. Journal of
Experimental Child Psychology, 2, 67-84.
Luiselli, J. K. (1988). Comparative analysis of sensory extinction treatments for
self-injury. Education and Treatment of Children, 11, 149-156.
Mace, F. C., Lalli, J. S., & Lalli, E. P. (1991). Functional analysis and treatment of
aberrant behavior. Research in Developmental Disabilities, 12, 155-180.
Mace, F. C., Page, T. J., Ivancic, M. T., & O'Brien, S. (1986). Analysis of
environmental determinants of aggression and disruption in mentally retarded children.
Applied Research in Mental Retardation, 2, 203-221.
Mackintosh, N. J. (1974). The psychology of animal learning. New York:
Matson, J. L., & DiLorenzo, T. M. (1984). Punishment and its alternatives: New
perspectives for behavior modification. New York: Springer.
Mazaleski, J. L., Iwata, B. A., Vollmer, T. R., Zarcone, J. R., & Smith, R. G.
(1993). Analysis of the reinforcement and extinction components in DRO contingencies
with self-injury. Journal of Applied Behavior Analysis, 26, 143-156.
Mellgren, R. L., & Elsmore, T. F. (1991). Extinction of operant behavior: An
analysis based on foraging considerations. Animal Learning and Behavior, 19, 317-325.
Neisworth, J. T., Hunt, F. M., Gallup, H. R., & Madle, R. A. (1985). Reinforcer
displacement: A preliminary study of the clinical application of the CRF/EXT effect.
Behavior Modification, 9, 103-115.
Nevin, J. A. (1974). Response strength in multiple schedules. Journal of the
Experimental Analysis of Behavior, 21, 389-408.
Nevin, J. A. (1979). Reinforcement schedules and response strength. In M. D.
Zeiler & P. Harzem (Eds.), Advances in analysis of behaviour: Vol. 1. Reinforcement
and the organization of behaviour (pp. 117-158). Chichester, England: Wiley.
Nevin, J. A. (1988). Behavioral momentum and the partial reinforcement effect.
Psychological Bulletin, 103, 44-56.
Nevin, J. A. (1992). An integrative model for the study of behavioral momentum.
Journal of the Experimental Analysis of Behavior. 57, 301-316.
O'Brien, S., & Karsh, K. G. (1990). Treatment acceptability, consumer, therapist,
and society. In A. C. Repp & N. N. Singh (Eds.), Perspectives on the use of
nonaversive and aversive interventions for persons with developmental disabilities (pp.
503-516). Sycamore, IL: Sycamore Publishing Co.
Paisey, T. J. H., Whitney, R. B., & Hislop, P. M. (1990). Client characteristics
and treatment selection: Legitimate influences and misleading inferences. In A. C. Repp
& N. N. Singh (Eds.), Perspectives on the use of nonaversive and aversive interventions
for persons with developmental disabilities (pp. 175-197). Sycamore, IL: Sycamore
Pavlik, W. B., & Flora, S. R. (1993). Human responding on multiple variable
interval schedules and extinction. Learning and Motivation, 24, 88-99.
Perin, C. T. (1942). Behavior potentiality as a joint function of the amount of
training and the degree of hunger at the time of extinction. Journal of Experimental
Psychology, 30, 93-113.
Peterson, L. P. (1956). Variable delayed reinforcement. Journal of Comparative
and Physiological Psychology, 49, 232-234.
Pittenger, D. J., & Pavlik, W. B. (1988). Analysis of the partial reinforcement
extinction effect in humans using absolute and relative comparisons of schedules.
American Journal of Psychology, 101, 1-14.
Rashotte, M. E., Ross, M., & Amsel, A. (1968). Generalization of the partial
reinforcement effect. Psychonomic Science, 11, 173-174.
Repp, A. C., Felce, D., & Barton, L. E. (1988). Basing the treatment of stereotypic
and self-injurious behaviors on hypotheses of their causes. Journal of Applied Behavior
Analysis, 21, 281-289.
Rincover, A., Cook, R., Peoples, A., & Packard, D. (1979). Sensory extinction
and sensory reinforcement principles for programming multiple adaptive behavior
change. Journal of Applied Behavior Analysis, 12, 221-233.
Romanczyk, R. G. (1977). Intermittent punishment of self-stimulation:
Effectiveness during application and extinction. Journal of Consulting and Clinical
Psychology, 45, 53-60.
Romanczyk, R. G., Kistner, J. A., & Plienis, A. (1982). Self-stimulatory and self-
injurious behavior: Etiology and treatment. In J. J. Steffan & P. Karoly (Eds.),
Advances in child behavior analysis and therapy (pp 189-254). Lexington, MA:
Salend, S. J., & Meddaugh, D. (1985). Using a peer-mediated extinction procedure
to decrease obscene language. The Pointer, 30, 8-11.
Schmid, T. L. (1986). Reducing inappropriate behavior of mentally retarded
children through interpolated reinforcement. American Journal of Mental Deficiency, 91,
Schroeder, S. R. (1972). Parametric effects of reinforcement frequency, amount of
reinforcement, and required response force on sheltered workshop behavior. Journal of
Applied Behavior Analysis, 5, 431-441.
Scobie, S. R., & Kaufman, A. (1969). Intermittent punishment of human
responding maintained by intermittent reinforcement. Journal of the Experimental
Analysis of Behavior, 12, 137-147.
Sidman, M. (1960). Tactics of scientific research. New York: Basic Books, Inc.
Slifer, K. J., Ivancic, M. T., Parrish, J. M., Page, T. J., & Burgio, L. D. (1986).
Assessment and treatment of multiple behavior problems exhibited by a profoundly
retarded adolescent. Journal of Behavior Therapy and Experimental Psychiatry, 17, 203-
Smith, R. G., Iwata, B. A., Vollmer, T. R., & Pace, G. M. (1992). On the
relationship between self-injurious and self-restraint. Journal of Applied Behavior
Analysis, 25, 433-445.
Steege, M. V., Wacker, D. P., Berg, W. K., Cigrand, K. K., & Cooper, L. J.
(1989). The use of behavioral assessment to prescribe and evaluate treatments for
severely handicapped children. Journal of Applied Behavior Analysis, 22, 23-33.
Stephens, C. E., Pear, J. L., Wray, L. D., & Jackson, G. C. (1975). Some effects
of reinforcement schedules in teaching picture names to retarded children. Journal of
Applied Behavior Analysis, 8, 435-447.
Sturmey, P., Carlsen, A., Crisp, A. G., & Newton, J. T. (1988). A functional
analysis of multiple aberrant responses: A refinement and extension of Iwata et al.'s
methodology. Journal of Mental Deficiency Research, 32, 31-46.
Taylor, J. C., & Carr, E. G. (1992). Severe problem behaviors related to social
interaction: I. Attention seeking and social avoidance. Behavior Modification, 16, 305-
Vollmer, T. R. (1994). The concept of automatic reinforcement: Implications for
behavioral research in developmental disabilities, Research in Developmental Disabilities,
Vollmer, T. R., & Iwata, B. A. (1993). Implications of a functional analysis
technology for the use of restrictive behavioral interventions. Child and Adolescent
Mental Health Care, 3, 95-113.
Vollmer, T. R., Iwata, B. A., Zarcone, J. R., Smith, R. G., & Mazaleski, J. L.
(1993). The role of attention in the treatment of attention-maintained self-injurious
behavior: Noncontingent reinforcement and differential reinforcement of other behavior.
Journal of Applied Behavior Analysis, 26. 9-21.
Wacker, D. P., Steege, M. W., Northup, J., Sasso, G., Berg, W., Reimers, T.,
Cooper, L., Cigrand, K., & Donn, L. (1990). A component analysis of functional
communication training across three topographies of severe behavior problems. Journal
of Applied Behavior Analysis, 2, 417-429.
Warren, A. B., & Brown, R. H. (1943). Conditioned operant response phenomena
in children. Journal of General Psychology, 28, 181-207.
Waters, W., & Knott, P. D. (1970). Tests of frustration theory extended to the
generalized partial reinforcement effect. Psychonomic Science, 20, 61-62.
Weeks, M., & Gaylord-Ross, R. (1981). Task difficulty and aberrant behavior in
severely handicapped students. Journal of Applied Behavior Analysis, 14, 449-463.
Wertheim, G. A., & Singer, R. D. (1964). Resistance to extinction in the goldfish
following schedules of continuous and variable interval reinforcement. Journal of the
Experimental Analysis of Behavior, 7, 357-360.
Zarcone, J. R., Iwata, B. A., Smith, R. G., Mazaleski, J. L., & Lerman, D. C.
(1994). Reemergence and extinction of self-injurious escape behavior during stimulus
(instructional) fading. Journal of Applied Behavior Analysis, 27, 307-316.
Zimmerman, J., & Ferster, C. B. (1963). Intermittent punishment of S^ responding
in matching-to-sample. Journal of the Experimental Analysis of Behavior. 6, 349-356.
Dorothea C. Lerman was born in Daytona Beach, Florida, on February 16, 1966.
She earned a Bachelor of Science degree in psychology from the University of Florida in
April 1988. Although she took a variety of courses, Dorothea was not exposed to
behavior analysis until her senior year, when she met Dr. Brian Iwata and asked him to
supervise her senior thesis. As a direct result of this experience and Brian Iwata's
encouragement, she decided to enter the field of applied behavior analysis. Dorothea
worked for two years as a program manager at a sheltered workshop for individuals with
developmental disabilities before enrolling at the University of Florida as a graduate
student in psychology (experimental analysis of behavior) in 1990. She has participated
in research activities throughout graduate school and served as a teaching assistant and
instructor. Dorothea's research has emphasized analysis and treatment of severe behavior
disorders in individuals with developmental disabilities. After graduation, she plans to
pursue a career in behavior analysis, including teaching and research.
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
an A. Iwata, hairprson
Professor of Psycholy
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy. ,
Timothy D. fickenberg
assistant Professor of Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Professor of Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy. i .
Shari A. Ellis
Assistant Professor of Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Cecil D. Mercer d
Professor of Special Education
This dissertation was submitted to the Graduate Faculty of the Department of
Psychology in the College of Liberal Arts and Sciences and to the Graduate School and
was accepted as partial fulfillment of the requirements for the degree of Doctor of
Dean, Graduate School
UNIVERSITY OF FLORIDA
I 1 111 I IIII5 91 III
3 1262 08553 9012