Available online at www.sciencedirect.

com

ScienceDirect
Behavior Therapy 47 (2016) 720 – 732
www.elsevier.com/locate/bt

An Operant Pursuit of Generalization – Republished Article

Trevor F. Stokes ⁎
Pamela G. Osnes
Florida Mental Health Institute, University of South Florida

ior change according to the goals of the therapist

We outline some principles and tactics which are recom- and the client. The behaviors focused upon during
mended as likely to facilitate the occurrence of generaliza- those changes, their specificity or breadth, are
tion and maintenance in programs of clinical importance. In determined largely by the nature of the problems
general, clinicians and researchers would do well to presented to the therapist and by the power and
implement and analyze procedures that follow the general- limitations of the therapist's analytic and interven-
ization programming principles of exploiting current tion techniques. It often is assumed that the
functional contingencies, training diversely, and incorpo- technology to effect the targeted changes is readily
rating functional mediators. More specifically, the tactical available, and it is. There are, however, some gaps
armamentarium should include contacting natural conse- in our ability to accomplish desired changes. Some
quences, recruiting natural consequences, modifying mal- of our shortcomings relate to limitations in dealing
adaptive consequences, reinforcing occurrences of with certain behaviors, some to the limitations of
generalization, using sufficient stimulus exemplars, using procedures, and some to limitations to our ability to
sufficient response exemplars, making antecedents less effect widespread and well maintained changes.
discriminable, making consequences less discriminable, The success in many behavior therapy endeavors
incorporating common salient physical stimuli, incorporat- may be analyzed in terms of the direct behavior
ing common salient social stimuli, incorporating changes that occur. There are many programs,
self-mediated physical stimuli, and incorporating however, that require more comprehensive and
self-mediated verbal and covert stimuli. durable changes in order for a demonstration of
success. This endeavor relates specifically to the
issues of generalization.
What is Generalization?
THERE CLEARLY IS ONE major focus in clinical The concept of generalization has a long history. An
behavior therapy: To accomplish beneficial behav- operant perspective was developed by Skinner
(1938, 1953). He emphasized the effect of stimulus
The authors thank Barry Edelstein for his many discussions with consequences on the frequency of the behaviors
us concerning the topic of generalization and for his encouragement those consequences followed. Antecedent stimuli,
to write this paper. Preparation of this paper is supported m part by which occur prior to behaviors and effect them,
U.S. Department of Education Grants #0008630340 and
#H024A80037 to the authors. Requests for reprints should be were also considered. The influences of these
sent to Trevor Stokes or Pamela Osnes, Florida Mental Health various stimuli were the result of the contingency
Institute, University of South Florida, 13301 Bruce B. Downs relationships between the antecedent stimuli, the
Boulevard, Tampa, FL 33612.
This article is a reprint of a previously published article. For behaviors, and the consequences for the occurrence
citation purposes, please use the original publication details; of the behaviors. The operant focus related to the
Behavior Therapy, 20 (1989), pp 337-355. fact that the behavior in question operated on the
DOI of original item: http://dx.doi.org/10.1016/
S0005-7894(89)80054-1. environment to produce or cause the stimulus
⁎ Corresponding author. consequence to occur. This influenced the frequen-
cy of occurrence or probability of the behavior (the
0005-7894/© 2016 Association for Behavioral and Cognitive Therapies.
Published by Elsevier Ltd. All rights reserved.
operant) or (more likely) the class of behaviors.
 generalization 721

If training or a program of intervention occurred tion appeared not to define or specify functional
in the presence of one set of circumstances (stimulus relations, in the traditional sense of an operational
class), that training may have an effect on respond- definition (Stokes & Osnes, 1988). There has been
ing in the presence of similar stimuli. This was discussion and controversy related to this topogra-
called generalization or induction (Skinner, 1953, hical rather than functional characterization (see
p. 132). As Skinner noted, "generalization is not an Baer, 1982; Johnston, 1979), but it reflected the
activity of the organism; it is simply a term which generalization and maintenance literature of the
describes the fact that the control acquired by a time. Much of that literature did not attend to the
stimulus is shared by other stimuli with common analysis of functional controlling variables. The
properties" (1953, p. 134). Another effect of some categories described provided an inductive guide
importance is that when one operant (a behavior) is through a frequently unanalyzed and unproven
reinforced, there is an increase in the frequency of maze of accumulated wisdom and data within
other behaviors without their being directly rein- applied behavior analysis and therapy endeavors.
forced. Skinner (1953) noted that this has been The Stokes and Baer (1977) characterization
variously referred to as "response generalization, proposed that after a behavior was taught or
transfer, or response induction (p. 93f)." Tradi- modified through the operations of reinforcement,
tional experimental psychology has documented punishment, and extinction, it may also occur in
such effects with measurement precision, demon- conditions different to those of the specific training.
strating the functional variables controlling the To advance our knowledge of generalization
generalization. For example, this is noted in a programming, it is important for researchers to
stimulus generalization gradient which specifies the describe the dimensions of their analyses and the
dimension of stimulus control. The technical scope of their generalization assessment. Two
precision and contingency control allowed in important questions are extant for any program:
applied behavior analysis and therapy is not as Did the behavior occur in generalized circum-
well developed, but the pursuit of generalization stances, and what are the functional variables
outcomes and the analysis of functional variables is which account for that generalization? As clinicians
nonetheless important. and researchers in the field appreciate, we are far
As the term that has maintained best in applied better at documenting the former than we are at
work, generalization is the term of choice in this demonstrating the latter.
paper. Reflecting the applied perspective, the
description of generalization here refers to the Principles and Tactics of Generalization
outcome of behavior change and therapy programs, Programming
resulting in effects extraneous to the original Stokes and Baer (1977) characterized the literature
targeted changes. This occurs in the absence of in applied behavior analysis and therapy as it
comprehensive programming across stimuli, re- related to the assessment and programming of
sponses, and time. In therapeutic activities, these generalization. Some of the categories they pro-
effects are sought across clients, stimulus conditions posed to describe the literature were not in fact
and settings, and behaviors. The durability of classifications of research related to the principles
effects across time is referred to as maintenance. of programming generalization. Train and hope,
Of course, a description of therapeutic outcome for example, was a reflection of the frequently
along these dimensions may be useful to document occurring consideration of generalization which
success of clinical interventions. However, descrip- included some measurement of generalization
tion alone may be insufficient if the science of outcome but did not specifically program or
human behavior is to benefit similarly. What is analyze the occurrence of generalization. Similarly,
relevant is the analysis of precisely how the sequential modification was more a description of
generalized and maintained changes were accom- typical experimental designs (usually within-subject
plished and how the variables producing that multiple baseline designs) than it was related to
generalization and maintenance can be functionally procedures for the programming of generalization.
related to the behavior change outcome. The category, natural communities of reinforce-
In their 1977 paper, Stokes and Baer described ment, is more difficult to classify as an
generalization as "the occurrence of relevant assessment-only procedure or as a specific tactic,
behavior under different nontraining conditions because it deals with naturally occurring direct
(across subjects, settings, behaviors, and/or time) consequences. Yet, even if it is not a legitimate
without the scheduling of the same events in those principle, there is much about the procedures that
conditions as had been scheduled in the training should be recommended (Baer, 1982; Stokes &
conditions (p. 350)." This topographical descrip- Osnes, 1986, 1988).
722 stokes & osnes

We have considered the classification system of examination. Contact with, and exploitation
Stokes and Baer (1977) and have tried to refine it by of, natural contingencies may also require
focusing upon the principles of generalization some initial therapeutic manipulations such
programming. Presented next is a description of as teaching the behaviors that may then
twelve programming tactics grouped into three contact the maintaining or decelerative con-
areas of general principles. Some selected research sequences or prompting an available yet
that represents each of the tactics will be described. inactive community of consequences to fol-
The sophistication of the experimental analyses of low the relevant behavior in a functional
the programming strategies described may vary. way.
The categories are as follows:
1. Contact natural consequences. Many conse-
A. Exploit Current Functional Contingencies: quences function to increase, maintain, or
1. Contact natural consequences. decrease the behavior they follow. These
2. Recruit natural consequences. consequences may be naturally occurring or
3. Modify maladaptive consequences. they may be artificially programmed by a
4. Reinforce occurrences of generalization. behavior change agent. If the existing conse-
quences that follow certain behaviors may be
B. Train Diversely: contacted in a functional way, then a power-
5. Use sufficient stimulus exemplars. ful entrapment for that behavior may occur.
6. Use sufficient response exemplars. Of course, these entrapment consequences
7. Make antecedents less discriminable. may increase and maintain either appropriate
8. Make consequences less discriminable. or inappropriate behaviors.

C. Incorporate Functional Mediators: Reinforcers function to increase the frequency of

9. Incorporate common salient physical stimuli. the behaviors they follow. Perhaps the most
10. Incorporate common salient social stimuli. fundamental guideline of behavior programming,
11. Incorporate self-mediated physical stimuli. as well as generalization programming, is to teach
12. Incorporate self-mediated verbal and covert behaviors that are likely to come into contact with
stimuli. powerful reinforcing consequences that do not need
to be programmed by a therapist or behavior
Now let us examine each of these principles more change agent. It may be very difficult to analyze
closely. experimentally and demonstrate that the conse-
quences are in fact reinforcers maintaining the
A. Exploit Current Functional Contingencies: A relevant behavior. In addition, as with all opera-
functional contingency is an arrangement of tions of reinforcement, individual variation as a
antecedents, responses, and consequences result of the consequences manipulations may be
that affects the frequency, magnitude, or observed.
duration of the relevant behavior (Bijou and For example, Paine, et al. (1982) examined the
Baer, 1978; Stokes, 1985). There are numer- maintenance of social interaction among with-
ous contingencies that occur naturally in the drawn elementary school children and their peers.
environment, some being more active than Maintenance was examined during repeated rever-
others. For example, touching a hot stove by sals subsequent to interventions using peer-involved
accident is a likely natural contingency which tutoring and a point system. Some of the children
will act as a powerful punisher with persis- showed increased social behavior during each
tent effects. When contingencies of reinforce- successive baseline condition, suggesting some
ment, punishment, and extinction, are entrapment and peer reciprocal control over the
operating potently without artificial manip- social interaction. Some children did not display
ulation there is an obvious advantage to such effects. Apparently, a relevant variable was the
exploiting their effects. Many of these con- extent of previous training. This may have been
tingencies are positive and have been referred related to the degree of facility of the skill, i.e., the
to as natural communities of reinforcement ability to contact the existing natural consequences
(Stokes and Baer, 1977). Natural communi- that would reinforce the performance of the
ties of punishment are also likely to exist for relevant social behaviors.
certain behaviors, but the documentation and The use of normalized least intrusive conse-
analysis of those contingencies are far less quences in programming is also relevant here.
extensive and deserving of more thorough Generalization programming seems to be well
 generalization 723

served by providing the least artificial, least the recruitment strategies, the dormant envi-
cumbersome, and most natural positive conse- ronment may be recruited to provide some
quences in programming interventions. Such pro- potent consequences. Some people are quite
gramming most closely matches naturally prepared to respond positively, yet their
occurring consequences and their entrapment powers of observation are inadequate to
potential. notice and pay off the appropriate behavior
The relationship between target behavior and the that they wish to develop. Active entrapment
reinforcer that follows its occurrence, was exam- of natural consequences by the recipient of
ined by Koegel and Williams (1980). Three young the treatment may be a suitable tactic in such
severely autistic children who displayed low rates of circumstances.
acquisition of new behaviors were studied under
two response-reinforcer contingencies. The first, For example, Stokes, Fowler, and Baer (1978)
which was an indirect relationship, involved a examined the interactions of both normal and
frequently typical practice of the delivery of the behavior disordered preschoolers with their
reinforcer by an external agent handing it to the teachers. These four-to-six year old children were
child following the performance of the behavior. taught first to increase the quality of their academic
The second, which was a direct relationship, performance. Then they were taught skillful ways
involved a sequence in which the child's perfor- to draw their teachers' attention to that work
mance of the behavior itself produced the reinforcer through the use of infrequent cueing statements
as a natural outcome, e.g., picking up an object such as, "How is this?" After this intervention, the
resulted in obtaining the consequence which had children successfully recruited positive attention
been placed under it. The results showed that and praise from their teachers. This previously
acquisition of the behaviors was more quickly dormant community of reinforcement responded
demonstrated when the target behaviors were part consistently and positively in a manner likely to
of the response chain leading to the reinforcer maintain the improved work and the cues that
rather than when the consequence was artificially facilitated increased positive attention to academic
provided in a manner less directly involved in a productivity. One other appealing feature of these
natural response-reinforcer relationship. These data procedures is that the children themselves became
suggest that generalization programming tactics active agents of their own treatment, rather than
should include both the choice of relevant behaviors being the recipient of procedures implemented and
that will meet natural consequences, but also controlled primarily by others.
attention to the naturalness and directness of the The maintaining effects of the recruitment of
contingencies of training. consequences was also examined by Mank and
There is a paucity of information about natural Homer (1987). Six persons with severe disabilities
communities of punishment, perhaps owing in large were taught to self-monitor their work performance
part to the ethical considerations which would and productivity in vocational settings. Subsequent-
make such investigations difficult. Given these ly, the clients were trained to evaluate their
circumstances, a conceptually-based discussion performance as it compared with generally accept-
may be warranted. Consider peer entrapment that able criteria of performance and to recruit feedback
is observed wherein adolescents reinforce poten- about that performance from their supervisors. The
tially maladaptive behaviors. For example, use of clients kept records of their activity, evaluated the
illicit drugs may result in social attention from records against the standards, and made a notation
peers. Society has, however, systematized a com- on their records regarding productivity, according
munity of punishment: police intervention and legal to whether it was above or below criterion
sanctions. Unfortunately, continued use of drugs is performance. The feedback received was both
frequently maintained because of the rich natural praise recognizing productivity and disapproval
reinforcers, physiological and peer mediated, which and encouragement following substandard perfor-
may be more potent than the intermittent schedule mance. Under these conditions of the recruitment of
of punishment provided by law enforcement. both positive and corrective feedback (perhaps
reinforcers and punishers), production rates main-
2. Recruit natural consequences. Perhaps a tained better than with the self-monitoring.
behavior is not occurring at a sufficient
frequency or with adequate skill to naturally 3. Modify maladaptive consequences. Inappro-
contact important consequences. However, priate behaviors may be maintained by
following suitable training focusing on both consequences which are quite powerful, yet
the frequency and skill of the behavior and the behavior itself is considered maladaptive.
724 stokes & osnes

The consequences that are occurring may be procedures. Following the reduction of the parent's
considered maladaptive also. As such, if the inappropriate behaviors, there was a generalized
behavior change agent can temporarily or increase in the mother's use of positive verbal and
permanently curtail, interrupt, or discontinue physical behaviors during interaction with the
the scheduling of these consequences, then the children. The frequency of these behaviors were
termination of reinforcer delivery will be an increased further during parent training that
effective extinction contingency and the fre- focused directly on those skills.
quency of the maladaptive behavior will
decrease. Many therapy programs endeavor 4. Reinforce occurrences of generalization.
to intervene to eliminate the maladaptive Called "train to generalize" by Stokes and
contingencies maintaining inappropriate be- Baer (1977), this tactic is probably the least of
havior so that more appropriate behaviors the current 12 tactics to be called a legitimate
can be developed and maintained through programming strategy. Equally so, it may be a
natural or temporarily artificial conse- most reliable and legitimate technique that
quences. facilitates generalization. What is being ad-
vocated is that any occurrence of generaliza-
For example, Sanson-Fisher, Seymour, Mont- tion that occurs, is noticed, and is valued,
gomery, and Stokes (1978) employed procedures to should be followed by consequences that are
directly intervene in the interactions of female likely to function as reinforcers. Take advan-
adolescent delinquents. Existing contingencies in tage of randomness and good fortune, and if
the institution were naturally arranged to support possible, set up the conditions for generaliza-
and attend to antisocial and delinquent talk and tion of responding and reinforce it when it
activity, with few documented positive conse- occurs.
quences for adaptive and prosocial activity. There-
fore, a self-recording and contingent token Parsonson and Baer (1978) provided an example
reinforcement procedure was implemented to facil- of reinforcement of the occurrence of generalization
itate the occurrence of positive conversation and to in the improvised use of tools by preschoolers.
decrease the frequency of attention that apparently When presented with a task for hammering, filling
reinforced inappropriate conversation. There was containers, or tying, new or creative solutions for
an increase in the frequency of the adolescents' completing the task in the absence of the typical
prosocial comments and their frequency of atten- material (e.g., hammer, bag, shoelace) were rein-
tion to the prosocial comments of peers. These forced. These novel solutions to the problem were
changes were documented both when staff mem- in fact generalizations that had not been seen before
bers were present, and when staff members were and their reinforcement by praise and feedback was
absent. However, the complexity of such interven- an efficient strategy to promote further generaliza-
tion in systems with maladaptive consequence tion.
relationships is emphasized by the additional In a similar fashion, Campbell and Willis (1978)
finding that even though antisocial comments developed creativity of essay writing by children in
decreased in frequency, the frequency of attention the fifth grade. Using procedures of social and
to those comments increased, consistent with the token reinforcement, they targeted originality and
overall increased attention to the comments of creative writing as compared to peers or to
peers. Fortunately, the intervention procedures previous individual performance. In this manner,
incorporating self-recording and tokens effectively variability, i.e., occurrences of generalization, was
reduced the frequency of antisocial comments even reinforced, and those behaviors increased and
though the schedule of competing consequences of maintained after the artificial consequences were
peer attention did thicken. faded.
Another example of an intervention to modify
maladaptive contingencies and to foster the devel- B. Train Diversely: A tradition of tightly con-
opment of appropriate behavior through more trolled training conditions for behavior anal-
adaptive consequences was provided by Wolfe, et ysis and therapy programs is changing to a
al. (1982). They provided an intensive program of more frequent recommendation of less rigid
parent training to decrease negative and hostile programming that will likely have a greater
verbal and physical behaviors directed by an impact on generalization outcomes. What has
abusive mother towards her children. Procedures been frequently documented is the fact that
included the manipulation of attention, timeout, focused training frequently has focused ef-
and therapist praise for correct implementation of fects. When the goals and procedures of
 generalization 725

training are more widespread, so are the all tables that needed attention and refrained from
outcomes more widespread. Diversity in bussing when the circumstances did not warrant
training involves the use of controlled and such activity. In order to manage the training,
frequent variation consistent with the direct multiple stimulus exemplars of appropriate bussing
and generalized targets for the intervention circumstances and multiple exemplars of inappro-
program, i.e., allowing variations in the priate conditions were presented as part of the
conditions of training and the antecedent general case programming procedures. Under these
stimuli, responses, and consequences conditions, during probe sessions in untrained
employed in the modification procedures. cafeterias, the clients correctly determined when
to bus when appropriate and refrain from such
5. Use sufficient stimulus exemplars. A stimulus activity when appropriate. In this way, a general-
exemplar is a training condition related to the ized repertoire with limits was developed.
circumstances of training, such as the person
who is conducting the therapy sessions or the 6. Use sufficient response exemplars. A response
room in which the therapy occurs. If gener- exemplar is a behavior employed in training.
alization obtained is related to the conditions If a response is considered as a representative
of training, then generalization that occurs is of a particular class of responses, then
more likely with that trainer or in similar training that incorporates a subset of those
rooms. However, if multiple stimuli condi- responses may be sufficient to accomplish
tions are the focus of training, e.g., multiple generalization to many previously untrained
trainers, multiple settings, then more general- behaviors. As the complexity of a response
ization is observed. In fact, it appears that if a class increases, the need to sample the
small number of stimulus exemplars is incor- diversity of the dimensions of that class in
porated into training, then an efficient pro- the choice of the training dimensions becomes
gramming of generalization occurs. It is a more exacting task.
important here to note that the most efficient Many skill based therapeutic programs em-
generalization occurs within the stimulus ploy multiple exemplars of the targeted
class that is represented in training. behavior as a routine part of the intervention.
For example, Plienis, et al. (1987) evaluated a
Haring (1985) trained four to seven year old program of conversational skills and prob-
handicapped children to play with various toys. lem solving training for emotionally dis-
Initial training was in the general use of the play turbed adolescents. During small group
skills, e.g., holding and moving an aeroplane training, multiple exemplars of the skills
through the air. Multiple classes of toys were were used during instruction, modeling,
used. After the training included the use of the skill rehearsal, feedback, and shaping procedures.
with different types of planes, for example, the In addition, the clients were prompted to
children were able to play appropriately with a identify, discuss, and practice skills using
range of plane types. That is, the training program examples of situations which occurred in
taught an initial skill, then expanded the scope of daily living circumstances. The clients' im-
the procedures to include multiple stimulus exem- proved conversational and problem solving
plars in which the skill could be used. The result skills showed generalization to new situa-
was the display of the generalized skill with toys not tions and topics as well as the maintenance of
previously incorporated into training. those skills across time.
The use of multiple stimulus exemplars to The notion of response classes is relevant
facilitate generalization is a tactic that may have here. That is, some classes of behaviors
some limitations if the exemplars are not chosen increase and decrease in frequency as a
carefully. Horner, Eberhard, and Sheehan (1986) function of interventions applied to some
emphasize this point by documenting training members (behaviors) in the class. Carr and
effects involving both appropriate and inappropri- Durand (1985) have extended the formula-
ate stimulus examples using general case program- tion of response class to include even dramat-
ming procedures. Four retarded persons were ically different responses that are controlled
taught to bus tables in cafeteria restaurants. The by the same functional consequences. In their
restaurants presented multiple stimulus exemplars research, developmentally disabled children
for bussing, i.e., multiple opportunities to bus or were taught effective (i.e., functional) com-
not to bus. The question was whether the client munication skills that, when used, allowed
generalized in a discriminated fashion, i.e., bussed the children to evoke task related attention
726 stokes & osnes

and decrease the display of problem behavior. vocabulary use. They also resulted in the use of
Such procedures based on functional equiva- more elaborate sentences in spontaneous language.
lences within response classes emphasize the McGee, Krantz and McClannahan (1986) provided
need to consider the control of diverse an additional example of the use of incidental
behaviors through the application of the teaching procedures to reading instruction for
same consequences. In a related analysis, autistic children.
Schreibman and Carr (1978) modified the A similar example of the use of the tactic of
echolalic responses of handicapped children making antecedents less discriminable was provid-
by teaching them to make an appropriate ed by Campbell and Stremel-Campbell (1982).
verbal response, i.e., "I don't know," to They taught handicapped language-delayed chil-
questions that previously were echoed. After dren to increase their use of certain syntactic
training on a small set of questions, the structures, such as the appropriate use of "is/are"
children appropriately responded to a large in response to questions. The training conditions
number of untrained questions without echo- allowed the child to initiate a verbal response in the
ing, either by giving an answer or by stating presence of a wide array of natural stimulus events
that the answer was not known. or in response to a trainer's prompts, questions, or
statements. Any correct response that occurred
7. Make antecedents less discriminable. Many under these diverse conditions was reinforced using
therapy and training protocols require tightly social and token consequences. Concurrent exam-
controlled conditions that maximize the ination of verbalizations in a play setting showed
control over all aspects of the training collateral changes that occurred after the training
conditions and the stimuli presented during began.
training. This may be a valuable procedure to In an examination of less discriminable anteced-
obtain maximum training effects, but it does ent training conditions, Panyan and Hall (1978)
not facilitate generalization. The current focused on the training of writing/tracing and vocal
recommended generalization programming imitation tasks under two types of interventions. In
tactic, of making antecedents less discrimina- the first, serial training, tracing was taught to a
ble, is to allow variety in the conditions of mastery criterion, then training on vocal imitation
training so that the client will not readily began. In the second, concurrent training, the two
discriminate performance to a particular set tasks, tracing and vocal imitation, were presented in
of circumstances. Stokes and Baer (1977) an alternating less predictable (hence less discrim-
have referred to such conditions as "loose inable) fashion. The retarded clients showed that
training." Conceptually, it is not being they learned equally well under either training
proposed that previously discriminated stim- format. In testing with untrained letters for tracing
ulus control be entirely lost by making the and untrained sounds for imitation, however,
antecedents indiscriminable. Making anteced- generalization was superior following the less
ents less discriminable broadens the stimulus discriminable antecedent conditions of concurrent
control by making the relations between the training.
antecedent stimuli and the behavior less
predictably tied to a narrow range of stimuli. 8. Make consequences less discriminable. Pre-
dictability is frequently the cause of discrim-
For example, Hart and Risley (1980) developed inated effects. Therefore, if the schedule or
incidental language teaching procedures that took circumstances of the delivery of behavior
advantage of naturally occurring interactions be- consequences is variable, then generalization
tween handicapped and nonhandicapped pre- may be enhanced. Various mechanisms for
schoolers and their teachers. The procedures making consequences less discriminable ap-
capitalized on the approaches made by the children pear in the literature, including the use of
by requiring elaborated verbalization related to the intermittent reinforcement schedules, the tim-
topic raised by the child prior to the delivery of the ing and delay of the presentation of the
attention, material, or activity consequence. This consequence, and the predictability of the
training occurred in natural, not specialized train- presence of the therapy agent to observe and
ing, settings which contained a rich variety of deliver the consequence. In Stokes and Baer
stimuli. It occurred at times unplanned and spread (1977), this tactic was referred to as "indis-
nonsystematically throughout the day and was in criminable contingencies." In the current
large part controlled by the child. These procedures taxonomy, contingencies are separated some-
resulted both in improved frequency of talking and what artificially into their antecedent and their
 generalization 727

consequences functions and recommendations not maintain over time. However, some schedule
are made for making both less discriminable. manipulations did enhance the maintenance (or
generalization across time). Thinning the continu-
Relevant here is the response-consequence rela- ous schedule of consequences in training did
tion. If the behavior that is reinforced, for example, facilitate better maintenance. Another successful
is allowed to vary around a standard, then this tactic of making the consequences less discrimina-
arrangement is likely to facilitate a less discrimina- ble was the occasional noncontingent presentation
ble consequence relationship. Dunlap and Johnson of the reinforcing consequence in the generalization
(1985) have provided an example of the effect of setting.
using less discriminable consequences in their
research with autistic children. They provided C. Incorporate Functional Mediators: A media-
training under conditions in which the therapist tor is a stimulus that occurs between the
would be present in the clinic some time and absent training and the occurrence of generalization
at other times. The major manipulation was the in such a way that it facilitates or mediates
predictability of the times when the therapist would that generalization, probably as a discrimi-
be present to provide consequences for being native stimulus for the performance of the
on-task and correct responding with the task. In behavior. The term mediator may be taken to
the less predictable condition, the therapist's imply something more important than mere
presence was intermittent and variable in length. environmental stimuli such as antecedents,
Under these conditions, when left unsupervised, the behaviors, and consequences. But it is, in fact,
children were on task more and completed more just another stimulus, no more, no less. A
response tasks during the conditions of less mediator assists another behavior by facili-
discriminable consequences. tating the generalization. A mediating stimu-
Timing in the delivery of consequences was also a lus is usually one that can be carried easily by
factor in the research by Fowler and Baer (1981), in the client to a diversity of extra-therapy
which delayed reinforcers were presented in such a conditions or is readily present in a diversity
manner that there was not a ready discrimination of of conditions. To be useful, the mediator
the settings or behavior contingencies. Pre- must both be present or produced in other
schoolers' social and academic performance was relevant conditions and must be functionally
targeted within multiple settings of the school day, discriminative for performance in those
an early directly targeted contingent setting and a stimulus conditions. Presumably, the devel-
later generalization setting. No contingencies ever opment of the discriminative control occurs
operated in the generalization setting. When as part of the training. Any procedures
reinforcers were presented after the increased relevant to the production of the controlling
behaviors occurring in the early contingent setting, stimuli in the generalization environments
improvements were demonstrated. Similar im- will often occur in the original therapy
provements were not seen in the later setting unless conditions as well.
there was a delay in the delivery of the reinforcer for
the behavior in the early setting until after the later 9. Incorporate common salient physical stimuli.
generalization setting. That is, the delay of rein- A common stimulus that facilitates generali-
forcement acted as a less discriminable contingency. zation may be a physical object that is present
Manipulations of the schedules of reinforcement in both the training and the generalization
are another example of procedures to effect setting, or at least very similar items are
generalization. Intermittent schedules are noted to present in both settings. Usually, the physical
facilitate maintenance. The discriminability of the stimulus is incorporated into the training
delivery of consequences is an important issue in situation in an obviously salient and func-
generalization programming. One examination of tional manner. That stimulus may then be
such schedule manipulations is the research by transported across relevant generalization
Koegel and Rincover (1977). Autistic children were environments. Alternatively, physical items
taught various imitation and instruction compli- common and obvious in natural and relevant
ance tasks in one setting and the generalization and generalization environments are incorporated
maintenance of those skills in an extra-therapy into the training. The therapist then takes
setting was assessed. The results highlight some advantage of the natural occurrence of those
differences between generalization across settings stimuli in settings to which generalization is to
and the maintenance of responding in that setting, be facilitated. As with all programming tactics
i.e., initial generalization often occurred but it did there is no guarantee that stimuli chosen by a
728 stokes & osnes

therapist will actually function as facilitators, of the therapeutic manipulation of environ-

nor is there any guarantee that nontargeted mental elements in a way to take advantage
stimuli will acquire stimulus control over the of discriminative properties of peers.
behaviors targeted.
For example, Stokes, Doud, Rowbury, and Baer
For example, Marholin and Steinman (1977) (1978) incorporated peers into the training of
examined the discriminative properties of the behavior disordered preschoolers completing an
teacher of 10-12 year old behavior problem academic task. During training, the peer presented
children as it related to academic productivity and stimulus materials and provided consequences to
remaining ontask during class times. Whenever the the child learning the skills. Implementation of the
teacher left the classroom, performance levels procedures was monitored by an adult instructor.
declined. However, if the teacher reinforced aca- The children learned the prepositional recognition
demic rate and accuracy rather than on-task responses but did not generalize those skills to a
behavior, academic productivity remained higher regular academic setting. When the peer-tutor was
during the teacher's absence. Apparently, the brought to the classroom during testing for
academic materials themselves acquired discrimi- generalization, no generalization was displayed.
native control that facilitated generalization to However, if the salience of the peer-tutor was
circumstances under which the most potent con- increased, either by the tutor's presentation of the
trolling stimulus (the teacher) was not present. The stimulus cards, or by the tutor's occasional
materials were the common salient physical stimuli. provision of a positive consequence, generalized
Charlop, Schreibman, and Thibodeau (1985) performance was facilitated. The dimensions of
also considered manipulation of common salient salience and discriminative control by peers may
stimuli. They used the reinforcer from training as require more explicit programming. There are
the common stimulus. That is, in language training various strategies for ensuring changes as a function
for autistic children, the children were taught to of peer involvement in training, but the develop-
name objects and ask for them. The stimuli used ment of discriminative control over generalized
were food and drink, which were also the rein- performance is not a routine matter (Odom,
forcers for the correct response. In other settings, Hoyson, Jamieson, and Strain, 1985). The use of
given the presence of these obviously salient stimuli, peers as agents of treatment and generalization
similarly correct and generalized performance programming has much to recommend it, including
could be facilitated. Generalization across settings the possible positive effects on the peer-trainers
and people was documented. themselves (Dougherty, Fowler, and Paine, 1985).
A salient stimulus may take many forms. Halle,
Marshall, and Spradlin (1979) used time delay to 11. Incorporate salient self-mediated physical
facilitate appropriate meal requests by retarded stimuli. A self-mediated stimulus is a stim-
children. The time delay took the form of postpon- ulus that is maintained and transported by
ing the delivery of the food tray until the correct the client as part of the treatment. An
request had been made. This salient procedure, example of a self-mediated physical stimulus
when used across settings, facilitated generalized is the use of a notebook that specifies how to
performance. The time delay itself may be consid- perform in a certain setting. This notebook
ered a common salient discriminative stimulus. may be referred to just prior to the behavior
performance opportunity and the reminders
10. Incorporate common salient social stimuli. facilitate appropriate performance. The so-
A social stimulus relates to the characteris- phistication of the self-mediated stimuli may
tics of a person, such as a certain gesture, or be as straightforward as the proverbial
the presence of the person. In some cases, the string on the finger or as complex as a
therapist is present in settings other than the detailed set of lecture notes that serves as a
original training setting. As common salient memory and prompts bridging the gap
stimuli, the therapist may act as a facilitator between task acquisition and refinement
of generalization because of acquired dis- and relevant generalization settings provid-
criminative properties. Another possibility is ing opportunities for performance of the
that whenever a therapist is not available in skill. Of course, no matter how sophisticated
multiple settings then important persons the program of generalization programming
from that setting may be incorporated into may be, if the common salient stimuli are not
training. The use of peers and peer-tutors in produced in the generalization settings, they
training and therapy is an obvious example are not going to facilitate performance.
 generalization 729

There may need to be a program that targets behavior was similarly effected. In fact, there was
the production of the facilitating stimulus in no spontaneous skill transfer or change in class-
relevant environments. room performance. At least two conclusions may be
drawn from this and related work (e.g., Bryant &
The use of procedures of self-recording were Budd, 1982). First, antecedent verbalizations can
demonstrated by Holman and Baer (1979). Chil- control subsequent behavior. Second, unless those
dren were taught to use a small bracelet with discriminative verbalizations are themselves pro-
moving beads to record the amount of academic duced in the relevant generalization setting, perfor-
work they completed in training. They later used mance there may not change. In the Guevremont et
this same bracelet to self-record performance in a al. (1988) study, academic performance and
regular classroom. This self-recording of task on-task activity did improve but only after the
completion facilitated improved (generalized) teacher in the classroom gave the instruction for the
ontask activity and some improved academic child to use the self-instruction procedures in the
measures. Use of the self-mediated common phys- classroom.
ical stimulus of the bracelet was also related to In a related examination of verbal regulation,
improved maintenance of performance. Guevremont, Osnes, and Stokes (1986) document-
Koegel, Koegel, and Ingham (1986) also used ed the development of correspondence between
self-recording procedures. Elementary school chil- what children said they would do and how they
dren involved in speech therapy recorded correct actually did act in settings remote from the time and
speech production by noting a plus (+) on a piece of place of the verbalization itself. The children were
paper. Self-recording resulted in improved articula- reinforced for the correspondence between an
tion of targeted phonemes and important generaliza- antecedent verbalization and subsequent perfor-
tion to natural settings. Occasional checks on the use mance in group and play activities. After training
of the procedures were made by parents and teachers had occurred in multiple settings and with multiple
in the natural settings. Improved articulation main- behaviors, a generalized verbal control developed
tained during extensive periods during which the such that control over behavior at home reliably
self-recording procedures were no longer used. came under the control of a preschool verbalization
alone. All such procedures are not straightforward.
12. Incorporate salient self-mediated verbal and Osnes, Guevremont, and Stokes (1987) found that
overt stimuli. A distinction is made for the a developmentally disabled child required a timeout
use of self-mediated verbal and covert procedure in addition to the regular positive
stimuli because they are sometimes more procedures for these typical procedures to work
difficult to monitor, not because they are effectively. Issues in the generalization of such
considered to be fundamentally different techniques are worthy of further investigation
processes of behavior control and generali- (Guevremont, Osnes, & Stokes, in press; Osnes,
zation programming (Stokes, Osnes, & Guevremont, and Stokes, 1986).
Guevremont, 1987). Many of these tactics Goal setting is another example of antecedent
relate to verbalization, language, and control. For example, Kelley and Stokes (1984)
thought. Many cognitively-oriented strate- employed contracting procedures to increase the
gies would be classified here. Self-mediated academic productivity of disadvantaged high
verbal and/or covert discriminative stimuli school dropouts. Performance following these
are produced by the client across relevant conditions did not maintain well. When the
settings in order to facilitate the demonstra- adolescents were able to set and monitor their
tion of skill acquired as part of a training or own goals and performance, however, this less
therapy program conducted elsewhere. intrusive, client oriented condition did facilitate
better maintenance of performance. It should be
For example, Guevremont, Osnes, and Stokes noted that contracting as a condition prior to goal
(1988) examined the use of self-instructions by setting may be an important procedural part of
preschoolers. They found that the children were treatment that enhances the discriminative proper-
able to learn the self-instructions related to aca- ties of the goal setting verbalizations and actions,
demic performance. They did not, however, use and thereby facilitates generalization (Guevremont,
those skills outside of the training environment. Osnes, and Stokes, 1988a).
Even though a functional relationship was estab-
lished between the children's use of the self­ In Pursuit of Generalization
instructions and their actual performance in train- We have outlined and given examples of principles
ing, that was no guarantee that natural classroom and tactics that may be incorporated into
730 stokes & osnes

therapeutic endeavors in order to program for Clearly, there should be a differential payoff for
generalization and maintenance. There are many researchers who are able to document more
such examples in the operant literature. Neverthe- comprehensive and thorough analyses of their
less, in the interest of a balanced perspective, it is generalization data. Increasingly, relevant generali-
necessary to make a few comments regarding the zation outcomes should be the specific dependent
general sophistication of the technology of gener- variables of the research and the analysis of the
alization programming, be it implicit or explicit, a independent variables effects on generalized out-
fledgling or in full flight. comes should be an essential aspect of data analysis.
Sometimes we do not need generalized changes Train and hope is merely a short term show-and-tell
because direct and focused change is the goal of our strategy which will not advance an inductive science
behavior analysis and therapy programs. However, much further than the (sometimes impressive) gains
there are many instances in which focused change is already made. Demonstration of valuable general-
insufficient. The relevant question here is whether ized outcome is no longer sufficient, even if it is most
the effects of the intervention spread, without direct welcome. Analysis of the variables that produce and
intervention, to other stimuli, responses, or across control the occurrence of generalization is essential if
time. Generalized outcomes are often sought and the science of generalization programming is to
are of critical importance in the work of a behavior thrive and expand. There is no doubt that general-
analyst and therapist. There has been a great ization is important to behavior analysis and therapy
volume of research over the past decade that has endeavors. When research is reviewed for publica-
shed further light on the principles and technology tion, there must be attention payed both to the scope
of generalization and its programming (see, for of generalization data, and the thoroughness of the
example, Horner, Dunlap & Koegel, 1988). Nev- specific analyses of the generalization promoting
ertheless, the lessons have been that the pursuit of procedures and the controlled evaluation of the
generalization is an arduous activity, with a few subsequent generalization effects.
impressive gains, some small triumphs, but mostly The literature is abundant with examples of the
the realization that, still, generalization should not occurrence of generalized outcomes, but many of
be automatically expected unless there are specific these describe generalization that may be quite
procedures implemented in order to facilitate its narrow in scope. To document the more wide-
occurrence. There frequently is a cost in obtaining spread and longlasting effects may require a major
generalization and many times the cost is worth the effort and the return for effort may rapidly
outcome, when considering a basic cost-benefit diminish. Many applied programs may, in fact,
analysis. There may be times, though, when the target large scale, widespread, and well maintained
only efficient way to ensure widespread change is to changes, but the documentation and analysis of
program in all relevant circumstances for as long as those changes is not easy, perhaps not feasible. This
the behavior control is desirable. See, for example, is where the occasional friction between meaningful
the discussion by Wolf, Braukman, and Ramp practice and systematic research occurs. Significant
(1987) regarding extended treatment of some effects are not yet established sufficiently. However,
delinquent behavior. we have cause for a more optimistic assessment
The occurrence of generalization itself seems to than this harsh conclusion. As Stokes and Osnes
be a sufficient reinforcer of research and clinical (1988) noted, we have come a long way even if we
activity. There are few natural positive conse- still have some distance to cover.
quences or costs associated with the analysis or Most practitioners do in fact consider generali-
absence of the analysis of the conditions and zation issues, but often as an adjunct procedure
circumstances of the occurrence versus the nonoc- near the end of a course of therapy. That should not
currence of generalization. Therefore, an unfortu- be the case. A diligent clinician should be attending
nate conclusion is that generalization and the to issues of generalization programming from the
variables related to its occurrence have not yet the outset of treatment. We hope that a pursuit of
been well enough researched. However, our field is generalization will be an operant behavior that
moving in the direction of greater sophistication, frequently meets functional reinforcers for practi-
allowing a greater specification of the relevant tioners. This paper has outlined a number of
variables and programming strategies that facilitate principles and tactics which are recommended as
generalization and maintenance. In fact it is likely likely to facilitate the occurrence of generalization
that parallel and complementary recommendations in programs of clinical importance. In general, a
will emerge delineating technologies of generaliza- clinician or a researcher would do well to imple-
tion of behavior change and the maintenance of ment procedures that incorporate the general
behavior change. principles of exploiting current functional
 generalization 731

contingencies, training diversely, and incorporating Guevremont, D. C., Osnes, P. G., & Stokes, T. F. (2016).
functional mediators. More specifically, the tactical Generalization and maintenance of verbal control using
correspondence training.Behavior Modification (In press)
armamentarium should include contacting and Halle, J. W., Marshall, A. M., & Spradllin, J. E. (1979).
recruiting natural consequences, modifying mal- Time-delay: A technique to increase language use and
adaptive consequences, reinforcing occurrences of facilitate generalization in retarded children. Journal of
generalization, using sufficient stimulus and re- Applied Behavior Analysis, 12, 431–439.
sponse exemplars, making antecedents and conse- Haring, T. G. (1985). Teaching between-class generali7lltion of
toy play behavior to handicapped children. Journal of
quences less discriminable, incorporating common Applied Behavior Analysis, 19, 127–139.
salient physical and social stimuli, and incorporat- Hart, B., & Risley, T. R. (1980). In vivo language intervention:
ing self-mediated physical, verbal, and covert Unanticipated general effects. Journal of Applied Behavior
stimuli. As the data base expands with future Analysis, 13, 407–432.
analyses of generalization and maintenance, these Holman, J., & Baer, D. M. (1979). Facilitating generalization of
on-task behavior through self­ monitoring of academic tasks.
concepts may well be expanded or modified. That is Journal of Autism and Developmental Disorders, 9(429), 416.
the hallmark of an inductive science that seeks new Horner, R. H., Dunlap, G., & Koegel, R. L. (1988).
knowledge to guide sound practice of behavior Generalization and Maintenance: Life-Style Changes in
analysis and therapy. Applied Settings. Baltimore, MD: Brookes Publishing.
Horner, R. H., Eberhard, J. M., & Sheehan, M. R. (1986).
Teaching generalized table bussing. Behavior Modification,
References 10, 457–471.
Johnston, J. M. (1979). On the relation between generalization
and generality. The Behavior Analyst, 2, 1–6.
Baer, D. M. (1982). The role of current pragmatics in the future Kelley, M. L., & Stokes, T. F. (1984). Student-teacher
analysis of generalization technology. In R. B. Stuart (Ed.), contracting with goal-setting for maintenance. Behavior
Adherence, compliance, and generalization m behav1oral Modification, 8, 223–244.
medicine (pp. 192–212). New York: Bruner/Mazel. Koegel, L. K., Koegel, R. L., & Ingham, J. S. (1986).
Bijou, S. W., & Baer, D. M. (1978). Behavior Analysis of Child Programming rapid generalization of correct articulation
Development. Englewood Cliffs, NJ: Prentice-Hall. through self-monitoring procedures. Journal of Speech and
Bryant, L. E., & Budd, K. S. (1982). Self-instructional training Hearing Disorders, 51, 24–32.
to increase independent work performance in preschoolers.- Koegel, R. L., & Rincover, A. (1977). Research on the
Journal of Applied Behavior Analysis, 15, 259–271. 1 difference between generalization and maintenance in
Campbell, C. R., & Stremel-Campbell, K. (1982). Programming extra-therapy responding. Journal of Applied Behavior
"loose training" as a strategy to facilitate language general- Analysis, 10, 1–12.
ization. Journal of Applied Behavior Analysis, 15, 295–301. Koegel, R. L., & Williams, J. A. (1980). Direct versus indirect
Campbell, J. A., & Willis, J. (1978). Modifying components of response-reinforcer relationships in teaching autistic chil-
"creative behavior" in the natural environment. Behavior dren. Journal of Abnormal Child Psychology, 8, 537–547.
Modification, 2, 549–564. Mank, D. M., & Horner, R. H. (1987). Self-recruited feedback:
Carr, E. G., & Durand, V. M. (1985). Reducmg behavior A cost-effective procedure for maintaining behavior. Re-
problems through functional communication training. search in Developmental Disabilities, 8, 91–112.
Journal of Applied Behavior Analysis, 18, 111–126. Marholin, D., & Steinman, W. M. (1977). Stimulus control in
Charlop, M. H., Schreibman, L., & Thibodeau, M. G. (1985). the classroom as a function of the behavior reinforced.
Increasing spontaneous verbal responding in autistic chil- Journal of Applied Behavior Analysis, 10, 465–478.
dren using a time delay procedure. Journal of Applied McGee, G. G., Krantz, P. J., & McClannahan, L. E. (1986). An
Behavior Analysis, 18, 155–166. extension of incidental teaching procedures to reading
Dougherty, B. S., Fowler, S. A., & Paine, S. C. (1985). The use instruction for autistic children. Journal of Applied Behavior
of peer monitors to reduce negative interaction during Analysis, 19, 147–157.
recess. Journal of Applied Behavior Analysis, 18, 141–153. Odom, S. L., Hoyson, M., Jamieson, B., & Strain, P. S. (1985).
Dunlap, G., & Johnson, J. (1985). Increasing the independent Increasing handicapped preschoolers' peer social interac-
responding of autistic children with unpredictable supervi- tions: Cross-setting and component analysis. Journal of
sion. Journal of Applied Behavior Analysis, 18, 227–236. Applied Behavior Analysis, 18, 3–16.
Fowler, S. A., & Baer, D. M. (1981). "Do I have to work all Osnes, P. G., Guevremont, D. C., & Stokes, T. F. (1986). "If I
day?" The timing of delayed reinforcement as a factor in say I'll talk more, then I will": Correspondence training to
generalization. Journal of Applied Behavior Analysis, 14, increase peer directed talk by socially withdrawn children.
13–24. Behavior Modification, 10, 287–299.
Guevremont, D. C., Osnes, P. G., & Stokes, T. F. (1986). Osnes, P. G., Guevremont, D. C., & Stokes, T. F. (1987).
Preparation for effective self-regulation: The development of Increasing a child's prosocial behaviors: Positive and
generalized verbal control. Journal of Applied Behavior negative consequences in correspondence training. Journal
Analysis, 19, 99–104. of Behavior Therapy and Experimental Psychiatry, 18,
Guevremont, D. C., Osnes, P. G., & Stokes, T. F. (1988). The 71–76.
functional role of preschoolers' verbalizations in the Paine, S. C., Hops, H., Walker, H. M., Greenwood, C. R.,
generalization of self-instructional training. Journal of Fleischman, D. H., & Guild, J. J. (1982). Repeated treatment
Applied Behavior Analysis, 21, 45–55. effects: A study of maintaining behavior change in socially
Guevremont, D. C., Osnes, P. G., & Stokes, T. F. (1988). withdrawn children. Behavior Modification, 6, 171–199.
Preschoolers' goalsetting with contracting to facilitate Panyan, M. C., & Hall, R. V. (1978). Effects of serial versus
maintenance. Behavior Modification, 12, 401–423. concurrent task sequencing on acquisition, maintenance,
732 stokes & osnes

and generalization. Journal of Applied Behavior Analysis, Behavior: Development, Assessment and Modification
11, 67–74. (pp. 407–443). Orlando, FL: Academic Press.
Parsonson, B. S., & Baer, D. M. (1978). Training generalized Stokes, T. F., & Osnes, P. G. (1988). The developing applied
improvisation of tools by preschool children. Journal of technology of generalization and maintenance. In R. H.
Applied Behavior Analysis, 11, 363–380. Horner, G. Dunlap, & R. L. Koegel (Eds.), Generalization
Plienis, A. J., Hansen, D. J., Ford, F., Smith, S., Stark, L. J., & and Maintenance: Life-Style Changes Applied Settings
Kelly, J. A. (1987). Behavioral small group training to (pp. 5–19). Baltimore, MD: Paul Brookes Publishing.
improve the social skills of emotionally-disordered adoles- Stokes, T. F., Fowler, S. A., & Baer, D. M. (1978). Training
cents. Behavior Therapy, 18, 17–32. preschool children to recruit natural communities of
Sanson-Fisher, B., Seymour, F. W., Montgomery, W., & Stokes, reinforcement. Journal of Applied Behavior Analysis, 11,
T. F. (1978). Modifying delinquent's conversation using 285–303.
token reinforcement of self-recorded behavior. Journal of Stokes, T. F., Osnes, P. G., & Guevremont, D. C. (1987). Saying
Behavior Therapy and Experimental Psychiatry, 9, and Doing: A commentary on a contingency-space analysis.
163–168. Journal of Applied Behavior Analysis, 20, 161–164.
Schreibman, L., & Carr, E. G. (1978). Elimination of echo1alic Stokes, T. F., Doud, C. L., Rowbury, T. G., & Baer, D. M.
responding to questions through the training of a general- (1978). Peer facilitation of generalization in a preschool
ized verbal response. Journal of Applied Behavior Analysis, classroom. Journal of Abnormal Child Psychology, 6,
11, 453–463. 203–209.
Skinner, B. F. (1938). The Behavior of Organisms. New York, Wolf, M. M., Braukman, C. J., & Ramp, K. A. (1987). Serious
New York: Appleton-Century. delinquent behavior as part of a significantly handicapping
Skinner, B. F. (1953). Science and Human BehaviOr. New condition: Cures and supportive environments. Journal of
York, New York: Macmillan. Applied Behavior Analysis, 20, 347–359.
Stokes, T. F. (1985). Contingency management. In A. S. Bellack, Wolfe, D. A., St. Lawrence, J., Graves, K., Brehony, K.,
& M. Hersen (Eds.), Dictionary of Behavior Therapy Bradlyn, D., & Kelly, J. A. (1982). Intensive behavioral
Techniques (pp. 74–78). Elmsford, NY: Pergamon. parent training for a child abusive mother. Behavior
Stokes, T. F., & Baer, D. M. (1977). An implicit technology of Therapy, 13, 438–451.
generalization. Journal of Applied Behavior Analysis, 10,
349–367.
Stokes, T. F., & Osnes, P. G. (1986). Programming the R E C E I V E D : August 15, 2016
generalization of children's social behavior. In P. S. Strain, A C C E P T E D : August 23, 2016
M. J. Guralnick, & H. Walker (Eds.), Children's Social