Вы находитесь на странице: 1из 7




We propose the use of a combined version of the alternating treatments and multiple baseline
designs in situations in which a traditional baseline (no treatment) condition either does not provide
an adequate contrast condition or is not feasible or practical due to clinical constraints. We refer to
this design as a sequential alternating treatments design because two treatments are initially im-
plemented in a random or counterbalanced fashion and are followed by a sequential change in one
or both treatments across settings, subjects, or tasks. The effects of the independent variables are
assessed first by analyzing the two series of data points representing the different treatments (relative
effects) and then by assessing changes in one or both series, as application of the alternative treatment
is introduced sequentially. The sequential application of treatment provides an analysis of control
in the same manner as the multiple baseline design; the initial alternating treatments phase provides
a contrast condition in much the same manner as a baseline condition. Applications of this design
to the assessment of peer training and self-injurious behavior are described.
DESCRIPTORS: within-subject design, clinical evaluation, baseline

In most within-subject designs, a nontreatment experimental design. For example, a pragmatic con-
or baseline phase precedes the initiation of treat- cern might occur in evaluating the occurrence of
ment (Barlow & Hersen, 1984). The baseline phase social interactions between handicapped and non-
provides a contrast condition or, as Risley and Wolf handicapped students in a dassroom setting under
(1972) pointed out, a predictor of the target be- two different conditions. In one condition, the non-
havior if untreated. In most cases, baseline contin- handicapped peer provides the statement, "Let's
ues until stability of responding is achieved (Baer, play a (specific) game"; in the other condition, the
Wolf, & Risley, 1968) so that the results of treat- peer says, "Let me show you how to play the
ment can be compared directly to baseline. A num- game." In this case, the initiating statement made
ber of authors (e.g., Barlow & Hersen, 1984; John- by the peer is the independent variable. Some stud-
ston & Pennypacker, 1981; Kazdin, 1982) have ies have suggested that social interactions will occur
discussed the importance of baseline for evaluating more often for the handicapped child if he or she
internal validity in within-subject designs. is asked rather than instructed to play (Haring,
Breen, Pitts-Conway, Lee, & Gaylord-Ross, 1987;
Occasions When a Traditional Baseline Is Sasso, Mitchell, & Struthers, 1986). A comparison
Not Feasible or Practical of these strategies as a research question poses sev-
Some evaluation situations pose feasibility and eral pragmatic problems relative to the indusion of
pragmatic concerns regarding the indusion of a a traditional baseline condition that indudes a series
traditional baseline (no treatment) condition in the of observations of unprompted interactions prior to
the initiating-statements condition. If a multiple
The preparation of this manuscript was supported, in part, baseline design across dyads is used for evaluation,
by the Iowa University Affiliated Facility. The authors grate-
fully acknowledge Agnes DeRaad for her expert editorial the uncontrolled behavior of the peer and the target
assistance. child always precedes both treatment conditions
Colleen McMahon is now at Georgetown University; Mark (play vs. instruct); this order could influence sub-
Steege is now at the University of Southern Maine. sequent treatment conditions in unknown direc-
Requests for reprints should be sent to David P. Wacker,
Division of Developmental Disabilities, The University of tions. In a worst-case scenario, the behavior of the
Iowa, Iowa City, Iowa 52242. peer or target child might require intervention due
334 DAVID WACKER et al.
to inappropriate behavior such as teasing, punishing In some cases, the use of an alternating treat-
statements, or the display of socially unacceptable ments design without a baseline phase is adequate
behavior. to evaluate functional control: Treatment A is su-
The indusion of baseline also is problematic if perior to Treatment B across sessions. For example,
the teacher or researcher is interested in evaluating in the study reported by Steege, Wacker, Berg,
the effects of two types of initiating statements Cigrand, and Cooper (1989), the use of an alter-
during the initial contact between the students. It nating treatments design was sufficient to demon-
is not typical for children to be paired or grouped strate changes in self-injurious behavior across func-
together without instructions regarding their activ- tional analysis conditions without preceding the
ities. Thus, a great deal of time and effort is some- treatment conditions with a baseline phase. The
times spent developing a baseline condition that alternating treatments design, as a comparison de-
seldom occurs in the natural context. Although a sign, is adequate for evaluating the relative effects
return to baseline following treatment (BAB) might of two or more treatments in the absence of a
be considered to evaluate control, a reversal is pos- baseline condition (Barlow & Hersen, 1984).
sible only if the effects of treatment are not main-
tained. Alternating Treatments Design
Similar concerns can arise in evaluating self-in- With Baseline
jurious behavior (SIB). For example, assume that In other cases, a baseline condition is necessary
a child in a hospital setting engages in SIB only to evaluate the effects of treatment, especially if
during demanding activities (e.g., self-help skills similar effects occurred for both treatments. For
training and physical therapy). If an activity varies example, Wacker, Berg, Wiggins, Muldoon, and
on some known dimension (e.g., passive vs. active Cavanaugh (1985) used a combination multiple
participation), it could be viewed as representing baseline (across subjects) with alternating treat-
more than one treatment (type of demand). In this ments design to evaluate potential reinforcers. In
case, the supposition could be made that it is not this investigation, two or more microswitch-acti-
demands per se, but distinct types of demands, that vated toys were evaluated as potential reinforcers
produce SIB. Baselines across other activities might for profoundly handicapped children within an al-
be recorded to replicate this finding (e.g., require- ternating treatments design. Performance during
ments for passive vs. active participation in other baseline demonstrated that presence of the micro-
activities), but one is still left with a baseline phase switches, when not connected to the toys, did not
that is comprised of two distinct treatments. increase responding (i.e., pressing the switch). In-
stead, responding increased only when the micro-
Alternating Treatments Design switches activated specific toys. Of equal impor-
Without Baseline tance, the inclusion of some toys resulted in decreased
One solution is to use an alternating treatments responding relative to baseline, indicating that these
design (Barlow & Hayes, 1979) that does not in- toys functioned as punishers to the children. Thus,
dude a baseline phase. The major question ad- the presence of baseline provided information need-
dressed by alternating treatments designs is whether ed for evaluating the reinforcing effects of each toy.
one treatment is more effective than another (Bar- A baseline condition is also useful for evaluating
low & Hersen, 1984; Hains & Baer, 1989). In carryover effects. For example, Shapiro, Kazdin,
both examples described above, the two treatment and McGonigle (1982) used a baseline condition
conditions (specific statements or specific demands) to evaluate carryover effects directly by changing
may be compared directly to determine whether the presentation of their treatments. The investi-
one treatment resulted in greater effects (higher gators compared response cost, token reinforce-
frequency of social behavior or self-injurious be- ment, and a baseline condition by conducting four
havior) relative to the other. alternating treatment evaluations, with each dyad

of treatment and baseline repeated twice. The to- "(Instruct statement). The same game was
kens produced similar results when compared to used for both peers; only the instructions changed.
either baseline or response cost, demonstrating min- In this investigation, the first condition of the
imal, if any, carryover effects. In both of the above design was an alternating treatments phase, with
investigations, a baseline condition, conducted ei- the statements made by a peer serving as the in-
ther prior to treatment or counterbalanced with dependent variable. The dependent variable was
treatment, functioned as a contrast condition against the percentage of social interactions occurring per
which the results of treatment could be compared. 10-min session. The initial condition of the design
In the above two examples, the use of a tradi- was then extended sequentially across target chil-
tional baseline condition permitted more precise dren to form a sequential alternating treatments
evaluation of the effects of the treatment conditions design. The results demonstrated good stability of
by providing a contrast condition. The contrast responding, suggesting that social interactions were
condition, in the first example, provided more in- facilitated more by the play statement than by the
formation on the absolute effects of treatment and, instruct statement.
in the second example, provided more information To replicate these results, one of the initiating
on interaction effects. It is probably for this rea- statements (counterbalanced across children) was
son-the provision of additional information-that changed to the alternative statement; in other words,
alternating treatments designs are seldom conduct- both peers asked the target child to play or both
ed without a baseline condition, even though a provided instruction. This was followed by a second
baseline condition is not a technical requirement change condition, in which both peers provided the
(Barlow & Hersen, 1984). opposite statement. The results show that (a) the
statement delivered by the peer controlled respond-
ing across target children, (b) the results were re-
Sequential Alternating Treatments Design versible, and (c) changes in behavior occurred with
A paradox confronts applied researchers: In some changes in statement and not as a function of ex-
applied situations, a naturalistic baseline is difficult posure to a particular peer. Extending the initial
to implement, yet the inclusion of a baseline or alternating treatments condition permitted evalu-
contrast condition is frequently necessary for more ation of the effects of time, exposure to a given
precise evaluations of experimental control. In these peer, practice, and other potential sources of ex-
situations, improved evaluation of control might ternal confounding. By changing the statement de-
be possible by first extending the alternating treat- livered by 1 peer and then by the 2nd peer, rep-
ments conditions across subjects, settings, or tasks lication of control over behavior by the independent
in a staggered fashion, identical to the multiple variable was evaluated.
baseline design. The alternating treatments condi- In the above example, evaluation of experimen-
tion then could be followed by the sequential ap- tal control by the independent variable was estab-
plication of one or both treatments to evaluate lished through two distinct design modifications:
changes in behavior associated with each treatment. (a) sequential modification of the treatment con-
An example of this approach is provided in Figure dition and (b) two change (or replication) condi-
1. In this simulated investigation (based on a study tions, each demonstrating that behavior was con-
by McMahon, Wacker, Sasso, & Melloy, 1988), trolled by the peers' verbal statements. Although
the investigators assigned 2 nonhandicapped peers both design modifications are desirable for exper-
each to play with 3 moderately mentally retarded imental control, pragmatic considerations (e.g.,
boys. One peer was told to approach his assigned maintenance) may limit the use of the replication
partner and say, "Let's play ; it's a favorite conditions (in the same way that standard reversal
game of mine" (Play statement). The second peer designs may not be possible). When replication or
was told to say, "Let me show you how to play reversal conditions are not possible, the sequential
Alternating Instruct Play
100 Treatments Only Only

r ~~4060 -
820 ¶iI<t~
, , . . .


Instru-tPlay Instruct
O I ~~~~~~~~~Only

Cw 80 7S

40 Peer Play
20 XPer e IeI

0 I
°, L - - -i Play Instruct Play
0 100
Only l
Only Only
40- *ee P lay Bil
- IBo -
0 I~~~

;4 60 1 b~Istruct II

1002 4 6 a 10 12 14 16
Figure 1. Example of a sequential alternating treatments design across subjects with peer training.

application of one of the treatments can still provide treatment are expected with minimal carryover ef-
increased information regarding experimental con- fects from one treatment to the next. In these sit-
trol. uations, the greatest utility of the sequential alter-
An example of this type of sequential alternating nating treatments design is probably in dinical or
treatments design is provided in Figure 2. In this educational situations in which two or more in-
example, an initial assessment demonstrated that dependent variables warrant evaluation and rela-
the subject, a profoundly mentally retarded child, tively brief periods of time are available for iden-
engaged in self-injurious behavior during demand- tifying the respective effects of each independent
ing activities. Within these activities, however, the variable. For example, in the evaluation of aberrant
behavior of the child was inconsistent, suggesting behavior discussed earlier, the control of behavior
that different types of demands during the activities by one treatment over another might first be eval-
were differentially affecting behavior. To better as- uated through direct comparison across situations
sess whether specific types of demands had differ- and then confirmed through sequential changes in
ential effects on the occurrence of self-injurious be- one or both treatments over time.
havior, separate occurrence data were collected for In other situations, the greatest utility might be
both passive and active demands. The active de- when implementation of baseline poses pragmatic
mands were defined as prompting the child to par- problems. For example, the applicability of a tra-
ticipate partially in the activity, and the passive ditional baseline may be problematic in situations
demands were defined as occurring when the ac- in which the student is receiving on-the-job train-
tivities were performed for the child. ing. Instead, the student might initially receive two
During the alternating treatments condition, a types of instruction (e.g., picture prompts vs. time
greater percentage of self-injurious behavior oc- delay prompting) across work tasks to determine
curred across all activities when the subject was the relative effects of each instructional approach.
provided with passive versus active demands (see Once relative differences have been identified, in-
Figure 2). However, as was the case for the first troduction of the preferred approach across work
example, possible confounding effects occurred be- tasks in a staggered fashion provides the teacher
cause the passive and active demands occurred in with evidence of functional control in the same
the context of the same activity. The possibility of manner as the multiple baseline design.
this type of confounding effect was reduced by In our previous research, the sequential alter-
showing the same effects across four activities (stan- nating treatments design would have been useful
dard alternating treatments design), but was vir- in our evaluations of self-injurious behavior (Steege
tually eliminated with the sequential application of et al., 1989), picture prompts in school settings
active demands across the activities. Because the (Wacker, Berg, Berrie, & Swatta, 1985), and social
modification of activities occurred in a sequential behavior (Sasso & Rude, 1987).
fashion, the effects of practice and exposure to treat- Hains and Baer (1989) provided a thorough
ment were evaluated in the same way that these evaluation of alternating treatments designs and
extraneous variables are evaluated within a multiple presented several options for evaluating interaction
baseline across tasks design. effects. We agree with Hains and Baer that eval-
uation of interaction effects can be important and
Applications and Limitations of the can be best evaluated within the design options
Sequential Alternating Treatments Design they proposed. Those options should be considered
The application of the sequential alternating when interaction effects are of interest, because the
treatments design is limited to the same situations sequential alternating treatments design is not as
recommended for the alternating treatments design, useful for evaluating interaction effects.
as discussed by Barlow and Hersen (1984) and Gast and Wolery (1988) also proposed a mod-
Hains and Baer (1989): Relatively quick effects of ification of the alternating treatments design, which
100 Alternating Treatments Active Demands Only
s0 Passive I
60 Active
40 _ _ _ _ _

30 -_ _ _ _ _

20 - IToothbrushing
0 ac C <6C->C> > )
0 2 4 6 I 10 12 14 16

Y ~~90-
. 70 -
30 - Drinking
* 0 ~20-
*"mob 0 2 4 6 8 10112 14 16

w 89 0
O 60 i
30 Recreation

; ~~0 2 4 6 8 10 12 14 16

50 Physical
30 - Therapy
0 2 4 6 8 10 12 14 16
Figure 2. Example of a sequential alternating treatments design across tasks with self-injurious behavior.

they termed the parallel treatments design, for treat- Gast, D., & Wolery, M. (1988). Parallel treatments design:
ments that produce durable results that are not A nested single subject design for comparing instructional
procedures. Education and Treatment of Children, 11,
easily reversible. In situations in which two or more 270-285.
treatments will be compared, and both treatments Hains, A., & Baer, D. (1989). Interaction effects in mul-
are expected to produce durable results, the use of tielement designs: Inevitable, desirable, and ignorable.
Journal of Applied Behavior Analysis, 22, 57-69.
the parallel treatments design should be considered. Haring, T., Breen, C., Pitts-Conway, V., Lee, M., & Gaylord-
Ross, R. (1987). Adolescent peer tutoring and special
Summary friends experiences. Journal of the Association for Per-
sons with Severe Handicaps, 12, 280-286.
We suggest that the sequential alternating treat- Johnston, J., & Pennypacker, H. (1981). Strategies and
ments design be considered in situations in which, tactics of human behavioral research. Hillsdale, NJ:
for a variety of reasons, conducting a traditional Erlbaum.
baseline condition may prove to be difficult. These Kazdin, A. (1982). Single-case research designs: Methods
for clinical and applied settings. New York: Oxford
situations may occur most often when two or more University Press.
independent variables are possible or are in effect McMahon, C., Wacker, D., Sasso, G., & Melloy, K. (1988).
during initial observation or assessment. In the se- Instructional control ofpeer interactions between hand-
icapped and nonhandicapped students. Unpublished
quential alternating treatments design, the alter- manuscript, University of Iowa.
nating treatments condition constitutes a contrast Risley, T., & Wolf, M. (1972). Strategies for analyzing
condition within which the relative effects of each behavioral change over time. In J. Nesselroade & H.
Reese (Eds.), Life-span developmental psychology:
variable are evaluated. This condition is then fol- Methodological issues (pp. 175-183). New York: Ac-
lowed by one or more change conditions, in which ademic Press.
one treatment is sequentially applied across subjects, Sasso, G., Mitchell, V., & Struthers, E. (1986). Peer tu-
tasks, or settings. toring versus structured interaction activities: Effects on
the frequency and topography of peer initiations. Be-
The absence of a baseline phase can still be havioral Disorders, 11, 249-259.
problematic with the sequential alternating treat- Sasso, G., & Rude, H. (1987). Unprogrammed effects of
ments design. Without a baseline, for example, it training high-status peers to interact with severely hand-
icapped children.Journal ofApplied Behavior Analysis,
is possible only to compare the relative effects of 20, 35-44.
two treatments and to demonstrate the sequential Shapiro, E., Kazdin, A., & McGonigle,J. (1982). Multiple
control established by one or both of those treat- treatment interference in the simultaneous or alternating
ments. In many situations, this may be adequate, treatments design. Behavioral Assessment, 4, 105-115.
Steege, M., Wacker, D., Berg, W., Cigrand, K., & Cooper,
as with the self-injurious behavior example. In other L. (1989). The use of behavioral assessment to prescribe
cases, this may not be adequate, because the central and evaluate treatments for severely handicapped chil-
question is concerned with the absolute magnitude dren. Journal of Applied Behavior Analysis, 22, 23-
of effects, or with interaction effects, in addition to Wacker, D., Berg, W., Berrie, P., & Swatta, P. (1985).
relative effects. Generalization and maintenance of complex skills by se-
verely handicapped adolescents following picture prompt
training. Journal of Applied Behavior Analysis, 18,
Wacker, D., Berg, W., Wiggins, B., Muldoon, M., & Cava-
Baer, D., Wolf, M., & Risley, T. (1968). Some current naugh, J. (1985). Evaluation of reinforcer preferences
dimensions of applied behavior analysis. Journal of Ap- for profoundly handicapped students.Journal of Applied
plied Behavior Analysis, 1, 91-97. Behavior Analysis, 18, 173-178.
Barlow, D., & Hayes, S. (1979). Alternating treatments
design: One strategy for comparing the effects of two Received January 17, 1989
treatments in a single subject. Journal of Applied Be- Initial editorial decision May 24, 1989
havior Analysis, 12, 199-210. Revisions received August 17, 1989; April 18, 1990
Barlow, D., & Hersen, M. (1984). Single case experi- Final acceptance April 20, 1990
mental designs (2nd ed.). New York: Pergamon Press. Action Editor, Susan A. Fowler