Exceptional Children

Vol. 71. Na. 2. pp. 165-179.

&2005 Councilfor hoxpmmil ChiUm.

The Use ofSingle-Subject

Research to Identify
Evidence-Based Practice
in Special Education
ROBERT H. HORNER
Vnivmiiy nfOrfaun

EDWARD G. CARR
State University of New York at Stony Brook

JAMES HALLE
Universiiy of/Hindis

GAIL MCGEE
Emory University

SAMUEL ODOM
liiiluirui ihiivmity

MARK WOLERY
Van^rbilt Univenity

ABSTRACT: Slnglesubject research plays an important role in the development of evidence-based

practice in special education. The defining features of single-subject research are presented, the con-
tributions oj single-subject research for special education are reviewed, and a specific proposal is of-
fered for using single-subject research to document evidence-based practice. This article allows
readers to determine if a specific study is a credible example of single-subject research and if a spe-
cific practice or procedure has been validated as "evidence-based" via single-subject research.

ingle-subject research is a rigor- the field of special education (Kennedy, in press;

ous, scientific methodology Odom & Strain, 2002: Tawney & Gast, 1984;
used to define basic principles Wolery & Dunlap, 2001). Since the methodol-
of behavior and establish evi- ogy was first operationalized over 40 years ago
dence-based practices. A long (Sidman, 1960), single-subject research has
and productive history exists in which single-sub- proven particularly relevant for defining educa-
ject research has provided useful information tor tional practices at rhe level ol the individual

fixceptional Children 16S

learner Educators building individualized educa- 1982; Ktatochwill & l.evin, 1992; Richard, Tay-
tional and support plans have beneHted from the lor, Ratnasamy &C Richards, 1999; Iawncy &
systematic forni of experimental analysis single- Gast, 1984), and our goal here is not to provide
subject research permits (Dunlap & Kern, 1997). an introduction to the single-subject research, but
Of special value has been the ability of single-stib- to clarify how single-subject research is u.scd to es-
ject research methods to provide a level of experi- tablish knowledge within special education and
mental rigor beyond that found in traditional case define the empirical support needed to document
studies. Because .single-subject research docu- evidence-based practices.
ments experimental control, it Is an approach,
like randomized control-group designs (Shavelson
& Towne, 2002), that may be used to establish SINGLE-SUBJECT RESEARCH
evidence-based practices. METHODOLOGY
The systematic and detailed analysis of in-
dividuals that is provided through single-subject Single-subject research is experimental rather than
research methods has drawn researchers not only correlational or descriptive, and its purpose is to
from special education, but also from a growing document causal, or functional, relationships be-
array of scholarly disciplines, with over 45 profe.s- tween independent and dependent variables. Sin-
sional journals now reporting single-subject re- gle-subject research employs within- and
•search (American Psychological Association, between-subjects comparisons to control for
2002; Anderson, 2001). Further, an array of efHec- major threats to internal validity and requires sys-
tive interventions is now in use that emerged tematic replication to enhance external validity
through single-subject research methods. Rein- (Martdia, Nelson, & Marchand-Martella, 1999).
forcement theory or operant psychology has been Several critical features define this methodology.
the substantive area that has benefited most from Each feature is described in the follt>wing sections
single-case research methodology. In fact, operanr and organized later in a table of quality indicaror.s
principles of behavior have been empirically that may be used to assess ii an individuaJ study is
demonstrated and replicated within the context of an acceptable exemplar of single-subject research.
single-subject experiments for more than 70 years.
i- INDIVIDUAL PARTICIPANT IS THE
However, the close association between operant
UNIT OF ANALYSIS
analysis of human behavior and single-subject ex-
perimental research is not exclusionary. That is, Single-subject designs may Involve only one par-
many procedures based on diverse theoretical ap- ticipant, but typically include multiple partici-
proaches to human behavior can be evaluated pants (e.g., 3 to 8) in a single study. Each
within the confines of single-subject research. In- participant serves as bis or her own control. Per-
terventions derived from social-learning theory, formance prior to intervention is compared to
medicine, social psychology, social work, and performajice during and/or after intervention. In
communication disorders arc but a sample of pro- most cases a research participant is an individual,
cedures that have been analyzed by single-subject but it is possible for each participant to be a
designs and methods (c^,, Hersen & Barlow, group whose performance generates a single score
1976; Jayaratne &c Levy. 1979; McReynolds & per measurement period (e.g., the rate of problem
Kearns, 1983), behavior performed by all children within a class-
The specific goals of this article are to (a) room during a 20 min period).
present the defining features of single-subject re-
PAHI'ICIPANT AND SFTTING DESCRIPTION
search methodology, (b) clarily the relevance of
single-subject research methods for special educa- Single-subject research requires operational de-
tion, and (c) otifer objective criteria for determin- scriptions of the participants, setting, and the pro-
ing when single-subject research results are cess by which participants were selected (Wolery
sufficient for documenting evidence-based prac- & E?,ell, 199.3). Another researcher should be able
tices. Excellent introductions to single-subject re- to use the description of participants and setting
.search exist (Hersen &c Barlow, 1976; Kazdin, to recruit similar participants who inhabit similar

Winter 2005
settings. For example, operational participant de-
scriptions of individuals with a disability would
Single-subject research is experimental
require that the specific disability (e.g., autism rather than correlational or descrip- j
spectrum disorder, Williams syndrome) and the tive, and its purpose is to document \
specific instrument and process used to determine causal, or fiinctional, relationships be-
their disability (e.g., the Autism Diagnostic Inter-
tween independent and dependent vari- \
view-Revised) be identified. Global descriptions
such as identifying participants as having develop- ahles.
mental disabilities would be insufficient.
• Dependent variable recording is assessed for con-
DEPENDENT VARIABLE sistency throughout the experiment by frequent
Single-subject research employs one or more de- monitoring of interobserver agreement (e.g.,
pendent variables that are defined and measured. the percentage of observational units in which
In most cases the dependent variable in single- independent observers agree) or an equivalent.
subject educational research is a form of observ- The measurement of interobserver agreement
able behavior. Appropriate application of should allow assessment for each variable
single-subject methodology requires dependent across each participant in each condition of
variables to have the following features: the study. Reporting interobserver agreement
only for the baseline condition or only as one
• Dependent variables are operationally defined to score across all measures in a study would not
allow (a) valid and consistent assessment of the be appropriate.
variable and (b) replication of the assessment
process. Dependent variables that allow direct • Dependent variables are selected for their social
observation and empirical summary (e.g., significance. A dependent variable is chosen
words read correctly per min; frequency of not only because it may allow assessment of a
conceptual theory, but also because it is per-
head hits per min; number of s between re-
ceived as important for the individual partici-
quest and initiation of compliance) are desir-
pant, those who come in contact with the
able. Dependent variables that are defmed
individual, or for society.
subjectively (e.g., frequency of helping behav-
iors, with no deBnition of "helping" provided) INDEPENDENT VARIABLE
or too globally (e.g., frequency of "aggressive"
The independent variable in single-subject re-
behavior) would not be acceptable.
search typically is the practice, intervention, or
• Depejideiit variables are measured repeatedly behavioral mechanism under investigation. Inde-
within and across controlled conditions to pendent variables in single-subject research are
allow (a) identification of performance pat- operationally defined to allow both valid interpre-
terns prior to intervention and (b) comparison tation of results and accurate replication of the
of performance patterns across conditions/ procedures. Specific descriptions of procedures
phases. The repeated measurement of individ- typically include documicntation of materials
ual behaviors is critical for comparing the per- (e.g., 7.5 cm x 12.5 cm card) as well as actions
formance of each participant with his or her (e.g., peer tutors implemented the reading cur-
own prior performance. Within an experimen- riculum in 3 1:1 context, 30 min per day, 3 days
tal pha.se or condition, sufficient assessment per week). General descriptions of an interven-
occasions are needed to establish the overall tion procedure (e.g., cooperative play) ihat are
pattern of performance under that condition prone to high variability in implementation
(e.g., level, trend, variability). Measurement of would not meet the expectation for operational
the behavior of the same individual across description of the independent variable.
phases or conditions allows comparison of per- To document experimental control, the in-
formance patterns under different environ- dependent luiriiible in single-subiect research is ac-
mental conditions. tively, rather than passively, manipulated. 'I'he

Exctptienal Childrtn 167

researcher must determine when and how the in- typically requires multiple data poitits (five or
dependent variable will change. For example, if a more, although fewer data points are acceptable
researcher examines the effects of hard versus easy in specific cases) without substantive trend, or
schuol work {independent variable) on rates of with a rrend in the direction opposite that pre-
problem behavior (dependent variable), the re- dicted by the intervention. Note that if the data
searcher would be expected to operationally de- in a baseline documents a trend in the direction
fine, and systematically introduce, hard and easy predicted by the intervention, then the ability to
work rather than simply observe behavior across document an effect following intervention is
the day as work of varying difficulty was naturally compromised.
introduced.
EXPERIMENTAL CONTROL
In single-subject research the fidelity of in-
dependent variable implementation is documented.Single-subject research designs provide experi-
Fidelity of implementation is a significant con- mental control for most threats to internal valid-
cern within single-subject research because the in- ity and, thereby, allow confirniiuion of a
dependent variable is applied over time. As a functional relationship between manipulation of
result, documentation of adequate implementa- the independent variable and change in the de-
tion fidelity is expected either through continuous pendent variable. In most cases experimental con-
direct measurement of the independent variable, trol is demonstrated when the design documents
or an equivalent (Gresham, Gansel, & Kurtz, three demonstrations of the experimental effect at
1993). three different points in time with a single partici-
pant (within-subject replication), or across different
BASF.LlNE/COMPAklSON CONDITION participants (inter-subject replication). An experi-
Single-subject research designs typically compare mental effect is demonstrated when predicted
the effects of an intervention with performance change in the dependent variable covaries with
during a baseline, or comparison, condition. The manipulation of the independent variable (e.g.,
baseline condition is similar to a treatment as the level, and/or variability of the dataset in a
usual condition in group designs. Single-subject phase decreases when a behavior-reduction inter-
research designs compare performance during the vention is implemented, or the level and/or vari-
baseline condition, and then contrast this pattern ability of the dataset in A phase increases when the
with performance under an intervention condi- behavior-reduction intervention is withdrawn).
tion. The emphasis on comparison across condi- Documentation of experimental control is
tions requires measurement during, and detailed achieved through (a) the introduction and with-
description ot, the baseline (or comparison) con- drawal (or reversal) ot the independent variable;
dition. Description of the baseline condition (b) the staggered introduction of the independent
should be sufTicicntly precise to allow replication variable at different points in time (e.g., multiple
of the condition by other researchers. baseline); or (c) the iterative manipulation ol^ the
Measurement of the dependent variable independent variable (or levels of the independent
during a baseline should occur until the observed variable) across observation periods (e.g., alternat-
pattern of responding is sufficiently consistent to ing treatments designs).
allow prediction of future responding. Documen- For example. Figure 1 presents a typical A
tation of a predictable pattern during baseline (Baseline)-B (Inrervention)-A (Baseline 2)-B (In-
tervention 2) single-subject research design that
establishes three demonstrations of the experi-
mental effect at three points in time through
To (document experimental control, the demonstration that behavior change covaries with
independent variable In single-subject manipulation (introduction and removal) of the
research is actively, rather than pas- independent variable between Baseline and Inter-
vention phases. Three demonstrations of an ex-
sively, manipulated.
perimental effect are documented at the three
arrows in Figure 1 by (a) an initial reduction in

Winter 2005
FIGURE 1

Example of a Single-Subject Reversal Design Demonstrating Experimental Control

Baseline Inten/enfloo Baseline Intervention

T I I
1 22 23 24 2 5 26 27 28

Note. Arrow.^ indicate [he ihrce points in ihc srudy where an experimental eftecr is confirmed.

[antrums between the first A phase (Baseline) and ships between manipulation of independent vari-
the first B phase (Intervention); (b) a second ables and change in dependent variables. Rival
change in response patterns (e.g., return to Base- hypotheses (e.g., passage of time, measurement
line patterns) with re-introduction of the Baseline effects, uncontrolled variables) must be discarded
conditions in the second A phase; and (c) a third to document experimental control. Traditional
change in response patterns (e.g., reduction in case study descriptions, or studies with only a
tantninis) with re-introduction of the interven- baseline followed by an intervention, may provide
tion in the second B phase. useful information for the field, but do not pro-
A similar logic for documenting experimen- vide adequate experimental control to qualify as
tal control exists tor multiple baseline designs single-subject research.
with three or more data series. The sta^ered in-
VISUAL ANALYSIS
troduction of the intervention within a multiple
baseline design allows demonstration of the ex- Single-subject research results may be interpreted
perimental ertect not only within each data series, with the use of statistical analyses (Todnian &
btit also across data series at the staggered times ot Dugard, 2001); however, the traditional approach
intervention. Figure 2 presents a (.ie.sign that in- ro analysis of single-subject research data involves
cludes three series, with introduction of the inter- systematic visual comparison of responding
vention at a different point in time for each series. within and across conditions of a study (Parson-
The results document experimental control by son & Baer, 1978). Documentation of experi-
demonstrating a covariation between change in mental control requires assessment of all
behavior patterns and introduction of the inter- conditions within the design. Each design (e.g.,
vention within three different series at three dif- reversal, multiple baseline, changing criterion, al-
ferent point.s in time.
Excellent sources exist describing the grow-
ing array of single-subject designs that allow doc- Single-subject designs provide experimen-
umentation of experimental control (Hersen &
Barlow, 1976; Kazdin, 1982, 1998; Kennedy, in tal documentation of unequivocal rela-
press; Kratochwill & Levin, 1992; McReynolds & tiomhips between manipulation of
Kearns, 1983; Richard, et al., 1999; Tawney & independent variables and change in de-
Gast, 1984). Single-subject designs provide exper- pendent variables.
imental documentation of unequivocal relation-

Exceptional Children
FIGURE 2

Example of a Multiple Baseline Design Across Participants That Demomtrates Experimental Control

Baseline Intervention
100

90

SO

70

60

50

40

30

20

10 Participant A
0
n I I r
100

90

80
O
70
O
Vi 60
CD
50
o 40
(D
30

20

a. 10 Participant B
0
I r i I I I \ I I r
100
90
80
70
60
50
40
30
20
Participant C
10
0
y
^ r I i I I 1 ; TTi
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

Sessions

Note. Arrows indicate the three demonstrations of effect at thtee ditlcrent points in time.

170 WmUr2005
ternating treatments) requires a specific data pat- tent to which an effect documented by one study
tern for the researcher to claim that change in the has relevance for participants, locations, materials,
dependent variable is, and only is, a function of and behaviors beyond those defined in the study.
manipulating the independent variable. External validity of results from single-subject re-
Visual analysis involves interpretation of search IS enhanced through replication of the ef-
the level, trend, and variability ot performance oc- fects across different participants, different
curring during baseline and intervention condi- conditions, and/or different measures of the de-
tions. Level refers to the mean performance pendent variable.
during a condition (i.e., phase) of the study. Although a study may involve only one par-
Trend references the rate of increase or decrease of ticipant, features of external validity of a .single
the best-fit straight line for the dependent variable study are improved if the smdy includes multiple
within a condition (i.e., slope). Variability refers participants, settings, materials, and/or behaviors.
to the degree to which performance fluctuates It i.s typical for single-subject studies to demon-
around a mean or slope during a phase. In vi.sual strate effects with at least three different partici-
analysis, the reader also judges (a) the immediacy pants. It also is expected that the generality
of effects following the onset and/or withdrawal and/or "boundaries" of an intervention will be es-
of the intervention, (b) the proportion of data tablished not by a single study, but through sys-
points in adjacent phases that overlap in level, (c) tematic replication of effects across multiple
the magnitude of changes in the dependent vari- studies conducted in multiple locations and across
able, and (d) the consistency of data patterns multiple researchers (Birnbrauer, 1981). External
across multiple presentations of intervention and validity in single-subject rese,irch also is enhanced
nonintervention conditions. The integration of through operational description of (a) the partici-
information from these multiple assessments and pants, (b) the context in which the study is con-
comparisons is used to determine if a functional ducted, and (c) the factors influencing a
relationship exists benveen the independent and participant's behavior prior to intervention (e.g.,
dependent variables. assessment and baseline respon.se patterns).
Documentation of a functional relationship The external validity for a program of sin-
reqtiires compelling demonstration of an effect gle-subject studies is narrowed when selection and
(Parsonson & Baer, 1992). Demonstration of a attrition bias (e.g., the selection of only certain
functional relationship is compromised when (a) participants, or the publication of only successful
there is a long latency between manipulation of examples) limit the range of examples available
the independent variable and change in the de- for analysis (Durand & Rost, in press). Having
pendent variable, (b) mean changes across condi- and reporting specific selection criteria, however,
tions are small and/or similar to changes within assist in defining for whom, and under what con-
conditions, and (c) trends do not conform to ditions a given independent variable is likely to
those predicted following introduction or manip- result in defined changes in the dependent mea-
ulation of the independent variable. sures. Attrition is a potent threat to both the in-
A growing set of models also exists for con- ternal and external validity of single-subject
ducting mcta-analysis of single-subjeci research studies, and any participant who experienced
(Busk & Serlin, 1992; Didden, Duker, & Korzil-
ius, 1997; Faith, Allison, & Gorman, 1996; Her-
shberger, Wallace, Green, & Marquis, 1999;
Marquis et al., 2000). This approach to analysis is External validity of results from single-
of special value in documentation of comparative subject research is enhanced through
trends in a field. replication of the effects across different
EXTERNAL VALIDITY participants, different conditions, and/or
Single-subject designs are used to (a) test concep-
different measures of the dependent vari-
tual theory and (b) identify and validate effective able.
clinical interventions. A central concern is the ex-

Exceptiottal ChiUren 171

both conditions (i.e., baseline and intervention) normatively would not have good social valid-
of a study should be included in reports of that ity.
study.
• Demonstration that the intervention produced
SOCIAL VALIDITY an effect that mei the defined, clinical need.
Within education, single-subject research has Within special education, single-subject re-
been used not only to identify basic principles of search has been used to examine strategies for
behavior (e.g., theory), but also co document in- building academic achievement (Greenwood,
terventions (independent variables) rhat are func- Iapia, Ahbotr & Walton, 2003; Miller, Gunter,
tionally related to change in socially important Venn, Hummel, & Wiley, 2003; Rohena, Jitendra
outcomes (dependent variables; Wolf, 1978). Tiie & Browdcr, 2002); improving social behavior and
emphasis on intervention has resulted in substan- reducing problem behavior (Carr et al., 1999;
tial concern about the social validity, or practical- Koegel & Koegel, 1986. 1990); and enhancing
ity, of research procedures and findings. I he the skills of teachers (Moore et al., 2002) or fami-
social validity of singlc-suhjcct research goals, pro- lies who implement interventions (Cooper,
cedures and findings is enhanced by: Wacker, Sasso, Reimers, & Donn, 1990; Hall et
al., 1972).
• Emphasis on the selection of dependent vari-
SIngle-suhject research also can be used to
ables that have high social importance.
emphasize important distinctions between, and
• Demonstration that the independent variables integration of, efficacy research (documentation
can he applied with fidelity by typical inter- that an experimental cflect can be obtained under
vention agents (e.g., teachers, parents) in typi- carefully controlled conditions) and effectiveness
cal intervention contexts across meaningful research (documentaiian that an experimental ef-
periods of time. fect can be obtained under typical conditions)
that may affect targe-scale implementation of a
" Demonstration that typical intervention
procedure (Flay, 1986).
agents (a) report the procedures to be accept-
able, (b) report the procedures to he feasible RESFAUCH QUESTIONS API'ROPRIATB FOR
within available rcsource.s, (c) report the proce- SlN(JLH-SUBJECT RHSEARCH METHODS
dure to he effective, and (d) choose to con-
The selection of any research methodology should
tinue use of the intervention procedures after
be guided, in pan, by the research question(s)
formal support/expectation of use is removed.
under consideration. No research approach is ap-
For example, an effective procedure designed
propriate for all research questions, and it is im-
for use hy young parents where the procedure
portant to clarify the types of research questions
fits within the daily family routines would
that any research method is organized to address.
have good social validity, whereas an interven-
Single-subject rt'search designs are organized to
tion that disrupted family routines and com-
provide fine-grained, time-series analysis of change
promised the ability of a family to function
in a dependent variable(s) across systematic intro-
duction or manipulations of an independent vari-
able. They are particularly appropriate when one
wishes to understand the performance of a specific
Within education, single-subject research individual under a given set of conditions.
has been used not only to identify basic Research questions appropriately addressed
principles of behavior (e.g., theory), but with single-subject methods (a) examine causal,
or functional, relations by examining the effects
also to document interventions (indepen- that introducing or manipulating an independent
dent variables) that are functionally re- variable (e.g., an intervention) has on change in
lated to change in socially important one or more dependent variables; (b) focus on the
outcomes (dependent variables). effects that altering a component of a multicom-
poncnt independent variable (e.g., an intervcn-

Wimer 2(tO5
tion package) has on one or more dependent vari-
ables; or (c) focus on the relative effects of two or Single-subject research designs are orga- '
more independent variable manipulations (e.g., nized to proi'idefine-grained,time-series
alternative interventions) on one or more depen- analysis of change in a dependent vari-
dent variables. Examples of re.search questions ap-
able(s) across systematic introduction or
propriately addressed by single-subject methods
include
manipulations of an independent vari-
able.
• Does functional communication training re-
duce problem behavior?
I M P O R T A N C E OF SINGLE-
• Do incidental reaching procedures increase so- SUBJECT RESEARCH METHODS
cial initiations by young children with autism? F O R R E S E A R C H IN S P E C I A L
EDUCATION
• Is time delay prompting or least-to-most
prompt hierarchy more effective in promoting Single-subject research methods offer a number of
self-help skills of young children with severe features that make them particularly appropriate
disabilities? for tise in special education research. Special edu-
cation is a field that emphasizes (a) the individual
• Does pacing of reading instruction increase the
student as the unit of concern, (b) active interven-
rate of acquisition of reading skills by third
tion, and (c) practical procedures that can he used
graders? in typical school, home, and community contexts,
• Does the use of a new drug for children with Special education is a problem-solving discipline,
AD/HD result in an increase in sustained at- in which ongoing research in applied settings is
tention? needed. Single-subject research matches well with
the needs of special education in the following
ways,
QUALITY INDICATORS FOR • Single-subject research focuses on the individual.
SINGUE-SUBJECT RESEARCH Causal, or functional, relationships can be iden-
tified without requiring the assumptions
By its very nature, research is a process of approxi- needed for parametric analysis (e.g., normal dis-
mations. The features listed previously define the trihution). Research questions in special educa-
core elements of single-subject research methodol- tion often focus on low-incidence or
ogy, but we recognize that these features will be heterogeneoas populations. Information about
met with differing levels of precision. We also rec- mean performanct- of these groups may be of
ognize that there are conditions in which excep- less value for application to individuals. Single-
tions are appropriate, It is important, therefore, to subject methods allow targeted analysis at the
offer guidance for assessing the degree to which unit of the "individual," the same unit at which
single-subject research methods have been applied the intervention will be delivered.
adequately within a study, and an objective stan- • Single-subject research allows detailed analysis of
dard for determining if a particular study meets "nonresponders'as well as "rcsponders." C,onuo\
the minimally acceptable levels that permit inter- group designs produce conclusions about the
pretation. generality of treatment effects as they relate to
Impressive efforts exist for quantifying the group means, not as they relate to specific indi-
methodological rigor of specific single-subject viduals. Even in the most successful group de-
studies (Busk & Serlin, 1992; Kratochwill, & signs, there are individuals whose behavior
Scoiber, 2002). In combination with the previous remains unaffected, or is made worse, by the
descriptions, we ofFer the information in Table 1 treatment (e.g., "nonresponders"). Single-suh-
as content for determining if a study meets the ject designs provide an empirically rigorous
"acceptable" methodological rigor needed to be a method for analyzing the characteristics of
credible example of single-subject research. these nonresponders, thereby advancing

Exceptional ChiUren 173

TABLE 1
Quality Indicators Within Single-Subject Research

Desfripiion of Participants and Setting

• Participants are dcscribL-d witli sufficient detail to allow others to select indiviJiials wiih simiUr characteristio;
(e.g., age, gender, disability, diagnosis).
• The process for selecting participants is described with leplicable precision.
• Critical features of the physical setting are described with siifficitnt precision to allow replication.
Dependent Variable
• Dependent variables are described with operational precision.
• F.ath dependent variable is measured witli a pnicedure that generates a quantifiable index.
• Measuremeiu ol the dependent variable is valid and described with replicable precision.
• Dependent variables are measured repeatedly over titne.
• Data are collected on the reliability or interobserver agreement associated with each depit-ndcnt variable, and
lOA levels meet tnlnimal standards {e.g., lOA = 80%; Kappa = 60%).
hidependent Variable
' Independent variable is described with replicable precision.
• Independent variable is systematically manipulated and under the control of the experimenter.
• Overt measurement of the fidelity of implementation for the independent variable is highly desirable.

• The majoriry of single-subject research studies will include a baseline pliase thai provides repealed measure-
ment of a dependent variable and establishes a pattern of responding that can be used to predict the pattern of
future performance, if introduction or manipulation of the independent variable did not occur.
• Baseline conditions are described with replicable precision.
Experimental Control/internal Validity
• The design provides at least three demonstrations of experimental efFcct at three different points in time.
• The design controls for common threats to internal validity (e.g., permits elimination of rival hypotheses).
• The results document a pattern that demonstrates experimental control.
External Validity
• Experimental effects are replicated across participants, settings, or materials to establish extetna! validity.
Social Validity
• The dependent variable is socially important.
• The magnitude of change in the dependent variable resulting from the intervention is socially important.
• Implementation of the independent variable is practical and cost effective.
• Social validity is enhanced by implementation of the independent variable over extended time periods, by typi*
cal intervention agents, in typical physical and social contexts.

knowledge about the possible existence of sub- Single-subject research provides a practical
groups and subjcct-by-treattiient interactions. methodology for testing educational and behav-
Analysis of notirespondcrs also allows identifi- ioral interventions. Single-subject methods
cation of intervention adaptations needed to allow unequivocal analysis of tbe relationsbip
produce intended outcomes with a wider range between individualized inicrventions and
of participants. change in valued outcomes. Tbtough replica-
tion, the methodology also allows testing of the
breadth, or external validity', of findings.
Single-subject research methods ojfer a
numher of features that make them par- Single-subject research provides a practical re-
tiaihirly appropriate for use in special ed- search methodology for assessing experimental ef-
ucation research. fects under typical educational conditions.
Single-subject designs evaluate interventions

174 Winter 2005

 (independent variables) under conditions simi- & Srrain, 2002; Shernoff, Kratochwill, & Stoiber,
lar to those recommended for special educa- 2002). This is a logical, bur not easy, task (Chris-
tors, such as repeated applications of^ a tenson, Carlson, & Valdez, 2002). We provide
procedtire over time. This allows assessment of here a context for using single-subject research to
the process of change as well as the product of document evidence-based practices in special edu-
change, and facilitates analysis of maintenance cation that draws directly from recommendations
as well as initial effects. by the Task Force on Evidence-Based Interven-
tions in School Psychology (Kratochwill &
Single-subject research designs allow testing of con- Stoiber, 2002), and the Committee on Science
ceptual theory. Single-subject designs can be and Practice, Division 12, American Psychologi-
used to tesr the validity oF theories of behavior cal Association (Weisz & Hawley. 2002).
that predict conditions under which behavior
A practice refers to a curriculum, behavioral
change (e.g., learning) should and should not
intervention, systems change, or educational ap-
occur.
proach designed for use by families, educators, or
Single-subjeet research methods are a cost-effective students with the expres.s expectation that itnple-
approach to identifying educational and behav- mentation will result in measurable educational,
ioral interventions that are appropriate for large- social, behavioral, or physical henefit. A practice
scale analysis. Single-subject research methods, may be a precise intervention (e.g., functional
when applied across multiple .studies, can be communication training; Carr & Durand, 1985),
used to guide large-scale policy direcrives. Sin- a procedure for documenting a controlling mech-
gle-subject research also can be used cost effec- anism (e.g., the use of high-probability requests to
tively to produce a body of reliable, persuasive create behavioral momentum; Mace et al., 1988),
evidence that justifies investment in large, often or a larger program with multiple compotients
expensive, randomized control group designs. (e.g., direct instruction; Gettinger, 1993).
The control group designs, in turn, can be used Within single-subject research methods, as
to further demonstrate external validity of find- with other research methods, the field is just be-
ings established via .single-^subject methodology. ginning the process of determining the profes-
sional standards that allow demonstration of an
evidence-based practice (KratochwitI & Stoiber.
T H E I D E N T I F I C A T I O N OF
2002). It is prudent to propose initial standards
that are conservative and draw from existing ap-
EVIDENCE-BASED PRACTICES
plication in the field (e.g., build from examples of
USING SINGLE-SUBJECT
practices such as functional communication train-
RESEARCH
ing that are generally accepted as evidence ba.sed).
(Current Icgislaiion and policy within education We propose five standards that may be applied to
etnphasize commitment to, and dissemination of, assess if single-subject research results document a
evidence-based (or research-validated) practices practice as evidence based. The standards were
(Shavelson & Towne. 2002). Appropriate concern drawn from the conceptual logic for single-sub-
exists that investment in practices that lack ade- ject methods (Kratochwill & Stoiber), and from
quate empirical support may drain limited educa- standards proposed for identifying evidence-based
tional resources and, in some cases, may result in practices using group designs (Shavelson &
the use of practices that are not in the best inter- Townc, 2002).
est of children (Bcutler, 1998; Nelson, Roberts, Single-subject research documents a prac-
Mathur. & Rutherfbrd, 1999; Whitehurst, 2003). tice as evidence based when (a) the practice is op-
To support tbe investtnent in evidence-based erationally defined; (b) the context in which the
practices, it is appropriate for any research practice is to be ttsed is deftned; (c) the practice is
method to define objective criteria that local, state implemented with fidelity; (d) results from single-
or federal decision makers may use to detertnine subject research document the practice to be
if a practice is evidence based (Chaniblcss & Hol- ftmctionally related to change in dependent mea-
lon, 1998: Chambless & Ollendick. 2001; Odotn sures; and (e) the experimental effects are rcpli-

Exceptianal Children
 " Experimental control is demonstrated across a
sufficient range of studies, researchers, and partic-
Appropriate concern exists that invest- ipants to alloiv confidence in the effect. Docu-
ment in practices that lack adequate em- mentation of an evidence-based practice
pirical support may drain limited typically requires multiple single-subject stud-
ies. We propose the following standard: A
educational resources and, in some cases,
practice may be considered evidence based
may result in the use of practices that are when (a) a minimum of five single-subject
not in the best interest of children. studies that meet minimally acceptable
methodological criteria and document experi-
mental control have been published in peer-re-
cated across a sufficient number of studies, re- viewed journals, (b) the studies arc conducted
searchers, and participants to allow confidence in by at least three different researchers across at
the findings. Each of these standards is elaborated lea.sr three diffetent geographical locations, and
in the following list. (c) the five or more studies include a total of at
least 20 participants.
• The practice is operationally defined. A practice
must be described with sufficient precision so An example of applying these criteria is
that individuals other than the developers can provided by the literature assessing functional
replicate it with fidelity. communication training (FCT). As a practice,
FCT involves (a) using functional assessment pro-
• TJje context and outcomes associated with a prac- cedures to define the consequences (hat function
tice are clearly defined. Practices seldom are ex-as reinforcers (or undesirable behavior, (b) teach-
pected to produce all possible benefits for all ing a socially acceptable, and equally efficient, al-
individuals under all conditions. For a practice ternative behavior tbat produces the same
to be considered evidence based it must be de- consequence as the undesirable behavior, and (c)
fined in a context. This means operational de- minimizing reinforcement of the undesirable be-
scription of (a) the specific conditions where havior. Doctimentation of this practice as evi-
the practice should be used, (b) the individuals dence-based is provided by the following
qualified to apply the practice, (c) the popula- citations, which demonstrate experimental effects
tion(s) of individuals (and their functional in eight peer-reviewed articles across five major
charactetistics) for whom the practice is ex- research groups and 42 participants (Bird, Dores,
pected to be effective, and (d) the specific out- Moniz, & Robinson, 1989; Brown et al., 2000;
comes (dependent variables) affected by the Carr & Durand, 1985; Durand & Carr, 1987,
practice. Practices that are effective in typical 1991; Hagopian, Fisher, Sullivan, Acquisto, &
performance settings such as the home, school, LeBlanc, 1998; Mildon, Moore, & Dixon, 2004;
community, and workplace are of special Wackeretal.. 1990).
value.

• The practice is implcmcntrd with documented fi-CONCLUSION

delity. Single-subject research studies should
provide adequate documentation that the We offer a concise description of the features that
practice was implemented with fidelity. define single-subject research, the indicators that
can be used to judge quality of single-subject re-
• The practice is functionally related to change in search, and the standards for determining if an in-
X'tdued outcomes. Single-subject research studies tervention, or practice, is validated as evidence
should document a causal, or functional, rela- based via single-subject tnctbods. Single-subject
tionship between use of the practice and research offers a powerful and useful methodology
change in a socially important dependent vari- for improving the practices that benefit individu-
able by controlling for the effects of extraneous als with disabilities and their families. Any sys-
variables. tematic policy for promoting the development

Winter 2005
and/or dissemination of evidence-based practices porrunities, challenges, and cautions. School Psychology
in education should include single-subject re- Quarterly 17, 466-474.
search as an encouraged methodology. Cooper, L J., Wacker, D. P., Sasso, G. M., Rcimers, T.
M.. & Donn, L. K, (1990). LIsing parenrs as therapists
to evaluate appropriate behavior of their children; Ap-
REFERENCES plication to a teniary diagnostic clinic. Journal of Ap-
plied Behavior Analysis, 23, 285-296.
American Psychological Association. (2002). Criteria Didden, R., Dukcr, P. C, & Korzilius, H. (1997).
for evaluating treatment guidelines. American Psycholo- Meta-analytic study on treatment effecriveness for
gist, 57, 1052-1059. problem behaviors with individuals who have mental
Anderson, N. (2001). Design and analysis: A new ap- retardation. American Journal on Mental Retardation,
proach. Mahwah, NJ: Eribaum. 101. 387-399.
Beutler, L. (1998). Identifying empirical supported Dunlap. G., & Kern, L. (1997). The relevance of be-
rrcatments; What if we didn't^ Journal of Consulting and havior analysis to special education. In J. L Paul, M.
Clinical Psychology, 66, 113-120. Churton, H. Roselli-Kostoryz, W. Morse, K. Marfo, C.
Bird, F., Dores, P. A., Moniz, D., & Robinson, J. Lavely, & D. Thomas (Eds.), Foundations ofspecial edu-
(1989). Reducing severe aggressive and self-injurious cation: Basic knowledge informing research and practice in
behaviors with functional communication training: Di- special education (pp. 279-290). Pacific Grove, CA:
rect, collateral, and generalized results. American Jour- Brooks/Cole.
nal on Mental Reutrd/ttion, 94. 37-48. Durand, V. M., & Carr, E. G. (1987). Social influences
Birnbrauer, J. S. (1981). External validity and experi- on "sclf-stimulatoty" behavior: Analysis and treatment
mental investigation of individual behavior Analysis application. Journal of Applied Behavior Analysis. 20,
and Intervention in Developmental Disabilities, I, 117-119-132.
132. Durand, V. M., & Carr. E. G. (1991). Functional com-
Brown, K. A., Wacker. D. P., Derby, K. M.. Peck, S. munication training to reduce challenging behavior:
M., Richman, D. M., Sasso, G. M. et al. (2000). Eval- Maintenance and application in new si:tfin^s. Journal of
uating rfie effects of functional communication training Applied Behavior Analysis. 24, 251-264.
in the presence and absence of establishing operations. Durand, V. M., & Rost, N. (in press). Selection and at-
Journal of Applied Behavior Analysis, 33, 53-71. trition in challenging behavior research. Journal of Ap-
Busk, P., & Serlin, R. (1992). Meta-analysis ft>r single- plied Behavior Analysis.
participant research. In T. R. Kratochwill & J. R. Levin Faith, M. S.. Allison, D. B.. & Gorman. B. S. (1996).
(Eds.), Single-case research design and analysis: New di- Meta-analysis of single-case research. In R. D. Franklin,
rections Jbr psychology and education (pp. 187-212). D. B. Allison, & B. S. Gorman (Eds.), Design and anal-
Mahwah, NJ: Eribaum. ysis of single-case research (pp. 256-277). Mahwah, NJ:
Carr, E. G., & Durand, V, M. (1985). Reducing behav- Eribaum.
ior problems through functional communication train- Flay, B. R. (1986). Efficacy and effectiveness trials (and
ing. y(j«n'Wo//l/'/'//WfifAdf/i:>r/ln*i^m. 18, 111-126. other phases of researcfi) in the development of health
Carr, E. G., Levin, L., McConnachie, G., Carlson, J. I., promotion programs. Preventive Medicine, 15. 451-
Kemp, D. C , Smitfi, C. E. et al. (1999). Comprehcn- 474.
.sive multisituarionl intervention for problem behavior Gettinger, M. (1993). Effects of invented spelling and
in the community. Journal of Positive Behavior Interven- direct instruction on spelling performance of second-
tions, I, 5-25. grade boys. Journal of Applied Behavior Analysis, 26,
Chambless, D., & Hollon, S. (1998) Dcfming empiri- 281-291.
cally supported therapies. Journal of Consulting and Greenwood, C , Tapia. Y., Abbott, M.. & Walton, C.
Clinical Psychology, 66, 7-18. (2003). A building-based case study of evidence-ba.scd
Chamhiess. D.. & Ollendick. T. (2001). Empirically literacy practices: Implementation, reading behavior,
supported psychological interventions: Controversies and growtfi in reading fluency. K-4. The Journal of Spe-
and evidence. Annual Review of Psychology. 52, 685-716. cial Education 37, 95-110.
Christenson. S., Carlson, C , & Valdez, C. (2002). Evi- Gresham. E M.. Gansel, K. A, & Kurtz, P. F. (1993).
dence-based interventions in school psychology: Op- Treatment integrity in applied behavior analysis with

Exceptional Children
children. Journal of Applied Behavior Analysis, 26, 257- Marquis. J. G., Horner. R. H., Carr, E. G., Turnbull,
263. A. P., Thompson, M,, Behrens. G. A. et al. (2000). A
Hagopiaii. L. P., Fisher, W. W., Sullivan, M, T , Ac- meta-analysis of positive behavior support. In R. M.
quisto, )., & LcBlanc, L. A. (1998). Effectiveness of Getston & E. P. Schiller (Eds.), Contemporary special
functional communication training with and without education research: Syntheses of the knowledge base on
extinction and punishment: A summary of 21 inpatient critical instructional issues {pp. 137-178). Mahwab, NJ;
i:3!i€i. Journal of Applied Behavior Analysis, 31, 211-235. Erlbaum.

Hall, V. R., Axclrod, S., Tyler, L., Grief, E., Jones, E C , Martella, R., Nelson, J. R., & Marchand-Manella, N .
& Robertson, R. (1972). Modification of behavior (1999). Research methods: Learning to become a critical
problems in the home with a parent as observer and ex- research consumer. Boston: Allyn & Bacon.
[lerimenter. Journal of Applied Behavior Analysis, 5, 53- McReynolds, L. V., & Kearns. K. P. (1983). Single-sub-
64. ject experimental designs in communicative disorders. Bal-
timore: University Park Press.
Hersen, M., & Barlow, D. H. (1976). Single-case exper-
imemal designs: Strategies for studying behavior change. Mildon. R. L., Moore, D. W., & Dixon, R. S. (2004).
New York: Pergamon. Combining nonconcingent escape and functional com-
municarion training as a treatment for negatively rein-
Hersbberger, S. L.,Wallace, D. D., Green, S. B., &
forced disruptive hcWA\'\o{. Journal of Positive Behavior
Marquis, J. G. (1999). Meta-analysis of single-case de-
Interventions. 6, 92-102.
signs. In R. H. Hoyle (Ed.), Statistical strategiesfi}rs?nall
sample research (pp, 109-132). Ncwbury Park, CA: Miller. K.. Gunccr. P I.., Venn. M.. Hummel. J., &
Sage. Wiley, L, (2003). Effects of curricular and materials
modifications on academic performance and task en-
Jayaratne, S., &c Levy, R. L. (1979). Empirical clinical
gagement of three students with emotional or bebav-
practice. New York: Columbia University.
ioral disorders. Behavior Disorders, 28. 130-149.
Kazdin, A, E. (l')S2). Single-case research designs: Meth-
Moore, J. W., Edwards, R. P., Sterling-Tutner, H. E.,
ods for clinical and applied settings. New York; t)xford
Riley, J., DuBard, M., & McGeorge, A. ( 2 0 0 2 ) .
University Press.
Teacber acqui.sition of functional analysis methodology.
Kazdin, A. E. (1998). Research design in clinical psychol- Journal of Applied Behavior Analysis, 35, 73-77.
ogy (.3rd ed.). Boston: Allyn & Bacon.
Nckon, R., Roberts, M., Matbur, S., & Rutherford, R.
Kennedy, C. H. (In press). Single case designs for edwa- (1999). Has public policy exceeded our knowledge
tional research. Boston: Allyn & Bacon. base? A review of the functional bebavloraJ assessment
Koegel, I,. K., & Kocgel. R. L. (1986). The effects of literature. Behavior Disorders, 24. 169-179.
interspersed maintenance tasks on academic perfor- Odom, S., & Strain, P. S. (2002). lividence-based prac-
mance in a severe childbood stroke victim. Journal of tice in early intervention/early childhood special educa-
Applied Behavior Analysis, 19, 425-430. tion: Single-subject design resea-vch. Journal of Farly
Koegel, R. L , & Koegel, L. K. (1990). Extended reduc- Intervention, 25. 1 ^ 1 -160.
tions in srcreotypic bebavior of students witb autism Parsonson, B., & Bacr, D. (1978). The analysis and
through a self-management treatment package./flttrn<i/ presentation of graphic data. In 1. Kratochwill (Ed.),
of Applied Behavior Analysis, 2.3, ! 19-127, Single-subject research: Strategies for evaluating change
Kratocbwill, T , & I^vin, J. R. (1992). Single-case re- (pp. 105-165). New York: Academic Press.
search design and analysis: New directions for psychology Parsonson, B., Sf Baer, D. (1992). Visual analysis of
and education. HilLsdale, NJ: li,rlbauni. data, and current research into the stimuli controlling
Kratochwill, T., & Stoiber, K., (2002). Evidence-based it. In T. Kratochwill & J. Levin (Eds.), Single-case re-
interventions in school psychology: Conceptual foun- search design and analysis: New directions far psychology
dations for the procedural and coding manual of Divi- and education (pp. 15-40). Hillsdale, NJ: Erlbaum.
sion 16 and Society for the Study of School Psychology Richard, S. B., Taylor, R,, Ramasamy, R., & Richards,
Task Force. School Psychology Quarterly 17, 341-389. R. Y. (1999). Single-subject research: Applications in edu-
Mace, R C , Hock, M. L , Lalli, j . S., West, B. J., cational and clinical setting. Belmonr, CA: Wadswortb.
Elclfiore, P, Pinter, E. et al. (1988). Behavioral momen- Robena, E., Jitcndra, A., & Browder. D. M., (2002).
tum in the treatment of non-compliance. Comparison of the effects of Spanish and English con-
plied Behavior Analysis, 21, 123-141. stant time delay instruction on sight word reading by
Hispanic learners with mental retardation. Journal of Brook. JAMES HALLE (CEC #•>!), Professor,
Special Education. 36, 169-184. Department of Special Education, University of
Shavelson, R., & Townc, L. (2002). Scientific research Illitiois, Champaign, GAIL MCGEE (CEC #685).
in education. Washington, DC: National Academy Professor, Emory Autism Resource Center, Emory
Press. University, School of Medicine, Atlanta, Georgia.
Shernoff, E., Kratochwill. T., & Stoibet, K. (2002). Ev- SAMUEL ODOM (CEC #407), Ptofessor. School
idence-based interventions in school psychology: An il- of Education, Indiana University, Bloomington.
lustration of task force coding criteria using MARK WOLERY (CEC #98), Professot, Depart-
single-participant research designs. School Psychology ment of Special Education, Vanderbilt University,
Quarterly, 17, .390-422. Nashville, Tennessee.
Sidman, M. (1960). Tactia of scientific research: Evalu-
ating experimental data in psychology. New York: Basic
Books. Address all corrcspuniicncc to Robert H. Homer,
Tawney.J. W., & Gast, D. L. (1984). Single-subject re- Educational and Community Supports, 1235
search in special education. Columbus, OH: Merrill. University of Oregon, Eugene, OR 97403-1235;
Todman. )., & Dugard. P. (2001). Single-case andsmall- (541) 346-2462 (e-mail: robh@uoregon.edu)
n experimental designs: A practical guide to randomiza-
tion tests. Mahwah, NJ: Eribaum. Manuscript received December 2003; mantiscript
Wacker, D. P.. Steege, M. W., Northup, J., Sasso, G., accepted April 2004.
Berg, W, Reimers. T. et al. (1990). A component anal-
ysis of functional communication training across three
topographies of severe behavior problems. Journal of
Applied Behavior Analysis. 23. 417-429.
Weisz. J. R., & Hawley, K. M (2002). Procedural Ufui
coding manuai for identification of beneficial treatments.
Washington, DC: American Psychological Association,
Societj- for Clinical Psychology Division 12 Committee
on Science and Practice.
Whitehurst, G. J. (2003). Evidence-based education
[PnwetPoint presentation]. Retrieved April 8, 2004,
from htrp://www.ed.gov/off!ces/C)ER,I/presentations/
evidencebase.ppt
Wolery, M., & Dunlap. G. (2001). Reporting on stud-
ies using single-subject experimental methods./flwrnrf/
of Early Intertvntion, 24, 85-89.
INDEX OF ADVERTISERS
Wolery, M , & E/.ell, H. K. (1993). Subject descrip-
tions and single-subject research. Journal of Learning
Council for Exceptional Children, cover 2,
Disabilities. 26, 642-647.
p 129, cover 4
Wolf, M. M. (1978). Social validity: The case for sub-
jective measurement or how applied behavior analysis is
finding its heart. Journal of Applied Behavior Analysis, DePaul University, p 134

East Carolina University, p 180

ABOUT THE AUTHORS

New York University, p 180

ROBERT H. HORNER (CEC OR Federation), Rider University, p 134

Professor, Educartonal and Comtnunit)' Supports,
University of Oregon, Eugene, E D W A R D G . Wadsworth, cover 3
CARR (CEC #71), Professor. Dep;!rtment of Psy-
chology, State University of New York at Stony

ExceptUttud Children