Descriptive Assessment Method To Reduce Overall Disruptive Behavior in a Preschool Classroom

Abstract

Legal mandate and limited resources provide an impetus to increase the utility of functional assessment. Studies have demonstrated the utility of descriptive analysis-based interventions on an individual level (Lalli, Browder, Mace, & Brown, 1993; Mace & Lalli, 1991). The goal of this study was to develop a brief assessment that could be conducted in the natural setting to identify naturally occurring, high-frequency subsequent events that may serve as maintaining consequences for disruptive behavior using the entire class as the unit of analysis. Procedures were conducted in two early childhood classrooms during regularly scheduled classroom activities. Descriptive analyses were conducted by rotating among students every 90 s. Conditional probabilities from the aggregate class data were calculated to identify the most frequently occurring subsequent event(s) for disruptive behavior in each classroom. Following the descriptive analysis, the identified subsequent event was manipulated in an alternating treatments design (contingent delivery of the subsequent event and contingent withholding of the subsequent event) to validate the results of the descriptive analysis. In both classrooms, the descriptive analysis-based contingency reversal treatment more successfully suppressed disruptive behavior than a contraindicated treatment.

Full text

548
School Psychology Review,
2001, Volume 30, No. 4, pp. 548-567
Please direct correspondence to: Amanda VanDerHeyden, Louisiana State University Health Sciences Center,
Early Intervention Institute, 1900 Gravier St., 8B16, New Orleans, LA 70112; email: avande1@lsuhsc.edu.
The authors wish to thank Natalie Slider for assisting with data collection and Ed Daly for his comments on
an earlier version of this manuscript.
Copyright 2001 by the National Association of School Psychologists, ISSN 0279-6015
RESEARCH INTO PRACTICE
Descriptive Assessment Method to Reduce Overall
Disruptive Behavior in a Preschool Classroom
Amanda M. VanDerHeyden
Early Intervention Institute
Louisiana State University Health Sciences Center
Joseph C. Witt and Susan Gatti
Louisiana State University
Abstract. Legal mandate and limited resources provide an impetus to increase the
utility of functional assessment. Studies have demonstrated the utility of descrip-
tive analysis-based interventions on an individual level (Lalli, Browder, Mace, &
Brown, 1993; Mace & Lalli, 1991). The goal of this study was to develop a brief
assessment that could be conducted in the natural setting to identify naturally oc-
curring, high-frequency subsequent events that may serve as maintaining conse-
quences for disruptive behavior using the entire class as the unit of analysis. Pro-
cedures were conducted in two early childhood classrooms during regularly sched-
uled classroom activities. Descriptive analyses were conducted by rotating among
students every 90 s. Conditional probabilities from the aggregate class data were
calculated to identify the most frequently occurring subsequent event(s) for dis-
ruptive behavior in each classroom. Following the descriptive analysis, the identi-
fied subsequent event was manipulated in an alternating treatments design (con-
tingent delivery of the subsequent event and contingent withholding of the subse-
quent event) to validate the results of the descriptive analysis. In both classrooms,
the descriptive analysis-based contingency reversal treatment more successfully
suppressed disruptive behavior than a contraindicated treatment.
ferred activities or items, food, reprimands)
contingent on the occurrence of a behavior may
function as positive reinforcement for that be-
havior, strengthening the behavior. No specific
form of treatment (e.g., response cost, medi-
cation) has been shown to be sufficiently ef-
fective across all behavioral topographies, con-
tingencies, and situations. Thus, function-based
treatments, which are individually determined,
The purpose of functional assessment is
to identify the antecedent and/or subsequent
variables that are consistently associated with
the occurrence of the behavior targeted for as-
sessment. These assessment data link directly
to treatment by identifying potential maintain-
ing variables that can be manipulated to alter
the response-reinforcer relationship. For ex-
ample, presentation of events (e.g., praise, pre-

549
Descriptive Classwide Assessment
are quite effective in reducing problematic
behaviors and increasing appropriate behav-
iors (Iwata, Vollmer, Zarcone, & Rodgers,
1993). Indeed, function-based treatments have
yielded effective treatments for aggression
(Mace, Page, Ivancic, & O’Brien, 1986), bi-
zarre speech (Mace & Lalli, 1991), disruption
(Carr & Durand, 1985), pica (Mace & Knight,
1986), self-injury (Iwata, Dorsey, Slifer,
Bauman, & Richman, 1994), and stereotypy
(Repp, Felce, & Barton, 1988). Several meth-
ods of functional assessment are possible (e.g.,
analogue functional analysis, in vivo functional
analysis, descriptive analysis).
In a descriptive analysis, the observer
records, usually in a naturalistic setting, oc-
currences of the target behavior along with
temporally proximate antecedent and subse-
quent events. When antecedent or subsequent
events are highly correlated with the occur-
rence of the target behavior, it is generally in-
ferred that these antecedent or subsequent
events predict the occurrence of the target be-
havior. The mathematical depiction of this re-
lationship is called a conditional probability
(Bakeman & Gottman, 1986). Conditional
probabilities provide an indication of the de-
gree to which one observed behavior tends to
occur in temporal proximity to the occurrence
of another observed behavior. That is, condi-
tional probabilities answer the question, “When
a target behavior has occurred (i.e., condition),
what is the probability that another behavior
will either precede or follow the target behav-
ior in close temporal contiguity (i.e., probabil-
ity)?” For example, if the occurrence of ag-
gression is highly correlated with escape from
demands, the presentation of a demand is said
to predict the occurrence of aggression. Fur-
ther, it is hypothesized1 that a functional rela-
tionship exists between aggression and escape,
so that the client engages in aggression to ob-
tain escape from demands (i.e., negative rein-
forcement). Several studies have used condi-
tional probabilities to identify naturally occur-
ring stimuli and quantify their relationship to
target behaviors in classroom settings (Gunter,
Jack, Shores, Carrell, & Flowers, 1993;
Hendrickson, Strain, Tremblay, & Shores,
1982). Gunter et al. (1993) state that hypoth-
eses concerning the function of behaviors de-
rived from conditional probability data must
be experimentally verified. In their study, these
authors used conditional probability data to
quantify the interactions of two students in
special education classrooms. The conditional
probability data were analyzed to yield hypoth-
eses concerning the variables maintaining stu-
dent disruption. The resulting hypotheses were
examined within a reversal design in three of
the four cases. In these three cases, student dis-
ruption was lower during the treatment condi-
tions compared to baseline conditions, al-
though the treatment effects were somewhat
modest.
Descriptive analyses are not without
their limitations (e.g., absence of experimen-
tal control, high frequency irrelevant variables
may mask low frequency meaningful variables,
intermittent schedules of reinforcement may
be difficult to discern) and may result in in-
conclusive data or inaccurate hypotheses
(Lerman & Iwata, 1993). Yet, descriptive
analysis may allow the practitioner to conduct
functional behavioral assessment in a time-ef-
ficient manner. Further, descriptive analysis
provides an opportunity to observe and quan-
tify naturally occurring, possibly idiosyncratic
variables that may influence responding. Hy-
pothetical behavior functions generated by
descriptive analysis data can be experimentally
verified using treatment probes in the applied/
natural setting.
Several studies have attempted to iden-
tify ways of enhancing the utility of descrip-
tive analysis methods. For example, Mace and
Lalli (1991) combined descriptive and experi-
mental analyses to identify variables maintain-
ing bizarre speech exhibited by a developmen-
tally disabled adult. A descriptive analysis sug-
gested a possible attention and/or escape func-
tion for the target behavior. An experimental
analysis supported an attention function, and
interventions designed to alter the bizarre
speech-attention relationship effectively sup-
pressed bizarre speech. Thus, a descriptive
analysis identified two potential maintaining
variables for the problem behavior and subse-
quent analyses confirmed one of the functions.
Lalli, Browder, Mace, and Brown (1993) con-

550
School Psychology Review, 2001, Volume 30, No. 4
ducted descriptive analyses of problem behav-
iors exhibited by 3 students in special educa-
tion classrooms to generate hypotheses con-
cerning the variables maintaining the problem
behaviors exhibited by each student. These
hypotheses were then tested directly by hav-
ing the teacher provide the hypothesized rein-
forcing event contingent on occurrence of the
targeted problem behavior and indirectly by
having the teacher provide an intervention de-
signed to withhold the hypothesized reinforcer
contingent on the target behavior (i.e., extinc-
tion) and deliver the hypothesized reinforcer
contingent on an appropriate alternative behav-
ior (i.e., differential reinforcement). These re-
sults supported the hypothetical functional
relationships suggested by the descriptive
analyses for each student. In all three cases,
the target behaviors were successfully sup-
pressed and targeted alternative adaptive be-
haviors were successfully established. Sasso
et al. (1992) trained teachers to record ante-
cedent-behavior-consequence sequences
during activities designed to simulate typi-
cal functional analysis conditions (e.g., low
attention, high demand) and compared the
results to experimenter-conducted func-
tional analyses and teacher-conducted func-
tional analyses. They found similar results
across analyses for each of the two subjects.
In each case, the authors of these studies rec-
ommended using descriptive analysis to iden-
tify potential maintaining variables in the natu-
ral environment and then implement functional
analysis procedures to directly test each po-
tential reinforcing function.
Based on earlier findings that child be-
havior influenced adult responding (Carr, Tay-
lor, & Robinson, 1991; Taylor & Carr, 1992),
Taylor and Romanczyk (1994) attempted to
generate hypotheses about the function of in-
dividual students’ problem behavior based
solely on the amount of teacher attention re-
ceived by students during small group instruc-
tional sessions. First, Taylor and Romanczyk
observed small groups of 3 children each and
recorded the amount of teacher attention de-
livered to each child. The authors predicted that
students receiving large amounts of teacher
attention relative to their group peers would
exhibit problem behavior maintained by
teacher attention (Carr et al., 1991). They fur-
ther predicted that students receiving lower
levels of teacher attention would exhibit prob-
lem behavior maintained by escape (Taylor &
Carr, 1992). In the second phase of their study,
the authors conducted a brief experimental
analysis based on the procedures described by
Cooper, Wacker, Sasso, Reimers, and Donn
(1990) to validate the findings of the group
descriptive assessment. Taylor and Romanczyk
(1994) found that students receiving low
amounts of teacher attention in the first phase
exhibited problem behavior most frequently
during the difficult task demand conditions of
the experimental analysis irrespective of the
level of attention provided for appropriate be-
havior (i.e., escape function). Students receiv-
ing high amounts of teacher attention relative
to their peers in the first phase exhibited prob-
lem behavior most frequently in the restricted
attention conditions of the experimental analy-
sis irrespective of task demands (i.e., attention
function).
Several authors have begun to examine
the possibility of extending these assessment
methods to groups as the unit of analysis.
Atwater and Morris (1988) conducted a group
descriptive analysis to characterize the relation-
ship between teachers’ instructions and
children’s compliance in a preschool class-
room. In this study, observers rotated among
teachers and students every 4 min, recording
the occurrence and type of teacher instruction
(e.g., imperative, question) and child behav-
ior. Teacher-provided consequences (coded as
either positive, negative, or prompting) for
compliance and noncompliance were recorded.
These authors found that child compliance was
related to the context (e.g., art activity, transi-
tion) in which the instruction was delivered,
but not the form of the instruction (e.g., direct
imperatives, etc.).
Hoier, McConnell, and Pallay (1987)
used a partial-interval recording system to note
the occurrence of teacher and child behaviors,
as well as the occurrence of three-term se-
quences (e.g., teacher instruct-child attend-
teacher praise). Instead of rotating among sub-
jects, Hoier et al. (1987) observed only 1 child

551
Descriptive Classwide Assessment
per session and then computed the mean oc-
currence of each behavior in the classroom. The
authors used the calculated means to compare
differences in teacher behaviors between class-
rooms. They also compared measured behav-
iors of a target student to mean student behav-
iors in a template-matching paradigm, the idea
being that if the target student’s behavioral pro-
file approximated the classroom behavioral
profile, then the student would be successful
in that classroom. They found that their obser-
vation system was differentially sensitive to
between-classroom differences in teacher and
child behavior, that observation of two-teacher-
nominated “index” students approximated ag-
gregate means for each behavior category, and
they made recommendations for placement and
programming changes to enhance child-envi-
ronment fit based on their assessment results.
Although several studies have observed
and recorded overall rates of group behavior,
the purpose of the studies has been to evaluate
differences in overall behavior under varying
conditions. The group assessment was not
linked to an intervention in any of the studies
mentioned; thus, it is unclear if a group de-
scriptive analysis can be used to generate a
valid treatment for reducing overall levels of
disruptive behavior in the classroom. Although
descriptive analysis data have been used to
identify effective interventions on an individual
level (Lalli et al., 1993; Mace & Lalli, 1991),
the extent to which such analyses will gener-
ate effective treatments for an entire group is
unclear. The goal of this study was to evaluate
a brief descriptive assessment that was con-
ducted in the natural setting to identify natu-
rally occurring, high-frequency subsequent
events that may serve as maintaining conse-
quences for disruptive behavior.
Method
Setting, Materials, and Participants
Classroom 1. Sessions were conducted
during the morning “circle” activity at a nurs-
ery school program for children with speech
delays. Circle occurred in a carpeted area of
the classroom with chairs. The teacher identi-
fied circle as both the most problematic activ-
ity and the most important instructional period
of the day. Circle activities were interactive and
typically included a song, a story, and a related
language-based activity. For example, one
story was The Three Little Pigs and the related
activity involved building the three types of
houses and knocking two of them down. Dur-
ing circle time, the children were given mul-
tiple opportunities to respond and they were
required to exhibit turn-taking skills, staying
in their seats, and choral responding.
This classroom consisted of 8 children
between the ages of 2 and 4, a head teacher
(master’s level speech pathologist), a gradu-
ate student in speech pathology from a local
university, and two classroom aides. All of the
children were diagnosed as having mild to
moderate speech delay. Two of the 8 children
were diagnosed with autism, and 1 child was
diagnosed with hypothyroidism. Circle time
was scheduled to occur during a 20-min block
of time. Thus, one 20-min session was con-
ducted daily during regularly scheduled circle
time. One researcher participated in the circle
activity during reprimand sessions, and a dif-
ferent researcher participated in the circle ac-
tivity during differential reinforcement of al-
ternative behavior (DRA) sessions in an at-
tempt to enhance discriminability of treatment
conditions. The researcher prompted the adults
to deliver the contingency according to the type
of session being conducted. One person was
designated to lead the activity and this adult
was instructed to continue with the lesson while
the other adults managed disruptive behavior.
Classroom 2. Sessions were conducted
in a classroom of 4-year-old students at a local
Headstart center. The participating teacher
identified all teacher-directed large and small
group activities as most problematic in her
classroom. She nominated morning and after-
noon circles as the most problematic activities
of the day. Because this classroom contained
22 students, it was not feasible to conduct ses-
sions sampling each student’s behavior (i.e.,
this would have required too much time). Thus,
the teacher conducted circle in her regular
classroom with a randomly selected subset of
students during center time (i.e., the circle ac-
tivity was treated as a center). Informed con-
sent forms were distributed to parents of 6 ran-

552
School Psychology Review, 2001, Volume 30, No. 4
domly selected children and the parents of all
6 consented for their children to participate.
Each session consisted of a story and songs,
held roughly equivalent across sessions. The
children were given multiple opportunities to
respond, including choral responding, indi-
vidual responses to teacher questions, and non-
verbal responses (e.g., turning the page of the
book). The children were required to attend to
the teacher, sit quietly in chair, and exhibit fre-
quent choral responding.
One researcher served as the classroom
assistant during baseline and reprimand ses-
sions. A different researcher served as the class-
room assistant during DRA sessions. During
baseline, the teacher was instructed to conduct
circle as she normally would. She was in-
structed to treat the researcher as a classroom
aide, prompting the researcher to assist her in
managing the children as she normally would.
During baseline, the researcher sat in circle and
interacted with the children as prompted by the
classroom teacher. One researcher acted as a
classroom aide during each phase of the study
in order to maintain a constant adult-to-child
ratio across phases. During reprimand sessions,
the researcher assisted the teacher in redirect-
ing students who exhibited disruptive behav-
ior. Two 10-min sessions were conducted daily.
Response Definitions
Behaviors were defined specifically for
each classroom. Two target child behaviors,
one peer behavior, and five to eight teacher
behaviors were recorded. Behaviors coded
were attention, tangible, demand, compliance,
escape, and disruptive behavior. Specific defi-
nitions are provided in Tables 1, 2, and 3. Ob-
servers coded the behavior of the teacher and
classroom aides collectively as teacher behav-
ior. Because the definition of disruptive behav-
ior was intended to encompass all instances of
problematic behavior as identified by prelimi-
nary observations and confirmed by teacher
report, appropriate behavior was defined as the
absence of disruptive behavior for data analy-
sis purposes. In treatment sessions, teachers
were instructed to deliver consequences for
Table 1
Ta r get Child Behaviors
Disruptive Behavior Compliance
Classroom 1 Talking out, getting out of seat, Target child initiates response within 10 s.
talking in louder than conversational
volume, hitting, biting, kicking,
removing clothes so that skin is
exposed, touching the book without
being invited by the teacher to do so.
Classroom 2 Out of seat for more than 2 seconds, Target child initiates response within 2 s.
talking out, touching book when not
invited to do so, talking in louder
than conversational volume (i.e.,
yelling), crying, tantruming,
aggression, removing clothes so
that skin is exposed.
Table 2
Peer Behaviors
Classroom 1 Peer speaks to or touches the target child.
Classroom 2 Target child touches or talks to peer or peer talks to or touches target.

553
Descriptive Classwide Assessment
instances of appropriate behavior that included
sitting in chair, not talking out or touching
neighbor, complying with teacher requests, and
task participation.
A partial-interval recording procedure
was used to record the occurrence of behav-
iors across 10-s intervals. Antecedent events
were defined as having occurred within the
same interval or one interval prior to disrup-
tive behavior. Specific antecedent events
coded included peer attention, teacher atten-
tion delivery and removal, demand, and tan-
gible delivery and removal. Although anteced-
ent events are frequently conceptualized as the
absence of an event that is then delivered sub-
sequent to the behavior (i.e., attention) or the
delivery of an event that is then removed fol-
lowing the behavior (i.e., demand), the possi-
bility was considered that peer attention and
tangible delivery might tend to precede dis-
ruptive behavior (e.g., tangible item becomes
available to child and child begins to play with
the item instead of participating in the activ-
ity). Specific subsequent events coded in-
cluded peer attention, teacher attention deliv-
ery and removal, demand, and tangible deliv-
Table 3
Teacher Behaviors
Attention Tangible Escape Demand
Classroom 1 Delivery—verbal Delivery—teacher Termination of Teacher makes a
statements made delivers a tangible an instruction or request of the
to the child by item (e.g., book) ongoing instruc- entire class (e.g.,
the teacher, to the target child. tional context line up) or a
statements of Removal—teacher specific request to
praise directed removes a tangible the child (e.g., I
toward the entire item from the need you to put
group, or hugging target child. away that toy), or
the child, patting prompting behavior.
the child on the Prompting—Teacher
shoulder, or other- reminds or guides
wise touching the the child to emit an
child. Excludes adaptive behavior.
demands.
Removal—Cessation
of attention that
had occurred con-
tinuously for at least
10 seconds.
Attention Tangible Escape Demand
Classroom 2 Adult touches or Delivery—tangible The following Request made to
talks to the target or edible item is sequence only: target child that
child specifically. delivered to the demand, disruptive requires an observ-
Excludes state- target child or behavior, then any able response from
ments directed to child has tangible behavior other than the child. Includes
the entire group. item in his or her another demand or requests made to
Excludes demands. possession or demand, compliance, the whole group and
child obtains the and any behavior prompting.
item. other than another Prompting—Teacher
Removal—Teacher demand reminding or guiding
removes a tangible the child to emit an
item from the adaptive behavior.
target child.

554
School Psychology Review, 2001, Volume 30, No. 4
ery and removal. Subsequent events were de-
fined as having occurred subsequent to disrup-
tive behavior within the same interval or the fol-
lowing interval and adults were prompted to wait
10 s following occurrence of disruptive behav-
ior to praise appropriate behavior during treat-
ment sessions. These data enabled a compari-
son of the specific antecedent and subsequent
events most frequently associated with disrup-
tive behavior. Percent interval occurrence of dis-
ruptive behavior (i.e., the dependent variable)
was calculated for each session and these data
constituted the pretreatment baseline data.
Data Analysis
Each response on the observational code
was assigned a number so that each number
represented one of the behaviors targeted for
observation. For example, disruptive behavior
was coded as “1” and peer attention was coded
as “9.” Observation sheets were divided into
60 small rectangular blocks. Observers re-
corded the corresponding numeric code in an
interval block as each response occurred, gen-
erating a sequence of numbers in each interval
block on the data sheet. From these series of
numbers, data were calculated to determine
which events occurred in temporal proximity
to disruptive behavior. In some cases, re-
sponses may have continued for more than 10
s (i.e., the duration of one interval). If a teacher
or student continued responding across inter-
vals, the relevant codes were scored at the start
of each subsequent interval until the behavior
ceased. Each behavior was coded a maximum
of one occurrence per interval. Thus, in some
cases, an event could count as antecedent or
subsequent to a maximum of two occurrences
of disruptive behavior (i.e., one occurrence of
attention could follow two instances of disrup-
tive behavior meeting criteria as a subsequent
event for both occurrences of disruptive be-
havior), whereas each occurrence of disruptive
behavior was counted as one occurrence. Se-
quences of behavior were recorded within-chil-
dren only and not across-children. Therefore,
conditional probabilities were not calculated
based on sequences of events across-children.
In other words, if one child was given a de-
mand in the final interval before the observa-
tion shifted to the next participant, and the next
participant exhibited disruptive behavior in the
subsequent interval, then demand was not con-
sidered antecedent to disruptive behavior for
this particular occurrence of disruption.
Conditional probabilities were calculated
according to the procedures described by
Lerman and Iwata (1993). Conditional prob-
abilities for antecedent events were calculated
as follows: the total of each antecedent cat-
egory (e.g., antecedent attention) was divided
by total interval occurrence of disruptive be-
havior resulting in a disruptive behavior prob-
ability for each respective antecedent category.
The total of each antecedent category for events
preceding disruptive behavior was divided by
the total interval occurrence of that event. The
result was the proportion of antecedent events
preceding disruptive behaviors of all occur-
rences of antecedent events. This proportion
was calculated for each antecedent category.
Conditional probabilities for subsequent events
were calculated the same way.
Together, the disruptive behavior prob-
abilities and proportions of antecedent or sub-
sequent events to all events described the de-
gree of correlation between a particular con-
tiguous event and the problem behavior. Spe-
cifically, the disruptive behavior values indi-
cated which events were most frequently as-
sociated with the occurrence of disruptive be-
havior; whereas, the antecedent and subsequent
event proportions indicated the extent to which
the relationship may be spurious (i.e., high de-
gree of association simply because the event
occurred across multiple intervals). An event
that tended to occur most frequently in rela-
tion to disruptive behavior and to not occur in
intervals where disruptive behavior did not
occur may be more likely to represent a func-
tional relationship between that event and dis-
ruptive behavior (Lerman & Iwata, 1993). For
example, if attention occurred subsequent to
disruptive behavior 93% of the times that dis-
ruptive behavior occurred, but occurred as a
subsequent event for disruptive behavior on
only 20% of the occasions that attention oc-
curred, then attention may be less likely to be
functionally related to disruptive behavior (i.e.,
may be spurious). Alternatively, if attention

555
Descriptive Classwide Assessment
occurred subsequent to disruptive behavior
93% of the times that disruptive behavior oc-
curred, and occurred as a subsequent event for
disruptive behavior 100% of the times that at-
tention occurred, then attention may be more
likely to be functionally related to disruptive
behavior. Therefore, disruptive behavior cal-
culations reflected the degree of sensitivity of
the assessment, or the degree to which disrup-
tive behavior was associated with a given event.
The antecedent and subsequent event proportions
reflected the degree of specificity of the assess-
ment, or the degree to which an antecedent or
subsequent event was specifically related to the
occurrence of disruptive behavior.
Experimental Design
The descriptive analysis data served as
baseline. Following baseline, an assessment
validation phase was implemented. In this phase,
two interventions, derived from a descriptive
analysis, were implemented in each class fol-
lowing an alternating treatments design (Hayes,
Barlow, & Nelson-Gray, 1999). An alternating
treatments design was chosen because disrup-
tive behavior was not a low-frequency behav-
ior in each classroom, multiple response op-
portunities existed for students to exhibit dis-
ruptive behavior, the interventions were expected
to produce fairly immediate effects on disrup-
tive behavior, and carryover effects were not
expected. Additionally, an alternating treatments
design allows for implementation and evalua-
tion of treatment effects when baselines are un-
stable, and treatment comparisons can be accom-
plished within a relatively short period of time.
Whereas alternating treatments designs avoid
the threat of sequence effects, the possibility
of multiple treatment interference cannot be
avoided. Thus, it is possible that the effect of
each treatment is influenced by its juxtaposi-
tion against the other so that the level of be-
havior obtained was different from what would
have been obtained if either treatment had been
presented in isolation (Kazdin, 1982).
Procedure
Teacher interview and observation.
The consultant conducted an interview with
each teacher to explain the steps of the project,
obtain informed consent from the teacher,
determine the best times to conduct classroom
observations, and obtain information regarding
the most problematic behaviors in each teacher’s
classroom. The consultant conducted one obser-
vation during the teacher-nominated problem-
atic activity (i.e., circle for each class) to define
disruptive behavior. After this observation the
consultant discussed the definition of disruptive
behavior with the teacher and the teacher was
asked to make any changes she thought were
necessary to adequately measure the occur-
rence of disruptive behavior in her classroom.
Descriptive analysis and baseline.
The descriptive analysis was conducted by ro-
tating observations from one student to another
in a predetermined, systematic manner every
90 s (Handen, McAuliffe, Janosky, Feldman,
& Breaux, 1994). Observers used a 10-s par-
tial-interval recording procedure to record be-
havior in continuous sessions. A session oc-
curred when at least 6 of the 8 children were
present in the first classroom. In the second
classroom, sessions were conducted when at
least 4 of the 6 children were present. In both
cases, the 2 children reported by the teacher to
most frequently exhibit disruptive behavior had
to be present for sessions to be conducted.
Once the children were seated in the
circle area, observers started with the left-most
chair and rotated their observations around the
circle of children. This process resulted in a
random order of observation because the chil-
dren sat in different places most days (i.e., there
was no assigned seating). Session duration was
20 min for the first classroom and 10 min for
the second classroom in all phases.
Assessment validation. Treatment type
was manipulated in this phase to evaluate the
efficacy of the classwide descriptive analysis.
Two treatments were compared during this
phase. One treatment procedure was an inter-
vention that would be expected to reduce suc-
cessfully disruptive behavior based on the re-
sults of the descriptive analysis (i.e., contin-
gency reversal). The second treatment strategy
involved a direct test of the influence of the
most frequently associated subsequent event
(i.e., providing the subsequent event contin-

556
School Psychology Review, 2001, Volume 30, No. 4
gent on disruptive behavior—a contraindicated
treatment). For example, if the descriptive
analysis indicated that attention was the most
frequent event associated with disruptive be-
havior, then praising instances of appropriate
behavior while ignoring disruptive behavior
(i.e., DRA) would have been an appropriate
treatment, whereas attending to disruptive be-
havior in the form of reprimands would have
constituted a direct test of the disruption-at-
tention relationship, and would have been a
contraindicated treatment strategy. DRA was
selected as the indicated treatment strategy
because it represented a mirror-image of the
contraindicated treatment and still provided for
attention delivery (so that amount of attention
could be yoked across sessions).
Total instances of attention (i.e., collaps-
ing teacher attention and demand delivery)
were averaged across baseline sessions to yield
an average percent interval occurrence score
for attention. In the second classroom, an at-
tempt was made to yoke the level of attention
delivered during assessment validation ses-
sions to baseline levels, while manipulating the
contingency according to the session type. That
is, during reprimand treatment sessions, an
amount of attention roughly equivalent to the
average amount of total teacher attention de-
livered in baseline sessions (i.e., for disruptive
and appropriate behavior) was provided con-
tingent on inappropriate behavior, and vice
versa during differential reinforcement of al-
ternative behavior (i.e., DRA) attention ses-
sions. An observer monitored attention deliv-
ery and prompted the consultant to deliver at-
tention as needed to ensure a more closely
matched schedule of attention delivery for the
second classroom.
Following the selection of two treatment
strategies, the consultant and teacher(s) imple-
mented the treatments in an assessment vali-
dation phase. During DRA sessions, adults
praised appropriate behavior (e.g., “good lis-
tening,” giving a thumbs-up sign, patting a stu-
dent on the back) and ignored disruptive be-
havior. During reprimand sessions, adults re-
directed students exhibiting disruptive behav-
ior (e.g., “I need you to sit in your chair”) and
physically guided compliance following a ver-
bal redirection. The consultant trained the
teacher to implement the indicated and con-
traindicated interventions in the classroom. The
consultant explained the intervention to the
teacher, answered questions, and role-played
the exact steps of each intervention. Prior to
each session, the consultant reviewed the pro-
cedure with the teacher. During the session,
the consultant provided assistance and prompt-
ing as needed to ensure treatment integrity. The
consultant was unaware of which child was
being observed at any given time. In the first
classroom, a consultant joined the circle ac-
tivity during the assessment validation phase
and prompted and assisted the teacher in de-
livering prescribed consequences. In Class-
room 2, a consultant joined circle during
baseline to control for the presence of an addi-
tional adult during the circle activity.
The teacher was encouraged to treat the
consultant as a classroom aide during baseline
sessions. During baseline, the consultant did
not interact with students unless the teacher
specifically directed the consultant to assist her
in managing a problem. During the assessment
validation phase, the consultant prompted and
assisted the teacher in delivering prescribed
consequences. A consultant participated dur-
ing assessment validation for two purposes.
One purpose was to achieve intervention in-
tegrity. The second was to allow the activity
and lesson to continue with minimal disrup-
tion. Thus, the teacher was encouraged to con-
tinue the lesson while simultaneously deliver-
ing consequences (e.g., allowing students to
turn the page of the book contingent upon good
behavior, giving a thumbs up sign while sing-
ing, physically placing children back in the
chair next to the teacher while singing). The
consultant assisted the teacher in delivering
consequences so that the teacher could con-
tinue teaching (e.g., placing children back in
their seats when the teacher was reading a
story). Thus, an attempt was made to compare
the alternating contingencies under very simi-
lar conditions (i.e., circle activity during which
similar activities were conducted in the same
order, and similar response requirements were
placed upon students in a structured and en-
riched environment) within each class.

557
Descriptive Classwide Assessment
Interobserver Agreement and
Procedural Integrity
Graduate students enrolled in a school
psychology doctoral program and one under-
graduate research assistant served as observ-
ers. Prior to the beginning of the study, observ-
ers were trained in a pilot classroom. First,
observers used a flashcard training procedure
to memorize the behavioral codes, and were
required to emit 40 correct responses per
minute to meet training criteria. Second, ob-
servers collected data in a pilot classroom and
discussed specific agreements and disagree-
ments in operational definitions. Observers
were required to achieve 80% reliability dur-
ing a training session to be considered success-
fully trained.
A second observer independently ob-
served and recorded behaviors for 50% of ses-
sions in the first classroom and 75% of ses-
sions in the second classroom. Reliability data
were recorded across all session types in all
phases for each classroom. Interobserver
agreement (IOA) was calculated on an inter-
val-by-interval basis by dividing the number
of agreements by number of agreements plus
disagreements and multiplying the result by
100%. Mean IOA for disruptive behavior was
94% (range = 90% to 97%) for Classroom 1
and 91% (range = 85% to 95%) for Classroom
2. IOA was also calculated for teacher behav-
iors in each classroom. Mean IOA for teacher
behaviors was 97% (range = 96% to 99%) for
Classroom 1 and 98% (range = 97% to 100%)
for Classroom 2. Percent agreement for the
occurrence and nonoccurrence of each behav-
ior code was also calculated using the same
method as that used to calculate IOA. Mean
occurrence agreement for disruptive behavior
was 83% (range = 72% to 92%) for Classroom
1 and 71% (range = 53% to 83%) for Class-
room 2. Mean occurrence agreement for all
teacher behaviors was 73% (range = 49% to
90%) for Classroom 1 and 73% (range = 53%
to 100%) for Classroom 2. Mean
nonoccurrence agreement for disruptive be-
havior was 90% (range = 87% to 96%) for
Classroom 1 and 87% (range = 84% to 94%)
for Classroom 2. Mean nonoccurrence agree-
ment for all teacher behaviors was 96% (range
= 94% to 98%) for Classroom 1 and 98%
(range = 97% to 100%) for Classroom 2. The
complex nature of the observation plan (i.e.,
large number of behaviors observed and rotat-
ing sequence) may have affected occurrence
agreement (Kazdin, 1982). Because some less
than desirable occurrence agreement estimates
were obtained, sequence reliability was calcu-
lated to indicate the degree to which conclu-
sions based upon the data collected were reli-
able. Specifically, sequence reliability was used
to determine the extent to which observers re-
corded the same subsequent events for each
occurrence of disruptive behavior. For each
occurrence of disruptive behavior, the number
of agreements for the recorded subsequent
event was divided by the number of agreements
and disagreements and multiplied by 100% to
yield percent agreement scores for each po-
tential behavior-subsequent event sequence.
Mean sequence reliability was 86% (range =
72% to 98%) for Classroom 1 and 84% (range
= 57% to 100%) for Classroom 2. Procedural
integrity scores, indicating the degree to which
experimental procedures were correctly imple-
mented, were calculated for each classroom for
all the treatment sessions by dividing the num-
ber of times the correct consequence was de-
livered (i.e., reprimand or ignore) by the num-
ber of intervals during which disruptive behav-
ior occurred and multiplying by 100%. The
mean procedural integrity score was 75%
(range = 41% to 94%) for Classroom 1 and
81% (range = 45% to 100%) for Classroom 2.
Results
Conditional probabilities of each subse-
quent and antecedent event for Classroom 1
are provided in Tables 4 and 5. Conditional
probabilities of the antecedent events were
undifferentiated across sessions and were, thus,
not included in further analyses or manipula-
tions (i.e., assessment validation). Figure 1
depicts the conditional probabilities of each
subsequent event given the occurrence of dis-
ruptive behavior for Classroom 1. The most
prevalent subsequent event for disruptive be-
havior, on average, was teacher attention de-
livery. The average probability of obtaining
teacher attention for disruptive behavior (.42)

558
School Psychology Review, 2001, Volume 30, No. 4
Table 4
Conditional Probabilities of Events Subsequent to Disruptive
Behavior for Classroom 1
Session Attention Peer Demand Tangible
Number Delivery Attention Removal Delivery
1 0.52 NA NA 0.06
2 0.45 0.03 0.05 0.05
3 0.29 0.07 0.03 0.07
4 0.22 0.04 0.06 0.08
5 0.37 0 0.06 0.03
6 0.67 0.07 0.05 0.05
11001
2 0.71 0.1 1 0.25
3 0.58 0.45 1 0.63
41 0.5 0.67 0.67
5 0.75 0 0.33 0.22
61 0.4 1 0.29
Calculated as a Proportion
of Disruptive Behavior
Calculated as a Proportion of All
Subsequent Events for Category
was greater than the probability of obtaining
attention for appropriate behavior (i.e., the
absence of disruptive behavior) (.30) during
baseline sessions. Additionally, in four of the
six baseline sessions, the probability of obtain-
ing teacher attention for disruptive behavior
was greater than the probability of obtaining
teacher attention for appropriate behavior.
Based on the high frequency sequence of dis-
ruption-attention, two treatments were derived
Note. NA= Event did not occur during session.
Figure 1. Conditional probabilities for each subsequent event given the occur-
rence of disruptive behavior for Classroom 1 and calculated as a proportion of
disruptive behavior.

559
Descriptive Classwide Assessment
Table 5
Conditional Probabilities of Events Occurring Prior to
Disruptive Behavior for Classroom 1
Calculated as a Proportion of
Disruptive Behavior
Calculated as a Proportion of All
Antecedent Events for Category
Note. NA = Event did not occur during session.
Session Attention Escape Tangible Tangible Peer
Delivery Removal Attention
1 0.43 NA 0.06 0 0
2 0.47 0.05 0.08 0.03 0.03
3 0.32 0.02 0.07 NA 0.07
4 0.1 0.04 0.06 0 0.04
5 0.29 0 0.03 0 0.03
6 0.5 0.05 0.05 NA 0.02
1 0.45 NA 0.75 0 0
2 0.41 1 0.38 0.33 0.09
3 0.43 0.5 0.57 NA 0.4
4 0.12 0.67 0.5 0 0.5
50.29 0 0.11 0 0.5
6 0.4 1 0.29 NA 0.2
from these results: the indicated treatment or
contingency reversal was DRA (i.e., provid-
ing attention contingent on appropriate behav-
ior while ignoring disruptive behavior). The
contraindicated treatment or direct test of the
influence of the subsequent event upon disrup-
tive behavior was providing a reprimand con-
tingent on inappropriate behavior (i.e., provid-
ing attention contingent on inappropriate be-
havior). That is, the descriptive analysis indi-
cated that attention frequently followed the
occurrence of disruption in the classroom. Fur-
ther analyses indicated that attention more re-
liably occurred following disruptive behavior
than appropriate behavior.
These results led to the hypothesis that
disruptive behavior may be maintained by
teacher attention in this classroom. When the
maintaining variable is known (or suspected),
several intervention strategies can be em-
ployed to suppress the targeted maladaptive
behavior and increase the targeted adaptive
behavior. For example, the intervention agent
may provide a very dense schedule of the hy-
pothesized reinforcer (in this case, attention)
presumably to reduce the child’s motivation
to engage in the behavior that will reliably pro-
duce that reinforcer (i.e., the reinforcer is pro-
vided free of cost, so there is no need for the
subject to engage in the behavior). The inter-
vention agent may also withhold the reinforcer
when the target behavior occurs, so that the
child learns that engaging in the behavior no
longer results in the reinforcer (i.e., extinc-
tion). The intervention agent may provide the
reinforcer contingent on an adaptive behavior
or contingent on the absence of the targeted
maladaptive behavior (i.e., differential rein-
forcement). In this case, the treatment team
(i.e., consultants and teacher) selected a com-
bined intervention of ignoring disruption (i.e.,
extinction) and praising appropriate behavior
(i.e., differential reinforcement).
Figure 2 depicts percent interval occur-
rence of disruptive behavior during baseline
and the assessment validation phase for Class-

560
School Psychology Review, 2001, Volume 30, No. 4
room 1. Each data point represents a 20-min
session. One session was conducted per day.
Assessment validation spanned seven sessions
beginning by random selection with the indi-
cated treatment. On average, disruptive behav-
ior occurred in 39% of intervals during
baseline sessions in a variable pattern. On av-
erage, disruptive behavior occurred in 32% of
intervals during the contraindicated treatment
sessions (i.e., reprimands) in a relatively stable
pattern and possible upward trend. On aver-
age, disruptive behavior occurred in 23% of
intervals during the indicated treatment ses-
sions (i.e., DRA attention) in a variable pat-
tern. There was one overlapping data point
within the first three sessions during the as-
sessment validation phase. Perfect treatment
integrity was never achieved during DRA ses-
sions. That is, children continued to access
attention contingent on some occurrences of
disruptive behavior, perhaps contributing to
the variable pattern observed across DRA ses-
sions. Nonetheless, disruptive behavior oc-
curred with less frequency during all DRA
sessions than during the lowest frequency
baseline session. In contrast, disruption oc-
curred at least as frequently as the lowest fre-
quency baseline session during all reprimand
sessions.
During baseline, teacher attention oc-
curred on average 41 times per session. Dur-
ing DRA sessions, attention was delivered
contingent on appropriate behavior 44 times
per session on average. During reprimand ses-
sions, attention was delivered contingent on
disruptive behavior 36 times per session on
average. Thus, the total occurrence of tempo-
rally proximate (baseline) and contingent (as-
sessment validation) attention was roughly
equivalent across phases. That is, the amount
of attention did not vary, whereas the contin-
gency did during the assessment validation
phase. Figure 2 also depicts delivery of atten-
tion for disruptive behavior across phases for
Classroom 1, which provides a representation
of the degree of treatment integrity. During
DRA sessions, disruption-contingent attention
should have approximated zero occurrences.
That is, attention that occurred during DRA
sessions should have occurred in the form of
praise contingent upon appropriate behavior.
During reprimand sessions, disruption-contin-
gent attention should have occurred approxi-
mately 41 times per session and always sub-
sequent to the occurrence of disruption. That
is, attention that occurred during reprimand
sessions should have occurred in the form of
redirective statements contingent upon disrup-
tive behavior.
Conditional probabilities of subsequent
and antecedent events for Classroom 2 are pro-
vided in Tables 6 and 7. Conditional probabili-
ties of the antecedent events were largely un-
differentiated across sessions and were, thus,
not included in further analyses or manipula-
tions (i.e., assessment validation). Figure 3
depicts the conditional probabilities of each
subsequent event given the occurrence of
Figure 2. Percent interval occurrence of disruptive behavior during baseline
and assessment validation for Classroom 1. Instances of disruption-contingent
adult attention across phases for Classroom 1.

561
Descriptive Classwide Assessment
Table 6
Conditional Probabilities of Events Subsequent to Disruptive
Behavior for Classroom 2
Session Attention Escape Tangible Peer
Attention
1 0.68 0.03 NA 0.23
20.64 0.05 NA 0.11
3 0.7 NA NA 0.1
4 0.5 0.04 NA 0.04
5 0.82 0 NA 0
6 0.5 0.14 NA 0.29
11 1NA1
2 0.71 1 NA 0.4
3 0.58 NA NA 0.5
41 1NA1
5 0.75 0 NA 0
61 1NA0.8
Calculated as a Proportion of
Disruptive Behavior
Calculated as a Proportion of All
Subsequent Events for Category
Note. NA = Event did not occur during session.
disruptive behavior for Classroom 2. The most
prevalent subsequent event associated with
disruptive behavior was teacher attention. Peer
attention was the second most prevalent sub-
sequent event associated with disruptive be-
havior. Further, the average probability of ob-
taining teacher attention for disruptive behav-
ior (.64) exceeded the average probability of
obtaining attention for appropriate behavior
(.07) in baseline sessions. In fact, in all
baseline sessions the probability of obtaining
teacher attention for disruptive behavior ex-
ceeded the probability of obtaining teacher
attention for appropriate behavior. Based on a
hypothesis of attention as the maintaining vari-
able for disruption, two treatments were de-
rived from these results. Again, attending to
appropriate behavior while placing disruptive
behavior on extinction was the indicated treat-
ment (i.e., DRA). The contraindicated treat-
ment was providing a reprimand for each in-
stance of disruptive behavior (i.e., providing
attention contingent on disruptive behavior).
Figure 4 depicts percent interval occur-
rence of disruptive behavior during baseline
and the assessment validation phase for Class-
room 2. Each data point represents a 10-min
session. Two sessions were conducted each
day. On average, disruptive behavior occurred
in 31% of intervals during baseline sessions
in a highly variable pattern with a possible un-
stable and perhaps even decreasing trend. The
assessment validation phase began by random
selection with the indicated treatment and ses-
sion order was counterbalanced across days.
In the assessment validation phase, DRA con-
sistently resulted in lower levels of disruptive
behavior. During DRA sessions, disruptive
behavior occurred on average during 16% of
intervals in a relatively stable pattern with a
possible decreasing trend. During reprimand
sessions, disruptive behavior occurred on aver-

562
School Psychology Review, 2001, Volume 30, No. 4
Table 7
Conditional Probabilities of Events Occurring Prior to
Disruptive Behavior for Classroom 2
Session Attention Escape Tangible Peer
Attention
1 0.26 0.03 NA 0.13
20.32 0 NA 0.11
3 0.1 0 NA 0.2
4 0.23 0.04 NA 0
5 0.55 0 NA 0
6 0.14 0 NA 0.14
1 0.57 1 NA 0.8
2 0.35 0 NA 0.4
3 0.08 0 NA 1
4 0.5 1 NA 0
5 0.5 0 NA 0
6 0.4 0 NA 0.4
Note. NA = Event did not occur during session.
Calculated as a Proportion
of Disruptive Behavior
Calculated as a Proportion of All
Antecedent Events for Category
Figure 3. Conditional probabilities for each subsequent event given the occur-
rence of disruptive behavior for Classroom 2 and calculated as a proportion of
disruptive behavior.
age during 27% of intervals in a relatively
stable pattern with a possible increasing trend.
There were no overlapping data points be-
tween session types during the assessment
validation phase. However, disruptive behav-
ior occurred with the same frequency as the
lowest frequency baseline session during two
of the three DRA sessions, indicating a more
modest overall effect. Thus, the primary ef-
fect of the indicated treatment for Classroom
2 may have been to stabilize disruptive be-
haviors at low levels.

563
Descriptive Classwide Assessment
During baseline, contingent attention
occurred on average 15 times per session. Dur-
ing DRA sessions, attention was delivered con-
tingent on appropriate behavior 17 times per
session on average. During reprimand sessions,
attention was delivered contingent on disrup-
tive behavior 15 times per session on average.
Thus, the total occurrence of temporally proxi-
mate (baseline) and contingent (assessment
validation) attention did not differ across
phases. Figure 4 depicts attention delivery for
disruptive behavior across all phases for Class-
room 2, which provides a representation of the
degree of treatment integrity. During DRA ses-
sions, disruption-contingent attention should
have approximated zero occurrences. That is,
attention that occurred during DRA sessions
should have occurred in the form of praise con-
tingent upon appropriate behavior. During rep-
rimand sessions, disruption-contingent atten-
tion should have occurred approximately 15
times per session and always subsequent to the
occurrence of disruption. That is, attention that
occurred during reprimand sessions should
have occurred in the form of redirective state-
ments contingent upon disruptive behavior,
which is precisely what happened.
Discussion
The purpose of this study was to exam-
ine the validity of basing treatment on the re-
sults of a whole-class descriptive analysis con-
ducted by rotating observations among all the
children in one classroom and a subset of chil-
dren in a second classroom. Although the re-
sults for Classroom 2 were more modest than
the results for Classroom 1, in each classroom,
the interventions derived from group descrip-
tive analyses produced differentiated results.
Further, a more successful suppression in dis-
ruptive behavior was obtained with the treat-
ment indicated by the descriptive analysis in
the assessment validation phase in both class-
rooms. These findings are preliminary and re-
quire replication, but suggest that whole-class
analyses merit further scrutiny and research.
The results of this study are similar to
studies that have successfully linked descrip-
tive analysis to effective treatments for indi-
vidual clients with multiple behavior problems
(Lalli et al., 1993; Mace & Lalli, 1991). That
is, this study formulated treatment hypotheses
based upon descriptive analysis data and vali-
dated the descriptive analysis results in an as-
sessment validation phase. Additionally, this
project extends the findings of previous stud-
ies that have attempted to measure classwide
levels of target behaviors by incorporating a
validation measure (Atwater & Morris, 1988;
Hoier et al., 1987; Taylor & Romanczyk,
1994). For example, Hoier et al. (1987) dem-
onstrated that aggregate observation data
could represent behavioral averages in a class-
room setting. The present study based treat-
Figure 4. Percent interval occurrence of disruptive behavior during baseline
and assessment validation for Classroom 2. Instances of disruption-contingent
adult attention across phases for Classroom 2.

564
School Psychology Review, 2001, Volume 30, No. 4
ment hypotheses upon aggregate data and
demonstrated that the treatment indicated by
the aggregate data resulted in greater reduc-
tions in the target behavior. It remains unclear
whether or not the event most frequently asso-
ciated with disruptive behavior across children
actually represented a functional relationship.
Future studies should examine the degree to
which the type of analysis described in this
paper increases the efficiency with which ser-
vices are provided to classroom teachers and
their students and the degree to which inter-
ventions developed from this type of assess-
ment are more effective than general interven-
tions that a school psychologist might recom-
mend without the benefit of observational data.
These findings are limited by several
additional important considerations. Less than
optimal levels of interobserver occurrence
agreement and potential sequencing effects due
to the order in which students were observed
each day may have served as sources of error.
The lower reliability estimates were obtained
during the descriptive analysis phase when the
highest number of behaviors were being ob-
served, and thus, may have affected the de-
scriptive analysis results, which, in turn, may
have affected the veracity of the analysis-based
hypothesis. This potential source of error may
have been somewhat attenuated by the rela-
tively reasonable sequence reliability esti-
mates that were obtained.
In the first classroom, a general suppres-
sion in disruptive behavior in the assessment
validation phase could have resulted from the
presence of an additional adult (i.e., the con-
sultant was present during assessment valida-
tion but not during baseline). This possibility
was controlled for in the second classroom by
adding the consultant during baseline. Al-
though differentiated patterns of responding
were obtained during the assessment valida-
tion phase for both classrooms, disruptive be-
havior was not completely eliminated using
contingency reversal. Low levels of disrup-
tive behavior may have persisted for several
reasons. One possible explanation is that com-
plete extinction did not occur. That is, teach-
ers occasionally attended to disruptive behav-
ior despite consultant prompts to ignore. Other
possible explanations are that the descriptive
analysis did not identify all sources of rein-
forcement in the natural setting and disruptive
behavior may have been multiply controlled.
Because our observation code did not separate
topographies of disruptive behavior, differen-
tial effects in severity of the disruptive behav-
iors that persisted cannot be quantified. How-
ever, teachers and observers noted anecdotally
that disruptive behaviors exhibited during
baseline and reprimand conditions were more
severe (i.e., aggression, jumping off chairs,
yelling, leaving circle) than disruptive behav-
iors exhibited during the DRA condition (i.e.,
getting out of seat, touching the book without
permission, talking during story). Because the
descriptive analysis indicated a possible con-
tingent relationship between attention and dis-
ruption in both classrooms, it is unclear
whether or not this type of analysis can suc-
cessfully identify potential maintaining vari-
ables other than attention. Future studies are
needed to resolve this question.
The indicated treatment in Classroom 1
achieved a more significant effect than in
Classroom 2; yet, responding during the as-
sessment validation phase in Classroom 1 was
more variable than responding in Classroom
2, despite similar levels of treatment integrity
across classrooms. At least two important
intraclass differences existed. First, Classroom
1 included children identified as exhibiting
some sort of developmental delay; whereas,
Classroom 2 included only typically develop-
ing children. Second, Classroom 1 maintained
a smaller teacher-to-child ratio. It is unclear
whether these or other intraclass differences
contributed to the different patterns of re-
sponding observed across classrooms. Further,
a limited number of sessions were conducted
in each phase and long-term follow-up data
were not collected. Thus, it remains unclear
whether or not intervention effects would
maintain and whether or not teachers could
conduct the intervention without researcher
assistance.
Direct testing of the hypothetical re-
sponse-reinforcer relationship in the natural
environment may have several advantages over
other methods of experimental verification

565
Descriptive Classwide Assessment
(e.g., analogue functional analyses). Given dif-
ferentiated descriptive analysis results, assess-
ment verification initially could include only
tests of those variables that appear to be re-
lated to the occurrence of the target behavior,
decreasing the amount of assessment time re-
quired and the potential risk of establishing
previously unfamiliar events as reinforcing.
Should the descriptive analysis-based treat-
ment prove unsuccessful, a functional analy-
sis could be conducted to examine the poten-
tial influence of unobserved events upon tar-
get behaviors. That is, a functional analysis
could be used to examine the influence of low-
frequency events (that might be missed during
descriptive analysis) upon the target behavior.
Additionally, the utility of the analysis
presented in this article would be enhanced by
recording the data in such a way that individual
student data could be analyzed separately
should the group intervention fail to produce
clinically meaningful changes in student be-
havior. For example, in this study, when stu-
dents chose a chair for circle, observers re-
corded students’ names next to the blocks dur-
ing which they would be observed prior to be-
ginning the observation. Names were recorded
to ensure that students were not observed in
the same order each day. However, these data
could be analyzed separately to develop hy-
potheses concerning the function of disruptive
behavior for individual students. Individual
student data were not analyzed separately be-
cause teachers in both classes expressed satis-
faction with the treatment effects obtained dur-
ing the group intervention.
This assessment process could lend it-
self to a model of assessment progressing from
less complex to more complex until differen-
tiated patterns of responding occur and a deci-
sion can be made (e.g., Vollmer, Marcus,
Ringdahl, & Roane, 1995). That is, practitio-
ners may begin with a classwide behavior ob-
servation during the problematic class activity
and graph time-series representation of the con-
ditional probabilities of antecedent and subse-
quent events. If no discernible pattern is ob-
served, the practitioner may examine individual
descriptive assessment data for each of the stu-
dents about whom the teacher has expressed a
concern. The next step might be brief analogue
experimental analysis, followed by treatment
probes conducted in the classroom setting to
validate the intervention, followed by extended
multi-element experimental analysis, followed
by additional manipulations until experimental
control is demonstrated. That is, the practitioner
need only proceed with further assessment until
differentiated response patterns are observed and
a hypothesis can be formulated.
Conducting sessions in the natural envi-
ronment may result in enhanced generalization
(e.g., training with stimuli common to the natu-
ral environment, capturing naturally occurring
discriminative stimuli, matching schedules of
programmed reinforcement during treatment
to naturally occurring schedules during
baseline). It is the belief of many early inter-
ventionists that multiple prompts embedded
within the natural environment occasion be-
haviors and that these events cannot be recre-
ated adequately in analogue settings (Bricker
& Cripe, 1992). Thus, assessment conducted
in the natural setting in the presence of the same
classroom conditions is likely to enhance ex-
ternal validity. Nonetheless, because this study
was conducted during circle time for both
classes, the results do not necessarily general-
ize to other stimulus conditions that commonly
occur in preschool classrooms (e.g., snack
time, free play, center time).
Because descriptive analysis includes
naturally occurring reinforcers in clients’ ev-
eryday environments with minimal alteration
of contextual stimuli, it offers the possibility
of enhanced treatment utility with typically
developing students (i.e., the possibility of
generating a treatment that is effective in the
relevant environment). Presumably, once the
probability of accessing the reinforcer for ap-
propriate responding exceeds the probability
of accessing the reinforcer for inappropriate
responding, the appropriate response can be
predicted to occur. However, this prediction is
complicated by several issues (e.g., dimensions
of reinforcement, learning history, behavioral
economics, rule-governed behavior, discrimi-
nation of contingencies, substitutability). The
goal seems to lie in the translation of the meth-
odological precision of functional analysis to

566
School Psychology Review, 2001, Volume 30, No. 4
relevant, everyday contexts. Some researchers
have suggested that one way to bridge this gap
in the methodology is to teach parents and
teachers to conduct descriptive analyses and
then conduct a series of treatment probes to
determine which treatment best achieves the
desired effect (Wacker, 1998). Such a process
would allow for ongoing assessment and revi-
sion as needed to produce effective treatments
that persist across time and settings. Further
research is needed to identify variations of de-
scriptive analysis procedures to increase their
utility in applied settings.
Footnote
1Note, however, that this relationship is in-
ferred and may not be entirely accurate in some cir-
cumstances. The presentation of demands does not
necessarily predict the occurrence of aggression in
all cases. For example, imagine that 100 demands
are presented and aggression occurs following 10
presentations of demands. In this case, the presen-
tation of a demand does not reliably predict the oc-
currence of aggression despite the sole occurrence
of aggression when demands are presented, because
the sequence of demand-aggression only occurs a
small number of times relative to total number of
demand presentations. This issue is dealt with more
thoroughly in the section titled “Data Analysis.”
References
Atwater, J. B., & Morris, E. K. (1988). Teachers’ instruc-
tions and children’s compliance in preschool class-
rooms: A descriptive analysis. Journal of Applied Be-
havior Analysis, 21, 157-167.
Bakeman, R., & Gottman, J. M. (1986). Observing inter-
action: An introduction to sequential analysis. New
York: Cambridge University Press.
Bricker, D., & Cripe, J. J. (1992). An activity-based ap-
proach to early intervention. Baltimore: Paul H.
Brookes Publishing.
Carr, E. G., & Durand, M. (1985). Reducing behavior prob-
lems through functional communication training. Jour-
nal of Applied Behavior Analysis, 18, 111-126.
Carr, E. G., Taylor, J. C., & Robinson, S. (1991). The ef-
fects of severe behavior problems in children on the
teaching behavior of adults. Journal of Applied Be-
havior Analysis, 24, 523-535.
Cooper, L. J., Wacker, D. P., Sasso, G. M., Reimers, T.
M., & Donn, L. K. (1990). Using parents as therapists
to evaluate appropriate behavior of their children: Ap-
plication to a tertiary diagnostic clinic. Journal of Ap-
plied Behavior Analysis, 23, 285-296.
Gunter, P. L., Jack, S. L., Shores, R. E., Carrell, D. E., &
Flowers, J. (1993). Lag sequential analysis as a tool
for functional analysis of student disruptive behavior
in classrooms. Journal of Emotional and Behavior
Disorders, 1 (3), 138-148.
Handen, B. L., McAuliffe, S., Janosky, J., Feldman, H., &
Breaux, A. M. (1994). Classroom behavior and chil-
dren with mental retardation: Comparison of children
with and without ADHD. Journal of Abnormal Child
Psychology, 22, 267-280.
Hayes, S. C., Barlow, D. H., & Nelson-Gray, R. O. (1999).
The scientist practitioner: Research and accountabil-
ity in the age of managed care. Needham Heights, MA:
Allyn & Bacon.
Hendrickson, J. M., Strain, P. S., Tremblay, A., & Shores,
R. E. (1982). Interactions of behaviorally handicapped
children: Functional effects of peer social initiations.
Behavior Modification, 6, 323-353.
Hoier, T. S., McConnell, S., & Pallay, A. G. (1987). Ob-
servational assessment planning and evaluating edu-
cational transitions: An initial analysis of template
matching. Behavioral Assessment, 9, 5-19.
Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K.
E., & Richman, G. S. (1994). Toward a functional
analysis of self-injury. Journal of Applied Behav-
ior Analysis, 27, 215-240. (Reprinted from Analy-
sis and Intervention in Developmental Disabilities,
2, 1-20, 1982).
Iwata, B. A., Vollmer, T. R., Zarcone, J. R., & Rodgers, T.
A. (1993). Treatment classification and selection based
on behavioral function. In R. Van Houten & S. Axelrod
(Eds.), Behavior analysis and treatment (pp. 101-125).
New York: Plenum Press.
Kazdin, A. E. (1982). Single-case research designs: Meth-
ods for clinical and applied settings. New York: Ox-
ford University Press.
Lalli, J. S., Browder, D. M., Mace, F. C., & Brown, D. K.
(1993). Teacher use of descriptive analysis data to
implement interventions to decrease students’ problem
behaviors. Journal of Applied Behavior Analysis, 26,
227-238.
Lerman, D. C., & Iwata, B. A. (1993). Descriptive and
experimental analysis of variables maintaining self-
injurious behavior. Journal of Applied Behavior Analy-
sis, 26, 293-319.
Mace, F. C., & Knight, D. (1986). Functional analysis and
treatment of severe pica. Journal of Applied Behavior
Analysis, 19, 411-416.
Mace, F. C., & Lalli, J. S. (1991). Linking descriptive and
experimental analyses in the treatment of bizarre
speech. Journal of Applied Behavior Analysis, 24, 553-
562.
Mace, F. C., Page, T. J., Ivancic, M. T., & O’Brien, S. (1986).
Analysis of environmental determinants of aggression
and disruption in mentally retarded children. Applied
Research in Mental Retardation, 7, 203-221.
Repp, A. C., Felce, D., & Barton, L. E. (1988). Basing the
treatment of stereotypic and self-injurious behaviors
on hypotheses of their causes. Journal of Applied Be-
havior Analysis, 21, 281-289.
Sasso, G. M., Reimers, T. M., Cooper, L. J., Wacker, D.,
Berg, W., Steege, M., Kelly, L., & Allaire, A. (1992).
Use of descriptive and experimental analyses to iden-
tify the functional properties of aberrant behavior in
school settings. Journal of Applied Behavior Analysis,
25, 809-821.

567
Descriptive Classwide Assessment
Taylor, J. C., & Carr, E. G. (1992). Severe problem be-
haviors related to social interaction: Attention seeking
and social avoidance. Behavior Modification, 16, 305-
335.
Taylor, J. C., & Romanczyk, R. G. (1994). Generating
hypotheses about the function of student problem be-
havior by observing teacher behavior. Journal of Ap-
plied Behavior Analysis, 27, 251-265.
Vollmer, T. R., Marcus, B. A., Ringdahl, J. E., & Roane,
H. S. (1995). Progressing from brief assessments to
extended experimental analysis in the evaluation of
aberrant behavior. Journal of Applied Behavior Analy-
sis, 28, 561-576.
Wacker, D. P. (1998, May). Functional analyses as “Bridge
Studies”: Further consideration of the social validity
of analogue assessments. In C. Piazza (Chair), The
current status and future directions of functional analy-
sis. Symposium conducted at the meeting of the 24th
Annual Convention of the Association for Behavior
Analysis, Orlando, FL.
Amanda VanDerHeyden, Ph.D., is Assistant Professor and Assistant Director of Research
at Louisiana State University Health Sciences Center in the Early Intervention Institute.
Her research interests include applied behavior analysis, early intervention, teacher and
parent training, and generally finding ways to assist children and families in achieving
greater habilitation. This project was completed when she was a student at Louisiana State
University.
Joseph Witt, Ph.D., is Professor of Psychology at Louisiana State University. He has
received recognition in the form of Editor of School Psychology Quarterly, LSU Alumni
Distinguished Professor, Editor of Guilford School Practitioner Series, and Associate Edi-
tor of the Buros Mental Measurements Yearbook and School Psychology Review. He is
interested in ways to enhance children’s learning.
Susan Gatti, M.A., is completing her Ph.D. in School Psychology at Louisiana State Uni-
versity under the direction of Dr. George Noell. Her research interests include functional
assessment of academic and behavior problems in high incidence populations.