Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

Generating Randomized Schedules for


Direct Observations in Microsoft® Office Excel®
Richard L. Azulay & Derek D. Reed

Abstract

The process of scheduling classroom observations—either for staff or for student performance data
collection—is often done haphazardly or out of simple convenience. However, slapdash approaches to scheduling
are rich with possible confounds to the validity of the data being collected. Due to the possibility of detrimental
outcomes such as reactivity or stimulus control, classrooms observations ought to be scheduled in a calculated
“randomized” approach. In this technical article, we outline the various ways in which unsystematic scheduling of
observation may evoke undesirable trends in data. We then present a task analysis detailing ways in which a
supervisor may use Microsoft ® Office Excel®—a widely available spreadsheet program—to create randomized
schedules to promote the integrity of the dependent variables being collected during scheduled observations.
Implications of using simple spreadsheets to assist supervisors in scheduling “random” observations are discussed.
Keywords: behavioral assessment, data collection, Microsoft ® Office Excel®, technology

The active treatment ingredient in any operant-based behavior change procedure is the provision
of some form of consequence made contingent upon a target response (see Skinner, 1953; 1957). Whether
in the form of an edible, a praise statement, or the simple provision of data regarding one’s performance,
the science of operant behavior dictates that this consequential feedback will impact future instances of
behavior in some capacity—either increasing or decreasing the likelihood that this behavior is emitted
again in the future. Moreover, this notion is echoes that of Skinner’s eloquently worded description in
Science and Human Behavior (1953, pg. 59), that “The consequences of behavior may "feed back" into
the organism. When they do so, they may change the probability that the behavior which produced them
will occur again.” Perhaps more importantly, the scheduling of such consequential feedback has proven to
be paramount in the kinds of behavioral patterns subsequently observed (see Ferster & Skinner, 1957).

Whether by design or act of nature, feedback schedules – similar to other consequential


schedules, such as reinforcement or punishment – come as either interval- or ratio-based forms. Interval-
based schedules deliver feedback contingent upon the first occurrence of some target response after some
pre-specified passage of time (e.g., a child reinforced for his/her first correct response after 5-min).
However, ratio-based schedules dictate that the target response must occur a pre-specified number of
times before it is provided feedback. (e.g., a child is reinforced after he/she emits five correct responses).
As such, interval-based schedules typically produce lower response rates, relative to ratio-based schedules
(see Ferster & Skinner, 1957). This is in part due to the fact that during ratio-based schedules, the
responding organism is to some extent able to govern the rate of feedback through his/her own behavior
since the feedback is directly contingent upon response rate.

In addition, all feedback schedules, whether interval- or ratio-based, feature an additional


requirement characteristic. Namely, as either being fixed (i.e., after a static number of required responses;
e.g., a child is reinforced after every fifth correct response) or variable (i.e., after a dynamic number of
required responses which averages to some specified value; e.g., on average, a child’s behavior is
reinforced for approximately every five correct responses) contingencies. Like interval- and ratio-
schedules, fixed- and variable -schedules have specific behavioral patterns associated with their schedule
type. Specifically, fixed-schedules tend to produce pauses in responding after the delivery of feedback,
while variable -schedules tend to produce steady response rates despite instances of feedback delivery (see
Ferster & Skinner, 1957). This phenomenon likely occurs because organisms may more accurately predict

349
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

feedback delivery in fixed-schedules, but remain persistent in variable -schedules since the next feedback
event cannot be predicted. Moreover, given these features, responding on fixed-schedules is often easier
to extinguish than responding on variable -schedules (see Ferster & Skinner, 1957).

While applied behavior analysts are typically well-versed in the programming of reinforcement
schedules for their clients, they may not be aware of the kinds of feedback schedules imposed on their
own behaviors concerning the active supervision and observations of other staff or clients on their
caseload (cf. Reed, Fienup, Luiselli, & Pace, in press). Such schedule requirements may result from
program/school policies and procedures, or be the result of competing priorities or responsibilities.
Nevertheless, not taking a systematic approach to scheduling may be costly due to the phenomenon
discussed above regarding schedule -induced behavioral patterns and the fact that the quality of
supervision feedback to staff is crucial to maintaining adequate levels of staff performance in human
service organizations (see Reid, 1998; Reid & Parsons, 2002). Like any organism responding on a
schedule of feedback, staff behavior will conform to the kinds of schedules and schedule combinations
imposed on them.

In the workplace, identifying natural feedback schedules may be difficult due to the fact that no
intentional reinforcement schedule may have been programmed. However, typical workplace schedules
are likely a combination of interval and ratio schedules in which staff must meet some quota by a
particular deadline (Duvinsky & Poppen, 1982). In particular, Duvinsky and Poppen describe such
workplace schedules as a conjunctive fixed-interval fixed-ratio (i.e., FI FR) schedule wherein some ratio
of responses (FR) is reinforced after a particular passage of time (FI). The experimental analysis of
behavior literature suggests that responding on conjunctive FI FR schedules features response rates that
are initially high, and then decelerate after a brief pause in responding (Ferster & Skinner, 1957;
Herrnstein & Morse, 1958). In practice, such conjunctive schedules may be imposed in situations where
supervisors must complete a certain number of feedback tools to his/her staff or complete some
prescribed number of observations in a given time frame.

While such discussion may seem superfluous, the supervisor’s response patterns on such
schedules may produce inadvertent effects on those individuals he/she is scheduled to supervise and/or
observe. For instance, consider a supervisor working in a school which dictates that s/he must deliver a
number of feedback tools (e.g., treatment integrity checks, observations of lesson plan or behavior support
plan implementation, etc.) to his/her staff by the end of the month. Such a prescribed schedule will likely
result in a passage of time where the supervisor completes no observations/supervisions, and then
completes all the requirements in one brief stint—this would thus be in accordance with what would be
expected on a conjunctive FI FR schedule or with a scallop effect in FI responding when the organism
waits until near the end of the interval before completing the response requirement (Ferster & Skinner,
1957). A recent study conducted by Reed and colleagues (in press) confirmed that staff under such
schedule requirements do indeed produce a scallop effect akin to FI or conjunctive FI FR schedules. The
unwanted effects of such schedule -induced observation patterns could be that the observed staff begin to
expect the observations at the same time each week/month, and that the presence of the supervisor in the
staff’s classroom becomes associated with supervision/feedback which could produce reactivity.
Reactivity would thus spoil the supervisor’s interpretation of the competence or performance of the
individual s/he is supervising (see Brackett, Reid, & Green, 2007). Thus, supervisors must take extra
precautions in their scheduling to avoid such confounds.

One such measure that supervisors may take in limiting the risk of reactivity from the individuals
they observe is to program observation times in both a well-distributed and pseudo-randomized format. In
particular, such scheduling would limit the degree to which the supervisor scallops his/her observations
within his/her caseloads. With less scalloping and more consistent presence in the observed setting, the
supervisor would reduce the possible stimulus control of the analyst over the staff’s adherence to their job

350
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

requirements and/or behavior support plan implementation. Behavior analysts must ensure that no
confounds are present in their behavioral assessments of students or staff alike, as any form of reactivity
will inherently violate the assumption that behavioral assessment samples are representative of all
possible observation opportunities (see Cone, 1977). Perhaps more crucial to this discussion is that
service delivery and placement decisions for students may be at stake contingent upon the results of an
analyst’s observation in the natural environment. Thus, this technical report provides behavior analysts
with a simple user-friendly schedule calculator which is easily programmable on Microsoft® Office
Excel® (versions 1997 through 2007)—a widely used and available computer spreadsheet software
program. This technical report provides a task analysis for creating a randomized scheduler on their home
or office computer, using simple Excel® programming techniques. Upon creation, behavior analysts can
provide themselves with an antecedent intervention to ensure their observations and subsequent feedback
are done on a randomized schedule —thereby reducing violations to behavioral assessment assumptions
and limiting possible confounds to observation, such as reactivity.

Method

1. To begin creating the randomized scheduler, first open Microsoft® Office Excel® on the
computer desktop. Next, save the document with the name “random schedule” by clicking FILE and
moving the cursor down to SAVE AS and selecting this option by left-clicking the mouse. In the FILE
NAME field, key the words “random schedule.” Move the cursor to the left side of the box and click
DESKTOP to save the random schedule file to the computer’s desktop. Click SAVE1 . Be sure to save this
file frequently throughout the task analysis to prevent accidental losses should the program freeze or crash
for any reason.

2. Next, create a list of observee names in the first column. To do so, first create an “Observee”
header by clicking on cell A1, and keying the word “Observee.” This specifies the names of the
clients/staff to be observed using the randomized schedule. Next, specify the first observee by keying
his/her name (or “Observee 1”) in cell A2. For instructional purposes, let us assume that there are five
observees with whom to schedule randomized observations for. Thus, in cells A3 to A6, key these
individuals’ names as previously completed for “Observee 1.”

3. Next, create a list of observation types in the second column. First, create an “Observation”
header by clicking on cell B1, and keying the word “Observation.” This specifies the type of observation
to be conducted. Next, specify the first observation by keying its name (or “Observation 1”) in cell B2.
Again, for instructional purposes, let us assume that there are five types of observations which you would
benefit from a random schedule. Thus, in cell B3 to B6, key the observations’ names as previously
completed for “Observation 1.” See Figure 1 for depiction of how spreadsheet should appear at
completion of step 3.

Figure 1, Next Page!

351
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

Figure 1. Screenshot of completion of step 3.

4. Next, click on the second worksheet (“Sheet2”). See Figure 2 for depiction of where the
“Sheet2” tab is located.

Figure 2. Screenshot of the location of the “Sheet 2” tab (see step #4).

5. Now you will set up a chart for the randomized schedule to appear in. Create a “Date” header
by clicking on cell A1, and keying the word “Date.” Create an “Observee” header by clicking on cell B1
and keying the word “Observee.” Create an “Observation” header by clicking on cell C1, and keying the
word “Observation.”

352
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

6. Next, fill in the dates of observations. For the purpose of this example, schedule one
observation for each day of the week. Click on cell A2 and key the word “Monday.” In cells A3 to A6,
key the words “Tuesday,” “Wednesday,” “Thursday,” and “Friday.”

7. Now randomize the “Observee” column. Click on cell B2. Next, click on the formula bar. See
Figure 3 for depiction of location of formula bar. Key the function command:
“=INDEX(Sheet1!A2:Sheet1!A6,RANDBETWEEN(1,COUNTA(Sheet1!A2:A6)),1)”.2 Repeat this
process for cells B3 through B6, first clicking on the cell, then clicking on the formula bar and keying the
same function.

Figure 3. Screenshot of the location of the formula bar referenced in step 7.

8. Now randomize the “Observation” column. Click on cell C2. Click on the formula bar and key
the function command:

“=INDEX(Sheet1!B2:Sheet1!B6,RANDBETWEEN(1,COUNTA(Sheet1!B2:B6)),1). Repeat this


process for cells C3 through C6, first clicking on the cell, then clicking on the formula bar and keying the
same function.

9. The random schedule generator is complete. See Figure 4 for depiction of completed schedule
with randomization. Currently, if you key anything into any cell in the document, notice that the random
entries will change. In order to stop the formulas from re-calculating, click on OPTIONS, FORMULA,
and “MANUALLY UPDATE FORMULAS.” Key the “F9” key on the keyboard to randomize entries
into the cells.

353
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

Figure 4. Screenshot of completed schedule with randomization.

10. If you see any errors in the cells (#NAME? or #NUM?), check all cells to make sure
everything is entered as written above.

Alternative Clinical Applications

While this task analysis was specifically designed to reduce the lackadaisical scheduling of
observations to prevent scalloping and its subsequent confounds to behavioral assessment, this
randomizer may be modified to complete many of the other scheduling tasks a behavior analyst might
frequently encounter in their clinical work. Such adaptations need only be applied to the entry types in the
columns (e.g., changing the variable labels from “Observee” or “Observation” to meet idiosyncratic needs
of the behavior analyst). For example, rather than specifying the day of the observation (e.g., Monday),
one could include specific dates, or even specify particular weeks (e.g., “third week of May”). More
complex random schedules could be created by adding more variables—such as assessment tool/type--or
by increasing the ratio of a particular observation/observe related to other observation/observees. If there
exists a desire for a particular observee to receive more observations, or to conduct a particular
observation more frequently, simply enter it more times into the column of entries to be randomized.

While this randomizer is set to schedule observations daily for the supervisor doing the
observations, it also randomizes the variable (e.g., student or staff) being assessed. However, by simply
substituting the names of the observees with other variables, the aforementioned scheduler may be
adapted to randomize a host of behavioral assessment objectives. For instance, substituting the column
label “Observee” with “Assessment Type,” the user can randomize the sequencing of various behavioral
assessment observations. For instance, one observation may be targeting the integrity in which a behavior
support plan is being implemented, while another observation may target the fidelity of an academic
lesson plan execution. Thus, simply replace the observees’ names with the terms “Behavior Support
Integrity” and “Lesson Plan Integrity.” Other variables which be of utility to behavior analysts to

354
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

randomize inc lude the times of day for planned activity checks (Doke & Risley, 1972), times to conduct
social validity observations (e.g., indices of happiness; Green & Reid, 1996), or the sequencing of brief
experimental analysis (e.g. Daly, Martens, Hamler, Dool, & Eckert, 1999) or functional analysis (see
Hanley, Iwata, & McCord, 2003) conditions (e.g., attention, demand, play, and alone). No matter the
function of the user’s utilization of this scheduler, the user should note that this does not create a fixed or
variable ratio of observations, nor does it guarantee an increase in the frequency of a particular
observation or the prevention of unwanted confounds such as reactivity or scalloping. Rather, this tool
simply increases the chance that a particular observation/observe may be scheduled at any given time—
which, by logic, reduces the chances that these confounds could arise.

Summary

The task analysis for creating a random schedule was created using Microsoft® Office Excel®
2007, but was successfully tested on version 2003. As users begin utilizing the random scheduler in their
clinical practice, it is recommended that the user save different schedules for each classroom or group of
clients on their caseload. Specifically, having individualized spreadsheets may aid in the prevention of
unwanted data input errors and/or the inadvertent loss of information. We anticipate that through
randomizing one’s observations of staff or students, the behavior analyst will protect his/her behavioral
assessment from possible reactivity effects or data sampling errors. It is our hope that other researchers
will formally evaluate the role of reactivity in non-randomized observation schedules, as well as the
investigation of our claim that randomized schedules may curtail these issues. One such avenue of future
research may be the investigation of a comparison of staff performance when provided with a fixed and
predictable schedule of reinforcement to a staff performance when provided with a random schedule.
Moreover, research on staff satisfaction or choice of different schedules of feedback may be completed
using this scheduler tool. Such investigations may help to identify schedules of feedback or supervision
that aid in perceived quality of work life – a topic area of growing interest in the field of organizational
behavior management (see Reid & Parsons, 2002).

References

Brackett, L., Reid, D. H., & Green, C. W. (2007). Effects of reactivity to observations on staff
performance. Journal of Applied Behavior Analysis, 40, 191-195.

Cone, J. D. (1977). The relevance of reliability and validity for behavioral assessment. Behavior Therapy,
8, 411-426.

Daly, E.J.,III, Martens, B.K., Hamler, K.R., Dool, E.J., & Eckert, T.L. (1999). A brief experimental
analysis for identifying instructional components needed to improve oral reading fluency. Journal
of Applied Behavior Analysis, 32, 83-94.

Doke, L. A., & Risley, T. R. (1972). The organization of day-care environments: Required vs. optional
activities. Journal of Applied Behavior Analysis, 5, 405-420.

Duvinsky, J. D., & Poppin, R. (1982). Human performance on conjunctive fixed-interval fixed-ratio
schedules. Journal of the Experimental Analysis of Behavior, 37, 243-250.

Ferster, C. B.; Skinner, B. F. (1957). Schedules of reinforcement. New York: Appleton-Century-Crofts.

355
The Behavior Analyst Today Consolidated Volume 10, Number 3 & 4

Green, C. W., & Reid, D. H. (1996). Defining, validating, and increasing indices of happiness among
people with profound multiple disabilities. Journal of Applied Behavior Analysis, 29, 67-78.

Hanley, G. P., Iwata, B. A., & McCord, B. E. (2003). Functional analysis of problem behavior: A
review. Journal of Applied Behavior Analysis, 36, 147-185.

Herrnstein, R. J., & Morse, W. H. (1958). A conjunctive schedule of reinforcement. Journal of the
Experimental Analysis of Behavior, 1, 15-24.

Reed, D. D., Fienup, D. M., Luiselli, J. K., & Pace, G. M. (in press). Evaluating the scalloping effect in
treatment integrity data collection in a behavioral healthcare organization. International Journal
of Behavioral Consultation and Therapy.

Reid, D. H. (Ed.). (1998). Organizational behavior management and developmental disabilities services:
Accomplishments and future directions. New York: The Haworth Press, Inc.

Reid, D. H., & Parsons, M. B. (2002). Working with staff to overcome challenging behavior among
people who have severe disabilities: A guide for getting support plans carried out. Morganton,
NC: Habilitative Management Consultants, Inc.

Skinner, B. F. (1953). Science and human behavio r. New York: Macmillan.

Skinner, B. F. (1957). Verbal behavior. New York: Appleton-Century-Crofts.

Footnotes

1
If using Microsoft ® Office Excel® 2007, in order to ensure that this document can be opened with older
versions of Microsoft ® Office Excel®, click the arrow to the right of Save as type and choose Excel 97-2003
document

2
Within this formula command “=INDEX(Sheet1!A2:Sheet1!A6,RANDBETWEEN(
1,COUNTA(Sheet1!A2:A6)),1)”, the items in bold should be modified to meet the needs outside of this task
analysis. “Sheet1!” references the worksheet where the entries to be randomized are located. By default, the first
worksheet where you entered information is named Sheet1. If the name of the worksheet has not been changed
before entering data, replace”Sheet1”in these formulas with the name of the sheet. If the name of the worksheet is
changed after the fact, the formula will be updated automatically. If the entries are in the same worksheet, this
WORKSHEETNAME! is not necessary. A2:A6 defines the range. A2 references the column (A) and row (2) of the
beginning of the range to be randomly chosen. A6 references the column (A) and row (6) of the end of the range. If
the range is shorter or longer than the 5 used in our example, update this as necessary (if there are only three
observes, the range could be A2:A4). This range begins at A2 to avoid randomizing the column header.

Author Contact Information:

Richard L. Azulay, M.Ed.


Melmark New England
461 River Road
Andover, MA 01810
Phone: 978.654.4309
E-mail: razulay@melmarkne.org

356
Copyright of Behavior Analyst Today is the property of Joseph D. Cautilli and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.

You might also like