JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2018, 51, 283–302 NUMBER 2 (SPRING)

MANIPULATING PARAMETERS OF REINFORCEMENT TO REDUCE

PROBLEM BEHAVIOR WITHOUT EXTINCTION
S. SHANUN KUNNAVATANA
UTAH STATE UNIVERSITY

SARAH E. BLOOM, ANDREW L. SAMAHA

UNIVERSITY OF SOUTH FLORIDA

TIMOTHY A. SLOCUM
UTAH STATE UNIVERSITY

AND

CASEY J. CLAY
UNIVERSITY OF MISSOURI

Differential reinforcement of alternative behavior (DRA) most often includes extinction as a

treatment component. However, extinction is not always feasible and it can be counter-
therapeutic if implemented without optimal treatment integrity. Researchers have successfully
implemented DRA without extinction by manipulating various parameters of reinforcement
such that alternative behavior is favored. We extended previous research by assessing three par-
ticipants’ sensitivities to quality, magnitude, and immediacy using arbitrary responses and rein-
forcers that maintain problem behavior. The results were used to implement an intervention for
problem behavior using DRA without extinction. Our findings indicate that arbitrary responses
can be used to identify individual and relative sensitivity to parameters of reinforcement for rein-
forcers that maintain problem behavior. Treatment was effective for all participants when we
manipulated parameters of reinforcement to which they were most sensitive, and, for two partic-
ipants, the treatment was less effective when we manipulated parameters to which they were
least sensitive.
Key words: DRA, function-based interventions, parametric analysis, problem behavior

Extensive research has established differential

S. Shanun Kunnavatana, Department of Special Educa- reinforcement of an alternative response (DRA)
tion and Rehabilitation, Utah State University, Sarah as an effective intervention for problem behav-
E. Bloom, Child and Family Studies Department, Univer- ior, particularly when problem behavior is
sity of South Florida, Andrew L. Samaha, Child and Fam-
ily Studies Department, University of South Florida, placed on extinction (e.g., Kelley, Lerman, &
Timothy A. Slocum, Department of Special Education Van Camp, 2002; Shirley, Iwata, Kahng,
and Rehabilitation, Utah State University, and Casey Mazaleski, & Lerman, 1997). Perfect treatment
J. Clay, Department of Health Psychology, University of
Missouri. integrity for DRA with extinction requires ther-
This study was conducted in partial fulfillment of the apists to deliver reinforcers every time an indi-
first author’s requirements for the doctoral degree at Utah vidual engages in an alternative response and
State University. We thank Thomas S. Higbee and Cour-
tenay Barrett for their thoughtful comments on an earlier refrain from delivering reinforcers following
version of this manuscript. any instance of problem behavior. However,
Correspondence concerning this article should be data from at least two studies suggest that
addressed to Sarah E. Bloom, 4202 E. Fowler Avenue,
Tampa, FL 33620. Email: [email protected] errors of omission (failing to deliver reinforcers
doi: 10.1002/jaba.443 consistently following alternative behavior) have
© 2018 Society for the Experimental Analysis of Behavior
283
284 S. SHANUN KUNNAVATANA et al.

relatively small effects on treatment outcome magnitude of reinforcement. Horner and Day
(St. Peter Pipkin, Vollmer, & Sloman, 2010; (1991) evaluated the effects of immediacy of
Vollmer, Borrero, Lalli, & Daniel, 1999). In reinforcement on response allocation between
comparison, errors of commission (delivering problem behavior and alternative behavior
reinforcers following problem behavior) more when all other parameters were held constant.
greatly reduce the effectiveness of DRA, partic- One participant, a woman diagnosed with
ularly when the rate of reinforcement no longer autism and severe mental retardation, engaged
favors the alternative behavior. in severe self-injury and aggression maintained
Resources may not always be available to by negative reinforcement in the form of
provide the necessary support to ensure DRA escape. When immediacy favored problem
with extinction is implemented with optimal behavior over alternative behavior, the partici-
treatment integrity, and there are some cases in pant continued to engage in a higher rate of
which extinction is not feasible. For example, problem behavior; however, her response allo-
when problem behavior is dangerous and rein- cation shifted to the alternative behavior when
forced by attention, it may be unreasonable immediacy was manipulated to favor that
(and unsafe) to prescribe an extinction proce- behavior. Neef, Mace, Shea, and Shade (1992)
dure. Additionally, if extinction is not possible, reported different patterns of responding for
and problem behavior occurs at a high rate, it three participants when both reinforcement rate
may be difficult to manipulate the rate of rein- and quality were manipulated. Although evalu-
forcement so that it favors alternative behavior. ated with academic tasks and not problem
Thus, alternative approaches to DRA without behavior, quality overrode rate of reinforcement
extinction should be considered to improve the for two individuals; that is, they emitted more
feasibility of behavior analytic treatments. of the response that produced higher quality
One potential approach is to conceptualize reinforcement at a lower rate than the response
DRA as a concurrent-operants arrangement resulting in lower quality reinforcement at a
(e.g., Athens & Vollmer, 2010), in which two higher rate.
independent schedules of reinforcement are in Numerous studies have found that sensitivity
effect simultaneously for two different to parameters of reinforcement varies across
responses (Ferster & Skinner, 1957). In tradi- individuals (Neef & Lutz, 2001a, 2001b; Neef,
tional DRA procedures, the alternative response Mace, & Shade, 1993; Neef, Shade, & Miller,
is initially reinforced on a continuous schedule 1994; Perrin & Neef, 2012), and therefore
and problem behavior is on extinction. How- similar manipulations are unlikely to be simi-
ever, this is only one of many different concur- larly effective for different individuals. Collec-
rent schedule arrangements. Other concurrent tively, these results suggest that it might be
schedules of reinforcement can be arranged in useful for parameter sensitivity to be assessed at
which problem behavior does result in rein- an individual level. If sensitivity to various
forcement but at a lower rate than that of alter- parameters of reinforcement could be assessed
native behavior, so that alternative behavior is in advance, interventions could be prescribed
differentially favored over problem behavior. that capitalize on individual differences in
Parameters of reinforcement other than rate, parameter sensitivity rather than relying on
such as magnitude, immediacy, and quality, extinction being carried out with perfect treat-
can also produce shifts in response allocation. ment integrity.
For example, Borrero, Vollmer, Borrero, and Neef et al. (1994) evaluated parameter sensi-
Bourret (2005) found that one participant’s tivity for six individuals using arbitrary
behavior appeared sensitive to manipulations of responses (i.e., math problems) rather than
 MANIPULATING PARAMETERS OF REINFORCEMENT 285

problem behavior. The authors used an auto- both problem behavior and alternative behavior
mated computer-based assessment that pre- were reinforced; however, parameters of rein-
sented pairs of math problems from which forcement either favored problem behavior or
participants could choose. Each option was alternative behavior. When the contingencies
associated with a reinforcement contingency were reversed, a shift in responding was
that could be manipulated across different observed such that more responding was allo-
parameters; that is, if students answered the cated to the response that produced the highest
problem on the left correctly, they contacted quality, largest magnitude, or most immediate
one contingency, and if students answered the consequence.
problem on the right correctly, they contacted Athens and Vollmer (2010) proposed an
a different contingency. The goal of the assess- assessment procedure that might reasonably be
ment was to identify relative sensitivity across applied to problem behavior. However, the
rate, quality, immediacy, and response effort by procedures employed by Athens and Vollmer
pitting each parameter against every other require the participants to engage in problem
parameter. For example, to compare sensitivity behavior, which could be problematic depend-
to quality, response allocation was recorded ing on the severity of the behavior. A blended
when high- and low-quality reinforcers were approach that combines the advantages of Ath-
combined with other parameters: high and low ens and Vollmer’s use of reinforcers maintain-
rate of reinforcement, high and low response ing problem behavior with Neef et al.’s (1994)
effort, and short and long delays to reinforce- use of arbitrary responses might be useful. One
ment. Responding differed across individuals benefit of using arbitrary responses to identify
when dimensions were combined and pitted sensitivities to parameters of reinforcement, as
against one another. Although this procedure in Neef et al. (1994), is that it does not require
identified relative sensitivities to a range of an individual to engage in problem behavior.
parameters, there are limitations to applying it Although not evaluated in the context of prob-
to assess problem behavior. It might not only lem behavior, basic researchers have previously
be difficult to capture the motivating opera- demonstrated that participant responses to
tions to assess the efficacy of reinforcers that hypothetical tests to determine sensitivity to
maintain problem behavior, but also it might immediacy of reinforcement is indicative of
be difficult to deliver reinforcement via an how the individuals will allocate responding in
automated program when problem behavior is real choice situations (Odum, 2011). Thus,
maintained by positive reinforcement, especially other parameter sensitivities identified in analog
in the form of attention. Thus, an alternative assessments may also predict sensitivities in
procedure appears necessary to assess parameter other contingency arrangements; that is, it may
sensitivity in the context of problem behavior. be possible to assess sensitivity to different
Athens and Vollmer (2010) evaluated indi- parameters in the context of arbitrary responses
vidual sensitivities of problem behavior to mag- rather than problem behavior, but then to
nitude, quality, and immediacy of apply the results to a DRA-without-extinction
reinforcement. After identifying the function of procedure.
problem behavior (aggression) for seven chil- Thus, the purpose of this study was to deter-
dren, independent assessments were conducted mine whether arbitrary responses (not problem
to determine whether individuals’ behavior was behavior) could be used to identify individual
sensitive to quality, magnitude, and immediacy and relative sensitivities to quality, magnitude,
manipulations. The assessments consisted of a and immediacy for individuals who engage in
concurrent-schedule arrangement in which problem behavior maintained by social-positive
286 S. SHANUN KUNNAVATANA et al.

reinforcement. Additionally, we evaluated 3

RUFUS
whether a DRA procedure in which problem
behavior was not placed on extinction was Esc
2
effective when the alternative behavior was dif-
ferentially favored using the parameter of rein-
forcement to which the participant was most 1 Tang
Attn
Play
and least sensitive.
0

GENERAL METHODS
2.0

PROBLEM BEHAVIOR
SABRINA
Participants and Setting

(PER MINUTE)
Three individuals participated in this study: 1.5
Rufus, Sabrina, and Max. Rufus was a 31-year-
old male diagnosed with cerebral palsy and a 1.0

visual impairment; however, he could differen-

0.5
tiate between colors, shapes, and large pictures
and objects. He was referred for aggression,
0.0
property destruction, and inappropriate vocali-
zations. Sabrina was a 24-year-old female diag-
nosed with mood disorder, autism spectrum 1.0
MAX
disorder, post-traumatic stress disorder, psy- 0.8
chotic disorder (not otherwise specified; NOS),
personality disorder (NOS), and paranoid and 0.6

antisocial traits. She was referred for inappro- 0.4

priate vocalizations. Max was a 10-year-old
0.2
male diagnosed with autism spectrum disorder.
He was referred for inappropriate vocalizations, 0.0
disruption, and aggression. All three partici- 2 4 6 8 10 12 14 16
pants’ problem behavior was determined SESSIONS
through functional analysis (Iwata, Dorsey, Sli-
fer, Bauman, & Richman, 1982/1994) to be Figure 1. Functional analysis data for Rufus, Sabrina,
multiply maintained by access to tangibles and and Max.
escape from demands (Figure 1). The tangible
items used in the functional analysis were based
on the results of a paired stimulus preference conducted sessions with Sabrina and Max in a
assessment (Fisher et al., 1992) for Rufus and a university-based clinic equipped with a one-
multiple stimulus without replacement prefer- way mirror. Both settings included a table and
ence assessment (DeLeon & Iwata, 1996) for chairs for the participants and therapists and
Sabrina and Max. Items included in the prefer- relevant materials for the condition being con-
ence assessments included familiar items com- ducted (e.g., highly preferred items, less pre-
monly used by the participants per caregiver ferred items, etc.). We used colored cards and
report. colored touch-lights as response manipulanda
We conducted sessions with Rufus in an in the individual parameter sensitivity assess-
empty room at a university-based day program ments and relative parameter sensitivity assess-
that he attended during the week. We ments (additional descriptions provided below).
 MANIPULATING PARAMETERS OF REINFORCEMENT 287

Response Measurement color. We used the same colored cards as the

Topographies of problem behavior included buttons for Rufus; however, the touch-light was
inappropriate vocalizations for all three partici- omitted to make his selections easier. Different
pants, aggression for Rufus and Max, property colored buttons and/or cards were used for each
destruction for Rufus, and disruption for Max. assessment. A highly preferred item (iPad for all
Inappropriate vocalizations were defined as yell- participants) was used as the high-quality item
ing and/or directing profanity toward another during the parameter sensitivity assessments.
person. For Sabrina, inappropriate vocalizations Rufus watched a variety of children’s television
also included name calling and making false shows on the iPad, Sabrina played various games
accusations about staff or therapists and watched movies, and Max played two spe-
(e.g., “You’re torturing me,” “You’re abusing cific games. Less preferred items (Sesame Street
me”). For Max, inappropriate vocalizations also cards for Rufus, a magazine for Sabrina, and an
included grunting vocal protest (e.g., “No you’re I Spy book for Max) were used as low-quality
not,” “I don’t have to,” “I’m not going to”), and items during the parameter sensitivity
threatening to leave the area or call the police or assessments.
another authority. Aggression was defined as hit-
ting, kicking, or grabbing. For Rufus, aggression
also included pushing, pinching, biting, and Token Economy
spitting at others. Property destruction was During the individual and relative parameter
defined for Rufus as throwing items (not in the sensitivity assessments, we observed both Rufus
direction of a person), hitting/swiping items off and Sabrina engage in problem behavior
a surface, or hitting any surface with his hand or between sessions and during exposure trials.
foot. Finally, disruption was defined for Max as Because both participants had multiply con-
stomping one or both of his feet. We collected trolled problem behavior (i.e., tangible and
data on the occurrence of problem behavior escape functions), it is plausible that attending
throughout the study. sessions (Sabrina) and being instructed to “pick
Additionally, in Experiment 1 and 2, we col- one” during sessions (Rufus) was evoking prob-
lected data on response allocation between vari- lem behavior maintained by escape. We there-
ous contingencies. A response was counted fore introduced a token economy in which
when the participant pressed one of the touch- Sabrina received tokens for entering and
lights such that it illuminated or touched one remaining in the room where sessions were
of the colored cards (Rufus only). Data collec- held and Rufus received a token for each
tors scored the response as the parameter value instance of compliance with the instruction to
selected (e.g., if the card associated with low make a selection (regardless of the selection
magnitude reinforcement was touched, “low made). We selected this intervention because
magnitude” was scored). we thought it could easily be implemented and
would not interfere with response allocation
during the parameter sensitivity assessments. It
Materials also allowed us to delay when back-up rein-
Each participant was presented with a set of forcement was delivered for compliance (Carr,
buttons that served as the response manipulanda Frazier, & Roland, 2005), such that we did not
during the parameter sensitivity assessments. adventitiously reinforce participants’ response
For Sabrina and Max, the buttons were 6.4-cm allocation during the parameter sensitivity
diameter colored touch-lights, which we placed assessments. A token economy was not neces-
on top of 8.9-cm cards that corresponded in sary for Max.
288 S. SHANUN KUNNAVATANA et al.

We introduced the token economy with multiple days), the primary therapist took her to
Rufus during the immediacy sensitivity assess- the campus food court where Sabrina could
ment (Experiment 1). Rufus did not have pre- choose to purchase one item that cost less than
vious experience with token economies; thus, three dollars. Sabrina consistently selected fruit
brief training was provided. Rufus earned stars smoothies as her reinforcer. Problem behavior
contingent on making a selection when was not observed presession after the token econ-
instructed to do so. The therapist increased the omy was implemented. The therapist implemen-
exchange requirement across sessions until he ted the token economy for the remainder of the
earned five stars before receiving the backup relative parameter sensitivity assessment (imme-
reinforcer (i.e., an edible reinforcer—ranch fla- diacy vs. quality and magnitude vs. quality).
vored chip—that was identified as highly pre-
ferred during a paired-stimulus preference
assessment). It took six training sessions to EXPERIMENT 1: INDIVIDUAL
establish the token economy. Problem behavior PARAMETER SENSITIVITY ASSESSMENT
only occurred during the initial training session The purpose of the individual parameter sensi-
and was placed on extinction. The therapist tivity assessment was to assess participants’ sensi-
implemented the token economy for the tivity to quality, magnitude, and immediacy
remainder of the immediacy sensitivity assess- using arbitrary responses and reinforcers that
ment and the tracking test. maintain problem behavior. We used arbitrary
We introduced the token economy with Sab- responses to decrease the likelihood of the partici-
rina during the relative parameter sensitivity pants engaging in problem behavior during the
assessment (Experiment 2) when immediacy was assessment. Although all three participants
tested against quality. Sabrina had prior experi- engaged in problem behavior maintained by both
ence with token economies; thus, rather than social positive reinforcement in the form of access
providing training, the therapist described the to tangibles and social negative reinforcement in
contingency at the beginning of each appoint- the form of escape, the individual parameter sen-
ment. Sabrina earned a happy face for each hour sitivity assessment exclusively focused on the
she remained in the therapy room, regardless of social positive function. Table 1 contains a sum-
the selections she made during the assessments. mary of the values used for each of the individual
When she had five happy faces (earned across parameter sensitivity assessments.

Table 1
Summary of Individual Parameter Sensitivity Assessment Values

Manipulated
Parameter Definition Consequence 1 Consequence 2 Constant Parameters
Quality Preference for stimulus High Quality Low Quality Magnitude
iPad Rufus: cards 30-s access
Sabrina: magazine Immediacy
Max: book 0-s delay
Magnitude Duration of access High Magnitude Low Magnitude Quality
90-s access 15-s access High
Immediacy
0-s delay
Immediacy Delay between behavior Immediate Delayed Quality
and reinforcer delivery 0-s delay Rufus: 10-s High
Sabrina: 280-s Magnitude
Max: 136-s 30-s access
 MANIPULATING PARAMETERS OF REINFORCEMENT 289

Method Procedure
Design. The individual parameter sensitivity Throughout the experiment, we collected
assessments consisted of giving the participants data on the occurrence of problem behavior
an opportunity to choose between two concur- (rate) and response allocation.
rently available buttons that were associated Exposure trials. Prior to each phase of the
with different parameters of a particular conse- individual parameter sensitivity assessments, we
quence. Each selection resulted in the participant conducted six exposure trials (three per switch).
experiencing the relevant consequence associated Each exposure trial consisted of presenting the
with that switch. We used an ABAB design in relevant antecedent for problem behavior by
which the contingency associated with each but- restricting highly preferred items and prompt-
ton was reversed across phases to test whether the ing the participant to touch one of the buttons.
participant tracked the preferred contingency. The therapist then delivered the relevant conse-
For example, if the purple button was associated quence associated with that button (e.g., 15-s
with a high-quality reinforcer and the green but- vs. 90-s access to tangibles). Sessions within a
ton was associated with a low-quality reinforcer phase sometimes took place across multiple
in the first phase, during the subsequent phase, days; therefore, therapists conducted two expo-
the green button was associated with the high- sure trials (one for each switch) prior to the first
quality reinforcer and the purple button was asso- session of the day to increase the likelihood of
ciated with the low-quality reinforcer. behavior in session coming under the control
Each session consisted of 10 trials in which a of the arranged contingencies.
therapist presented the participant with the two Quality sensitivity assessment. The purpose of
buttons and instructed them to “pick one.” Dur- the quality sensitivity assessment was to evalu-
ing the trials, all other materials were kept off the ate sensitivity to quality of reinforcement.
table (i.e., reinforcers were not present) until the Thus, we made a high-quality reinforcer
participant made a selection and the therapists (i.e., highly preferred item) and a low-quality
positioned their timers facing away from the par- reinforcer (i.e., less preferred item) available.
ticipants so that there were no visible cues associ- The high-quality stimulus was defined as a
ated with the different contingencies. stimulus selected in more than 80% of trials in
Contingent on a selection, the therapist delivered the tangible preference assessment (Koehler,
the corresponding consequence. Throughout all Iwata, Roscoe, Rolider, & O’Steen, 2005). The
individual parameter sensitivity assessments all low-quality stimulus was defined as a stimulus
instances of problem behavior were ignored and that was selected in 10% to 30% of trials dur-
the therapist honored all bids for attention. ing the tangible preference assessment (Koehler
A second trained observer collected data for et al., 2005). We hypothesized that stimuli that
59% of the individual parameter sensitivity fell within this range would still function as
assessment sessions. An agreement was defined reinforcers despite being identified as less pre-
as both data collectors scoring the same selected ferred (Roscoe, Iwata, & Kahng, 1999).
parameter value (e.g., low magnitude) during a During the quality sensitivity assessment,
given choice opportunity. Reliability was calcu- either the high-quality item or the low-quality
lated for each session by taking the number of item was delivered contingent on the button
agreements and dividing by the total number of selected. We kept magnitude and immediacy
agreements plus disagreements and multiplying constant; that is, therapists delivered the
by 100 to yield a percentage. Mean reliability selected item immediately and for 30 s, regard-
across sessions was 99% (range, 95%-100%). less of the item was selected.
290 S. SHANUN KUNNAVATANA et al.

Magnitude sensitivity assessment. The purpose for Max. We doubled the delay for Rufus to
of the magnitude sensitivity assessment was to 20 s (four times the median IRT) after we did
evaluate sensitivity to reinforcer magnitude not observe sensitivity to immediacy.
(i.e., duration of access to reinforcer). We set During the immediacy sensitivity assessment,
the magnitude values based on values used in therapists either delivered reinforcement imme-
Athens and Vollmer (2010). Athens and Voll- diately or after a delay, contingent on the but-
mer used a 1:6 ratio in setting magnitude ton selected. We kept quality and magnitude
values; thus, low magnitude was 15-s access constant; that is, both selections produced the
and high magnitude was 90-s access. For Rufus highly preferred item for 30 s.
and Max, we doubled the high magnitude to Tracking test. Due to Rufus’ insensitivity to
180 s (a 1:12 ratio) after not observing sensitiv- magnitude and immediacy, we conducted a
ity to the high magnitude at 90 s. tracking test to assess side and color bias that
During the magnitude sensitivity assessment, might have masked the effects of manipulating
therapists delivered either the high-magnitude these parameters. Using the same colors used in
or low-magnitude consequence contingent on the immediacy sensitivity assessment, we tested
the button selected. Subsequent trials were con- immediacy against quality (a parameter to
ducted immediately following the termination which Rufus showed sensitivity). One button
of the low or high magnitude reinforcement was associated with the low-quality item deliv-
interval. We kept quality and immediacy con- ered immediately and the other button was
stant; that is, both selections resulted in the associated with the high-quality item delivered
therapists delivering the high-quality item after a 20-s delay. We used an ABA design to
immediately. demonstrate that he tracked contingencies
Immediacy sensitivity assessment. The purpose when a parameter to which he was sensitive
of the immediacy sensitivity assessment was to was manipulated.
test for sensitivity to immediacy of reinforce-
ment. The immediate value was a 0-s delay.
For the delay condition, we wanted to set a Results
delay that would both affect response allocation Rufus. Rufus’ parameter sensitivity data are
and be anchored to an aspect of behavior main- depicted in Figure 2. In the quality sensitivity
tained by that individual’s reinforcer. We felt assessment, Rufus consistently allocated more
this approach might be more favorable than responding to the card associated with the
selecting an arbitrary but equal delay for each high-quality item, even when the contingencies
participant. Hence, we selected a delay that was were switched across phases. These data suggest
twice as long as the median interresponse time he was sensitive to quality of reinforcement.
(calculated with reinforcement intervals omit- When we assessed sensitivity to magnitude,
ted) in the tangible condition of the functional data were less consistent. Initially, Rufus alter-
analysis. Previous research suggests interre- nated between the high- and low-magnitude
sponse time is sensitive to individual motiva- options. We increased the high-magnitude
tion for reinforcement and is commonly used value to 180 s (session 7), but he continued to
as a basis for determining schedules of rein- allocate more responding to the button associ-
forcement (e.g., Kahng, Iwata, DeLeon, & ated with low magnitude. When we switched
Wallace, 2000). Median values were selected the contingencies across phases, Rufus contin-
over means because the former are less sensitive ued to allocate his responding to the pink but-
to outliers. This resulted in a delay value of ton (positioned on the right) regardless of the
10 s for Rufus, 280 s for Sabrina, and 136 s contingency associated with it, indicating either
 MANIPULATING PARAMETERS OF REINFORCEMENT 291

Figure 2. Individual parameter sensitivity assessments for Rufus (left column), Sabrina (center column), and Max
(right column). PB = Problem behavior; HQ = high quality; LQ = low quality; HM = high magnitude; LM = low
magnitude; I = immediate; D = delay. The asterisk on Rufus’ immediacy sensitivity assessment graph denotes when the
token economy was introduced. The bottom row depicts Rufus’ tracking test.

a lack of sensitivity to magnitude at these values button associated with immediate reinforcement.
or a potential position or color bias. However, when we switched the contingencies,
When we assessed sensitivity to immediacy, he continued to select the red button (positioned
he initially allocated more responding to the on the right). The therapist introduced the token
292 S. SHANUN KUNNAVATANA et al.

economy with Rufus at session 11 when we higher rate of problem behavior (range, 0.2-
observed an increase in problem behavior 0.9) during the last two phases of the assess-
between sessions and during exposure trials. The ment. Nonetheless, her response allocation sug-
token economy did not appear to affect his gests she was also sensitive to immediacy as a
response allocation, as he continued to select the parameter of reinforcement.
button associated with the delayed conse- Max. Max’s data are depicted in Figure 2.
quence more frequently. We increased the When we assessed sensitivity to quality, he con-
delay to 20 s (session 14) to see if he would sistently allocated more responding to the but-
be sensitive to a longer delay; however, he con- ton that produced the high-quality reinforcer.
tinued to select the red button (positioned on This was consistent across phases, showing that
the right) regardless of the consequence associ- he tracked contingencies, and suggesting he is
ated with it. These results suggest that he was sensitive to quality of reinforcement. For magni-
not sensitive to immediacy of reinforcement at tude of reinforcement manipulations, Max ini-
the delays we tested or that there was a posi- tially allocated more responding to the button
tion or color bias. that was associated with low-magnitude rein-
A tracking test was conducted after we failed forcement. Similar to Rufus, we increased the
to observe sensitivity to two parameters to value of the high-magnitude option from 90 s
examine whether extraexperimental features to 180 s (session 4). At these values, we saw sen-
(e.g., side bias, color bias) were overriding the sitivity to magnitude of reinforcement, as he
effects of the parameter manipulations. Rufus consistently allocated his responding to the but-
consistently allocated more responding to the ton that produced high-magnitude reinforce-
button associated with the high-quality delayed ment. Finally, when we evaluated sensitivity to
consequence, regardless of the button color or immediacy of reinforcement, Max consistently
side on which it was presented. This indicates allocated more responding to the button that
that Rufus tracked consequences when parame- produced immediate reinforcement. Thus, he
ters to which he was sensitive were manipu- was found to be sensitive to all three parameters.
lated. Thus, we concluded he was sensitive to
only quality of reinforcement at the parameter
values we manipulated. Discussion
Sabrina. The results of Sabrina’s quality sen- The results of Experiment 1 indicate that one
sitivity assessment are depicted in Figure 2. participant, Sabrina, was sensitive to all three
Sabrina consistently selected the button associ- parameter manipulations at their initial values.
ated with the high-quality item more than the Max was also found to be sensitive to all three
button associated with the low-quality item. parameters; however, his sensitivity to magni-
Additionally, she tracked the contingencies tude was detected only when we increased the
across phases, suggesting that Sabrina was sensi- values such that the high-magnitude value was
tive to quality as a parameter of reinforcement. 12 times that of the low-magnitude value. Rufus
Similar patterns of responding were observed was found to be sensitive to only quality. When
during the magnitude sensitivity assessment, we assessed immediacy and magnitude, he con-
suggesting she was also sensitive to magnitude tinued to allocate responding to whichever con-
of reinforcement. During the immediacy sensi- tingency was associated with the button on his
tivity assessment, Sabrina allocated more right. One potential explanation for this
responding to the option that produced imme- response pattern is that there was a position bias
diate reinforcement. She tracked the contingen- that was potentially a function of differential
cies across phases; however, we observed a response effort; Rufus is right-hand dominant
 MANIPULATING PARAMETERS OF REINFORCEMENT 293

Table 2
Summary of Relative Parameter Sensitivity Assessment Values

Manipulated Parameters Consequence 1 Consequence 2 Constant Parameter

Magnitude vs. Immediacy Low Magnitude-Immediate High Magnitude-Delayed Quality High
15-s access 0-s delay Sabrina
90-s access, 280-s delay
Max
180-s access, 136-s delay
Immediacy vs. Quality Low Quality-Immediate High Quality-Delayed Magnitude 30-s access
Low quality 0-s delay Sabrina
High quality, 280-s delay
Max
High quality, 136-s delay
Magnitude vs. Quality Low Magnitude-High Quality High Mag-Low Quality Immediacy 0-s delay
15-s access High quality Sabrina
90-s access, low quality
Max
180-s access, low quality

and therefore it is possible that when we manip- assessments, with the exception that buttons
ulated parameters to which he was not sensitive, were associated with a combination of parame-
selecting the button on his right may have been ter manipulations (e.g., high-quality, delayed
slightly less effortful. reinforcement vs. low-quality, immediate rein-
forcement). Materials and values were identical
to those used in Experiment 1. Table 2 con-
EXPERIMENT 2: RELATIVE PARAMETER tains a summary of the values used for the rela-
SENSITIVITY ASSESSMENT tive parameter sensitivity assessment.
The purpose of the relative parameter sensi- A second trained observer collected data for
tivity assessment was to obtain a hierarchy of 26% of the relative parameter sensitivity assess-
parameters to which participants were sensitive, ment sessions, and interobserver agreement was
for participants who demonstrated sensitivity to calculated using the same procedures as in the
more than one parameter during Experiment parameter sensitivity assessment. Mean reliabil-
1. Sabrina and Max participated in Experiment ity across sessions was 99.7% (range,
2 because the results of their individual param- 99%-100%).
eter sensitivity assessments indicated they were Magnitude versus immediacy. The purpose of
sensitive to multiple parameters. We used the this assessment was to determine whether Sab-
same materials as those used in the individual rina and Max were more sensitive to magnitude
parameter sensitivity assessments and presented or immediacy of reinforcement when specific
them in the same concurrent arrangement. values were tested. The two consequences we
evaluated in this assessment were low-
Method magnitude reinforcement delivered immediately
Design and procedures. We conducted ses- and high-magnitude reinforcement delivered
sions using an ABAB design as in the individ- after a delay. We kept quality consistent across
ual parameter sensitivity assessments, reversing both options; that is, both choices resulted in
the consequences associated with the buttons Sabrina and Max receiving the iPad.
across phases. We conducted sessions identical Immediacy versus quality. The purpose of this
to the individual parameter sensitivity assessment was to determine whether Sabrina
294 S. SHANUN KUNNAVATANA et al.

MAGNITUDE/IMMEDIACY SENSITIVITY ASSESSMENT

A B A B A B A B
100 1.0 100 1.0

80 0.8 80 0.8
LM-I
60 0.6 60 0.6

40 0.4 40 0.4
SABRINA MAX
20 HM-D PB 0.2 20 0.2

0 0.0 0 0.0
2 4 6 8 10 12 2 4 6 8 10 12 14 16 18 20 22 24

IMMEDIACY/QUALITY SENSITIVITY ASSESSMENT

A B A B A B A B
*
% OF TRIALS SELECTED

100 1.0 100 1.0

PROBLEM BEHAVIOR
80 0.8 80 0.8

(PER MINUTE)
60 D-HQ I-LQ 0.6 60 0.6

40 0.4 40 0.4

20 0.2 20 0.2
PB
0 0.0 0 0.0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12

MAGNITUDE/QUALITY SENSITIVITY ASSESSMENT

A B A B A B A B
100 1.0 100 0.8

80 0.8 80
LM-HQ 0.6

60 0.6 60
0.4
40 HM-LQ 0.4 40
0.2
20 0.2 20
PB
0 0.0 0 0.0
2 4 6 8 10 12 2 4 6 8 10 12

SESSIONS

Figure 3. Relative parameter sensitivity assessments for Sabrina (left column) and Max (right column). PB = Prob-
lem behavior; HM = high magnitude; LM = low magnitude; I = immediate; D = delay; HQ = high quality; LQ = low
quality. The asterisk on Sabrina’s immediacy versus quality sensitivity assessment graph denotes when the token econ-
omy was introduced.

and Max were more sensitive to immediacy or low-quality reinforcement delivered immedi-
quality of reinforcement. The two conse- ately and high-quality reinforcement delivered
quences we evaluated in this assessment were after a delay. We kept magnitude consistent
 MANIPULATING PARAMETERS OF REINFORCEMENT 295

across both options; that is, both choices reinforcement, with the exception of session
resulted in Sabrina and Max having access to 19. These data indicate that she was more sen-
the reinforcer for 30 s. sitive to quality than immediacy.
Magnitude versus quality. The purpose of this Finally, we compared magnitude and quality.
assessment was to determine whether Sabrina and Sabrina consistently allocated more responding
Max were more sensitive to magnitude or quality. to the button associated with the low-magni-
In this assessment, we compared high magnitude tude, high-quality consequence, suggesting she
of a low-quality reinforcer to low magnitude of a was more sensitive to quality than magnitude.
high-quality reinforcer. We kept immediacy con- Together, these results suggest Sabrina was
sistent across both options; that is, the therapist most sensitive to quality and least sensitive to
delivered both consequences immediately after magnitude.
Sabrina or Max made a selection. Low rates of problem behavior were observed
during three sessions (range, 0.04 to 0.2) when
sensitivity to immediacy and quality of rein-
Results forcement was assessed.
Sabrina. The results of Sabrina’s relative Max. The results for Max are depicted in
parameter sensitivity assessment are in Figure 3. When we compared magnitude and
Figure 3. When we compared magnitude and immediacy, Max did not consistently allocate
immediacy, Sabrina responded more frequently his responding to either switch. Across condi-
on the button associated with low-magnitude, tions, his response allocation sometimes favored
immediate reinforcement, suggesting her magnitude and sometimes favored immediacy.
behavior was more sensitive to immediacy than These results suggest that neither parameter
magnitude. Problem behavior was not observed manipulation was strong enough to alter
during this parameter sensitivity assessment. response allocation.
When we compared immediacy and quality, To rule out color bias, we used the same colors
Sabrina initially allocated more responding to when assessing quality and immediacy. We
the button associated with high-quality rein- observed clear differentiation in his response
forcement delivered after a delay. However, allocation—he consistently selected the button
when we changed phases, she started selecting associated with the high-quality delayed rein-
the immediate low-quality option more fre- forcer, regardless of position of the button. These
quently. After additional sessions, she returned results suggest Max was more sensitive to quality
to allocating more responding to the delayed, than immediacy.
high-quality option. The therapist introduced Finally, we compared magnitude and quality.
the token economy at the beginning of session Max consistently allocated more responding to
14, because Sabrina began engaging in more the button associated with the low-magnitude,
problem behavior between sessions and during high-quality consequence, suggesting he was
exposure trials, and she made statements about more sensitive to quality than magnitude.
choosing the immediate option so that she Together, the results indicate Max was most sen-
could go home sooner. After we incorporated sitive to quality and less sensitive to immediacy
the token economy, Sabrina tracked the option and magnitude. Relative sensitivity to immedi-
that produced the delayed, high-quality rein- acy and magnitude could not be determined.
forcement (in the third phase). In the final Problem behavior was not observed during
phase, she again allocated more responding to any of the relative parameter sensitivity assess-
the option that produced delayed, high-quality ment sessions.
296 S. SHANUN KUNNAVATANA et al.

Discussion independently emitted the responses within

Both participants were most sensitive to four training sessions (data available upon
quality of reinforcement. Although there was request). During the training, Sabrina and Max
clearly differentiated sensitivity for Sabrina, did not engage in any instances of problem
Max allocated his responding such that relative behavior; Rufus engaged in one occurrence,
sensitivity could not be determined between which was ignored. Rufus’ alternative response
magnitude and immediacy. Thus, relative sensi- consisted of handing the primary therapist a
tivity was not the same across participants, and communication card (8.9-cm and lined in
the participants were not equally sensitive to all bright yellow) that said, “Share with me.” Sab-
parameters of reinforcement. These results pro- rina and Max’s alternative response was to say,
vide additional support for identifying individ- “Can I have a turn, please?” Approximations of
ual relative sensitivities to parameters of this response (e.g., leaving off the word
reinforcement. “please”) were not reinforced.
Design and procedures. In Experiment 1 for
Rufus and Experiment 2 for Sabrina and Max,
EXPERIMENT 3: DRA WITHOUT
all three participants were most sensitive to
EXTINCTION
quality of reinforcement. Sabrina was least sen-
In Experiments 1 and 2, we assessed sensitiv- sitive to magnitude of reinforcement. Rufus
ity to parameters of reinforcement using arbi- was insensitive to both magnitude and immedi-
trary responses. Experiment 3 was designed to acy of reinforcement; however, there was less
determine whether the results of the sensitivity variability in his selections during the magni-
assessments could be used to inform an effective tude parameter sensitivity assessment, suggest-
intervention for problem behavior. The purpose ing he may have been slightly less sensitive to
of Experiment 3 was to determine whether a magnitude. Therefore, we manipulated magni-
DRA-without-extinction procedure that uses tude as the parameter to which Rufus was least
the parameter to which the participant is most sensitive. Max was equally sensitive to magni-
sensitive is more effective than a DRA without tude and immediacy; however, during the ini-
extinction procedure that uses a parameter to tial magnitude sensitivity assessment, when we
which the participant is less sensitive. tested 90-s access as the high-magnitude value,
we did not see sensitivity. Thus, we also manip-
Methods ulated magnitude for Max as the parameter to
Materials and response definitions. We used which he was least sensitive.
the high-quality and low-quality tangible items We used a multiple baseline design across
during Experiment 3. Prior to starting this the three participants to determine the effec-
phase of the study, we hypothesized that Rufus tiveness of magnitude and quality manipula-
was habituating to the iPad because he would tions for treating problem behavior. In all
often try to engage in conversation with thera- phases, we programmed reinforcement for both
pists while he had access to the iPad rather than problem behavior and alternative behavior;
engage with the iPad. As such, we included however, the parameters of reinforcement var-
additional preferred items (i.e., nesting blocks ied across phases. Sessions were 10 min.
and a Sight-and-See puzzle) for Rufus. A second trained observer collected data for
We selected simple alternative responses for 69% of the treatment evaluation sessions. An
the participants. Prior to baseline, mand train- agreement was defined as both data collectors
ing was conducted, and all participants scoring the occurrence of the alternative
acquired the alternative responses and response or problem behavior in a 10-s interval.
 MANIPULATING PARAMETERS OF REINFORCEMENT 297

Reliability was calculated for each session by BASELINE MAGNITUDE QUALITY

taking the number of agreements and dividing 4 90 s-HQ (PB) vs.
90 s-HQ (Alt)
15 s-LQ (PB) vs.
90 s-LQ (Alt)
30 s-LQ (PB) vs.
30 s-HQ (Alt)

by the total number of agreements plus dis-

3 PB
agreements and multiplying by 100 to yield a
percentage. Mean reliability across sessions was 2
RUFUS
99% (range, 95%-100%). 1
Alt
Baseline. We used baseline procedures similar
0
to those used in Athens and Vollmer (2010) and

RESPONSES PER MINUTE

the establishing operation for problem behavior 2.0

was presented by restricting access to high pre- 1.5

ferred items. Problem behavior and alternative
1.0
behavior both resulted in high-quality, high-
magnitude reinforcement; that is, the pro- 0.5 SABRINA

grammed consequence for problem behavior and

0.0
alternative behavior was 90-s access to the iPad.
2.0
Magnitude manipulation. During the magni-
tude manipulation, reinforcement favored the 1.5

alternative behavior; that is, therapists delivered 1.0

high-magnitude reinforcement (i.e., 90 s) con-
0.5
tingent on alternative behavior and low- MAX

magnitude reinforcement (i.e., 15 s) contingent 0.0

on problem behavior. We kept immediacy of 2 4 6 8 10 12 14

reinforcement constant, such that therapists SESSIONS

always delivered reinforcement immediately.

We also kept quality of reinforcement consis- Figure 4. Treatment evaluation. PB = problem
tent for both problem behavior and alternative behavior; Alt = alternative behavior; HQ = high quality;
behavior; however, we used low-quality items LQ = low quality.
to assess the effectiveness of the magnitude
manipulation to ensure that high-quality rein- Results
forcers (a parameter to which both participants The results of the treatment evaluation are
were sensitive) would not overshadow the effect depicted in Figure 4. During baseline, when
of magnitude. No prompts were provided to problem behavior and alternative behavior both
engage in the alternative response. resulted in the same consequences, we observed
Quality manipulation. The quality manipula- increasing trends in problem behavior for all
tion consisted of favoring alternative behavior participants and no alternative behavior. When
using differential reinforcer quality. The pro- we manipulated magnitude to favor alternative
grammed consequence for alternative behavior behavior, we initially observed instances of
was access to a high-quality reinforcer, and the alternative behavior for all three participants,
programmed consequence for problem behavior but it decreased to zero for Rufus and Sabrina.
was access to a low-quality reinforcer. We kept The alternative behavior maintained for Max.
magnitude and immediacy consistent; that is, We did not observe problem behavior with
therapists delivered reinforcers immediately, Rufus or Max, but we did observe a few
and participants had access for 30 s. No instances with Sabrina. In the last phase, when
prompts were provided to engage in the alter- we manipulated quality, we did not observe
native response. any problem behavior for any of the
298 S. SHANUN KUNNAVATANA et al.

participants but saw an increase in alternative participant was most sensitive and led to the
behavior for all three. most effective treatment.

Discussion GENERAL DISCUSSION

For Sabrina, the magnitude manipulations The results of this study indicate that it may
resulted in a decrease in problem behavior, but be possible to use arbitrary responses to assess
alternative behavior was not maintained. How- individual sensitivity to different parameters of
ever, the magnitude manipulation provided suf- reinforcement that maintain problem behavior.
ficient reinforcement to suppress problem In addition, when an individual is sensitive to
behavior for Rufus and Max and maintain multiple parameters of reinforcement, relative
alternative behavior for Max, despite both parameter sensitivity can also be determined
showing insensitivity to these parameter values using arbitrary responses rather than problem
during Experiment 1. In comparison, the qual- behavior. The results of the intervention for all
ity manipulation suppressed problem behavior three participants validate the results of the
and increased alternative behavior for all three individual parameter sensitivity assessment
participants. Although both the magnitude and (Rufus) and the relative parameter sensitivity
quality manipulations were effective for Max, assessment (Sabrina and Max). Moreover, a
we observed a higher rate of alternative behav- parameter sensitivity assessment using arbitrary
ior during the quality manipulation, suggesting responses may be used to develop effective
that it was more effective to manipulate the treatments for problem behavior in situations
parameter to which he was most sensitive. in which extinction is not feasible. It is impor-
The results of Experiment 3 provide further tant to note, however, that problem behavior
support for identifying individual sensitivities was not reinforced during the individual or rel-
to parameters of reinforcement as not all ative parameter sensitivity assessments, and
parameter manipulations were equally success- therefore, it is still unclear whether this assess-
ful in shifting responding away from problem ment is appropriate if extinction is also not fea-
behavior, despite the individual being sensitive sible during the assessment itself. These results
to the parameter. For example, Sabrina was must be interpreted cautiously; future research
sensitive to magnitude during the individual should include the implementation of these
parameter sensitivity assessment; however, procedures with additional participants, includ-
when we manipulated magnitude during DRA ing those with problem behavior maintained by
without extinction, it did not alter her response attention and/or escape. Although our partici-
allocation, suggesting that magnitude only mat- pants’ problem behavior was also maintained
ters when the reinforcer is of sufficient quality. by escape, these parameter sensitivity assess-
Although both magnitude and quality manipu- ments were not directly applied to that
lations were effective for Max, manipulating function.
the parameter to which he was most sensitive This study extends previous research in sev-
produced a higher rate of alternative behavior, eral ways. First, this study extends the work of
which may indicate more robust treatment Athens and Vollmer (2010) and Neef
effects. Thus, individual sensitivity assessments et al. (1994) by using arbitrary responses and
alone may not be sufficient in identifying reinforcers that maintain problem behavior to
which parameter to manipulate. Relative assess sensitivity to parameters of reinforcement
parameter sensitivity assessments, however, for the purpose of developing a treatment for
identified the parameter to which each problem behavior. We were able to identify
 MANIPULATING PARAMETERS OF REINFORCEMENT 299

parameter sensitivity during an assessment intervention. There may be cases in which the
using arbitrary responses, and the results parameter to which an individual is most sensi-
allowed us to predict which parameter manipu- tive cannot easily be manipulated
lations would be more and less effective when (e.g., delaying reinforcement for problem
applied to a treatment designed to reduce prob- behavior). Future researchers may want to
lem behavior and increase functionally alterna- investigate the extent to which various parame-
tive responses. ters can be manipulated to produce positive
Second, our use of arbitrary responses also treatment effects; that is, can more than one
made it possible to expose participants to con- parameter to which the individual is less sensi-
tingencies prior to conducting the sensitivity tive be combined to outweigh a single parame-
assessments. In using problem behavior to ter to which an individual is more sensitive?
assess sensitivity to parameters of reinforce- It is possible that had we tested additional
ment, it is unclear whether the participants in magnitude and immediacy values, we may have
Athens and Vollmer (2010) were exposed to been able to detect sensitivity for Rufus. For
the contingencies as intended because preexpo- example, basic researchers have assessed sensi-
sure sessions were not conducted. This is par- tivity to immediacy of reinforcement
ticularly important for the immediacy (i.e., impulsivity) by testing a variety of values
parameter sensitivity assessment. Participants and determining an indifference point
may have experienced the delay in reinforce- (e.g., Madden & Johnson, 2010). Procedures
ment delivery as extinction rather than as a such as these could be useful in determining
delayed consequence, potentially resulting in exact values necessary to completely shift
extinction-induced side effects (e.g., high rate response allocation from one option to another;
of responding). Thus, ensuring that participants however, doing so would likely be time con-
are exposed to the relevant contingencies prior suming and therefore less feasible for clinicians
to assessment could more accurately identify and practitioners. Although we tested only two
sensitivity to immediacy. or three values per parameter, it is likely that
Third, we sought to identify one parameter we can make certain extrapolations beyond the
to which each participant was most sensitive so tested values. For example, if a participant is
that an intervention could be developed using sensitive to magnitude when 15-s access and
only one parameter manipulation rather than 90-s access are compared, it is likely that they
multiple parameter manipulations, as other would also be sensitive to larger magnitudes
researchers have done. The results of Rufus’ (e.g., 120-s access) when compared to 15-s
individual parameter sensitivity assessment indi- access. We extended previous research by
cate that some individuals will be insensitive to selecting values for magnitude and immediacy
certain parameters of reinforcement in some that were potentially relevant for the partici-
contexts. Thus, it should not be assumed that pants instead of using arbitrary values. For
all individuals would be equally sensitive to all immediacy values, we extended previous
parameters of reinforcement. We successfully research by selecting values based on individual
identified the most influential parameter for all participant behavior (i.e., IRT). Therefore, we
three participants and subsequently implemen- would expect that the delays tested would have
ted effective treatments. By reducing the num- been sufficient and any sensitivity to immedi-
ber of treatment components to the most acy should have been detected. Similarly, we
important parameter for each participant, it attempted to choose values for the magnitude
may be possible to increase overall treatment assessment that would be detectable and mean-
integrity and long-term effectiveness of the ingful to the participants but also feasible for
300 S. SHANUN KUNNAVATANA et al.

implementation. An alternative could be to procedurally we implemented DRA without

determine what is feasible for the behavior extinction, we cannot be certain that reinfor-
change agents to implement. Additional cing both problem behavior and alternative
research should investigate various methods for behavior but favoring the alternative behavior
efficiently selecting values to be evaluated in would be effective. However, it is important to
the parameter sensitivity assessments. note that manipulating the parameter to which
One potential limitation of this study is that, the participants were most sensitive produced
although problem behavior was relatively infre- the highest rate of alternative behavior for all
quent for all participants during the parameter three participants. Thus, we can conclude that
sensitivity assessments, all occurrences were manipulating parameters for which participants
ignored while the button touches were rein- are most sensitive may increase alternative
forced. During the intervention phase of the behavior, but we cannot make strong conclu-
study, we reestablished baseline rates of prob- sions about the parameter manipulation effects
lem behavior; however, it is possible that the on problem behavior.
arrangement used during the parameter sensi- All three participants had behavior main-
tivity assessments inadvertently contributed to tained by social positive reinforcement in the
the rapid reduction of problem behavior during form of access to tangibles as well as social neg-
the intervention phase of the study. One ative reinforcement in the form of escape. It
potential explanation is that problem behavior was necessary to address the escape function
contacted extinction during the parameter during the parameter sensitivity assessments for
assessments, thus weakening the behavior. Rufus and Sabrina because participants’
Alternatively, it is possible that we established a attempts to escape from the sessions could have
strong reinforcement history for alternative otherwise interfered with the results. The inclu-
behavior in general. Additional participants are sion of token economies sufficiently addressed
needed to determine whether such rapid reduc- the escape function for both participants, allow-
tions are common after exposure to the param- ing us to obtain results on parameter sensitivity
eter sensitivity assessments. If such reductions for tangible items. The token economy was in
are common, it may be useful to conduct place for all conditions once it was introduced;
future research on whether similar effects are therefore, it cannot directly account for differ-
observed when parameter assessments are con- ential results across conditions. However, the
ducted in a brief format. extent to which the token economy interacted
A second limitation is the low rate of prob- with the experimental contingencies is unclear.
lem behavior observed during the intervention Although we identified the parameters to which
parameter manipulations. The purpose of the both participants were most sensitive and
intervention phase was to evaluate a DRA-with- implemented an effective intervention, it is pos-
out- extinction procedure using the parameters sible that, without treating both functions, the
to which each participant was most and least parameter manipulations would not be as effec-
sensitive. However, Rufus and Max did not tive. That is, practitioners are cautioned against
engage in problem behavior during the magni- treating only one function of problem behavior
tude manipulation phase or the quality manip- through parameter manipulations. Additional
ulation phase; Sabrina also did not engage in research is needed on how multiple functions
problem behavior during the quality manipula- for problem behavior might influence sensitiv-
tion phase. Thus, they did not contact the pro- ity to parameters of reinforcement and subse-
grammed contingencies for problem behavior quently influence the effectiveness of an
in those conditions. Therefore, although intervention.
 MANIPULATING PARAMETERS OF REINFORCEMENT 301

Finally, social negative reinforcement was allocated to such activities increases the odds of
excluded from this investigation because a meaningful treatment effect when implemen-
research thus far has not directly manipulated ted by teachers and parents in classrooms and
quality in parameter manipulations for nega- homes.
tively reinforced behavior. For example, Peter-
son, Frieder, Smith, Quigley, and Van Norman
REFERENCES
(2009) manipulated quality of demand by pro-
viding a break from the demand with high Athens, E. S., & Vollmer, T. R. (2010). An investigation
of differential reinforcement of alternative behavior
quality (or high-preference) items. This manip- without extinction. Journal of Applied Behavior
ulation enriches the break with a combined Analysis, 43, 569-589. https://doi.org/10.1901/jaba.
positive and negative reinforcement contin- 2010.43-569
Borrero, C. S. W., Vollmer, T. R., Borrero, J. C., &
gency rather than altering the quality of the Bourret, J. (2005). A method for evaluating parame-
break. Rather, quality of negative reinforcement ters of reinforcement during parent-child interactions.
varies based on the aversiveness of the demand Research in Developmental Disabilities, 26, 577-592.
(e.g., Knighton, Bloom, & Clark, 2014). Addi- https://doi.org/10.1016/j.ridd.2004.11.010
Carr, J. E., Frazier, T. J., & Roland, J. P. (2005). Token
tional research is needed on quality manipula- economy. In A. M. Gross & R. S. Drabman (Eds.),
tions for behavior maintained by social negative Encyclopedia of behavior modification and cognitive
reinforcement so that procedures, such as those behavior therapy – Volume 2: Child clinical applica-
investigated here, can be applied to and evalu- tions (pp. 1075-1079). Thousand Oaks, CA: Sage.
DeLeon, I. G., & Iwata, B. A. (1996). Evaluation of a
ated with problem behavior with various multiple-stimulus presentation format for assessing
functions. reinforcer preferences. Journal of Applied Behavior
The participants who took part in this study Analysis, 29, 519-533. https://doi.org/10.1901/jaba.
were referred for problem behavior with rela- 1996.29-519
Ferster, C. B., & Skinner, B. F. (1957). Schedules of rein-
tively low severity and therefore it may be feasi- forcement. New York, NY: Appleton-Century-Crofts.
ble to place their problem behavior on Fisher, W., Piazza, C. C., Bowman, L. G.,
extinction if those implementing the interven- Hagopian, L. P., Owens, J. C., & Slevin, I. (1992).
tion are able to do so consistently. It is possible A comparison of two approaches for identifying rein-
forcers for persons with severe and profound disabil-
that individuals with more severe topographies ities. Journal of Applied Behavior Analysis, 25, 491-
of problem behavior would engage in more 498. https://doi.org/10.1901/jaba.1992.25-491
problem behavior during the parameter sensi- Horner, R. H., & Day, H. M. (1991). The effects of
tivity assessments. This could be problematic if response efficiency on functionally equivalent com-
peting behaviors. Journal of Applied Behavior Analysis,
the behavior interferes with the assessment. To 24, 719-732. https://doi.org/10.1901/jaba.1991.
combat this risk, the assessments were designed 24-719
to provide a reinforcing outcome for either Iwata, B. A., Dorsey, K. F., Slifer, K. J.,
response, decreasing the motivation to engage Bauman, K. J., & Richman, G. S. (1982/1994).
Toward a functional analysis of self-injury. Journal of
in problem behavior once the selection Applied Behavior Analysis, 27, 197-209. https://doi.
response was acquired. Nonetheless, additional org/10.1901/jaba.1994.27-197
research should be conducted to evaluate this Kahng, S. W., Iwata, B. A., DeLeon, I. G., &
assessment procedure with other populations. Wallace, M. D. (2000). A comparison of procedures
for programming noncontingent reinforcement
The use of parameter manipulations to schedules. Journal of Applied Behavior Analysis, 33,
implement DRA-without-extinction procedures 223-231. https://doi.org/10.1901/jaba.2000.33-223
is still relatively new. Additional research is Kelley, M. E., Lerman, D. C., & Van Camp, C. M.
needed to determine the most efficient and (2002). The effects of competing reinforcement
schedules on the acquisition of functional communi-
effective procedures to determine how best to cation. Journal of Applied Behavior Analysis, 35, 59-
manipulate parameters and whether the time 63. https://doi.org/10.1901/jaba.2002.35-59
302 S. SHANUN KUNNAVATANA et al.

Knighton, R., Bloom, S. E., & Clark, D. (2014). The use Perrin, C. J., & Neef, N. A. (2012). Further analysis of
of progressive-ratio schedules to assess negative reinforcers. variables that affect self-control with aversive events.
Manuscript in preparation. Journal of Applied Behavior Analysis, 45, 299-313.
Koehler, L. J., Iwata, B. A., Roscoe, E. M., https://doi.org/10.1901/jaba.2012.45-299
Rolider, N. U., & O’Steen, L. E. (2005). Effects of Peterson, S. M., Frieder, J. E., Smith, S. S.,
stimulus variation on the reinforcing capability of Quigley, S. P., & Van Norman, R. K. (2009).
nonpreferred stimuli. Journal of Applied Behavior The effects of varying quality and duration of rein-
Analysis, 38, 469-484. https://doi.org/10.1091/jaba. forcement on mands to work, mands for break, and
2005.38-469 problem behavior. Education and Treatment of
Madden, G. J., & Johnson, P. S. (2010). A delay- Children, 32, 605-630. https://doi.org/10.1353/
discounting primer. In G. J. Madden & W. K. Bickel etc.0.0075
(Eds.), Impulsivity: The behavioral and neurological sci-
ence of discounting (pp. 11-37). Washington, DC: Roscoe, E. M., Iwata, B. A., & Kahng, S. (1999). Relative
American Psychological Association. versus absolute reinforcement effects: Implications for
Neef, N. A., & Lutz, M. N. (2001a). Assessment of vari- preference assessments. Journal of Applied Behavior
ables affecting choice and application to classroom Analysis, 32, 479-493. https://doi.org/10.901/
interventions. School Psychology Quarterly, 16, 239- jaba.1999.32-479
252. https://doi.org/10.1521/scpq.16.3.239.19887 Shirley, M. J., Iwata, B. A., Kahng, S., Mazaleski, J. L., &
Neef, N. A., & Lutz, M. N. (2001b). A brief computer- Lerman, D. C. (1997). Does functional communica-
based assessment of reinforcer dimensions affecting tion training compete with ongoing contingencies of
choice. Journal of Applied Behavior Analysis, 34, 57- reinforcement? An analysis during response acquisi-
60. https://doi.org/10.1901/jaba.2001.34-57 tion and maintenance. Journal of Applied Behavior
Neef, N. A., Mace, F. C., & Shade, D. (1993). Impulsiv- Analysis, 30, 93-104. https://doi.org/10.1901/jaba.
ity in students with serious emotional disturbance: 1997.30-93
The interactive effects of reinforcer rate, delay, and St. Peter Pipkin, C., Vollmer, T. R., & Sloman, K. N.
quality. Journal of Applied Behavior Analysis, 26, 37- (2010). Effects of treatment integrity failures during
52. https://doi.org/10.1901/jaba.1993.26-37 differential reinforcement of alternative behavior: A
Neef, N. A., Mace, F. C., Shea, M. C, & Shade, D. (1992). translational model. Journal of Applied Behavior
Effects of reinforcer rate and reinforcer quality on time Analysis, 43, 47-70. https://doi.org/10.1901/jaba.
allocation: Extensions of matching theory to educa- 2010.43-47
tional settings. Journal of Applied Behavior Analysis, 25,
Vollmer, T. R., Borrero, J. C., Lalli, J. S., & Daniel, D.
691-699. https://doi.org/10.1901/jaba.1992.25-691
(1999). Evaluating self-control and impulsivity in
Neef, N.A., Shade, D., & Miller, M. S. (1994). Assessing
children with severe behavior disorders. Journal of
influential dimensions of reinforcers on choice in stu-
Applied Behavior Analysis, 32, 451-466. https://doi.
dents with serious emotional disturbance. Journal of
org/10.1901/jaba.1999.32-451
Applied Behavior Analysis, 27, 575-583. https://doi.
org/10.1901/jaba.1994.27-575
Odum, A. L. (2011). Delay discounting: I’m a k, you’re a Received September 4, 2015
k. Journal of the Experimental Analysis of Behavior, 96, Final acceptance May 1, 2017
427-439. https://doi.org/10.1901/jeab.2011.96-423 Action Editor, Jason Bourret