Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2014, 47, 314–324 NUMBER 2 (SUMMER)

A LABORATORY COMPARISON OF TWO VARIATIONS OF


DIFFERENTIAL-REINFORCEMENT-OF-LOW-RATE PROCEDURES
JOSHUA JESSEL AND JOHN C. BORRERO
UNIVERSITY OF MARYLAND, BALTIMORE COUNTY

We compared 2 variations of differential-reinforcement-of-low-rate (DRL) procedures: spaced-


responding DRL, in which a reinforcer was delivered contingent on each response if a specified
interval had passed since the last response, and full-session DRL, in which a reinforcer was presented
at the end of an interval if the response rate was below criterion within the specified interval. We used
a human-operant procedure and analyzed within-session responding to assess any similarities or
differences between procedures. Data revealed a positive contingency between responding and
reinforcement under the spaced-responding DRL schedule and a negative contingency under the
full-session DRL schedule. Furthermore, 60% of the participants discontinued responding by the
last full-session DRL session. Implications for the appropriate procedural and taxonomical usage of
both DRL schedules are discussed.
Key words: differential reinforcement of low rates, differential reinforcement of other behavior,
human operant, interresponse time, translational research

Differential-reinforcement-of-low-rate (DRL) responding has elapsed (Catania, 2013). The


schedules are temporally based reinforcement interval between each response is known as the
schedules that arrange the delivery of reinforcers interresponse time (IRT), and this schedule is
contingent on reduced rates of responding sometimes called an IRT > t arrangement because
(Ferster & Skinner, 1957). Two commonly the time between responses (IRT) must be greater
conducted iterations of DRL schedules differ than the programmed interval (t) to produce a
based on the unit of analysis; one programs a reinforcer. These temporally sensitive schedules
contingent relation between each response, result in the subsequent reinforcement of both the
whereas the other programs a contingent relation relative IRT and response, respectively, character-
between an overall count within a particular time izing patterns of behavior that are referred to as
frame that can be intervals, sessions, or days spaced responding. This schedule has been aptly
(Deitz & Repp, 1973). named a spaced-responding DRL in the applied
In the first DRL variation, a response produces literature (Deitz, 1977).
a reinforcer only after a specified time of no The second DRL variation defines a contin-
gency between responding and a reinforcer
following the elapse of a predetermined interval
This study was conducted in partial fulfillment of the
requirements for the master’s degree of the first author at the as long as the overall rate within that interval is
University of Maryland, Baltimore County. We thank Iser below a predetermined criterion (Catania, 2013).
G. DeLeon and SungWoo Kahng for their insightful In some cases, the absence of responding is
contributions to earlier versions of this manuscript. Joshua
Jessel is now affiliated with Western New England
considered an acceptable dimension, and the
University. Preparation of this article was supported by characteristic features of maintaining low rates
Grant RO1HD049753 from the Eunice K. Shriver National during the DRL schedule are ignored (e.g., Bird,
Institute of Child Health and Human Development Hepburn, Rhodes, & Moniz, 1991; Hagopian,
(NICHD). Its contents are solely the responsibility of the
authors and do not represent the official views of NICHD. Kuhn, & Strother, 2009; Shaw & Simms, 2009;
Address correspondence to J. C. Borrero, Department of Turner, Green, & Braunling-McMorrow, 1990).
Psychology, University of Maryland, Baltimore County, For example, Deitz and Repp (1973) specifically
1000 Hilltop Road, Baltimore, Maryland 21250 (e-mail:
jborrero@umbc.edu). termed this variation a full-session DRL and arranged
doi: 10.1002/jaba.114 criteria whereby reinforcers were presented, not

314
VARIATIONS OF DIFFERENTIAL REINFORCEMENT 315

contingent on a response, but at the end of the are acceptable or valued, but only when the
interval if the target behavior (i.e., talking out responses occur at a low to moderate rate. These
during class without permission) occurred at a include responses such as hand raising in
rate less than a specified criterion. This procedure classroom settings (Austin & Bevan, 2011) and
inherently permitted reinforcer delivery given no independent eating or drinking during mealtimes
responding. Because zero responding is sufficient to (Anglesea, Hoch, & Taylor, 2008; Lennox,
produce reinforcer delivery, Deitz and Repp’s Miltenberger, & Donnelly, 1987; Wright &
example of a full-session DRL may have more in Vollmer, 2002). Thus, the goal is to reduce the
common with a differential-reinforcement-of-other- frequency of the target response but not to
behavior (DRO) schedule in that (a) reinforcers are completely extinguish responding. The spaced-
presented on an interval basis, (b) there is only an responding DRL schedule lends itself to cases for
indirect relation between reinforcer deliveries and which high rates of the target behavior could be
IRTs, and (c) there is a possible negative contingency hazardous (e.g., rapid eating could lead to an
between reinforcer presentation and the target increased chance of choking) but complete
response. The combined schedule could instead elimination of the response would alternatively
be described as an alternative DRO DRL schedule lead to dangerous complications (e.g., starvation).
(Ferster & Skinner, 1957). This description incor- In contrast, full-session DRL schedules, or the
porates the two alternative contingencies available alternative DRL DRO schedules, have been
for contacting reinforcement, in that the organism implemented in cases in which the target behavior
can either not respond at all or respond x or fewer is inappropriate and reduced rates are acceptable
times. To draw attention to this detail may seem but complete elimination is ideal. Common
trivial, but the effects on behavior could be as examples of these behaviors include talking out
substantial. Specifically, the schedule arrangement during class without permission (Deitz & Repp,
supports both low-rate and zero responding. 1973) or engaging in stereotypy (Singh, Dawson,
Therefore, the spaced-responding DRL and & Manning, 1981). The DRL component
full-session DRL might establish disparate accounts for the permissible (tolerable) rate of
patterns of responding and might be properly inappropriate behavior to occur, whereas the
used in different contexts. Of these two methods alternative DRO component schedules reinforce-
of programming DRL schedules, the majority ment during intervals of no responding.
of basic research has been conducted using Austin and Bevan (2011) used an amalgam-
spaced-responding DRL schedules that main- ation of procedures from both the spaced-
tain responding. However, in application, the responding and full-session DRL schedules to
spaced-responding DRL is used infrequently decrease requests for help by three typically
and is supplanted by the full-session DRL that developing elementary school students. The
might actually be more likely to eliminate authors reported considerable decreases in re-
responding. In other words, basic and applied quests during the treatment component. How-
research that involves DRL arrangements has ever, some features of the results warrant
focused on different contingencies that may consideration. Although the ability to eliminate
promote very different response patterns. The the target response completely has often been
explanation for this difference may be elucidated regarded as a strength of the DRO schedule, the
when considering why applied researchers target response (requests for help) selected for this
would select a DRL arrangement as a means study would likely not fall under that category.
of clinical intervention. The full-session DRL schedule has often been
Spaced-responding DRL schedules have often preferred over the spaced-responding DRL
been implemented to reduce response forms that schedule because of the relative simplicity of
316 JOSHUA JESSEL and JOHN C. BORRERO

the nonresetting interval for teachers in classroom each participant can be obtained from the
settings (Deitz, 1977; Deitz & Repp, 1973); Supporting Information on the Wiley Online
however, this may not be clinically appropriate Library.)
when considering the possibility of extinguishing
an appropriate classroom response. Apparatus and Settings
The purpose of the current study was to Participants were situated in a room (3 m by
compare the effects of the spaced-responding 3 m) with a desk (with laptop computer) and
DRL schedule and the full-session DRL schedule chair. The participant was asked to be seated
in a preliminary human-operant investigation while the instructions were read to him or her.
using college students as participants. A DRO- The participant was asked to read along with the
based schedule was specifically targeted because of instructions on the computer screen, and to begin
the frequency with which DRO procedures are the session when he or she was ready. The
represented in the research literature related to instructions included the following statement:
reducing problem behavior (Kahng, Iwata, & Thank you for your participation in this
Lewin, 2002). The study was designed to provide study. Your goal is to earn as many points as
laboratory (translational) evidence for possible possible before time is up. There is a
similarities or differences between the underlying possibility of earning up to $50 (with other
mechanisms of the procedures and whether or not monetary rewards for second and third place).
the change in a procedural taxonomy is warranted. There are different ways to earn points.
That is, if the full-session DRL schedule results in Clicking on the colored buttons in different
maintained responding similar to that of the patterns could add to your earnings, not affect
spaced-responding DRL, then there would be your earnings, or subtract from your earnings.
little evidence to suggest change, because the All of your earnings will be visible throughout
implications for application will not differ for the experiment at the top of the screen, and a
tone will sound with each distribution. Your
practitioners and applied researchers.
time here will approximate 1 hour with a
minute break every 5 minutes. You are free to
METHOD leave at any point during this study; however,
you will only be eligible to win the monetary
Participants prizes on completion. Remember, you are
Sixteen university students (seven women, trying to beat other participants so do your
nine men), with an age range of 18 to 29 years best! Click the START button when you are
old, were recruited for participation. All partic- ready and good luck!
ipants were sufficiently proficient in the manip-
ulation of a computer mouse and had experience The program was created using Microsoft
using computers. Three participants served as Visual Basic and consisted of 24 colored squares
pilots during the initial stages of program (100 by 100 pixels) in the center of the screen
development. Two of the remaining 13 partic- with a text bar at the center top that displayed
ipants engaged in similar response rates during real-time point accumulation. The colors of the
the variable-ratio (VR) and extinction condi- squares differed depending on the programmed
tions, and one participant refused to wear the reinforcement schedule. The squares were sta-
headphones and could not hear when points tionary, and each click on a square made the
were being delivered. Therefore, 10 data sets square disappear. All the squares that were clicked
from the original 16 participants were produced; regenerated after 6 s. Therefore all 24 squares
one participant (P-8) completed two sessions. were visible every 6 s, and at no time was the
(The reinforcement schedules and parameters for participant left without any squares to click.
VARIATIONS OF DIFFERENTIAL REINFORCEMENT 317

Design and Response Measurement Participants could not earn points during the
The computer program automatically re- extinction phase. This arrangement included
corded mouse clicks. Every 1 s, a preset automat- blue squares that did not disappear when clicked
ed timer recorded the current frequency of clicks (the squares disappeared contingent on each click
and points to a notepad file. Clicks were recorded during the reinforcement phases). We elected not
if they occurred on the squares. All other clicks on to remove squares contingent on clicks during
the gray background produced no differential extinction, because three pilot participants
consequences and were not recorded. The continued to click on the squares during sessions
primary dependent variable was rate of mouse in which points were never delivered.
clicks expressed as responses per second. During the VR phase, points were presented
ABA(C þ D) reversal designs were conducted on a VR 15 ( 5) schedule of reinforcement. The
to assess possible effects between baseline and the scheduled mean number of responses was
variations of the differential-reinforcement con- selected based on pilot data that indicated that
ditions. The initial reinforcer assessment (ABA) responding would persist under a VR 15 ( 5).
consisted of three VR (A) blocks, followed by The algorithm used to generate the VR schedule
three blocks of extinction (B), and a return to the was based on that provided by Dixon and MacLin
VR blocks. The imbedded multielement (C þ D) (2003). This schedule was correlated with green
design was implemented within treatment con- squares. The VR schedule constituted the
ditions to assess possible differentiating results of reinforcement condition in the reinforcer assess-
the DRL IRT (C) and the DRO rate (D). The ment and the baseline condition in the compara-
design was extended to an ABA(C þ D)ACAD tive differential reinforcement assessment.
for the one participant who opted to participate During the spaced-responding DRL phase,
for two sessions. However, all were given the points were presented contingent on the first
opportunity to participate for both sessions. instance of a mouse click that followed the
completion of a preset interval in which no
responding occurred. The initial IRT interval was
Procedure calculated as twice the mean IRT interval during
Sessions lasted a minimum of 60 min and the last VR phase. If clicks occurred at any point
consisted of three or more sequential 5-min before the interval elapsed, the automated timer
blocks, for a total of 12 to 15 blocks each session. reset to the original IRT and no points were
An optional 1-min break followed each block, delivered. Points were not presented following
and a 5-min break followed every six blocks. the interval if no response occurred and were
Points were delivered for mouse clicks based on withheld, without restarting the timer, until the
the relevant schedule of reinforcement in place. target response was emitted. This schedule was
After completion of the study, monetary rewards correlated with yellow squares.
of $50, $40, and $10 were awarded to the Both the spaced-responding DRL IRT and the
participants who earned the most, second-most, full-session DRL interval were calculated from
and third-most amount of points, respectively. responding during the VR phase. During the full-
The initial nine blocks of each session session DRL phase, the interval duration was
consisted of a reinforcer assessment. The rein- calculated as four times the average IRT during
forcer assessment was comprised of three blocks the VR phase. Tolerance for the full-session DRL
of alternating extinction and VR schedules in an was defined as the maximum frequency of
ABA reversal design. The reinforcer assessment responding that could occur without resetting
was conducted to ensure that point delivery the reinforcer-delivery interval. Tolerance was
increased responding. calculated as half the mean response rate of the
318 JOSHUA JESSEL and JOHN C. BORRERO

target response during the VR condition. For frequency of clicks exceeded that of the set
example, if the mean IRT during the VR phase tolerance within the interval, points were not
was 2 s, the spaced-responding DRL would be delivered and the interval was restarted immedi-
calculated as 4 s (i.e., IRT > 4 s), the full-session ately after the violation of tolerance. This
DRL interval would be set as 8 s, and tolerance condition was correlated with red squares.
would be set as one response. Therefore, the
scheduled probability of reinforcer delivery Data Analysis
during the full-session DRL sessions remained Data from a cumulative record were analyzed
proportional to the spaced-responding DRL across 300 1-s bins per session to determine
condition. Although this method resulted in an comparative optimal and allowable response rates
interval substantially longer than DRO intervals during the last sessions of the spaced-responding
typically conducted in applied contexts (see DRL and full-session DRL conditions, respec-
Vollmer & Iwata, 1992), without this modifica- tively. The optimal response rate refers to the rate
tion comparative results between the spaced- of responding during the spaced-responding
responding DRL and full-session DRL proce- DRL condition in which the most reinforcers
dures would not be mutually interpretable. can be produced within the allotted session time.
Decreased response rates during the spaced- The allowable response rate refers to the rate of
responding DRL are a function of increased IRTs responding during the full-session DRL condi-
and longer intervals, relative to some baseline tion in which all reinforcers can be delivered
response rate. On the other hand, short intervals within the allotted session time without penalty
are preferred during DRO arrangements to of point loss. The slope of the cumulative data
reduce the negative side effects of extinction during the last session of each condition (i.e., VR,
and increase contact with the scheduled rein- spaced-responding DRL, full-session DRL) was
forcement. Thus, we elected to increase the full- calculated by taking the response frequency at
session DRL interval rather than decrease the 10 s (Y1), subtracting it from the response
spaced-responding IRT because decreases in frequency at 290 s (Y2), and dividing the
spaced-responding DRL durations relative to a difference by 280 (X2–X1).
full-session DRL interval would have contra- A contingency strength analysis (Luczynski &
indicative or no effects on responding that was Hanley, 2009, 2010) was conducted to provide a
already occurring slower than the imposed rate. quantifiable value of the contingent relations
For example, if a participant is already responding during the spaced-responding DRL phase and
at a pace of one response every 10 s, a minimum the full-session DRL phase. The contingency
IRT of 5 s would not likely affect behavior. In value was defined as the difference of two
addition, in comparison to the full-session DRL disparate conditional probabilities: response
intervals commonly conducted in applied set- conditional probability and point conditional
tings, the currently calculated intervals were probability. A positive contingency value from
relatively small. The intervals from Dietz and the contingency strength analysis supports a
Repp (1973) consisted of entire class periods of correlation between a response and a reinforcer;
50 min. The negative effects often associated with the higher the positive value the stronger the
long intervals may be avoided by the addition of correlation. Therefore, responses with positive
tolerance. contingencies are more likely to be strengthened,
During the full-session DRL condition, points whereas negative contingencies are more likely to
were delivered following the elapse of the interval, be weakened. By definition, for example, DRO
whether or not any clicks occurred at any time arrangements promote a negative contingency
during the session block. However, if the between responding and reinforcer presentation
VARIATIONS OF DIFFERENTIAL REINFORCEMENT 319

(Vollmer, Borrero, Wright, Van Camp, & DRL condition, whereas 60% of participants
Lalli, 2001). ceased responding during the final blocks of the
The response conditional probability was the full-session DRL condition.
quotient of the number of responses followed by Seven of the 10 participants produced response
a point (1-s window) divided by the total number rates below optimal performance during the
of responses. A 1-s window was selected for two spaced-responding DRL conditions and did not
reasons: (a) Previous research that has compared maximize the total possible points that could have
window sizes suggests that short window sizes been earned (mean difference from optimal ¼
(i.e., 2 s to 5 s) provide a more conservative 36.7%; range, 13.3% to 85.6%). However,
measure (Luczynski & Hanley, 2009), and (b) P-19 (Figure 2, bottom left) exhibited near
little variance was observed between 1 s and 3 s optimal responding (0.74 responses per second)
for the current data. The point conditional during the final spaced-responding DRL session
probability was defined as the quotient of the (difference from optimal ¼ 2.8%). Three par-
number of points not preceded by a response (1-s ticipants’ rate of responding was above optimal
window) divided by the total number of points (mean difference from optimal ¼ 40.8%; range,
delivered. Therefore, a continuous reinforcement 14.6% to 63.5%) with P-12 (Figure 1, bottom
schedule will result in a positive contingency right) exhibiting near optimal responding (0.7
value of 1 because the response will always responses per second) during the final spaced-
precede reinforcer delivery. In contrast, a DRO responding DRL session (difference from
schedule without tolerance will result in a optimal ¼ 1.3%).
contingency value of 1 because the numerator The full-session DRL schedule resulted in all
of the response conditional probability will but P-4 (Figure 1, top left) producing mean
always be 0 (i.e., zero responses will be followed response rates (M ¼ 0.11 responses per second;
by a point within 1 s). The contingency strength SD ¼ 0.13) below allowable performance (mean
analysis was conducted to determine whether or difference from allowable ¼ 62.7%; range,
not the full-session DRL schedule resulted in 100% to 86.5%). Furthermore, there was a
positive contingencies similar to most differential 20.4% difference from the allowable perfor-
reinforcement schedules or in a negative contin- mance for the only participant (P-4) who
gency, as is the case with the DRO schedule. continued to respond near allowable (0.51
responses per second).
Results of the contingency strength analyses for
RESULTS
all 10 participants were positive (M ¼ 0.71;
Figures 1 through 3 depict the results of the SD ¼ 0.26) during the spaced-responding DRL
reinforcer assessment and the comparison of condition, and greater than that of the VR schedule
response-reducing differential reinforcement value (0.07). The top panel of Figure 4 shows data
techniques for all participants as responses per for P-21, a representative example of the contin-
second. The solid horizontal line represents gency strength analysis. That is, during the spaced-
the optimal performance during the spaced- responding DRL condition, the response condi-
responding DRL condition, and the dashed tional probability exceeded the point conditional
horizontal line represents the allowable perfor- probability. In addition, negative contingency
mance during the full-session DRL condition. values (M ¼ 0.82; SD ¼ 0.28) were obtained
Both differential reinforcement techniques re- for 9 of 10 participants during the full-session DRL
duced responding compared to baseline for all condition. The bottom panel of Figure 4 shows
participants. Furthermore, all participants con- data for P-4, who was the only participant to
tinued to respond during the spaced-responding continue to respond at near allowable rates during
320 JOSHUA JESSEL and JOHN C. BORRERO

3 3
VR 15(±5) EXT VR 15( ±5) VR 15(±5) EXT VR 15( ±5)

Optimal
Allowable
2 2
CLICKS (RPS)

1 1 DRL-s

P-4 P-11 DRL-f


0 0
5 10 15 5 10 15
3 3

2 2
CLICKS (RPS)

1 1

P-7 P-12
0 0
5 10 15 5 10 15
BLOCKS BLOCKS

Figure 1. Response rates during the variable ratio (VR), extinction (EXT), spaced-responding DRL (DRL-s), and full-
session DRL (DRL-f ) conditions across sessions for P-4, P-11, P-7, and P-12. The solid horizontal line depicts the optimal
response rate for producing the most points during the spaced-responding DRL condition. The dashed horizontal line
depicts the allowable response rate without resetting point delivery during the full-session DRL condition. RPS ¼ responses
per second.

the full-session DRL condition, and who showed DISCUSSION


the highest mean negative contingency (M ¼ These findings replicate previous research on
0.11; SD ¼ 0.16) with the last session a positive DRL and DRO in that low-rate spaced-response
value (0.06) comparable to that of the positive VR patterns and the elimination of responding were
value (0.07). obtained in the spaced-responding DRL and full-
We also evaluated responding during the VR, session DRL conditions, respectively (Deitz &
spaced-responding, and DRL conditions as a Repp, 1973). Although the full-session DRL
proportion of responding during each participant’s schedule permitted a predetermined response rate
extinction condition. The greatest proportional derived from the participants’ baseline perform-
increase in responding compared to extinction ances, many participants stopped responding. We
occurred in the VR for all participants. In addition, also observed an overall negative contingency value
90% of the participants displayed higher rates of consistent with a DRO-based schedule during the
proportional responding during the spaced-respond- full-session DRL. Thus, the two schedules reduced
ing DRL condition (M ¼ 2.96; SD ¼ 1.72) com- responding compared to the VR schedule of
pared to extinction. Notably, 30% of the reinforcement, with the full-session DRL arrange-
participants displayed higher rates of proportional ment resulting in the larger reduction.
responding to extinction during the full-session Although previous research has supported the
DRL condition (M ¼ 0.54; SD ¼ 0.51). use of full-session DRL schedules over spaced-
VARIATIONS OF DIFFERENTIAL REINFORCEMENT 321

3 3
VR 15(±5) EXT VR 15( ±5) VR 15(±5) EXT VR 15( ±5)

Optimal
Allowable
2 2
CLICKS (RPS)

DRL-s
1 1

P-15 P-17 DRL-f


0 0
5 10 15 5 10 15
3 3

2 2
CLICKS (RPS)

1 1

P-19 P-21
0 0
5 10 15 5 10 15
BLOCKS BLOCKS

Figure 2. Response rates during the variable ratio (VR), extinction (EXT), spaced-responding DRL (DRL-s), and full-
session DRL (DRL-f ) conditions across sessions for P-15, P-17, P-19, and P-21. The solid horizontal line depicts the optimal
response rate for producing the most points during the spaced-responding DRL condition. The dashed horizontal line
depicts the allowable response rate without resetting point delivery during the full-session DRL condition. RPS ¼ responses
per second.

responding DRL schedules due to the ease of grammed IRT) may be assuaged when consider-
implementation (Deitz, 1977; Deitz & ing the possibility of the use of widespread
Repp, 1973), results of the present study suggest technology such as tablets in classroom settings.
that functional differences in these procedures are Applications could be created on tablets with
not simply a matter of structural semantics. which teachers could select the IRT for each
Evaluation and implementation of spaced-re- student immediately before the start of classroom
sponding DRL and full-session DRL should be activities. The teacher would only have to click on
considered in the context of clinical or research the child’s name after each response to determine
goals. The more effortful approach (spaced- whether or not he or she met the criteria for
responding DRL) may be required when the goal reinforcer presentation. Future research could
is to sustain responding, albeit at rates lower than assess the use of programmatic schedules of
those produced under baseline conditions (e.g., reinforcement on handheld devices to reduce
applications to rapid eating, excessive hand work-related effort needed to create individual-
raising, or tattling). In addition, the limitations ized interventions for classwide implementation.
of using spaced-responding DRL schedules (e.g., The results of the contingency strength
requires individual timers for each student to be analysis also provide insight into the possible
reset following responding below the pro- quantitative differences between the reductions
322 JOSHUA JESSEL and JOHN C. BORRERO

VR 15(±5)
3 3
VR 15(±5) EXT VR 15( ±5) EXT

Optimal
Allowable
2 2
CLICKS (RPS)

1 DRL-s 1

P-18 DRL-f P-8


0 0
5 10 15 5 15 25
BLOCKS BLOCKS

Figure 3. Response rates during the variable ratio (VR), extinction (EXT), spaced-responding DRL (DRL-s), and full-
session DRL (DRL-f ) conditions across sessions for P-18 and P-8. The solid horizontal line depicts the optimal response rate
for producing the most points during the spaced-responding DRL condition. The dashed horizontal line depicts the
allowable response rate without resetting point delivery during the full-session DRL condition. RPS ¼ responses per second.

in responding under the spaced-responding and be equivalent could result in misapplication (i.e.,
full-session DRL schedules. The VR and spaced- inadvertently maintaining inappropriate behavior
responding DRL schedules both maintained with the spaced-responding DRL or inadvertently
higher rates than those observed during extinc- eliminating appropriate behavior with the full-
tion. This should not be surprising, considering session DRL). We suggest that the full-session
that a positive contingency between responding DRL could be expressed as a variation of a DRO
and point delivery existed for both VR and schedule and that the term DRO-with-tolerance
spaced-responding DRL schedules. Reductions schedules could be considered when reducing
were observed during the spaced-responding inappropriate behavior. In a very literal sense, a
DRL sessions when mean rates were analyzed; predetermined tolerated level of inappropriate
however, the contingency value was actually behavior is overlaid on top of a DRO schedule.
greater than the comparative VR reinforcement Not only does the definition of DRO with
schedule from a within-session perspective. An tolerance incorporate both components in the
artificial reduction in response rates was created basic arrangement (alternative DRO DRL), but it
in spaced-responding DRL schedules by reinforc- also implicitly suggests targeting undesirable
ing larger IRTs relative to baseline response rates. behaviors. This simplifies the categorical usage
In contrast, the full-session DRL schedule for practitioners as DRL schedules for appropriate
produced a negative contingency between re- behavior and DRO schedules for inappropriate
sponding and point delivery. behavior. Furthermore, modifying its classifica-
A change in procedural taxonomy may be tion to a DRO schedule (rather than the
warranted. Currently, little distinction is made previously reported DRL schedule) may improve
between the two variations of DRL schedules the selection of interventions by clarifying its
(e.g., Cooper, Heron, & Heward, 2007) although effects, by highlighting the units of analysis, and
it is evident that they differ in process and by underscoring the behavioral mechanisms that
application. Assuming the two DRL schedules to are responsible for effects obtained therein.
VARIATIONS OF DIFFERENTIAL REINFORCEMENT 323

1.0 The brevity of our procedures and emphasis on


P-21 DRL-s temporal properties of responding also lend
CONTINGENCY VALUE

0.5
themselves to evaluations for which mechanisms
of timing are thought to be operative or
important. Dube and McIlvane (2002) used
0.0 this type of assessment approach to evaluate
sensitivity to concurrent reinforcement schedules
-0.5 by individuals with intellectual disabilities. Inter-
estingly, most of the participants’ responding in
DRL-f
the current assessment was above optimal perfor-
-1.0
mance, resulting in mean pauses that were longer
1.0
P-4 than the set IRT. This is directly juxtaposed with
previous basic research with rats, in which subjects
CONTINGENCY VALUE

0.5 were more likely to engage in below optimal


performances, resulting in mean pauses that were
shorter than the set IRT (Mazur, 1994). Although
0.0
little can be responsibly deduced from these
differences due to the procedural variance
-0.5 between the present human operant arrangement
and basic research with nonhumans, it may still be
-1.0
important to note for those interested in studying
2 4 6 temporal dimensions of behavior.
BLOCKSKS One limitation of the present study was the lack
of multiple tolerance settings. It is therefore
Figure 4. Data in the top panel illustrate a strong unknown if an increased ratio of allowable
positive contingency between responding and point delivery responses to the same interval duration would
(DRL-s) and a strong negative contingency between
responding and point delivery (DRL-f ). These are
result in responding more similar to that under the
representative data for 9 of 10 participants. Data in the spaced-responding DRL. The tolerance may have
bottom panel are those from P-4, the only participant who been too stringent and resulted in increased contact
continued to respond at near allowable rates during the full- with the DRO-related contingency of the full-
session DRL condition. session DRL schedule. However, many partici-
pants who discontinued responding reported the
ability to continue without resetting the timer and
Our brief experimental arrangement also could in some cases reported the approximate full-session
be used to examine two previously reported DRL interval and tolerance setting. Therefore,
explanations for the suppressant effects of DRO even though the participants could tact the
and whether similar outcomes would be observed contingencies with surprising accuracy, they chose
with DRO schedules implemented with tolerance. not to respond at all. Of further interest was the
One explanation is that “other behavior” that fact that many of the studies that have imple-
occurs contiguously with the presentation of the mented the full-session DRL schedule used
reinforcer displaces target responding (Ecott & arbitrary reinforcers intended to compete with
Crtichfield, 2004; Poling & Ryan, 1982). The multiple unknown sources of qualitatively dissim-
second explanation conceptualizes the DRO ilar reinforcement (e.g., Austin & Bevan, 2011;
schedule as a form of negative punishment because Deitz & Repp, 1973). The current study used only
responding delays reinforcer onset (Lattal, 2013). one source of reinforcement and one value of
324 JOSHUA JESSEL and JOHN C. BORRERO

tolerance for each participant, but future research T. D. Hackenberg, G. P. Hanley, & K. A. Lattal (Eds.),
could extend the generality of these findings to APA handbook of behavior analysis (pp. 33–63).
Washington, DC: American Psychological Association.
multiple sources of reinforcement and multiple doi: 10.1037/13937-002
values of tolerance. Lennox, D. B., Miltenberger, R. G., & Donnelly, D. R.
(1987). Response interruption and DRL for the
REFERENCES reduction of rapid eating. Journal of Applied Behavior
Analysis, 20, 279–284. doi: 10.1901/jaba.1987.20-279
Anglesea, M. M., Hoch, H., & Taylor, B. A. (2008). Luczynski, K. C., & Hanley, G. P. (2009). Do children prefer
Reducing rapid eating in teenagers with autism: Use of a contingencies? An evaluation of the efficacy of and
pager prompt. Journal of Applied Behavior Analysis, 41, preference for contingent versus noncontingent social
107–111. doi: 10.1901/jaba.2008.41-107 reinforcement during play. Journal of Applied Behavior
Austin, J. L., & Bevan, D. (2011). Using differential Analysis, 42, 511–525. doi: 10.1901/jaba.2009.42-511
reinforcement of low rates to reduce children’s requests Luczynski, K. C., & Hanley, G. P. (2010). Examining the
for teacher attention. Journal of Applied Behavior generality of children’s preference for contingent
Analysis, 44, 451–461. doi: 10.1901/jaba.2011.44-451 reinforcement via extension to different responses,
Bird, F., Hepburn, S., Rhodes, K., & Moniz, D. (1991). reinforcers, and schedules. Journal of Applied Behavior
Multiple reinforcement contingencies to reduce aggres- Analysis, 43, 397–409. doi: 10.1901/jaba.2010.43-397
sion, self-injury, and dysfunctional verbal behaviors in Mazur, J. E. (1994). Learning and behavior. Englewood
an adult who is sensory impaired. Behavioral Inter- Cliffs, NJ: Prentice Hall.
ventions, 6, 367–383. doi: 10.1002/bin.2360060506 Poling, A., & Ryan, C. (1982). Differential-reinforcement-
Catania, A. C. (2013). Learning (5th ed.). Upper Saddle of-other-behavior schedules. Behavior Modification, 6,
River, NJ: Prentice Hall. 3–21. doi: 10.1177/01454455820061001
Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Shaw, R., & Simms, T. (2009). Reducing attention-
Applied behavior analysis (2nd ed.). Upper Saddle River, maintained behavior through the use of positive
NJ: Prentice Hall. punishment, differential reinforcement of low rates,
Deitz, S. M. (1977). An analysis of programming DRL and response marking. Behavioral Interventions, 24,
schedules in educational settings. Behaviour Research 249–263. doi: 10.1002/bin.287
and Therapy, 15, 103–111. doi: 10.1016/0005-7967 Singh, N. N., Dawson, M. J., & Manning, P. (1981). Effects
(77)90093-6 of spaced responding DRL on the stereotyped behavior of
Dietz,[sic], S. M., & Repp, A. C. (1973). Decreasing profoundly retarded persons. Journal of Applied Behavior
classroom misbehavior through the use of DRL Analysis, 14, 521–526. doi: 10.1901/jaba.1981.14-521
schedules of reinforcement. Journal of Applied Behavior Turner, J. M., Green, G., & Braunling-McMorrow, D.
Analysis, 6, 457–463. doi: 10.1901/jaba.1973.6-457 (1990). Differential reinforcement of low rates of
Dixon, M. R., & MacLin, O. H. (2003). Visual basic for responding (DRL) to reduce dysfunctional social
behavioral psychologists. Reno, NV: Context Press. behaviors of a head injured man. Behavioral Inter-
Dube, W. V., & McIlvane, W. J. (2002). Quantitative ventions, 5, 15–27. doi: 10.1002/bin.2360050103
assessments of sensitivity to reinforcement contingen- Vollmer, T. R., Borrero, J. C., Wright, C. S., Van Camp, C.,
cies in mental retardation. American Journal of Mental & Lalli, J. S. (2001). Identifying possible contingencies
Retardation, 107, 136–145. during descriptive analyses of severe behavior disorders.
Ecott, C. L., & Critchfield, T. S. (2004). Noncontingent Journal of Applied Behavior Analysis, 34, 269–287. doi:
reinforcement, alternative reinforcement, and the 10.1901/jaba.2001.34-269
matching law: A laboratory demonstration. Journal of Vollmer, T. R., & Iwata, B. A. (1992). Differential
Applied Behavior Analysis, 37, 249–265. doi: 10.1901/ reinforcement as treatment for behavior disorders:
jaba.2004.37-249 Procedural and functional variations. Research in Develop-
Ferster, C. B., & Skinner, B. F. (1957). Schedules of mental Disabilities, 13, 393–417. doi: 10.1016/0891-
reinforcement. East Norwalk, CT: Appleton-Century- 4222(92)90013-V
Crofts, doi: 10.1037/10627-000 Wright, C. S., & Vollmer, T. R. (2002). Evaluation of a
Hagopian, L. P., Kuhn, D. E., & Strother, G. E. (2009). treatment package to reduce rapid eating. Journal of
Targeting social skills deficits in an adolescent with pervasive Applied Behavior Analysis, 35, 89–93. doi: 10.1901/
developmental disorder. Journal of Applied Behavior Analysis, jaba.2002.35-89
42, 907–911. doi: 10.1901/jaba. 2009.42-907
Kahng, S., Iwata, B. A., & Lewin, A. B. (2002). Behavioral
treatment of self-injury 1964 to 2000. American Journal
on Mental Retardation, 107, 212–221. Received June 15, 2012
Lattal, K. A. (2013). The five pillars of the experimental Final acceptance December 12, 2013
analysis of behavior. In G. J. Madden, W. V. Dube, Action Editor, Michael Kelley

You might also like