Literature review on behavior analysis

EFFICACY OF AND PREFERENCE FOR REINFORCEMENT AND RESPONSE COST IN TOKEN ECONOMIES

ERICA S. JOWETT HIRST SOUTHERN ILLINOIS UNIVERSITY

CLAUDIA L. DOZIER UNIVERSITY OF KANSAS

AND

STEVEN W. PAYNE STATE UNIVERSITY OF NEW YORK

Researchers have shown that both differential reinforcement and response cost within token economies are similarly effective for changing the behavior of individuals in a group context (e.g., Donaldson, DeLeon, Fisher, & Kahng, 2014; Iwata & Bailey, 1974). In addition, these researchers have empirically evaluated preference for these procedures. However, few previous studies have evaluated the individual effects of these procedures both in group contexts and in the absence of peers. Therefore, we replicated and extended previous research by determining the individual effects and preferences of differential reinforcement and response cost under both group and individualized conditions. Results demonstrated that the procedures were equally effective for increasing on-task behavior during group and individual instruction for most chil- dren, and preference varied across participants. In addition, results were consistent across partici- pants who experienced the procedures in group and individualized settings. Key words: differential reinforcement, independent group contingency, preference, response

cost, token economy

The token economy is a common behavioral intervention that has been demonstrated to be effective for increasing appropriate behavior and decreasing inappropriate behavior for many populations across different settings (Doll, McLaughlin, & Barretto, 2013; Hackenberg, 2009; Kazdin, 1977). Token economies involve delivery, removal, or both delivery and removal of conditioned reinforcers (e.g., tokens and points) that can be exchanged for back-up rein- forcers (e.g., prizes, treats, and leisure activ- ities). When tokens are delivered contingent on appropriate behavior or for the absence of inap- propriate behavior, these procedures are termed

differential reinforcement of alternative behavior (DRA) or differential reinforcement of other behavior (DRO), respectively. When tokens are removed contingent on inappropriate behavior or for the absence of appropriate behavior, this procedure is termed response cost (RC). An advantage of token economies is that

they can be implemented with a group of indi- viduals as a general behavior-management strat- egy during small-group instruction or as a classwide intervention. Classwide behavior- management strategies such as token economies should be considered to address minor disrup- tive behavior, to increase motivation for learn- ing, or as a complement to an individualized intervention. However, general behavior- management strategies may not be effective in isolation for some individuals who engage in severe problem behavior or have more intense

Correspondence concerning this article should be addressed to Claudia L. Dozier, Department of Applied Behavioral Science, University of Kansas, Lawrence, Kan- sas 66045 (e-mail: [email protected]). doi: 10.1002/jaba.294

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2016, 49, 329–345 NUMBER 2 (SUMMER)

329

deficits in learning. These individuals may require more individualized, function-based assessment, intervention, and additional sup- port. Regardless, token economies are common in classrooms and numerous other environ- ments because they are likely to create motiva- tion for changes in behavior for most individuals in the group, creating a more man- ageable and effective learning environment. After numerous studies were conducted to

demonstrate the effectiveness of reinforcement and RC procedures in token economies, researchers began to compare the effectiveness of these two procedures (e.g., Brent & Routh, 1978; Broughton & Lahey, 1978; Iwata & Bai- ley, 1974; Panek, 1970). Overall, most studies that have compared differential reinforcement (DR) to RC have demonstrated equal effective- ness of the two procedures (e.g., Capriotti, Brandt, Ricketts, Espil, & Woods, 2012; Donaldson, DeLeon, Fisher, & Kahng, 2014; Iwata & Bailey, 1974; McGoey & DuPaul, 2000). However, these results are limited in two important ways. First, most studies involved the use of group contingencies (i.e., the implementation of the procedures in the context of a group in which others are present), which may have influenced responding. For example, comments made or behaviors mod- eled by others in the group may have influ- enced target responding. Second, most studies reported only group averages with respect to target behavior, which does not allow analysis of individual differences. For example, Iwata and Bailey (1974) compared DRO and RC for decreasing rule violations and increasing on- task behavior of 15 children in a classroom. During DRO, tokens were delivered at the end of a 3- to 5-min interval if no rule violations occurred during that interval. During RC, tokens were removed at the end of an interval if any rule violations occurred during that inter- val. The children could earn or lose up to 10 tokens throughout a 30-min math period, and the tokens could be exchanged for snacks

and free time. Results showed that the proce- dures were similarly effective for reducing rule violations and off-task behavior. However, the authors reported group averages, which may not be representative of individual responding. Furthermore, because the study was conducted as a group intervention, the influence of peer behavior on target responding is unknown. More recently, Donaldson et al. (2014) com-

pared DRO and RC for decreasing the disrup- tive behavior of 12 first-grade students. Although the procedures were implemented in a group context, the authors reported both group-average outcomes and individual out- comes. Group-average data showed low to zero levels of problem behavior; however, an analysis of individual data showed that responding dur- ing DRO was somewhat variable for four of the 12 participants. Although this study, along with Iwata and Bailey (1974) and most others, provides preliminary evidence regarding the effectiveness of reinforcement and RC when used in a token economy, because the proce- dures were implemented in a group context, the influence of peers on target responding is unknown. For example, individuals may show an increase or decrease in target behavior because their peers are (a) engaging in a target behavior, (b) prompting them to engage in a target behavior, (c) providing reinforcers (e.g., attention) for them to engage in appropriate target behavior, (d) implementing punishers (e.g., reprimands) for not engaging in a target behavior (Salend & Kovalich, 1981), or (e) extinguishing previously reinforced target behavior (e.g., no longer delivering attention). Therefore, to further isolate the effects of rein- forcement and RC contingencies in token economies, conducting the comparison while students work independently or are otherwise not in the presence of others might be important (Capriotti et al., 2012; Sindelar, Honsaker, & Jenkins, 1982). Furthermore, comparing responding of a single individual when in the presence and absence of peers to

ERICA S. JOWETT HIRST et al.330

determine whether changes in responding are associated with the presence or absence of peers would be useful. In addition to comparing the effectiveness of

DR and RC procedures in individual and group contexts, considering preference is also important; however, only two studies that have compared DR and RC in token economies have empirically evaluated preference (Donaldson et al., 2014; Iwata & Bailey, 1974). Iwata and Bailey (1974) compared the effects of DRO and RC for reducing disruptive classroom behavior displayed by 15 elementary school special-education students. To deter- mine preference across the procedures, the experimenters conducted a choice assessment during which each child was given the opportu- nity to select which token procedure would be implemented for a particular session. After all children made a selection, the chosen token procedure was implemented for each child. The results showed that four students chose DRO most often, five students chose RC more often, and six students switched their selection across opportunities. Donaldson et al. (2014) used a similar procedure and found that six of the 12 children preferred RC, four children pre- ferred DRO, and two children had approxi- mately equal preference. These studies provide evidence that prefer-

ence varies among individuals; however, the results are limited, at least in Donaldson et al. (2014), because children made selections vocally and in the presence of their peers (Iwata & Bailey, 1974, did not provide infor- mation regarding how or where children made a selection). Therefore, some children’s selec- tions may have been influenced by the presence or behavior (e.g., choices or comments) of their peers (Donaldson et al., 2014). To isolate indi- vidual preference, it is important to conduct a preference assessment when the child is not in the presence of his or her peers (e.g., Layer, Hanley, Heal, & Tiger, 2008). For example, Layer et al. (2008) presented choices on an

upright board in front of each child with the choices facing the child (not visible to other children) and then had the child use a motor response (i.e., pointing), rather than a vocal response (i.e., stating which procedure he or she liked best), to make his or her selection. This procedure controlled for both visual and auditory observation of other children’s choice. Overall, given the demonstrated effectiveness

of DR and RC but unknown influence of peers and lack of empirical data for preference in the absence of peers, further research is warranted. The current study involved several evaluations that replicate and extend previous research. The purpose of the first evaluation was to replicate research directly comparing the effectiveness of DR and RC procedures in a group setting. The second purpose was to provide a direct compar- ison of the effectiveness of DR and RC proce- dures for the on-task behavior of individual children engaged in a solitary work task. The third purpose was to evaluate individual prefer- ence of all children in the absence of peers. Finally, responding of individuals who partici- pated in both the small-group activity and the solitary work task was compared to determine if the presence of peers influenced responding.

STUDY 1: DR VERSUS RC (GROUP)

Method Participants and setting. Three groups of

three typically developing preschool-aged (3 to 5 years old) children who attended a university- based preschool program participated. All chil- dren could follow multistep instructions (e.g., walk to your cubby, hang up your jacket, and come sit on the floor) and communicated using vocal speech. We conducted sessions 3 to 5 days per week, once or twice per day, in a quiet area of the classroom separate from all other chil- dren. During each session, only one group of participants was present. Participants sat next to one another on the floor on designated mats across from the experimenter, and one to two

331REINFORCEMENT AND RESPONSE COST

data collectors and relevant session materials were present. Materials. During all sessions, small-group

activity materials were present. Materials included plastic letters and numbers for expres- sive labeling and individual bingo boards with various items (i.e., plastic buttons and jewels) for matching. During some sessions, tokens (i.e., pennies) were present that could be earned or lost. Tokens were attached to and removed from laminated strips of paper (approximately 10.2 cm by 30.5 cm) with 10 square pieces of Velcro. Participants earned access to a toy room with tangible items (e.g., stickers, plastic rings, spin tops, sticky hands), edible items (e.g., gummies, Smarties, Skittles, and M&Ms), and leisure activities (e.g., video games and DVDs) via token exchange following some sessions (DR and RC). Different-colored materials (pos- ters and token boards) were present during each of the different conditions to aid in dis- crimination between conditions. Response measurement and interobserver agree-

ment. Trained graduate and undergraduate stu- dents collected data using paper and a pencil. The dependent variable was percentage of intervals with on-task behavior. We defined on- task behavior as sitting on a mat (i.e., bottom on the mat), keeping hands to oneself (i.e., keeping hands in lap unless instructed to manipulate activity materials), and sitting quietly (i.e., talking only when the experi- menter asked or called on the participant to respond). We partitioned sessions into 5-s intervals and scored on-task behavior for each child using a momentary-time-sample proce- dure. That is, at the end of every 5-s interval (signaled by an auditory cue), the data collector scored whether each child was on task at that moment. After each session, we collected data for on-task behavior of an individual child by dividing the number of intervals on task by the total number of intervals in the session and converting the result to a percentage. In addi- tion, for two groups, experimenters collected

data on the number of tokens that remained on each participant’s board at the end of a DR ses- sion or the number of empty spaces on each participant’s board at the end of each RC ses- sion. We later subtracted the number of empty spaces counted after RC sessions from 10 to compare number of net tokens in each session. Two independent observers collected data

for at least 30% of sessions and then calculated interobserver agreement for on-task behavior by dividing the number of 5-s intervals during which both observers agreed by the total num- ber of intervals and converting the result to a percentage. We defined an agreement for on- task behavior as both observers scoring or not scoring the occurrence of the behavior in a given interval. We calculated interobserver agreement for token count using the total method. That is, we divided the smaller num- ber of tokens that remained on a board (at the end of each DR session) or were missing from the board (at the end of each RC session) by the larger number and converted the result to a percentage. Interobserver agreement averaged 93% (range, 73% to 100%) for on-task behav- ior and 99% (range, 88% to 100%) for token count. Procedure. All sessions lasted 5 min. During

all sessions, the participants sat next to one another and in front of the experimenter in a small area away from the other children in the classroom. In addition, the experimenter placed bingo boards with pieces and token boards (in some sessions) in front of each participant and a colored poster board on the wall in front of the children. Before the start of the first ses- sion of each condition, the experimenter described the rules and the session contingen- cies and required each participant to practice engaging in related behaviors (e.g., sitting quietly, talking out of turn, keeping hands in lap, and touching materials) to experience the consequences associated with each behavior. During the 5-min sessions, the experimenter provided continuous individual and group

ERICA S. JOWETT HIRST et al.332

instructions to name letters and numbers (e.g., the experimenter held up a plastic letter and said, “Caroline, what letter is this?” and “Can everybody tell me what letter this is?”) and place a marker on a specific bingo board letter or number (e.g., “Ok everyone, put a gem on the letter d”). The experimenter delivered sev- eral instructions during a session in a way that was similar to instructions delivered during a classroom activity; however, the rate at which instructions were provided varied depending on responding. During all sessions, if a child (or children) responded correctly, the experi- menter delivered praise, and if any child did not respond correctly, the experimenter prompted the correct response and then moved on to another instruction. First, the experimenter conducted baseline

sessions to determine the level of on-task behavior in the absence of programmed conse- quences. Next, the experimenter practiced token trading with the participants. That is, the experimenter gave each child tokens and the opportunity to trade the tokens for various items (e.g., prizes and snacks). Next, we com- pared DR and RC to determine their effects on on-task behavior. During DR and RC sessions, the experimenter observed each participant in the group at the same moment every 30 s on average (ranging from 15 to 45 s) according to a schedule based on a pseudorandom number generator in Excel. We created three versions of the schedule and rotated across sessions to reduce the likelihood that the participants would learn a schedule. During each scheduled observation and depending on the condition, the experimenter quietly delivered a token to every child who was on task at that moment (DR) or removed a token from any child who was off task at that moment (RC). The experi- menter did not say anything when delivering or removing a token. We used the same schedules across both conditions; therefore, the possible number of net tokens across conditions was equal (i.e., 10 tokens). In addition, the last

opportunity to earn or lose a token was at the last second of each session; therefore, no partic- ipant could earn or lose all tokens before the end of the session. After each DR and RC session, an experi-

menter took the participant to a room that contained many different toys, leisure activities, edible items, and trinkets that were not found in the preschool classroom and gave the partici- pant the opportunity to trade tokens for edible items or trinkets or engagement with a toy or leisure activity. A participant could trade one token for 1 min to play with a toy or leisure activity, one token for one edible item to con- sume, or three tokens for one trinket to take home. Each participant could spend the num- ber of tokens he or she had for any combina- tion of the above. All participants traded all tokens at the end of a session. We used a mul- tielement design in which we rapidly alternated baseline, RC, and DR conditions to compare the effects of the different procedures on on- task behavior. Baseline. Before the start of all baseline ses-

sions, the experimenter described the rules and contingencies for the session and posted a white board on the wall in front of the participants. The experimenter stated the rules as follows: “Today it’s white, and there are no tokens. When we start, you need to sit on your mat, keep your hands to yourself, and raise your hand to talk.” During the session, the experi- menter did not provide any programmed con- sequences for any behavior, with the exception of responses to correct and incorrect responding (as mentioned above). Differential reinforcement. Before the start of

all DR sessions, the experimenter described the rules and contingencies for the session, posted a green poster board on the wall in front of the participants, and placed a green board with no tokens on the floor in front of each participant. The experimenter stated the rules as follows: “Today you get the green board, and it doesn’t have any tokens. If you stay on your mat, keep

333REINFORCEMENT AND RESPONSE COST

your hands to yourself, and raise your hand to talk, you will get a token. If you get off your mat, touch your friends, or talk during some- one else’s turn, you will not get any tokens. When small group is done, you can trade your tokens for prizes and candy. If you don’t have any tokens, you don’t get anything.” Each participant had his or her own token board. Throughout the session, the experi- menter watched a timer, and during a sched- uled observation, placed a token on the token board of any participant who was on task. The experimenter did not deliver any pro- grammed consequences for participants who were not on task. Response cost. Before the start of all RC ses-

sions, the experimenter described the rules and contingencies for the session, posted a red poster board on the wall in front of the partici- pants, and placed a red board with 10 tokens in front of each participant. The experimenter stated the rules as follows: “Today you get the red board, and it has 10 tokens. If you stay on your mat, keep your hands to yourself, and raise your hand to talk, you will keep your tokens. If you get off your mat, touch your friends, or talk during someone else’s turn, you will lose tokens. When small group is done, you can trade your tokens for prizes and candy. If you don’t have any tokens, you don’t get anything.” During the session, the experi- menter followed the variable momentary obser- vation schedule as in the DR condition; however, when a scheduled observation occurred, the experimenter did not deliver con- sequences for any participant who was on task and removed a token from any participant’s token board who was not on task. Choice. When we observed stable levels of

responding in the DR and RC phases for each participant, we conducted a preference assessment to determine the procedure that each participant preferred. We conducted this evaluation with Groups 2 and 3 only because one participant in Group 1 left the preschool

before evaluation of preference. We used a pro- cedure similar to that used by Layer et al. (2008) to evaluate preference. Before each session, the experimenter placed the stimuli (i.e., different-colored token boards and materi- als) associated with each type of condition (i.e., baseline, RC, and DR) on the floor where the experimenter conducted sessions. We presented the DR token board without tokens present and the RC token board with all tokens on the board. Near each of the token boards was a small strip of paper that matched the color of the stimuli (e.g., a green strip of paper was placed in front of the the DR token board). The experimenter called each participant to the small-group area one at a time and reminded him or her of the contingencies associated with each set of materials. Next, the experimenter asked the participant to pick which session he or she liked best by placing the colored strip of paper associated with the selected condition into a canvas bag. When the participant made a selection, he or she was asked to go play in another area of the classroom until this proce- dure was repeated with each participant. This method reduced the likelihood that a partici- pant’s choice would be influenced by other children’s prompts or comments or by obser- ving the choices of other members in the group. Although it is possible that children could have discussed their choices with a peer before his or her selection, informal observa- tions suggest that this did not occur. However, we did observe participants occasionally discuss their choices after all participants had made a selection. After all participants independently made a selection, the experimenter called them to the small-group area, drew a color from the bag, then explained the contingencies in place for the chosen session. After the experimenter had explained the contingencies for the chosen procedure, the experimenter implemented the type of session chosen as described above. We determined individual preference by counting the number of selections of each procedure; the

ERICA S. JOWETT HIRST et al.334

procedure that an individual selected most often was identified as the preferred procedure. During the choice phase, we calculated inter-

observer agreement for selection of a procedure using a total agreement method. That is, we scored an agreement if both observers agreed which procedure the participant selected and a disagreement if the two observers disagreed. Thus, interobserver agreement for selection of a procedure for a particular session was either 100% (the two observers agreed) or 0% (the two observers disagreed). Interobserver agreement for selection was 100% for all participants.

Results Figure 1 displays graphs of the percentage of

intervals of on-task behavior for all participants in Groups 1, 2, and 3 and individual cumula- tive selections and experimenter-selected proce- dures during the choice phase for Groups 2 and 3. During the initial baseline, most parti- cipants engaged in moderate to low levels of on-task behavior, although participants in Group 1 engaged in somewhat higher levels of on-task behavior. When we compared DR and RC, we observed similarly high levels of on-task behavior for six of the nine participants (93% during DR and 95% during RC) and higher levels of on-task behavior during RC for three participants (Adam, Molly, and Carl). When we evaluated preference, one participant switched his selections but selected DR more than RC (Paul), two participants switched their selections but selected RC more than DR (Judy and Molly), and three participants selected RC exclusively (Carl, Jack, and Lance). Table 1 provides a summary of results with

respect to percentage of selections during the choice phase and average net tokens yielded during the DR and RC comparison phase. We did not evaluate preference or calculate net tokens for Group 1; therefore, Table 1 includes data only for participants in Groups 2 and

3. Preference results show that one participant chose DR more than RC (Paul), and the other five participants chose RC more than DR. Also, three of six participants had an aver- age difference of at least 0.5 tokens between the two procedures, and all three participants (Molly, Carl, and Lance) preferred response cost, which was the procedure for which more net tokens were yielded.

STUDY 2: DRA VERSUS RC (INDIVIDUAL)

Method The purposes of Study 2 were twofold. The

first purpose was to replicate Study 1 by com- paring the effectiveness of and preference for DR and RC in the context of an independent work task. The second purpose was to compare responding of participants in Studies 1 and 2 to evaluate the influence of the presence of peers. Participants and setting. Thirteen typically

developing preschool-aged (3 to 5 years old) children (three of whom participated in Study 1) and one child with cerebral palsy (Brianna), who were enrolled in a university-based pre- school program, participated. All children could follow multistep instructions and communi- cated using vocal speech. We conducted ses- sions 3 to 5 days per week, once or twice per day, in session rooms that contained tables, chairs, and relevant session materials. The experimenter, one participant, and one or two data collectors were present for each session. Materials. During all sessions, we placed

worksheets with printed letters and shapes and markers on a child-sized table, and two chairs were available for the child and experimenter. In addition, we placed toys from the preschool classroom (e.g., puzzles, dolls, toy cars, coloring book, and crayons) on the floor on the opposite side of the session room. Tokens were identical to those used in Study 1. We also used different-colored token boards and poster

335REINFORCEMENT AND RESPONSE COST

025507510 0

B L

D R

v s

R C

A da

m

B L

D R

v s

R C

C ho

ic e Pa

ul 0246810

B L

D R

v s

R C

C ar

l

C ho

ic e

025507510 0

L uk

e

Ju dy

0246810

Ja ck

0 5

10 15

20 025507510 0

A nn

a 0

5 10

15 20

25 30M ol

ly

5 10

15 20

25 30

0246810

L an

ce

G ro

u p

1 G

ro u

p 2

G ro

u p

3 % Intervals (On task)

S es

si on

s

Cumulative Selections

F ig u re

1. P er ce n ta ge

of on -t as k be h av io r fo r A da m , L u ke , an d A n n a (G

ro u p 1) ; P au l, Ju dy , an d M ol ly

(G ro u p 2) ; an d C ar l, Ja ck , an d L an ce

(G ro u p 3)

du ri n g

R C

(fi lle d ci rc le s) an d D R (fi lle d tr ia n gl es ), ba se lin

e (o pe n sq u ar es ), an d cu m u la ti ve

se le ct io n s (o pe n ci rc le s fo r R C

se le ct io n s an d op en

tr ia n gl es

fo r D R se le ct io n s) .

T h e fi lle d da ta po in ts gr ap h ed

al on g th e le ft y ax is du

ri n g th e ch oi ce

ph as e (G

ro u ps

2 an d 3)

re pr es en t pe rc en ta ge

of in te rv al s on

ta sk

du ri n g th e co n di ti on

th e ex pe ri -

m en te r se le ct ed .

ERICA S. JOWETT HIRST et al.336

boards to aid in the discrimination between the conditions as in Study 1. Furthermore, partici- pants earned access to the same toy room used in Study 1 after some sessions; however, some of the toys changed over time. Response measurement and interobserver agree-

ment. Trained graduate and undergraduate stu- dents collected data using handheld computers. The dependent variable during all sessions was percentage of intervals of on-task behavior. We defined on-task behavior as the first instance of walking to the work table, the first instance of removing the lid of the marker, moving the marker approximately within the boundaries of the printed lines of a worksheet, and turning over pages to access a new worksheet. We did not score on-task behavior if the participant was scribbling or drawing pictures on the work- sheet or making patterns (e.g., dashed lines or dots) within the printed boundaries of the let- ters or shapes. We partitioned sessions into 5-s intervals and scored on-task behavior using partial-interval recording. That is, we scored on-task behavior if it occurred during any por- tion of the 5-s interval. Next, we converted data to a percentage by dividing the number of intervals during which the child was on task by the total number of intervals in the session. We also collected data on the frequency of token delivery (i.e., when the experimenter placed a token on the token board) and token removal (i.e., when the experimenter removed a token from the token board).

We calculated interobserver agreement for on-task behavior as in Study 1 and calculated interobserver agreement coefficients for token delivery or removal by dividing the session time into 5-s intervals and comparing observer data on an interval-by-interval basis. If exact agree- ment occurred (i.e., both observers scored or did not score a token delivery or removal within a 5-s interval), we gave a score of 1 for that interval. For any disagreements, we divided the smaller score in each interval by the larger. We then summed interval scores, divided them by the total number of observation intervals, and converted the result to a percentage. Inter- observer agreement for on-task behavior was 93% (range, 73% to 100%) and for token delivery or removal it was 96% (range, 78% to 100%). Design. We used a multielement design for

10 participants to compare the effects of the different procedures on on-task behavior, and we conducted sessions in a quasirandom order. In addition, for two of these participants, we used a reversal design following the multiele- ment design to rule out discrimination failure or carryover effects during the multielement comparison. However, because we conducted the reversal designs after the participants had a history of both procedures, we used a reversal design with four participants to determine levels of responding during DRA before and after a history of RC. Procedure. All sessions lasted 5 min. Before

the first session of each condition, the experi- menter described the session contingencies and required the participant to practice engaging in related behaviors (i.e., tracing or playing with toys) to experience the consequences associated with each behavior, as in Study 1. For example, the experimenter required the participant to practice tracing by providing a vocal and model prompt (i.e., “Try tracing like this,” while demonstrating tracing), and used physical guid- ance as necessary. After the participant prac- ticed tracing, the experimenter provided the

Table 1 Percentage of Selections and Average Net Tokens Yielded

for Participants in Study 1 (Group Analysis)

% selections Average net tokens

Participant Group DR RC DR RC

Paul 2 67 33 9.8 9.9 Molly 2 22 78 8.5 9.4 Judy 2 11 89 9.4 9.0 Carl 3 0 100 7.3 9.1 Jack 3 0 100 9.1 9.1 Lance 3 0 100 8.9 9.6

337REINFORCEMENT AND RESPONSE COST

relevant consequences and repeated the contin- gency for that particular phase (e.g., “Look, you got a token because you were tracing.”). Before the start of each subsequent session dur- ing a particular phase, the experimenter described the session contingencies (see condi- tion descriptions below). First, we conducted baseline sessions to

determine the level of on-task behavior in the absence of programmed consequences. Next, the experimenter practiced token trading with the participant, as in Study 1. During DRA and RC sessions, the experimenter deliv- ered or removed tokens according to the same variable momentary schedule used in Study 1; however, the experimenter conducted observations on a fixed 30-s schedule for four participants (Brianna, Mark, Zoey, and Sam), who participated later in the study, to simplify data collection. In addition, after each DRA and RC session, participants traded tokens for prizes, candy, and access to leisure items. Baseline. Before the start of all baseline ses-

sions, the experimenter described the rules and contingencies for the session and placed a white board with no tokens near the participant. The experimenter stated the rules as follows: “Today you get the white board, and there are no tokens. When we start, you can either work on tracing or play with toys. If you are working (i.e., tracing), nothing will happen, if you are not working, nothing will happen.” During the session, the experimenter did not provide programmed consequences for any behavior. Differential reinforcement of alternative behav-

ior. Before the start of all DRA sessions, the experimenter described the rules and contin- gencies for the session and placed a green board with no tokens near the participant. The exper- imenter stated the rules as follows: “Today you get the green board, and it doesn’t have any tokens on it. When we start, you can either work on tracing or play with toys. If you are

working, you will get a token; if you are not working, you will not get a token. At the end, you can trade your tokens for prizes and snacks. If you don’t have any tokens, you don’t get anything.” Throughout the session, the experimenter watched a timer. If the partici- pant was on task at the time of a scheduled observation, the experimenter placed a token on the token board. If the participant was not on task at the time of the scheduled observa- tion, the experimenter did not provide any pro- grammed consequences. Response cost. Before the start of all RC ses-

sions, the experimenter described the rules and contingencies for the session and placed a red board with 10 tokens near the participant. The experimenter stated the rules as follows: “Today you get the red board, and it has 10 tokens on it. When we start, you can either work on trac- ing or play with toys. If you are working, you will keep your tokens; if you are not working, you will lose tokens. At the end, you can trade your tokens for prizes and snacks. If you don’t have any tokens, you don’t get anything.” Throughout the session, the experimenter watched a timer. If the participant was on task at the time of a scheduled observation, the experimenter did not provide any programmed consequences. If the participant was not on task at the time of a scheduled observation, the experimenter removed a token from the token board. Choice. When we observed stable levels of

responding in the DRA and RC evaluations, we conducted a preference assessment to deter- mine the procedure that each participant pre- ferred. Before each session, the experimenter placed the stimuli (i.e., poster and token boards) associated with each type of condition (i.e., baseline, RC, and DRA) near the partici- pant and reminded him or her of the contin- gencies associated with each set of materials. For example, the experimenter reminded the participant that the white board means that there are no tokens; the green board means that

ERICA S. JOWETT HIRST et al.338

he or she can earn tokens if he or she is tracing; and the red board means that he or she could keep his or her tokens if he or she is tracing. The experimenter switched the placement of the different sets of stimuli and materials each session. After the experimenter reminded the participant of the contingencies associated with each set of materials, the experimenter asked the participant to pick (by pointing to or touching a set of materials) which session he or she wanted to do. When the participant made the selection, the experimenter explained the contingencies in place for the session (e.g., “You picked green, you will get a token when I see that you are working on tracing.”). After the participant chose a procedure, the experi- menter implemented the chosen type of session as described above. The experimenter con- ducted sessions until we observed a stable pat- tern of selections. During the choice phase, we calculated interobserver agreement as in Study 1; it was 100% for all participants.

Results Figure 2 shows the results for 10 of the

14 participants. During the initial baseline, all participants engaged in moderate to low levels of on-task behavior, and these levels remained low throughout the evaluation (with the excep- tion of Adam, Frank, and Martin, who engaged in variable levels of on-task behavior during baseline). When we compared DRA and RC using a multielement design, we observed (a) similar levels of on-task behavior for eight of the 10 participants (average of 88% during DRA and 85% during RC), (b) higher levels of on-task behavior during DRA for one partici- pant (Emily; 94% during DRA and 82% dur- ing RC), and (c) higher levels of on-task behavior during RC for one participant (Adam; 47% during DRA and 65% during RC). When we compared DRA and RC using a reversal design for two participants (Anna and Caro- line), we observed similar and high levels of on-

task behavior as during the multielement evalu- ation. When we evaluated preference, two par- ticipants selected DRA exclusively (Paul and Frank), three participants switched their selec- tions but selected DRA more than RC (Martin, Emily, and Adrianna), three participants switched their selections but selected RC more than DRA (Elisa, Adam, and Anna), and two participants selected RC exclusively (Collin and Caroline). Figure 3 shows the results for Brianna,

Mark, Zoey, and Sam. During baseline ses- sions, all participants engaged in low to zero levels of on-task behavior. When we compared DRA and RC using a reversal design only, we observed similar and high levels of on-task behavior for three of the four participants (Brianna, Mark, and Zoey); however, we observed higher levels of on-task behavior dur- ing RC for one participant (Sam; 62% during DRA and 90% during RC). These data suggest that a history of response cost is not likely to influence responding during DRA. Table 2 provides a summary of results from

Study 2 with respect to the percentage of selec- tions in the choice phase and the net tokens yielded for participants during the DRA and RC comparison phases. We evaluated prefer- ence for 10 of the 14 participants and calcu- lated net tokens for all participants. Preference results show that five participants chose DR more than RC and five chose RC more than DR. Although these results are similar to those of previous studies (e.g., Donaldson et al., 2014; Iwata & Bailey, 1974), these results were somewhat different than those of Study 1. That is, the majority of participants preferred RC in Study 1, but only half of the participants pre- ferred RC in Study 2. Also, five of the 10 parti- cipants in Study 2 for which we also assessed preference had an average difference of at least 0.5 tokens between the two procedures, and four of these five participants (Frank, Paul, Adam, and Anna) preferred the procedure that yielded more net tokens.

339REINFORCEMENT AND RESPONSE COST

GENERAL DISCUSSION

Overall, DR and RC were effective proce- dures for increasing the on-task behavior of the majority of children who participated in a group activity (Study 1), and these findings replicated those of previous research (e.g.,

Donaldson et al., 2014; Iwata & Bailey, 1974). However, similar to Donaldson et al. (2014) and Tanol, Johnson, McComas, and Cote (2010), the procedures were differentially effec- tive for some individuals in the group, which suggests that analyzing individual data is

5 10 15 20 25 30 35 40 45 50

0

20

40

60

80

100 BL

Paul

DRA vs RC Choice

5 10 15 20 25 30 35 40

0

20

40

60

80

100 BL

Elisa

DRA vs RC Choice

5 10 15 20 25 30

0

20

40

60

80

100 BL

Frank

ChoiceDRA vs RC

10 20 30 40 50 60 70

0

20

40

60

80

100 BL

Adam

DRA vs RC Choice

5 10 15 20 25

0

20

40

60

80

100 BL

Martin

DRA vs RC Choice

5 10 15 20 25 30 35 40

0

20

40

60

80

100 BL

Collin

DRA vs RC Choice

5 10 15 20 25 30 35 40 45

0

20

40

60

80

100 BL

Emily

DRA vs RC Choice

10 20 30 40 50 60 70 80

0

20

40

60

80

100

BL

Anna

DRA vs RC Choice RC RC D D

5 10 15 20 25 30

0

20

40

60

80

100 BL

Adrianna

DRA vs RC Choice

10 20 30 40 50 60 70 80 90 100 0

20

40

60

80

100

Caroline

BL

DRA vs RC Choice D D DRA RC RC

% I

n te

rv al

s (O

n t

as k

)

Sessions

Figure 2. Percentage of on-task behavior for Paul, Frank, Martin, Emily, Adrianna, Elisa, Adam, Collin, Anna, and Caroline during RC, DRA (also denoted as D during the short reversal phases for Anna and Caroline), and baseline in the comparative analysis and choice phases. The symbol used for each data point during the choice phase represents the condition selected by the participant for that session.

ERICA S. JOWETT HIRST et al.340

important because these differences may not have been observed if we reported only group averages. The importance of analyzing individ- ual data is further supported by the results of Study 2, which showed differential effects for three participants (Adam, Emily, and Sam), whereas the overall results suggest that the two procedures are equally effective. Several variables might have influenced

results of the current study, including the type of contingency used (individual vs. group oriented) and the experimental design. Results showed that the comparative effectiveness of the procedures was the same for all three

participants who participated in Studies 1 and 2 (Adam, Anna, and Paul). That is, RC was more effective than DR for Adam during the group activity and solitary work task, and the procedures were equally effective for Anna and Paul under both conditions. These results sug- gest that the presence of peers did not influ- ence the comparative effectiveness of DR and RC. However, an analysis of the results for Adam and Anna shows that these participants engaged in 10% to 20% higher levels of on- task behavior during the group evaluation than in the individual evaluation. These results ten- tatively suggest that the presence of peers may enhance the effectiveness of the procedures for some children. Because both procedures resulted in equally higher levels of responding in the presence of peers, it could be that obser- ving a peer receiving a token increases the value of the token or functions as a discrimina- tive stimulus for on-task behavior (during DR conditions). In addition, the aversiveness of token loss might also be enhanced when tokens are removed in the presence of peers (during RC). Although the relative efficacy of DR and RC

was not influenced by the use of group- oriented contingencies, the overall effectiveness of the procedures was greater during the group activity. These higher levels of on-task behavior during the group activity may have been due to the differential effort or task difficulty across tasks in the group activity and individual activ- ity (i.e., it may have been more effortful to trace letters than to keep one’s hands in one’s lap and sit on the mat). In addition, higher levels of on-task behavior in the group activity may have been due to the absence of a salient alternative task, as was provided in the individ- ual activity (i.e., toys were available). However, there were many alternative tasks available dur- ing the group activity, such as playing with or manipulating the bingo boards and pieces and leaving the mat to join other activities in the classroom.

2 4 6 8 10 12 0

20

40

60

80

100

% I

nt er

va ls

( O

n T

as k)

Brianna

DRA RC

BL

DRA

2 4 6 8 10 12 0

20

40

60

80

100

% I

nt er

va ls

( O

n T

as k)

Mark

DRA RC

BL

DRA

2 4 6 8 10 12 0

20

40

60

80

100

% I

nt er

va ls

( O

n T

as k)

Zoey

DRA

BL

RC DRA

5 10 15 20 25 30 0

20

40

60

80

100

% I

nt er

va ls

( O

n T

as k)

BL

Sam

DRA RC DRA

Sessions

RC

Figure 3. Percentage of on-task behavior for Brianna, Mark, Zoey, and Sam during RC, DRA, and baseline.

341REINFORCEMENT AND RESPONSE COST

We used a multielement design in Study 1 and for 10 participants in Study 2. Thus, similar effects observed across DR and RC may have been due to multiple-treatment interfer- ence because of the rapid alternation of condi- tions that were similar in numerous respects. Although we attempted to control for multiple- treatment interference by including session rules and discriminative stimuli, we also attempted to address this concern by evaluating the effects when a different design was used. For two participants in Study 2 (Anna and Car- oline), in which we used both a multielement design and a reversal design to compare the effects of DR and RC, we found similar results regardless of which design was used. In addi- tion, for four participants in Study 2 (Brianna, Mark, Zoey, and Sam), in which we used only a reversal design to compare DR and RC, we showed similar levels of on-task behavior across the two procedures as well as similar levels of on-task behavior regardless of whether DR was conducted before or after RC. These data sug- gest that the use of a multielement design was unlikely to influence the results. With respect to preference, five of the 15 par-

ticipants in the choice evaluation preferred DR,

and the other 10 participants preferred RC. As suggested in previous research (e.g., Donaldson et al., 2014), several variables may have influ- enced preference for the different procedures. Participants may select the reinforcement pro- cedure to avoid the loss condition, as observed by Pietras, Brandt, and Searcy (2010), who found that when they equated net tokens, par- ticipants avoided the procedure that involved token loss. In addition, participants may prefer reinforcement, specifically when reinforcer delivery is spaced evenly throughout the ses- sion, because token delivery signals time pro- gression through the session. That is, token delivery provides feedback regarding the dura- tion of the session, which may be valuable, especially with young children. With respect to preference for RC, the

potential aversion associated with RC may have been eliminated because participants did not contact loss often; as Donaldson et al. (2014) noted, one participant mentioned preference for RC due to losing few tokens. However, additional variables also warrant consideration. First, some participants may have preferred RC because selection of the RC procedure results in the delivery of all tokens; therefore, access to all tokens may function as a reinforcer for selec- tion of that procedure. In addition, selection of RC over DR may be because, from the child’s perspective, starting with tokens is viewed as not having to work for the tokens. That is, the procedure appears to be less effortful. To rule out influence of the presence of tokens, future researchers might evaluate preference under conditions in which the tokens are present for DR and RC (i.e., a cup of tokens next to the DRA token board and tokens attached to the RC board) or the tokens are not present (i.e., placing colored strips of paper representing each procedure or asking the participant which procedure he or she would like to do). Other variables that might influence prefer-

ence in the current study are the consequences that followed selection of a particular condition

Table 2 Percentage of Selections and Average Net Tokens Yielded

for Participants in Study 2 (Individual Analysis)

% selections Average net tokens

Participant DR RC DR RC

Frank 100 0 8.9 8.3 Paul 100 0 9.6 9.1 Martin 82 18 9.3 9.3 Adrianna 75 25 9.6 9.7 Emily 67 33 8.5 8.2 Adam 28 72 4.1 5.3 Elisa 21 79 8.4 8.2 Anna 18 82 7.6 8.8 Collin 0 100 9.8 9.1 Caroline 0 100 5.7 5.4 Brianna 8.7 8.7 Mark 9.4 9.6 Zoey 9.4 8.7 Sam 6.1 8.8

ERICA S. JOWETT HIRST et al.342

(DRA vs. RC) and the net tokens earned within a particular condition. Participants in the group evaluation may have chosen a differ- ent procedure the next time they were offered a choice if the experimenter did not implement the procedure they had chosen in a given ses- sion. However, an evaluation of data for parti- cipants in Study 2 showed that participants switched their selection during subsequent choice opportunities when the session that the experimenter implemented after a selection did not match the initial selection on 38% (Paul), 38% (Molly), and 50% (Judy) of selections. These results suggest that switches in selections were not influenced by whether the session that was implemented matched the procedure they had selected, and these findings are consistent with those of Layer et al. (2008). Previous researchers have evaluated the

potential influence of net tokens across DR and RC conditions. Iwata and Bailey (1974) calcu- lated the average number of net tokens for the class, and Donaldson et al. (2014) calculated individual net token averages; both studies found that net tokens were similar across proce- dures. Although the number of net tokens was similar, because some participants preferred one procedure over another, it could be that even slight differences may influence preference. In the current study, we were able to evaluate preference for 15 participants (twice with Paul) and found that seven of the 14 children who participated once (and Paul on one occasion in Study 2) yielded an average difference of at least 0.5 tokens between the two procedures. Of these eight participants, seven preferred the procedure for which more net tokens were yielded during the comparison phase. However, in previous research and in the current study, experimenters did not manipulate the number of net tokens. Therefore, the influence of net tokens on preference is unknown, and research on this variable is warranted. Another point of discussion relates to

best practice guidelines. The general

recommendation is to use reinforcement-based procedures when possible (Bailey & Burch, 2005). Therefore, because RC is a negative punishment procedure (Kazdin, 1977), RC often is not recommended before implementa- tion of positive reinforcement procedures. However, given that (a) RC is just as effective as reinforcement, (b) RC has limited side effects (Kazdin, 1972), (c) more participants preferred RC in the current study, and (d) previous researchers have also found prefer- ence for punishment procedures (e.g., Hanley, Piazza, Fisher, & Maglieri, 2005), reconsidera- tion of best practice appears to be warranted. Perhaps the use of effective and preferred pro- cedures should be considered best practice (e.g., Hanley, 2010). There are several areas for future research.

First, we were able to compare responding of only three individuals who participated in both the group activity and solitary work task; there- fore, our conclusions about the effects of peer presence are limited, and future researchers should consider conducting this evaluation with a larger number of participants. Second, because we conducted both preference evalua- tions in Studies 1 and 2 in the absence of peers, we were unable to compare choice in the pres- ence versus absence of peers. Third, we did not collect data on side effects

of the procedures, which may be important, specifically with the possibility of negative side effects (e.g., emotional responding or increases in problem behavior) when RC procedures are used. However, little to no negative side effects have been reported during the use of RC proce- dures (Conyers et al., 2004; Kazdin, 1972) nor were negative side effects observed in the cur- rent study. Fourth, future researchers should include a

measure of accuracy. In the current study, we selected on-task behavior because it was age appropriate, but we did not measure the accu- racy of responding. Iwata and Bailey (1974) showed decreases in rule violations without

343REINFORCEMENT AND RESPONSE COST

increasing correct responding. Because on-task behavior is a prerequisite for accurate respond- ing in many situations, correct responding should increase as children are attending; there- fore, future researchers should measure changes in accuracy when reinforcement and punish- ment contingencies are in effect for on-task behavior. Fifth, we arranged individual contingencies,

rather than interdependent group-oriented con- tingencies or dependent group-oriented contin- gencies. Individual and interdependent group- oriented contingencies require that the teacher monitor the behavior of each child and then deliver consequences based on the behavior of each child individually or for the behavior of the group, respectively; on the other hand, a dependent group-oriented contingency requires that a teacher monitor the behavior of only one child in the group. Herman and Tramontana (1971) found no difference in the effectiveness of individual and group contingencies and sug- gested that group contingencies may be easier for teachers. Therefore, future researchers should compare DR and RC using dependent and interdependent group-oriented contingen- cies (see Litow & Pumroy, 1975, for a brief review of group contingencies). Finally, because we associated specific colors

with the different procedures, children’s choices for procedures may have been based on prefer- ence for color rather than procedure. However, anecdotal reports do not suggest that partici- pants had strong preferences for colors (i.e., it was not common for participants to report color preference during the choice evaluation). Future researchers might control for the influ- ence of color preferences by using low or mod- erately preferred colors for the stimuli used for the DR and RC procedures (e.g., Luczynski & Hanley, 2009) or changing the colors associ- ated with the procedures throughout the study. In summary, there are several important

implications of the current study. First, the results suggest that both DR and RC are

similarly effective; therefore, teachers might use the procedure that more children prefer or that is easier to implement in a classroom setting. Second, the presence of peers does not appear to influence the relative efficacy of the proce- dures; therefore, future researchers might con- tinue to conduct comparisons of DR and RC in group settings for more efficient data collec- tion. Finally, considerations for best practice should take into account preference, given the large number of participants who preferred RC.

REFERENCES

Bailey, J. S., & Burch, M. R. (2005). Ethics for behavior analysts. Mahwah, NJ: Erlbaum. doi: 10.4324/ 9781410613738

Brent, D. E., & Routh, D. K. (1978). Response cost and impulsive word recognition errors in reading-disabled children. Journal of Abnormal Child Psychology, 6, 211–219. doi: 10.1007/bf00919126

Broughton, S. F., & Lahey, B. B. (1978). Direct and col- lateral effects of positive reinforcement, response cost, and mixed contingencies for academic performance. Journal of School Psychology, 16, 126–136. doi: 10.1016/0022-4405(78)90051-1

Capriotti, M. R., Brandt, B. C., Ricketts, E. J., Espil, F. M., & Woods, D. W. (2012). Comparing the effects of differential reinforcement of other behavior and response-cost contingencies on tics in youth with Tourette syndrome. Journal of Applied Behavior Analysis, 45, 251–263. doi: 10.1901/ jaba.2012.45-251

Conyers, C., Miltenberger, R., Maki, A., Barenz, R., Jurgens, M., Sailer, A., … Kopp, B. (2004). A com- parison of response cost and differential reinforce- ment of other behavior to reduce disruptive behavior in a preschool classroom. Journal of Applied Behavior Analysis, 37, 411–415. doi: 10.1901/ jaba.2004.37-411

Doll, C., McLaughlin, T. F., & Barretto, A. (2013). The token economy: A recent review and evaluation. International Journal of Basic and Applied Science, 2, 131–149.

Donaldson, J. M., DeLeon, I. G., Fisher, A. B., & Kahng, S. (2014). Effects of and preference for condi- tions of token earn versus token loss. Journal of Applied Behavior Analysis, 47, 537–548. doi: 10.1002/jaba.135

Hackenberg, T. D. (2009). Token reinforcement: A review and analysis. Journal of the Experimental Analy- sis of Behavior, 91, 257–286. doi: 10.1901/ jeab.2009.91-257

ERICA S. JOWETT HIRST et al.344

Hanley, G. P. (2010). Toward effective and preferred pro- gramming: A case for the objective measurement of social validity with recipients of behavior-change pro- grams. Behavior Analysis in Practice, 3, 13–21.

Hanley, G. P., Piazza, C. C., Fisher, W. W., & Maglieri, K. A. (2005). On the effectiveness of and preference for punishment and extinction compo- nents of function-based interventions. Journal of Applied Behavior Analysis, 38, 51–65. doi: 10.1901/ jaba.2005.6-04

Herman, S. H., & Tramontana, J. (1971). Instructions and group versus individual reinforcement in modify- ing disruptive group behavior. Journal of Applied Behavior Analysis, 4, 113–119. doi: 10.1901/ jaba.1971.4-113

Iwata, B. A., & Bailey, J. S. (1974). Reward versus cost token systems: An analysis of the effects on students and teacher. Journal of Applied Behavior Analysis, 7, 567–576. doi: 10.1901/jaba.1974.7-567

Kazdin, A. E. (1972). Response cost: The removal of con- ditioned reinforcers for therapeutic change. Behavior Therapy, 3, 533–546. doi: 10.1016/S0005-7894(72) 80001-7

Kazdin, A. E. (1977). The token economy: A review and evaluation. New York, NY: Plenum Press. doi: 10.1007/978-1-4613-4121-5

Layer, S. A., Hanley, G. P., Heal, N. A., & Tiger, J. H. (2008). Determining individual preschoolers’ prefer- ences in a group arrangement. Journal of Applied Behavior Analysis, 41, 25–37. doi: 10.1901/ jaba.2008.41-25

Litow, L., & Pumroy, D. K. (1975). A brief review of classroom group-oriented contingencies. Journal of Applied Behavior Analysis, 8, 341–347. doi: 10.1901/ jaba.1975.8-341

Luczynski, K. C., & Hanley, G. P. (2009). Do children prefer contingencies? An evaluation of the efficacy of and preference for contingent versus noncontingent social reinforcement during play. Journal of Applied

Behavior Analysis, 42, 511–525. doi: 10.1901/ jaba.2009.42-511

McGoey, K. E., & DuPaul, G. J. (2000). Token rein- forcement and response cost procedures: Reducing the disruptive behavior of preschool children with attention-deficit/hyperactivity disorder. School Psychol- ogy Quarterly, 15, 330–343. doi: 10.1037/h0088790

Panek, D. M. (1970). Word association learning by chronic schizophrenics on a token economy ward under conditions of reward and punishment. Journal of Clinical Psychology, 26, 163–167. doi: 10.1002/1097-4679(197004)26:2<163::aid-jclp2270 260208>3.0.co;2–5

Pietras, C. J., Brandt, A. E., & Searcy, G. D. (2010). Human responding on random-interval schedules of response-cost punishment: The role of reduced rein- forcement density. Journal of the Experimental Analysis of Behavior, 93, 5–26. doi: 10.1901/jeab.2010.93-5

Salend, S. J., & Kovalich, B. (1981). A group response- cost system mediated by free tokens: An alternative to token reinforcement. American Journal of Mental Deficiency, 86, 184–187.

Sindelar, P. T., Honsaker, M. S., & Jenkins, J. R. (1982). Response cost and reinforcement contingencies of managing the behavior of distractible children in tutorial settings. Learning Disability Quarterly, 5, 3–13. doi: 10.2307/1510610

Tanol, G., Johnson, L., McComas, J., & Cote, E. (2010). Responding to rule violations or rule following: A comparison of two versions of the Good Behavior Game with kindergarten students. Journal of School Psychology, 48, 337–355. doi: 10.1016/j. jsp.2010.06.001

Received December 2, 2014 Final acceptance October 8, 2015 Action Editor, Jeanne Donaldson

345REINFORCEMENT AND RESPONSE COST