PREFERENCES AMONG ARRAY SIZES FOR BACK UP REINFORCERS: By NATHALIE FERNANDEZ A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2019
2019 Nathalie Fernandez
3 ACKNOWLEDGMENTS I would like to thank my advisor, Dr. Iser DeLeon, for his support and admirable levels of patience throughout this and previous studies I would also like to thank the other members of my committee Dr. Tim Vollmer and Dr. Nicolas Gage for their feedback I would like to thank Izzy Schieber for her assistance with data collection. Additionally, I would like to thank my colleagues in the DeLeon lab and the staff at the Behavior Analysis Support S ervices who helped me throughout this project. Finally, I wo uld like to thank my family for their unwavering support, even though they do not understand what my research is about.
4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 3 LIST OF TABLES ................................ ................................ ................................ ........................... 5 LIST OF FIGURES ................................ ................................ ................................ ......................... 6 ABSTRACT ................................ ................................ ................................ ................................ ..... 7 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................... 8 2 EXPERIMENT 1 ................................ ................................ ................................ .................... 16 Participants, Settings, and Materials ................................ ................................ ....................... 16 Response Measurement and Interobserver Agreement ................................ .......................... 17 Leisure Item Preference Assessment ................................ ................................ ...................... 18 Discrimination Probe ................................ ................................ ................................ .............. 18 Procedures ................................ ................................ ................................ ............................... 18 Results and Discussion ................................ ................................ ................................ ........... 22 3 EXPERIMENT 2 ................................ ................................ ................................ .................... 24 Participants, Settings, and Materials ................................ ................................ ....................... 24 Response Measurement and Interobserver Agreement ................................ .......................... 24 Token Training ................................ ................................ ................................ ....................... 25 Procedures ................................ ................................ ................................ ............................... 27 Results and Discussion ................................ ................................ ................................ ........... 28 4 GENERAL DISCUSSION ................................ ................................ ................................ ..... 31 APPENDIX TABLES AND FIGURES ................................ ................................ ........................ 35 LIST OF REFERENCES ................................ ................................ ................................ ............... 42 BIOGRAPHICAL S KETCH ................................ ................................ ................................ ......... 46
5 LIST OF TABLES Table page A 1 Participant names and stimuli identified as high and low preference. ............................... 35 A 2 Token production and exchange production schedules across participants in Experiment 2 ................................ ................................ ................................ ..................... 36
6 LIST OF FIGURES Figure page A 1 Array size preference assessment across presentation formats for Andre. ........................ 37 A 2 Array size preference assessment across presentation formats for Kim. ........................... 38 A 3 Array size preference assessment across presentation formats for Tyler. ......................... 39 A 4 Array size preference assessment across presentation formats for Lauren. ....................... 40 A 5 Comparison of preference for array size across no work and work conditions for all participants. ................................ ................................ ................................ ........................ 41
7 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science PREFERENCES AMONG ARRAY SIZES FOR BACK UP REINFORCERS: N ECONOMIES By Nathalie Fernandez August 2019 Chair: Iser G. DeLeon Major: Psychology Researchers recommended that clinicians establish tokens as generalized conditioned reinforcers when working with individuals with autism spectrum disorder ( ASD ). During token exchanges, clinicians may present individuals with a token store, which displays an array of available backup reinforcers. Prior research with typically developing populations has reported that large array sizes, in conjunction with the manner in which arrays are presented, can produce oidance, post choice preference reversals, and other negative collateral effects. However, the preference for array sizes and the format in which arrays are presented has not been evaluated with individuals with ASD Additionally, it is unknown if preferen ces for array size within the context of a token economy is consistent with preferences displayed during a selection based preference assessment. We sought to examine preference for array size across four presentation formats and when a work requirement pr eceded array selection. Results indicated that, overall, participants selected the large array condition most often across all presentation formats. Furthermore, two of three participants selected the large array condition most often, regardless of whether a work requirement proceeded array selection.
8 CHAPTER 1 INTRODUCTION Token systems are frequently implemented when the immediate delivery of reinforcers, such as edible and tangible items, is not feasible (Kazdin & Bootzin, 1972). Graff and Karsten (2012) found that tokens are the second most delivered consequence in the treatment of individuals with intellectual and developmental disabilities (IDD). Despite their ubiquity, clinicians may encounter several barriers to establishing an effective token economy in applied settings, one of which is that the effectiveness of terminal reinforcers, or backup reinforcers, may vary across time (Hine, Ardoin, & Call 2017). One solution to this barrier is to establish tokens as generalized conditioned reinforcers by pairing them with a variety of backup reinforcers. Hackenberg (2018) described tokens as conditioned reinforce rs, the value of which is established through a relationship with backup reinforcers. However, the effectiveness of backup reinforcers can be affected by shifts in motivational states, which can also decrease the reinforcing efficacy of tokens. Thus, o ne o f the advantages frequently attributed to generalized tokens is that they maintain their effectiveness as a reinforcer despite fluctuations in motivational states (Kazdin & Bootzin, 1972; Moher, Gould, Hegg, & Mahoney, 2008; DeFulio, Yankelevitz, Bullock, & Hackenberg, 2014). For example, Moher et al. demonstrated that responding for commodity specific tokens, or nongeneralized tokens, varied greatly across states of deprivation and satiation for the backup reinforcer in individuals with ASD Specifically, responding occurred only under states of deprivation. However, the authors showed that levels of responding for generalized tokens remained elevated, independent of motivational states. Thus, the reinforcer efficacy of generalized tokens is resistant to th e effects of satiation. These results suggest that, ideally, individuals with ASD should earn tokens paired with a variety of backup reinforcers.
9 Another potential advantage of generalized tokens is that they create an opportunity for individuals to choose among backup reinforcers. Researchers have demonstrated that choice may enhance the effectiveness of interventions ( Powell & Nelson, 1997; Dyer, Dunlap, & Winterling, 1990, Kern, Mantegna, Vorndran, Bailin, & Hilt, 2001; Vaughn & Horner, 1997; Rispoli, La ng, Neely, Camargo, Hutchins, Davenport, & Goodwyn, 2013 ), and that choice making opportunities are preferred to conditions in which an experimenter or clinician selects the reinforcer (Tiger, Gregory, & Hanley, 2006; Sran & Borrero, 2010). For example, Sr an and Borrero (2010) assessed whether the opportunity to choose among backup reinforcers within a token economy affected rates of responding in four typically developing children. During reinforcement phases, experimenters delivered tokens contingent on a cademic task completion. During token exchanges, participants either (a) received one type of edible reinforcer, (b) selected from an array of five identical edible reinforcers, or (c) selected from five different edible reinforcers. The experimenters also conducted a concurrent chains evaluation to determine preference among token exchange conditions in three participants, meaning participants were able to choose to work for any of the three alternatives. Although all three token exchange conditions produc ed similar rates of responding for three of the four participants, all three participants exposed to the concurrent chains evaluation preferred to choose from a variety of backup reinforcers. Despite the advantages of providing choice opportunities within the context of behavioral interventions, the number of choice options presented to individuals with ASD may need to be carefully considered. Research with typically developing adults indicates that an excessive number of choice options may produce a pheno menon termed choice overload. Choice overload occurs when a choice array contains too many options such that it produces negative collateral
10 effects including reluctance to choose, choice dissatisfaction and decision regret (Schwartz, 2004a; 2004b). For e xample, Iyengar and Lepper (2000) examined the effects of limited and extensive arrays on consumer purchases and satisfaction. The experimenters presented grocery store customers an array of six or twenty four different flavored jams to sample. The results suggested that customers who experienced the extensive array were less likely to purchase jam (i.e., less likely to make a choice) and were also less likely to report satisfaction with their purchase. Behavior analysts have also recently begun to consider the potential for choice overload effects. Reed, DiGennaro Reed, Chok, and Brozyna (2011) examined preference for array size by asking human service workers to consider a hypothetical scenario in which they would have to find a suitable treatment facility for a student. Participants selected from (a) single option array, (b) a limited option array containing two options, and (c) an extensive option array, in which available options were increased across trials from three options to 384 options. Reed et al. found that workers were more likely to select the limited option and extensive option arrays than the single option array, but that preference for the extensive option array decreased as the number of available options increased. That is, when the number of options become increasingly larger, the participants were more likely to relinquish the opportunity to choose. Having too many options seemingly made choosing aversive. Following this line of reasoning, Reed and Tiger (2015) suggested that overly lar ge array sizes might similarly impact willingness to choose, or the quality of choices, in persons with intellectual disabilities. The authors further suggested that the number of options presented m aking repertoire and that children with autism may be more likely to benefit from smaller arrays sizes than typically developing
11 children. However, researchers have yet to examine how the number of options influences during token exchanges. Perhaps for individuals with ASD there is a point at which there are too many options in the token store, similarly to what Reed et al. (2011) observed. For tokens to function as reinforcers, the token exchange context should not be aversive to the individual; therefore, preference for the size of arrays displaying backup reinforcers should be assessed. Another area that requires consideration is the way in which clinicians present the available backup reinforcers. In applied setti ngs, clinicians frequently present these arrays in token stores (Kazdin, 1977). Token stores can take the form of a physical store, like a canteen, or a menu. In a recent meta analysis, Chernev, Bckenholt, and Goodman (2015), found that four antecedent va riables moderated the effect of large array sizes on choice overload. These y (e.g., absence of a highly preferred option), and the difficulty of the decision task (e.g., presentation format). Of these moderating variables, the effects of presentation format on choice overload seems of particular relevance in how one should arran ge token stores for individuals with ASD Researchers have demonstrated that the structure of arrays, in terms of organization and assortment variety, may impact the effects of array size on choice overload (see Chernev, 2012 for review). Results from Hoch, Bradlow, and Wansink (1999) indicated th at consumers were less likely to prefer choice arrays that presented options in a random fashion rather than in an organized fashion. The authors reported that disorderliness within an array likely increased search costs and, thus, increased the difficulty of making a selection. In a series of experiments, Kahn and Wansink (2004) investigated the effects of array organization and assortment variety
12 on the consumption of edible and leisure items with elementary school children. Participants selected as many items from the presented array as they wanted to take. Consumption was defined as the selection of an item. In one experiment, the authors assessed the interaction between array size (i.e., 6 vs. 24 items) and array organization (i.e., orderly vs. disorder ly) on consumption. In the orderly array conditions, items were organized on a tray by color or flavor but were randomly arranged in the disorderly conditions. Results suggested that participants consumed less of both commodities when the 24 item array was presented in a disorganized manner compared to the 6 item array. The opposite was true when both arrays were presented in an organized manner. Additionally, the authors evaluated the interaction between array organization and assortment variety, described as the number of distinct versus duplicate options within the array. The participants selected items from an array which contained either six distinct items or twelve items, including duplicates of each of the six items in the six item array. These arrays were either organized or disorganized. Results of this experiment demonstrated that participants consumed the same number of items across array sizes when the arrays were presented in a disorganized manner but consumed more items when the large array was presented in an organized manner. Miller, Kaplan, Reed, and White (2017) assessed the effects of assortment variety on toy engagement under free operant conditions in preschool children. The authors presented an array of six distinct items or 30 items, wh ich included five, duplicate sets of the items available in the 6 item array. Participants gained access to each array for two minutes and were instructed to engage with one item at a time. The authors found that the large array influenced response allocat ion among highly preferred and least preferred items. In the large array condition, engagement with the top preferred item decreased by 25% 40% compared to the small array
13 condition, and engagement with the least preferred items increased. Considering thes e results with those of Hoch et al. (1999) and Kahn and Wansink (2004), these studies suggest that disorganized arrays which include duplicate items influence various aspects of responding. However, it is unknown how a disorganized array or an array which includes duplicates may affect preference for array size with individuals with ASD Although, in clinical settings, it is unlikely that a token store would include duplicate items, including duplicate items (rather than additional items) increases the numb er of items in the array without the potential confound of differences in quality in the larger array, thus isolating the effect of array size. Another important consideration for clinicians is whether to present the array of available backup reinforcer s pictorially or textually within a token store. In a series of studies, Townsend and Kahn (2014) examined the effects of pictorial and textual presentation of choice options on a variety of factors that may contribute to choice overload, such as perceived assortment variety. In one of the studies, the authors examined preference for pictorial versus textual presentation of choice options and array size for two product categories (i.e., crackers and mutual funds) with nical Turk. Participants selected whether they preferred their choice options presented pictorially or textually and whether they preferred a small array (i.e., eight options) or a large array (i.e., 27 options). The order of preference questions was count erbalanced across participants. The authors then presented participants with Results indicated that most participants preferred pictorial presentation, regardl ess of their preference for array size or the product category presented. Thus, it seems important to examine the preferences of individuals with ASD for array size when choice options are presented pictorially or textually.
14 Finally, it may be important to consider preference for array size under conditions involving a work requirement. Response effort is inherent to token economies used to increase appropriate responding, such as completion of academic tasks. Individuals must work to produce tokens, whic h they accumulate until they meet the exchange production schedule (i.e., the number of accumulated tokens required to produce an exchange opportunity). Researchers have demonstrated that preference assessed under low work requirements may shift under more effortful schedules (Tustin, 1994; DeLeon, Iwata, Goh, & Worsdell, 1997; Roane, Lerman, & Vorndran, 2001). For example, DeLeon et al. separately examined preference between categorically similar (i.e., two food items) and dissimilar (i.e., a food item and a leisure item) reinforcers with two individuals with intellectual and developmental disabilities The authors assessed preference for reinforcers using a concurrent operant arrangement in which two microswitches were presented. Responding on eit her microswitch resulted in access to a different reinforcer Participants chose equal amounts of the reinforcers used in each condition when the work requirement schedule was fixed ratio ( FR ) 1. The researchers then increased the schedule requirement to F R2, FR5, FR10, and for dissimilar reinforcers only, FR20 a cross phases. N o clear preference emerged between dissimilar reinforcers across increases in schedule values for either participant. However, the authors found that a clear preference for one of the similar reinforcers emerged for both participants when the schedule requirement increased to FR5 and maintained at FR10. Hence, increases in the amount of effort may influence preferences, perhaps even preferences for array size, particularly when the ite ms available in those arrays are categorically similar (e.g., tangible items). To date, no studies have examined preference for array size with individuals with ASD Furthermore, researchers have yet to examine the effects of presentation format within tokens
15 stores or response effort on preference for array size. Therefore, in Experiment 1, we assessed preference for array size when the large array is presented in a n organized or disorganized manner. Specifically, we manipulated whether the position of the backup items within the array were fixed or randomized across trials. Next, we assessed preference for array size when the number of items in the large array was i ncreased to 30 items, which included duplicate items in a manner consistent with prior research We then assessed preference for array size when the backup items were presented textually across all array conditions. The purpose of Experiment 2 was to asses s preference for array size within the context of a token economy, when a work requirement proceeded array selection. To do so, we compared preference for array size with in a simple choice format (hereafter, the selection based preference assessment ) to a work requirement preference assessment, in which participants worked to produce and accumulate tokens prior to array selection.
16 CHAPTER 2 EXPERIMENT 1 Participants, Settings, and Materials Four individuals, who attended a behavioral services clinic part icipated in the present experiment. Andre was a 7 year old male diagnosed with ASD and attention deficit hyperactivity disorder. Kim was a 9 year old female diagnosed with ASD Lauren was a 4 year old female diagnosed with ASD chromosomal deletions, and F ragile X syndrome. Tyler was a 19 year old male diagnosed with ASD Kim, Andre, and Tyler had extensive verbal repertoires, were able to read, and communicated in full sentences. Lauren had a limited verbal repertoire, was unable to read, and communicated in three to four word sentences. All participants were able to follow simple one to two step instructions. Sessions took place in session rooms or a conference room, which contained a small table and two chairs or a conference table and three or more chai rs, respectively. Choice condition materials remained consistent across participants. In each choice condition, a white laminated sheet displayed the corresponding number of leisure items as either small pictures or textual labels affixed to the sheet with Velcro The large and small array condition sheets included a textual label at the top of the sheet indicating that the participant indicated the therapist would mak e a choice. Lauren could not read; therefore, pictures of Lauren and the therapist replaced textual labels to indicate who would make the item selection in each condition. A total of 16 leisure items were assessed and remained consistent across participant s. Participants had access to these items only during research sessions.
17 Response Measurement and Interobserver Agreement During the leisure item preference assessment (described below), the primary dependent variable was item selection, defined as touchin g or verbally requesting one of the presented items. The selection percentage for each leisure item was calculated by dividing the number of times the participant selected a given item by the number of times it was presented, multiplied by 100. During the array size preference assessment, the primary dependent variable was condition selection, defined as touching a condition sheet with any part of the hand. Item selection was a secondary dependent variable defined as touching the picture or textual label of a leisure item within the array with any part of the hands. For Kim, Lauren, and Tyler, additional dependent variables included array position, which was recorded as the position of the selected condition (i.e. left middle right) and item position, whic h was recorded as the position of the selected leisure item within the large array condition. Specifically, the large array was divided into 12 regions and data were collected on which region the selected item was located. In the no choice and small array conditions, item position was not recorded because the items remained in a fixed position throughout the assessment. Trained observers, which included graduate and undergraduate students and therapists at the behavioral services clinic, recorded data in vivo using paper and pencil. A secondary observer independently recorded reliability data on all portions of the experiment across participants. Agreement was defined as both data collectors reporting that the same selection within a given trial. Interobse rver agreement scores were calculated by dividing the number of agreements across trials by the total number of trials within a session and then converted into a percentage. Reliability data were collected on 100% of leisure item preference assessment sess ions and agreement scores were 100% across all participants. During the array size preference assessment, reliability data was collected for a mean of 89% (range: 68 100%) of
18 sessions across participants and produced means scores of 98% (range: 80 100%), 9 2% (range: 60 100%), 96% (range: 80 100%), and 97% (range: 80 100%) for array selection for Andre, Kim, Lauren, and Tyler, respectively. Leisure Item Preference Assessment All participants completed a 16 item paired stimulus preference assessment (PSPA), similar to Fisher et al. (1992), to identify highly preferred leisure items used in the small and no choice arrays. Prior to the assessment, participants received 30 s of free exposure to each item. Then, the experimenter presented the leisure items in pai rs and prompted participants to select one item. Item selection produced 30 s access to that item and removal of the non selected item. Attempts to select both items were blocked, and the experimenter subsequently removed and represented both items. Partic ipants completed the PSPA twice to produce an average percent selection for all items. The two items with the greatest percent selection were identified as high preference item 1 (HP1) and high preference item 2 (HP2). The item with the lowest percent sele ction was identified as low preference item 1 (LP1). If two or more items were selected equally, the experimenter referred to trials in which those items appeared concurrently and identified the item the participant selected in these trials as more preferr ed than the other. Discrimination Probe Prior to examining array preference, each participant completed a discrimination probe, in which the participant tacted the picture and textual label of each leisure item (only pictures for Lauren). The purpose of th e probe was to confirm that participants were able to identify each leisure item through the stimulus modalities subsequently used throughout both experiments. At the beginning of each trial, the experimenter pointed to a picture or textual label and provi ded a
19 next trial. If the participants emitted an incorrec t response or no response within 5 s, the therapist modeled the correct response and presented the leisure item. The therapist then removed the leisure item, represented the picture or textual label of the item, and prompted the participant it?). Discrimination probes ended when the participant independently tacted each picture or textual label on 100% of trials in one session. All participants were able to tact the items and did not require a second session. Procedures Experiment 1 assesse array, small array, and no choice array). In each condition, a laminated sheet pictorially or textually displayed the prescribed number of available leisure items. Prior to each session, t he therapist prompted participants to identify who would select items from each array condition If the participant correctly tacted the condition, the therapist delivered a praise statement and presented the next condition sheet. The participant had to tact correctly all three condition sheets before the session began. However, if the participant did not correctly tact the condition, the therapist modeled the correct response, prompted the participant to t act the condition again, and then presented the next condition. The condition that produced the incorrect tact was represented prior to starting the session. Sessions started with the participant sitting at a table while the experimenter presented random order across trials. At the beginning of each choice trial, the experimenter instructed partici pants to select one of the three conditions sheets. If the participant did not select a condition sheet within 5 s, a verbal prompt participant did not select and re pre sented the selected condition sheet for the participant to select a leisure item from within the sheet. In the large and small array conditions, the participant
20 selected one of the available leisure items from the array, which produced 30 s access to that item. If the participant manded to change their selection prior to item delivery, the array was re presented, and the participant selected a new item. Therapists allowed participants to change the selected item only once per trial. If the participant sele cted the no choice condition sheet the therapist pseudo randomly selected from the two available HP items, to which the participant gained access for 30 s. However, if the participant manded for a specific item prior to item delivery, the therapist delive red an item other than the one requested. This was done to avoid that they could make a choice after selecting the no choice condition. Each session consisted of f ive choice trials. In the Pictorial/Fixed phase of the array size preference assessment, all leisure items were displayed pictorially and remained in a fixed position within the arrays throughout the phase. In the large array condition, all 16 leisure ite ms were available. In the small array and no choice conditions only the two most preferred leisure items, HP1 and HP2, were available (see Table A 1 for list of HP1, HP2, and LP1 items for each participant). Preference for a choice condition was defined as selection of that condition on 80% or more of choice trials across three consecutive sessions. A preference for the no choice condition would have resulted in termination of the preference assessment but this did not occur If the participant displayed a preference for the large array condition or if a clear preference (one meeting the criterion) was not demonstrated within 10 sessions, a new phase was conducted with a different presentation format. In the Pictorial/Random phase, the condition sheet disp lay of leisure items was in a pictorial format across conditions but the position of leisure item pictures was pseudo randomized within the large array across choice trials. Therefore, the participant had to re scan
21 the array to select the same item acros s trials ostensibly increasing the effort of selecting the most preferred item The position of items within the small array and no choice conditions remained fixed across phases. The Pictorial/Random/Duplicate phase was similar to the previous phase wit h the exception that the number of pictures displayed in the large array condition increased to 30. Specifically, the display did not include the least preferred leisure item, LP1, and included duplicate pictures of the remaining 15 items. In the Textual/ Random phase, the display presented the leisure items textually instead of pictorially across all three conditions. As in the previous phase, the large array condition included 30 options, the positions of which were randomized across trials. Lauren did no t participate in this phase because she could not read. Data collected on the position of the conditions Tyler selected during the Textual/Random phase indicated that he often selected the condition displayed in the middle of the array of three, regardless of the condition type. Tyler therefore completed an additional array positional analysis to assess whether the Textual/Random presentation format produced sub optimal array selection. For Tyler, sub optimal array selection was defined as selection of the least selected condition in previous phases (i.e., the no choice condition). The array position analysis was similar to the Textual/Random phase with the exception that the no choice condition was always in the middle position. The position of large and s mall array conditions was counterbalanced across choice trials. Kim also completed an additional analysis, because data collected on the position of selected items during the Textual/Random phase suggested that she was exclusively selecting items from the top and bottom rows within the large array condition. Therefore, she completed
22 an additional within array positional analysis to determine whether the Textual/Random presentation format produced sub optimal selections. A sub optimal selection for Kim was defined as selection of an item previously identified as low pref erred through the PSPA. The within array positional analysis was similar to the previous phase, with the exception that the the large array condition and their position within these rows was randomized across trials. Results and Discussion Figures A 1 through A 4 show the results of the array size preference assessment for each participant (Andre, Kim, Tyler, and Lauren, respectively). In the Pictorial/Fixed pre sentation phase, all participants preferred the larger array condition. In the P ictorial/ R andom presentation phase, only Andre and Kim displayed a preference for the large array condition. P reference for an array condition was not determined for Tyler or Lauren within this phase (i.e., neither participant selected one condition in over 80% of c hoice tr ia ls across two consecutive) ; however, both participants selected the large array condition most often ( 58 % and 48 % across the phase for Tyler and Lauren respectively) In the P ictorial/ R andom/ D uplicate presentation phase, Andre, Kim and Lauren pr eferred the larger array condition. A preference was not technically determined for Tyler in this phase, as per the previously mentioned preference criteria, but he selected the large array condition most often ( 50 % across the phase) In the Textual/Random /Duplicate phase, Andre expressed a preference for the large array, but required 10 sessions before his preference stabilized enough to meet criterion. Preference for a choice condition was not determined for Kim or Tyler before 10 sessions were completed but again, these participants were more likely to select the large array condition within this phase. Overall, w e observed little evidence of effects consistent with choice overload across the presentation formats evaluated That is, no participant prefer red the small array nor did any
23 relinquish their opportunity to choose (i.e., choose the no choice condition) consistently T he small array and no choice conditions only contained HP1 and HP2 stimuli that we identified pre experimentally as most preferred Given that preferences may shift over time, it is possible that the items displayed in the small array and no choice conditions were no longer hi ghly preferred. Therefore, selection of the large array condition may have been due, in part, to the availability of momentarily more preferred items in that array. Still preferences beca me more inconsistent as the complexity of the presentation format in creased for some of the participants (Andre, Kim, and Tyler) particularly wh en duplicate items were in the large array size which increased the array to 30 items. It remains possible that the increase in array size increased the effort of scanning the arr ay for a preferred item. Thus, the increasingly inconsistent selections across conditions may reflect increased scanning effort.
24 CHAPTER 3 EXPERIMENT 2 Participants, Settings, and Materials Kim, Lauren, and Tyler participated in Experiment 2. All sessions were conducted in a therapy room in a behavioral services clinic. The materials were the same as those from Experiment 1. However, during the work requirement phases, token delivery was cont ingent on completion of target responses. Therefore, tokens, a clear container for token accumulation, and materials related to the target responses were present during the work requirement phases. Task materials varied across participants due to variation responses were acquisition tasks currently targeted in their academic programming. Target responses for Kim included discrete listener responding, intraverbal responses, tacting picture cards, object to pictur e matching, and picture to included discrete listener responding, intraverbal responses, tacting picture cards, verbal imitation, and gross and fine motor imitation. Tyler did not have an academic program and was instead, acquiring occupational skills. The experimenter conducted an informal interview with task that would be beneficial to him. The target responses selected for Tyler were discrete intraverbal respon ses in Spanish (e.g., days of the week). Response Measurement and Interobserver Agreement During token training, the primary dependent variable was independent token exchanges, defined as the participant placing one token in the hand of the therapist in t he absence of verbal, model, or physical prompts. Trained observers recorded data using paper and pencil and a secondary observer collected reliability data on 100% of sessions. Interobserver agreement scores were calculated as the total number exchange tr ials in which both observers reported an
25 independent exchange divided by the total number of exchange trials within each session, multiplied by 100. Mean agreement was 100% During the response effort assessment, condition selection was the primary depen dent variable. Trained observers also collected data on array position, item selection, and item position as in Experiment 1. Interobserver agreement was assessed for a mean of 34 % 39 %, and 100 % of sessions and produced a mean score of 100 %, 100 % and 96 % (range: 80 100 %) for Kim, Lauren, and Tyler, respectively. Token Training All participants had a previous history with token economies as part of their educational programs Kim had a well established token economy, which was used in the present experiment Lauren had experience with a token economy recently introduced into her academic programming, but therapists reported that tokens did not function as conditioned reinforcers. Tyler had an ongoing token economy in which he earned money for compl eting occupationally related tasks. For logistical reasons, monetary outcomes were not available as backup reinforcers in the context of the present study. Therefore, Tyler and Lauren required training of a novel, qualitatively different token (i.e. green poker chip). During the work requirement phases of the array preference assessment, the token production and exchange current or previous token economies ( see Table A 2 for token produ ction and exchange production schedules by participant) All participants exchanged tokens on an FR1 schedule where each token was exchanged for 30 s access to the selected backup item During the first phase of token training, participants were asked to verbally state the contingencies of their respective token economies. If a participant was unable to describe the contingencies in place, they were exposed to the second phase of token training Both Kim and Tyler were able to
26 correctly state the contingen cies of their token economies and did not require further training. Lauren was exposed to the remaining phases of the token training, which continued until the mastery criteria (i.e., independently exchanged tokens for 80% or more trials for three consecut ive sessions) were met at the programmed token production and exchange production schedule. During the stimulus stimulus pairing phase of token training with Lauren the experimenter delivered a token noncontingently into a clear container and immediately after provided access to a leisure item for 30 s. All 16 leisure items followed token delivery twice in a pseudorandom order for a total of 32 pairing trials. During token exchange training, the experimenter delivered one token noncontingently into a clear The therapist then held out their hand to prompt the participant to exchange the token to access a backup leisure item. If Lauren did not exchange the token within 5 s, the therapist followed a least to most prompting sequ ence and represented the token. Contingent upon token exchange, the therapist prompted Lauren to select a backup item from a pictorial array of the 16 available leisure items, which produced 30 s access to that item. Sessions consisted of a single trial an d t he criteria to move onto the next phase of token training was independent token exchange across three consecutive sessions During token production training, token delivery and exchange procedures were similar to those in token exchange training with t he exception that token delivery was contingent on completion of a target response. Specifically, the target response was an academic task that target response was matching color cards in an array of three. If an incorrect response or no response occurred within 5 s of task presentation, the therapist followed a least to most
27 prompting sequence and delivered the token contingent on task completion at any prompting le vel. Following token delivery, Lauren exchanged the token, selected a backup leisure item from the pictorial array, and gained access to that item for 30 s. Upon completion of the reinforcement interval, the experimenter presented a new work trial. Each se ssion consisted of five work trials. A minimum of three sessions were conducted per token production schedule. The criteria to increase the token production schedule was two consecutive sessions in which the participant independently exchanged tokens on 80 % of trials. The token production schedule increased to FR2, FR3, and finally variable ratio ( VR ) 3. Once Lauren met the mastery criteria at a VR3 token production schedule, she began token accumulation training. In this phase, Lauren earned tokens contingent on task completion on a VR3 schedule and the number of tokens Lauren had to accumulate to produce an exchange opportunity (i.e., exchange production schedule) was increased steadily Once Lauren met the exchange production schedule, she exchanged tokens one at a time on an FR1 schedule for 30 s of access to the selected backup item. Attempts to exchange more than one token at a time were blocked. Sessions ended once Lauren exchange d all tokens. A minimum of three sessions were conducted per exchange production schedule. The exchange production schedule increased to FR2, FR3, FR4, and FR5 contingent on independent token exchanges on 100% of exchange trials across two consecutive sessio ns. Token training ended when Lauren met the mastery criteria at the FR5 exchange production schedule. Procedures The response effort assessment evaluated the effects of a work requirement versus no work requirement on preference for choice conditions in a reversal design For Kim and Tyler, leisure items appeared textually in all choice conditions throughout the assessment. For Lauren leisure items appeared pictorially within each choice condition. For all participants, the large
28 array condition included the same 30 items (i.e., 15 unique items and a duplicate of each) as previously described in Experiment 1, and the position of these items was randomized across choice trials. The small array and no choice conditions included the same two highly preferred leisure items used in Experiment 1. Sessions within the no work phases were similar to those in Experiment 1. The no work phase consisted a maximum of ten sessions and each session consisted of five choice trials The primary dependent variable was preference for choice condition, defined as selection of one condition on 80% or more of trials across three consecutive sessions. When a preference for a choice condition emerged or no preference emerged within ten ses sions, the work requirement phase began. During the work requirement phase, participants produced tokens contingent on completion of academic tasks. Token production and exchange production schedules were med token economy schedule, thus these schedules varied across participants (see Table A 2 for token production and exchange production schedules by participant). Once participants met the exchange production schedule, they selected a choice condition in w hich they would exchange their tokens for 30 s access to the selected leisure item on an FR1 schedule. We observed that completing the exchange production schedule took up to 45 minutes for Kim and Tyler. Therefore, o nly one choice trial was presented in e ach session and the criteria for determining preference in this phase was changed. The primary dependent variable was preference for a choice condition, defined as selection of one choice condition across five consecutive sessions. A reversal to the no wor k phase began when a preference emerged or when no preference emerged across 10 sessions. Results and Discussion Figure A 5 shows the results of the response effort assessment for Kim, Tyler, and Lauren as a proportion of total choices in each phase. Data are displayed as proportional bar
29 graphs because the criteria to identify a preference for a choice condition was different across phases due to the nature of the sessions. That i s, sessions in the no work phases consisted of 5 choice trials and typically took less than five minutes to complete S essions within the work phases consisted of one choice trial per session and sessions took upwards of 45 minutes to complete the exchange production schedule. Thus, for practical reasons, the criteria to identify the preference within work phases was adjus ted to selection of the same condition across five consecutive sessions. To make phases comparable, data were analyzed as a proportion of selections of each condition over the total number of choice opportunities. Additionally, data from the T extual/ R andom / D uplicate phase from Experiment 1 for Kim and Tyler, and the P ictorial/ R andom/ D uplicate phase for Lauren were included in the results graphs for E xperiment 2 as the first no work phases for each participant Across phases, both Kim and Lauren were more li kely to select the large array condition independent of work requirement. In fact, Lauren demonstrated a very strong preference for the large array condition large array condition seemed to be heightened slightly by the work re quirement. Tyler was also more somewhat likely to select the large array condition during the no work phases. By contrast during the work phases he never selected the array of 30. Rather, he selected the small array condition most often in the first work phase and in the second work phase, he most often opted to relinquish choosing. Selection patterns during work phases are the only outcomes we found that seem consistent with the c hoice overload concept. It seems possible that scanning the large array could function as additional demand, which becomes more aversive following the pre choice work requirement. Therefore, Tyler may have preferred the small array and no choice condition s during work requirement phases because both require d
30 minimal scanning effort (i.e., less subsequent effort after already having to exert effort to get to the choice).
31 CHAPTER 4 GENERAL DISCUSSION Given that tokens are valuable only insofar as they are exchangeable for preferred backup items, it is important that the context in which they are exchanged does not produce negative collateral effects, such as those reported in the choice overload literature. In these two experiments, we a ssessed preference for array size under varying presentation formats and under conditions both with and without a work requirement in individuals with ASD In Experiment 1, a ll participants selected the large array condition more often across all presentat ion formats. Additionally, Kim and Lauren selected the large array condition most often in the selection based arrangement and within the context of their token economies. Tyler also selected the large array condition with greater frequency than the other two conditions in the selection based arrangement; however, Tyler preferred the small array condition or the no choice condition when preference was assessed within the context of his token economy (Figure A 5). For the most part these results are incongr uent with findings within the choice overload literature I n the discussion that follows, we propose several factors that may account for the disparity in findings. Although research within choice overload has generally focused on typically developing pop ulations, we examined preference for array size with individual s with ASD which presents a unique set of constraints. Individuals with ASD may show scanning deficits, which may preclude selecting items from a large array. Researchers have suggested that in dividuals with ASD may benefit from smaller stimul us arrays (Reed & Tiger, 2015; Maurice, Green, & Foxx, 2001; Lovaas, 2003). However, all the participants in the present study were more likely to select the large array condition, with the exception of Tyl er and only in the work requirement phases. Scanning large arrays may not be particularly problematic for this population. For example, Kodak et al. (2019) evaluated the efficacy of tact training when participants were
32 presented with picture cards in array s of three, four, six, and twelve. Results indicated that arrays of six and twelve required fewer exposures to mastery, thus larger arrays were associated with more rapid acquisition. Although further investigation on the effects of array size on preferenc e are warranted, preliminary results suggest that individuals with ASD may benefit from larger array sizes. Researchers have also suggested that scanning an array may be an essential skill for the acquisition of conditional discriminations (Green, 2001; Grow & LeBlanc, 2013). In the present study, we did not assess the extent to which participants could scan an array. Interestingly, Tyler did not prefer the large array during work requirement phases. As already suggested, i t may have been possible that se lecting from a large array functioned as an additional demand in that it requires increased effort to scan the items. Future research should examine the extent to which participants can scan increasingly larger arrays, as described in Kodak et al. (2015), and assess whether preference for array size varies across participants with different scanning repertoires. Furthermore, one of the behavioral outcomes reported as a product of choice overload is that individuals are more likely to switch their initial ch oice to an alternative option (Chernev, 2003). Within the present study, participants were able to change their selection of backup item once per choice trial. Anecdotally, Lauren made several requests to change the selected backup item throughout both exp been a result of not scanning the array prior to making her initial selection; however, data were not collected on the frequency of preference reversals. Therefore, future research shoul d examine whether individuals who do not have extensive scanning repertoires are more likely to experience post choice preference reversals and whether these preference reversals are more likely to occur with large arrays.
33 Research investigating choice ove rload effects typically uses single decision scenarios when assessing preference rather than repeated measures, which are used conventionally in behavior analytic research. We assessed preference for the backup items to be displayed in the small array and no choice conditions pre experimentally. However, preference can shift over time such that the items available in the small array and no choice conditions may become less preferred than items available exclusively in the large array condition an effect tha t might not be captured in a single decision scenario Future research c ould assess preference for backup items prior to each session block to assure that items which are displayed in the small array and no choice conditions remain highly preferred. Additi onally, the size of the small and no choice arrays (i.e., 2 items) may have been too limiting, such that these arrays were unable to support moment to moment shifts in preference for backup items. Future research should also attempt to parametrically incre ase the number of items available in the small and no choice arrays to identify the optimal number of backup reinforcers that should be used for each individual. In relation the size of the large array conditions (16 or 30 items) may have been insufficien tly large to produce effects consistent with choice overload I ncreasing it beyond those values seemed ecologically inconsistent with how token stores are arran g ed in actual practice. For three participants, we assessed preference for array size when bac kup items were presented textually. Heinicke, Carr, and Copsey (2019) reviewed frequently used stimulus modalities in published stimulus preference assessments (SPA) with individuals with ASD Of the studies which included a validation measure, such as cor respondence with a tangible SPA or reinforcer assessment, 75%, 61%, and 37% reported accurate reinforcer identification through video, pictorial, and verbal modalities, respectively. However, the authors reported that they did not identify any studies that employed a textual based SPA. We did not assess whether items
34 selected during the textual presentation phase corresponded with highly preferred items identified through the tangible PSPA. Given that token stores may be presented textually in applied setti ngs, future research should examine whether results from textual based SPAs correspond with SPAs employing other stimulus modalities. Overall, despite not having observed choice overload effects in our participants, there is still ecological validity in op timizing token stores for individuals with ASD There are several research questions that the current studies raised with the goal of enhancing the efficacy of token economies in this population. There are also ethical considerations in making sure that to ken economies are as effective as possible and do not evoke any potential negative collateral effects.
35 APPENDIX T ABLES AND FIGURES Table A 1. Participant names and stimuli identified as high and low preference. Participant High Preference (HP1 & HP2) Low Preference (LP1) Andre slime and k inetic sand blocks Kim bubble wand and soccer ball toy cars Tyler slime and stress ball playdoh Lauren slime and playdoh stress ball
36 Table A 2. Token production and exchange production schedules across participants in Experiment 2. Participant Token Production Schedule Exchange Production Schedule Kim VR7 FR20 Lauren VR3 FR5 Tyler FR5 FR10
37 Figure A 1. Array size preference assessment across presentation formats for Andre
38 Figure A 2. Array size preference assessment across presentation formats for Kim
39 Figure A 3 Array size preference assessment across presentation formats for Tyler.
40 Figure A 4. Array size preference assessment across presentation formats for Lauren.
41 Figure A 5 Comparison of preference for array size across no work and work conditions across all participants.
42 LIST OF REFERENCES Chernev, A. (2003). When more is less and less is more: The role of ideal point availability and assortment in consumer choice. Journal of Consumer Research 30 (2), 170 183. Chernev, A. (2012). Product assortment and consumer choice: An interdisciplinary review. Foundations and Trends in Marketing 6 (1), 1 61. doi: 10.1561/1700000030 Chernev, A., Bckenholt, U., & Goodman, J. (2015). Choice overload: A conceptual review and meta analysis. Journal of Con sumer Psychology 25 (2), 333 358. doi: 10.1016/j.jcps.2014.08.002 Reward value is directly related to the effort or time required to obtain the reward. Psychonomic Bulletin & Review, 7(1), 100 106. DeFulio, A., Yankelevitz, R., Bullock, C., & Hackenberg, T. D. (2014). Generalized conditioned reinforcement with pigeons in a token economy. Journal of the Experimental Analysis of Behavior 102 (1), 26 46. DeLeon, I. G., Iwata, B. A., Goh, H. L., & Worsdell, A. S. (1997). Emergence of reinforcer preference as a function of schedule requirements and stimulus similarity. Journal of Applied Behavior Analysis 30 (3), 439 449. Diehl, K. (2005). When two righ ts make a wrong: Searching too much in ordered environments. Journal of Marketing Research 42 (3), 313 322. Dyer, K., Dunlap, G., & Winterling, V. (1990). Effects of choice making on the serious problem behaviors of students with severe handicaps. Journal of Applied Behavior Analysis 23 (4), 515 524. Fisher, W., Piazza, C. C., Bowman, L. G., Hagopian, L. P., Owens, J. C., & Slevin, I. (1992). A comparison of two approaches for identifying reinforcers for persons with severe and profound disabilities. Journa l of Applied Behavior Analysis 25 (2), 491 498. Graff, R. B., & Karsten, A. M. (2012). Assessing preferences of individuals with developmental disabilities: A survey of current practices. Behavior Analysis in Practice 5 (2), 37 48. Green, G. (2001). Behavior analytic instruction for learners with autism: Advances in stimulus control technology. Focus on Autism and Other Developmental Disabilities, 16 72 85. doi:10.1177/108835760101600203 Grow, L., & LeBlanc, L. (2013). Teaching receptive lang uage skills: Recommendations for instructors. Behavior Analysis in Practice, 6 56 75. doi:10.1007/BF03391791 Hackenberg, T. D. (2018). Token reinforcement: Translational research and application. Journal of Applied Behavior Analysis 51 (2), 393 435.
43 Heini cke, M. R., Carr, J. E., & Copsey, C. J. (2019). Assessing preferences of individuals with developmental disabilities using alternative stimulus modalities: A systematic review. Journal of Applied Behavior Analysis Hine, J. F., Ardoin, S. P., & Call, N. A (2017). Token economies: Using basic experimental research to guide practical applications. Journal of Contemporary Psychotherapy 48 1 10. Hoch, S. J., Bradlow, E. T., & Wansink, B. (1999). The variety of an assortment. Marketing Science 18 (4), 527 54 6. Iyengar, S. S., & Lepper, M. R. (2000). When choice is demotivating: Can one desire too much of a good thing? Journal of Personality and Social Psychology 79 (6), 995. doi: 10.1037//0022 3518.104.22.1685 Kazdin, A. E. (1977). The token economy: A review and evaluation. New York: Plenum Kazdin, A. E., & Bootzin, R. R. (1972). The token economy: An evaluative review. Journal of Applied Behavior Analysis 5 (3), 343 372. Kern, L., Mantegna, M. E., Vorndran, C. M., Bailin, D., & Hilt, A. (2001). Choice of task sequence to reduce problem behaviors. Journal of Positive Behavior Interventions 3 (1), 3 10. Kodak, T., Clements, A., Paden, A. R., LeBlanc, B., Mintz, J., & Toussaint, K. A. (2015). Examination of the relation between an assessment of skills and performance on auditory visual conditional discriminations for children with autism spectrum disorder. Journal of Applied Behavior Analysis, 48 52 70. doi:10.1002/jaba.160 Kodak, T., Halbur, M., Bergmann, S., Costello, D. R., Benitez, B., Olsen, M., ... & Cliett, T. (2019). A comparison of stimulus set size on tact training for children with autism spectrum disorder. Journal of A pplied B ehavior A nalysis Lovaas, O. I. (2003). Teaching individuals with developmental delays: Basic intervention techniques. Austin, TX: PRO ED. Maurice, C., Green, G., & Foxx, R. M. (2001). Making a difference: Behavioral intervention for autism. Austin, TX: PRO ED. Miller, J. R., Kaplan, B. A., Reed, D. D., & White, L. D. (2017). Effects of array size on preschoole overload. Behavior Analysis: Research and Practice 17 (1), 74. generalized conditioned reinforcers: Establi shment and validation. Behavioral Programs 23 (1), 13 38.
44 Powell, S., & Nelson, B. (1997). Effects of choosing academic assignments on a student with attention deficit hyperactivity disorder. Journal of Applied Behavior Analysis 30 (1), 181 183. Reed, D. D., & Tiger, J. H. (2015). Basic choice research. In F. D. DiGennaro Reed & D. D. Reed (Ed.), Autism service delivery: Bridging the gap between science and practice (pp. 175 191). Ne w York: Springer. doi:10.1007/978 1 4939 2656 5 overload as a possible instance of effort discounting. The Psychological Record 61 (4), 547 560. Rispoli, M., L ang, R., Neely, L., Camargo, S., Hutchins, N., Davenport, K., & Goodwyn, F. (2013). A comparison of within and across activity choices for reducing challenging behavior in children with autism spectrum disorders. Journal of Behavioral Education 22 (1), 66 83. Roane, H. S., Lerman, D. C., & Vorndran, C. M. (2001). Assessing reinforcers under progressive schedule requirements. Journal of Applied Behavior Analysis 34 (2), 145 167. Schwartz, B. (2004a). The paradox of choice: Why more is less. New York, NY: HarperCollins. Schwartz, B., & Ward, A. (2004b). Doing better but feeling worse: The paradox of choice. Positive Psychology in Practice 86 104. Sran, S. K., & Borrero, J. C. (2010). Assessing the value of choice in a token system. Journal of Applied Behavior Analysis 43 (3), 553 557. Tarbox, R. S., Ghezzi, P. M., & Wilson, G. (2006). The effects of token reinforcement on attending in a young child with autism. Behavioral Interventions: Theory & Practice in ical Programs 21 (3), 155 164. Tiger, J. H., Hanley, G. P., & Hernandez, E. (2006). An evaluation of the value of choice with preschool children. Journal of Applied Behavior Analysis 39 (1), 1 16. versus verbal depiction on assortment processing, perceived variety, and choice overload. Journal of Consumer Research 40 (5), 993 1015. Tustin, R. D. (1994). Preference for reinforcers under varying schedule arrangement s: A behavioral economic analysis. Journal of A pplied B ehavior A nalysis 27 (4), 597 606. ethic effect in pigeons. Journal of the experimental analysis of beha vior, 91(1), 143 152. Journal of the Experimental Analysis of Behavior, 87(3), 383 399.
45 Vaughn, B. J., & Horner, R. H. (1997). Ident ifying instructional tasks that occasion problem behaviors and assessing the effects of student versus teacher choice among these tasks. Journal of Applied Behavior Analysis 30 (2), 299 312.
46 BIOGRAPHICAL SKETCH Nathalie Fernandez received her the University of Florida in 200 7 Sh e received 201 2 from Florida International University Sh e is currently pursuing a degree in p sychology, in the area of b ehavior a nalysis, under the advisement of Dr. Iser DeLeon