Abstract
The current study investigated the effectiveness of stating and modeling contingencies in increasing food consumption for two children with food selectivity. Results suggested that stating and modeling a differential reinforcement (DR) contingency for food consumption was effective in increasing consumption of two target foods for one child, and stating and modeling a DR plus nonremoval of the spoon contingency was effective in increasing consumption of the remaining food for the first child and all target foods for the second child.
Keywords
Feeding problems are a particular concern for children with autism spectrum disorder (ASD; Ahearn, Castine, Nault, & Green, 2001; Schreck, Williams, & Smith, 2004). One such problem is food selectivity, which is characterized by consumption of a small range of foods and rejection of less preferred and/or novel foods (Levin & Carr, 2001; Penrod, Gardella, & Fernand, 2012). Although the likelihood of developing a feeding problem has been reported to be higher in children with autism compared with other populations (Ahearn, 2002; Ahearn et al., 2001), there is a relatively small literature base evaluating treatment procedures for this population.
Escape extinction (often in the form of a nonremoval of the spoon procedure; NRS) has received the most attention in the literature and has been well documented as an effective procedure for expanding the number of foods in a child’s diet, while decreasing inappropriate mealtime behavior (Bachmeyer, 2009; Freeman & Piazza, 1998). However, due to potential side effects of NRS, such as extinction bursts and emotional responding, this treatment procedure may not always be feasible; thus, evaluating alternative treatment procedures, specifically for children with autism, is much needed. The focus of the current study was to evaluate the effects of stating and modeling contingencies as a treatment for food selectivity in children with autism, given that modeling has been shown to be effective in other applications of behavior analysis with this population (e.g., LeBlanc et al., 2003; Sherer et al., 2001; Shipley-Benamou, Lutzker, & Taubman, 2002)
Although human learning often involves interaction with environmental variables, a large portion of human behaviors are learned not through direct interaction with the environmental contingencies, but through observations of others who have contacted those contingencies. In a classic experiment conducted by Bandura (1965), children who observed a model receiving punishment for his aggressive behavior toward a Bobo doll exhibited less aggression than those who saw the model receiving reinforcement or no consequence for his aggressive behavior. Bandura’s experiment showed that whether a learner imitates the behavior of a model depends on the indirect contact with the consequences of the model’s behavior. If Person A sees Person B engaging in a behavior that is then followed by reinforcement, the likelihood of imitation will increase. However, if Person B’s behavior is punished, the likelihood of imitation will decrease.
Pierce and Cheney (2008) referred to this type of observational learning as complex observational learning. Over time, multiple similar learning experiences teach the child differential imitation based on the modeled consequences. The child learns to imitate behaviors of others when the modeled behaviors have been reinforced and avoid imitating the same behaviors when the behaviors of others have been punished. By contrast, a person who only has simple generalized imitation in his or her repertoire (i.e., ability to imitate others as a repertoire) may not respond differentially based on the situation. A child may learn to match the model’s response; however, the child may not learn to do so only when the model’s response is reinforced.
As mentioned earlier, although modeling has been evaluated in several applications of behavior analysis (e.g., LeBlanc et al., 2003; Sherer et al., 2001; Shipley-Benamou et al., 2002), limited behavioral research has been done to evaluate the effects of modeling in the treatment of pediatric feeding disorders such as food selectivity. One notable exception was a study conducted by Greer, Dorow, Williams, McCorkle, and Asnes (1991) evaluating the effectiveness of peer modeling on swallowing in a child with dysphagia. During baseline, the experimenter presented the same portion of the same food to both the target child and his sister (model). No programmed reinforcement was given to the sister contingent on swallowing. During treatment, the therapist first presented a spoonful of the target food to the sibling, after which the target child observed his sister swallowing the food and receiving reinforcement in the form of verbal praise and a token. The target child was then presented with a spoonful of the target food, and he was given 5 s to swallow the food independently. If he did not consume the food after 5 s had elapsed, the presented spoon was removed, and his sister was given another opportunity to consume the food and receive reinforcement. For the remaining trials, eating opportunities between the sister and the target child were rotated in this fashion until the sister finished her assigned portion. Results showed that modeling was effective when opportunities to eat and receive reinforcement were alternated between the target child and his sister.
In a more recent study conducted by Sira and Fryling (2012), researchers evaluated the effectiveness of a treatment package including peer modeling and differential reinforcement (DR) with a child with autism. During baseline, the child was presented with 10 bites of the target food and an instruction to “take a bite” in the absence of the peer model. During treatment, the child’s peer model (his sister) was seated at the dining table with him. The presentations of food and instructions were alternated between the model and the participant. The peer model contacted the programmed reinforcers for each bite of food she consumed. Results showed that a treatment package involving modeling and DR was effective in increasing the variety of foods in the participant’s diet. However, it should be noted that given the absence of a peer model in the baseline condition, it is difficult to conclude whether the consumption during treatment was due to the treatment procedures or the mere presence of the model.
Although results of these studies are promising, the scarcity of research in this area warrants further investigation. Thus, the purpose of the current study was to evaluate the effects of a treatment package that involved stating and modeling contingencies for consumption and inappropriate mealtime behavior. A baseline condition in which consumption was modeled in the absence of any programmed consequences was included. It should be noted that in contrast to the aforementioned studies, in the current study, eating opportunities were presented to both the participant and the model simultaneously; this was done to resemble a more natural mealtime situation.
Method
Participants, Models, Therapists, and Setting
Two children with autism and no history of feeding therapy participated. Larry was a 10-year-old boy whose diet was limited to yogurt, chicken nuggets, grapes, and watermelon. Adam was a 9-year-old boy whose diet was limited to pancakes, waffles, chicken, steak, and pork roast. Both had extensive verbal repertoires, imitated others, and followed multi-step instructions. One model and two feeding therapists, trained graduate students in applied behavior analysis, implemented the intervention and served as data collectors. The graduate students involved had prior clinical and research experience in implementing the procedures we used in the current study (DR and NRS) under the direction of the second author. For the current study, graduate students were trained by the first author using Behavioral Skills Training (BST), including written instructions, modeling, rehearsal, and feedback. Sessions did not commence until after the graduate students demonstrated competency implementing the procedures (rehearsed the procedures with no errors requiring feedback from the first author).
Sessions were conducted in the Pediatric Behavior Research Laboratory at Sacramento State University, equipped with a table, chairs, microwave, timers, foods, toys and games, utensils, napkins, serving dishes, and camcorder. Participants sat across from an adult model. The therapist sat at the end of the table. The therapist presented foods in pieces or prescooped, approximately 1.5 cm × 1.5 cm in size and conducted three to four sessions with 10-min breaks in between, 2 days per week. Each session had 12 trials. Four bites of three different foods were presented one at a time; in other words, the therapist served four bites of a single food, and plates were cleared prior to the therapist serving a new food. Mean session duration in minutes for Group A foods was 14 min (range = 11-19) for modeling, 19 min (range = 14-22) for modeling DR, and 21 min (range = 13-44) for modeling DR and NRS. Mean session duration for Group B foods was 12 min (range = 9-14) for modeling, 10 min (range = 7-15) during the modeling DR phase, and 9 min (range = 5-11) during the modeling DR and NRS phase.
Response Measurement
Session data were scored from videotapes. Data were recorded bite-by-bite, using prepared data sheets. Observers scored independent acceptance when the child placed a bite of food past the plane of his lips, using his hands or a utensil any time after the therapist presented the food and within 5 s of the model accepting the last of the four bites on his plate. Independent acceptance was scored for every bite the child placed past the plane of his lips. For example, if a participant put two bites in his mouth at the same time, independent acceptance would be scored twice. Observers scored bite consumption for each bite the child independently accepted when no food was in the child’s mouth after approximately 30 s from the time the bite was accepted. Percentage of bites consumed was calculated by dividing the number of bites consumed by the number of bites presented. Observers collected 10-s partial interval data on inappropriate mealtime behavior scored whenever participants engaged in any of the following behaviors: negative vocalizations about the foods (e.g., “Ew,” “Yuck”), covering the mouth or the foods, turning the head or body away from the table, pushing the plate or therapist’s hand, elopement, throwing objects, aggression, or self-injurious behavior.
Treatment Integrity and Interobserver Agreement
Treatment integrity and interobserver agreement data were collected during all sessions and calculated by dividing the total number of correct implementations or agreements, respectively, by correct plus incorrect implementations or agreements plus disagreements, respectively, and converting the ratio to a percentage. Treatment integrity for bite presentation, prompting, consequence delivery, bite removal, and correct modeling procedure and interobserver agreement for independent acceptance and bite consumption were 100% across participants. Mean interobserver agreement for inappropriate mealtime behavior was 97.64% (range = 82.86%-100%).
Experimental Design
We used a non-concurrent multiple-baseline design across participants combined with a multi-element design consisting of alternating baseline and treatment conditions, which were associated with different therapists. During baseline (modeling), two different feeding therapists alternated conducting sessions to control for any experimenter specific effects. Then, during the treatment phase, one therapist continued conducting baseline sessions while the other therapist conducted treatment sessions.
Paired-Choice Preference Assessment
We assessed relative preferences for nonpreferred and high-preferred foods identified in a parent interview separately in two paired-choice preference assessments (Fisher et al., 1992). The purpose of doing preference assessments for the nonpreferred foods was to ensure that foods identified in the parent interview were indeed nonpreferred; therefore, they could be used as treatment and baseline foods for the study. We identified six nonpreferred foods (foods that were never selected during the preference assessment), which we divided into Groups A and B. For Adam, Group A foods were cucumber, honeydew, and celery, and Group B foods were zucchini, watermelon, and carrot. For Larry, Group A foods were pasta, mushroom, and Havarti cheese, and Group B foods were sweet potato, bell pepper, and Greek yogurt. The purpose of doing preference assessments for high-preferred foods was to identify the top three high-preferred foods to be used as reinforcers during treatment phases of the study. For Adam, M&M’s®, Girl Scout® Thin Mints cookies, and Kellogg’s® Rice Krispie Treats were used as reinforcers. For Larry, Cheetos®, Crunch®, Lays® chips, and Kit Kat® were used as reinforcers.
Procedure
Modeling
Foods in Groups A and B were alternated across sessions, using the same procedure. The therapist presented three foods (one food at a time) in a systematic, randomized order, placing two plates, with four bites of a single food (e.g., carrots) on the table within arm’s reach of the model and the participant while saying, “Let’s try some [food]” or “How about some [another food].” After 5 s had elapsed without independent acceptance or food-approaching behavior (e.g., picking up a bite), the model consumed the first bite on the plate; if the participant had not put a bite in his mouth but was approaching the food after 5 s had elapsed, the model waited an additional 5 s before consuming the first bite on the plate. The same procedure was followed for all remaining bites; in other words, if the participant had not put a bite in his mouth but was approaching the food after 5 s had elapsed from the time the model consumed the first bite on the plate, the model waited an additional 5 s before consuming the second bite. If the participant was not looking at the model, the model prompted an observing response, defined as the participant making eye contact for at least 3 s, by making verbal comments (e.g., “I feel tired” and “Look at me”).
Once the model consumed the first bite (i.e., chewed and swallowed the bite), he or she then waited 5 s before accepting another bite on the plate. After the model consumed all four bites of the first food, the therapist waited 5 s if there was still food on the participant’s plate. If independent acceptance or food-approaching behavior did not occur after 5 s, the therapist removed the remaining bites from the participant’s plate and presented another four bites of a different food to the model and the participant. If the participant put a bite in his mouth within 5 s of the model consuming the last bite on the plate, independent acceptance was scored and the therapist provided more time to allow the participant to consume the remaining bites; for the remaining bites, independent acceptance was scored if the participant put a bite in his mouth using his hands or a utensil within 5 s of swallowing the previous bite. During this time, the model sat quietly. At any point during the session, if the model had more bites on his or her plate than the participant, the model accepted two or more bites simultaneously such that the number of bites on the model’s and participant’s plates were equal. Sessions ended when the therapist presented all three foods consecutively, with a total of 12 bites per session.
There was no programmed contingency for consumption. The therapist and model made neutral statements or engaged the participant in conversation not related to eating. A cup of water was next to the therapist, model, and child to resemble a typical mealtime.
Modeling DR
The therapist implemented treatment with Group A foods. Group B foods were presented in alternating sessions under baseline (modeling) conditions. The procedure for Group A foods was the same as modeling, except as follows. The therapist presented each plate of food by saying, “Let’s try some [food]. If you finish all your [food], you can pick one of your favorite treats, and you can also play with [preferred item].” The therapist said “Good job eating. You only have [number] more bites” to the model or participant after consumption of every bite. Each time the model or participant consumed all four bites on the plate, the model or participant selected one 3-cm × 3-cm piece of one of the participant’s preferred foods followed by approximately 3 min of access to the participant’s preferred item. The therapist removed the participant’s plate if the participant showed no independent acceptance or food-approaching behavior for 5 s when the model was accessing reinforcement.
Modeling DR and NRS
The therapist implemented treatment with Group A foods. Group B foods continued to be presented in alternating sessions under baseline (modeling) conditions. Treatment was identical to modeling DR except as follows. The therapist added the statement, “But if you don’t eat your [food], I will have to help you” to the cue for the presentation of each plate of food. If the participant did not show independent acceptance or food-approaching behavior within 5 s of the initial presentation of the plate of food or model’s consumption of a bite, the model engaged in an inappropriate mealtime behavior he or she had observed previously from the participant, independent of whether the participant was currently engaging in inappropriate mealtime behavior or passive refusal behavior. Inappropriate mealtime behaviors that were modeled included negative vocalizations about the foods (e.g., “Ew,” “Yuck”), covering the mouth or the foods, turning the head or body away from the table, and pushing the plate or therapist’s hand. The model did not exhibit severe inappropriate mealtime behavior (e.g., elopement, throwing objects, hitting). The model continued to engage in inappropriate mealtime behavior, and in sequential 5-s intervals, the therapist gave a model prompt and warning of the upcoming NRS procedure, then the therapist implemented NRS. During NRS, the therapist placed the spoon with food near the model’s lips. The model showed inappropriate mealtime behavior for approximately 30 to 60 s, after which he opened his mouth such that the therapist could deposit the bite into his mouth. After the bite had been deposited, the model and the therapist waited 5 s. If the participant did not show independent acceptance or food-approaching behavior, the model demonstrated inappropriate mealtime behavior, and the therapist implemented NRS. If the participant showed independent acceptance or food-approaching behavior within the 5-s opportunity before the modeling procedure, the model did not show inappropriate mealtime behavior. After the model consumed all four bites of the same food, the therapist gave him the programmed reinforcement (even when the therapist used NRS).
While the model was contacting reinforcement, the therapist waited 5 s for independent acceptance or food-approaching behavior from the participant. If independent acceptance or food-approaching behavior did not occur within 5 s, the therapist, in sequential 5-s intervals, would give a model prompt and warning of the upcoming NRS procedure and then implement NRS in the same manner described above. However, it should be noted that the therapist never had to implement NRS with either participant.
Modeling DR and NRS with Food Groups A and B
In the final treatment phase of the study, modeling DR and NRS was applied to Group B foods that were previously associated with baseline (modeling) conditions. Thus, Food Groups A and B were presented in alternating sessions, with the same treatment procedures in place for both food groups.
Follow-up
The therapist conducted follow-up 8 and 4 weeks after completion of treatment for Larry and Adam, respectively. The therapist taught the parents to implement treatment in their home with all six foods, with one parent as the therapist and the other as the model. The same conditions in effect during the final treatment phase of the study (modeling DR and NRS) were in place during follow-up sessions, and parents were trained to implement the procedures using BST. Parents were provided with written instructions that were reviewed by the first author, who then modeled implementation of the procedures with another trained graduate student. The parent then rehearsed the procedures with a graduate student, and the first author provided corrective feedback as necessary.
Results
Percentage of bite consumption for Larry and Adam is depicted in Figure 1. During modeling, neither participant consumed any foods. During the first session of modeling DR (Session 5), Larry independently accepted and consumed 67% (8 of 12 bites) of the foods presented. The first four bites were consumed after the therapist stated the contingency and before the modeling procedure was implemented. The next four bites were consumed after Larry contacted the modeling procedure (i.e., after he observed the model contact the programmed reinforcement). During subsequent sessions of modeling DR (Sessions 7, 9, and 11), Larry again independently accepted and consumed 8 of 12 bites; during Session 7, all 8 bites were independently accepted and consumed after the therapist stated the contingency and before the modeling procedure was implemented; during Session 9, the 8 bites were independently accepted and consumed after Larry observed the model contact the programmed reinforcement; and during Session 11, half of the bites were independently accepted and consumed after the therapist stated the contingency and before the modeling procedure was implemented, and the other half were consumed after Larry observed the model contact the programmed reinforcement.
Percentage of bites consumed for Larry (top panel) and Adam (bottom panel).
During modeling DR, Adam did not consume any foods with the exception of the first session during which he consumed one bite. When we added modeling NRS to the treatment, Larry and Adam consumed 100% of target foods. For Larry, all bites consumed during modeling DR and NRS sessions were consumed after the therapist stated the contingency and before the modeling procedure was implemented. For Adam, during the first modeling DR and NRS session, all 12 bites were independently accepted and consumed after Adam contacted the modeling procedure (i.e., after he observed the model be exposed to the NRS procedure and subsequently contact the programmed reinforcement upon finishing all bites on his plate). Thereafter, Adam independently accepted and consumed all 12 bites after the therapist stated the contingency and before the modeling procedure was implemented.
Consumption for both participants remained at zero with Group B foods until the therapist implemented modeling DR and NRS, at which time consumption increased to 100%. For Larry, during the first session during which the treatment was applied to Group B foods, all 12 bites were independently accepted and consumed after Larry contacted the modeling procedure. Thereafter, Larry independently accepted and consumed all bites after the therapist stated the contingency and before the modeling procedure was implemented. For Adam, when treatment was applied to Group B foods, all bites were independently accepted and consumed after the therapist stated the contingency and before the modeling procedure was implemented.
For Larry, the average percentage of intervals with inappropriate mealtime behavior for Group A foods was 16.19 for modeling, 8.23 for modeling DR, and 1.39 for modeling DR and NRS. For Group B foods, average percentage of intervals with inappropriate mealtime behavior was 19.49 for modeling, 8.59 during the modeling DR phase, and 4.94 during the modeling DR and NRS phase. During the final phase, when we applied modeling DR and NRS to Group B foods, the average percentage of intervals with inappropriate mealtime behavior was 2.23 for Group A foods and 13.61 for Group B foods.
For Adam, the average percentage of intervals with inappropriate mealtime behavior for Group A foods was 22.57 for modeling, 35.72 for modeling DR, and 0.69 for modeling DR and NRS. For Group B foods, average percentage of intervals with inappropriate mealtime behavior was 4.05 for modeling, 0.99 during the modeling DR phase, and 0.04 during the modeling DR and NRS phase. During the final phase, when we applied modeling DR and NRS to Group B foods, the average percentage of intervals with inappropriate mealtime behavior was 0.31 for Group A foods and 1.94 for Group B foods.
Discussion
Results of the current investigation demonstrated that a treatment package in which contingencies were both stated and modeled was a viable procedure in treating food selectivity for both participants. Modeling consumption alone was not sufficient in increasing consumption and decreasing inappropriate mealtime behaviors. These results suggest that stating and modeling the consequences of a behavior (as opposed to only modeling the behavior) is more likely to produce imitation (i.e., consumption). Of particular interest, modeling DR for consumption was not as effective as suggested by Greer and colleagues (1991) and Sira and Fryling (2012); this procedure was only successful in increasing consumption of two of the three foods for one participant (Larry). Modeling DR and NRS were found to be more effective in increasing consumption than modeling DR alone. The current study differed from the aforementioned studies in two ways: The confederate peer was an adult model as opposed to a sibling, and the model and participant were presented with the target foods simultaneously as opposed to eating opportunities being rotated between the participant and the model. It is possible that these procedural differences influenced the outcome. Future research should evaluate whether particular features of a model are relevant (e.g., age, gender) to the effectiveness of modeling contingencies as a treatment for food selectivity.
Results of the current study are promising; if stating and modeling contingencies effectively increases consumption without directly exposing the child to NRS, this would offer an effective intervention clinicians and parents could easily implement. One explanation for the increased consumption observed with Adam in the modeling DR and NRS phase is that, through generalized imitation, he had learned that what others do and its corresponding consequence signal the onset of direct contact with the same consequences. Thus, consumption is an avoidance response (a response that occurs prior to the onset of an aversive stimulus and is negatively reinforced through prevention of the aversive stimulus). This also may explain why less inappropriate mealtime behavior was observed during modeling DR and NRS, insofar as such behaviors are incompatible with consumption. This interpretation is supported by anecdotal observations of Adam’s vocalizations (e.g., stating, “Don’t worry about me”) after the modeling procedure and before consuming the target foods during modeling DR and NRS.
It is important to note that the component responsible for the increase in consumption cannot be determined given that two treatment components, statement of the contingency and modeling of the contingency, were implemented together as part of a treatment package. Therefore, one may argue that it was a combination of treatment components that involved both rules and modeling or that participants’ behavior was entirely under the control of rules. The latter explanation may have especially been true for Larry who independently accepted and consumed all bites during the modeling DR and NRS phase, after the therapist stated the contingency and before the modeling procedure was implemented. The likelihood of Larry’s behavior being solely under the control of rules is further supported by anecdotal observations of his vocalizations after the therapist stated the contingency in the beginning of the modeling DR and NRS sessions (e.g., “So if I don’t eat it, does that mean I have to sit here forever?”). Future research should isolate the effects of these treatment components and/or investigate whether modeling procedures alone are effective in increasing independent acceptance and consumption.
It is also important to note that the effects of modeling DR (without NRS) were not evaluated with Group B foods, and so we have limited information about the effectiveness of modeling DR alone. As previously mentioned, modeling DR was effective in increasing consumption of two of the three foods for one participant. It is possible that this procedure may be differentially effective based on the types of foods presented. In other words, some foods may be more disliked than others. The paired-choice preference assessments used in the current study may not be the best measure of preference when evaluating foods reported to be nonpreferred. Results of the preference assessments corroborated parents’ reports but did not provide information on relative distaste for the array of foods that were included.
Another limitation of the current study is that session duration was not held constant across conditions. Although, coincidentally, session durations for both treatment conditions were similar, modeling sessions were 6 to 8 min shorter than treatment sessions, raising the question of whether the behavior change was due to the modeling DR and NRS procedure or longer access to the treatment foods. Future research should hold session duration constant across conditions.
Footnotes
Authors’ Note
This research was completed in partial fulfillment of thesis requirements by the first author. A special thanks to Cathleen Piazza for her feedback on a previous version of this manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.