The sorts of reinforcer assessments described above are important in validating the predictions of preference assessments. Preference assessments are methods used to identify stimuli that may function as reinforcers. Stimuli shown to be more preferred are predicted to be more effective reinforcers . Thus, a typical course in applied research on preference assessment is to conduct the assessment to determine its predictions about relative reinforcer efficacy, then test those predictions using one of the reinforcer assessment methods just described.

Behavioral researchers have evaluated numerous methods of identifying stimulus preferences . The methods vary along many dimensions, including the effort required and accuracy of outcomes. Prior to the development of these methods, clinicians relied on staff or parent report or checklists (e.g., Atkinson et al. 1984; Cautela and Kastenbaum 1967) and similar methods sometimes collectively termed indirect preference assessments. These methods are more efficient in terms of time and effort than others, but their outcomes often correspond poorly with the outcomes of more rigorous assessments (Cote et al. 2007; Windsor et al. 1994). However, there may be some benefit to conducting informal assessments to identify stimuli to include in more systematic preference assessments (e.g., Fisher et al. 1996).

Pace et al. (1985) were among the first to describe a systematic preference assessment methodology. The researchers used a single-stimulus (SS) presentation format to assess the preferences for and reinforcer efficacy of various stimuli for six individuals with intellectual and developmental disabilities (IDD). Sixteen items thought to produce different forms of stimulation were included. During each trial, one item was placed in front of the participant and approach responses (i.e., moving hand or body toward the item within 5 s of presentation) were recorded. Preference hierarchies were established by calculating the percentage of approach responses for each stimulus. During the reinforcer assessment, the reinforcer efficacy of high-preference stimuli (those approached on 80 % or greater of trials) and low-preference stimuli (items approached on 50 % or fewer of trials) was assessed. The authors found that, in general, high-preference stimuli were more likely to function as reinforcers than low-preference stimuli.

The SS preference assessment was a relatively efficient method for directly measuring the preferences of individuals with severe learning deficits. One drawback, however, was that some participants approached the majority of stimuli that were presented. This pattern of indiscriminate responding implied that all stimuli were equally preferred by some participants. Alternatively, participant learning histories and subtle demand features of the SS preference assessment may have evoked approach responding regardless of the specific item presented. Thus, SS assessments may not yield information on relative preferences, causing clinicians to select some non-preferred stimuli as reinforcers.

To address the issue of false-positive findings, Fisher et al. (1992) developed a forced-choice or paired-stimulus (PS) assessment. In this preparation, two items were simultaneously presented to the participant, who could only approach one item during each trial. This methodological variation ensured that not all stimuli would be consumed during 100 % of trials, which increased the odds of the assessment generating differentiated preference hierarchies. In a comparison of the SS and PS methods, all items determined to be high preference (selected on 80 % or greater of trials) in the PS assessment were also identified as high preference in the SS assessment. However, items classified as moderate (50–79 %) to low (50 % or below) preference in the PS assessment were also frequently classified as high-preference stimuli in the SS assessment. Thus, the PS assessment generated more differentiated preference hierarchies than the SS assessment. During subsequent reinforcer assessments, stimuli determined to be highly preferred during both types of preference assessments supported higher rates of responding than stimuli identified as highly preferred during the SS assessment but low to moderately preferred in the PS assessment. These findings suggest that the PS assessment may offer a more accurate measure of relative preference than the SS assessment.

In an attempt to develop an assessment method that required less time to implement than a PS assessment while still providing information about relative preferences, DeLeon and Iwata (1996) proposed the multiple stimuli without replacement (MSWO) assessment. At the beginning of each session, the experimenter sat across a table from a participant and placed seven stimuli in a straight line approximately 5 cm apart and 0.3 m in front of the participant. The experimenter verbally instructed the participant to approach one item. After the participant approached one item, he or she was allowed to consume or play with that item. During the next trial, the selected stimulus was removed from the array and the remaining items were again laid out in front of the participant. Trials continued in this manner until the last item was approached, or the participant did not approach any of the remaining items within 30 s. Results obtained from the MSWO were compared to those obtained from a PS preference assessment. The researchers found that PS and MSWO methods generated similar preference hierarchies, but the MSWO assessment required far fewer trials.

Although most preference assessment procedures measure approach responding to stimuli presented across a series of trials, Roane et al. (1998) developed a brief duration-based, free-operant (FO) preference assessment. The authors noted that a brief FO assessment potentially had advantages over the traditional approach-based assessments like the PS and MSWO. They suggested it was quicker to administer, allowing for more frequent assessments; stimuli were never withheld or withdrawn, which might evoke problem behavior for some individuals; and although not specifically acknowledged by the authors, the FO method allows for the assessment of larger items that cannot be presented on the tabletop. During the FO assessment, sessions were 5 min in duration. Items were placed in a circle on the tabletop, and participants were free to engage with any of the items during that 5 min. Object manipulation was measured using 10-s partial interval recording. Preference hierarchies were established by ranking items according to the percentage of intervals in which object manipulation occurred. A brief concurrent-schedule reinforcer assessment followed the preference assessment. The researchers found that highly preferred stimuli (i.e., manipulated at the highest rates) were more likely to serve as reinforcers than less-preferred items. Furthermore, when compared to results obtained from a PS assessment, it was observed that the FO assessment was less likely to generate a distinct preference hierarchy (i.e., identification of at least one high-preference stimulus and at least one relatively less-preferred stimulus). However, the FO assessment was faster to administer and was associated with less problem behavior.

The previous sections have discussed different types of preference assessments and considerations in selecting a preference assessment method. Next we consider factors that may influence the types of stimuli one chooses to include in a preference assessment, and subsequently deliver in applied settings. Again, reinforcers can take many forms and some types are more commonly used than others. In a survey of the types of stimuli commonly delivered by individuals who work with persons with ASD and other special needs, Graff and Karsten (2012) found that 91.5 % of all respondents reported using social attention or praise. Tokens were used by 65.6 % of respondents, followed by breaks (65 %), edible stimuli (50.2 %), and toys (49 %). Community-based activities were least likely to be delivered (19.2 %).

When considering which types of these commonly delivered stimuli to include in a formal preference assessment, one may begin by conducting informal assessments, such as asking a caregiver to nominate highly preferred stimuli. For example, Fisher et al. (1996) sought to determine whether a structured interview form assessing caregiver nominations of preferred items would be useful in helping to construct a stimulus array for preference assessments. Six parents of children with disabilities were given a standard list of items and asked to rank order those items from most to least preferred. Parents were also provided with a carefully structured interview form called the Reinforcer Assessment for Individuals with Severe Disabilities (RAISD). The interview form instructed parents to name items they thought their child preferred in a number of categories, such as auditory stimuli, visual stimuli, edibles items, social stimuli, etc. Then, parents were asked to rank order items from the RAISD which they thought were most to least preferred for their child. PS preference assessments were conducted, and the preference hierarchies generated from the assessments were compared to parent rankings from both the standard list and RAISD assessments. The authors found that the top-ranked items identified by parent predictions based upon the RAISD were more preferred than the top-ranked items identified by parent predictions based upon a standard list of items. Thus, while caregiver reports may not consistently identify the most preferred items, they may play an important role in constructing a stimulus pool that includes the most effective reinforcers.

Although the effectiveness of a stimulus as a reinforcer is a critical consideration in arranging reinforcement contingencies, it is not the only consideration. As suggested in Fig. 11.2, another might be termed the ecological fit of the stimulus. By ecological fit, we refer to how well a stimulus fits into the use environment in which that stimulus will be delivered. Ideally, the stimulus used are those that fall into the upper right quadrant of Fig. 11.2, stimuli that are both effective and a good fit. A variety of characteristics of the stimulus determine how well it fits. For example, one should consider whether the stimulus is easily replenishable and relatively inexpensive, especially if one anticipates frequent delivery of the stimulus. The stimulus should be one that will remain effective for long periods of time. Duration of access to the stimulus should also be considered. If stimuli can only be delivered for short periods of time, then one may not want to include stimuli for which effectiveness hinges upon extended access (e.g., videos or games). In addition, one should ask whether it is likely to disrupt the environment or ongoing behavior. Clinicians understandably want to include the most effective reinforcers in their treatment plans. However, what is most effective may not always be the best fit for the individual’s environment. For example, community outings may motivate a student to complete his/her academic tasks and may effectively decrease problem behavior. However, they have the potential to be costly, can be difficult to arrange, and may not always be available. Similarly, provision of a movie contingent upon compliance may be disruptive to other students in the classroom. On the other hand, praise may arguably be associated with the greatest ecological fit (it is abundant and cheap, easy to deliver, is appropriate for almost any environment, and is not incredibly disruptive to ongoing behavior), but may not always be the most effective reinforcer. Therefore, it is important to balance both effectiveness and ecological fit.

DeLeon et al. (2013) recently created a flowchart to aid in reinforcer selection (Fig. 11.3) designed to be sensitive to this notion of ecological fit. They suggested that one begin by evaluating the effectiveness of social consequences (e.g., praise) as reinforcers because they may have the best ecological fit. If they are effective (or can be established as effective) under simple and remain effective under more stringent conditions (e.g., under thinner schedules of reinforcement), then they should be used as reinforcers. If social consequences are ineffective, nonedible tangible stimuli should be assessed next. Although food is easily delivered, one may prefer to use nonedible tangibles for a number of reasons. For example, frequent delivery of food may be associated with mounting costs and health concerns. Furthermore, food may not be appropriate in all environments, such as in the bathroom or during some medical procedures. If nonedible tangible stimuli are ineffective reinforcers, edible stimuli should be considered last. Token systems offer a number of advantages over immediate delivery of the actual reinforcer, including not disrupting ongoing responding, mediating delays between responding and reinforcer delivery, and being less subject to satiation because they can be exchanged for a variety of back-up reinforcers. Furthermore, tokens allow for accumulated access to reinforcers. For these reasons, one may wish to assess effective nonedible and edible stimuli within token systems. If token systems are ineffective, distributed reinforcement arrangements should be used.

Although relatively little research has been conducted on the effects of different rules for constructing the pool of stimuli for preference assessments, results from several studies suggest that composition of the assessment array can influence outcomes. For example, DeLeon et al. (1997b) conducted preference assessments that included food and leisure items in the same assessment. They then repeated the preference assessment without the food and assessed whether leisure items identified as low preferred during the mixed array functioned as reinforcers. Results suggested that food items often displace leisure items in mixed arrays but, when assessed separately, those leisure items may still be effective reinforcers. This study demonstrated that simultaneously assessing stimuli from multiple stimulus classes (edible, activity, social) can influence the obtained preference value of stimuli, and, importantly, may hinder the identification of other potentially effective reinforcers. Therefore, if different classes of stimuli are to be assessed, one should conduct separate preference assessments for each class.

Duration of access to the selected stimulus should also been taken into consideration when choosing stimuli to include in the preference assessment as it has also been shown to affect preference assessment outcomes. Steinhilber and Johnson (2007) conducted two MSWO assessments of seven activities. In the MSWO-short assessment, activities were available for 15 s following each selection. In the MSWO-long assessment, activities were available for 15 min. For one participant, the assessments produced disparate hierarchies. During a subsequent concurrent-chains assessment, when items were available for 15 min in the terminal link, the items identified as high preference on the MSWO-long assessment were more preferred than the item identified as high preference on the MSWO-short assessment. In contrast, when items were available for only 15 s in the terminal link, the items identified as high preference on the MSWO-short assessment were more preferred than items identified as high preference on the MSWO-long assessment. Thus, if the duration of access to an item in a preference assessment is substantially different than the amount of time that item will be available in the natural environment, preference assessments may not identify the most effective reinforcers.

The effectiveness of a reinforcer can refer to both its momentary capacity to support responses that produce it and its utility in producing long-term behavior change. The difficulty in actually quantifying effectiveness lies in that reinforcer effectiveness is dynamic; it changes as a function of a number of factors. For example, Neef and colleagues have extensively examined reinforcement parameters that affect response allocation, including delay to reinforcement (e.g., Neef and Lutz 2001; Neef et al. 1993), rate of reinforcement (Mace et al. 1994; Neef and Lutz 2001), and reinforcer quality (Neef and Lutz 2001; Neef et al. 1992). All else being equal, individuals will typically allocate a greater amount of responding towards the response option associated with more immediate reinforcement, higher rates of reinforcement, and better quality reinforcers.

In an extension of this line of research, DeLeon et al. (2011b) assessed the influence of contingency and amount of effort on the preference for and reinforcing value of four reinforcers for seven individuals with IDD. In this study, moderately preferred stimuli were assigned to one of four conditions: FR1, escalating FR (increasing effort across weeks), noncontingent delivery (without any earning requirement), and restricted access. Preference and PR assessments were conducted prior to and following 4 weeks of training with each of these conditions (with the exception of the restricted access stimuli, which were stored away during the 4-week training and only presented during subsequent preference and progressive-ratio assessments). Results were mixed across participants in that contingent stimuli (i.e., FR 1 and escalating FR conditions) and those stimuli associated with greater effort (i.e., escalating FR condition) were not always associated with increases in preference or reinforcer efficacy. However, consistent across all participants was a decrease in preference for the stimuli presented without an earning requirement. Furthermore, the smallest increase in reinforcer efficacy (i.e., lowest percentage change in PR breakpoints) obtained for the stimuli in the NCR condition. These results suggest that although effort may not necessarily increase the value of stimuli as reinforcers in persons with IDD, it is possible that noncontingent delivery may devalue stimuli more rapidly.

Results of studies such as those conducted by Trosclair-Lasserre et al. (2008) do provide some support for the notion that reinforcer magnitude may affect the value of a reinforcer. However, other researchers have observed little effect of magnitude on responding (e.g., Lerman et al. 1999, 2002). Trosclair-Lasserre et al. suggested that the effects of magnitude on reinforcer efficacy may depend on the schedule arrangement and schedule of reinforcement used. Specifically, magnitude may affect responding under concurrent-operant (e.g., Hoch et al. 2002; Steinhilber and Johnson 2007) or PR (e.g., Trosclair-Lasserre et al. 2008) arrangements and under schedules of reinforcement associated with increased response rates, such as VR schedules (e.g., Reed 1991). Many applied studies have used response rates under single-operant schedules to evaluate the relative potency of reinforcers. However, as previously mentioned, sensitivity to relative reinforcer value may be limited by ceiling effects. Individuals may respond as fast as possible regardless of the reinforcer (or magnitude) provided. Concurrent-operant and PR arrangements are not subject to these same ceiling effects and may therefore be more sensitive to differences in the relative reinforcer value of stimuli that differ with respect to magnitude. In addition, results of studies that incorporate concurrent-operant arrangements may be more clinically relevant than those using single-operant arrangements, particularly when one considers that individuals are constantly faced with multiple response options in the real world (Trosclair-Lasserre et al. 2008) .

The results of studies assessing the effects of magnitude on reinforcer value generally suggest that magnitude is an important variable to consider when it comes to reinforcer value. Furthermore, given that preference and reinforcer efficacy may vary as a function of reinforcer duration (e.g., Steinhilber and Johnson 2007), preference assessments should be conducted under conditions that more closely approximate how the reinforcer will be used in the treatment context. Lastly, magnitude may also play an important role when thinning schedules of reinforcement to make treatments more practical for use in the natural environment (e.g., Roane et al. 2007). One may wish to adjust reinforcer magnitudes as reinforcement becomes less frequent .