Descriptive analyses were conducted for problem behavior and potential precursors to problem behavior for 2 participants. Data from the descriptive observations were transformed into static (based on the entire descriptive observation) and dynamic (based on moment-to-moment changes) probability values. Results showed that conditional probability values of a precursor given problem behavior and of problem behavior given a precursor exceeded the unconditional probability values of a precursor or problem behavior for both participants. Results of the lag-sequential analyses, using the same descriptive data, showed that the probability of a precursor increased substantially in the 1-s intervals immediately preceding an instance of problem behavior. Following instances of problem behavior, the probability of a precursor decreased until it became relatively stable. In addition, the probability of problem behavior reached its highest absolute value in the 1-s interval following the precursor for both participants. Based on the results of the lag-sequential analyses, we concluded that the vocalizations of both participants were reliable precursors to problem behavior. Separate functional analyses of precursors and problem behavior showed that precursors and problem behavior were most elevated in the same functional analysis conditions, suggesting that both problem behavior and vocalizations were members of the same response class.
The present study involved the use of descriptive analysis methods to assess the extent to which changes in the probability of one response (vocalizations) were associated with changes in the probability of a second response (problem behavior). This type of analysis was recently employed by Hagopian, Paclawskyj, and Contrucci-Kuhn (2005)
to assess the relation between two responses (eye poking and stereotypy) exhibited by a young man with Down syndrome. Hagopian et al. compared the probability of eye poking given instances of stereotypy to the probability of eye poking given no stereotypy during observations in which treatment (protective goggles, continuous access to toys, and response blocking) was in place. Results showed that eye poking and stereotypy were most suppressed when blocking was implemented for both responses. Results of the study by Hagopian et al. demonstrated the effectiveness of an intervention that was informed by the analysis of response–response relations.
In addition to the results presented, we also evaluated the probability of a known reinforcer (tangible items for Paolo and tangible items or escape from instructions for Bruno) following instances of problem behavior and following instances of the precursor during naturally occurring interactions. Results of those analyses showed that, in their classroom environments, teachers did not provide access to tangible items (for Paolo and Bruno) or escape from instructional activities (for Bruno) following instances of the precursor or problem behavior at levels that exceeded the unconditional probability of those events. There are at least two possible explanations for this finding. First, the primary data collector also served in the capacity of supervisor for the implementation of classroom management strategies. It is possible that the teachers' behavior was influenced by the presence of the data collector; however, such reactivity effects might be expected to wane over repeated observations. Second, all classroom teachers were exposed to extensive behavior management training that emphasized minimizing attention, continuing presentation of instructions, and maintaining restricted access to tangible items following problem behavior. Had either the precursor or problem behavior reliably produced reinforcers in the current investigation, results may have been altered substantially. For example, had the precursor produced reinforcers during the descriptive observations, the conditional probability of problem behavior given a precursor would likely have been much lower. Future research along these lines might involve explicit comparisons between cases in which precursors are and are not reinforced during descriptive observations. Such comparative analyses may highlight the conditions under which probability analyses of the sort applied in the current investigation are most suggestive of response-response relations.
Results of the current study both extend and replicate findings reported by Smith and Churchill (2002)
. First, we conducted static and dynamic probability analyses based on extensive (formal) descriptive observations to assess the relation between potential precursors and problem behavior. As noted previously, Smith and Churchill concluded that responses were precursors based on informal observations. Second, the present results substantiate those reported by Smith and Churchill in that problem behavior and precursors were shown to be reinforced by the same events (i.e., access to tangible items for Paolo and access to tangible items and escape from instructional demands for Bruno). This suggests that under some circumstances, when supported by lag-sequential analyses of descriptive data, functional analyses of less severe response topographies may be used as a basis for designing function-based treatments to address more severe topographies of problem behavior (e.g., intense SIB). Third, as reported by Smith and Churchill, we also found that levels of problem behavior were suppressed when contingencies during the functional analysis were withheld for problem behavior and arranged for the precursor. Clinically, this is an encouraging finding because the potential for harm incurred during functional analyses of severe problem behavior might be reduced if the functional properties of problem behavior can be inferred from outcomes of a functional analysis of precursors.
Limitations of the current study also should be considered when interpreting the results. Unconditional probability values were calculated based on intervals of 1 s, but conditional probability values were based on intervals of 10 s, which in most cases would not produce similar outcomes. For example, if we assume a 60-s observation period in which 12 instances of behavior are equally distributed in each of six 10-s intervals (i.e., two instances per 10-s bin), the obtained probability values differ substantially. By assessing the unconditional probability of behavior in 10-s intervals, the obtained probability would be 1 (i.e., at least one instance of behavior occurred in each 10-s interval). On the other hand, by assessing the unconditional probability of behavior in 1-s intervals, the obtained probability would be .2 (i.e., 12/60). Because the two probability values were calculated using different methods, the unconditional probability of behavior may have been suppressed artificially (as evidenced by the aforementioned example). On the other hand, the analysis in terms of 1-s intervals may be a more representative account of behavior in that each instance of behavior and each opportunity to emit behavior are taken into consideration. Future research in this area may involve comprehensive comparative analyses of various methods of calculating descriptive data (e.g., methods of calculating unconditional probability values as described) to elucidate the merits and limitations of each.
An additional limitation of the current investigation involves the omission of a treatment component based on the results of the functional analyses. Thus, it is currently unknown whether a treatment based on the results of the functional analyses of precursors would have produced clinically significant reductions in either or both response categories. However, the extensive literature on function-based treatments suggests that such treatments would have produced improvements in behavior. Future research may be designed to evaluate treatments based on functional analyses of precursors identified via probability analyses of the sort evaluated in the current investigation.
In addition, the absence of inferential statistical analyses of the relations between precursors and problem behavior may be viewed as a limitation of the current investigation. Instead, we relied on visual analysis of comparative probability values. The rather pronounced differences were considered sufficient to draw conclusions from these data; however, researchers interested in the sequential association of response–response relations may evaluate odds ratios or Yule's Q,
as suggested by Yoder and Feurer (2000)
, in future investigations.
Although the functional analysis of severe problem behavior has proven extremely useful as a basis for treatment development, there remain circumstances in which such assessments are not possible due to the risk of harm to the participant or therapists. We do not suggest that precursor analyses supplant functional analyses of problem behavior, but that they may represent useful alternative procedures for minimizing risk during the assessment of particularly harmful response topographies.