Evaluations of heart rate during functional analyses of destructive behavior

Abstract

Previous research has examined the predictive validity of heart rate on destructive behavior; however, such research has yet to improve clinical practice or enhance our understanding of the relation between physiology and destructive behavior. The purpose of this study was to examine the predictive validity of heart rate on varying topographies and functions of destructive behavior while controlling antecedent and consequent events through functional analysis. We first demonstrated the reliability of the Polar H10 heart rate monitor and assessed the feasibility of its use in simulated functional analysis sessions. However, across four consecutively enrolled patients, heart rate was not found to be a reliable predictor of destructive behavior, regardless of its topography or function. Instead, functional reinforcer presence and absence was sufficient to predict socially reinforced destructive behavior. This study may provide a framework for the future assessment of other biological measures in relation to destructive behavior occurrence and nonoccurrence.

The prevalence of developmental disabilities in the United States has increased over the last 25 years (Boyle et al., 2011; Zablotsky et al., 2019). Due to this rise in prevalence, further research is required to understand the unique behavioral challenges present in this population. One such challenge is destructive behavior (Emerson et al., 2001). Destructive behavior often poses a safety risk (Crocker et al., 2006), a financial burden (National Institute of Health, 1989), and a threat to successful community integration (Harris, 1993). With destructive behavior being dangerous, costly, and a threat to accessing a higher quality of life, effective intervention of such behavior is a clear priority.

The field of applied behavior analysis has been successful in assessing and prescribing effective interventions for destructive behavior for individuals with developmental disabilities (Campbell, 2003). Using the functional analysis approach (Iwata et al., 1982/1994), behavior analysts uncover the environmental conditions responsible for destructive behavior occurrence and nonoccurrence through the manipulation of possible reinforcement contingencies. After identifying the function of destructive behavior, clinicians can bring about effective behavior change during treatment.

Functional analysis methodology has guided the prescription of effective function-based treatments for the last 40 years (Melanson & Fahmie, 2023). In a large-scale review of published functional analyses, Melanson and Fahmie (2023) identified 1,214 functional analyses published since a review conducted by Beavers et al. (2013). Although the majority (≥ 945, or at least 78%) of those functional analyses indicated that destructive behavior was maintained by social reinforcement, automatically reinforced destructive behavior was detected in a substantial number of functional analyses. Automatically reinforced destructive behavior poses unique clinical challenges because the inability to identify, and therefore control, the specific source of reinforcement complicates the development of function-based treatment (Vollmer et al., 1994). In cases where the antecedent conditions surrounding both destructive behavior occurrence and nonoccurrence are difficult to identify (e.g., Subtype 2 automatically reinforced self-injurious behavior [SIB]; Hagopian et al., 2015), physiological arousal may function as a predictor of destructive behavior onset for some individuals. As such, there is a small but growing body of research that incorporates measurement of physiological arousal in relation to destructive behavior in hopes that such data will improve the assessment and treatment process.

Incorporating biological measurement into the assessment process may provide added layers of information that could prove useful in better predicting, and potentially mitigating, occurrences of destructive behavior in situations difficult to do so currently based on the observation and manipulation of behavior–environment interactions alone. Such outcomes could be highly advantageous for at least some types of automatically reinforced destructive behavior but also could be helpful for the treatment of socially reinforced destructive behavior, particularly if changes in physiological arousal reliably precede occurrences of destructive behavior (i.e., function as a biomarker). Even a few seconds of change in a predictive measure of physiological arousal may provide sufficient time for caregivers to prepare for the onset of destructive behavior (e.g., clearing the environment of unsafe materials, moving to a safe room). Such forewarning also may allow for the possibility of adapting differential-reinforcement procedures to biologically based precursors to destructive behavior (e.g., providing escape contingent on a rise in heart rate that reliably precedes escape-maintained aggression).

Literature on the predictive relations between physiological arousal and the occurrence of destructive behavior suggests that such forewarning may be possible (Freeman et al., 1999; Goodwin et al., 2018). For example, Goodwin et al. (2018) examined the predictability of changes in physiological arousal prior to the onset of bouts of aggression for autistic children. Twenty participants who displayed aggression were observed naturalistically while admitted to an inpatient unit for individuals with developmental disabilities. All participants wore a device that recorded heart rate, heart rate variability, electrodermal activity, and body movement. Behavioral and physiological data were entered in real-time into a ridge-regularized logistic regression for binary decision-making in which data recorded across different units of time (e.g., last 3 min) estimated whether a bout of aggression was likely in the near future (e.g., next 3 min). Goodwin et al. found that this aggregated approach to data collection and analysis predicted bouts of aggression up to 1 min prior to its occurrence. However, these results were only achieved with complex data collection and analysis procedures, possibly decreasing the clinical usefulness of this approach. Furthermore, the naturalistic design of the study did not rule out alternative explanations of their results, nor were the function(s) of each participant’s destructive behavior reported or considered in relation to their findings. These limitations aside, the study by Goodwin et al. suggests that physiological arousal may predict occurrences of destructive behavior, but whether these findings are replicable and clinically useful given the complexity of the measurement and analysis approach used remains unknown.

Barrera et al. (2007) examined the heart rates of three participants who engaged in automatically reinforced SIB under the controlled context of a functional analysis. The researchers examined heart rate in relation to SIB by examining heart rate within 5-s windows of SIB occurrence. Heart rate was found to increase just prior to SIB and decrease just after SIB. This pattern of physiological arousal surrounding SIB occurred across participants, functional analysis conditions, and topographies of SIB. Thus, the data from Barrera et al. suggest that individuals who display automatically reinforced SIB may have a relatively uniform and predictable pattern of heart rate (i.e., a reliable heart rate waveform) surrounding SIB. However, it is important to acknowledge that only three individuals with SIB participated, and all three participants engaged in automatically reinforced SIB. Therefore, the generality of these findings should be assessed with additional participants who engage in similar topographies and functions of destructive behavior, as well as those who present differently.

Although studies examining the predictive validity of heart rate on destructive behavior are scarce, they have yielded some potentially fruitful results. However, such research has yet to improve clinical assessment or treatment procedures or improve our understanding of physiology and destructive behavior. The purpose of the present study was to address some of the limitations in prior research on this topic by examining the predictive validity of heart rate on varying topographies and functions of destructive behavior in the context of a functional analysis while using more clinically feasible data collection and analysis procedures. In a pre-experiment evaluation, we tested a common and cost-effective physiological measurement device by examining device reliability and assessing the feasibility of continuous heart rate data collection during simulated functional analysis sessions. We then examined the predictive validity of heart rate on the destructive behavior of four consecutively enrolled patients as they experienced the test and control conditions of a functional analysis of destructive behavior.

METHOD

Pre-experiment evaluation

Prior to conducting our experiment, we sought to independently evaluate the Polar H10 heart rate monitor, which is a cost-effective and commercially available device that has already been shown to be an accurate and reliable measure of heart rate¹ when compared to an electrocardiogram (ECG) Holter monitor (Gilgen-Ammann et al., 2019), which is considered the gold standard of empirically validated heart rate monitors. The Polar H10 heart rate monitor has a sampling and recording rate of once per second, and it consists of an elastic chest strap that contains two electrodes and a Bluetooth module that transmits data wirelessly to a separate electronic device (e.g., a smartphone). Data transmission and subsequent analysis requires the use of two affiliated applications—one called Polar Beat that is installed on a Bluetooth-compatible electronic device for collecting the heart rate data and another called Polar Flow that is a web-based program for extracting the data. We used the web-based Polar Flow application to export heart rate data streams as .csv files for subsequent data analysis.

Because the Polar H10 heart rate monitor is not a research-grade medical device, we conducted additional tests of the device to assess its reliability and the feasibility of its use with a clinical population by testing two identical devices under situations designed to mimic differing levels of physical activity (i.e., rest, walk, jog, sprint) and varying bouts (2 s, 6 s, 18 s, continuous) of each activity. The first author wore two Polar H10 heart rate monitors during one 5-min session for each of the following conditions: rest (continuous), walk (2 s, 6 s, 18 s, continuous), jog (2 s, 6 s, 18 s, continuous), and sprint (2 s, 6 s, 18 s, continuous). All non-continuous conditions programmed 20-s pauses between bouts of activity to simulate periods of reinforcer consumption and reduced levels of destructive behavior. These parameters were selected to simulate the range of intensities and durations of destructive behavior that may be seen in the experiment proper.

Results of this evaluation showed that both the primary and reliability Polar H10 heart rate monitors maintained electrode connection and transmitted heart rate data continuously across all conditions, regardless of activity level or duration. Inter-device reliability during these tests was exceedingly high, with Pearson’s r correlation coefficients averaging .99 (range: .98 to 1.0), indicating near-perfect correspondence between the two independently generated heart rate data streams. From these data, we concluded that the Polar H10 heart rate monitor was reliable and feasible for use with patients who engaged in destructive behavior.

Participants and setting

Participants were outpatients recruited from a hospital-based, university-affiliated program specializing in the assessment and treatment of destructive behavior. Inclusion and exclusion criteria stipulated that participants had to be between the ages of 3–17 years old and not wearing an implanted biological device (e.g., a pacemaker, gastronomy tube) that could interfere with heart rate data collection, wireless data transmission, or potentially jeopardize the correct placement or functioning of the implanted biological device. No prospective participants were denied participation due to these criteria. We enrolled four consecutive participants, who each happened to be 8 years old and male. All four participants carried a diagnosis of autism spectrum disorder. Additional demographic information for each participant is displayed in Table 1. All sessions were conducted in 3-m by 3-m padded treatment rooms within the clinic.

TABLE 1.

Participant demographics

Participant	Age	Race	Ethnicity	Diagnoses	Target Behavior	Functions Identified	Active Medications
Ash	8	White	Not Hispanic/Latino	ASD, ADHD, DMDD	PD	Automatic	Adderall, Risperidone, Clonidine
Clifford	8	White	Not Hispanic/Latino	ASD, ADHD	PD, AGG, SIB	Attention, Tangible, Social Control	Adderall, Risperidone
Leon	8	White	Not Hispanic/Latino	ASD, ADHD, Tourette’s Syndrome	AGG, PD	Tangible, Attention	—
Jean	8	Unknown/Not Reported	Unknown/Not Reported	ASD	AGG, PD	Tangible, Escape	Risperidone, Clonidine

Note. ASD = autism spectrum disorder; ADHD = attention deficit hyperactivity disorder; DMDD = disruptive mood dysregulation disorder; AGG = aggression; PD = property destruction; SIB = self-injurious behavior.

Materials

We used the Polar H10 heart rate monitor with all participants. The chest straps compatible with the Polar H10 heart rate monitor come in multiple sizes to accommodate differently sized individuals. The best-fitting chest strap for all four participants was the small size, which measured 51 cm to 66 cm. Therefore, all participants wore the Polar H10 heart rate monitor on the small-sized chest strap.

Measurement

Trained staff and students from the clinic and from the affiliated university collected frequency data on participant destructive behavior using BDataPro (Bullock et al., 2017). Self-injurious behavior was defined as any response directed toward the individual’s own body that could cause harm. Aggression was defined as any forceful response that could harm another individual. Property Destruction was defined as any forceful response that could damage an object or part of the environment. Data collectors also measured the duration of putative reinforcer presence. Reinforcer present was defined as the participant having access to the putative reinforcer programmed in the functional analysis condition.

Participants wore the Polar H10 heart rate monitor across all functional analysis sessions. No training or desensitization procedures were necessary, as all participants readily wore the device.

Interobserver agreement

Two independent observers recorded frequency and duration data simultaneously on participant destructive behavior and putative reinforcer presence, respectively, and did so for at least 25% (range: 25%–53%) of each participant’s functional analysis sessions. For frequency measures, we calculated exact agreement within 10-s intervals by scoring a 0 for all intervals without exact agreement between the two observers and a 1 for all intervals with exact agreement. We then averaged these scores across all observation intervals of the session. Interobserver agreement for frequency measures averaged at least 92% when separated by topography and participant. For duration measures, we calculated proportional agreement within 10-s intervals by dividing the smaller duration recorded by the larger duration recorded within each interval. Interobserver agreement for duration measures averaged at least 95% when separated by participant. Mean interobserver agreement coefficients are displayed in Table 2.

TABLE 2.

Mean interobserver agreement coefficients

Participant	Aggression	Disruption	Self-Injurious Behavior	Reinforcement Interval
Ash	100%	99%	100%	95%
Clifford	99%	92%	99%	99%
Leon	94%	97%	99%	96%
Jean	96%	98%	99%	97%

Design

This study is a prospective, consecutive controlled case series in which incoming patients meeting the study’s inclusion criteria (and failing to meet the exclusion criteria) were offered to participate. All eligible participants (n = 4) participated, and all consented participants completed participation.

Participation in this study was designed not to interfere with the routine assessment of each patient’s destructive behavior. Therefore, the specific test and control conditions comprising each participant’s functional analysis differed, as did the experimental design used to analyze the function of destructive behavior. These decisions were left to the provider responsible for each patient’s admission. The only departure from these routine clinical decisions was that each participant wore the Polar H10 heart rate monitor throughout their functional analysis. Relevant portions of each participant’s functional analysis are summarized herein for the purpose of assessing the predictive validity of heart rate on occurrences of destructive behavior.

Procedures

We conducted a functional analysis of each participant’s destructive behavior using procedures similar to those described by Iwata et al. (1982/1994). Common test conditions included no interaction, attention, escape, tangible, and social control. Common control conditions included toy play and test-specific controls (e.g., noncontingent access to highly preferred toys). All functional analysis sessions lasted 5 min, and reinforcement intervals lasted 20 s across all social test conditions.

Test conditions

During the no-interaction condition, the therapist provided no consequences following participant destructive behavior while monitoring participant safety. The treatment room was barren. During the attention condition, participants had access to a low-preference item and received high-quality attention from the therapist for approximately 1 min prior to the session. The session began with the therapist stating, “I have some work to do” and removing their attention. Destructive behavior produced 20 s of attention, during which time therapists honored participant requests to modify their play and placed no demands on the participant. During the escape condition, therapists used a three-step prompting sequence (i.e., vocal, model, physical) to guide the participant through completing academic or household instructions. Destructive behavior produced a 20-s break. During the tangible condition, participants had access to a high-preference item for approximately 1 min prior to the session. Sessions began with the therapist removing the item. Destructive behavior produced 20-s access to the previously withheld item. During the social-control condition (Clifford only), the participant had access to high-preference items and activities and could direct play with the therapist for approximately 1 min prior to the session. Sessions began with the therapist stating, “It’s my way” and initiating therapist-directed play. Destructive behavior produced 20 s of child-directed play.

Control conditions

During the toy-play condition, participants received high-quality attention from the therapist at least every 20 s and had continuous access to a high-preference item. Therapists honored participant requests to modify their play and placed no demands on the participant. Destructive behavior produced no stimulus change.

Clifford experienced test-specific control conditions. During the attention (control) condition, Clifford had continuous access to high-quality attention without demands or access to high-preference items throughout the session. Destructive behavior produced no stimulus change. During the tangible (control) condition, Clifford had continuous access to a high-preference item without therapist attention throughout the session. Destructive behavior produced no stimulus change. Finally, during the social-control (control) condition, Clifford had access to high-preference items and activities and could direct play with the therapist throughout the session. Destructive behavior produced no stimulus change.

Data analysis

We analyzed the data in multiple ways to explore the predictive validity of heart rate on destructive behavior. As a point of comparison, we also examined the predictive validity of a commonly measured event (i.e., presence and absence of the functional reinforcer) on socially reinforced destructive behavior to determine the degree to which heart rate predicted destructive behavior in relation to an established predictor (see Fisher et al., 2023, and Shahan, 2022, for discussion of reinforcer consumption as a competing response). As such, we examined differences in heart rate and destructive behavior given the presence or absence of the functional reinforcer for participants with socially reinforced destructive behavior.

To assess the predictive validity of heart rate on destructive behavior, we graphed and visually analyzed heart rate at each second from 15 s before to 15 s following each instance of destructive behavior, similar to a lag-sequential analysis (see Borrero & Borrero, 2008, for example applications). We then calculated the median heart rate at each second within these windows. To examine heart rate in relation to functional reinforcer presence or absence, we graphed the maximum and minimum heart rate values across within-session intervals of functional reinforcer presence or absence. We also assessed rates of responding across these same within-session intervals of functional reinforcer presence or absence. Finally, we conducted randomization tests with 100,000 repetitions each (see Craig & Fisher, 2019, for elaboration) to identify statistically significant differences in both heart rate and rates of destructive behavior when functional reinforcers were present or absent. All randomization tests were conducted in R (version 4.1.2; R Core Team, 2021).

RESULTS

Functional analysis results for all participants are displayed in Figure 1. Ash experienced an extended series of no interaction, and we observed variable but generally elevated rates of destructive behavior across sessions. From these data, we concluded that Ash’s destructive behavior was maintained by automatic reinforcement. For Clifford, we tested for attention, tangible, escape, and social-control functions using a series of reversal designs with test-specific control conditions. Clifford’s functional analysis results indicated that his destructive behavior was maintained by access to attention, tangibles, and social control, but not escape. Leon experienced the tangible, attention, escape, and toy-play conditions in a series of pairwise designs. Leon’s functional analysis results indicated that his destructive behavior was maintained by access to tangibles and attention but not escape. Jean experienced the typical test and control conditions in a multielement design, followed by a reversal design for further assessing escape. Jean’s functional analysis results indicated that his destructive behavior was maintained by access to tangibles and escape but not attention or automatic reinforcement.

FIGURE 1.

Functional analysis results. Asterisks denote sessions for which loss of heart rate data occurred (e.g., participant removal of the monitor, electrode misalignment resulting from participant movement and an imperfectly fitted strap, data transmittal error). RPM = responses per minute.

Figure 2 displays the heart rate data from 15 s before to 15 s following each instance of destructive behavior observed during the relevant test conditions of each participant’s functional analysis. Across participants, we did not observe a consistent pattern of heart rate surrounding instances of destructive behavior. When examining these heart rate data within-participant, Leon was the only participant who showed a discernable pattern across the 30-s windows. In both the tangible and attention conditions, Leon’s heart rate tended to be lower prior to instances of destructive behavior and higher following instances of destructive behavior. However, this pattern was not replicated with any other participant, regardless of function.

FIGURE 2.

Heart rate surrounding instances of destructive behavior. Black data paths represent primary measures of heart rate expressed as beats per minute (BPM) and are anchored around each instance of destructive behavior in the associated test condition of the functional analysis. Heart rates extend −15 s to +15 s around instances of destructive behavior. Black data paths represent median heart rate.

Figure 3 displays the maximum and minimum heart rate values measured within and across within-session intervals of functional reinforcer presence or absence for Clifford, Leon, and Jean. When examining both the maximum and minimum heart rate values across social functions of destructive behavior, a clear pattern of heart rate rarely emerged. Using the randomization tests described above, we observed a statistically significant difference between heart rate values when reinforcers were present or absent in only five of 14 comparisons (35.7%; see Figure 3 for results). In four of these five comparisons, heart rate was higher during periods of functional reinforcer presence. However, visual inspection of the data suggests that statistical significance in most of these cases was driven by the large number of data points within and across groupings, not by large effect sizes.

FIGURE 3.

Within-session analysis of heart rate. Minimum and maximum heart rate expressed as beats per minute (BPM) during within-session intervals of reinforcer absence (i.e., SR+ Absent, SR− Absent) and presence (i.e., SR+ Present, SR− Present). Open circles represent heart rates, black lines indicate median heart rates, and asterisks indicate presence and level of statistical significance (* = p < .05; ** = p < .01; *** = p < .001).

Figure 4 displays rates of destructive behavior across the same within-session intervals of functional reinforcer presence or absence displayed in Figure 3. Destructive behavior occurred almost exclusively when functional reinforcers were absent with the three participants for whom we identified a social function of destructive behavior. Unlike the heart rate data in Figure 3, destructive behavior was well predicted by the presence or absence of the functional reinforcer for these participants, with all seven comparisons (100%) showing a statistically significant difference using the randomization tests described above (see Figure 4 for results). Importantly, these differences appeared not to arise simply from large numbers of observations within and across groupings, indicating much larger effect sizes.

FIGURE 4.

Within-session analysis of destructive behavior. Minimum and maximum rates of destructive behavior expressed as responses per second (RPS) during within-session intervals of reinforcer absence (i.e., SR+ Absent, SR− Absent) and presence (i.e., SR+ Present, SR− Present). Open circles represent rates of destructive behavior, black lines indicate median rates of destructive behavior, and asterisks indicate presence and level of statistical significance (* = p < .05; ** = p < .01; *** = p < .001).

DISCUSSION

Despite taking several steps to ensure that the Polar H10 heart rate monitor was reliable and feasible for use with a clinical population that engaged in destructive behavior, results of the present study showed no reliable, predictive relation of heart rate on destructive behavior, regardless of its topography or function. Leon was the only participant for whom heart rate appeared to show a pattern before and after instances of destructive behavior. For both his tangible and attention functions of destructive behavior, heart rate was generally lower in the 15 s that preceded destructive behavior and higher in the 15 s that followed—a heart rate waveform opposite that found by Barrera et al. (2007). Similarly discernable patterns in heart rate surrounding instances of destructive behavior were not observed with the other participants.

Leon was the only participant who was not receiving psychotropic medication at the time of the study. Ash, Clifford, and Jean were each on a stable regimen of psychotropic medications, such as risperidone and Adderall, during the study. The degree to which such medications affect heart rate patterning under these and similar conditions is unclear, but it does seem reasonable that such medications could blunt otherwise reliable changes in heart rate. Stimulant medications (e.g., Adderall) often used to manage attention deficit hyperactivity disorder may be especially prone to moderating heart rate. Given the widespread use of one or more psychotropic medications to manage destructive behavior, future research on heart rate and other potential biomarkers for destructive behavior should report on medication use and consider the potential impact of each participant’s medication regimen on the associated findings.

We found that simple heart rate measurement was not a reliable predictor of destructive behavior; however, a more advanced (e.g., measuring heart rate variability) or comprehensive (e.g., measuring heart rate and heart rate variability) approach may have yielded a different result. As previously noted, Goodwin et al. (2018) collected various measures of the autonomic nervous system, and when analyzed using an advanced algorithm, this more advanced and comprehensive approach could predict the occurrence of destructive behavior within the upcoming minute. Future researchers should attempt to identify the least cumbersome approach to physiological measurement that maintains a reliable and clinically meaningful degree of predictive validity for destructive behavior. Identifying which measures and analyses are necessary and sufficient to reliably predict various topographies and functions of destructive behavior could prove to be an interesting and important line of research. Directly replicating the study by Goodwin et al. and then systematically paring down the sophistication of their approach could be one promising strategy.

We assessed the predictive validity of heart rate as a biomarker for destructive behavior using functional analysis methodology as a means of controlling relevant environmental variables (i.e., functionally relevant antecedent and consequent events) known to occasion and suppress destructive behavior. However, one limitation of doing so is that the continuous reinforcement schedule programmed during functionally relevant functional analysis test conditions confounded occurrences of destructive behavior with the delivery of the functional reinforcer for destructive behavior. As such, we cannot be certain that changes in heart rate observed following instances of destructive behavior in these sessions were due to imminent or recent destructive behavior or were the result of delivering the functional reinforcer. One point worth noting is that because destructive behavior often ceased following the delivery of the functional reinforcer (see Figure 4), further occurrences of destructive behavior during this time rarely confounded changes in heart rate. In this sense, control over the availability of the functional reinforcer imposed by the functional analysis procedures discouraged one potential confound (i.e., heart rate changes from increased activity associated with nonreinforcement and continued destructive behavior) but created another (i.e., heart rate changes affected by decreased activity associated with reinforcer delivery and the absence of destructive behavior). There is no clear solution to this problem, but one possibility could entail measuring the co-occurrence of heart rate and destructive behavior under yoked schedules of intermittent reinforcement that produce different rates of destructive behavior but similar schedules of obtained reinforcement. However, the data resulting from such an arrangement may not be particularly useful in guiding future research or practice.

One limitation of the present study is that heart rate data loss occurred in a minority of sessions for three of the four participants. The reasons for heart rate data loss varied within and across participants but often involved participant removal of the heart rate monitor, electrode misalignment resulting from participant movement and an imperfectly fitted strap, or a data transmittal error. Fortunately, the heart rate data that were lost did not meaningfully affect whether we could address the aims of the present study. Future researchers looking to extend this and related lines of research should anticipate similarly imperfect collection of physiological data. Heart rate data loss across participants averaged 13.3% of sessions (range: 0%–26.5%). Although doing so in the context of the present study would have been difficult given our desire not to deviate from the routine functional analysis procedures conducted with each participant, future investigators should consider overestimating the number of sessions needed to assess other potential biomarkers to account for similar levels of data loss (e.g., conservatively increasing the number of sessions by approximately 25%).

Although heart rate was not predictive of destructive behavior regardless of the various forms of data analysis conducted, we found that within-session intervals of functional reinforcer presence or absence reliably predicted the destructive behavior of all three participants whose responding was socially reinforced. Our findings suggest that simple measurement of the availability of a functional reinforcer may provide information on the best-known predictor of socially reinforced destructive behavior. Despite the lack of a consistent, predictive relation between heart rate and occurrences of destructive behavior, future research that parses out the potentially mediating effects of psychotropic medication may be an important next step in this line of research. Additionally, future researchers should assess other potential biomarkers of destructive behavior. Neurological measurement specifically in individuals with automatically reinforced destructive behavior appears to be a particularly worthwhile area for future research. In conducting such work, the procedures and methods of data analysis outlined in the current study may provide a procedural framework from which future researchers may continue to assess biological processes in relation to destructive behavior.