CARDIOVASCULAR AROUSAL IN INDIVIDUALS WITH AUTISM SPECTRUM DISORDER: AN IDIOGRAPHIC ANALYSIS

Stress in individuals with autism spectrum disorder (ASD) is poor ly understood, yet can be detrimental to the functioning of these individuals. Stressrelated problems are more common in ASD than the typical population, and individuals with ASD of ten have poorer coping skills. It is crucial to understand stress responses in these individuals, to help them better learn, cope, and prevent problem behavior associated with stressful events and heightened arousal. However, traditional measures of s tress (e.g. self-reports) are often unreliable in this population, due to communication deficit s in ASD. Studying physiological responses is an alternative, potentially more acc ur te, way to study stress in ASD. This idiographic study systematically examines heart rate ( HR) responses to six stressors in 39 individuals with ASD. Patterns of response for eac h individual are discussed. Examples of four hypothesized physiological subtype re sponders were identified. These subtypes include: hyperarousal (characterize d by high baseline HRs, with low variation in response to different stressors), hyporesponsiv e (characterized by low/normal baseline HR, with low variation in response to different s tres ors), reactive responsivity (characterized by HR that increases significa tly throughout the assessment and fail to return to baseline level), and normal responsivity (chara terized by normal baseline HR that varies during stressor phases, but returns to base line level during subsequent baseline phases). Clinical and general implications of t hese findings are discussed, as well as directions for future research.

and is the fastest growing developmental disability (California Health & Human Services, 2003). Because of its heterogeneous nature, and purely behavioral definition, ASD is likely to have multiple possible etiologies that are not fully understood (Eigsti & Shapiro, 2003). Studying ASD not only sheds a light on the disorder itself, but can also improve understanding of normal functioning and development (Cicchetti & Rogosch, 1996).

Comorbidity in ASD
Individuals with ASD are more likely than the general population to have a range of comorbid diagnoses. Seventy-five percent of individuals with ASD also have mental retardation (MR), while 25% have intellectual abilities that range from low average to above average (Eigsti & Shapiro, 2003). These individuals are at higher risk for seizure disorders (20-30% lifetime prevalence), which is even more likely for those with MR (Rapin, 1996). Individuals with Fragile X are also at greater risk (3-25% incidence) for ASD (Baileyet al. 1993). In addition, multiple studies have shown higher rates of stressrelated problems in ASD than the general population, including: anxiety (Bellini, 2004;Gillot, Furniss, & Walter, 2001;Gillot & Standen, 2007;Kim, Szatmari, Bryson, Streiner & Wilson, 2000;Muris, Steerneman, Merckelbach, Holdrinet, & Meesters, 1998), depression (Kim et al., 2000), and fears and phobias (Evans, Canavera, Kleinpeter, Maccubbin, & Taga, 2005;Knapp, Barrett, Groden & Groden, 1992;Matson & Love, 1990). Gillot & Standen (2007) note that compared to typical adults, adults with ASD have more difficulty coping with change, anticipation, sensory stimuli, and unpleasant events. Wood & Gadow (2010) suggest that stress may moderate ASD symptom severity (e.g. social skills deficits, and repetitive behaviors). Since individuals with ASD are likely to experience stress-related problems, it is crucial to understand how individuals with ASD experience stress. Due to the heterogeneity of ASD, and the likelihood of comorbid diagnoses, it is also necessary to acknowledge that stress experience varies by individual.

Stress and ASD
According to Selye (1974), stress is the physiological reaction of the body to either positive or negative events, or stressors. Stressors are events that place a demand on an organism and require an organism to make an adjustment to maintain homeostasis (Lazarus & Folkman, 1984). Groden, Cautela, Prince, & Berryman (1994) propose that individuals with ASD are at greater risk for experiencing high stress levels, and respond to stressors differently than the typically developing population. This may be due to social and communication deficits, as well as difficulty adapting to new situations. As many as 50% of individuals with ASD fail to develop spoken language (Bryson, Clark, & Smith, 1988) making it very difficult to communicate feelings of anxiety. Stress and anxiety can affect the cognitive, behavioral, and physiological responses of people with ASD (King, Hamilton & Ollendick, 1994). Therefore, it is crucial to understand how stress affects these individuals, in order to improve their quality of life, create targeted interventions and prevention programs, and better understand the nature of ASD.

Assessment of Stress in ASD
Self-Reports. While self-reports are a commonly used tool to assess stress in typical populations, many individuals with ASD have communication deficits that make self-report measures unreliable (Hill, Berthoz, & Frith, 2004). Two alternatives for measurement are parental and caretaker ratings, and physiological measures.
The Stress Survey Schedule. The Stress Survey Schedule (SSS) developed by Groden & colleagues (2001) and validated by Goodwin & colleagues (2007) Physiological Measures. Another alternative to stress measurement is physiological measurement. It has been suggested that passive, physiological measurement is especially appropriate to use with this population due to heterogeneity in ASD in regard to chronological age, developmental level, and linguistic and sensorimotor skills and capabilities, and potential behavioral/physiological dysynchrony . Autonomic nervous system (ANS) arousal is a good physiological indicator of one's stress level at rest and in the presence of different stimuli. If the system becomes aroused, changes in the cardiovascular system, immune system, endocrine glands, and brain regions involved in memory and emotion occur (Sapolsky, 1998). Cardiovascular activity (including HR) is a commonly measured ANS stress indicator (Andreassi, 2000).
HR quickens to more intense stimulation and slows to less intense stimulation, which is presumed to be a defensive response to perceived danger (Lacey & Lacey, 1958). Kootz & Cohen (1981) suggested that a heightened ANS activity is indicated by high HR.
Lower HR indicates focused attention, and blockade of external stimuli, also called an orienting response (Cohen, & Johnson, 1977). Romanczyk & Matthews (1998) proposed physiological state could be an antecedent to problem behavior often seen in ASD, and Freeman, Horner, & Reichle (1999) demonstrated HR changes before, during and after episodes of self-injury, aggression, and other problem behaviors in individuals with developmental disabilities.

The ANS
The ANS is comprised of two separate systems: the sympathetic nervous system (SNS) and the parasympathetic nervous system (PNS). SNS responses in the presence of stressors include increased HR and respiration, pupil dilation, increased perspiration, inhibition of salivation and digestion, increased respiration, blood pressure increase, inhibition of reproductive organs, and adrenaline discharging into the system. Once the perceived threat has passed, the PNS constricts the pupils, stimulates salivation, decreases HR, slows respiration, stimulates digestive activity, and stimulates reproductive organs enabling a return to a homeostatic state (Sapolsky, 2002). Gellhorn (1957) suggested that trying to maintain balance between the SNS and PNS activates either system, but also excitation in one system may result in activation of the complementary one. He suggests that PNS activity could directly relate to the specific intensity, frequency, and duration of the preceding SNS stimulation. This is called the "principle of reciprocity" and involves maintaining neurochemical homeostasis between dynamic branches of the ANS. (1962) suggests that individuals with ASD may have deficiencies in regulation between the two ANS branches. Compared to children with ASD, typicallydeveloping children were found to have significantly greater capacity for SNS activity, greater reactivity to their environment (specifically to changes in stimulation), and greater capacity to inhibit this reactivity and return to a state of homeostasis. Hirstein, Iverson, and Ramachandran (2001) suggested that the ANS in individuals with ASD cannot regulate itself appropriately, and requires additional behaviors for regulation (i.e. selfinjurious and stereotyped behavior). Porges (1976) suggests that studying this autonomic imbalance using physiological measures in ASD, early in a child with ASD's development, may facilitate positive and successful intervention.

Responsivity to Stressors in ASD
Difficulty in modulation, or hyper-/hyporesponsivity to stimuli can lead to a range of problems in ASD. Hyperresponsivity (responding inappropriately with high arousal levels to innocuous stimuli) or hyperarousal (being in a chronically high state of heightened arousal) may lead to behavior problems such as self-injury and aggression, which interferes with learning and attention, and may require pharmacological intervention (King, 2000). Hyporesponsivity may make an individual appear to be lethargic or unfocused, and could also interfere with learning. Understanding how people with ASD experience stress is integral to improving their quality of life. If stress responses are better understood in this population, it may be possible to help these individuals better deal with stress, so that they are able to better focus their attention, learn, and reduce problem behaviors and the likelihood of developing other diagnoses (such as anxiety, and mood disorders).

Physiological Subtypes in ASD
Findings from a few physiological studies suggest subgroups exist in ASD. Cohen and Johnson (1977) identified three subgroups. One small subgroup had normal greater within group differences among individuals with ASD than between group differences in ASD compared to a control group Kootz & Cohen, 1981). These findings suggest that it may be more appropriate to analyze physiological data in ASD idiographically, rather than nomothetically. Baranek (2002) suggested that identifying specific individual physiological patterns that differentiate responder types would be very useful when planning interventions in ASD.

Limitations of Prior Research
While informative, previous physiological studies in ASD consist of small samples and vary widely in their use of physiological measures and experimental stimuli, making it difficult to generalize findings. Experimental stimuli also usually consisted of one or few potential stressors. Older instruments used to measure physiology often required that the participant must restrict movement (Kootz & Cohen, 1981;Stevens & Gruzelier, 1984), which would most likely prove quite difficult for the majority of participants with ASD, resulting in error. Many studies were also published before publication of the DSM-III, potentially resulting in non-ASD individuals being included in their samples. Most studies use group-level analyses, which can wash out effects (i.e. high and low responders will be averaged together, and look "normal"). The current study attempts to overcome these limitations by analyzing data from a larger sample of individuals with ASD (N=39) assessed on one physiological measure (HR) while exposed to a standardized variety of potential stressors. Use of time series analysis (TSA) at the idiographic (individual) level can provide detailed data on individual stressresponse patterns.

An Idiographic Analysis of HR in ASD
No studies to the author's knowledge look at multiple stimuli with many replications across many individuals, but one study examines cardiovascular responses to a variety of potential stressors identified by the SSS, in a small sample of individuals with ASD and a typically-developing age-/sex-matched control group (Goodwin et al., 2006).
They found that individuals with ASD have higher baseline HR and less HR variability to different stressors than the control group. Also, individuals in the control group had more significant responses to different stressors than the individuals with ASD. This indicates that some people with ASD may be in a constant state of cardiac over-arousal, and may experience high levels of stress on a more continuous basis than those in the typical population. Another explanation could be that the individuals with ASD in this study were a subset of individuals who exhibit hyperarousal, and that other response patterns exist in ASD.

This Investigation
The present study, a secondary data analysis, replicates and extends the Groden et al. (2005) and the Goodwin et al. (2006) studies by examining clinical HR assessment data from 39 students enrolled at the Groden Center Day School, Providence, RI; a program serving the academic and behavioral needs of children with developmental disabilities. These assessments contributed to a functional behavior assessment by identifying stressors that may serve as antecedents to problem behavior for each of the 39 participants. These assessments provide 39 replications, more than the typical 5 or 6 replications typically recommended in single-case design research (Barlow & Hersen, 1982).
This study will: 1) Explore variation in individual HR responses to stressors specifically by number, type, and combination of significant responses.
2) Examine individual patterns of responses reflective of four predicted subtypes: hyperarousal-high baseline HR and low variation in response across stressors; hyporesponsive-low/normal baseline HR and low variation in response across stressors; reactive responsive-HR increases throughout the assessment, and fails to return to baseline level; and normal responsivenormal baseline HR with some variation in HR during stressor phases, but HR returns to baseline level during subsequent baseline phases.

Participants
Participants included 39 former and current clients (males=33, females=6) from the Groden Center Day School. Written consent from guardians of each participant was obtained to collect these data. Participants ranged in age from 3 years 2 months to 19 years 11 months (m= 11 years 10 months, median= 12 years 7 months) (See Figure 1 for frequency of participants in each age range). Only participants who had a primary or secondary diagnosis of ASD made by a licensed psychologist familiar with the DSM-IV were included in this study. Thirty-six participants (92%) had a primary diagnosis of ASD, and 3 (8%) had a secondary diagnosis of ASD (See Table 1). All participants with available blood pressure data were normotensive (<90 mmHg diastolic blood pressure) (See Table 2  Socialization it was 2 years 2 months. Twenty participants were on at least one medication at the time of the assessment (not all of which affected ANS arousal), 10 were on none, and data were not available for nine. No participants had low baseline HRs, and 6 had baselines that were high for their age (Participants 4,7,11,25,33,and 36), and three of these participants only had slightly above average heart rates (Participants 25, 33, and 36). Of these six participants, only Participants 7 and 11 were on medications that could have raised their heart rate. Ways to compensate for this will be discussed further.

Multicultural Representation
Participants include individuals with ASD, as the goal is to see individual patterns of cardiovascular response to a variety of potentially stressful stimuli. According to the American Psychological Association (2000), males are four to five times more likely than females to have ASD. This accounts for more males (n=33) being included in this study than females (n=6). Seventy-nine percent of participants were Caucasian (n=31), and 21% participants were racial/ethnic minorities (African American (n=3), Latino (n=4), and Asian American (n=1)) (See Figure 2).

Setting
Assessments took place in a sound-attenuated laboratory room with plain white walls, low incandescent lighting, a neutral-colored carpet, and a one-way mirror (to allow discrete viewing from an adjacent observation room). The glass was covered by a blind, so that participants were not distracted by their reflections.
This non-invasive vest telemetrically recorded HR, respiration, electrocardiograph (ECG) data, and motor movement. Data were continuously stored (i.e., beat-to-beat) on a portable battery-powered electronic recorder worn on the body. Motor movement and posture changes were recorded by a dual-axis accelerometer inside of the Lifeshirt positioned on the anterior surface of the ribcage. Movement data were collected to control for HR changes due to increased physical demands. See Wilhelm, Roth & Sackner (2003) and Heilman & Porges (2007) for a more complete description of this system, including reliability and validity data. Groden et al. (2005) found that individuals with ASD could tolerate the Lifeshirt system well. Data were collected and transferred onto a personal computer, were exported into Excel, and were later analyzed in SAS.

Materials
A familiar staff was present during the assessments and was given a sheet listing the phases. A vacuum, remote control car, edible, two small dish towels, and stationary bike were used during the phases of the assessments. Researchers recorded start and end times of each phase using a data sheet and stopwatch in the adjoining room to the laboratory.
All assessments were videotaped using a discrete camera mounted in the upper corner of the lab room. A cushioned chair was provided for participants. Across the room was another chair for the familiar staff. There was a rectangular table pushed against the wall during most of the assessment (except for "Difficult Task," which required the table to be moved between the participant and staff).

Original Data Collection Procedure
The current study is a secondary data analysis of HR data that were collected as part of a clinical assessment that is a regular part of the intake assessment at the Groden Center. The intake assessment is a two month period where students become acclimated to the center. HR analysis became a regular part of this assessment to integrate HR data into a functional analysis of behavior; to provide a physiological and behavioral baseline for evaluating program interventions over time; and to increase understanding of individual differences in individuals with autism.
Most participants had been assessed during their first two months at the Center (n=24, 62%). In some cases, assessments were delayed by unavailability of equipment.
The majority of assessments were done within the participants' first year at the Center (n=33, 85%). Only two participants (5%) had their assessment between one year to one year and six months at the Center, two (5%) had it between two to four years at the Center, and 2 (5%) had it after 10 years at the Center (See Table 2).
Before the assessments, all participants went through a familiarization period. This served to increase comfort level related to the lab room and the Lifeshirt before participants had their assessments, and also to control for the novelty of these factors accounting for HR changes. Accompanied by a familiar staff, all participants were introduced to the lab room as well as the Lifeshirt, at least one time before their assessment. Number of visits varied depending on the needs of the participant. Once the participant had at least one exposure, and appeared to grow comfortable with the room and Lifeshirt, they underwent the full HR assessment.

HR Assessment Description
For the HR assessment, the client entered the now familiar lab room with a familiar staff. A researcher put on the Lifeshirt, and connected three adhesive electrodes to the participant (on the left and right upper part of the chest, and on the left side of the torso, below the ribcage). The client was then seated in a chair across the room from their staff. Participants were instructed to sit quietly, at which point an initial baseline phase began. This phase served as a comparison to subsequent phases (stressors and rests) during statistical analyses.
The assessment consisted of 14 phases and followed an A1BA2CA3DA4EA5FA6GA7H design, where A represented a baseline phase, B through G represented stressor phases, and H was an additional phase described below (See Figure 3 for a detailed description of each phase). A fixed order was used across all assessments to maximize comparability of exposures. Six stressor phases were adapted from five domains of the SSS to be examined empirically. These domains included: and consisted of a pool of all time in-between stressor and baseline phases. This HR assessment was well tolerated in individuals with ASD (Groden et al, 2005).
Each phase was two minutes long, with the exception of the initial baseline, which was five minutes long. The first three minutes of the initial baseline were not included in the analyses to allow participants time at the beginning of the assessment to grow acclimated with the environment. Once these data were discarded, all phases were equal in length allowing for TSA to be performed. While the length of phases was standard for 21 participants, 18 participants had shortened phases presented in the same order (after confirming that HR responsivity was not statistically significantly different between two minutes and one minute). For these participants, the initial baseline was two minutes, while each subsequent baseline and stressor phase was one minute. This shortened assessment was given to very young participants, or when a familiar staff requested this assessment due to special behavioral concerns for the participant. One participant (29) required Unstructured Time phase to occur later in the assessment than typical for safety reasons.

Analyses
Data analyses consisted of 39 separate univariate interrupted time series analyses (Crosbie, 1993;Glass, Willson, & Gottman, 1975;Velicer & Colby, 1997;Velicer & Fava, 2003) performed on each participant for the dependent variable HR. Time series analysis can model change over time and requires a large number of observations at equally spaced intervals. In TSA, sample size is the number of observations over time rather than the number of subjects. Each full-length HR assessment generated over 3,000 data points per participant. However, since all data were taken from a single participant, there is serial dependency in the data. Group level analyses assume that data are independent. Therefore, for the present study, traditional group-level analyses are inappropriate unless the data are transformed to be independent. TSA addresses the issues of dependency in the data by determining the degree of autocorrelation that transforms the data to be independent. After this transformation dependency is removed and standard general linear model procedures can be employed.
Time series is a regression-based technique that uses an autoregressive integrated moving average (ARIMA) models of the order (p, d, q) to model the serial dependence of the data. The p represents the autoregressive term that shows the degree to which the data are dependent on previous observations. The d term represents the number of times a series has to be differenced in order to make it stationary. The q represents the moving average term that describes the persistence of a previous shock to the system (Box & Jenkins, 1970). Velicer & Harrop (1983) caution that the correct ARIMA model underlying a time series is difficult to determine. Therefore, this study employs the General Transformation Approach (Velicer & McDonald, 1984). This approach uses an ARIMA (5, 0, 0) model for all TSA and has been shown to adequately approximate most commonly encountered time series analyses in the behavioral sciences (Velicer & McDonald, 1984). Missing data were handled using the maximum likelihood procedure, which has been identified to best approximate missing data, when compared to other procedures, with up to 40% of data missing (Velicer & Colby, 2005).
For the present study, PROC ARIMA was used in SAS. The dependent variable was HR. Shape, level, and variability were examined. T-tests were done on all data to test for significant differences between stressor phases and the initial baseline phase.
Results for significant phases are reported as a change in level (stressor phase mean HRbaseline mean HR=change in level).
Power Calculation. Based on Goodwin et al. (2006)

Individual Results
Participant 1

Chapter 4. Discussion
This study went beyond the current body of related literature by idiographically examining 39 replications of cardiovascular responsivity in individuals with ASD to a variety of systematically selected stressors. This is the only study, to the author's knowledge, that has idiographically examined physiological responses in a large group of individuals with ASD, and identified potential subtype responders (warranting future confirmatory analyses). Much of the prior research in this area examined a small number of participants using nomothetic methods to compare individuals with ASD to control groups. Group level analyses washes out high and low responders. Idiographic analyses allow examination of each responder. This allows for tailored interventions for the needs of individuals, before, during, or after exposure to a stressor to help the individual cope, learn, and reduce problem behaviors.
Based on mixed research findings, prior research suggestive of subtype physiological responders in ASD (Cohen, & Johnson, 1977;Hirstein et al, 2001), and findings of larger intraindividual than interindividual variation in ASD when compared to a control group when using group-level statistics Kootz & Cohen, 1981), idiographic analyses to identify individual patterns of physiological response to stimuli was warranted. As expected, individual HR patterns varied. Typically very high autocorrelations were found for participants (around .95 for the first lag). This result would likely be the expected pattern for HR data taken from very short intervals.
There were four hypothesized responder types in this study. Examples of each were identified. These subtypes include: hyperarousal (i.e. have a high baseline HR, and low variation in response across stressors), hyporesponsive (i.e. have a low/normal baseline HR, and low variation in response across stressors), reactive responsive (i.e. HR increases throughout the assessment, and fails to return to baseline level), and normal responsive (i.e. normal baseline HR that varies during stressor phases, but returns to baseline during subsequent baseline phases). These findings warrant future investigation, discussed below.
Individuals who fit the hyperarousal subtype could include teaching relaxation techniques, and to prompt using these techniques multiple times in a day. Individuals who fit the hyporesponsive subtype may need to do physical or sensory activities to get their arousal up so that they are better able to focus throughout the day. Individuals who fit the reactive responsive subtype would also probably need to learn relaxation techniques that are used throughout the day, especially before and after events that are known stressors. Individuals who fit the normal responsive subtype could also benefit from relaxation strategies that are used specifically prior to exposure to known stressors.
One limitation of the current study was that it was a secondary data analysis.
Demographic and medical information (i.e. medications) were difficult to gather (as many participants had left the Center, and some data had not been collected close to the time of the assessment), which makes it difficult to compare individual results with respect to these data. Follow-up studies will further examine individual results with respect to different participant characteristics outlined in Table 2. Medications are one factor that can affect HR. Although many participants were on medications, the majority of participants (n=33) had HRs that were in the normal range for their age at baseline.
Only six had high baseline HRs (participants 4, 7, 11, 25, 33, and 36), and none had low baseline HRs. All participants with high baseline HRs were over 10 years old. Follow-up investigation could attempt to partial out HR changes due to medication, or could include only participants not on medications (however, this will substantially decrease participant pool, and may not be representative of individuals with ASD).
It is possible that HR changes may have been due to being in a laboratory setting, rather than the different stressor phases, per se. However, this may have been controlled by the familiarization period with the lab and Lifeshirt, and also by being accompanied by a familiar staff at all times. It is possible that baseline HR here was not a true indicator of one's resting HR. The first three minutes in the lab may not have been enough time for all individuals to habituate to their environment, or HR may have been artificially high due to being observed in an artificial setting. Attempts were made to control for this with the familiarization period, and the three minutes at the beginning of baseline that are discarded. Also, sitting quietly in a comfortable chair with a noninvasive vest and a familiar staff were other attempts to control for this. Only six individuals had high baseline HRs, so it doesn't appear that this was a potential problem for most. This study does not correlate overt behavioral responses with physiological responses, although individuals with high HRs during the assessment often showed little to no overt behavioral signs of distress. It would be informative to systematically investigate if there is synchrony or dysynchrony between behavior and physiology in ASD, and a follow-up study could be done examining the correlation between behavior and physiological measures. .
HR is a robust measure of arousal, however, HR alone does not reveal which system (SNS or PNS) is controlling HR responses. HR variability (HRV) is a measure that allows one to infer which system is working (or may be deficient). HRV data were collected during all sessions, and a follow-up study will be done examining HRV responses in these 39 individuals.
With ideographic analyses, it is important to see if findings generalize across time, stimuli, and settings. Follow-up studies to examine this can involve testing a subset of this sample again at a six month follow-up session to assess generalization across time.
Different stimuli representing the same construct could be used at a follow-up session with a subset of this sample. Finally, assessments could be done in a classroom setting, or other real-world setting to see if results generalize across setting with a subset of this sample.
Since ASD is solely defined by behavioral characteristics, it is useful to be able to break down this group into subtypes based on other characteristics, to better understand various phenotypes and to tailor interventions and prevention programs. A follow-up study will be conducted exploring different endophenotypes that may exist in ASD, using the four hypothesized responder types as a guide. These patterns may offer a better understanding of how stress operates in individuals with ASD, and may have direct clinical applications. Dynamic cluster analysis will be performed on the current data in a future study to see exactly what subtypes of responders exist. Cluster analysis categorizes inter-individual heterogeneity in intra-individual change as indices of different sub-populations that are characterized by different trajectories (Dumenci & Windle, 2001). This method allows researchers to identify patterns of change when group membership is not known a priori. Three reasons identified by Hoeppner, Goodwin, Velicer, Mooney, & Hatsukami (2008) as to why this method is so useful is that it "(1) parsimoniously represent(s) individual differences in intra-individual stability and change, (2) evaluate(s) taxonomic developmental theories of change, and (3) facilitate(s) the development of models for early intervention and prevention programs by determining predictors and outcomes specific to a certain growth pattern." (p. 625).
Typically, cluster analysis is used for data collected from many people at a single time point. This future study will use dynamic cluster analysis, since it is based on a single variable measured on multiple occasions over time (Norman, Velicer, Fava, & Prochaska, 1998;Norman, Velicer, Fava, & Prochaska, 2000;Prochaska, Velicer, Guadagnoli, Rossi & DiClemente, 1991      C h a n g e in S ta ff R e s t P h y s ic a l E x e rtio n T ra n s itio n Phase M e a n H e a rt R a te (b p m ) Figure 58. Normal Responsivity Exemplar Example 4