This is a replication, randomized control trial, that investigated the therapeutic effects of a 12-week equine-assisted (EA) intervention on the social and sensory functioning of children with autism. Reliability and stability of parent and teacher reports of children’s social and sensory functioning across three assessment times were assessed, in support of the validity of observed outcomes. Furthermore, it was hypothesized that children in the EA group (n = 25) would significantly improve, relative to a wait-list control group (n = 25), in both domains of functioning. Results indicated that reports were reliable, and children in the experimental group improved in overall social and sensory functioning, as well as within specific subdomains, with “unblinded” assessment methods. Relative to the pre-assessment scores, children improved in functioning in specific areas at post-assessment and 8-weeks post-intervention. Therefore, results of the study suggest EA activities may be a beneficial modality for delivering autism-specific treatment strategies.
For centuries, qualitative reports have suggested nonhuman animals can positively impact human well-being. From Hippocrates, who described horseback riding as a universal exercise, to more recent scientific literature, researchers continue to be interested in this topic (Fine, 2006; Frewin & Gardner, 2005; Klontz, Bivens, Leinart, & Klontz, 2007; Macaulay & Gutierrez, 2004; Nimer & Lundahl, 2007; Notgrass & Pettinelli, 2014; Riede, 1987; Souter & Miller, 2007; Trotter, Chandler, Goodwin-Bond, & Casey, 2008). Animals are purported to benefit humans through a variety of modalities (e.g., service animals). Specifically, animal-assisted therapies (AAT) incorporate an animal in a therapeutic treatment plan to promote interactions and achieve individualized goals (Chandler, 2005; Nimer & Lundahl, 2007).
AAT have shown promise across many disabilities (e.g., cerebral palsy, visual impairments, language deficits), psychological dysfunctions (e.g., depression, anxiety), and developmental challenges (e.g., motor difficulties) (Macauley & Gutierrez, 2004; Nimer & Lundahl, 2007; Tseng, Chen, & Tam, 2013). AAT improves cognitive and social functioning (Fine, 2006), decreases depressive symptoms (Souter & Miller, 2007; Klontz et al., 2007), decreases anxiety (Fine, 2006; Morrison, 2007), and increases social functioning (Fine, 2006). As a result, AAT, and more specifically, equine-assisted activities (EAA), are now integrated into the treatment of many populations, including those with autism spectrum disorders (ASD; Pavlides, 2008).
The Professional Association of Therapeutic Horsemanship (PATH) International defines EAA as interventions that incorporate equine activities/the equine environment to achieve specific rehabilitative or therapeutic goals (PATH, 2011; Pavlides, 2008). Using a horse as a medium for delivering treatment components purportedly stimulates multiple domains of functioning via the “multidimensional movement” of the animal and the “therapeutic bond” between patient and animal (Bliss, 1997; American Hippotherapy Association, 2011).
Individuals within the autism field have shared their personal experiences with animals and supported their beneficial effects for children with ASD (see Grandin & Johnson, 2005). However, there is currently a paucity of rigorous scientific work evaluating the efficacy of the intervention for this population (Zane, 2010). Though some studies have demonstrated favorable outcomes for both children in EAA and a comparative non-equine program, (Lanning, Baier, Ivey-Hatz, Kreneck, & Tubbs, 2014), they did not use a randomized control trial (RCT) approach. Thus, the primary goal of this study was to evaluate the effects of an equine-assisted intervention on the sensory and social functioning of children with ASD, via a more robust replication of our prior research and utilizing an RCT approach.
To our knowledge, one RCT has been conducted to evaluate the treatment efficacy of EAA for children with ASD. Bass, Duchowny, and Llabre (2009) examined the effects of a 12-week therapeutic horseback riding intervention on the sensory and social functioning of 34 children with ASD. Subjects were randomized to either an EAA condition (n = 19) or a wait-list control group (n = 15). Participants were assessed at pre- and post-assessment points with “unblinded” parent reports on the Social Responsiveness Scale (SRS; Constantino, 2002) and the Sensory Profile (SP; Dunn, 1999). Results suggested that compared to the wait-list control group, participants in the experimental condition improved in the domains of sensory integration, directed attention, social motivation, and attention/distractibility. It was concluded that EAA may be an efficacious and viable treatment option for these children because they are stimulating, provide a rewarding multisensory experience that promotes social engagement and motor skills, and may positively impact sensory functioning.
Although this previous study demonstrated positive effects, there were several methodological shortcomings with the initial experiment conducted by Bass et al. (2009), namely the following: the experimental group only consisted of 19 participants and the control consisted of 15; only two time points were utilized in the outcome assessment, thus, questions pertaining to maintenance of the effects were not addressed; and the previous work relied on “unblinded” parent reports as the only assessment of outcomes which may have been subjective, unreliable, and unstable across the two assessment time points (i.e., no measurement of inter-rater reliability). While all of these methodological shortcomings were could not be addressed with this current project (e.g., reports were not “blinded,” there were no treatment fidelity measurements, and our study utilized a wait-list control versus a placebo group or an “attention-activity control group”), we aimed to address several of these factors.
First, our project is a replication, and the replication of research designs is vital in the establishment of treatment effects and evaluations of overall treatment efficacy (Nathan & Gorman, 2002). We also recruited a larger number of participants that were randomized into groups; we used clear inclusion/exclusion criteria and sound assessment measures; collected data across three time points (i.e., pre, post, and follow-up) from two informants; and evaluated the inter-rater reliability and stability of the parent and teacher reports. Additionally, we evaluated the measurements within the study, as prior research has demonstrated poor reliability between different informants (Achenbach, McConaughy, & Howell, 1987; De Los Reyes & Kazdin, 2005). We hypothesized that parent and teacher reports of children’s social and sensory functioning would demonstrate reliability and stability across all three time points. Furthermore, we hypothesized children exposed to EAA would exhibit significantly more improvements, relative to the wait-list control group, in both domains of functioning as assessed by both reports, and that effects would be maintained across time.
Materials and Methods
Participants were recruited from organizations (e.g., Parent to Parent) serving the autism community in Southeastern Florida. Inclusion/exclusion criteria for participants were: (a) children needed a DSM-IV-TR ASD diagnosis (American Psychiatric Association, 2000); and (b) children could not have had prior experience with hippotherapy nor any EAA provided by a certified PATH instructor. All diagnoses were confirmed by the referring agency and/or a family physician. Parents first received letters describing the study, followed by a phone call to screen for eligibility. Out of 61 potential participants contacted, 11 were deemed ineligible due to age and/or a non-ASD diagnosis. A total of fifty children between the ages of 7 and 12 years (M = 8.70, SD = 1.60) were recruited. All were enrolled in a school setting. None of the children enrolled into the study had physical disabilities. The majority of participants were verbal (approximately 76%) and had been previously exposed to conventional treatments for ASD (see Table 1). All participating families were bilingual (English and Spanish) and approximately 20% requested to receive measures in Spanish.
Social Responsiveness Scale (SRS)
The SRS (Constantino, 2002) is a 65-item questionnaire assessing the severity of symptoms associated with ASD. Parents/teachers rate participants on a 4-point Likert scale ranging from 0 (never true) to 3 (almost always true). The measure includes five subscales—social awareness, social cognition, social communication, social motivation, and autistic mannerisms—and a total score. Higher scores are indicative of higher levels of impairment, with criteria for ASD being met at a total T-Score of 60. The SRS has overall internal consistency (α = .97) and a retest temporal stability in males and females (r = .85 and r = .77, respectively). In addition, the internal consistency for each subscale yields high Cronbach alphas (e.g., α = .92 for social communication subscale, see Constantino, 2002).
Sensory Profile (SP)
The SP (Dunn, 1999) is a 125-item questionnaire asking parents for an assessment of overall sensory functioning and the level of concern related to children’s difficulties in (a) sensory processing, (b) modulation, and (c) behavioral and emotional responses. Raters use a 5-point Likert scale ranging from 1 (always) to 5 (never). Lower scores are indicative of higher levels of impairment. The SP is scored on nine separate subscales including: sensation seeking, emotionally reactive, low endurance/tone, oral sensory sensitivity, inattention/distractibility, poor registration, sensory sensitivity, sedentary, and fine motor/perception. The internal consistencies for the subscales range from .47 to .91 (Dunn, 1999).
Sensory Profile School Companion (SPSC)
The SPSC (Dunn, 2006), a teacher questionnaire composed of 62 items, provides an assessment of students’ responses to common sensory experiences in school. Raters respond to items using a 5-point Likert scale ranging from 1 (always) to 5 (never). The SPSC has four quadrant scores—(1) registration, (2) seeking, (3) sensitivity, and (4) avoiding—corresponding to Dunn’s Model of Sensory Processing. Lower scores are also indicative of higher levels of impairment. Only the four quadrant scores were used as outcomes in this study. Internal consistency estimates for the four quadrants range from .89 to .93, and the test-retest reliability coefficients range from .84 to .94 (Dunn, 2006).
Setting and Materials
All sessions took place at the Good Hope Equestrian Training Center (GHETC) in Homestead, Florida. GHETC is the only Miami-Dade County Premier Accredited PATH facility with PATH-certified riding instructors, all of whom had a minimum of 3 years of experience working with children with ASD. Volunteers assisting as “side walkers” previously completed GHETC volunteer training and mastered the duties set forth by GHETC and PATH. All of the children attended every session (i.e., no drop-outs). A total of twelve horses were used and were assigned to a particular rider for each session.
After screening for eligibility, parents consented to the pre-testing (i.e., pre), 12 weeks of EAA at no cost, a post-testing session after the 12 weeks of intervention (i.e., post), and a follow-up testing session conducted eight weeks after the conclusion of the intervention (i.e., follow-up). Pre-assessment measures and instructions were mailed to and completed by parents and teachers prior to randomization, and before the intervention sessions were initiated. Parents delivered and returned the teacher measures. Parents were also told not to inform teachers of their child’s participation in the project; however, this was not formally monitored. Subsequently, participants were randomly assigned via a computer-generated randomization schedule to either the experimental equine-therapy group or a wait-list control group.
After the 12-week intervention, parents and teachers of participants in both the experimental and control conditions completed post-test measures. Next, parents and teachers completed the same battery of measures 8 weeks after the conclusion of the intervention (i.e., follow-up). Upon completion of the follow-up assessment, participants in the wait-list control condition received 12 consecutive weeks of 1-hour sessions of the equine-assisted intervention. No additional data were collected on the control participants once they received the intervention.
The current study replicated the EAA intervention as reported in Bass et al. (2009). (For specific details of the EAA intervention and its components, refer to this previous study.) Provided here is a general overview of the intervention. The 12 weeks of the intervention incorporated mounting/dismounting, warm-up exercises, riding skills, and mounted games. All processes underlying these techniques were verbalized and modeled by the trainers using step-by-step instructions, as well as a picture book (see Appendix A). Children performed approximately 10 minutes of warm-up exercises (e.g., trunk twists, toe touches), followed by 25 minutes of riding skills, and then approximately 20 minutes of individualized and group games while on the horse (i.e., “musical stalls,” which is reminiscent of musical chairs although on horseback and with the use of horse stalls).
The final segment of the sessions involved 15 minutes of grooming activities with their horse, where instructions were also provided through visual supports (see Appendix B). Participants were verbally and physically (e.g., hugs, high-fives, praise, etc.) praised for their effort and engagement. Furthermore, instructors and assistants made concerted efforts to facilitate eye contact and demonstrate or model adaptive socialization and communication skills. The director of GHETC supervised each of the sessions to insure there were no deviations from fidelity intervention protocol. The instructors/assistants also had weekly meetings to debrief and discuss progress of the participants. No known deviations from the intervention protocol were noted.
One aim of the current study was to evaluate the inter-rater reliability between parent and teacher reports on each of the SRS domains as well as their stability across three time points (i.e., pre [T1], post [T2], and follow-up [T3]). Analyses were completed using the control group only. Specifically, a parallel measurement model (refer to Figure 1) was specified for each SRS domain to determine the stability across the three time points and the reliability between the parent and teacher reports. This model yielded estimates of the true score variance and error variance used to estimate the intraclass correlation as a measure of stability. For each model run, error variances (i.e., ε1 = ε2 = ε3) were set equal across time, and factor loadings were set to 1.
All analyses were done in Mplus version 5.21 (Muthen & Muthen, 2004) within a structural equation modeling framework. As depicted in Figure 1, the latent variable entitled Parent Report was defined by three indicators which included the parents’ report on the SRS domain at each of the three times. The latent variable entitled Teacher Report was defined by three indicators which included the teachers’ reports on the SRS domain at each of the three times. Section A of Figure 1 illustrates the specific model component used to estimate the stability of the parent report on each domain across time, and section B illustrates the specific model component used to estimate the stability of the teacher report on each domain across time. In order to evaluate the inter-rater reliability of the SRS domains between the parent and teacher reports, the correlations between the latent variables in sections A and B were estimated with the aforementioned constraints on model parameters. For the SP and SPSC domains, only the stability across time could be estimated, since parent and teacher instruments do not result in comparable domains.
A mixed 2 (experimental; control) × 2 (pre; post) repeated measures analysis of variance (ANOVA) was employed to evaluate differences between the experimental and control groups in the change from pre- to post-assessment. Significant group by time interactions were followed up with paired sample t-tests. A similar design was employed to evaluate the maintenance of all significant treatment effects and differences in treatment effects from post to follow-up. We used η2 as a measure of effect size in the ANOVA, representing the proportion of variance in the dependent variable explained by the treatment.
Following the paired sample t-tests, we used d as a measure of effect size, indicating the standardized increases from pre to post, pre to follow-up, and post to follow-up. We applied a correction for the Type-I error rate at two levels. First, since there were three measures and three time comparisons, we corrected the analyses of the total scores by dividing the .05 level of significance by 9, resulting in α = .0056 for the initial analyses. Only when the total score results were significant did we follow up with analyses of the subscales. Subscale scores were tested at a level of significance that corrected for the number of subscales within a specific pair of time points. Thus for the SRS (either parent or teacher), we corrected for 5 subscales (α = .01); for the SP, we corrected for 9 subscales (α = .0056); and for the SPSC, we corrected for 4 subscales (α = .0125), when comparing pre to post. Only when there was an effect at post were the pre to follow-up comparisons or the post to follow-up performed and similarly corrected.
Refer to Table 2 for all descriptive data. The experimental group had a mean age of 8.84 (SD = 1.72), and the wait-list control group had a mean age of 8.56 (SD = 1.50). Results indicated no significant group differences on gender, ethnicity, or age. Additionally, groups did not differ at pre on parent SRS total score, SP total score, teacher SRS total score, or any subscales.
Reliability and Stability
The stability coefficients for all domains and reliability coefficients for the SRS domains for the control group are listed in Table 3. All domains including the SRS total score, social awareness, social cognition, social communication, social motivation, and autistic mannerisms demonstrated robust stability across time for both parent and teacher reports. Additionally, all SRS domains, with the exception of autistic mannerisms (r = .33, p = .09), demonstrated reliability between the parent and teacher reports. With respect to the SP instrument, all subscales including the SP total score, sensation seeking, emotionally reactive, low endurance/tone, oral sensory sensitivity, inattention/distractibility, poor registration, sensory sensitivity, sedentary, and fine motor perception demonstrated robust stability across time for the parent reports. In addition, all quadrants of the SPSC demonstrated robust stability across time points for the teacher reports.
Means and standard deviations of parent reports on all subscales administered at pre, post, and follow-up for both the experimental and the control group are reported in Table 4. With respect to the SRS total score, there was a significant group × time interaction from pre to post, F(1, 48) = 25.34, p < .001, η2 = 0.35. The experimental group mean significantly decreased from pre and post, t(24) = 6.58, p < .001, d = 1.31, while the control group mean remained unchanged, t(24) = -.639, p = .53. In support of the maintenance of these effects, a group × time interaction was also significant when comparing pre to follow-up, F(1, 48) = 8.89, p = .005, η2 = 0.16. The mean of the experimental group at follow-up remained significantly lower than the mean at pre, t(24) = 4.76, p < .001, d = 0.95. However, a group × time interaction comparing post to follow-up indicated a significant decrease in treatment effects from post to follow-up, F(1, 48) = 21.85, p < .001, η2 = 0.31. The mean of the experimental group significantly increased from post to follow-up, t(24) = -4.39, p < .001, d = -0.88.
To further analyze the significant treatment effects demonstrated within the SRS total score, group × time interactions were examined for each of the SRS subscales separately. Significant group × time interaction effects were revealed for pre and post for four out of the five subscales: social cognition, F(1, 48) = 13.88, p = .001, η2 = 0.22; social communication, F(1, 48) = 19.47, p < .001, η2 = 0.29; social motivation, F(1, 48) = 24.40, p < .001, η2 = 0.34; and autistic mannerisms, F(1, 48) = 20.69, p < .001, η2 = 0.30. There was no significant interaction for the social awareness subscale (p = .153). Significant pre- to post-treatment effects for the four subscales were observed in the experimental group: social cognition, t(24) = 4.40, p < .001, d = 0.88; social communication, t(24) = 6.49, p < .001, d = 1.29; social motivation, t(24) = 7.27, p < .001, d = 1.45; and autistic mannerisms, t(24) = 4.61, p < .001, d = .92. No significant difference between pre and post was observed within the control group for any subscale.
A group × time interaction comparing pre and follow-up was also significant for the following subscales: social motivation, F(1, 48) = 9.47, p = .003, η2 = 0.17; and autistic mannerisms, F(1, 48) = 7.67, p = .008, η2 = 0.14. No significant interactions were revealed for the social communication (p = .013) or social cognition (p = .09) subscales. The mean of the experimental group at follow-up remained significantly different from the mean at pre for social motivation, t(24) = 5.58, p < .001, d = 1.11; and autistic mannerisms, t(24) = 2.40, p = .025, d = 0.48. However, a group × time interaction comparing post to follow-up was also significant, indicating a decrease in the treatment effect from post to follow-up for the subscales social motivation, F(1, 48) = 6.42, p = .015, η2 = 0.12; and autistic mannerisms, F(1, 48) = 13.91, p < .001, η2 = 0.23. The mean of the experimental group significantly increased from post to follow-up for the social motivation, t(24) = -2.85, p = .009, d = -0.57; and autistic mannerisms, t(24) = -4.20, p < .001, d = -0.84, subscales.
Regarding the SP total score, there was a significant group × time interaction from pre to post, F(1, 48) = 21.09, p < .001, η2 = 0.31. The experimental group mean significantly increased from pre- to post-testing, t(24) = -5.19, p < .001, d = -1.04; while the mean of the control group remained unchanged, t(24) = .930, p = .36. However, a group × time interaction was not significant when comparing pre to follow-up, F(1, 48) = 8.24, p = .006 (at the determined α = .0056), suggesting no maintenance of these effects at follow-up.
Group × time interactions were also examined for each of the SP subscales separately. Significant group × time interaction effects were revealed for pre and post for four out of the nine subscales: emotionally reactive, F(1, 48) = 16.21, p < .001, η2 = 0.25; low endurance/tone, F(1, 48) = 21.85, p < .001, η2 = 0.31; inattention/distractibility, F(1, 48) = 31.84, p < .001, η2 = 0.40; and sedentary, F(1, 48) = 19.57, p < .001, η2 = 0.29. There were no significant interactions for the sensory sensitivity (p = .011), sensation seeking (p = .06), oral sensory sensitivity (p = .314), poor registration (p = .313), and the fine motor perception (p = .113) subscales.
Significant pre- to post-treatment effects for the four subscales were observed in the experimental group: emotionally reactive, t(24) = -4.80, p < .001, d = -0.96; low endurance/tone, t(24) = -4.18, p < .001, d = -.84; inattention distractibility, t(24) = -8.15, p < .001, d = -1.63; and sedentary, t(24) = -5.04, p < .001, d = -1.01. No significant difference between pre and post was observed within the control group for any subscale, with the exception of the low endurance/tone subscale where a significant decrease was observed, t(24) = 2.47, p = .021, d = 0.49, indicating an increase in symptomatology. However, a group × time interaction comparing pre and follow-up was not significant for the emotionally reactive (p = .044), low endurance/tone (p = .034), inattention distractibility (p = .006), and the sendentary (p = .051) subscales.
Means and standard deviations of teacher reports on all subscales administered at pre, post, and follow-up for both the experimental and the control groups are reported in Table 5. Regarding the SRS total score, there was a significant group × time interaction from pre to post, F(1, 48) = 29.52, p < .001, η2 = 0.38. The experimental group mean significantly decreased from pre- to post-testing, t(24) = 6.39, p < .001, d = 1.23; while the control group mean remained unchanged, t(24) = -1.58, p = .13. A group × time interaction was also significant when comparing pre to follow-up, F(1, 48) = 13.48, p = .001, η2 = 0.22. The experimental group mean at follow-up remained significantly lower than the mean at pre, t(24) = 5.72, p < .001, d = 1.14. However, a group × time interaction comparing post to follow-up indicated a significant decrease in treatment effects, F(1, 48) = 12.32, p = .001, η2 = 0.20. The mean of the experimental group significantly increased from post to follow-up, t(24) = -3.39, p = .002, d = -0.68.
Group × time interactions were also examined for the teacher reports on each of the SRS subscales. Significant group × time interaction effects were revealed for pre and post for all five subscales: social awareness, F(1, 48) = 10.54, p < .002, η2 = 0.18; social cognition, F(1, 48) = 17.50, p < .001, η2 = 0.27; social communication, F(1, 48) = 31.34, p < .001, η2 = 0.40; social motivation, F(1, 48) = 15.56, p < .001, η2 = 0.25; and autistic mannerisms, F(1, 48) = 21.98, p < .001, η2 = 0.31. Significant pre- to post- treatment effects were observed in the experimental group for the five subscales: social awareness, t(24) = 4.83, p < .001, d = 0.97; social cognition, t(24) = 5.34, p < .001, d = 0.82; social communication, t(24) = 6.31, p < .001, d = 1.26; social motivation, t(24) = 4.87, p < .001, d = 0.97; and autistic mannerisms, t(24) = 5.12, p < .001, d = 1.02. No significant difference between pre and post was observed within the control group for any subscale.
A group × time interaction comparing pre and follow-up was also significant for the following subscales: social cognition, F(1, 48) = 7.37, p < .009, η2 = 0.13; social communication, F(1, 48) = 15.76, p < .001, η2 = 0.25; and autistic mannerisms, F(1, 48) = 11.09, p = .002, η2 = 0.19. A significant interaction was not observed with the social motivation (p = .013) or social awareness (p = .06) subscales. The mean of the experimental group at follow-up remained significantly different from the mean at pre for social cognition, t(24) = 4.21, p < .001, d = 0.81; social communication, t(24) = 5.69, p < .001, d = 1.14; and autistic mannerisms, t(24) = 4.38, p < .001, d = 0.88. However, a group × time interaction comparing post to follow-up was also significant, suggesting a significant decrease in the treatment effect from post to follow-up for the following subscales: social communication, F(1, 48) = 12.10, p = .001, η2 = 0.20; and autistic mannerisms, F(1, 48) = 5.03, p = .03, η2 = 0.10. The mean for the social communication subscale of the experimental group significantly increased from post to follow-up, t(24) = -3.22, p = .004, d = -0.64, but this was not observed on the autistic mannerisms subscale, t(24) = -1.85, p = .08.
Group × time interactions were examined for each of the four SPSC quadrants as well. Significant group × time interaction effects were revealed for pre and post for all four quadrants: registration, F(1, 48) = 19.08, p < .001, η2 = 0.28; seeking, F(1, 48) = 6.84, p = .012, η2 = 0.13; sensitivity, F(1, 48) = 16.97, p < .001, η2 = 0.26; and avoiding, F(1, 48) = 21.97, p < .001, η2 = 0.31. In the experimental group, there were significant pre- to post-treatment effects for the four quadrants: registration, t(24) = -6.05, p < .001, d = -1.21; seeking, t(24) = -3.71, p = .001, d = -0.74; sensitivity, t(24) = -5.64, p < .001, d = -1.13; and avoiding, t(24) = -6.15, p < .001, d = -1.23. No significant difference between pre and post was observed within the control group for any subscale.
A group × time interaction comparing pre and follow-up was significant for only one of the four quadrants: registration, F(1, 48) = 8.39, p < .006, η2 = 0.15. There were no significant interactions for the sensitivity (p = .04), avoiding (p = .02), or seeking (p = .15) quadrants. The mean of the experimental group at follow-up remained significantly different from the mean at pre for registration, t(24) = -4.52, p < .001, d = -.90. A group × time interaction comparing post to follow-up was also significant, indicating a significant decrease in the treatment effect from post to follow-up for the registration quadrant, F(1, 48) = 9.54, p = .003, η2 = 0.17. The mean of the experimental group significantly decreased from post to follow-up for registration, t(24) = 3.02, p = .006, d = 0.60.
The results of this replication study corroborate the findings of Bass et al. (2009), and provide further support that EAA may be utilized in improving the social and sensory functioning of children with ASD. Specifically, each measure demonstrated robust stability across time, and the reliability between both parent and teacher reports on the SRS was robust, with the exception of one subscale (autistic mannerisms, r = .33). Furthermore, parents and teachers reported that children in the experimental group improved in specific areas of social and sensory functioning relative to the wait-list control group. Moreover, relative to the pre-assessments scores, children maintained a proportion of these improvements in functioning 8-weeks post-intervention. Lastly, improvements in social cognition and autistic mannerisms as reported by teachers were maintained 8-weeks post-intervention.
There are two common arguments against the utilization of EAA for individuals with ASD: (1) few well-designed studies (i.e., RCTs); and (2) measurement of outcomes with survey instruments that may be unreliable and/or subjective (Fraenkel & Wallen, 2009; Zane, 2010). The first argument is apparent following a review of the extant literature on EAA. The second argument stems from prior research that demonstrated poor reliability between different informants of treatment outcomes (see Achenbach, McConaughy, & Howell, 1987; De Los Reyes & Kazdin, 2005). Nevertheless, a review of the literature on informant discrepancies provides a clear indication that reliability, as well as stability of outcome measurements, should be considered when using survey instruments within treatment research.
The current study addresses both of these arguments. First, this study employed an RCT design and evaluated the efficacy of EAA versus a wait-list control group. No significant differences between groups on demographics or pre-assessment scores emerged. Secondly, the measurement model was evaluated within the control group, showing both informants’ reports on each measure were stable across time. In addition, inter-rater reliability between informants on the SRS was robust for the total score and all but one of five subscales (autistic mannerisms). Our estimate of parent and teacher agreement was based on a latent variable that controlled for stability over time.
It is possible that the autistic mannerisms subscale of the SRS was not reliable across reports because of the differences between home and school environments. Low inter-rater agreement is often attributed to different situational factors (e.g., the structure of the environment) that produce different behaviors (De Los Reyes & Kazdin, 2005; Winsler & Wallace, 2002). Nonetheless, the results of the measurement model provide strong evidence for both the reliability and stability of these measurements, and the validity of observed treatment effects.
According to both informants, significant improvements emerged in overall social functioning within the experimental group, including social cognition, social communication, social motivation, and autistic mannerisms. Parents did not report significant improvements in social awareness; however, teachers reported such improvements. Additionally, parents indicated that relative to the pre-assessments scores, children exhibited improvements in functioning 8-weeks post-intervention, specifically in overall social functioning, including social motivation. In contrast to parents, teachers reported this same pattern within overall social functioning, including social cognition and communication. Moreover, all of the treatment effects from the pre- to post- and from pre- to follow-up assessments demonstrated moderate to large effect sizes for the intervention (d = 0.48 to d = 1.45; Cohen, 1992; 1998). In addition, the observed intervention effects on autistic mannerisms and social cognition as reported by teachers were maintained 8 weeks after intervention. Overall, results of this study demonstrate more areas of improvement relative to the previous study conducted by Bass et al. (2009).
Parents reported significant improvements in overall sensory functioning (emotionally reactive, low endurance/tone, inattention/distractibility, and sedentary). However, parents reported that relative to the pre-assessment scores, children’s functioning did not improve in these areas assessed at follow-up. Similarly, teachers reported significant improvements in registration, seeking, sensitivity, and avoiding. Relative to their pre-assessment scores, all of the children showed improvements in the registration domain at follow-up. All significant treatment effects demonstrated moderate to large effect sizes for the intervention (d = -.84 to d = -1.63; Cohen, 1992; 1998). Once again, the results of this study demonstrate improvement in more areas relative to the previously conducted study, leading us to the question(s) of why and/or how EAA may be beneficial for children with ASD.
Bass et al. (2009) attributed increases in social and sensory functioning to a variety of factors. First, exposure to the horse provided a multisensory experience, plausibly eliciting high levels of social motivation and engagement. Therapeutic horseback riding is an activity that demands motor-learning skills, motor control, and social engagement; therefore, Bass et al. purported that some, if not most, observed increases within the areas of social and sensory functioning could be attributed to cerebellar stimulation, a structure of the brain that has been implicated in children with ASD (Pierce & Courchesne, 2001).
Our current conceptualization of why we observed the treatment effects mirrors these attributions. However, we also strongly believe that EAA may simply be a beneficial medium for providing empirically supported treatments. White, Koenig, and Scahill (2007) provide strategies for social skills development for children with ASD. For example, to increase social motivation for this population, the authors recommended fostering self-awareness; self-esteem; developing a nurturing and fun environment; interspersing new skills with mastered skills; and starting with simple, easily learned skills. Moreover, Schaaf (2011) described the principles that guide sensory-integrative approaches to treating sensory dysfunction which include helping the child attain appropriate levels of alertness; presenting the child with different types of sensory opportunities including, tactile, vestibular, and proprioceptive, to develop self-regulation and sensory awareness; and challenging postural, ocular, oral, and bilateral motor control.
We believe the equine-assisted intervention studied here incorporated several of these components within its primary treatment goals. For example, the goal of mounting/dismounting was to model and facilitate receptive and expressive language abilities as well as vestibular processing. Additionally, the goal of riding skills was to stimulate sensory seeking, promote gross and fine motor development, and to facilitate verbal and/or nonverbal social and communication skills. Therefore, it is plausible that using animals may serve as a beneficial modality for providing other empirically supported, widely used, and accepted strategies for improving the functioning of children with ASD.
In conclusion, we believe EAA may be a beneficial vehicle by which effective treatment strategies can be provided. It is likely that the components embedded within this particular equine-assisted treatment elicited improvements, and there may be an interactive effect due to the stimulation and therapeutic bond with an animal that facilitates engagement with these active ingredients. In other words, EAA may serve as a conduit for other empirically supported treatment strategies and may help engage clients more successfully during treatment. Many of us can relate to the special companionship that animals provide and their contributions to our overall well-being—this phenomenon in combination with evidenced-based treatment components may prove to be an important delivery method for autism-specific interventions.
There are several important limitations to this study, thus, further research is strongly warranted. First, and perhaps most importantly, parents and teachers were not “blind” raters to the intervention group. Secondly, the EAA intervention was not evaluated against a non-active treatment (e.g., a leisure horseback-riding group) to assess a placebo effect nor an “active” control condition, as was employed in the study conducted by Lanning et al. (2014). This limitation certainly brings into question whether the same or better results could be achieved with an equal amount of attention and novel exercise, sans a horse, or perhaps with other animal-assisted activities (e.g., canine-assisted therapy).
The effects of human interactions (i.e., positive social interactions with the trainers) from companion animal interactions are overall difficult to isolate. However, due to the limited resources, the limits of recruitment, and staff availability, we were unable to include this third comparison group. Therefore, given the importance of such a group when conducting efficacy studies, a third placebo or an “active” control condition should be employed in future research. In addition, the 8-week follow-up period is relatively short, which does not permit conclusions about the stability of treatment over longer time periods and there was, in fact, a decline in effects of treatment for several domains even during this relatively short follow-up period. There were also no measures of fidelity collected on the intervention. Furthermore, this study did not independently evaluate autism severity or cognitive functioning. Therefore, moderators affecting treatment response could not be investigated and the generalizability of this treatment is unknown. Lastly, accessibility to EAA programs is likely restricted, as horseback riding is an expensive and geographically limited activity.
Future research should identify mechanisms of change and active treatment components by assessing mediators of outcomes. Moderators (e.g., severity and IQ) should also be assessed to further elucidate for whom this treatment modality may be beneficial. Overall, more replication studies with inclusion of a placebo group and unblinded objective assessments (e.g., clinician ratings or standardized assessments) are strongly indicated here, as well as a line of research completed by an entirely independent camp of researchers.
The development of this project and paper was supported with funding from the Horses and Humans Research Foundation (HHRF). The opinions expressed by the authors are not necessarily reflective of the position of, or endorsed by, the HHRF.
Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications of cross-informant correlations for situational specificity. Psychological Bulletin, 101, 213-232.
American Hippotherapy Association. (2011). About AHA. Retrieved from the AHA Website: http://www.americanhippotherapyassociation.org/.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: American Psychiatric Association.
Bass, M., Duchowny, C., & Liabre, M. (2009). The effect of therapeutic horseback riding on social functioning in children with autism. Journal of Autism and Developmental Disorders, 39(9), 1261-1267.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (Second Edition). Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers.
Courchesne, E., Yeung-Courchesne, R., Press, G. A., Hesselink, J. R., & Jernigan, T. L. (1988). Hypoplasia of cerebellar vermal lobules VI and VII in autism. The New England Journal of Medicine, 318, 1349-1354.
De Los Reyes, A., & Kazdin, A. E. (2005). Informant discrepancies in the assessment of childhood psychopathology: A critical review, theoretical framework, and recommendations for further study. Psychological Bulletin, 131, 483-509.
Fine, A. H. (2006). Handbook on animal-assisted therapy: Theoretical foundations and guidelines for practice (2nd ed.). New York: Academic Press.
Fraenkel, J. R., & Wallen, N. E. (2009). How to design and evaluate research in education (Seventh Edition). New York: McGraw-Hill.
Frewin, K., & Gardiner, B. (2005). New age or old sage? A review of equine assisted psychotherapy. The Australian Journal of Counseling Psychology, 6, 13-17.
Klontz, B., Bivens, A., Leinart, D., & Klontz, T., (2007). The effectiveness of equine-assisted experiential therapy: Results of an open clinical trial. Society & Animals, 15, 257-267.
Lanning, B. A., Baier, M. E. M., Ivey-Hatz, J., Krenek, N., & Tubbs, J. D. (2014). Effects of equine assisted activities on Autism Spectrum Disorder. Journal of Autism and Developmental Disorders, 44, 1897-1907.
Macaulay, B., & Gutierrez, K. (2004). The effectiveness of hippotherapy for children with language-learning disabilities. Communication Disorders Quarterly, 25, 205-217.
Mesibov, G. B., & Shea, V. (2010). The TEACCH Program in the era of evidence-based practice. Journal of Autism and Developmental Disorders, 40(5), 570-579.
Muthen, L., & Muthen, B. (1998-2004). Mplus (Version 5) [Computer Software]. Los Angeles: Muthen & Muthen.
Nimer, J., & Lundahl, B. (2007). Animal-assisted therapy: A meta-analytic review. Anthrozoos: A Multidisciplinary Journal of the Interactions of People & Animals, 20(3), 225-238.
Notgrass, C. G., & Pettinelli, J. D. (2014). Equine assisted psychotherapy: The Equine Assisted Growth and Learning Association’s model overview of equine-based modalities. Journal of Experiential Education, 1053825914528472.
Pierce, K., & Courchesne, E. (2001). Evidence for a cerebellar role in reduced exploration and stereotyped behavior in autism. Biological Psychiatry, 49(8), 655-664.
Professional Association of Therapeutic Horsemanship International (PATH). (2011). The PATH premiere accredited center program. Retrieved from the PATH Website: http://www.pathintl.org/.
Riede, D. (1987). The relationship between man and horse with reference to medicine throughout the ages. People-Animals-Environment, 5(2), 26-28.
Rogers, S., & Vismara, L. (2008). Evidence-based comprehensive treatments for early autism Journal of Clinical Child & Adolescent Psychology, 37 (1), 8-38.
Schaaf, R. C. (2011). Interventions that address sensory dysfunction for individuals with Autism Spectrum Disorders: Preliminary evidence for the superiority of sensory Integration compared to other sensory approaches. In B. Reichow, P. Doehring, D. V. Cicchetti, & F. R. Volkmar (Eds.), Evidence-based practices and treatments for children with autism. New York: Springer.
- Search Google Scholar
- Export Citation
( Schaaf, R. C. ). 2011 Interventions that address sensory dysfunction for individuals with Autism Spectrum Disorders: Preliminary evidence for the superiority of sensory Integration compared to other sensory approaches. In (Eds.), Evidence-based practices and treatments for children with autism. , B. Reichow , P. Doehring , & D. V. Cicchetti F. R. Volkmar New York: Springer.
Souter, M. A., & Miller, M. D. (2007). Do animal-assisted activities effectively treat depression? A meta-analysis. Anthrozoos, 20, 167-180.
Trotter, K. S., Chandler, C. K., Goodwin-Bond, D., & Casey, J. (2008). Comparative study of the efficacy of group equine assisted counseling with at-risk children and adolescents. Journal of Creativity in Mental Health, 3(3), 254-284.
Tseng, S. H., Chen, H. C., & Tam, K. W. (2013). Systematic review and meta-analysis of the effect of equine assisted activities and therapies on gross motor outcome in children with cerebral palsy. Disability and Rehabilitation, 35(2), 89-99.
Vitaro, F. Gagnon, C., & Tremblay, R. E. (1991). Teachers’ and mothers’ assessment of children’s behaviours from kindergarten to grade two: Stability and change within and across informants. Journal of Psychology and Behavioural Assessment, 13, 325-343.
White, S.W, Koenig, K., & Scahill, L. (2007). Social skills development in children with Autism Spectrum Disorders: A review of the intervention research. Journal of Autism and Developmental Disorders, 37, 1858-1868.
Winsler, A., & Wallace, G. L. (2002). Behavior problems and social skills in preschool children: Parent-teacher agreement and relations with classroom observations. Early Education & Development, 13, 41-58.
Zane, T. (2010). A review of the effectiveness of therapeutic horseback riding. Retrieved from the Operation Autism: A Resource Guide for Military Families Website: http://www.operationautismonline.org/blog/a-review-of-the-effectiveness-of-therapeutic-horseback-riding/.
- Search Google Scholar
- Export Citation
( Zane, T. ). 2010 . Retrieved from the Operation Autism: A Resource Guide for Military Families Website: A review of the effectiveness of therapeutic horseback riding http://www.operationautismonline.org/blog/a-review-of-the-effectiveness-of-therapeutic-horseback-riding/.