!'--'+/)/4'2/#4+0/#-!'--'+/)/4'2/#4+0/#-
!45&+'3'103+4028!45&+'3'103+4028
0-1*+/33+34'&*'2#1802'-#7'##/&02'-#7'&0-1*+/33+34'&*'2#1802'-#7'##/&02'-#7'&
0/%-53+0/30/%-53+0/3
02+#2+/0
.028 /+6'23+48
%044+-+'/('-&
.028 /+6'23+48
0--074*+3#/&#&&+4+0/#-702,3#4*44137777'--$'+/)+/4-345&+'32'103+402802)#%71"##4
#240(4*'/+.#-33+34'&*'2#180..0/3/+.#-45&+'30..0/3#/&4*'4*'2
/4*2010-0)80..0/3
'%0..'/&'&+4#4+0/'%0..'/&'&+4#4+0/
#2+/0+-+'/('-&0-1*+/#33+34'&4*'2#1802':#7'&##/&.02':#7'&
%0/%-53+0/3/4*2090035-4+&+3%+1-+/#28052/#-0(*'/4'2#%4+0/30('01-'/+.#-3
*+3.#4'2+#-+3$205)*440805(02(2''#/&01'/#%%'33
$8!'--'+/)/4'2/#4+0/#-4*#3$''/#%%'14'&(02
+/%-53+0/$8#/#54*02+9'&#&.+/+342#4020(4*'!
45&+'3'103+402802.02'+/(02.#4+0/1-'#3'%0/4#%4
7$+32+/(07'--$'+/)+/4-02)
Dolphin-Assisted Therapy: More Flawed Data and More Flawed Conclusions
Lori Marino and Scott O. Lilienfeld
Emory University
KEYWORDS
DAT, dolphin-assisted therapy, dolphins, therapy, validity
ABSTRACT
Dolphin-Assisted Therapy (DAT) is an increasingly popular choice of treatment for illness and
developmental disabilities by providing participants with the opportunity to swim or interact with live
captive dolphins. Two reviews of DAT (Marino and Lilienfeld [1998] and Humphries [2003]) concluded
that there is no credible scientific evidence for the effectiveness of this intervention. In this paper, we offer
an update of the methodological status of DAT by reviewing five peer-reviewed DAT studies published in
the last eight years. We found that all five studies were methodologically flawed and plagued by several
threats to both internal and construct validity. We conclude that nearly a decade following our initial
review, there remains no compelling evidence that DAT is a legitimate therapy or that it affords any more
Dolphin-Assisted Therapy (DAT) is an increasingly popular choice of treatment for illness, disability, and
psychopathology in children and adults (Whale and Dolphin Conservation Society 2006). It involves
swimming and interaction with dolphins, typically in captivity. DAT formally began in the 1970s and over
the years has grown into a highly lucrative business with facilities all over the world, including the United
States, Mexico, Israel, Russia, Japan, China, and the Bahamas, to name only a few countries. The claims
made by these facilities have been subject to little or no scientific scrutiny. Moreover, there has been no
significant increase in the rate of peer-reviewed papers on DAT from the 1970s to the present. Yet DAT
programs continue to proliferate. As a consequence, DAT’s popularity greatly outstrips its meager
research base.
Eight years ago we published a review (Marino and Lilienfeld 1998) of the available peer-reviewed DAT
literature at the time, focusing on two papers by David Nathanson and his colleagues (Nathanson et al.
1997; Nathanson 1998). These authors advanced several extremely strong claims concerning the
efficacy of DAT for treating severely disabled children: (1) DAT significantly increases attention span,
motivation, and language skills; (2) DAT achieves these results more rapidly and more cost-effectively
than conventional therapies; and (3) DAT produces positive treatment effects that are maintained over a
long-term period (i.e., at least one year).
In our original paper (1998) we presented a methodological analysis of these two studies and the claims
derived from them by applying standard criteria for scientific validity from four established sources (Kazdin
and Wilson 1978; Cook and Campbell 1979; Kendall and Norton-Ford 1982; Shaughnessy and
Zechmeister 1994). We found no fewer than eleven independent methodological weaknesses in both
studies that seriously undermined their scientific validity. These shortcomings included the absence of
adequate comparison or control groups, unreliable, subjective and potentially biased raters, and analytic
methods that did not allow the reader to ascertain whether any children were harmed by DAT. We
concluded:
“...serious threats to validity and flawed data analytic procedures render the findings of
Nathanson and colleagues uninterpretable and their conclusions unwarranted and
premature... the current evidence for the efficacy of DAT can at best be described as
thoroughly unconvincing” (Marino and Lilienfeld 1998, p. 199).
At the time of our review, these two studies were the only peer-reviewed investigations of the therapeutic
effects of DAT. Therefore, as of 1998 there was no credible scientific evidence for the effectiveness of
this intervention.
A more recent and similar critical review of DAT was published five years later by Humphries (2003). In
her excellent analysis of the peer-reviewed DAT literature, Humphries evaluated six DAT studies and
found that all six lacked experimental controls and neglected to control adequately for major threats to
validity and alternative explanations for the results. She concluded that:
“...the available research evidence, as examined in this synthesis, does not conclusively
support the claims that DAT is effective for improving the behaviors of young children
with disabilities. More specifically, the results of the synthesis do not support the notion
that using interactions with dolphins is any more effective than other reinforcers for
improving child learning or social-emotional development.” (Humphries 2003, p. 6)
Finally, Brensing, Linke and Todt (2003) evaluated the claim that DAT works through healing by
ultrasound. The authors conducted an observational study of the behavior of dolphins in two DAT
programs to determine if their behavior was consistent with the hypothesis that they were echolocating on
human participants (and thus ostensibly providing “healing energy” through ultrasound). They concluded
that the dolphins’ behavior was not consistent with this hypothesis and therefore did not meet the minimal
requirements for common ultrasound therapies. Therefore, there is no scientific evidence that
echolocation can heal, nor that dolphins echolocate on humans in a manner even minimally consistent
with that claim.
In this paper, we update the DAT literature by reviewing the five peer-reviewed DAT studies published
since 1998. We focus on peer-reviewed papers because these articles presumably represent the “best”
evidence for the efficacy of DAT. We argue that an updated analysis of DAT is warranted given the
increasing popularity of this activity. Humphries (2003) reviewed peer-reviewed studies that ranged from
1989 to 1999, two of which had already been evaluated by Marino and Lilienfeld (1998). Brensing, Linke
and Todt (2003) based their conclusions on data taken from their studies of DAT at two facilities and did
not provide a methodological evaluation of the peer-reviewed literature. In the present article, we pick up
where our 1998 paper left off by evaluating the methodological validity of the five peer-reviewed DAT
studies that have been published since then.
Methods
We examined a total of five papers describing studies using DAT. These are Antonioli and Reveley
(2005), Iikura et al. (2001), Lukina (1999), Servais (1999), and Webb and Drummond (2001). The authors
of four of the studies reported improved behavioral outcomes for participants receiving DAT The lone
exception was Servais (1999), who reported positive outcomes in only one of the two experimental
groups in her study. Beyond the ambiguous results reported by Servais (1999), we found no other
published studies in which negative findings were noted.
We searched for peer-reviewed studies on DAT in several ways. First, we submitted the search terms
“dolphin assisted therapy” and “dolphin therapy,” to the Google Scholar search engine. Second, we used
these same terms to search Google for the websites of DAT facilities and searched any bibliographies
listed on these websites. Third, we used these terms in a search of the Current Contents database.
Fourth, we conducted a comprehensive search of all papers on DAT from 1999 to the present in the
following peer-review journals: Anthrozoös, Society & Animals, Applied Animal Behaviour Science, and
Zoo Biology. Fifth, we reviewed the reference sections of articles identified through the above means for
further relevant studies.
As in our 1998 paper, we assessed the validity of each study according to the standard criteria put forth
by four sources: Cook and Campbell (1979), Shadish, Cook and Campbell (2002), Kendall and Norton-
Ford (1982), and Shaughnessy and Zechmeister (1994). These sources describe a set of threats to
experimental validity that should be avoided in experimental research. The presence of even one major
threat to validity can render a study’s findings difficult, or in some cases even impossible, to interpret.
Table 1 displays several important threats to validity, their definition, and whether they are present in
each of the five studies. Most of these threats relate to either internal validity, that is, the methodological
soundness of the study, or to construct validity, that is, the soundness of the measures as indicators of
the constructs examined by the investigators. In the interest of space, we limit ourselves here to the most
serious threats to validity.
Results
Table 1 shows that each of the five studies we examined violated several important criteria for validity.
Moreover, because of inadequate experimental control, two threats to construct validity, that is,
nonspecific effects and construct confounding, plagued all of these studies. Because these threats to
validity are so ubiquitous in the DAT literature, we discuss them in depth in separate sections and follow
with more specific points concerning each of the individual studies.
Nonspecific Effects
Nonspecific effects are improvements from influences that are not specific to the intended treatment, and
that are shared by a wide variety of other treatments. They are generic effects of the treatment rather
than the result of the intended therapeutic ingredient(s). Two relevant subcategories of nonspecific effects
are placebo effects and novelty effects. The placebo effect is the well-documented but little understood
improvement that derives from participants’ expectation of improvement. Novelty effects are the general
energizing and uplifting effects of a new, exciting experience. Because of its nature, DAT is particularly
prone to these two nonspecific effects. Furthermore, many nonspecific effects are notoriously transient.
Therefore, any study of DAT must incorporate rigorous procedures for minimizing these nonspecific
effects and, in particular, must include ways to assess longer-term effects of the treatment.
DAT is vulnerable to placebo effects in part because DAT is typically marketed to participants and their
family members as highly efficacious and in part because the nature of the treatment is evident to
participants. Moreover, none of the studies we reviewed used a control that eliminates or substantially
minimizes this effect by eliminating cues to treatment condition. DAT is vulnerable to novelty effects
because of the obviously new and exciting experience of swimming with and interacting with a large,
intelligent, charismatic animal. The proper control for novelty would be exposure of the control group to
another novel, attractive animal, while keeping all else equal. In this way, both groups would have similar
reason to believe that they had received the relevant treatment, and both would be subject to the
excitement of interacting with an exotic animal. At the very least, DAT should be compared with other
animal-assisted therapies in addition to a no-treatment control group. If one found differential effects of
dolphins versus other large charismatic animals, it would be important to seek possible specific
therapeutic ingredients inherent in DAT. In contrast, if both dolphin and other animal groups improved
equally, this would suggest a generic characteristic of DAT as an example of a temporary “feel-good
effect” (activating effect) received from any animal therapies—and indeed any therapies involving exciting
and novel features.
Table 1. Major threats to validity in each of the five studies.
Validity Threat Definition
Antonioli
and
Reveley
(2005)
IIkura
et al.
(2001)
Lukina
(1999)
Servais
(1999)
Webb
and
Drummond
(2001)
Construct Validity
Nonspecific Effects (Improvement from effects not specific to the intended treatment)
Placebo
Improvement from expectation of improvement
X X X X X
Novelty
Effects of energy, excitement, and enthusiasm
not specific to the intended treatment
X X X X X
Construct
Confounding
Failure to take into account the fact that the
procedure may include more than one active
ingredient
X X X X X
Resentful
Demoralization
Participants aware of not receiving the active
treatment may be resentful and respond more
negatively than the treatment group
X
Demand
Characteristics
Tendency of participants to alter their responses in
accord with their suspicions about the research
hypothesis
X X X
Experimenter
Expectancy
Effects
Tendency for experimenter to unintentionally bias
the results in accordance with the hypothesis
X X X
Internal Validity
History
Occurrence of potentially therapeutic events other
than the intended treatment during the course of
the study
X
Testing
Improvements due to testing itself (e.g., practice
effects)
X X
Regression
Tendency of extreme scores to become less
extreme on re-testing
X X X
Instrumentation
Changes in the dependent measure at different
times in the study
X X
Multiple
Intervention
Interference
Administration of treatments other than the
intended treatment during the course of the study
X X
Maturation
Changes over time due to natural developmental
effects
X X
Informant Bias
Tendency of informants to selectively recall
improvement in accord with their hopes and
expectations (retrospective bias) or unintentional
distortion of improvement due to effort justification
X X
Construct Confounding
Construct confounding occurs when there is a failure to take into account the fact that the experimental
procedure may involve more than one active (effective) ingredient. In DAT, the experimental treatment
typically consists of a complex assortment of ingredients in addition to interacting with a dolphin per se,
such as swimming in water, being near water, being outside, and receiving attention from human
professionals. Moreover, the dolphin itself is a complex stimulus that can be deconstructed into various
potentially therapeutic components, such as the size and touch of the animal and the opportunity for
interaction with the animal. Because none of the DAT studies we examined adequately controlled for
these possibilities, they are all subject to construct confounding. In the psychotherapy literature, construct
confounding is typically decomposed by means of dismantling studies (Kazdin 1994), which separate the
potential effects of different treatment ingredients by creating different experimental conditions containing
these effects. And although we are not suggesting there is one single, ideal control for DAT, no DAT
study has included an adequate subset of the many control groups that would be required for even a
minimally effective dismantling strategy.
Antonioli and Reveley (2005)
Antonioli and Reveley (2005) conducted a single, blind, randomized controlled experiment to determine if
swimming and interacting with dolphins at a captive facility in Honduras lowered scores on depression
and anxiety scales. The total sample consisted of 30 men and women (prior to participant drop outs) with
mild to moderate depression scores on the ICD-10 (i.e., the International Classification of Diseases, 10th
edition) who also scored at least 11 on the modified Hamilton rating scale for depression at baseline after
four weeks without medications. The experimental group (the animal care program) consisted of 13
participants who played with, swam with, and “took care of” dolphins while in the water with them. The
control group (the outdoor nature program) consisted of 12 participants who swam and snorkeled in the
barrier reef and experienced a similar degree of individualized human contact as those in the
experimental group, but in the absence of dolphins. Both programs ran simultaneously from Monday to
Friday for one hour a day for two weeks. The authors conducted pre-treatment and post-treatment ratings
of depression with the Hamilton Rating Scale and the Beck Depression Inventory, and measured anxiety
with the Zung Self Rating Anxiety Scale.
The authors found a significantly greater improvement in depression scores in the experimental group
than in the control group. They also found no differences in the anxiety scores (although the authors
argued the lack of statistical significance could be due to the fact that only a subset of the sample was
clinically anxious prior to treatment). It was concluded that DAT is more effective than “water” therapy for
treating mild to moderate depression. However, this conclusion is compromised by several shortcomings.
First, the authors acknowledged that a limitation of their study is that the participants were not blind to
treatment. Therefore, demand characteristics might have influenced participants’ responses. The authors,
did, to some extent, guard against potential demand characteristics by emphasizing to participants that
they were taking part in a research study rather than a clinical intervention and should therefore not
expect any improvement. Still, it is not known whether participants may have discerned the true nature of
the study despite this information.
The second major shortcoming is that the control condition did not account for the potential effects of
interacting with any charismatic animal, in this case, a dolphin. The experimental group swam with
dolphins, whereas the control group did not interact with any other analogous animal. Participants in the
control group swam along a coral reef and interacted with the experimental personnel. However, there
was apparently no introduction of other large mobile animals, such as fish, that could have served as a
salient stimulus control for the dolphin. Therefore, because of the lack of an appropriate control group,
participant outcomes could have been due to nonspecific effects, including placebo and novelty effects.
Therefore, the authors’ inference that the dolphin was the effective therapeutic ingredient is premature.
The third potential threat to validity in this study is informant bias. Because the authors’ relied on self-
report measures of depression, the participants, who were not blind to treatment, could have selectively
recalled the amount of their improvement as a consequence of expectations, hopes, and even effort
justification (i.e., the psychological need to justify to oneself the time, expense, and energy invested in the
treatment).
Fourth, the authors acknowledged that they did not perform a follow-up study of the post-treatment
ratings. Therefore, the only justifiable conclusion they can draw is that there was a difference between the
experimental and control groups immediately following treatment (and even that conclusion is suspect
given the flaws discussed above). It is not known whether this difference existed even one day after the
post-treatment assessment and, moreover, whether any participants deteriorated. This lack of follow-up
assessment is a major shortcoming, as it renders the results susceptible to novelty or short-term
activating effects.
Related to this point, the authors touted the benefits of DAT over conventional drug therapy or
psychotherapy by claiming that depression is relieved after only two weeks of dolphin therapy but
requires at least four weeks to improve with mainstream therapies. The authors neglected to point out,
however, that the conventional treatments often exert long-term effects. In contrast, the authors provided
no evidence that dolphin treatment produces any effects beyond the immediate aftermath of the
treatment.
Finally, Shadish, Cook, and Campbell (2002) refer to a threat to construct validity known as “resentful
demoralization,” which may occur when members of a control group become aware that they are
receiving a less desirable treatment or no treatment, thereby becoming disappointed and resentful. These
negative emotions can then affect their responses. Resentful demoralization may be responsible for
negative control group outcomes that falsely produce the appearance that the treatment was effective.
Antonioli and Reveley appeared to be aware of this possibility and attempted to mitigate it by allowing the
control group to swim with dolphins at the conclusion of the treatment. However, because the control
group was allowed to swim with the dolphins only after the final evaluation, the possibility of resentful
demoralization was not adequately eliminated.
Iikura et al. (2001)
Iikura et al. (2001) sought to determine whether the presence of dolphins enhanced the effectiveness of
seawater therapy for atopic dermatitis. Thirty-six patients with atopic dermatitis were subjected to
seawater therapy (bathing in seawater at a beach) with dolphins present for six days, and 27 received the
seawater treatment without dolphins for six days. No other details were offered about the seawater
therapy, the condition of the patients, the nature of exposure to the dolphins, or any other methodological
components of the study. Yet the authors concluded that all parameters of the atopic dermatitis improved
in both groups and, more important, that the dolphin group experienced “less stress and pain” during the
seawater therapy than the non-dolphin group. Furthermore, the authors claimed that the “patients also
enjoyed swimming with the dolphins.” They concluded that dolphins were “very useful as a pain reliever
during therapy...” (p. 390).
It is difficult to evaluate the methodology and conclusions of Iikura et al. given the marked paucity of
information in the paper regarding how the study was conducted. Moreover, the authors’ conclusions are
vague and subjective. In the absence of appropriate methodology, it can only be concluded that the Iikura
et al. study is wholly uninterpretable and offers no credible support for the efficacy of DAT. Therefore, as
far as can be determined, most of the potential threats to validity described in Table 1 apply to Iikura et al.
Lukina (1999)
Lukina (1999) employed a single-group pretest-posttest design to determine the effects of DAT on
“psychoneurological” functioning in healthy children and those with various diseases. The participants
were 57 healthy children, 30 with “infantile neurosis,” 25 with mental retardation and autism, and 35 with
other unspecified diseases. All of the children swam with and interacted with captive dolphins for 10–15
minute periods over 5–10 sessions. Although Lukina indicated that a number of psychological tasks were
presented to the children during these sessions, she did not describe them. Likewise, although she stated
that her outcome measures included many tests and observations, none were described in sufficient
detail. Therefore, most of the threats to validity in Table 1 cannot be ruled out.
Lukina (1999) reported that the variability of cardiac rhythms in all groups increased after the dolphin
exposure and claimed that “this confirms the redistribution of the psychoemotional dominants in the
course of contacts with the dolphin, a fact that opens possibilities for rehabilitation measures and
psychotherapy” (p. 678), but it is unclear what the author meant by “psychoemotional dominants” or how
they are related to cardiac rhythms. Furthermore, the author claimed that after the dolphin sessions,
parents and facility workers noted the emergence of “new-individual personal qualities,” such as kindness
and self-control. Nevertheless, these observations were subjective and unquantified. Also, Lukina claimed
that many disease symptoms in the “infantile neurosis” group, such as depression, night phobias,
hysteria, and enuresis, diminished and that all the children responded “positively” to the dolphin sessions.
Nevertheless, she offers no data or description of the assessment instruments to support these purported
outcomes. Finally, because psychotherapy was a part of the system of treatment, it is possible that any
improvement in the participants was due to interventions other than the dolphin component per se,
thereby making multiple interference intervention a potential threat to validity.
In addition to these flaws, the pivotal weakness of the Lukina study is the absence of a control group
consisting of children who did not swim with dolphins. Therefore, the study does not meet the minimal
criteria for basic experimental design. This flaw alone renders the Lukina study difficult to interpret even
without the myriad other threats to validity. Although sophisticated single-subject designs for drawing
causal conclusions exist (Shadish, Cook and Campbell 2002), simple A-B (pretest-posttest) designs are
typically extremely limited in internal validity.
Servais (1999)
The design of this study included two experiments, lasting 16 and 14 months, respectively. The first
included three groups of 3 autistic children each. The dolphin group was taught a cognitive task while
interacting with a dolphin and trainer at dockside. The two control groups were a classroom group and a
computer group in which the same cognitive task was taught in their respective settings. The second
experiment included a dolphin group and a classroom group. Servais did not report the actual age of the
participants but indicated that the developmental ages were 1–3 years. All sessions were held on an
individual basis and lasted 15–20 minutes. The dolphin and computer groups were given 10–35
“habituation sessions” to familiarize themselves with the new setting and were required to master some
simple behaviors before moving on to the next phase of the study. All groups received pre-tests followed
by 10–15 “learning sessions” in which other cognitive tasks were given, and then post-tests.
It is not clear whether the pre-tests and the post-tests were the same across or within groups. Therefore,
instrumentation effects (the effects of changing dependent measures at different times in the study) might
be present. Besides post-test performance, the other outcome measure was a rating of attention based
on videotaped observations by the author. Because the author was the only person to code behavioral
outcomes, it is possible that experimenter expectancy effects influenced the findings.
Servais (1999) found that the first dolphin group learned the tasks better than the other groups, but
reported no other differences between groups. In particular, the second dolphin group did not perform
better than the control group. Servais concluded that the positive results were due to the emotional
interaction between the experimenters and children and that better-designed studies may “make the
‘animal effect’ disappear” (p. 14).
Despite Servais’ appropriate cautions about her findings and the effects of DAT, it is worth noting several
threats to validity. In addition to experimenter expectancy effects and potential instrumentation effects,
demand characteristics are impossible to rule out. Moreover, the authors’ conclusion that the dolphin may
not have been as important for improvement as other components of the treatment points saliently to the
possibility that other aspects of the treatment procedure (such as swimming outdoors) besides the
dolphin interaction were responsible for the reported effects.
Webb and Drummond (2001)
Webb and Drummond (2001) investigated the psychological effects of swimming with dolphins at a
marine park in Australia. Participants consisted of a wide age range of male and female teenagers and
adults without known or, at least, measured clinical pathology. They were paying participants and had
been on a waiting list for up to six months. In the well-being study, the experimental group consisted of 74
females and 25 males who swam with four dolphins in groups of four for 25–30 minutes. The control
group consisted of 14 females and 15 males who swam at a beach adjacent to the marina in the absence
of dolphins. Both groups self-reported their feelings of well-being pretreatment and post-treatment. In the
anxiety study, self-reported anxiety ratings were obtained from 12 females and 7 males who swam at a
beach resort for 20–30 minutes in the presence of wild dolphins (the experimental group). The control
group consisted of 13 females and 8 males who swam at the same beach after the dolphins left.
In the well-being study, participants completed a self-report questionnaire designed to examine their
perceived levels of well-being immediately before and after their swim. Psychological wellbeing was
defined as how “positive” each participant felt at the moment. Physiological well-being was defined as
how energetic each participant felt. The participants reported these feelings on a scale of 1 to 100 for
each questionnaire item. In the anxiety study, participants completed the “state” component of the State-
Trait Anxiety Inventory immediately before and after their swim.
Webb and Drummond found that, in the well-being study, the experimental group rated their well-being
significantly higher than the control group prior to the treatment. This finding was presumably due to the
positive anticipation of swimming with dolphins in the experimental group. The well-being ratings of both
groups increased post-treatment, but ratings were significantly higher for the experimental group than the
control group. However, the authors found, following an analysis of covariance, that the pre-treatment
difference between the groups accounted for the differences that persisted after swimming. Therefore,
there was no significant effect of swimming with dolphins on well-being. In the anxiety study, there were
no significant pre-treatment differences in anxiety between the experimental and control groups. The
authors found a significant decrease in self-reported anxiety in the experimental group, but not in the
control group. From these results, Webb and Drummond concluded that anticipation of a new and
exciting experience, and swimming itself, increase well-being. In addition, they concluded that swimming
specifically with dolphins may lower anxiety.
Because, of the two studies, only the anxiety study revealed significant differences between the
experimental and control groups, we focus on the anxiety study here. However, all of the threats to
validity in the anxiety study also apply to the well-being study. In the anxiety study, participants swam in
the ocean during a visit by wild dolphins. Prior to swimming with the dolphins, the participants were told
that dolphins appeared at the site on a daily basis. Therefore, the participants were led to anticipate a
swim with dolphins as they waited on the beach along with members of the public who were also hoping
for a dolphin encounter. The control group swam in the same area after the dolphins left and did not
expect to see dolphins because they were swimming after the period when dolphins usually arrived.
Nevertheless, there are several problems with comparisons of the experimental and control groups. The
authors acknowledged that there was no limit on the number of people allowed in the water while the
dolphins were present. Given that they reported groups of people waiting on the beach to swim with the
dolphins, it seems plausible, if not likely, that there were more people in the water accompanying the
experimental group than the control group. Moreover, there was no apparent effort to make the
experimental and control group comparable on that dimension. Therefore, decreased anxiety in the
experimental condition could have been attributable to interacting with more people.
In addition, if all the participants were swimming together during the dolphin encounter, then it is likely
that they experienced the emotional contagion of positive feelings and excitement that accompany seeing
dolphins. This implies that the “therapeutic agent” in this condition may have been the experience of
swimming with other happy and excited people. The reported effects may have had nothing to do with
dolphins per se. Because the authors did not control for this confounding factor (or, at least did not offer
evidence that they did), we cannot conclude that the significant decrease in anxiety in the experimental
group was due to the presence of the dolphins. Moreover, the authors offered no information concerning
how participants interacted with the wild dolphins or even if any interacted directly with a dolphin.
Because this information was not provided, there is no means of determining to what “treatment” the
experimental group was actually exposed.
As in Antonioli and Reveley (2005), participants were not blind to condition and were asked to self-report
on their anxiety level. Consequently, it is not possible to rule out either demand characteristics or
informant bias. Informant bias based on effort justification may have played a significant role because the
participants were visitors to a local marine park who paid to participate in a dolphin swim experience;
moreover, some waited six months for the opportunity. If the participants in the experimental group were
those who waited longer for their opportunity to participate than those in the control group, then bias due
to effort justification becomes possible.
Conclusions and Summary
In summary, the abundance of serious threats to validity in the five studies we examined renders each of
their conclusions questionable at best, and entirely unwarranted at worst. All of the studies are vulnerable
to nonspecific effects, including placebo and novelty effects, and construct confounding. In addition, each
of the studies contained several other methodological weaknesses that render the conclusions doubtful.
However, these five studies varied considerably in methodological rigor. Of the studies reviewed,
Antonioli and Reveley (2005) is the most methodologically rigorous, as it includes controls for more
potential confounds than the other studies. This study implemented a number of commendable
methodological features, including randomized assignment of participants to conditions, pre- and post-
treatment blind raters of outcomes, procedures to minimize demand characteristics, the use of standard
validated assessment instruments (the Hamilton Depression Rating Scale, the Beck Depression
Inventory, and the Zung Self-Rating Anxiety Scale), and the application of appropriate statistical tests. In
this respect, Antonioli and Reveley’s study is clearly methodologically superior to Nathanson et al. (1997)
and Nathanson (1998). However, even the laudable attempts of Antonioli and Reveley do not eliminate
the presence of important validity threats, such as resentful demoralization and informant bias, nor the
ever-present problems of nonspecific effects and construct confounding. Therefore, Antonioli and Reveley
(2005) falls short of providing a valid test of DAT efficacy and stands as an important reminder of how far
the DAT literature must progress to meet minimal standards of methodological quality.
Of the remaining four studies, none come close to the methodological quality of Antonioli and Reveley.
Most of these studies are plagued by potential experimenter expectancy effects that could have been
eliminated by the inclusion of raters blind to experimental condition. In addition, these studies were
bedeviled by a host of other threats to validity (e.g., history, maturation, testing, regression), as well as a
paucity of reported information concerning basic methods and procedures. Moreover, none of the studies
included a follow-up assessment of the reported short-term improvements.
Most problematic are construct confounding and nonspecific effects (including placebo and novelty
effects), both of which appear to be ubiquitous in the DAT literature. None of the five studies incorporated
adequate controls for these two validity threats. Minimization or elimination of construct confounding
would require that both the experimental group and the control group be exposed to the same or at least
highly similar procedures and stimuli with only the key ingredient—the dolphin per se—as the differential
treatment component between groups.
Placebo effects can be minimized or controlled by a blind study in which participants are not afforded any
information that would provide them with clues regarding their assignment to treatment condition. Novelty
effects can be controlled for by exposure of the control group to another novel, attractive animal (e.g., a
horse, dog, or another aquatic mammal), while keeping as many other variables as possible equal.
Therefore, before concluding that DAT possesses specific efficacy that could not be attained with a
variety of alternative treatment conditions, authors must offer sufficient evidence that the same results
would not be achieved with another large, charismatic, attractive animal used in the same procedures. If
the proper tests were conducted and the control and treatment groups were found to improve equally,
one would need to conclude that there is nothing special or necessary about the dolphin per se in DAT. If
the DAT group improved significantly more than the control group, then perhaps DAT could be
considered a potentially therapeutically interesting intervention, although pragmatic issues of accessibility,
expense, and risk would remain. So far, however, no studies have met this challenge and compared DAT
with an appropriate control group—one exposed to identical procedures using a similar animal—within
the same study.
Given that dolphins are highly attractive and interesting animals to most people, the likelihood of novelty
effects raises particularly troubling concerns regarding the DAT literature. Most worrisome is the
conspicuous absence of evidence for long-term improvement from DAT. Despite DAT’s extensive
promotion to the general public, the evidence that it produces enduring improvements in the core
symptoms of any psychological disorder is nil. Occam’s razor suggests that it is probably most
parsimonious to interpret improvements from DAT as a temporary “feel good effect” of a highly positive
and exciting experience. From this perspective, there is little reason to believe that DAT is a legitimate
therapy or that it constitutes much more than entertainment.
The surprising paucity of scientific evidence for the long-term effects of DAT raises profoundly troubling
ethical questions regarding its widespread use and promotion. There is abundant evidence for injuries
sustained by participants in DAT programs ((Frohoff and Packard 1995; Samuels and Spradlin 1995;
Webster, Neil and Madden 1998). Moreover, interactions between dolphins and humans carry a
significant risk of infections and parasitism for both humans and dolphins (Geraci and Ridgway1991).
Therefore, DAT poses important ethical questions from the standpoint of human and captive dolphin
welfare. At the very least, we believe that DAT practitioners should be required to inform parents and,
when relevant, participants, of the absence of evidence for DAT’s enduring effects on psychological
symptoms. Only then can consumers of DAT make adequately informed decisions regarding the costs
and benefits of this unsubstantiated intervention.
References
Antonioli, C. and Reveley, M. A. 2005. Randomized controlled trial of animal facilitated therapy with
dolphins in the treatment of depression. British Medical Journal 331: 1231–1234.
Brensing, K., Linke, K. and Todt, D. 2003. Can dolphins heal by ultrasound. Journal of Theoretical Biology
225: 99–105.
Cook, T. D. and Campbell, D. T. 1979. Quasi-Experimentation: Design and Analysis Issues for Field
Settings. Boston, MA: Houghton Mifflin.
Frohoff, T. G. and Packard, J. M. 1995. Interactions between humans and free-ranging and captive
bottlenose dolphins. Anthrozoös 8: 44–54.
Geraci, J. R. and Ridgway, S. H. 1991. On disease transmission between cetaceans and humans. Marine
Mammal Science 7: 191–194.
Humphries, T. L. 2003. Effectiveness of dolphin-assisted therapy as a behavioral intervention for young
children with disabilities. Bridges: Practice-Based Research Synthesis 1: 1–9.
Iikura, Y, Sakamoto, Y., Imai, T., Akai, L., Matsuoka, T., Sugihara, K., Utumi, M. and Tomikawa, M. 2001.
Dolphin-assisted seawater therapy for severe atopic dermatitis: an immunological and
psychological study. Archives of Allergy and Immunology 124: 389–390.
Kazdin, A. E. 1994. Methodology, design, and evaluation in psychotherapy research. In Handbook of
Psychotherapy and Behavior Change. 4th edn, 19–71, ed. A. E. Bergin and S. L. Garfield. New
York: Wiley.
Kazdin, A. E. and Wilson, G. T. 1978. Evaluation of Behavior Therapy: Issues, Evidence and Research
Strategies. Cambridge, MA: Ballinger.
Kendall, P. C. and Norton-Ford, J. D. 1982. Therapy outcome research methods. In Handbook of
Research Methods in Clinical Psychology, 429–460, ed. P. C. Kendall and J. N. Butcher. New
York: John Wiley and Sons.
Lukina, L. N. 1999. Influence of dolphin-assisted therapy sessions on the functional state of children with
psychoneurological symptoms of diseases. Human Physiology 25: 676–679.
Marino, L. and Lilienfeld, S. O. 1998. Dolphin-assisted therapy: flawed data, flawed conclusions.
Anthrozoös 11: 194–200.
Nathanson, D. E. 1998. Long-term effectiveness of dolphin-assisted therapy for children with severe
disabilities. Anthrozoös 11: 22–32.
Nathanson, D. E., de Castro, D., Friend, H., and McMahon, M. 1997. Effectiveness of short-term dolphin-
assisted therapy for children with severe disabilities. Anthrozoös 10: 90–100.
Samuels, A. and Spradlin, T. 1995. Quantitative behavioral study of bottlenose dolphins in swim-with-the-
dolphin programs in the United States. Marine Mammal Science11: 520–544.
Servais, V. 1999. Some comments on context embodiment in zootherapy: the case of the Autodolfijn
project. Anthrozoös 12: 5–15.
Shadish, W. R., Cook, T. D. and Campbell, D. T. 2002. Experimental and Quasi Experimental Designs for
Generalized Causal Inference. Boston: Houghton: Mifflin.
Shaughnessy, J. J. and Zechmeister, E. B. 1994. Research Methods in Psychology. New York: McGraw-
Hill.
Webb, N. L. and Drummond, P. D. 2001. The effect of swimming with dolphins on human well-being and
anxiety. Anthrozoös 14: 81–85.
Webster, L. S., Neil, D. T. and Madden, C. A. 1998. Dolphin-initiated inter- and intra-specific contact and
aggression during provisioning at Tangalooma. Special Topic report, Department of Geographical
Sciences and Planning and School of Marine Science, The University of Queensland.
Whale and Dolphin Conservation Society. 2006. Dolphin therapy in the headlines.
<http://www.wdcs.org.au>
