Abstract
Hans Eysenck was head of Psychology at the Maudsley Hospital during the heyday of Sir Aubrey Lewis. Eysenck established the Maudsley as a powerhouse of behaviourism. His clear and lucid style of writing, and his incisive mind, made him a formidable advocate. In 1952, he offered a challenge to the field of psychotherapy. In considering a number of outcome studies he concluded that: They failed to prove that psychotherapy, Freudian or otherwise, facilitates the recovery of neurotic patients. They show that roughly two-thirds of a group of neurotic patients will recover to a marked extent within about 2 years of the onset of the illness, whether they are treated by means of psychotherapy or not [1].
In this article, which marks the landmark fiftieth anniversary of his paper, we reflect upon some effects of the Eysenckian challenge.
Eysenck was born in Berlin in 1916. He left Germany in 1934, first for France and then England where he studied English history and literature at Exeter University and then psychology at London University. He became Director of Psychology at the Institute of Psychiatry at the Maudsley in 1950 and Professor of Psychology at the University of London, in 1955. His history suggests his intellectual scope. He was not a behaviourist of the fundamentalist, closed-mind, kind. His range was wider and included the important notion that a psychology involves an understanding of its basis in brain function. His interest in literature persisted so that, in one of his books, each chapter is headed by a quotation from Alice in wonderland or Through the looking glass [2].
Although seen as an antagonist to psychodynamic thought, he did not begin that way. One of us (RM), was examined by Eysenck nearly 40 years ago. Only one question was asked. It concerned the works of Jung. Jung, of course, formed an important background to the work of Eysenck who took the terms ‘introversion’ and ‘extraversion’ from him, and attempted to make of them scientific constructs.
Jung had worked with Pierre Janet in 1902 and 1903 and was influenced by his idea that non-psychotic illnesses fell essentially into two main groups (i) those making up the hysteria syndrome, (i.e. certain personality disorders, conversion and dissociative disorders, and what is now called somatization disorder), and (ii) psychasthenic illnesses, which are largely anxiety based. Depression could occur in either category. Jung proposed that patients in the first group tended to be extraverted and in the second, introverted. Eysenck wrote that: ‘Psychology owes a debt of gratitude to Jung for this discovery’ [2], p.53].
Like many who become disenchanted with psychodynamic thought through what they perceive as its unscientific basis, Eysenck was not dispassionate in his evaluation of psychotherapy outcome. The offensive he launched against the dominance of psychoanalytic thought in the field of psychotherapy was not strongly based in scientifically gathered data. At that time, there was little appropriately acquired information available. Eysenck used it for the purposes of what was, in essence, a polemic. Although Eysenck provoked censure for the use of inadequate data upon which to mount his argument, the consequence was that, for the first time, serious attempts began to be made to estimate the effect of various kinds of therapeutic intervention.
Some of the earliest responses to Eysenck's challenge came from Bergin [3] and Lambert [4], both of whom remain important figures in the field of psychotherapy outcome. On the basis of their reviews they suggested that Eysenck's figure of 60–70% for spontaneous remission was inflated, and likely to be half of what he claimed. One of the important outcomes of Bergin's response was his discovery of the ‘deterioration’ effect. He found that, at least in some studies, therapy did not produce an outcome that was better than no treatment [5, 6]. However, when the results of psychotherapy were looked at individually, he was able to show that the scatter of change following psychotherapy was much greater than in cases who had not received treatment. The treated group showed larger positive effects in some patients, and larger negative effects in other patients, than would have been expected if they had followed the course of the untreated group. This suggested, perhaps unsurprisingly, that psychotherapy can not only do good but also harm. This confirmed clinical observations that certain kinds of psychotherapeutic practise, often dependent upon rigid adherence to a theory and method, have a potential for the infliction of harm through retraumatization or the undermining of a fragile personal reality [7]. In responding to one challenge, Bergin had implicitly posed another.
Although Lambert and Bergin [8], p.176–180] continue to pursue the issue of the deterioration effect, and have reviewed mounting evidence of its occurrence, this secondary challenge arising from Eysenck's tilt at those forms of therapy to which Anna O. gave the ‘good and serious name’ of ‘talking cure’ [9], p.20] has not been adequately confronted by the psychoanalytic movement in which deterioration is conceived as a ‘negative therapeutic reaction’, (i.e. a reflection of the patient's psychopathology rather than the treatment itself).
Behaviourists also have failed to address the possibility of deterioration as a consequence of their treatments, generally neglecting to report negative effects, such as retraumatization, which might be experienced by drop-outs from therapeutic procedures such as exposure. An important exception to this disregard of the possible malign effects of therapy was the Koach project [10], which was a model of design and reporting transparency. Forty subjects suffering posttraumatic stress disorder following one of the Israeli/Arab wars, were compared with another 40 subjects who received no treatment. The treated subjects were given what was thought to be ‘state of the art’ therapy. Both subjects and therapists were pleased with the results. However, objective measures, taken one year later, showed that the treated group had deteriorated relative to the untreated group.
The main effect of the Eysenckian challenge has been positive. A large number, perhaps 500, of outcome studies show that a variety of psychotherapies, including cognitive behaviour therapy, promote beneficial change. These data are reviewed in volumes such as Roth and Fonagy [11] and in Bergin and Garfield's Handbook [12].
Although the effect is clear when comparisons are made between psychotherapy and wait-list or no-treatment controls, the effect is less clear when the comparison group have ‘placebo’ or ‘non-specific’ treatment.
The placebo effect is a powerful one, beyond mere acquiescence to the hopes and wishes of therapists. For example, a study of the effect of particular medications on depression outcome measures included event-related potentials (ERP), certain amplitudes of which are raised during depression – the greater the depression [13, 14] the higher the amplitude [15]. In order to exclude the so called placebo responders, all patients were given placebo for one week. Both before and after this week, assessment included ERP recording in a laboratory. An impression was thus created of professionalism and of an organized and dedicated system of care. At the end of the week, about one-third of the sample had remitted in terms of depression. Moreover, the ERP amplitudes had changed to the non-depressed level. It seems improbable that such an alteration in sensory processing is the manifestation of ‘spontaneous remission’. Rather, it is likely that the social environment created by the treatment programme had this effect.
Lambert and Bergin conclude, on the basis of their review, that psychotherapy is superior to non-specific (or placebo) treatment which, in turn, is superior to no-treatment [8] p.149–152]. These conclusions lead to the possibility that formal psychotherapy consists of a skilled, systematic and disciplined use of the so called non-specific factors, (i.e. the relationship, which is seen as the vehicle is the treatment ‘technique’). This is supported by a body of evidence reviewed by Orlinsky et al. [16] who concluded, on the basis of a comprehensive review, that the association between good outcome and the quality of the therapeutic relationship was one of the strongest findings in the outcome literature, seeming to confirm Rogers’ [17] view that the therapeutic relationship provides the ‘necessary and sufficient’ conditions for beneficial change.
The significant effect of the therapeutic bond is evident even in those therapies which emphasize a particular technique. For example, Luborsky et al. [18] set up a study designed to minimize the effect of a therapist and how he or she related to patients. The therapists were rigorously trained to conform to a manual in order to maximize the effect of technique. Despite this, profound differences were discovered in the therapists’ success with the patients in their case load. A second example is provided by a study of cognitive behaviour therapy in depression. Therapist empathy was robustly associated with clinical improvement [19].
The possibility arises, on consideration of data such as these, that something beyond the specific technique characteristic of a particular treatment is the main therapeutic factor. If this were the case, most psychotherapies would have a similar effect. This appears to be supported by studies that compare well-validated forms of psychotherapy. For example, in the National Institute of Mental Health (NIMH) collaborative study on the treatment of depression [20], compared interpersonal therapy (IPT), cognitive behavioural therapy (CBT), medication, and placebo. There was little difference between the treatment groups at the end of treatment. All three were more effective than placebo. In another example, Fairburn et al. [21, 22] compared behaviour therapy, CBT and IPT in cases of bulimia and found at follow-up, one to six years later, that the effect of behaviour therapy was short-lived but that IPT and CBT had positive and roughly equivalent therapeutic benefits. Shapiro et al. [23, 24] compared CBT with an abbreviated form of the Conversational Model, which they termed psychodynamic-interpersonal, in the treatment of depression and found similar effects. This kind of evidence led Stiles, Shapiro and Elliot [25] to consider that ‘currently validated’ therapies are equally effective, a later version of Luborsky's verdict [26], recalling that of the dodo bird in Alice in wonderland, who declared that: ‘Everyone has won and all must have prizes’. Despite these early conclusions, it seems likely that, as more evidence accumulates, differences will be found in the effect of specific treatments of particular conditions.
The validated psychotherapies do not include psychoanalysis. It is unlikely that they ever will since validation studies must be prospective; concern a particular and well-defined condition; have a clearly described treatment method; have a specific duration; have an appropriate comparison group; use valid and reliable methods of assessment at entry, conclusion and follow-up; and involve audio taping of therapeutic sessions in order to judge therapist adherence to the prescribed method. Despite the difficulty of validating psychoanalysis, important studies have been made of the effect of the technique which is seen as central to its practice, namely interpretation of the transference.
It is a widely held psychoanalytic view that insight through interpretation is the ‘supreme agent in the hierarchy of therapeutic principles’ [27]. Yet it is only in recent years that empirical studies have been conducted into its effect. Henry et al. [27] reviewed these studies. They drew upon work such as that of Piper et al. [28], who in a very detailed study, necessarily involving audio tapes of sessions, rated a total of 22 500 therapist interventions. These workers reported an ‘inverse relationship between the proportion of transference interpretations (out of the total amount of therapist intervention) and patient-reported alliance as well as outcome’. Such investigations make a contribution to an understanding of the ‘deterioration effect’.
In concluding their review of this and similar studies, Henry and his colleagues remarked that the data ‘paint an emergent picture that challenges some long-held assumptions’. An appropriate response to these findings is neither to reject them as irrelevant to clinical practice or to accept them as ‘fact’. Rather, if one believes, as we do, that the management of transference phenomena is a central aspect of therapeutic work, the challenge leads to questions that will be the basis for future research. How are transference phenomena to be understood? What theraputic responses to them appear to be beneficial?
Eysenck's challenge was overtly an appeal to the scientific ideal but covertly a polemic. The overt challenge has been reasonably fruitful. However, the covert one has had a negative consequence. At a time when research is showing the essential principles of different therapeutic approaches, partisan advocacy of an approach tends to diminish the value of the scientific data. Eysenck used inadequate data to mount his argument. Although current data are superior they also have deficiencies which are often overlooked in the debate on the efficacy of various psychotherapies. The deficiencies are not, at first sight, obvious. Without being able to cover their range in this brief space, something of their nature is illustrated by considering of two well-known studies.
In the first, the aforementioned NIMH depression study [29], 239 patients entered treatment and were divided into four groups. Only 47 (i.e. 20%) completed the 16-week treatment, recovered, and maintained that recovery 18 months later. These 47 patients included 24% of the CBT group, 23% of the IPT group; 16% of the impramine group and 16% of the placebo group. Although the psychotherapies did slightly better than the others, the difference was not significant. Thus, all three treatments were judged effective at the end of treatment but were no better than placebo and clinical management at follow-up. The second illustration is provided by a CBT based treatment of parasuicidal behaviour in borderline patients [30, 31]. At the end of 12 months treatment, good results were achieved. However, one year later the parasuicidal behaviour did not differ from controls, while suicidal ideation and depression were undiminished.
These studies are models of current design. They are prospective, controlled and the therapeutic methods are clearly described. Most importantly, they include followup and, as in the admirable Koach study, they do not fail to report negative findings. Not all current studies are of this standard.
Much of what is said about the effectiveness is based on findings at the end of treatment rather than at followup. Furthermore, much current data relates to short-term therapies which may not have enduring effects or resemble day-to-day practice. The evidence of Orlinsky et al. [16] suggested that duration, or ‘dosage’, is positively related to outcome. There is a dearth of studies of longer-term therapy involving more complex kinds of psychiatric problems. Inferences from incomplete outcome evaluations mitigate against the support of such longer term therapy.
In conclusion, although the effect of Eysenck's paper has had lingering negative consequences, the overall effect of his challenge has been valuable, leading to the production of significant data which have led to further challenges, and to the stimulation of a specialized field of inquiry.