Behar. Res. Thu. Vol. 29, No. 3. pp. 283-292. 1991 0005.7967191 $3.00 + 0.00 Printed in Great Brilain. All rights reserved Copyright <c 1991 Pergamon Press plc

THE TRIPLE RESPONSE APPROACH TO ASSESSMENT: A CONCEPTUAL AND METHODOLOGICAL REAPPRAISAL

GEORC H. EIFERT’* and PETER H. WILSON*

‘Division of Psychology, James Cook University of North Queensland, Townsville. Queensland 481 I and lDepartment of Psychology, University of Sydney, Australia

(Received 16 August 1990)

Summary-Despite its positive effects on the increased use of multiple assessments and improved assessment validity, the triple-response concept has led to some conceptual and practical confusion. This is mainly due to two problems: (I) a confounding of the content and method of assessment; and (2) an imprecise and vague use of the ‘verbal-subjective mode’ which has been expanded to include cognitive elements since the introduction of cognitive-behavioural theories and treatments. A new matrix is proposed that clearly distinguishes content and method of assessment. It also defines a separate cognitive/information-processing content area and introduces affect as an additional content area. Thus, four content areas are suggested: behavioural, physiological, cognitive, and affective. which can be measured in three different ways: by means of self-report, observation, and instruments or technical equipment. We point out the implications of these changes for (I) a more appropriate selection of assessment procedures and outcome measures in clinical research; (2) a more adequate individualisation of treatment through matching individual response profiles to specific treatments; and (3) an improved understanding of the interrelationship between behavioural, physiological, cognitive, and affective processes in anxiety and depression. Finally, we suggest that the lack of agreement between measures of physiological, cognitive, behavioural and affective changes in some studies may be as much a reflection of the lack of agreement arising from spurious sources of variance within content areas as it is a reflection of the operation of different processes and systems.

In behavioural research it has become customary to define emotional problems as involving changes in three ‘response systems’ or ‘modalities’: overt behaviour, physiological activity, and verbal-cognitive components. Overt behaviour typically refers to directly observable aspects of the problem, for instance, frequency of panic attacks, number of drinks consumed per day, engagement in certain activities. Physiological activity refers to those features of the problem which involve some physiological change, such as perspiration, heart rate, respiration, or muscle tension. The verbal-cognitive mode typically refers to the verbal report of thoughts which precede, accompany or follow the occurrence of events or behaviours but it may also include the verbal report of distress, arousal or behaviour.

The wide acceptance of the triple-response mode concept of fear and other emotional disorders has not only resulted in more comprehensive assessments but has also generated useful hypotheses about the interrelationship between the behavioural, physiological, and cognitive aspects of emotional problems and has thereby led to more sophisticated and carefully targeted treatment programs. As many psychological problems involve changes in more than one of these systems, assessment and treatment can be directed at the evaluation of all three systems. For example, a person with a public-speaking phobia may manifest avoidance of speaking situations, thoughts about appearing foolish prior to being in a speaking situation, and an increase in heart rate when exposed to the speaking situation. Such information is helpful in designing specific assessment devices and in selecting treatment components which are likely to be relevant to the manifestation of the problem in a given individual. In addition, the clinical validity of assessment and outcome measures has improved because clinical researchers have tended to include a multitude of measures tapping into all relevant responses systems rather than just relying on one measure or a set of measures all employing the same method (e.g. self-report).

The central statement of this article is that these positive developments have been obscured by a frequent confounding of what is measured (content of assessment) with how it is measured

*TO whom all correspondence should be addressed: Division of Psychology, University of Mississippi Medical Center, 2500 North State Street, Jackson, MS 39216, U.S.A.

283

284 GEORG H. EIFERT and INTER H. WILSON

(method of assessment). This confusion has been exacerbated by a vague and inconsistent definition and operationalisation of the verbal-cognitive content area. We will therefore present a matrix model that will allow a clear distinction between the content and method of assessment. Consequently, the verbal-cognitive-subjective ‘mode’ will be redefined by distinguishing between a cognitive/information-processing content area and self-report as a measurement technique. In addition, we will introduce a fourth content area relating to affect and mood changes. Finally, the implications of these clarifications and changes will be indicated for (I) the design of research methodology, particularly the selection and implementation of assessment procedures in treatment outcome studies; (2) the use of a conceptually driven specific matching of treatment techniques to the particular response profiles of individual clients; and (3) the development of an improved understanding of response interrelationships in emotional dysfunctions such as depression and anxiety.

PROBLEMS OF THE CURRENT TRIPLE-RESPONSE MODE CONCEPT

The confusion of content and method of measurement

When Lang (1968) originally introduced the triple response concept, he argued that emotional behaviours are multiple system responses. He maintained that three different systems, controlied by partially independent brain centers, may be involved in the expression of emotion: motor, physiological, and verbal-cognitive. Despite the advantages of a loose definition of these response systems and their flexible use in treatment studies, a serious problem has arisen which appears to have confused researchers in the design of studies and the selection of outcomes measures. The problem has been the lack of a clear distinction between content and method of assessment. Evans (1986) indicated that this confounding may have also contributed to increasing the discordance between modes in the sense that it could be the confounding of different methods, and not differences between the modes per se, which may account for the existing evidence of low correlations between modes. The general problem was clearly recognised as early as 1979 in an article by Cone entitled “Confounded comparisons in tripie response mode assessment research,” but his arguments have failed to make the impact that they deserve. Cone provided a matrix which basically included two modes or content areas (motor and physiological) and two methods of assessment (self-report and direct observation) for measuring different behaviours. Using the example perspire, a response that falls in the physiological content system, he argued that it can be measured either by self-report (‘I often perspire’) or by direct observation (‘the client perspires’). Evans (1986) pointed out that even though perspiring can certainly be regarded as a psychophys- iological response, content and method have been so confounded in previous work that seff- reported perspiration would probably be classified by most investigators as a cognitive measure. This simple example shows the confusion which may arise when content and method are not clearly separated.

Although Cone’s matrix is very useful, it is incomplete because he did not include the cognitive content area in his matrix. In addition, he only incorporated two methods of assessment (self-report and direct observation) and omitted an important type of method which uses instruments or other technical and computer equipment to assess behavioural, physiological or cognitive activity. This type of method could also be described as “indirect observation’ in the sense that these measures rely on technical instruments or computer equipment to produce observable data. Hence, the matrix presented in this article includes this important third type of method and it also redefines the verbal-cognitive content area, as will be indicated.

The ~erbal-cog~~t~L~e-sMbject~~e mode: ~~~-de~ned and in need of reaision

When the notion of a triple-response mode of assessment was initially translated into measures for treatment outcome studies, the motor mode was typically some form of direct observation of overt behaviour (e.g. a behaviour avoidance test) and the physiological mode a polygraph based assessment of arousal. While these two modes appeared to be relatively straightforward in their operationalisation, the definition of the verbal-cognitive mode was much more difficult and problematic. This ambiguity was not so evident in early studies (e.g. Miller & Bernstein, 1972) where the verbal-cognitive mode generally referred to subjective anxiety (e.g. self-report ratings of

Triple response approach to assessment 285

fear such as the SUDS scale) and was usually equated with self-report measures (see Hodgson &

Rachman, 1974). With a growing interest in cognitive dysfunctions and cognitive therapy, and hence cognitive

measures, in the late 1970s and early 1980s the definition and operationalisation of the verbal-cognitive mode became more ambiguous and varied. This ‘mode’ frequently tended to encompass all measures involving any form of self-report. For instance, self observation of physiological activity is frequently assessed with questionnaires asking clients to report and rate their body sensations (e.g. the Body Sensation Questionnaire; Chambless, Caputo, Bright & Gallagher, 1984) whereas overt behaviour may be assessed with questionnaires such as the Mobility Inventory (Chambless, Caputo, Jasin, Gracely & Williams, 1985) which measures the degree of avoidance behaviour of persons with agoraphobic problems. In addition, the verbal- cognitive mode appears to refer to a variety of measures of thoughts and styles of thinking such as causal attributions, self-efficacy judgments, distorted or irrational thinking, and so forth, all of which can be assessed with formal questionnaires as well as informal self-monitoring procedures. It is therefore noteworthy that, in a more recent article, Lang and Cuthbert (1984) no longer refer to a verbal-cognitive area but merely describe this ‘general category of measurable responses’ as “verbal reports of distress” (p. 371). While self-ratings of severity or verbal reports of distress are indeed very important and should be part of every clinical assessment, they are neither measures merely of behaviour or physiological activity nor are they exclusively measures of cognitive functioning or information-processing. They are, however, part of an increasing number of attempts to assess the affective state of individuals-an area that has received considereable attention in recent times (e.g. Barlow, 1988; Salkovskis & Clark, 1990; Eifert, 1990).

The situation has been further complicated by the categorisation of deficits in attention and hypervigilance as “observable fear-related behavioural acts” (Lang & Cuthbert, 1984). This appears to be another example of the confounding of what is measured with how it is measured. In our view, deficits in attention and hypervigilance refer to information-processing phenomena. Hence, they belong into the cognitive content area which may be assessed in a number of different ways, but we would not refer to them as behavioural acts or measures of behaviour. This is not to say that such deficits have no behavioural consequences or effects. Recently, there has indeed been a very promising development of what one might call behavioural measures of cognitive processing. For instance, methods and instruments which have been employed for years in experimental cognitive psychology (e.g. the STROOP test) have been used to assess deficits in attention and the encoding and recall of relevant information of persons with specific phobias (Watts, McKenna, Sharrock & Trezise, 1986) agoraphobia (Watts, 1989) obsessive-compulsive problems (Sher, Frost, Kushner, Crews & Alexander, 1989) and depression (Derry & Kuiper, 1981; Teasdale, 1983; see also Williams, Watts, MacLeod & Mathews, 1988).

In order to avoid further confusion, we propose to disentangle the ‘verbal-cognitive-subjective’ mode by defining a separate cognitive content area that only includes elements which refer to information-processing phenomena and their classification (e.g. cognitive styles). Furthermore, the matrix in Table 1 proposes that self report (verbal and nonverbal) is not a content area or ‘mode’

Table 1. Classification of content x methods-of-assessment matrix. Measures are only listed as examples to illustrate the distinction between content and method of assessment

Method of assessment

Content to be assessed

Motonc

Physmlogical

Cognitive

Affective

Self report (verbal and nonverbal)

Mobibty Inventory, daily activity logs

Body Sensation Questionnaire. verbal reports of arousal

Attributional Style Questionnaire. thought-listing

Adjective Check Lists Mood Visual Analog Scales Self Assessment Manikins

Observation

BAT, behaviour samplmg

Perspiring, blushing, goose bumps

Response latency (e.g. in recall tasks)

Facial expressions. body posture

hStW”E~t!i\p~d~dtUS

Pedometer, activity meter

GSR. EMG, ECG etc.. Plethysmogrdph

STROOP Test. PET scans

Not yet available

286 GEORG H. EIFERT and PETER H. WILSON

but rather a method of assessment. Moreover, subjective information that refers to the experience of affective or mood states should be subsumed under the rubric of affect.

Behaviour, physiology, cognition, and afeet: a new matrix for their assessment

Another problem with the traditional triple response mode concept is the difficulty of classifying measures which are designed to assess a person’s affective state such as dysphoria and the subjective experience of fear, anger, and other forms of distress. Can these affective states be represented adequately and sufficiently in the existing three systems or do they require a separate category? In earlier studies subjective anxiety was considered to be part of the verbal-subjective mode, but as indicated in the previous section, this verbal-subjective mode has subsequently been transformed into a verbal-cognitive mode. This development was logical and inevitable with the advent of cognitive-behavioural theories and treatments. On the other hand, it has widened the scope of that mode so much that is has become conceptually almost meaningless. Unless one is prepared to argue that any form of verbal self-report, styles of thinking and information-processing, and affective mood states are somehow part of the same ‘system’, we need to separate the verbal-cognitive content area from an affect area. Without rekindling the argument about whether affect and cognition are separate entities or systems (cf. Lazarus, 1984; Zajonc, 1980, 1984). there appears to be some agreement among researchers and clinicians that affective states are important and sufficiently different from cognitive styles, information-processing, and also physiological arousal and overt behaviour to warrant the introduction of a separate measurement content area relating to affecting states. In addition, several researchers have called for a more careful scrutiny of affect in the understanding of anxiety (e.g. Barlow, 1988; Eifert, 1990) and depression (Heiby & Staats, 1990), and have given affective concepts and processes a central place in their models of such disorders. In addition, Rachman (1981, 1984) has argued for the modification of affect to become a direct intervention target in behaviour therapy, and studies are beginning to emerge which attempt a more direct modification of affect (Eifert, Craill, Carey & O’Connor, 1988). In the matrix presented in Table 1, affect has therefore been included as a fourth content area in addition to the traditional content areas of behaviour, physiology, and cognition.

As pointed out earlier, in much of the discussion and application of the triple-response approach to assessment, there has been a failure to disentangle the content to be assessed from the method of assessment. For example, individuals may be asked to keep daily diaries of their avoidance behaviour, anxiety-related thoughts and their physiological responses, representing different content areas but confounded with the assessment method, in this case, self-report. We therefore propose that a clear distinction needs to be made between the content to be assessed and the method of assessment, as illustrated in the matrix presented in Table 1.

Within the motoric content area, methods may include self-report (e.g. daily log of activities or recording of problem situations), direct observation (e.g. behaviour avoidance test or behavioural sampling of interactions of spouses or children and parents), or technical instruments and equipment (pedometer, activity meter). In the physiological content area, methods may include self-report (e.g. the Body Sensation Questionnaire), the direct observation of person’s blushing or perspiring, and indirect observation methods that require the use of instruments or technical equipment such as a polygraph to measure GSR, EMG, sexual arousal and so on. Within the cognitive content area, methods may include self-report (e.g. daily recording of thoughts, Automatic Thoughts Questionnaire; Hollon & Kendall, 1980) or instrument and computer-based techniques such as the STROOP Test that allow an indirect assessment of cognitive processing. Finally, the affective state a person experiences may be assessed by self-report such as visual analogues scales of mood and affective adjective check lists, but may also include verbal self-reports of distress and symptom severity. However, studies have also successfully attempted the direct observation of affective states or experiences by means of facial expressions (Ekman. Levenson & Friesen, 1983) and body posture and movements.

Although the matrix provides a clear separation of method and content of assessment, not all content areas are equally suited and accessible to all three methods of assessment. For instance, it is currently impossible to observe cognitive events or information-processing directly. On the other hand, general levels of cognitive activity may become more ‘visible’ in future with the increasing availability of sophisticated computer technology such as PET scans. Likewise. it is

Triple response approach to assessment 287

currently only possible to use self-report or observation to gain information on a person’s affective state. At present most self reports are of a verbal nature, but nonverbal means of assessing affective state have also been devised. A good example are the Self Assessment Manikins developed by Hodes, Cook and Lang (1985). These ‘SAMs’ are computer-generated graphic displays such as facial expressions of different affective dimensions varying from smiling to frowning which subjects can change and select to represent their current affective state. The development of other instruments to measure affect by nonverbal means is certainly highly desirable and remains a challenge for the future.

Furthermore, there can be some overlap in categorising the same assessment device as

representing two types of method or even two content areas. An example of the latter type of classification problem are response latency measures in lexical decision tasks (the identification by subjects that a set of letters is either a word or a non-word). Speed of response is clearly part of the topography of behaviour and could therefore be regarded as a behavioural measure, but the crucial component of the measurement is usually the difference in response latency to different types of emotional word stimuli. As different response latencies reflect differences in the accessing of different cognitive networks, we would regard this measure as an assessment of cognitive content or processes. Nevertheless, these classification problems do not diminish the conceptual and practical clinical usefulness of the matrix presented here; in fact, they only strengthen the case for a clearer separation between content and method.

IMPLICATIONS FOR TREATMENT AND CLINICAL RESEARCH

Selection and implementation of assessment procedures in treatment outcome studies

Much of the research on the triple-response mode concept has revealed a lack of agreement between the three modalities (see Hugdahl, 1981). The lack of agreement has been evident both in studies investigating concordance and discordance-that is, correlations between measures at a particular point in time-as well as in studies investigating synchrony and desynchrony which refer to the degree of correlations between change scores. As a result of these observations, it has been proposed that each system may be controlled by different organismic and environmental factors. An alternative view is that the lack of agreement between the modalities could be expected to occur for methodological reasons. First, the agreement between variables within one response system or content area is often low, placing a restriction on the potential size of agreement between response systems when only one measure is selected from each content area, a problem which is only amplified by the confusion between content and method. This point is particularly evident in the assessment of physiological activity, such as heart rate, muscle tension or galvanic skin response and neuroendocrinological changes. Any one of these measures may provide a poor reflection of the physiological aspect of the particular problem under investigation because there are variations both within and between individuals in the propensity of the various physiological indices to change (Abelson & Curtis, 1989). Similarly, there are often low correlations within sets of measures of overt behaviour and cognitive content. One might generally expect correlations within content areas to be greater than those between content areas. Thus, the researcher who selects only one measure from each area will be unlikely to find high correlations between the areas and may prematurely conclude that desynchrony or discordance exist in the disorder under investigation.

Second, as we have indicated above, the variety of assessments available under the rubric of cognitive measurement is also problematic in understanding the relationships between content areas. Many of these types of assessments can be conducted in several different ways and at different points in time in relation to behaviours and environmental events. Self-report assessment of cognitions may include measures of attributions about events which have already occurred (e.g. Attributional Style Questionnaire; Peterson, Semmel, von Baeyer, Abramson, Metalsky & Seligman, 1982) predictions about events or behaviours that may occur in the future (e.g. self-efficacy scales), automatic thoughts that have occurred in the past (e.g. Dysfuntional Attitude Scale; Weissman & Beck, 1978) or are occurring at the time of testing (e.g. thought sampling techniques), and numerous other procedures such as self observation of physiological activity or behaviour and self-ratings of symptom severity. As already mentioned, different degrees of reliability and validity of the measures within a particular type of assessment may affect the

288 GEORG H. EIFERT and PETER H. WILSON

correlations of measures between different contents of assessment. In addition, the different temporal relations between the implementation of any of these assessment methods and other potentially important environmental, behavioural, and physiological events will presumably contribute to variation. Thus, the failure to find agreement between measures of physiological, cognitive, behavioural, and affective changes may be as much a reflection of the lack of agreement arising from unwanted sources of variance of measures within systems as it is a reflection of the operation of different systems and processes.

The present discussion highlights the need for some changes to be made in the methodology used in treatment-outcome research so that instances of true process desynchrony can be discerned from those which result from methodological desynchrony. Future research should focus on the inclusion of more than one measure of each content area preferably measured with more than one method. This step would enable variance related to the method to be distinguished from variance related to the content. Researchers may also need to develop strategies for combining several measures of the same ‘content’ into a single measure, possibly by using factor analytic, multiple regression. and canonical correlation methods. Thus, measures of different content and methods could be subjected to a factor analysis which would provide several linear combinations of the measures, resulting in both a reduction in the number of variables and an increase in the coherence and reliability of each set of measures. The factor scores, representing each S’s score on each factor, can then be subjected to an ANOVA in the usual way. Alternatively, the researcher could make use of canonical correlation analysis. In this approach, the linear combination of measures representing the four content areas suggested in our matrix could be entered as the ‘criterion’ or ‘dependent’ variable side of a canonical correlation analysis. As in multiple regression (or, in fact, in ANOVA), the contrast representing the comparisons between various treatment and control conditions becomes the ‘predictor’ or ‘independent variable’ side of the equation.

Greater care should also be taken in the selection of different measures to ensure that they assess the most relevant aspects of the clinical problem under investigation. The selection of relevant self-report cognitive measures is particularly problematic because of the great variety of measures which have been proposed in this domain. For example, care needs to be taken to match the particular type of cognitive dysfunction that is supposedly responsible for a particular problem with the assessment instrument that most directly targets that particular type of cognitive dysfunction. This argument also exemplifies the close relationship between a clear understanding of the dysfunctional processes involved in a given disorder and the choice of the most appropriate assessment and effective treatment. In addition, the time-scale over which the measurements are taken should be given more consideration. For example, a physiological measure may relate to a very discrete moment in a laboratory setting, while a self-report measure may make reference to a retrospective period of one week or more. Closer matching of time frames across content areas is essential for a better understanding of the relationship between physiological, cognitive, behavioural, and affective components of a particular disorder, since discrepancies in information may be as much due to differences in the time scale as to genuine discordance between response systems. Measures such as the real-time computer diary reported by Taylor, Fried and Kennardy (1990), which can be used by Ss in the natural environment and subsequently interfaced with laboratory based computer systems, provide great potential for further developments and improvements in this area.

Another aspect influencing the degree of concordance and synchrony between measures is the potential influence of one measurement procedure on other measures in the total set. It is possible that there is an increase in validity and reliability of one measure when Ss are implicitly receiving feedback from other measures. For example, the accuracy of daily reports of sleep onset latency may be increased when an objective device, such as a sleep monitor in which subjects are required to press a button when they are awake in response to a faint auditory signal (Birrel, 1983) is included in the assessment battery. It could be argued that Ss receive a prompt for estimating their sleep-onset latency when they can recall the approximate number of button presses. Even without feedback. Ss may alter self-ratings when they are aware that other assessments, such as physiological ones, are being utilised. Similarly, behavioural tests may be affected by prior completion of a self-efficacy questionnaire because the S may attempt to display consistency in responding. The problem highlights the need for random or balanced administration of assessment

Triple response approach to assessment 189

procedures whenever possible and for the careful consideration of conclusions when such controls have not been implemented.

&latching of treatments to individuai client response profiles

Despite the proliferation of large-scale and very carefully controlled treatment outcome studies, the results of these studies have frequently been disappointing (cf. Wolpe, 1989). Studies typically yield significant pre-post within-group changes, that is, clients in most or all treatment groups improve but there are often no differences between treatment groups. Eifert, Evans and McKendrick (1990) have recently pointed out that this frustrating ‘state of the art’ is to some extent due to a neglect of individual differences between clients falling into broad diagnostic catergories and a resultant lack of treatment individualisation.

Rather than a careful specific matching of treatments to individually assessed problems in the various response repertoires, all clients are randomly assigned to one of the treatment conditions with all clients in a given condition receiving the same treatment. In view of the interindividual differences between response profiles of clients falling into the same diagnostic group, the unimpressive and inconclusive results of many outcome studies are not surprising.

Commenting on findings that clients in the same diagnostic category differ in their primary or predominant response pattern, Michelson (1984) indicated that one of the main clinical implications and advantages of the tripartite model of fear was the possibility of matching a client’s predominant response profile with a treatment which is specifically aimed at changing that particular response profile. In the prototypical treatment matching experiment, subjects are divided into two groups on a variable which is hypothesised to be important in determining the outcome of a particuIar intervention, and then randomly allocated to one of two treatments. For example, in a study by &t, Jerremalm and Johansson (1981) social phobic 5’s were divided into those presenting with mainly behavioural or mainly physiological manifestations of anxiety. These two groups then randomly received either a behavioural treatment (social skills training) or a physiological based treatment (applied relaxation). Thus, two cells in the resulting 2 x 2 design are seen to represent appropriate matching of clients to treatments, and the other two cells represent inappropriate matching. The interaction term in the ANOVA enables the researcher to determine the extent to which treatment response is related to the hypothesised client variables. It was indeed encouraging to read the results of these early studies by &t and his colleagues that treatment gains were enhanced when the specific treatment was consonant with the individual’s specific anxiety profile identified prior to treatment. Yet subsequent studies investigating the matching and mismatching of response profiles and treatments yielded inconclusive results, with some studies providing mixed support (e.g. Mackay & Liddell, 1986) and others not supporting such matching (e.g. Mersch, Emmelkamp, BGgels & van der Sleen, 1989; ijst, Jerremalm & Jansson, 1984).

It is unfortunate that such results may have contributed to the current widespread neglect of response-profile matching and treatment individualisation because they do not necessarily show that response profile-treatment matching is unnecessary and superfluous. These inconclusive results may merely indicate (1) there is more than one basis for matching clients to particular treatments; and (2) the triple response system model, as it was used, was too imprecise and stili flawed with methodological and practical problems which, we argue, can be solved, In addition, the matching of profiles and treatments may have been too broad and crude in some previous studies. For instance, the appropriate matching of ‘physiological responders’ in an agoraphobia study is not necessarily guaranteed by assigning them all to a relaxation training group. This treatment may be appropriate for those clients who are predominantly tense and overaroused, but it could be argued that clients who display a large number of physiological symptoms as a result of hyperventilation would probably benefit more from a breathing retraining program rather than a treatment that focuses on muscle relaxation. Similarly, the appropriate treatment for some ‘cognitive responders’ could be attribution retraining by means of exposure to body sensations, whereas the treatment of other cognitive responders could include restructuring with a focus on dysfunctional beliefs and catastrophising tendencies. Other cognitive responders may benefit more from a self-instructional program in which they are taught to use specific self-statements to prevent avoidance/escape behaviours and are encouraged to engage in approach behaviour.

290 GEORG H. EIFERT and PETER H. WILSON

In other words, the results of response profile-treatment matching could be more positive if there was no confounding of content and methods of assessment, a more careful definition and operationalisation of the content areas, particularly the cognitive and affective content area. Most importantly, however, will be a more direct matching of treatment techniques with the specific problems identified Gthin each response class as they relate to the specific dysfunctional processes of particular clients. For example, in their paradigmatic behavioural model of depression, Heiby and Staats (1990) have identified some specific relationships between biological variables, emotional, cognitive, and behavioural deficits, and environmental conditions for various subtypes of depression. The treatment is then specifically matched to the particular type and combination of dysfunctional processes identified in a given client. Heiby (1986) demonstrated in several single-case design comparisons that if treatment (e.g. social skills vs self-control skills) matched the particular deficits of depressive clients, better therapeutic results were obtained than if such matching did not occur. Another study with depressive persons by McKnight, Nelson. Hayes and Jarrett (1984) also found positive results for matching treatments with deficits in individually assessed response classes (social skills vs irrational cognitions). In other words, this approach to treatment matching is not so much concerned with matching treatments to symptoms in broadly defined response systems but to the specific deficits and dysfunctional processes of particular groups of clients.

Finally, it should be noted that treatmentxlient matching will only be successful if the assessment of the systems or content areas is itself valid and maximises variances related to the assessment content rather than variance which may be due to some artefact of the assessment procedures. The researcher may fail to identify an important effect for matching of clients to treatments simply as a result of the selection of less reliable or less valid assessment devices. Thus, there is a risk that the client-treatment matching procedure may be abandoned prematurely.

Improved understanding of response interrelationships in emotional disorders

The study of relationships between behavioural, physiological, and cognitive processes has assumed an important place in understanding the nature of emotional disorders (Barlow, 1988; Eifert & Craill, 1989). It is generally assumed that a comprehensive account of the causes and maintenance of psychological disorders requires an account of the role of each separate content or system of processes as well as an understanding of the relationships that may exist between them. For example, phobic anxiety is sometimes considered to be simply the affective state that arises from a certain set of cognitive (anticipation of danger), behavioural (avoidance and escape behaviours), and physiological responses (perspiration, increased heart rate). However, it can be argued that behaviour, cognitions, and physiological activity may change at different rates and in different sequences (Rachman & Hodgson, 1974). It is also possible that some of these responses may be more influenced by environmental factors than others which might be influenced more by biochemical or ‘internal’ factors.

The definition of ‘affect’ and its role in anxiety and depression represents a related issue. Some conceptual models have considered affect to be essentially synonymous with the disorder, that is, a linear combination of the three response systems and their interactions. Other theorists (e.g. Zajonc, 1980, 1984) have considered the concept of affect to be at least a partially separate and independent entity which could be argued to represent a fourth ‘mode’ or content area and response system. This is also reflected in some recent models of anxiety and depression where affective processes and concepts have either been given a central role (Barlow, 1988; Eifert et al., 1988; Heiby & Staats, 1990) or they are at least considered to be of equal importance as behavioural, physiological, and cognitive processes (for example, in the recent model of hypochondriasis by Warwick and Salkovskis, 1990). Research which is capable of testing such propositions about the nature and role of affect requires that the content of measurement is clearly distinguished from the method of measurement.

CONCLUSION

The main argument presented in this article is that the potentially positive contributions of the

triple-response-mode conception of assessment has been obscured by a frequent confounding of

Triple response approach to assessment 291

content and method of assessment. This problem has been exacerbated by a vague and inconsistent definition of the verbal-cognitive content area and the failure to take account of affective responses as a separate class. We have presented a matrix model in which an attempt is made to rectify these problems by making a clear distinction between the content and method of assessment. We have also disentangled the ‘verbal-cognitive-subjective mode’ by distinguishing between cognitive/information-processing as a content area and instances of self-report as a method for the assessment of behaviour, physiological activity, cognitive processes, and affective experiences; and by suggesting a fourth content area relating to affect and mood. We have indicated the implications of these clarifications and changes for (1) the design of research methodology, particularly the selection and implementation of assessment procedures in treatment outcome studies; (2) the use of a more appropriate and specific matching of treatment techniques to the particular response profiles of individual clients; and (3) the development of an improved understanding of response interrelationships in key emotional disorders such as depression and anxiety. Specifically, we have suggested that the failure to find agreement between measures of physiological, cognitive, behavioural, and affective changes in some studies may be as much a reflection of the lack of agreement arising from spurious sources of variance within content areas or response systems as it is a reflection of the operation of different processes and systems. Changes need to be made in the assessment methodology so that true desynchrony can be discerned from methodological desynchrony. We hope the use of the type of content x method matrix suggested herein will improve the clarity of conclusions reached about the synchrony/desynchrony issue as it relates to the understanding of various psychological problems and their treatment. The benefits of this approach should become particularly evident when combined with treatment-client matching research strategies.

Acknoi&dgemenf-Parts of this article were prepared while the authors were on a research visit at the University of London and the Center for Stress and Anxiety Disorders, SUNY Albany. The support by a travel grant from James Cook University and the Special Studies Program of the University of Sydney is gratefully acknowledged.

REFERENCES

Abelson. J. L. & Curtis, G. C. (1989). Cardiac and neuroendocrine responses to exposure therapy in height phobics: Desynchrony within the ‘physiological response system’. Behaoiour Research and Therapy, 27, 561-567.

Barlow, D. H. (1988). Anxiety and its disorders. New York: Guilford Press. Beck, A. T.. Ward, C. H.. Mendelson, M.. Mock, J. & Erbaugh, .I. (1961). An inventory for measuring depression. Archires

qf‘ General Psychiatry. 4, 56 IL57 1. Birrell, P. C. (1983). Behavioral, subjective, and electroencephalographic idices of sleep onset latency and sleep duration.

Journal of Beharioral Assessment, 5, 1799190. Chambless, D. L.. Caputo, G. C., Bright, P. & Gallagher, R. (1984). Assessment of fear in agoraphobics: The Body

Sensations Questionnaire and the Agoraphobic Cognitions Questionnaire. Journal of Consulting and C/inical Psychology, 52, 1090-1097.

Chambless. D. L., Caputo, G. C., Jasin. S. E. Gracely, E. J. &Williams, C. (1985). The mobility inventory for agoraphobia. Behariour Research and Therapy, 23, 3544.

Cone, J. D. ( 1979). Confounded comparisons in triple response mode assessment research. Behavioral Assessmenr, I, 85-95. Derry. P. & Kuiper. N. (1981). Schematic processing and self-reference in depression. Journal of Abnormal Psychology. 90,

286-297. Eifert. G. H. (1990). The acquisition and treatment of phobic anxiety: A paradigmatic behavioral perspective. In Eifert,

G. H. & Evans. 1. M. (Eds), Un$ving behacior therapy: Contributions of paradigmatic behaviorism (pp. 1733200). New York: Springer.

Eifert. G. H. & Craill. L. (1989). The relationship of affect, behaviour, and cognition in behavioural and cognitive treat- ments of depression and phobic anxiety. Beharriour Change, 6, 96-103.

Eifert. G. H.. Evans, I. M. & McKendrick. V. (1990). Matching treatments to client problems not diagnostic labels: A case for paradigmatic behavior therapy. Journal of Behavior Therapy and Experimental Psychiatry. 21, 163-172.

Eifert. G. H.. Craill, L.. Carey, E. & O’Connor, C. (1988). Affect modification through evaluative conditioning with music. Behariour Research and Therapy, 26, 321-330.

Ekman. P.. Levenson. R. W. & Friesen, W. V. (1983). Autonomic nervous systemactivity distinguishes among emotions. Science. 22f. 1208-1310.

Evans, 1. M. (1986). Response structure and the triple-response-mode concept, In Nelson, R. 0. & Hayes, S. C. (Eds), Conceprual,fbundarions of behacioural assessment (pp. 131-155). New York: Guilford Press.

Heiby. E. M. (1986). Social versus self-control deficits in four cases of depression. Behavior Therap),. /7, 158-163. Heiby. E. M. & Staats. A. W. (1990). Depression: Classification, explanation, and treatment. In Eifert. G. H. & Evans 1. M.

(Eds). Unif.iing behavior therapy: Contributions of paradigmatic behaciorsm (pp. 220-246). New York: Springer. Hodes. R. L.. Cook, E. W. & Lang. P. .I. (1985). Individual differences in autonomic reponse: Conditioned association or

conditioned fear? Pqvhophysiology, 22, 545-560. Hodgson. R. & Rachman. S. (1974). II. Desynchrony in measures of fear. Behariour Research and Therapy. 12, 319-326.

292 GEORG H. EIFERT and PETER H. WILSON

Hollon, S. D. & Kendall. P. C. (1980). Cognitive self-statements in depression: Development of an Automatic Thoughts Questionnaire. Cognitive Therapy and Research, 4, 383-395.

Hugdahl, K. (1981). The Three-Svstems-Model of fear and emotion-a critical examination. Behariour Research and -Therapy, 19. 15-85.

Lang, P. J. (1968). Fear reduction and fear behavior: problems in treating a construct. In Schlien J. M. (Ed.), Research in psvchafherapv (Vol. III). Washington, D.C.: APA.

Lang.-PI J. & Cuthbert, M. .I. (1984). Affective processing in anxiety. Journal 01 Beha~,~ora~ Arsessmenr. 6, 370-379. Lazarus, R. S. (1984). On the primacy of cognition. American Psychoiogisf, 39, 124-129. Mackay, W. & Liddell, A. (1986). An investigation into the matching of specific agoraphobic anxiety response characteristics

with specific types of treatment. Behariour Research and Therapy, 24, 361-364. McKnight, D. L., Nelson, R. O., Hayes, S. C. & Jarrett, R. B. (1984). Importance of assessing individually assessed response

classes in the amelioration of depression. Behavior Therapy, 15, 315-335. Mavissakal~an, M. (1987). Tr~modai assessment in agoraphobia: Further observations on heart rate and synchron}/!

desynchrony. Journal of Psychoparholog~ and Behavioural Assessmenf, 9, 88-98. Mersch, P. P. A., Emmelkamp, P. M. G.. Biigels, S. M. & van der Sleen, J. (1989). Social phobia: Individual response

patterns and the effects of behavioral and cognitive interventions, Behaciour Research and Therupy. 27. 421434. Michelson. L. (1984). The role of individual differences, response profiles, and treatment consonance in anxiety.

Journal qf Behavioral Assessmenf. 6, 349-361. Miller, B. V. & Bernstein, D. A. (I972). Instructiu~al demand in a behavioral avoidance test for claustrophobic fear.

Journal OS Abnormal Ps_~cho~ogy, 80. 206-210. 6st, L. G., Jerremalm, A. & Johansson, J. (1981). Individual response patterns and the effects of different behavioral

methods in the treatments of social phobia. Behaviour Research and Therapy, 19, 1-16. &t, L. G., Jerremalm, A. & Jansson, L. (1984). Individual response patterns and the effects of different behavioral methods

in the treatment of agoraphobia. Behaviour Research and Therapy, 22, 697-707. Peterson, C., Semmel, A., von Baeyer, C., Abramson, L., Metalsky, G. & Seligman, M. (198.2). The attributional style

questionnaire. Cognifitje Therap:~, and Research. 6, 287-300. Rachman, S. (1981). The nrimacv of affect: Some theoretical imnlications. Behaviour Research and TheraD,. 19. 279-290. Rachman, S. (1984). A reassessment of the ‘primacy of affect’.‘C’ognirive Therapy and Research, 8. 579r584. Rachman. S. & Hodgson, R. (1974). I. Synchrony and desynchrony in fear and avoidance. Behariour Research and Therapy.

1.2. 311-318. Salkovskis, P. M. & Clark, D. M. (1990). Affective responses to hyperventilation A test of the cognitive model of panic.

Behaviour Research and Therapy, 28. 5161. Sher, K. J.. Frost, R. O., Kushner, M., Crews, T. M. & Alexander, J. E. (1989). Memory deficits in compulsive checkers:

Replication and extension in a clinical sample. Behaviour Research and Therapy, 27, 65-69. Taylor, C. B.. Fried, L. & Kennardy. J. (1990). The use of a real-time computer diary for data acquisition and processing.

Behaciour Research and Therapy, 28, 93-97. Teasdale. J. D. (1983). Negative thinking in depression: Cause, effect or reciprocal relationship? Adt~unces in Behuriour

Research and Therapy, 5, 3-26. Warwick. H. M. C. & Salkovskis, P. M. (1990). Hypochondriasis. Behaviour Research and Therapy, 28, lOSSI 17. Watts. F. N. (1989). Attentional strategies and agoraphobic anxiety. Behaaioural Psychotherapy, 17, 15-26. Watts, F. N.. McKenna, F. P.. Sharrock, R. & Trezise. L. (1986). Colour naming of phobia-related words. Brifish Journal

of Psychology. 7% 97-108. Weissman. A. N. & Beck, A. T. (1978). Development and validation of the Dysfunctional Attitude Scale: A preliminary

investigation. Paper presented at the American Educational Research Association, Toronto, March, 1978. Williams. M., Watts, F. N.. MacLeod, C. & Mathews, A. (1988). Cognitive psyehoiog~ and emofjona~ disorders. Chichester:

Wiley. Wolpe, J. (1989). The derailment of behavior therapy: A tale of conceptual misdirection. Journal of Behavior Therapy and

Esperimental Psychiatry. 20, 3-l 5. Zajonc, R. (1980). Feeling and thinking: Preferences need no inferences. American Psychologixf. 35. 151-175. Zajonc, R. (1984). On the primacy of affect. American Psyeho~og~f, 39, 117~123.