Psychodiagnosis Based on Verbal Data

In psychological testing, the majority of psychodiagnostic approaches are strictly psychometric. These approaches have been criticized because of their lack of theoretical grounding in cognitive science and their limited use for decision making. Nichols, Chipman, and Brennan (1995) introduced a cognitive science approach that provides a theoretical basis for the development of methods for cognitively diagnostic assessment. The approach has drawn attention to the relevance of individual knowledge representations and to the assessment of qualitative verbal data as a basis for diagnosing declarative knowledge of an individual about aspects of his or her mental state, for example, structures of knowledge that refer to a particular subject matter to be learned, aspects of individual mental disorders, or personality features. According to this approach the focus is on what rather than how much a person knows. The present paper proposes both a theoretical framework and a methodology for assessing qualitative aspects of knowledge based on verbal data. It is applicable to psychodiagnosis in general but we focus specifically on educational and clinical diagnosis because of strong commonalities and correspondences between both approaches. As will be shown, educational and clinical diagnosis can be integrated into a general theoretical framework of cognitive psychodiagnosis based on verbal data. The benefits that result from this integrated theoretical framework and methodology for the diagnostic process and for decisions regarding instructional and therapeutic interventions are discussed. Psychodiagnosis Based on Verbal Data 3 A Cognitive Approach to Psychodiagnosis Based on Verbal Data Introduction The psychometric approach of psychological testing has been prevalent in many domains of psychodiagnosis (e.g., Aiken, 1997). A strictly psychometric approach has been criticized because of its lack of theoretical grounding in cognitive science and its limited use for practical decision making (Nichols, Chipman, & Brennan, 1995). For the field of education Nichols et al. (1995), for example, pointed out that “traditional tests are well behaved for ranking and comparing examinees, for grading, and for predicting who will do well in some future activity. Typically, they do not provide useful diagnostic information about specific content that should be studied or taught in order to improve performance” (p. 1). The traditional psychometric approach also falls short because “many want testing to be an integral part of instructional activity, leading to qualitative data to be used for learner modelling and adaptive teaching and helping to guide teachers and students to the eventual attainment of substantive educational goals” (p. 1). In the diagnosis of mental disorders, traditionally either psychometric scales (i.e., symptom scales) are used, or clinicians use free exploratory interviews for deriving diagnoses. Due to the high degree of standardization, scores on symptom scales tend to be reliable. Symptom scales, however, do not allow for deriving clinical diagnoses of mental disorders according to the major diagnostic systems of mental disorders (e.g., DSM or ICD). They do not yield sufficiently fine-grained, qualitative clinical diagnostic data. In contrast, free exploratory interviews provide rich clinical information. However, the reliability of diagnoses based on exploratory interviews is low. What is lacking for a psychodiagnosis based on verbal data is a new rationale and methodology strongly oriented toward modern cognitive theory. In recent research, the cognitive science approach has been suggested for providing a theoretical basis for the development of Psychodiagnosis Based on Verbal Data 4 methods in cognitively diagnostic assessment (Nichols, Chipman, & Brennan, 1995). This approach has drawn attention to the relevance of individual knowledge representations and to the assessment of qualitative verbal data as a basis for diagnosing declarative knowledge about aspects of an individual’s mental state. The term “mental state” is used here as a paraphrase for mentally represented subject-matter contents, such as knowledge about cognitive competence, emotional states, past behaviour, personal experience, certain personality traits, or aspects of health. A shift in rationale and methodology becomes particularly apparent in the field of learning and instruction. Here, the focus of interest has shifted away from both the assessment and analysis of quantitative aspects of an individual’s performance to qualitative aspects of individual knowledge (Nichols, Chipman, & Brennan, 1995; Tergan, 1986, 1988, 1989 a, b), and from statistically based assessment toward theory-based assessment (Frederiksen, Mislevy, & Bejar, 1993; Lohman & Ippel, 1993; Mislevy, 1993; Snow & Lohmann, 1989; Sternberg, 1991; Wittrock & Baker, 1991). Similarly, in recent decades clinical diagnosis has shifted from exploratory interviews to standardized diagnostic interview techniques. Modern clinical approaches use extensive, standardized interviewing techniques in an attempt to collect verbal data in a systematic and reliable manner. The correspondence between the educational and the clinical approach is that both are grounded on the assumption that a person has privileged access to knowledge of his or her mental state. Furthermore, this approach further is grounded on the implicit assumption that a person is able to verbalize this knowledge, provided that he or she is being asked the appropriate questions (cf. Robins, 1989). In sum, educational as well as clinical diagnosis appears to shift towards a cognitive approach that makes systematic use of individual language production (verbal data). From a cognitive perspective, it is assumed that the study of verbal data assists the diagnostician in Psychodiagnosis Based on Verbal Data 5 accessing an individual’s representation of conceptual knowledge on aspects of his or her mental state. In this paper we will show how a language-based cognitive approach may apply to psychodiagnosis. The benefits that may result for the diagnostic process and for instructional and therapeutic interventions in educational as well as clinical settings will be discussed. Firstly, the conceptual rationale of a language-based cognitive approach to psychodiagnosis is described. Secondly, an emphasis is placed on the role of verbal data in educational and clinical psychodiagnosis. Thirdly, the theoretical foundation of a methodology of psychodiagnosis based on verbal data is outlined and discussed with respect to its suitability for both educational diagnosis and clinical diagnosis. Finally, the implications of a cognitively oriented psychodiagnostic approach are discussed especially with regard to the theoretical foundations and benefits of its application to psychodiagnostic praxis. Theoretical Foundation Assumptions Central to a cognitive approach to psychodiagnosis based on verbal data are theories of language production on the one hand and models of representation of conceptual knowledge on the other (Bühler, 1932; Engelkamp, 1983; Kintsch, 1974; Levelt, 1989; Paivio, 1986). It has been suggested that language production is the end result of a series of mental processes based on representations at different levels of the cognitive system. These include the activation of a particular conceptual network, the choice of words, the decision for a syntactic structure, and its realization in the form of oral or written language (Herrmann & Grabowski, 1994; Levelt, 1989). The processes and representations are influenced by the interaction of internal states of the communicator (e.g., cognition, interests, emotions) and relevant conditions of a communication situation (Ericsson & Simon, 1990). In language production mental structures and processes are “frozen”. They may be conceived as externalisation of internal states Psychodiagnosis Based on Verbal Data 6 and processes. In the organon model of signs (Bühler, 1932, p. 28) the “Sprachzeichen” indicates internal states and processes of the language producer. A psychodiagnosis based on language productions is based on the assumption that there are indicators of mental states in the language products of an individual, and that the indicators may be used for diagnosis. Imagine, for example, that individuals are asked to answer questions about their current emotional state. This task will lead them to activate appropriate knowledge stored in long-term memory, and construct a mental representation in working memory. Constructing a representation of declarative knowledge often requires a change of existing representations stored in long-term memory. These can, for example, be recodings from a non-verbal (pictorial) to a verbal format in cases where people have double coded information (see Paivio, 1986) or from a procedural format into a propositional format (see Anderson, 1983, 1993). For instance, in order to verbalize knowledge about their mental well-being., individuals may have to recode knowledge about emotions which may be represented episodically, into a propositional format in order to verbalize it (Engelkamp, 1983; see also Tesser & Martin, 1996). (1) Based on general assumptions concerning the human mind (e.g. Engelkamp & Zimmer, 1994), mental representation, and speech production Eckert (1998), Nichols (1994), and Tergan (1988, 1989 a,b) suggested a conceptual rationale for diagnosing knowledge structures. Similarly, Kluwe (1988) had conceptualized assumptions involved in assessing knowledge and illustrated these in a model. For the present article we adapted this model and applied it to the domain of psychodiagnosis based on verbal data. The adapted model is shown in Figure 1. The rationale of this adapted model is based on the four general assumptionsIndividuals, when induced to verbalize aspects of their individual mental state activate knowledge stored in long-term memory. Psychodiagnosis Based on Verbal Data 7 (2) They construct a propositional knowledge representation about this state, which is stored in working memory. (3) They can externalise this knowledge by using language when asked to do so. (4) Language products can be analyzed in such a way that the diagnostician may infer theindividual’s mental representation of knowledge concerning aspects of his or her mental state. Insert Figure 1 about here Verbal data in educational diagnosis In the domain of education the diagnostician may be interested in the cognitive processes and structural aspects of individual declarative knowledge as determinants for future learning (Dochy, 1992) and performance in new tasks (Spada & Wiechmann, 1998). Or he or she may be interested in processes of knowledge acquisition (Rumelhart & Norman, 1978). A special focus in research on knowledge acquisition is on aspects of knowledge change, for example conceptual change, as a result of cognitive development or pedagogical intervention (Ewert & Thomas, 1996; Glynn & Duit, 1995; Jacobson & Archodidou, 2000; Mandl & Ballstaedt, 1986; Rumelhart & Norman, 1978; Vosniadou, 1994; Vosniadou & Brewer, 1992). Another central focus of psychodiagnosis based on verbal data is the study of expert-novice differences in domain knowledge and problem solving strategies. One goal is to design instructional measures that may overcome shortcomings in knowledge representations of novices and foster the construction of knowledge representations similar to experts. Prominent historical examples in the educational domain that may highlight both the way in which verbal data may be assessed and used in educational psychodiagnosis and the implicit rationale and the procedure of psychodiagnosis are the studies of Larkin (1979, 1983), Larkin, McDermott, D. P. Simon, & H. A. Simon. (1980), and Chi et al. (1981). In specific, Larkin (1979, 1983) and Larkin et al. (1980) have applied the think aloud method (see below) in their Psychodiagnosis Based on Verbal Data 8 research on differences in problem solving between novices and experts. The authors’ research aimed at contributing to the development of a theory of physical problem solving. In their analysis of the construct “problem solving competence” they started from the general assumption that explaining differences in problem solving cannot solely be done on the basis of cognitive processes but, additionally, has to take into account differences in the mental representation of the problem situation as constructed by individuals. The focus therefore was the analysis of the differences in features of mental representations which novices and experts construct in making sense of a problem situation before they start solving a problem. The subject-matter domain under consideration was elementary mechanics. In analyzing the task requirements for novices and experts of typical problems in elementary mechanics, the general assumption was that mental representations of novice learners differ from the representations of experts with respect to the abstract character of entities they are based on. In order to assess data the authors chose the think aloud method. The subjects verbal statements were tape recorded and transcribed. The implicit assumption was that this method is appropriate for inducing subjects to utter thought processes and that an analysis of verbal protocols would facilitate the identification of indicators representing conceptual and structural aspects of mental representations. Protocol analysis followed the procedure as described by Newell and Simon (1972). The data analyzed were interpreted in the light of the theoretical assumption concerning mental representations. In reconstructing the mental representation behind the verbal statements, Larkin and colleagues found that naive problem solvers constructed “naive” mental representations based on well known physical objects and entities. There was hardly any relation to formal physical entities. Expert problem solvers, on the other hand, constructed additional problem representations based on abstract physical schemata (e.g., forces schema or work-energy-schema). Larkin (1983) used these results for refining her theory of Psychodiagnosis Based on Verbal Data 9 problem solving based on mental representations. This approach differs from earlier approaches which focused on problem solving processes rather than on knowledge structures (e.g., Newell & Simon, 1972). In several follow-up studies Chi et al. (1981) and Chi, Glaser, and Rees (1982) modified the procedure of analyzing protocol data. “We focussed specifically on those qualitative analysis statements that seemed to generate knowledge not explicitly stated in the problem (e.g., inferences)” (Chi et al., 1982, p. 35f). The authors found that problem representations of experts are based on context independent physical principles (e.g., “preservation of energy”, Newton’s 2 law), whereas the representations of novices are based on familiar and situation specific concrete objects of the problem situation. The authors concluded from their results that not only experts but also novices try to construct an adequate physical representation of a given problem. In sum, this example from educational diagnosis shows how qualitiative psychodiagnosis based on verbal data can be used to examine cognitive processes and structural aspects of individual declarative knowledge as determinants for future learning. In more applied educational settings, qualitative psychodiagnosis based on verbal data can be used to guide educational and instructional decisions for individuals. Psychometric methods for generating, assessing and analyzing verbal data are used for decisions of selection as well as for modification and individualizing learning. For example, tests for assessing verbal intelligence may be used for selection decisions with regard to school types or professions. Decisions that involve modification measures are often based on verbal performance data when instructors wish to divide their students into different groups according to their level of conceptual preknowledge. For decisions that concern the individualization of instruction, psychometric methods are no longer considered appropriate (Anderson, 1993; Ohlsson, 1993). Instead of quantitative performance data, qualitative data that involve the structure of individual conceptual knowledge Psychodiagnosis Based on Verbal Data 10 are more appropriate for use in adaptive teaching and for helping to guide teachers and students to the eventual attainment of substantive educational goals (Chipman, Nichols, & Brennan, 1995; Leutner, 1997). Verbal data in the Clinical Domain Verbal data are used in clinical diagnosis for a variety of purposes. They include the classification of mental disorders based on clinical symptoms, the assignment of therapeutic interventions for patients, andthe prediction of the onset and course of mental disorders (cf. Robins, 1988). For example, mental disorders are characterized by a variety of symptoms at different levels, for example the behavioural, affective, physiological, or social level (Farmer & McGuffin, 1989; Seidenstücker & Baumann, 1978). In order to obtain a complete picture of a mental disorder, data need to be collected from alllevels (Bastine & Tuschen, 1996). Clinical diagnosticians try to select and apply appropriate methods for identifying the disorder which most likely underlies the clinical picture the patient presents. The systematic gathering and analysis of verbal data is currently the most important method used in clinical diagnosis. In contrast to most physical diseases, symptoms of mental disorders cannot yet be reliably identified and differentiated from one another on the basis of laboratory tests (biological markers), or behaviour observation. Rather, many symptoms of mental disorders can only be perceived by the patient him or herself. Hence, diagnostic decisions about mental disorders are usually based on verbal reports of the patients. Examples of methods based on verbal data include qualitative as well as quantitative approaches. Qualitative approaches includeprojective tests (e.g., Rorschach, TAT, sentence completion tests), free exploration or unstructured interviews and content-analysis of language productions (e.g., Gottschalk-Gleser content analysis). Diagnoses derived from projective tests have traditionally been found to be less reliable, and their validity unproven (see Chapman & Psychodiagnosis Based on Verbal Data 11 Chapman, 1969). Recent attempts to develop higher standardized protocols for projective tests promise improved diagnostic reliability (e.g., Exner, 1986 a,b). Using these improved protocols, projective tests such as the Rorschach test, may be successful in differentiating disorders from each other and from non-pathological states. This has, for example, recently been demonstrated by Sanders and colleagues (Sanders, Adams, Tager-Flusberg, Whenton, et al., 1995) who compared language production of patients with schizophrenia or manic depression with those of normal controls. Using a battery of linguistic measures for analyzing the Rorschach protocols, theyfound meaningful differences in the speech of schizophrenics as compared with that of the other groups. Schizophrenics’ language production, for example, had fewer clearly referenced elements, seen as a measure of cohesion. Improved techniques for content-analysis of language productions have also been suggested (e.g., Gottschalk, 1995), and the protocols have become scorable by computerized technology (Gottschalk, 1995). A recent example of clinical diagnosis oriented toward assessing declarative knowledge has been presented by Wenzel and Holt (2000). The authors applied a script methodology from cognitive psychology to investigate schema content for thread in two specific phobias. In this study, 17 anxious subjects listed prototypical sequences of events, or scripts, that most people experience in threatening situations. The scripts were then analyzed and compared with the scripts of 30 sub jects who were not anxious. Quantitative approaches in clinical diagnosis using verbal data include psychometric methods (tests, self-administered symptom scales), and structured as well as standardized interview techniques. Examples of the latter are the Diagnostic Interview Schedule (DIS, Robins, 1988) or the Composite International Diagnostic Interview (CIDI; WHO 1990). These interviews are grounded in complex classification systems such as the Diagnostic and Statistical Manual of Diseases (DSM) or the International Classification of Diseases (ICD). In these systems the diagnostic criteria are operationally defined (Robins, 1989; Westmeyer, 1990), and the Psychodiagnosis Based on Verbal Data 12 categorization of data into classes of disorders is regulated according to well defined rules. The rules, however, are often insufficiently grounded in a theory of measurement (cf., Westmeyer, 1990). Rather, it is (in most cases implicitly) assumed that a patient has declarative knowledge about the symptoms and their relations, and can access this knowledge during the interview. Standardized diagnostic interviews attempt to “translate” the diagnostic criteria as directly as possible into questions (Robins, 1989). The questions are fully worded, and have to be read to the respondent exactly as written, and in the order in which they were provided in the interview schedule. As has been demonstrated in numerous studies, this has increased the reliability of diagnoses, rendering them much more reliable than exploratory or semi-structured diagnostic interviews (Andreasen et al., 1981; Keller et al., 1981; Semler et al., 1987; Wittchen, 1994). Presently, common practice in clinical settings is still very different from the common practice in research settings (cf., Shaffer, Fisher, & Lucas, 1999). In research settings, for example epidemiology, standardized diagnostic interviews are now predominantly used due to their proven high reliability, even though the validity of diagnoses derived using these methods is still a matter of debate (for recent discussions, see Brugha, Bebbington, & Jenkins, 1999; Wittchen, Üstün, & Kessler, 1999). In contrast, clinicians still prefer free, exploratory interviews, most likely because of the higher flexibility and input of clinical experience that they grant (Grove & Meehl, 1996). Using free interviews still is widespread , even though numerous biases have been identified for clinical judgements based on free exploration (Dawes, Faust, & Meehl, 1989; Meehl, 1954). Applying the methodology for psychodiagnosis based on verbal data suggested in the present paper could substantially improve the reliability and validity of clinical diagnosis. Psychodiagnosis Based on Verbal Data 13 A Methodology of Psychodiagnosis Based on Verbal data The following presentation of the methodological foundations of a cognitive and clinical diagnosis is based on a conceptual model of qualitative diagnosis as proposed in the field of knowledge diagnosis (Nichols, 1994; Tergan, 1988; 1989a,b). It has been further elaborated and implemented into a computer-based methodology by Eckert (1998). The methodology presented well matches the stepwise approach which is implicitly used in empirical psychodiagnostic research (e.g. Larkin, 1979, 1983; Weber, 1994; Wenzel & Holt, 2000). The following closely related steps of the diagnostic processes may be distinguished: (1) Psychological construct analysis (2) Analysis and design of the diagnostic setting (3) Assessment of verbal data (4) Data analysis and interpretation (5) Evaluation of results (6) Application of results Although these steps are key elements of a systematic psychodiagnostic approach they may not be exhaustively carried out in the diagnostic praxis because of pragmatic considerations. They may often, however, implicitly guide a diagnostician’s actions. (1) Psychological Construct Analysis Psychological construct analysis in psychodiagnosis is concerned with individual representations of knowledge in long-term memory. During construct analysis it is the diagnostician’s task to analyze and describe in detail the representation of a mental state, for example, of a concept or the features of clinical syndrome (e.g., “anxiety”). This is performed on the basis of a theory concerning the particular mental construct that he or she has in mind or is considering. In knowledge diagnosis, the focus of mental construct analysis is on cognitive task Psychodiagnosis Based on Verbal Data 14 analysis (Greeno, 1980). The attention focuses on general features of knowledge. This knowledge has to be activated or constructed by an individual under the constraints of a particular task situation in a particular diagnostic setting. Taking a child’s knowledge of dinosaurs as an example, a diagnosis of this knowledge would involve studying its conceptual format and the network structure by analyzing the child’s verbal statements about dinosaurs (Chi & Koeske, 1983). Another example would be the comparison of maths experts with novices with regard to their categorization and representation of mathematical problems (Chi, Feltovich, & Glaser, 1981). The focus of construct analysis in psychodiagnosis is on the construction of a conceptual model of the knowledge assumed relevant for both the individual representation of knowledge of a particular mental state and the cognitive requirements necessary for coping with a particular diagnostic task. The construction of the conceptual model in language-based psychodiagnosis will be based on a psychological framework theory for the representation of concepts, for example a theory or model of semantic networks (e.g., Kintsch, 1974). The conceptual model serves as a framework for those aspects of individual knowledge about a mental state that best represents diagnostically relevant aspects in the speech products of the individual (see Figure 2). Insert Figure 2 about here In the context of learning and instruction, the conceptual model must be grounded on an appropriate criterion. For example, it may be oriented toward the knowledge structure of an expert or a layperson (e.g., pupil or student). If the goal is to assess the pre-knowledge structure of a student it may be worthwhile to construct a conceptual model oriented toward the knowledge structure of a prototypical “naive” student. When the knowledge structure of a student after instruction is to be assessed, a conceptual model oriented towards a prototypical advanced student or an expert would probably be more appropriate. Psychodiagnosis Based on Verbal Data 15 The described rationale of a cognitive approach to knowledge diagnosis has been applied, for example, to developing the diagnostic components of computer-based Intelligent Tutoring Systems. Intelligent Tutoring Systems seek to automatically model the learner’s knowledge in order to individualize instruction. In an “overlay-approach”, for example, the individual learning progress is compared (modelled) to the knowledge of an expert approach (Sleeman & Brown, 1982; Wenger, 1987). Where discrepancies arise, instructional measures are applied to help students overcome learning problems and to acquire the expert knowledge. As a theoretical framework, theories of mind (Engelkamp & Zimmer, 1994), models on the mental representation of knowledge (Rumelhart & Norman, 1985) and of the process and structure of language production (Levelt, 1989) may be helpful. Ideally in the case of language-based psychodiagnosis these framework-theories should be integrated into an overall theory of language production. In language-based clinical diagnosis the focus of construct analysis is on the mental disorder(s) under consideration, and what declarative (propositional) knowledge a person may have mentally represented about the disorder. Part of the construct analysis then is to define the possible verbal indicators of the construct in individual speech productions, for example, in answers to questions in a standardized diagnostic interview. As far as the theoretical background is concerned, a cognitive approach to clinical diagnosis based on verbal data may be grounded in the same theoretical framework as the cognitive diagnosis. However, to date clinical approaches have rarely taken cognitive theories into account as a basis for construct analysis (Knäuper, 1999). Crucial for the foundation of a cognitive psychodiagnostic methodology based on verbal data are theoretical assumptions concerning changes in mental representations. As mentioned in the foregoing, knowledge may be represented in different formats in the human brain (Rumelhart & Norman, 1985; Strube & Schlieder, 1998). Humans, in order to verbalize or write down what Psychodiagnosis Based on Verbal Data 16 they know or what they feel, may have to change an existing non-propositional format of mental representation into a propositional one. Researchers such as Winograd (1975) or Engelkamp and Zimmer (1994) have suggested that representational change is a characteristic competence of the human cognitive system. According to Winograd (1975), it may be assumed that changing a procedural representation into a declarative one is possible. Nevertheless, based on everyday experience, there is considerable doubt as to whether emotional states and symptoms of mental disorders can be adequately represented propositionally and expressed verbally, e.g. in terms of propositions. In most cases the connotation of meaning associated with concepts will be less rich than the individual perception of the mental states to which they refer (Engelkamp, 1983; Engelkamp & Zimmer, 1994; Herrmann, 1985). (2) Analysis and Design of the Diagnostic Setting The second step is closely related to psychological construct analysis. The focus is here on analyzing and designing appropriate task requirements. The goal is to create a diagnostic task situation in an overall diagnostic setting that induces individuals to (a) focus their attention on exactly those aspects of the psychological construct (i.e. the mental state) that the diagnostician is interested in (e.g., a specific knowledge structure or a symptom typical for a specific mental disorder), and (b) to construct a mental representation of knowledge on the individual mental state. In psychodiagnosis based on verbal data the second goal is to create the task setting such that the mental representation represents the individual’s declarative knowledge which may easily be expressed in verbal mode. For example, a questionnaire must be carefully designed in such a way that the individual will construct a valid mental representation in working memory that represents declarative knowledge relevant to the diagnosis. Usually, this can only be achieved by pre-testing the questionnaire or task setting. The diagnostician in this phase of the planning process also has to make sure that the induced language products really provide verbal indicators Psychodiagnosis Based on Verbal Data 17 of the hypothesized psychological construct (Herrmann, 1985). In other cases, relevant verbal data already exist. These can be written documents such as letters or diaries or verbal documents such as records of conversations (Ballstaedt, 1982, 1987). The disadvantage is that the conditions and intentions of the individuals at the time of language production are not well known. This is why in most cases diagnosticians prefer designing the diagnostic setting by themselves. In educational diagnosis, often the focus is only on specific aspects of knowledge, for instance, the individual structure of a concept (Mandl & Ballstaedt, 1986). In this case, subjects are instructed to freely associate concepts which they feel are related to a given concept or to draw a mind map and label the relations between the mapped concepts. Psychological constructs relevant in clinical diagnosis are often very complex. Therefore, the diagnostic process involves several steps in which symptoms and their course are assessed. An appropriate diagnostic task situation (i.e. a questionnaire or interview) must be created that allows for data assessment which covers a broad range of behavior, cognition, emotion, and perceived symptoms, their severity and duration, and the contexts and conditions in which the symptoms are present or disappear. Often, a combination of different diagnostic methods may be appropriate: The patient is asked to describe in his or her own words typical conditions of appearance and features of the symptoms, fill out a standardized questionnaire and answer questions put to them by different medical specialists in a clinical interview. Based on the results of construct analysis, the diagnostician must then evaluate the relevance of the various verbal utterances and structural features in the individual’s language products. Decisions have to be made as to which products constitute indicators for the cognitive or clinical construct in question. However, in many cases there is a lack of well established criteria as well as rules for determining what qualifies an indicator for the respective construct and what its relation to the construct is. It is, therefore, not surprising that a decision about Psychodiagnosis Based on Verbal Data 18 relevant indicators is made mostly implicitly in applied settings and is based on the diagnosticians expertise or intuition. However, particularly in clinical diagnosis, in recent years there has been a strong trend for converting implicit decision rules into explicit descriptions of qualifying indicators and their relation to the construct. The development of standardized diagnostic interviews is part of this process, and helps to decrease the impact of clinical intuition on diagnostic judgement. The analysis and design of the diagnostic setting includes reflections of the diagnostician with regard to the appropriate method for data assessment (see Jäger, 1983). One criterion for choosing a method is the modality of language production. Generally, language productions can be in a verbal or written form. It is usually acknowledged that written text (e.g. letters, scripts) has a higher level of cognitive reflection and a higher level of obligation on the side of the language producer when compared to verbal productions. As written text produc tions are more differentiated and can be collected more easily than verbal text productions, diagnosticians rely mostly on them (Grabowski-Gellert & Winterhoff-Spurk, 1989). Sometimes such productions are the only data to which the diagnostician has access. This is, for example, the case in historical and biographical psychology. Written language productions of deceased individuals are analyzed in order to diagnose cognitive or clinical features of the person’s personality or infer mental disorders he or she may have had. When verbal language products are used in psychodiagnosis, they must be transcribed before being analyzed. What is problematic about transcriptions is that prosodic information gets lost unless the diagnostician represents them by using an adequate symbol system (Herrmann & Grabowski, 1994). For deciding on the adequacy of a psychodiagnostic approach for the assessment of verbal data, the diagnostician has to analyze the degree and in which way a particular method is “sensitive” to the assessment of data concerning the psychological construct of interest. For Psychodiagnosis Based on Verbal Data 19 example, when declarative knowledge is the focus of diagnosis, those methodological approaches seem to be most appropriate which not only allow the assessment of relations between the concepts of a particular domain but also the representation of knowledge by means of network representations or schema structures (Kintsch, 1974; Scheele & Groeben, 1984; Tergan, 1986). Methods for the assessment of verbal data may be used for assessing causal background knowledge (e.g., Vosniadou & Brewer, 1992). Drawings in combination with verbal data may be used for assessing those aspects of mental models which correspond to their inherent structural features. When procedural knowledge is the focus, verbal data may be gathered by using the think-aloud method (Claparède, 1932; Duncker, 1935; Nisbett & Wilson, 1977). The resulting verbal protocols may be analyzed with respect to inherent cognitive processes and can be represented as condition-action units (productions). The latter has been suggested by artificial intelligence researchers (e.g. Anderson, 1983; Newell & Simon, 1972). Existing methods for the assessment of verbal data for a qualitative cognitive diagnosis have been put together and critically reviewed, for example, by Eckert (1998), Kluwe (1988), Nichols (1994), and Tergan (1986). A review of clinical methods for assessing and analyzing verbal data has been presented by Wittchen and Lachner (1996). These methods, for example, the Rorschach test, TAT-technique, psychometric scales, or standardized diagnostic interviews, belong to the standard repertoire of clinical and psychiatric diagnosis, or did so in the past. They differ largely with respect to the degree of structuring and standardization. As mentioned earlier, one problem with clinical diagnosis is the complexity of symptoms and the complexity of the representation formats involved (emotional, behavioral, cognitive, and social). Another problem is that patients may not be able to successfully “translate” a non-propositional knowledge representation into a propositional format for purposes of verbalization, as mentioned earlier. Therefore, Psychodiagnosis Based on Verbal Data 20 diagnosticians need to assist individuals in their attempt to verbalize their symptoms. They can do so by asking systematic questions that try to describe relevant symptoms as clearly and “vividly” as possible. (3) Assessment of Verbal Data The assessment of verbal data is based on the aforementioned assumption that individuals can externalise knowledge about a particular mental state of their own by using language when they are asked to do so. During the assessment phase, decisions made about the design of the diagnostic task situation are applied to a concrete diagnostic setting. The diagnostician must fix the concrete wording of the instruction with reference to the application of a chosen diagnostic method, the details for registering or recording the language productions suggested as relevant for diagnosing the psychological construct. The methods differ in terms of the nature of the responses expected from the subjects. One type of method prescribe that subjects respond to pre-verbalized responses, for example in a multiple choice questionnaire. These are responses which the diagnostician has developed on the basis of his or her knowledge of the characteristics of the psychological construct at the focus of the diagnosis. Other methods induce subjects to verbalize knowledge regarding the construct by means of instructions and diagnostic tasks presented by the diagnostician. In a third type of method subjects must answer predefined questions of the diagnostician or respond to a diagnostic task intended to encourage them to verbalize aspects of knowledge which are at the focus of the diagnosis. A concrete example is the Struktur-LegeTechnik (SLT) developed by Scheele and Groeben (1984). In this task, subjects first answer open questions about the structure of a concept and then construct a mental map by organizing cards labelled with concepts and possible relations between concepts. Psychodiagnosis Based on Verbal Data 21 Responding to vicarious verbalization When subjects are required to respond to verbalizations presented by the diagnostician, their task is to agree or disagree on a presented statement on the basis of their knowledge. For example, the statements are presented in the format of a questionnaire. Subjects are instructed to select and mark those statements that match individual cognition, perception or personal experience with reference to the respective construct. In this sense, the verbalizations offered by the diagnostician substitute that which the subjects themselves would have spontaneously verbalized . The method of responding to verbal utterances presented by the diagnostician has been introduced for increasing diagnostic reliability. When responding to verbalizations, for example as presented in standardized questions with fully worded response options, additional verbal utterances of the subjects are not accepted. Subjects must stick to the provided response format. Apart from general instructions on how to use the method appropriately, there are no additional hints for the subject on how to respond to the items. The non-verbal responses of the subjects are taken as having the same relevance for the construct which is the focus of the diagnosis as subject-generated verbalizations. In the domain of cognitive diagnosis, the methods that are often used in the way just described are structured questionnaires, psychometric self-administered scales, semantic differentials (Osgood, Suci, & Tannenbaum, 1957), and structure mapping techniques (Novak, 1990). In clinical diagnosis the methods most often used are standardized diagnostic interviews and psychometric self-assessment scales (Wittchen & Lachner, 1996). As is the case of cognitive diagnostic methods, the task of the patient is to check which of the verbal utterances presented comes closest to the one he or she could have verbalized spontaneously with regard to his or her knowledge, perception, emotion, or experience. Psychodiagnosis Based on Verbal Data 22 All forms of vicarious language productions have the advantage that all facets of the construct are assessed. This has a positive effect on the objectivity and reliability of a method and may contribute to enhancing the validity of diagnosis. The problem is that in a standardized questionnaire subjects first have to understand the meaning of a given verbal utterance before they can respond. This is not always easy as the person in question may lack domain specific preknowledge. This again stresses the importance of pre-testing of diagnostic testing instruments. Provocated verbalizations When the diagnostician is interested in more elaborated language productions, it becomes necessary to induce them. Some methods that are frequently used are outlined below. Think aloud methods. These are methods used to assess cognitive processes and structures relevant for coping with tasks and for solving problems. Subjects are instructed to verbally utter every image and reflection that occurs to them during task performance. The verbal reports are tape-recorded and transcribed. The analysis of think aloud data is called protocol analysis (for details, see Ericsson & Simon, 1984). Examples of the think aloud method date back to the early days of cognitive psychology (Bühler, 1907; Claparède, 1932; Duncker, 1935). Some decades later the method experienced a renaissance in problem solving research (Larkin, 1979, 1983; Larkin et al., 1980; Newell & Simon, 1972) (see example above). Variants of the method were used by Chi and Koeske (1983) and Chi (1985). The validity of protocol data as indicators of cognitive structures and processes is currently being debated. Ericsson and Simon (1984) hold the view that thinking aloud is a direct articulation of the thought units and processes in working memory deemed relevant by the subject for coping with the diagnostic task situation. The authors distinguish two levels. At level 1 subjects directly produce verbal utterances on the basis of an already existing conceptual representation. At level 2, units and processes of the mind first have to be translated into a Psychodiagnosis Based on Verbal Data 23 conceptual format, because they may only be existent as a mental image. Criticism concerning the validity of the method has been put forward by Nisbett and Wilson (1977). Among other things, they were criticized of the notion that thinking aloud only allows for assessing cognitive processes retrospectively. According to their view the think aloud method yields individual interpretation and verbalization of cognitive processes but not the processes themselves. Interview methods. Interview methods may be distinguished according to their degree of standardization. In unstandardized explorative interviews (either verbal interviews or questionnaires) the content and sequencing of questions the subjects are asked is left to the interviewer and may only be structured by some general guidelines fixed in written format (Baumann & Stieglitz, 1994; Robins, 1989; Wittchen & Lachner, 1996). Grove and Meehl (1996) comment on various internal and contextual factors which may bias subject’s responses and may lead to systematic errors in the diagnostic procedure when open question formats are used. In recent decades, in educational as well as clinical diagnosis it has increasingly become the norm to use highly formalized and standardized methods of interviewing, at least in research settings. A characteristic feature of these methods is that both the kind of questioning and the kind of answers are strictly regularized. In interviews the questions subjects are required to answer are posed exactly as written. Variants of interview methods that have been used in various research contexts are, for example, the interview-about-instances-technique (Gilbert, Watts, & Osborne, 1985), the interrogation technique (Broadbent, 1977; Vosniadou, 1994), and the tutorial dialogue (Stevens & Collins, 1980; Williams, Hollan, & Stevens, 1983). In clinical diagnosis, in principle, the questions in standardized diagnostic interviews are more or less direct translations of the diagnostic criteria as listed in the DSM and ICD (Knäuper, 1999; Knäuper & Wittchen, 1994). The wording of the questions itself is the result of a consensus process between diagnostic experts (Robins, 1989). In this way all clinical forms of judgement Psychodiagnosis Based on Verbal Data 24 are left up to the patients themselves and systematic distortions of the clinical judgement by different persons are avoided. The questions posed to patients in a standardized diagnostic interview in combination with the individual`s responses are treated diagnostically as if they were original language productions of the patients. The rationale for this procedure is empirical. This is also true for the interpretation of answers to certain questions as indicators for certain mental disorders. In order to improve the concurrent validity of the instruments, empirical tests have been carried out as to whether or not answers to specific questions are given by persons with certain mental diseases (previously diagnosed with the use of a different diagnostic method). Probing techniques. In text processing, research specific tasks are applied to subjects with the intention of inducing language production. The production is taken as a basis for a diagnosis of text comprehension and knowledge acquisition. Since Bartlett’s (1932) studies on text processing and comprehension, free recall techniques are used for inferring aspects of mental representations and (re)constructive processes in comprehension. For example, summaries are used for inferring which kind of reconstructive processes have been used to produce the compressed version of the original text and whether a coherent macrostructure has been constructed (van Dijk, 1980). Answering questions at different levels of complexity is a method often used to analyze the depth and breadth of comprehension. The types of questions that a reader puts to a text may be used for an analysis of his or her cognitive dispositions and abilities (Flammer, 1981; Miyake & Norman, 1979). The Cloze-Procedure technique – a technique based on filling gaps in a text – is used to assess the degree of reading comprehension (Taylor, 1953). Objectivity and reliability In assuring objectivity and reliability of verbal data, there is often a conflict between the need for standardizing a diagnostic method and the aim of creating a situation that allows for an unrestrained externalization of an individual’s knowledge about his or her mental states. When Psychodiagnosis Based on Verbal Data 25 psychodiagnosis is based on a qualitative approach of data assessment and analysis the use of traditional psychometric methods based on quantitative data are not appropriate. Another reason for this is that the statistical conditions for using certain psychometric methods for data analysis are often unknown or not fulfilled. For example, the stability of the state which comprises the focus of the diagnosis is a statistical presupposition when assessing test-retest reliability. This assumption may be appropriate when stable mental states are the focus of attention. It is, however, inappropriate in some contexts, for example, in the context of instruction because cognitive states such as an individual knowledge structure are subject to change as a consequence of learning and instruction (Messick, 1989). Instead of measuring retest-reliability it would be more appropriate to measure inter-rater reliability (e.g., Weber, 1994). Psychodiagnostic methods based on the concept mapping technique which allows for a sufficient standardization of data assessment have recently been suggested by Baumann and Stieglitz (1994), Eckert (1998), and Weber (1994). (4) Data analysis and interpretation Data analysis and interpretation is based on the fourth assumption of our diagnostic model which holds that the diagnostician can infer the individual’s mental representation of knowledge which concerns aspects of his or her mental state on the basis of the latters’ language products. Data analysis. The basis for the analysis of verbal data in psychodiagnosis may comprise different kinds of language products. These are, for example, texts such as verbal or written summaries in the context of diagnosing text comprehension, isolated statements (propositions) as responses to open questions in an interview or questionnaire, concepts and relations as a result of concept mapping, or quasi-verbal data such as responses in a highly standardized questionnaire. In most cases, raw data must be prepared according to the constraints of both the kind of cognitive or clinical construct the diagnostician is interested in and the data analysis method. Psychodiagnosis Based on Verbal Data 26 Language products may be broken down and categorized into measurement units at different levels of aggregation. These are, for example, turns as units of communication, passages in written documents, sentences, propositions, words, syllables, or sounds. The decision about which units should be analyzed, which level of aggregation should be used, which prescription the diagnostician should allocate to what measurement value (score) to a unit of a certain category of analys, is is constrained by the conceptual model as a result of the first step in psychodiagnosis: the construct analysis (see Nichols, 1994). The methods used in cognitive and clinical diagnosis for analyzing language products are the same: They are interpretative, content analytical, and formal analytical methods. When interpretative methods are used, language products are interpreted by the diagnostician in a holistic manner (Brunner, 1982). Content analytical methods may be viewed as a systematization of interpretative methods. When applying these methods, the diagnostician must break down existing language products into smaller units. These units are then analyzed according to their relevance for the psychological construct and allocated to certain construct relevant categories. Formal analytical methods are based on countable features of language production as, for example, word length, sentence length, type of words used. The score assigned for the respective psychological construct is then described in quantitative terms. Data interpretation. For data interpretation in a qualitative diagnostic approach the diagnostician has to look for indicators allocated to certain construct relevant categories and interpret them with reference to the psychological model of the construct assumed to have affected the content and structure of a particular language product. This process is non-trivial and depends extensively on the validity of the indicators at the level of language products. The problems associated with the interpretation of verbal data based on verbal reports on individual problem solving processes have been outlined in detail by Nisbett and Wilson (1977). In Psychodiagnosis Based on Verbal Data 27 psychometric diagnostic approaches quantitative data representing the value of a construct are often compared with other individuals or with group norms. In contrast, in qualitative diagnostic approaches data interpretation is a process of understanding the analyzed verbal data in light of the psychological model of the construct. In language-based knowledge diagnosis certain individual utterances or aggregated and categorized utterances are taken as indicators of an individual cognitive state, i.e. the structure of an individual`s knowledge. This process has been labelled “reconstruction of the individual knowledge representation” (Tergan, 1988, 1989 b). In clinical diagnosis there are various approaches which depend on the theoretical paradigm the diagnostician is using for interpreting verbal data. These could be, for example, the interpersonal, the system-oriented, or the cognitive paradigm. Examples for cognitive clinical methods are Beck’s or Lorenzer’s approaches (see Engel, 1986; Hautzinger, 1982). (5) Evaluation of results The evaluation of the results of psychodiagnostic analysis has two foci. On the one hand, diagnostic approaches must be evaluated with reference to quality criteria such as validity, reliability, and objectivity. As discussed in the above, classical psychometric quality criteria are often ill-suited to the methods described here due to the dominance of qualitative data in assessing mental representations and the dynamic character of the construct in focus. The suitability of criteria for assessing the validity of a diagnostic measurement concerning aspects of an individual’s knowledge representation is still a matter of debate (see Eckert, 1998). The proposed approach may contribute to providing a basis for a sound construct evaluation. On the other hand, the diagnostic results must be evaluated with reference to the goal of diagnosis and the context of application. A typical goal of diagnosis in the instructional context is the decision about the kind of instructional treatment for fostering learning. In clinical diagnosis, a typical Psychodiagnosis Based on Verbal Data 28 goal is to decide on the appropriate treatment of psychological problems. The diagnostician must evaluate the practical relevance of diagnostic results with reference to criteria and norms as a basis for decisions. The goals of diagnosis and the context of application determine to a large degree which kind of criteria and norms should be used for the evaluation of the results of diagnosis. When results are compared to an external criterion the evaluation may follow either a criterion referenced approach or a norm referenced approach. In a criterion referenced approach, the criterion may be a score indicating a certain level of a cognitive or mental affective state. In the context of cognitive diagnosis such a score may be, for example, a certain amount of relevant concepts and relations used in a concept mapping task. It may also be an expert rating indicating the degree of correspondence between the individual knowledge representation and an expert’s representation. For example, in the context of cognitive research on knowledge acquisition, the researcher may be interested in changes to the structure of knowledge as a result of cognitive development or to a certain instructional measure. He or she may have assessed the amount of concepts and relations used by an individual in a concept mapping task. Depending on the psychological model selected as a reference model, the researcher interprets existing changes as indicators for cognitive development that have taken place within a certain period of life, or as indicators for successful knowledge acquisition and conceptual change (e.g. Eckert, 1998; Mandl & Ballstaedt, 1986; Vosniadou & Brewer, 1992; Weber, 1994). In the clinical context, a typical example of a criterion referenced approach is a comparison of the individual verbal ability with a criterion defined by the diagnostician as a basis for diagnosing mental health or a mental disorder, for example dementia. In a norm-referenced approach the results of an individual diagnosis are compared to a reference or control group. Or results of novices are compared to the results of an expert or a Psychodiagnosis Based on Verbal Data 29 group of experts (Larkin, 1979). In clinical diagnosis, a norm referenced approach is used when comparing clinical symptoms of an individual with those of a group of individuals assumed to have the same mental disorder. A norm-referenced approach is also used when the diagnosed individual is compared with a group of healthy individuals so as to evaluate the degree and level of differences with respect to certain aspects of the disorder. (6) Application of results The goal of psychodiagnosis is to support diagnosticians in making decisions concerning instructional or therapeutic measures. As has been outlined in the above, in the field of instruction the decision has to do with which instructional measure should be taken for fostering learning. In clinical diagnosis decisions have to be made about which therapy is appropriate for a specific mental disorder (Krapp, 1979). A prominent example of cognitive diagnosis in the domain of instruction is the diagnosis of individual knowledge structures. The focus of knowledge diagnosis is to help instructors tailor instruction to the cognitive needs of the students and to help designers of technology-based learning environments to design adaptive instruction to make learning environments adaptable to the cognitive deficits of learners. Whereas it may be sufficient for decisions of selection to rely on quantitative data, for example, the performance of individuals, it is necessary to include qualitative data for planning modifications of individual cognitive or mental states at a more detailed level. Summary and Conclusion Psychodiagnostic methods based on language products are very important for diagnoses in educational as well as clinical settings. Their results make the inference and explanation of mental states of individuals possible. In educational diagnosis, the conceptual rationale for this inference is often grounded on theories of the mind and models of mental representation of knowledge. In clinical diagnosis it is based on extensive empirical evidence. In the present paper Psychodiagnosis Based on Verbal Data 30 we have sought to outline commonalities of both approaches. We used a cognitive conceptual rationale proposed by Tergan (1988, 1989 a, b) and Nichols (1994) in the context of knowledge diagnosis which has been elaborated and transferred to applied contexts by Weber (1994) and Eckert (1998). The rationale was chosen for integrating cognitive and clinical language-based psychodiagnostic approaches into a general conceptual framework. As has been shown, such an approach appears fruitful for developing a systematic theoryoriented methodology. The use of a cognitive approach to psychodiagnosis is said to be advantageous with respect to its theoretical grounding and its practical usability in a variety of diagnostic settings when compared to a strictly psychometric approach (Bejar, 1993; Nichols, Chipman, & Brennan, 1995). It is also advantageous in practical contexts where instructional and therapeutic measures must closely match the conditions of an individual’s representation of his or her mental state. In knowledge diagnosis these are ideosyncratic aspects of a concept structure which represent the meaning of an individual’s attributes to a concept. Knowing the individual concept structure may be important for designing tailored ins tructions for fostering learning or conceptual change. In clinical diagnosis, a cognitive approach could be advantageous for supplementing the traditional psychometric approach by identifying ideosyncratic facets of a patient’s representation of a mental disorder, as well as relations between different clinical symptoms. It may also be useful for developing a more differentiated system of symptom classes, for classifying individuals and assigning individualized therapeutic measures. There are, however, shortcomings that have yet to be overcome. Approaches in languagebased psychodiagnosis still lack a strong theoretical and empirical basis. This is particularly true for approaches in new and complex research domains where theories of the effects of knowledge on mental states have not yet been developed. Tergan (1989b) comments on the status of existing background-theories for a qualitative knowledge diagnosis. He is critical of the fact that while Psychodiagnosis Based on Verbal Data 31 cognitive science has developed models for representing knowledge and modelling knowledgebased cognitive processes per computer, there is often a lack of theory with respect to the definition and operationalization of verbal data as indicators of the psychological construct. There is also a lack of methodology for the identification, aggregation, analysis as well as rules and procedures for the reconstruction of individual features of the psychological construct as based on the identified indicators in individual language production. Furthermore, there appears to be a lack of empirical evidence concerning the construct validity of the assumptions underlying cognitive representational models (see Tergan, 1989b). For a theoretical grounding of a language-based cognitive psychodiagnosis it would seem important to integrate assumptions about theories of the mind, models for the representation of knowledge and assumptions about processes of externalising internal representations into a general theory of language-production. For this purpose, existing theories of language production (e.g. Levelt, 1989) must be elaborated and adapted. Individual preconditions, social and contextual conditions have to be taken into consideration to a far greater extent than hitherto. Without a sound theoretical basis, the process of inferring cognitive or mental states from verbal data has only face value. The conceptual framework outlined may help to overcome existing theoretical and methodological shortcomings in language-based psychodiagnosis. It may also help to initiate and guide empirical research in a framework of construct validation (Campbell & Fiske, 1959; Cronbach & Meehl, 1955; Herrmann, 1985), to develop a more systematic and theory-oriented diagnostic procedure and, thus, contribute to the enhancement of the validity and applicability of language-based psychodiagnostic approaches as well as to theory development. Psychodiagnosis Based on Verbal Data 32 ReferencesAiken, L. R. (1997). Mental testing and assessment. Boston, MA: Allyn and Bacon, Inc. Anderson, J. R. (1983). The architecture of cognition. Cambridge, Mass.: Harvard UniversityPress. Anderson, J.R. (1993). Rules of the mind. Hillsdale, NJ: Lawrence Erlbaum . Andreasen, N. C., Grove, W. M., Shapiro, R. W., Keller, M. B., Hirschfeld, R. M. A., & McDonald-Scott, P. (1981). Reliability of lifetime diagnosis. Archives of General Psychiatry, 38, 400-405.Ballstaedt, S.-P. (1982). Dokumentenanalyse. In G.L. Huber & H. Mandl (Eds.), Verbale Daten.Eine Einführung in die Grundlagen und Methoden der Erhebung und Auswertung (165-176).Weinheim/Basel: Beltz Verlag.Ballstaedt, S.-P. (1987). Zur Dokumentenanalyse in der biographischen Forschung. In G. Jüttemann& H. Thomae (Eds.), Biographie und Psychologie (S. 203-214). Berlin/Heidelberg: Springer.Bartlett, F. C. (1932). Remembering. A study in experimental and social psychology. Cambridge:University Press. Bastine, R., & Tuschen, B. (1996). Klinisch-psychologische Diagnostik [Clinical-mentaldiagnosis]. In A. Ehlers, & K. Hahlweg (Eds.), Enzyklopädie der Psychologie,Themenbereich D, Praxisgebiete, Serie II, Klinische Psychologie, Vol. 1, Grundlagen derKlinischen Psychologie (pp. 195-268). Göttingen: Hogrefe. Baumann, U., & Stieglitz, R.-D. (1994). Psychodiagnostik psychischer Störungen[Psychodiagnosis of mental disorders]. Stuttgart: Enke. Psychodiagnosis Based on Verbal Data 33 Bejar, I.I. (1993). A generative approach to mental and educational measurement. In N.Frederiksen, R. J. Mislevy, & I.I. Bejar (Eds.), Test theory for a new generation of tests(pp. 323-358). Hillsdale, NJ: Lawrence Erlbaum. Broadbent, D.E. (1977). Levels, hierarchies, and the locus of control. Quarterly Journal ofExperimental Psychology, 29, 181-201. Brugha, T. S., Bebbington, P. E., & Jenkins, R. (1999). A differences that matters: Comparisonsof structured and semi-structured psychiatric diagnostic interviews in the generalpopulation. Mental Medicine, 29 (5), 1013-1020. Brunner, E. J. (1982). Interpretative Auswertung. In G.L. Huber, & H. Mandl (Eds.), VerbaleDaten (p. 197-219). Weinheim: Beltz. Bühler, K. (1907). (I) Tatsachen und Probleme einer Psychologie der Denkvorgänge. Archiv fürPsychologie, 9, 297-305. (II) Über Gedanken. Archiv für Psychologie, 12, 1-23. Bühler, K. (1932). Sprachtheorie. Die Darstellungsfunktion der Sprache. Stuttgart: GustavFischer. Campbell, D. A., & Fiske, D. W. (1959). Convergant and discriminant validation by themultitrait-multimethod-matrix. Mental Bulletin, 56, 81-105. Chi, M. T. H., & Koeske, R. D. (1983). Network representation of a child’s dinosaur knowledge.Developmental Psychology, 19, 29-39. Chi, M. T. H. (1985). Interactive roles of knowledge and strategies in the development oforganized sorting and recall. In S. Chipman, J. Segal, & R. Glaser (Eds.), Thinking andlearning skills (pp. 457-484). Hillsdale, NJ: Lawrence Erlbaum. Psychodiagnosis Based on Verbal Data 34 Chi, M. T. H., Feltovich, P.J., & Glaser, R. (1981). Categorization and representation of physicsproblems by experts and novices. Cognitive Science, 5, 121-152. Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in problem solving. In R. J. Sternberg(Ed.), Advances in the psychology of human intelligence (Vol. 1, pp. 73-96). Hillsdale,NJ: Lawrence Erlbaum. Claparède, E. (1932). Die Entdeckung der Hypothese. In C.F. Graumann (Ed.) (1969), Denken(pp. 109-115). Köln: Kiepenheuer & Witsch. Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in mental tests. Mental Bulletin, 52(4), 281-302. Dawes, R. M., Faust, D., & Meehl, P. E. (1989). Clinical versus actuarial judgment. Science, 243,1668-1674. Dochy, F.J.R.C. (1992). Assessment of prior knowledge as a determinant for future learning.Center for Educational Technology and Innovation. Utrecht: Open University of theNetherlands.Duncker, K. (1935/1966). Zur Psychologie des produktiven Denkens. Berlin: Springer. van Dijk, T. A. (1980). Macrostructures. Hillsdale, NJ: Lawrence Erlbaum. Eckert, A. (1998). Kognition und Wissensdiagnose. Die Entwicklung und empirischeÜberprüfung des computerunterstützten wissensdiagnostischen InstrumentariumsNetzwerk-Elaborations-Technik (NET). Lengerich: Pabst. Engel, K. (1986). Zur theoretischen Einordnung des Gottschalk-Gleser-Verfahrens (unterbesonderer Berücksichtigung der psychoanalytischen Theorie und derKommunikationstheorie). In U. Koch, & Schoefer, G. (Eds.), Sprachinhaltsanalyse in der Psychodiagnosis Based on Verbal Data 35 psychiatrischen und psychosomatischen Forschung: Grundlagen und Anwendungsstudienmit den Affektskalen von Gottschalk und Gleser (pp. 19-34). Weinheim: PsychologieVerlags Union. Engelkamp, J. (1983). Sprache und Emotion. In H. A. Euler, & H. Mandl (Eds.),Emotionspsychologie: Ein Handbuch in Schlüsselbegriffen (pp. 262-267). München:Urban & Schwarzenberg. Engelkamp, J., & Zimmer, H. D. (1994). The human memory. A multi-modal approach. Bern:Huber. Ericsson, K.A. & Simon, H.A. (1980). Retrospective verbal reports as data. PsychologicalReview, 87, 215-251. Exner, J. E. (1986a). The Rorschach: A comprehensive system (Vol. I: Basic foundations). NewYork: Wiley.Exner, J. E. (1986b). Some Rorschach data comparing borderline with schizophrenics andschizotypal personality disorder. Journal of Personality Assessment, 50, 455-471.Ericsson, K. A., & Simon, H. A. (1984). Protocol analysis. Cambridge, Mass: MIT Press. Ewert, O., & Thomas, J. (1996). Das Verhältnis von Theorie und Praxis in derInstruktionspsychologie. In F. E. Weinert (Ed.), Enzyklopädie der Psychologie.Psychologie des Lernens und der Instruktion (pp. 89-118). Göttingen: Hogrefe. Farmer, A. E., & McGuffin, P. (1989). The classification of the depressions: Contemporaryconfusion revisited. British Journal of Psychiatry, 155, 437-443.Flammer, A. (1981). Towards a theory of question asking. Psychological Review, 43, 407-420. Frederiksen, N., Mislevy, R. J., & Bejar, I. I. (1993) (Eds.). Test theory for a new generation oftests. Hillsdale, NJ: Lawrence Erlbaum. Psychodiagnosis Based on Verbal Data 36 Glynn, S. M., & Duit, R. (1995). Learning science in the schools. Research reforming practice.Mahwah, NY: Lawrence Lawrence Erlbaum. Gilbert, J. K., Watts, D. M., & Osborne, R. J. (1985). Eliciting student views using an interview-about-instances technique. In L. H. T. West, & A. L. Pines (Eds.), Cognitive structure andconceptual change (pp. 11-27). Orlando: Academic Press Inc. Gottschalk, L. A. (1995). Content analysis of verbal behavior: New findings and clinicalapplications. Hillsdale: Lawrence Erlbaum. Grabowski-Gellert, J., & Winterhoff-Spurk, P. (1989). Schreiben ist Silber, Reden ist Gold. EineUntersuchung zur Äquivalenz von mündlicher und schriftlicher Erhebungsmethode beiExperimenten zur Sprachproduktion. Arbeiten aus dem Sonderforschungsbereich 245„Sprechen und Sprachverstehen im sozialen Kontext“. Heidelberg/Mannheim. Greeno, J.G. (1980). Some examples of cognitive task analysis with instructional implications. InR. E. Snow, P. A. Federico, & W. E. Montague (Eds.), Aptitude, learning and instruction(Vol. 2, pp. 1-22). Hillsdale, NJ: Lawrence Lawrence Erlbaum. Grove, W., & Meehl, P. (1996). Comparative efficacy of informal (subjective, impressionistic)and formal (mechanical, algorithmic) prediction procedures: The clinical-statisticalcontroversy. Psychology, Public Policy, and Law, 2, 293-323. Hautzinger, M. (1982). Ein Kategorienssystem zur Erfassung kognitiver Veränderungen.Zeitschrift für personenzentrierte Psychologie und Psychotherapie, 1, 47-61. Herrmann, T. (1985). Allgemeine Sprachpsychologie. Grundlagen und Probleme. München:Urban & Schwarzenberg. Psychodiagnosis Based on Verbal Data 37 Herrmann, T., & Grabowski, J. (1994). Sprechen. Psychologie der Sprachproduktion. Heidelberg:Spektrum. Jacobson, M., & Archodidou, A. (2000). The design of hypermedia tools for learning: fosteringconceptual change and transfer of complex scientific knowledge. The Journal of theLearning Sciences, 9 (2), 149-199. Jäger, R. S. (1983). Der diagnostische Prozeß: Eine Diskussion psychologischer undmethodischer Randbedingungen. Göttingen: Hogrefe. Keller, M. B., Lavori, P. W., McDonald-Scott, P., Scheftner, W. A., Andreasen, N. C., Shapiro, R.W., & Croughan, J. (1981). Reliability of lifetime diagnoses and symptoms in patients with acurrent psychiatric disorder. Journal of Psychiatric Research, 16 (4), 229-240.Kintsch, W. (1974). The representation of meaning in memory. Hillsdale, NJ: LawrenceLawrence Erlbaum Ass. Kluwe, R. H. (1988). Methoden der Psychologie zur Gewinnung von Daten über menschlichesWissen. In H. Mandl & H. Spada (Eds.),Wissenspsychologie (pp. 359-385). München:Psychologie Verlags Union. Knäuper, B., & Wittchen, H.-U. (1994). Diagnosing Major Depression in the elderly: Evidencefor response bias in standardized diagnostic interviews? Journal of Psychiatric Research,28 (2), 147-164. Knäuper, B. (1999). Relevante Fragen, valide Antworten: Eine sprachund sozialpsychologischeAnalyse verbaler subjektiver Einschätzungen. Kurzmonographie im Rahmen desHabilitationsverfahrens, Freie Universität Berlin. Krapp, A. (1979). Prognose und Entscheidung. Zur theoretischen Begründung undDifferenzierung der pädagogisch-psychologischen Prognose. Weinheim: Beltz. Psychodiagnosis Based on Verbal Data 38 Larkin, J. H. (1979). Processing information for effective problem solving. EngineeringEducation, 70, 285-288. Larkin, J. H. (1983). The role of representation in physics. In D. Gentner, & A. Stevens (Eds.),Mental models (pp. 75-93). Hillsdale: Lawrence Erlbaum. Larkin, J. H., McDermott, J., Simon, D. P., & Simon, H. A. (1980). Models of competence insolving physics problems. Cognitive Science, 4, 317-345. Levelt, W. J. M. (1989). Speaking. From intention to articulation. Cambridge, Mass.: MIT Press. Leutner, D. (1997). Adaptivität multimedialer Lehrund Informationssysteme. In L. J. Issing, &P. Klimsa (Eds.), Information und Lernen mit Multimedia (2nd ed., pp. 139-149).Weinheim: Beltz-PVU. Lohmann, D. F., & Ippel, M. J. (1993). Cognitive diagnosis. From statistically based assessmenttoward theory-based assessment. In N. Frederiksen, R. J. Mislevy, & I. I. Bejar (Eds.),Test theory for a new generation of tests (pp. 41-71). Hillsdale, NJ: Lawrence Erlbaum. Mandl, H., & Ballstaedt, St.-P. (1986). Assessment of concept building in text comprehension. InF. Klix, & H. Hagendorf (Eds.), Human memory and cognitive capabilities. Mechanismsand performances (pp. 861-870). Amsterdam: Elsevier Science Publishers. Mandl, H., & Huber, G. L. (Eds.) (1978). Kognitive Komplexität. Bedeutung,Weiterentwicklung, Anwendung. Göttingen: Hogrefe. Meehl, P. E. (1954). Clinical vs. statistical prediction: A theoretical analysis and a review of theevidence. Minneapolis, MN: University of Minnesota Press. Messick, S. (1989). Validity. In R. J. Linn (Ed.), Educational measurement (3. ed., pp. 3-104).New York: Macmillan. Psychodiagnosis Based on Verbal Data 39 Mislevy, R. J. (1993). Foundations of a new test theory. In N. Frederiksen, R. J. Mislevy & I. I.Bejar (Eds.), Test theory for a new generation of tests (pp. 19-40). Hillsdale, NJ:Lawrence Erlbaum. Miyake, N., & Norman, D. A. (1979). To ask a question, one must know enough to know what isnot known. Journal of Verbal Learning and Verbal Behavior, 18, 35-364. Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs: Prentice Hall. Nichols, P. D. (1994). A framework for developing cognitively diagnostic assessments. Reviewof Educational Research, 64 (4), 575-603. Nichols, P. D., Chipman, S. F., & Brennan, R. L. (1995). Cognitively diagnostic assessment.Hillsdale, NJ: Lawrence Erlbaum. Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mentalprocesses. Mental Review, 84, 231-259. Novak, J. D. (1990). Concept mapping. A useful tool for Science Education. Journal of Researchin Science Teaching, 27 (10), 937-949. Ohlsson, S. (1993). The interaction between knowledge and practice in the acquisition ofcognitive skills. In S. Chipman & A.L. Meyrowitz (Eds.), Foundations of knowledgeacquisition: Cognitive models of complex learning (pp. 147-208). Boston: KluwerAcademic. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana:University Illinois Press. Paivio, A. (1986). Mental representations. A dual coding approach. New York: OxfordUniversity. Psychodiagnosis Based on Verbal Data 40 Robins, L.N. (1988). An overview of the Diagnostic Interview Schedule and the CompositeInternational Diagnostic Interview. In J.E. Mezzich & M. von Cranach (Eds.), InternationalClassification in Psychiatry. Unity and Diversity (pp. 205-220). Cambridge: CambridgeUniversity Press.Robins, L. N. (1989). Diagnostic grammar and assessment: Translating criteria into questions.Mental Medicine, 19, 57-68. Rumelhart, D. E., & Norman, D. A. (1978). Accretion, tuning and restructuring: three modes oflearning. In J. W. Cotton, & R. L. Klatzky (Eds.), Semantic factors in cognition (pp. 37-53). Hillsdale: Lawrence Erlbaum. Rumelhart, D. E., & Norman, D. A. (1985). Representation of knowledge. In A. M. Aitkenhead,& J. M. Slack (Eds.), Issues in cognitive modeling. Hillsdale, NJ: Lawrence Erlbaum.Sanders, L. M., Adams, J., Tager-Flusberg, H., Shenton, M. E., et al. (1995). A comparison ofclinical and linguistic indices of deviance in the verbal discourse of schizophrenics. AppliedPsycholinguistics, 16 (3), 325-338.Scheele, B., & Groeben, N. (1984). Die Heidelberger Struktur-Lege-Technik (SLT). Weinheim:Beltz. Seidenstücker, G., & Baumann, U. (1978). Multimethodale Diagnostik. In U. Baumann,H. Berbalk, & G. Seidenstücker (Eds.), Klinische Psychologie: Trends inForschung und Praxis (pp. 134-182). Bern: Huber. Semler, G., Wittchen, H.-U., Joschke, K., Zaudig, M., von Geiso, T., Kaiser, S., von Cranach, M., &Pfister, H. (1987). Test-retest reliability of a standardized psychiatric interview (DIS/CIDI).European Archives of Psychiatric Neurological Science, 236, 214-222. Psychodiagnosis Based on Verbal Data 41 Shaffer, D., Fisher, P. W., & Lucas, C. P. (1999). Respondent-based interviews. In D. Shaffer, C. P.Lucas, & P. W. Fisher (Eds.), Diagnostic assessment in child and adolescentpsychopathology (pp. 3-33). New York: NY: Guilford Press.Sleeman, D., & Brown, J. S. (1982) (Eds.). Intelligent tutoring systems. London: AcademicPress. Snow, R. E., & Lohman, D. F. (1989). Implications of cognitive psychology for educationalmeasurement. In R. L. Linn (Eds.), Educational measurement (3. ed., pp. 263-331). NewYork: Macmillan. Spada H., & Wiechmann, S. (1998). Kognitive Determinanten der Lernleistung. In F. E. Weinert(Ed.), Enzyklopädie der Psychologie. Psychologie des Lernens und der Instruktion (Vol.2, pp. 119-152). Göttingen: Hogrefe. Sternberg, R. J. (1991). Cognitive theory and psychometrics. In R. K. Hambleton, & J. N. Zaal(Eds.), Advances in educational and psycho logical testing (pp. 367-394). Boston: KluwerAcademic Publishers. Stevens, A. L., & Collins, A. (1980). Multiple conceptual models of a complex system. In R. W.Snow, P. A. Federico, & W. E. Montague (Eds.), Aptitude, learning and instruction (pp.177-198). Hillsdale, NJ: Lawrence Erlbaum. Strube, G., & Schlieder, C. (1998). Wissensrepräsentation im Symbolverarbeitungsansatz. In F.Klix, & H. Spada (Eds.), Enzyklopädie der Psychologie: Wissen (Vol. 6, pp. 501-530).Göttingen: Hogrefe. Taylor, W. L. (1953). Cloze-procedure: a new tool for measuring reasoning, JournalismQuarterly, 30, 415-433. Psychodiagnosis Based on Verbal Data 42 Tergan, S.-O. (1986). Modelle der Wissensrepräsentation als Grundlage qualitativerWissensdiagnostik. Opladen: Westdeutscher Verlag GmbH. Tergan, S.-O. (1988). Qualitative Wissensdiagnose – Methodologische Grundlagen. In H. Mandl,& H. Spada (Eds.), Wissenspsychologie (pp. 400-422). München: Psychologie VerlagsUnion. Tergan, S.-O. (1989 a). Psychologische Grundlagen der Erfassung individuellerWissensrepräsentationen. Teil I: Grundlagen der Wissensmodellierung. Sprache undKognition, 8 (3), 152-165. Tergan, S.-O. (1989 b). Psychologische Grundlagen der Erfassung individuellerWissensrepräsentationen. Teil II: Methodologische Aspekte. Sprache und Kognition, 8(4), 193-202. Tesser, A. & Martin, L. (1996). The psychology of evaluation. In E. T. Higgins & A. W.Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 400-432). NewYork: NY: Guildford Press. Vosniadou, S. (1994). Capturing and modelling the process of conceptual change. Learning andInstruction, 4 (1), 45-69. Vosniadou, S. & Brewer, W. F. (1992). Mental models of the earth: A study of conceptualchange in childhood. Cognitive Psychology, 24, 535-585. Weber, S. (1994). Vorwissen in der betriebswirtschaftlichen Ausbildung. Eine strukturundinhaltsanalytische Studie. Wiesbaden: Deutscher Universitätsverlag. Wenger, E. (1987). Artificial Intelligence and tutoring systems. Los Altos, Calif: MorganKaufmann Publ., Inc. Psychodiagnosis Based on Verbal Data 43 Wenzel, A., & Holt, C. S. (2000). Situation-specific scripts for threat in two specific phobias.Journal of Psychopathology and Behavioral Assessment, 22 (1), 1-21. Westmeyer, H. (1990). Wissenschaftstheoretische Grundbegriffe für Klassifikation, Ätiologieund Diagnostik. In U. Baumann, & M. Perrez (Eds.), Lehrbuch Klinische Psychologie(Vol. 1, pp. 77-87). Bern: Huber. Williams, M., Hollan, J., & Stevens, A. L. (1983), Human reasoning about a simple physicalsystem. In D. Gentner, & A. L. Stevens (Eds.), Mental models (pp.131-153). Hillsdale NJ:Lawrence Erlbaum. Winograd, T. (1975). Frame representations and the declarative/procedural controversy. In D. G.Bobrow & A. M. Collins (Eds.), Representation and understanding: Studies in cognitivescience. New York: Academic Press. Wittchen, H.-J. (1994). Reliability and validity of the WHO-Composite International DiagnosticInterview (CIDI): A critical review. Journal of Psychiatric Research, 28 (1), 57-84.Wittchen, H.-U. (1994). Klassifikation. In R.-D. Stieglitz, & U. Baumann (Eds.),Psychodiagnostik psychischer Störungen (pp. 47-66). Stuttgart: Enke. Wittchen, H.-U., & Lachner, G. (1996). Klassifikation. In A. Ehlers, & K. Hahlweg (Eds.),Enzyklopädie der Psychologie: Klinische Psychologie (pp. 3-68). Göttingen: Hogrefe. Wittchen, H.-U., Üstün, T. B., & Kessler, R. C. (1999). Diagnosing mental disorders in thecommunity: A difference that matters? Mental Medicine, 29 (5), 1021-1027. Wittrock, M.C. & Baker, E.L. (1991), Testing and cognition. Englewood Cliffs, NJ: PrenticeHall. Psychodiagnosis Based on Verbal Data 44 Wottawa, H., & Hossiep, R. (1987). Grundlagen psychologischer Diagnostik: Eine Einführung.Göttingen: Hogrefe. Figure 1. Hypothetical processes in memory in language-based cognitive psychodiagnosis. Whencoping with a particular diagnostic task, after having processed the sensory input (SR = sensoryregister) aspects of an individual’s knowledge about the mental state in focus (the structure of aconcept, the feelings in a state of depression) are activated in long-term memory (LTM) andprocessed in working memory (WM) to construct a mental representation which is appropriatefor generating language products (see Kluwe, 1988, p. 361) Figure 2. Components and processes involved in cognitive diagnosis (see Tergan, 1989 b, p. 121)