Components of reading comprehension key function to discourse analysis & Big Data analytics need a specific mindset 'intricacy skills'

Components of reading comprehension and scholastic achievement

Chiara Meneghetti ?, Barbara Carretti, Rossana De Beni

Department of General Psychology, University of Padova, Italy

1. Introduction

Reading comprehension is a complex cognitive ability requiring the capacity to integrate text information with the knowledge of the listener/reader and resulting in the elaboration of a mental representation. Current models of reading comprehension highlight the importance of considering the role of different cognitive processes during text comprehension. For example, memory both in its short- and long-term components is broadly considered to have a fundamental role. Indeed, the reader has to store and manipulate information in working memory during the processing of the text, but at the same time in order to construct a coherent representation of the text usually he/she has to refer to his/her prior knowledge (van den Broek, 1994). The studies on reading comprehension have often adopted an individual differences viewpoint as an attempt to account for the processes and components that might differentiate skilled and less skilled readers, the latter usually named ‘poor comprehenders’ (Oakhill, Cain, & Bryant, 2003). Poor comprehenders are those individuals who have an average IQ but are specifically impaired in understanding the meaning of a text. The comparisons between good and poor comprehenders are usually made using tasks measuring either global or specific aspects of reading comprehension, such as the ability to make inferences (Cain & Oakhill, 1996, 1999; Ehrlich & Remond, 1997; Nation & Snowling, 1998, 1999; Stothard & Hulme, 1992; Yuill & Oakhill, 1991). The results of these studies have shown that poor comprehenders are at a particular disadvantage when they are required to execute a process that requires integrating newly encountered information with information encountered earlier in the text or retrieved from long-term memory. So, for example, poor comprehenders have problems interrelating successive topics (Lorch, Lorch, & Morgan, 1987) and integrating information to derive the overall gist or main theme of a passage (Palincsar & Brown, 1984).

Furthermore, poor comprehenders differ from good comprehenders in their ability to integrate text information, understand story structure and monitor their own understanding (Cain & Oakhill, 1996, 1999; Cain, Oakhill, Barnes, & Bryant, 2001; Ehrlich & Remond, 1997; Yuill & Oakhill, 1991). Poor comprehenders are less sensitive to semantic inconsistencies in the text compared to good comprehenders (Garner, 1981). At the same time less skilled comprehenders are more prone to allow the intrusion of irrelevant information (Gernsbacher & Faust, 1991; Gernsbacher, Varner, & Faust, 1990). Also from a metacognitive point of view, poor comprehenders have difficulties. It has been shown that they seem less aware of the importance of giving meaning to a text stressing the relevance of the decoding phase to the detriment of comprehension (Baker & Brown, 1984; Pazzaglia, De Beni, & Cacciò, 1999). Moreover, they sometimes do not realize when they do not understand (Garner & Reis, 1981), they rarely use reading strategies (Brown, Armbruster, & Baker, 1986) and they have more difficulty in adjusting the strategy to the task (Palincsar & Brown, 1984). Also the ability to evaluate the complexity of the text, to detect text structure, to monitor comprehension and to spot anomalies are factors which can discriminate between good and poor comprehenders (August, Flavell, & Cliff, 1984; Brown et al., 1986; Garner, 1981). The short review of the main findings on poor comprehenders' performance reveals a heterogeneous pattern of difficulties.

However this does not mean that all these aspects necessarily occur simultaneously in all cases of comprehension difficulty. In agreement with this view, Cornoldi, De Beni, and Pazzaglia (1996) discussed the idea that poor comprehenders show a high level of variability in their cognitive and metacognitive profiles. Cornoldi et al. (1996) longitudinally studied a group of poor and good comprehenders comparing their performance in a number of learning measures (e.g. reading and listening comprehension, decoding skills), cognitive measures (such as the Primary Mental Abilities by Thurstone and Thurstone (1963) and working memory) and metacognitive aspects related to reading comprehension (such as knowledge of reading goals, strategy use and text sensitivity).

 The single cases analysis highlighted that a reading comprehension failure generally implied a lower metacognitive control on reading comprehension, whereas in some participants it implies also a listening comprehension difficulty or a poor use of strategies. Poor comprehenders also failed, as a group, in a series of working memory tasks, but in some cases this deficit was limited to those tasks that need to rely on sentence processing. These results led the authors to claim that all these aspects are necessary prerequisites contributing, at higher levels, in the role of reading performance facilitators. Thus, a lower performance in metacognitive and cognitive tasks could produce a lower performance in reading comprehension. Conversely, low reading comprehension could affect some or all cognitive and metacognitive abilities. Furthermore, the mental model studies offer the possibility of understanding the kind of differences between the effective and ineffective reader (Lorch & van den Broek, 1997; McNamara, Kintsch, Songer, & Kintsch, 1996). Perrig and Kintsch's (1985) model proposes the distinction between a text-based representation, which maintains the verbal characteristics of the message, and a situation model, representing the situation described therein in which the text content is integrated into the comprehender's knowledge system. Following this line of research, some studies found that good comprehenders differ in a number of reading-related skills allowing them to build better-integrated and richer situation model (e.g. Lonka, Lindblom-Ylanne, & Maury, 1994; van Dijk & Kintsch, 1983).

This variability in the reading comprehension performance suggests the need to device tools that offer a composite measure of reading comprehension. Indeed, Palincsar and Brown (1984) identified six different component skills, which, they claim, make up the comprehension ability, including the activation of relevant background knowledge, generation of inferences, and monitoring of both ongoing comprehension and the internal consistency of the text. More recently, Hannon and Daneman (2001) proposed a composite tool that measures the ability to access and integrate longterm-memory knowledge, to make text-based inferences and recall new text information from long-term memory. The authors used these components to predict the performance of university students on a typical test of global reading comprehension, demonstrating that their new tool was better in predicting a global measure of reading comprehension than for example a typical working memory task.

Furthermore, Oakhill et al. (2003) investigated the relevance of some abilities in accounting for reading comprehension in children between 7 and 9 years old. Oakhill et al. (2003) included in their longitudinal study tasks measuring aspects of reading comprehension that result from the individual differences approach and have been revealed to be important for a good reading comprehension, i.e. the ability to make inferences, to follow the story structure of a text and comprehension monitoring. Moreover they added an evaluation of the working memory capacity (the classic Reading Span Test devised by Daneman & Carpenter, 1980). The results in both phases of the longitudinal study revealed that a significant portion of variance in the comprehension skill is accounted for by measures of text integration, metacognitive monitoring, and working memory capacity.

In the present study, we extended this line of research exploring the components of the reading comprehension ability. The first aim of the study was to identify the structure of reading comprehension processes testing three theoretical models:

a first model in which reading comprehension could be considered as a unique factor, a second model in which two factors (basic and complex aspects of reading comprehension) could be identified and finally a model in which metacognitive, cognitive and very basic aspects of reading comprehension could be distinguishable. Such a research approach has implications for both the theoretical knowledge on reading comprehension as well as for educational practice. From the theoretical point of view, it sheds light on the factors that influence a good reading comprehension performance. Indeed, with this kind of research it is possible to highlight which factor is mainly involved during text processing, more precisely than considering good comprehension as doing well in general comprehension tasks. Moreover, understanding which aspect is better in predicting a global measure of reading comprehension is also important for knowing how to improve reading comprehension. Indeed in a training context, one needs to have information about which components of comprehension are failing (e.g., literal or inferential skills, strategy to understand different kind of texts) because it is much easier to see how such components could be trained than to see how the unanalyzed ability (performance on a comprehension test) could be trained.

If reading comprehension can be divided into two or three aspects we might expect a difference in the involvement of these components in students with different scholastic achievement. Thus, testing which component could be considered as a predictor of academic achievement is the second main aim of this research. A number of other studies have attempted to analyze the specific contribution of reading comprehension to scholastic achievement. Actually, the role of reading comprehension is pretty obvious. Indeed, it is not possible to learn something about history or geography without understanding the textbook, however what is more interesting is whether or not it is possible to highlight the role of some aspects over others. Royer, Abranovic, and Sinatra (1987), for example, stressed above all the importance of prior knowledge. Indeed in their study the course-relevant reading comprehension performance was a significant predictor of course performance but not of the grade point average (GPA) of a group of college students.

However, more recently, Taraban, Rynearson, and Kerr (2000) highlighted that college students could be differentiated on the basis of global scores of academic proficiency, a reading comprehension score or a specific course score (English) reported to differently use reading comprehension strategies. These results converge towards the importance of certain components of reading comprehension (in this case the metacognitive aspects) in predicting scholastic achievement. Taken together, these results suggest that the good reader may have more skills in controlling and monitoring the strategy necessary to elaborate a mental model.

Thus in the present study we might expect that the complex components of reading comprehension, including metacognitive abilities, may be better predictors of scholastic achievement. To summarize, this study intends to analyze the unity or distinctiveness of reading comprehension and the relationship between different components of reading comprehension and scholastic achievement. To test these hypothesis a large sample of students were administered a battery of 10 tasks measuring different aspects of reading comprehension (De Beni, Cornoldi, Carretti, & Meneghetti, 2003). The ten aspects measured are: 1. Characters, Times and Events; 2. Events and Sequences; 3. Syntactic Structure; 4. Connections between parts of the text; 5. Inferences; 6. Text Sensitivity; 7. Text Hierarchy; 8. Mental Model; 9. Text Flexibility; 10. Errors and Inconsistencies.

4. Discussion and conclusion

The current study aimed to carry out a depth analysis of reading comprehension. A battery of tasks, recently devised in Italy (De Beni et al., 2003), allowed us to measure ten different aspects of reading comprehension, thus considering a broader range of abilities compared to studies carried out so far. For example, in some tasks participants were required to identify the main and secondary characters of a text, the temporal and causal structure of a text or the ability to make conjunctions between different parts of the text. Together with these basic aspects, other tasks measured the ability to recognize relevant information within the text and detecting incongruent elements and the success in the construction of the final product of reading comprehension, i.e. a coherent mental model. In addition, some metacognitive components related to reading comprehension are taken into account. For example, the flexibility in strategy use and the ability to foresee text content are estimated.

Using this articulated battery, in the first part of the study we tested whether or not reading comprehension could be considered a unique construct. Three theoretical models were compared with confirmatory factor analysis (CFA). In the first model reading comprehension is conceived of as a unique construct. All the measures obtained from the battery were supposed to load on a single factor. The second model contemplates two latent factors, a first factor that involves aspects requiring a continuous work on and from the text and a second one in which readers mostly rely on selfregulated aspects of cognition. Finally, the third model distinguishes three factors within the ten measures of reading comprehension. In addition to a factor that considers the very basic aspects of reading comprehension, the two additional factors differentiate between metacognitive and cognitive operations on the text.

The results showed that the two-components model better accounted for the data compared to the other models tested. This finding suggests that reading comprehension, as measured by the tasks used, could not be considered as a unique construct. Actually, the distinction into two factors probably relies on the nature of the skills involved in the tasks. The skills loading into the first factor offer to a reader a basic level of the understanding of the text, probably reached even in the first phases of the development of this ability. In these phases the reader is not completely aware of all the characteristics of the reading comprehension tasks. For example, he/she could ignore that a text can contain irrelevant or wrong information, or he/she is less aware of the fact that a text can be differentiated in function of its complexity allowing a modulation of the effort spent in the understanding process. The skills representing the second factor are, thus, the representation of a higher expertise and understanding of the meaning of the reading comprehension processes. This factor, named here “Complex aspects”, refers to refined or more sophisticated aspects of reading comprehension.

The purpose of the second part of the study was to understand the relationship between the reading comprehension model that includes the ‘Basic’ and the ‘Complex’ components and scholastic achievement. In particular, we wanted to analyze the specific role of one latent factor (Basic aspects) with respect to the other (Complex aspects). To this aim, the previous model was used to predict scholastic achievement, measured with evaluations given by the teachers at the end of the academic year. The results highlighted that scholastic achievement is better predicted by the latent factor referring to the more complex aspects than by that measuring the basic aspects of reading comprehension. It is worth noticing that basic aspects of reading comprehension made a lower level contribution in explaining scholastic performance. The findings of this second part of the study confirm the reflections made when discussing the two-factor model of reading comprehension obtained. The “Complex aspects” factor with its metacognitive mark, give the reader an opportunity of reaching a more refined level of reading comprehension and seems to be more strictly linked to general measure of scholastic achievement. The results of both models can be interpreted in relation to mental model studies which evidenced that the use of the generative strategy during text comprehension increases the active transformation of knowledge for the construction of an accurate mental model (Lahtinen et al., 1997; Slotte & Lonka, 1998, 1999). In this view, ‘Basic aspects’ should structure the abilities to elaborate the text-base representation, which refers to a level strictly linked with the text structures. In fact, the ability to identify the characters in a story, as well as the succession of facts and events requires comprehension based on the structure of the text. While, the ‘Complex aspects’ structure the abilities of the mental model representation that involve active constructive processing, elaboration, or efforts to understand. The second model highlighted these more elaborate aspects as the ‘key’ components in predicting scholastic performance. These students are aware and able to monitor and use the appropriate comprehension strategies for a specific text; the efficient comprehension skills consent the building up of an accurate situation model of the scenario described and offer the elements to recall and memorize it successfully (Pressley, 2000).

The worst CFA indexes for the three componential model in comparison with the two-componential model does not necessarily signal the absence of a contribution of the metacognitive aspects during the elaboration of the mental model. The potential metacognitive aspects measured with the Text Sensitivity, Text Hierarchy, Text Flexibility, Errors and Inconsistencies tasks are included in the ‘elaborated’ aspects of the second model. These results indicate that the metacognitive aspects are not independent but included in what we called the ‘Complex’ component. These results confirm the importance of metacognition for learning. Although the fact that the aspects investigated in the five tasks, loading in the complex aspect factor, are specific for reading comprehension, they refer to the more general components of metacognitive control. The regression analysis gives further information about which are the most relevant predictors within the five aspects of scholastic achievement. The fact that text sensitivity and text flexibility are the better predictors of scholastic achievement highlights the fundamental role of metacognition in facilitating learning. It suggests that it is important to incorporate activities on text sensitivity (i.e. title identifications task or text genre identification) in the school reading curricula (van den Broek, Lynch, Naslund, Ievers-Landis, & Verduin, 2003). These skills would be likely candidates for training. In fact, the statistical verification of these models has significant educational applications; allowing for a better planning of the training required to promote reading comprehension. Thus following from the results obtained here, it appears clear that training for better reading comprehension should consider both the ‘Basic’ and ‘Complex’ aspects of this ability, since these aspects result in specific consequences to scholastic performance.

 The main purposes of this article connecting the dots of million of million research and studies, integrate prior knowledge with each other which enables the integration of structural and expression data.

In March 2012 TIME magazine reported that text mining might be "the next big thing" (BELSKY, 2012). 

Analyzing Qualitative Interview Data: The Discourse Analytic Method

Sanna Talja* University of Tampere, Finland

The article presents discourse analysis as a method of analyzing qualitative interview data. Using examples from a study of users’ library conceptions, it is argued that participants’ interpretations are much more context-dependent and variable than normally recognized, and that this has important implications for the use of interview data. Instead of producing definitive versions of participants' action or beliefs, interview data may be used to reveal regular interpretative practices through which participants construct versions of actions, cognitive processes, and other phenomena. This method does not take the individual as the principal unit of analysis, but strives to recognize cultural regularities in participants’ accounts in order to examine the phenomena studied at a macrosociological level.

In 1993, Bradley and Sutton wrote that “the serious cultivation of the potential of qualitative research has yet to emerge” (p. 405). Four years later, Vakkari (1997, p. 458) concludes that the development in library and information studies (LIS) is methodological proliferation: both quantitative and qualitative techniques are applied, although qualitative techniques have become more popular, at least for information needs and seeking research. A recent trend in qualitative research in LIS is the shift of attention from data-gathering methods to the methods of data description, analysis, and interpretation. Qualitative methods are increasingly being understood as explicitly theory-dependent ways of describing, analyzing, and interpreting data. Such approaches as discourse analysis (Frohmann, 1994), frame analysis (Chelton, 1998), phenomenography (Bruce & Klaus, 1998), conversation analysis (Solomon, 1997), and deconstructionism (Olson, 1997) require specific norms and guidelines for data analysis; they do not use a generally descriptive, hermeneutic, or naturalistic method.

In the 1980s and early 1990s, methodological discussion in the social sciences was still essentially bound up in the contrasting of qualitative and quantitative methods. While the contrastive model was “neither accurate nor particularly helpful” (Sutton 1993, p. 411), it is the reason why assessments of the merits and weaknesses of qualitative research have been almost entirely connected to the data-gathering phase, while the methods of data analysis and interpretation have often not been discussed in similar detail. The contrastive model is based on the conception that researchers have to choose between a “humanistic,” subject-centered approach aiming at capturing participants’ indigeneous meanings and experiences, or a “hard,” statistical approach describing concrete facts or society’s larger structural processes. In LIS, qualitative methods have mainly been viewed as a welcome alternative to survey-dominated user studies conducted from the information systems’ or institutional point of view (see, e.g., Dervin & Nilan, 1986).

__________________________________________

* Direct all correspondence to: Sanna Talja, Department of Information Studies, University of Tampere, P.O. Box 607, FIN-33101 Tampere, Finland.

This article introduces a method of qualitative analysis in which the basic analytic unit is the interpretative repertoire (see Gilbert & Mulkay, 1984; Potter & Wetherell, 1987; Wetherell & Potter, 1988), and attempts to show the value of a method of qualitative analysis which does not aim at capturing participants’ authentic intentions, meanings, or experiences. In discourse analysis, interview data are analyzed at a macrosociological level, as social texts. Discourse analysis is an approach which surpasses the dichotomy between subjective meanings and objective reality, as well as the dichotomy between user-centered and system-centered research (see Talja, 1997). It concentrates on the analysis of knowledge formations, which organize institutional practices and societal reality on a large scale. 

Discourse analysis is a part of the linguistic turn in the social sciences and the humanities which emphasizes the role of language in the construction of social reality. It is one of the dominant or mainstream research approaches in communication, sociology, social psychology, and psychology. Although several articles have discussed the application of discourse analysis in information studies (e.g., Budd & Raber, 1996; Frohmann, 1994; Talja, 1997; Talja et al., 1997; Tuominen & Savolainen, 1997), it has, thus far, been used relatively little as a concrete research method. This article demonstrates, with examples from a study of library users’ library talk (Talja, 1998), how the discourses existing in a particular field can be identified.

Some critics have argued that discourse analysis directs LIS researchers’ attention to “private language use.” It should be noted, however, that the term “discourse” is used within diverse research approaches, which do not necessarily have common theoretical footings. The version of discourse analysis described in this article differs in focus from the interest in interaction and mundane talk dominant in discursive social psychology and conversation analysis. As Frohmann (1994, p. 120) emphasizes, Foucaultinfluenced discourse analysis does not study the rules and conventions of mundane talk; rather, it examines “serious speech acts,” institutionalized talk or practices. This does not mean that the participants of the study should be institutionally privileged speakers. Instead, regardless of the roles and positions of the participants, talk is studied as an example of more general interpretative practices.

Discourse analysis studies practices of producing knowledge and meanings in concrete contexts and institutions - be they in library organizations, information studies, information society strategies, database interfaces, or the Dewey Decimal Classification. Discourse analysis systematizes different ways of talking in order to make visible the perspectives and starting points on the basis of which knowledge and meanings are produced in a particular historical moment. It pays attention to the way in which discourses produce and transform social reality, and makes it possible to evaluate the practical consequences of different ways of approaching a particular phenomenon.

The article introduces the basic research strategy of the analysis of interpretative repertoires: the systematic examination of context-dependent variability in talk and texts. Second, it explains how the discourses existing in a particular field can be identified. The terms “discourse” and “interpretative repertoire” are synonyms. The identification of interpretative repertoires is the endpoint of discourse analysis, and a repertoire is a named discourse. Third, the article explains how discourse analysis differs from the hermeneutic reading of qualitative data. Fourth, the criteria of validity and reliability in qualitative research are discussed. Finally, the article discusses the application of discourse analysis in LIS research. 

Context-Dependent Variation In Interview Talk

The constructivist method of interpretation used in discourse analysis problematizes some traditional approaches in qualitative analysis. Interview talk is approached with very different expectations from how we have learned, as members of culture, to interpret people’s talk in everyday life. Participants’ accounts, or verbal expressions, are not treated as descriptions of actual processes, behavior, or mental events. Interview talk is by nature a cultural and collective phenomenon. The meaning of an answer is not a straightforward matter of external or internal reference, but also depends on the local and broader discursive system in which the utterance is embedded (Wetherell & Potter, 1988, p. 169).

Qualitative analysis is often started by analyzing and counting the distribution of answers question by question. The researcher selects some sections of participants’ discourse as providing the satisfactory answers to his or her questions, whereas other parts of participants’ discourse are ignored or treated as unimportant. It is assumed that this procedure will result in a logical and coherent picture of the researched groups’ actions or views, and can be generalized to classes of social action (e.g., informationseeking behavior) and to whole groups of actors (Gilbert & Mulkay, 1984). The difficulty with taking a collection of similar statements produced by participants as literally descriptive of social action is the variability in participants’ statements about a particular topic. Not only do different actors tell different stories, but over an entire interview, it is often exceedingly difficult to reconstruct or summarize the views of one participant, because each actor has many different voices (Gilbert & Mulkay, 1984, p. 2). The following illustrative example is from an interview with a user of a public library music department. Participants were asked to give their views about the main functions and principles of music library services (see Talja, 1998).

Example One

Extract 1 Q: What do you think should be taken into account when building the library collection?

A: Real people, quite simply. It should be the starting point that library services are for the ordinary people. Although I don’t listen to mass music, to the Rolling Stones, or to Madonna, it does not mean that I can deny others such music. If the majority of people like the music, then the library should offer it. And if I want something special, I can ask for it, and they can order it as a remote loan. The library should take care of ordinary people and their needs, their cultural needs, and recreational needs. It is the starting point that the library should serve the majority of people, and it should not be turned into a money machine, like our government is aiming to do. (Interviewee 8)

Extract 2

Q: Do you think this (music) department’s collection or profile is different from the music you hear on the radio? Or, do you think the profile should be somehow different?

A: I think the library should be an alternative to mass culture. It should be an outlet, so that you can find - if you listen only to Radio City or Radio One, you can’t get a picture of the world of music as a whole. I think the library should definitely provide alternatives. The library should be - it can be said that it is - the cradle of counter culture. If it didn’t exist, people would be much poorer culturally and spiritually. (Interviewee 8)

In the first interview extract, the interviewee’s response is formulated in the context of public discussion about library fees, an issue which tends to surface especially in times of economic recession. In the second extract, the answer is equally clearly a reaction to the contextualization of the question. The interviewee’s views could naturally be summarized as “the respondent thinks that the library must serve the majority of people, but that it is also important to provide alternatives”. However, such summary solutions are problematic, because consistency is an achievement of the researcher rather than a feature of the participant’s discourse, and the context-dependent nature and cultural logic of the answers are missed. In different sections of the interview, the interviewee approaches the topic from different angles, and expresses mutually contradictory views (“if I want something special, I can ask for it as a remote loan,” and “the library should definitely provide alternatives”). In each section of talk, one view is convincingly given as the participant’s authentic, fundamental view about the topic in question, and powerful, persuasive arguments are presented in support of that view (“the library should not be turned into a money machine,” “people would be much poorer culturally and mentally”). Let us take another example.

Example 2 

 Extract 1

Q: In your opinion, what should be considered when selecting materials to the library collection? If you were a librarian, what would you emphasize?

A: My starting point would be that the responsibility of the library is not to represent - in my mind it absolutely may NOT represent - any school or any direction in particular. To make acquisitions just because we want to guide people’s tastes. It is not - it must be based on the demand. It must answer to the USER’S needs. If I were a librarian, I would register the queries. And then, of course, the ideal situation would be that the computer system would do this. And one way would be to conduct research on these users’ needs. (Interviewee 22)

 Extract 2

Q: Is there any kind of music you think is not as necessary in the library collection?

A: No, I think the collection must be balanced. But, of course, it is quite a well-founded choice now, with all these local radio stations, many of which function just as - the ether is full of rock and other kinds of light music, there is so much of that available, from so many channels, that in comparison to the general societal availability of classical music, it is not balanced at all. Therefore, that group is in a minority position when the general availability of music is considered. And, therefore, in the name of impartiality, this unbalanced situation could be balanced by investing to classical music. And, it must be considered that the collection in the library dictates to a considerable extent the general profile and atmosphere of the library. If, for instance, some library specialized in rock music, the atmosphere there would correspond to that. And, the library is, after all, an institution that is above shopping centers. Yes, I think to offer classical music, in my opinion the library could guide the person to reach for higher destinations than what he or she might have at this moment, arouse curiosity. (Interviewee 22)

In the first extract, the interviewee emphasizes that the library should not strive to cultivate users’ tastes. Collection work should be systematically based on demand or researched user needs. In the second extract, the library is an institution that encourages people to aim at higher goals and educates them by its very atmosphere and the nature of its collections.

Participants' Interpretation Work

The kind of variation and inconsistency seen in the extracts is not an exception, nor is it a product of the interview situation (for similar examples, see Gilbert & Mulkay, 1984; Machin & Carrithers, 1996; Potter & Mulkay, 1985; Potter & Wetherell, 1987). Inconsistencies can also easily be found in the answers to survey questionnaires.

However, such variation is usually managed by analytic strategies of restriction, for instance categorization, coding and selective reading, because researchers are accustomed to regarding the individual as a coherent, consistent unit, and the natural starting point for their investigations (Potter & Wetherell, 1987; Wetherell & Potter, 1988). Variability between the accounts that people give is not part of this picture. Moreover, it is customary to assume that the object of talk, for instance the library, exists as a permanent and coherent whole of which different people simply have different opinions and experiences.1 

In the discourse analytic approach, the researcher abandons the assumption that there is only one truly accurate version of participants’ action and belief. Interview talk is, by nature, interpretation work concerning the topic in question. It is reflexive, theoretical, contextual and textual, because the objects of talk (e.g., the library) are not abstract, ideal entities everyone sees in the same way. When talking about the library, participants do not only produce a neutral description and express their opinion. They produce a version of the library (“a cradle of counter culture,” or “an institution above shopping centers”), and this version contains an evaluation. As Wittgenstein (1971) has noted, in normal language each expression not only states, but also evaluates. The nature and essence of the library institution are defined in interterview talk, and speakers’ opinions concern these specific formulations, context-dependent versions of the library. Descriptions, evaluations, and large-scale cultural models of accounting are inseparably bound together (Potter & Wetherell, 1987). A third example represents a library professional’s talk: 

Example 3

· Users make the library what it is through their own presence. They also select the library’s collections, if libraries are user-centered.

· At this moment we in the library have to consider very carefully what is most important for our customers. Is it education, research, and study; is it recreation and entertainment; or is it the everyday following of events? We have come to the conclusion that the first consideration is, by far, the most important. If we gave up investing in education and study, the consequences would be serious.

· The diversity of the collection has not yet badly suffered. If the library wants to function as a forum for alternative views, it should not eliminate the alternatives. We must be able to explain to the customers why this opinion magazine comes to the library, but not every ladies’ journal. (From a librarian’s interview, see Verho, 1993, p. 131). 

First, there is the user-centered library, then the institution supporting education and study, and finally the forum for alternatives. The versions contain mutually contradictory views about users’ position and influence (“users select the library’s collections if libraries are user-centered,” and “we must be able to explain to the customers why this opinion magazine comes to the library, but not every ladies’ journal”). Again, the example shows that when the topic is approached from different angles, different aspects of it come into sight and get a privileged position. The “truth” about the library is different and is supported by different kinds of “facts” (“users make the library what it is,” “if we gave up investing in education, the consequences would be serious,” and “the diversity of the collection has not yet badly suffered”).

__________________________________________

1 This is particularly evident when considering the tradition of social surveys and opinion polls, which is based on the assumption that neither the questionnaire’s formulation and contextualization of the topic nor the persuasiveness and cultural status of particular statements in a specific moment in time and space significantly affects the answering. It is assumed that people actually hold such onedimensional and context-independent opinions as described by the conclusions the research enabled.

In discourse analysis, this kind of variability and inconsistency in explanations is not seen as a potential source of error when trying to make coherent sense of participants’ views. Interview talk is the resourceful, context-dependent application of common interpretative resources. The variability of interpretations does not mean that there is no regularity at all in participants’ discourse; it only signifies that regularity cannot be pinned at the level of the individual speaker (Wetherell & Potter, 1988, p. 172). There are considerable similarities in ways of making sense of the library as an institution (e.g., all three examples contain the “user-centered” library version and the “forum of alternatives”), but all interpretations and arguments are not equally logical and acceptable in a particular speech situation. In similar conversational contexts, similar arguments tend to be used. 2

As seen in the extracts, the inconsistencies in answers are not necessarily evident or a problem for the speakers themselves, because, in a normal conversation, the speakers are simultaneously able to retain in memory only two or three of their latest turns (Gilbert & Mulkay, 1984). If, however, two different viewpoints are taken into consideration in the same section of talk, the speakers usually show an awareness of their possible inconsistencies, and attempt to resolve them. 3 If this orientation towards the resolvement of potential contradictions does not appear when different versions of the topic are produced in different sections of talk, it is a clear sign of the existence of different interpretative repertoires (Gilbert & Mulkay, 1984; Wetherell & Potter, 1988).

__________________________________________

2 When interpreting things it is always possible to bring up several different aspects. The chosen or privileged interpretation depends on the context of talk and to whom the talk is directed. For instance, if an adolescent reader is asked to give her opinion about a Nancy Drew book, description to her best friend is likely to be different from the one offered to a teacher in a classroom setting. Both interpretations are equally authentic, because the reading experience is, in itself, seldom unambiguous. Usually, it consists of multiple, both positively and negatively experienced, impulses. In the linear process of reading, the reader and the text are in symbiosis. The reader identifies with the implicit reader of the book (Vainikkala, 1989). The spoken description of the reading experience is necessarily an afterwards rationalization, in which some of all the manifold impulses are chosen as important, and the reading experience itself receives a form and a meaning.

3 For instance, “the library is a supporter of education” -discourse, the general education repertoire, and the user-centered demand repertoire are linked together by compromise talk: the library should make compromises between demand and quality in order not to exclude any user group from library services. It is noteworthy that this emphasis does not neutralize the differences between the two repertoires. The general education repertoire assumes that “demand” mainly consists of adolescent users’ requests for light pop music, which is not to be acquired because of its short-lived character. In the demand repertoire, “demand” is defined as the interest structure of the local user community, something to be systematically researched in order for the collection work to be successful and economical (Talja, 1998).

The Identification Of Interpretative Repertoires

The analysis of interpretative repertoires is like putting together a jigsaw puzzle. Interviews are not interpreted as stories having a clear and distinguishable message and meaning; instead, all the accounts produced by the participants are taken into consideration and analyzed in order to identify significant patterns of consistency and variation in them. Thereafter, the researcher starts to ask: What is the starting point behind this account? On what kinds of limitations of perspective is this particular description based? What other statements in participants’ interview talk are based on the same perspective? The endpoint of analysis is the systematic linking of descriptions, accounts, and arguments to the viewpoint from which they were produced, and the naming of the different interpretative repertoires - usually by concepts which repeatedly occur in participants’ talk and which tend to be used when the topic is approached from a particular angle.

The discourses existing in a particular field can be discerned on the basis of the interpretative conflicts, or points of incompability, present in the texts under study (Foucault, 1972, p. 65; Parker, 1992, p. 13). The search for the pattern of repertoires includes three phases. The first phase consists of the analysis of inconsistencies and internal contradictions in the answers of one participant.

The second phase consists of the identification regular patterns in the variability of accounts: repeatedly occurring descriptions, explanations, and arguments, in different participants’ talk (Potter & Wetherell, 1987).

The third phase consists of identifying the basic assumptions and starting points (in Foucauldian language, “statements”), which underlie a particular way of talking about a phenomenon.

In discourse analysis, inconsistencies in participants’ accounts are interpreted as differences between relatively internally consistent interpretative repertoires. What is, then, an interpretative repertoire? Wetherell and Potter (1988, p. 172) define interpretative repertoires as “bounded language units” constituted out of a restrictedrange of terms used in a specific stylistic and grammatical fashion. These terms are derived from one or more key metaphors and the presence of a repertoire will be signalled by certain trophes or figures of speech (e.g., “cradle of counter culture,” and “reach for higher destinations”) (Wetherell & Potter, 1988, p. 172). Foucault (1972, p. 49), on the other hand, defines a discourse as a practice which systematically forms the objects of which it speaks. He emphasizes that discourses do not consist merely of single meanings or interpretations: they are knowledge formations, entities that provide an effective and limited lens for producing knowledge about a topic.

According to Foucault (1972), the internal coherence of a discourse is not based on: (1) the object of talk, (2) the style or manner of speech, (3) a coherent and logical system of terms, or (4) established themes. When it is possible to discern a limited viewpoint on the basis of which the objects, style, and themes of talk are selected and common concepts are defined, one can speak of a discourse (Foucault, 1972).

Interpretative repertoires cannot be “bounded language units” consisting of a restricted range of terms, since the same terms are used in different discourses, in which their meanings are constituted differently (see Foucault, 1972, pp. 34-35; 1981, pp. 100-101). Language contains only a limited number of concepts, and distinctions like “alternative,” “commercial,” or “mainstream,” are widely used in different fields and contexts. Words are many-accented: words having strong positive connotations, such as “knowledge,” “quality,” or “diversity,” receive divergent social meanings in different discourses and contexts of speaking (Volosinov, 1986). Discourses interpret common concepts (like “demand,” see footnote 3) in a way that corresponds to the viewpoint on which the discourse is based.

Thus, the different starting points of discourses are discernible from the way common concepts are understood and defined. Terms that have been linked together on the basis of a particular background assumption, loose their link on the basis of a different assumption, and are linked to other words. Discourses are also classification practices: analyzing discourses involves analyzing the selection, linkage, and ordering of terms.4 Words are articulated with other words differently in discourses, and they implicate different ideas and ideologies. As Volosinov (1986) puts it, all the minuscule and passing moments of social changes are crystallized in words.

There are simultaneously several, more or less conflicting discourses existing in a particular field of knowledge or a particular institution at a certain point in time, because novel or alternative interpretations emerge as corrections to prior discourses. These earlier discourses appear in some respects erroneous or one-sided. Changes in social experiences and possibilities slowly render historically strong discourses less valid and accurate. These discourses seem to misinterpret some of the essential features or the “true nature” of the discussed phenomenon. Established ways of conceptualizing and approaching phenomena do not, however, vanish as their validity begins to be questioned. Novel interpretations gradually become established, and alternative discourses exist side by side in the same field of knowledge. That is why discourses are internally relatively coherent, but mutually contradictory and alternative. 

Statements And Absences

According to Foucault (1972), each discourse is based on a few background assumptions, or statements, as he calls them. Statements are unspoken theories about the nature of things, and they are the necessary and implicit starting points behind a particular way of speaking about a topic. On the basis of the statements, a particular state of things is assumed (e.g., “if we gave up investing in education, the consequences would be serious”). The statements building a discourse provide a particular angle from which a topic is approached, and they limit other ways in which the topic could be constructed (Hall, 1992, p. 291).5 This limiting effect of statements leads to absences, some possible interpretations or explanations are not voiced or even thought of when the topic is approached from a particular viewpoint (Foucault, 1972). The perspective that builds a discourse involves, above all, strategic selection of meanings. Discourses differ from one another in what kinds of (factual) statements about the nature of things they legitimize, and what kinds of meanings are absent or ignored.

_________________________________________________

4 Saussure (1983) has noted that the signs of language describe values, not real and permanent “essences.” Language is basically a system of distinctions, in which the meaning of a single word depends on its difference from other words. Language consists of distinctions, but, in permanent use, a word can get cemented as one-accented, so that it seems to have only one unchangeable meaning (Volosinov, 1986). That is why it is not necessarily noticed that talk about “quality,” for instance, can be meaningful only if it has opposites: entities that are not considered to represent quality. On the other hand, “quality” must also be linked with attributes, which specify the meaning of the term by their presence. One significant point of incompatibility in library talk can be found in interpretations of what kinds of entities represent and which attributes define “quality.”

In the examples, the speakers legitimized their views by appealing to generally accepted “facts”: “the ether is full of rock and other kinds of light music,” “you can’t get a picture of music as a whole by listening only to the radio,” and “if we gave up investing in education, the consequences would be serious.” It could equally well be claimed that there are radio channels (at least in Finland) which only play classical music, that the supply of publicly-funded radio channels aims at broadness, and that the function of school and academic libraries is to support education and study. However, the facts legitimizing a particular version of reality are rarely ever simply true or false (Foucault, 1972). The variability in participants’ accounts results from the fact that even generally accepted facts or empirically supported truths, and logical, well-founded opinions can be in conflict with one another (Billig et al., 1988). The constitution, interpretation and weighing of facts always depends on what is considered as important, valuable, or desirable. Volosinov (1986) emphasizes it is always a particular viewpoint, or horizon of evaluation, that brings “the facts” into speakers’ sight.

When made visible, the statements building a discourse are always relative, susceptible to dispute and denial (Foucault, 1972). However, because statements are a part of an established and naturalized way of speaking, they are not normally, in themselves, taken under scrutiny. The utterances produced on the basis of established discourses are normally received simply as “grammatical,” that is, as logical and believable descriptions of “how things are” (Foucault, 1972).

Discourses, in the context of their rules and conceptualizations, provide the space in which things can be talked about (Hall, 1992, p. 291). On one hand, they limit the ways in which it is possible to make sense of things, but, on the other hand, they provide conflicting and variable viewpoints. As discourses provide the language for talking about a topic, for presenting knowledge and views, in a profound sense, they construct the lived reality (Hall, 1992; Wetherell & Potter, 1988). The perspectives and horizons of interpretation that give structure to social action, practices, and relationships are not normally consciously reflected, because they are part of the everyday order of things. The choices of perspectives also have practical consequences, but analyzing these consequences does not mean speculating about individual speakers’ intentions. It means exploring the connotations, allusions, and implications which particular discursive forms evoke (Foucault, 1972, 1980; Parker, 1992). Discourses are not individuals’ creations: they have taken their shape with the passage of time, they reflect the whole history of the societal form, and they have effects which no one has consciously meant.

_______________________________________________

5 For instance, the “guide for higher destinations” library metaphor is based on the statement that there are cultural products which educate, or lead to spritual growth, and which merely entertain. The “cradle of counter culture” metaphor is based on the classification of cultural products to commercial and conventional “mainstream” and noncommercial, deviant “counter culture” or “alternatives.”

Interviews As Social Texts

Discourse analysts are not interested in processes taking place either in individuals’ minds or in reality. They concentrate on the regularities of language use: what kinds of descriptions and accounts of a topic are possible, what kinds of evaluations are they based on, how do different modes of accounting construct different versions of the topic or produce different kinds of truths, and what do these versions accomplish (Wetherell & Potter, 1988). It is seen that discourses are “historical facts,” but the speakers using them, topics to which they are applied, and attitudes towards them can, and do, change continuously (Foucault, 1972; Machin & Carrithers, 1996). The basic assumption of discourse analysis is that interview answers are manufactured out of pre-existing linguistic resources with properties of their own, “much as a bridge is put together using girders, concrete and cable” (Wetherell & Potter, 1988, p 171). Historically-formed discourses are repositories of starting points, definitions and themes that position the speakers as they give meanings to phenomena (Hall, 1982; Parker, 1992). Many researchers find this idea objectionable, because it seems to render individual subjects into “opinion automats.” If people change their attitudes from moment-to-moment and from setting-to-setting, what is left to give direction to their actions? When looked at more closely, the idea that discourses produce both the objects of which they speak and the speaking subjects, is no more objectionable than Wittgenstein’s (1971) well-taken critique of private languages. For example, the interview extracts showed how the speakers employed culturally strong interpretations about the societal function of the library institution in a way that enabled them to present their views, argue for them, and defend them in an effective and convincing way. In this process, the participants allowed themselves to embody different kinds of persons, or, to put it more precisely, the subject positions offered by discourses provided them with varying identities.

In the first section of extract two, the interviewee first aligned herself against librarians wanting to guide tastes.

In the second section, the same interviewee aligned herself with librarians who could encourage people to reach for goals higher than their present aspirations. Similarly, the library professional first took distance from a library-centered perspective by saying that it is users who make the library what it is. This person then talked about “we in the library” who must make the important decisions and be able to explain them to the users.

This kind of navigation between different subject positions, or temporary identities and categories of person, strongly clashes with the traditional view that qualitative research should aim at capturing the speakers’ authentic intentions, experiences, meanings, or behavior. This aim is based on a conception of individuals holding a static set of attitudes, values, or knowledge structures across different occasions, unaffected by local conversational settings or cultural resources of interpretation.

Discourse analysis emphasizes that subjects are not as unidimensional, sovereign, and static as is commonly assumed, since in different social contexts and speech situations the individual uses variable linguistic resources and moves between different discourses quite naturally and skillfully (Potter & Wetherell, 1987).

The starting point of discourse analysis is that meanings, values, and ethical principles are not individual creations, but entities that people create together in communication and social action. This view of language, mind, meaning and self hood is dialogic, emphasizing that we are not “self-contained” (Sampson, 1993) selves, but “owe our character as the individuals we are to our living, embodied relations to the others and othernesses around us” (Shotter, 1998, p.1). As Bakhtin (1984, p. 287) puts it: “To be means to communicate... To be means to be for another, and through the other, for oneself. A person has no internal, sovereign territory, he is wholly and always on the boundary; looking inside himself, he looks into the eyes of another or with the eyes of another.”

It may seem irrelevant whether the research object is defined as the subject’s sense making or linguistic practices. Surely, talk expresses what is in an individual’s mind. In normal life people do have to act on the basis of the assumption that language straightforwardly and accurately describes both mental events and the outside reality. However, language is not just a tool to be taken up and put down at will, when we have something to communicate (Williams, 1977). Language is also an indispensable part of the subject’s self-understanding, since words are present in every act of interpretation (Volosinov, 1986).

According to Bakhtin (1981), the words of language are always half someone else’s: when subjects use words, they formulate themselves and their thoughts from the point of others, from the point of view of their community. Words become speakers’ “own” as they use them for their own purposes, include their own intentions in them (Bakhtin, 1981). Individuals cannot, however, invent new words to express their intentions. They have to use the same expressions that have been used countless times before (Volosinov, 1986). For instance, libraries are routinely defined as providers of education and recreation, even if at closer reflection people probably do not think that serious reading could not be entertaining, or that entertainment could not educate.

It is as if the words called for quotation marks to be set around them. “Language’s own talk” is, however, supreme in its power compared to individual speakers’ views, since pre-existing conceptualizations and ways of classifying phenomena have to be used even by speakers whose conscious intention it is to oppose them (Hall, 1982). Individuals are not able to modify the resources of interpretation freely, since they are limited by the episteme of a specific cultural and historical phase. However, discourses, like individual subjects, are variable, conflicting, and continuously changing and developing.

Validity And Reliability In Discourse Analytic Researcher

Alasuutari (1995) distinguishes between two different approaches in the analysis of qualitative interview data, the factist and specimen perspectives. In the factist approach, researchers want to find out about the actual behavior or attitudes of the participants. The analysis concentrates on the contents of interview answers, which reveal something about phenomena or processes occurring either in participants’ inner realities or in external reality. These phenomena or processes are the true object of study, and therefore the factist perspective makes a clear distinction between research data and the reality it gives information about (Alasuutari, 1995, p. 54). If the object of study is information-seeking behavior, for instance, it is seen that observation or self observation provide more direct and reliable research data than interviews. Interviews provide secondary, interpreted data, which can be influenced by the research situation and participants’ ability to remember past information-seeking situations. In the factist perspective, the reliability of research results depends on how unbiased and accurate information the interview answers provide about the phenomenon studied.

The criteria of validity and reliability are very different in the specimen perspective. Interview answers are analyzed as linguistic expressions, not as facts about how users think or behave. Participants’ expressions are examined from not only the point of view of their content and meaning, but also their implications and effects in constructing different versions of reality. The reliability of research results does not depend on the trustworthiness of participants’ answers, since even a speaker who lies applies cultural forms and interpretative resources which, in themselves, are neither true or false, but simply exist (Silverman, 1985). In the specimen perspective, no research data are in themselves more authentic, unbiased, or accurate description of reality. All forms of talk and texts represent situated speech which provides evidence of the various ways in which a particular phenomenon can be approached. Research data do not describe reality, they are specimens of interpretative practices.

In the specimen perspective, the validity and reliability of research results do not depend on the empirical level, the nature of research materials, or on the nature of researcher- researched interactions. The reliability of findings depends on the verifiability of the researcher’s interpretations. The interpretations must, in a consistent and identifiable way, be based on the research data. In discourse analysis, extracts from the texts studied have a different role in the research report than in the factist perspective. The texts are not descriptions of the object of research; they are the object of research. Text extracts are a necessary basis for the researcher’s argumentation in the research report, and they also provide the linguistic evidence for the researcher’s interpretations (Potter & Wetherell, 1987).

In the factist perspective, research results are usually more generalizable when the quantity of data is large. In the specimen perspective, the question of generalizability is approached from a different direction: a key concept is possibility (Per?kyl?, 1997). Social practices that are possible, that is, possibilities of language use, are the central objects of analysis. The possibility of a particular model of interpretation can be considered generalizable even when it cannot be shown how widely the model is used across different settings. Usually, there is no logical reason to doubt that a particular model of argumentation could not be used by any competent member of society. The research results are not generalizable as descriptions of how things are, but as how a phenomenon can be seen or interpreted.

The criteria of validity and reliability are very different in the specimen perspective. Interview answers are analyzed as linguistic expressions, not as facts about how users think or behave. Participants’ expressions are examined from not only the point of view of their content and meaning, but also their implications and effects in constructing different versions of reality. The reliability of research results does not depend on the trustworthiness of participants’ answers, since even a speaker who lies applies cultural forms and interpretative resources which, in themselves, are neither true or false, but simply exist (Silverman, 1985). In the specimen perspective, no research data are in themselves more authentic, unbiased, or accurate description of reality. All forms of talk and texts represent situated speech which provides evidence of the various ways in which a particular phenomenon can be approached. Research data do not describe reality, they are specimens of interpretative practices.

In the specimen perspective, the validity and reliability of research results do not depend on the empirical level, the nature of research materials, or on the nature of researcher- researched interactions. The reliability of findings depends on the verifiability of the researcher’s interpretations. The interpretations must, in a consistent and identifiable way, be based on the research data. In discourse analysis, extracts from the texts studied have a different role in the research report than in the factist perspective. The texts are not descriptions of the object of research; they are the object of research. Text extracts are a necessary basis for the researcher’s argumentation in the research report, and they also provide the linguistic evidence for the researcher’s interpretations (Potter & Wetherell, 1987).

In the factist perspective, research results are usually more generalizable when the quantity of data is large. In the specimen perspective, the question of generalizability is approached from a different direction: a key concept is possibility (Per?kyl?, 1997). Social practices that are possible, that is, possibilities of language use, are the central objects of analysis. The possibility of a particular model of interpretation can be considered generalizable even when it cannot be shown how widely the model is used across different settings. Usually, there is no logical reason to doubt that a particular model of argumentation could not be used by any competent member of society. The research results are not generalizable as descriptions of how things are, but as how a phenomenon can be seen or interpreted.

The aim of discourse analysis is to produce interpretations which are intrinsically macrosociological. This does not mean the quantity of research data should be large. Even one interview may suffice to indicate what kinds of interpretations are possible. As discourse analysis is very labor-intensive, it makes sense to start with a thorough analysis of few interviews. The model of interpretative repertoires identified is then tested against a larger set of data. In this process, the model also gets more detailed and rich, and the researcher’s understanding of the starting points and statements behind different ways of talking increases.

In the factist approach, the reliability of research findings can be increased by methodological triangulation using multiple data-gathering methods (e.g, observation, interviews, and diary techniques). In the study of interpretative practices, the reliability and generalizability of research findings can be enhanced by combining different types of research materials, interviews, and written texts; and by contextual triangulation. According to Foucault (1972), one criterion for the existence of a “discourse” is that it is used in a variety of contexts and that it can be applied in the handling of a variety of themes. If the analysis of interviews can be extended by making explicit comparisons between different settings or contexts of discussion, the research does not comprise a case study with restricted generalizability.

In the study used as an example in this article, the interpretative repertoires found in users’ library talk were compared to handbooks of music library activity and library professionals’ writings. The themes and concepts that were disputed in professional discourse were also themes of discussion in users’ talk. Next, library talk was compared to texts outlining broadcasting and radio politics, as well as