LINGUIST List 12.2220

Tue Sep 11 2001

Review: Malrieu, Evaluative Semantics

Editor for this issue: Terence Langendoen <>

What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in.

If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Simin Karimi at or Terry Langendoen at


  • Kerstin Fischer, review of Malrieu: Evaluative Semantics

    Message 1: review of Malrieu: Evaluative Semantics

    Date: Fri, 7 Sep 2001 11:59:10 +0200 (MEST)
    From: Kerstin Fischer <>
    Subject: review of Malrieu: Evaluative Semantics

    Jean-Pierre Malrieu (1999) Evaluative Semantics: Cognition, Language and Ideology. Routledge, hardback ISBN 0-415-19671-9, $110.00 (US), $165.00 (Canada), 316+xiipp, Frontiers of Cognitive Science series

    Kerstin Fischer, University of Bremen

    SUMMARY The topic of the book is the relationship between language and ideology. The main project followed in 'Evaluative Semantics' is to develop a (formal) method to estimate the consistency of a text with respect to a given ideology.

    The author, Jean-Pierre Malrieu, takes an extraordinarily broad approach. In particular, he combines insight from

    - sociology - discourse analysis - lexical semantics - formal semantics - psychoanalysis and psycholinguistics - research on affect and emotion - artificial intelligence

    For all of these areas, he attempts to provide overviews of the current state of the discussion in order to position his own approach in the whole spectrum. That is, at no point in the book does the author argue 'X is suitable for my purposes, that is why I decided for it'. Instead, he attempts at arguing 'U, V, W, Y, Z are not suitable, that is why I chose X'. This methodological attitude gives rise to the fascinating breadth of the book. However, even more compelling is that many of the approaches presented are not well-known in the English-speaking world. So the book allows the non-French speaking linguist to learn a lot about French sociological and linguistic research. However, this attitude is also responsible for a number of problems: the book is, especially in the first half in which the other approaches are discussed, extremely difficult to follow because it is often not clear what the author is aiming at. The presentations themselves often remain sketchy, and the selection of approaches presented was sometimes puzzling to me, which may be partly due to the francophone background of the author (and the anglophone background of the reviewer). Occasionally, his sources furthermore remain hidden. Thus he criticizes Althusser and Pecheux for not having properly understood Lacan's psychoanalytic theory (p.99), without providing us with a reference to Lacan's writings himself, neither in his presentation of this theory (p.61-65), nor elsewhere in the book. Finally, some side topics, such as compositionality, have been dealt with rather vaguely, which may also be due to the enormous range of topics tackled.

    Jean-Pierre Malrieu takes the following steps, which correspond to the individual chapters of the book, to present us with his considerations about evaluative semantics:

    1. The first chapter consists in a discussion of the notion of ideology. The approaches on which the discussion is based are mainly those of Foucault, Althusser and Bourdieu. The first major question Malrieu deals with is how ideology can be defined. He therefore compares internal (ideologies are systems with their own logic and rigor), external (ideologies are real (material) phenomena), and so-called 'cold' (ideologies have a subjective rationality) theories of ideology. Topics of this chapter include the problems of external theories of ideology, the scarcity of discourse argument, discursive legitimation, and the role of the content of ideologies. Malrieu argues that because content is necessary to reject particular ideologies, and because ideologies often have no practical consequences for individuals, ideological discourse can neither be defined as completely external, nor as entirely 'cold'. The second major question concerns in what ways ideologies are consistent, and he discusses proposals such as that ideological statements group together because of the practical objectives they pursue (p.44), or that generative schemes and ideological patterns produce consistent ideological discourse. In contrast to these approaches Malrieu argues that consistency is achieved through a harmony of values, objects, and actions (p.42). Finally Malrieu describes an internal account of ideological categories developed by Boltanski and Thevenot (1991) which he uses in his model outlined in chapter eight.

    2. In the second chapter, Malrieu reviews some approaches to the relationship between affect and cognition, arguing that there are three possible positions: a) evaluation belongs to the affective sphere, b)evaluation belongs to the cognitive sphere, and c) evaluation is indissolubly affective and cognitive. Topics discussed include emotional development, Edelman's model, psycholanalysis, and of the role of evaluation in text processing. Malrieu concludes that evaluation is to a large extent autonomous from cognition.

    3. Then, Malrieu discusses different approaches to discourse analysis. His main focus is on content analysis. Furthermore, he includes some notes on the contribution of Anglo-American and French discourse analysis to the study of ideological language. The question he addresses in this chapter is whether there are linguistic criteria for isolating particular linguistic properties as ideologically relevant per se. Not surprisingly he concludes that ideological relevance of particular linguistic items depends on the respective ideology.

    4. In the fourth chapter, Malrieu reviews several lexical semantic approaches, in particular, markedness theory, the theory of argumentation within language (Ducrot), cognitive semantics, differential semantics (by which he understands the structuralist tradition (Saussure, Greimas, Pottier)), and what he calls interpretative semantics (Rastier, Halliday). These theories he rejects because, in his view, they "have remained, as far as evaluation is concerned, at the lexical level" (p.147). Denotational semantic approaches, which apparently have not paid much attention to evaluation at all, are then discussed in order to arrive at an account of evaluative meaning effects at the level of phrases, sentences and discourse. Malrieu proposes that "a semantics of contextual modification, in which the default evaluation of an expression is transformed into a contextual evaluation under the influence of its linguistic context" (ibid.) is the most suitable way to account for evaluative meaning.

    5. Then, Malrieu introduces semantic networks and different discourse representation formalisms. This chapter is certainly valuable as an overview for readers not previously acquainted with such representation formalisms. Sowa's Conceptual Graph Model is described in great detail, however, only to be rejected as showing that "the complexity required to achieve logical correctness often turns out to be as dangerous as the lack of logical foundations which characterizes ad hoc methods" (p.177). The author therefore concludes to develop a formalism which uses aspects of conceptual graphs, but which dispenses with logical foundations in favour of a representation that is closer to the linguistic surface.

    6. The sixth chapter introduces this kind of formal representation of text; the graph representation Malrieu developed is called Styled Semantic Networks (SSNs), 'styled' because he uses colours and fonts to express particular aspects of the semantics of a text. The networks are very close to the linguistic surface, syntactic structure plays no role, and determiners etc. are not interpreted but just represented in the nodes of the network. The links are constituted by relations very close to thematic roles or the links in frame semantics, etc. Malrieu argues that the SSNs can be tested for semantic validity by a projection game similar to that developed by Sowa, however, it did not become clear to me whether this game is actually part of the computational model of evaluative semantics developed step by step throughout the book. Because of the closeness of the representation to natural language utterances, the author argues that inter-coder reliability, another way to test the validity of a representation, should be high (p.203).

    7. In the chapter 'Dynamic Semantic Networks', Malrieu proposes his definition of ideological consistency as "the stability of evaluations in a dynamic semantic network" (p.209) and of evaluative meaning effects as "the dynamic modification between units of discourse" (ibid.). That is, ideological consistency is estimated by measuring the distance between initial and stable states of the representation. 'Dynamic' is the calculation of inertia and default values on the basis of the three relations attraction, repulsion, and attraction-repulsion that are argued to hold between linguistic items.

    8. Chapter eight comprises the complete model of ideology with its three layers ontology, relations, and ideological knowledge. While the ontology and the relations are held to be the same for all ideologies, the last level codes the particular ideology, the opinion. The relations are described as weighted connections between, for instance, actions and entities, coding information such as "The evaluative influence of an action on its actor is stronger than the reverse influence" (p.231). The third, ideological, level is represented by a networked analysis of each ideology in terms of Boltanski and Thevenot's categories. Malrieu then adds some general semantic properties to his model, such as accounts of negation, quantification and modality.

    9. The ninth chapter presents an example analysis of Antony's funeral speech in Shakespeare's Julius Cesar. The phrases of this monologue are analyzed as SSNs, the weights to the connections are added according to the general representation of the relations developed in the previous chapter, inertia and default values are assigned and then the representation is compared to the two different representations of the two conflicting ideologies 'tyranny' and 'republic'. For each sentence the consistency with either one ideology is estimated. That is, what the model does is to judge for each sentence of a text how likely it supports a particular given ideology.

    EVALUATION The book presents very stimulating and complex discussions on the nature of ideologies and their relationship to language. While developing his argument by means of discussing previous approaches and perspectives in the above listed research domains, Malrieu also offers some interesting points on aspects which can be seen as welcome side effects of his model, for example, the treatment of opaque contexts and the account of connectives, such as 'but'. Thus, the book presents a very rich and complex picture, and it is certainly worth reading because of this complexity and breadth.

    The model developed itself is of course very restricted. The decision whether a given sentence supports a given ideology or not constitutes only a minor aspect of the much more general problem of ideological language. The author argues that solving this problem helps to identify, for instance, ironical statements, thus being useful for the automatic disambiguation of sentences. However, though aspects of the model seem to be implemented, most parts are not, and since Malrieu holds that we can decide on the 'default' values of individual (lexical) items (the initial state which is compared to the terminal state after all relations have been calculated by their weights) only in the context of a given ideology, the decision being taken by the sociologist, it seems that the model relies on many intuitive judgements taken by the researcher. There are furthermore several aspects of the model developed that remained unintelligible to me, and this is probably not due to my sparse mathematical background; it seemed to me that the author did not pay too much attention to letting the reader participate in all aspects of his model. For instance, in the SSNs, tense is represented by colour, of which the author plainly notes that this is invisible in print (p.182). In spite of all shortcomings, the application of his model to the data in chapter nine is stunning. As it often is the case in artificial intelligence research, the path that leads to a particular model may be much more interesting than the formal representation resulting. This is definitely true of Malrieu's book.

    ABOUT THE REVIEWER Kerstin Fischer teaches English Linguistics at the University of Bremen, Germany. In her PhD thesis, she developed a model of the functional polysemy of discourse particles. Currently, she works on linguistic aspects of human-computer and human-robot communication. Her research interests lie in the areas emotionality in human-computer communication, register theory, recipient design, and the role of context in the interaction with artificial communication partners.