Genre- and Register-related Discourse Features in Contrast [1 ed.] 9789027266804, 9789027242754

Citation preview

s.) r d e e f ( Le e l e e r e d u A og V e i a r Ma Svetlan and

B E N J A M I N S C U R R E N T TO P I C S

rg i s te d Re urse e - an Genr Disco ntrast d o e re l at re s i n C u Feat



Genre- and Register-related Discourse Features in Contrast

Benjamins Current Topics issn 1874-0081 Special issues of established journals tend to circulate within the orbit of the subscribers of those journals. For the Benjamins Current Topics series a number of special issues of various journals have been selected containing salient topics of research with the aim of finding new audiences for topically interesting material, bringing such material to a wider readership in book format. For an overview of all books published in this series, please see http://benjamins.com/catalog/bct

Volume 87 Genre- and Register-related Discourse Features in Contrast Edited by Marie-Aude Lefer and Svetlana Vogeleer These materials were previously published in Languages in Contrast 14:1 (2014).

Genre- and Register-related Discourse Features in Contrast Edited by

Marie-Aude Lefer Université Saint-Louis - Bruxelles, Université catholique de Louvain

Svetlana Vogeleer Université catholique de Louvain, Université Saint-Louis - Bruxelles

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

doi 10.1075/bct.87 Cataloging-in-Publication Data available from Library of Congress: lccn 2016018281 (print) / 2016026353 (e-book) isbn 978 90 272 4275 4 (Hb) isbn 978 90 272 6680 4 (e-book)

© 2016 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Company · https://benjamins.com

Table of contents Introduction Marie-Aude Lefer and Svetlana Vogeleer Using multi-dimensional analysis to explore cross-linguistic universals of register variation Douglas Biber

1

7

Cross-linguistic register studies: Theoretical and methodological considerations Stella Neumann

35

A lexical bundle approach to comparing languages: Stems in English and French Sylviane Granger

59

Discourse-structuring functions of initial adverbials in English and Norwegian news and fiction Hilde Hasselgård

73

Canonical tag questions in English, Spanish and Portuguese: A discourse-functional study María de los Ángeles Gómez González

93

Loving and hating the movies in English, German and Spanish Maite Taboada, Marta Carretero and Jennifer Hinnell

127

Index

161

Introduction Marie-Aude Lefer and Svetlana Vogeleer

In his seminal book Seeing through Multilingual Corpora, Johansson (2007: 304) convincingly argued that “it is desirable to extend contrastive studies by taking into account the variation across registers within languages”. Contrastive linguists are now increasingly aware of the need to revisit and refine the early-day corpus-based studies, which were chiefly based on literary or news corpora and often tended to consider the languages under investigation as monolithic entities, thereby overlooking the influence of different registers on cross-linguistic contrasts. The role played by register variation is also emerging in corpus-based translation studies, as illustrated by the following concluding remark in Olohan’s (2004) Introducing Corpora in Translation Studies: We need to recognize that a substantial part of translation activity in the world involves non-literary texts. If we are to study the activity of translation and translation process generally, we need to delve deeper into those genres and texts too. This will have the consequence that we will be able to make more cross-genre comparisons and study the extent to which features of translation may be influenced by genre, text type, etc. (ibid., 191).

As a result, register- and/or genre-controlled multilingual corpora have recently been compiled for a number of language pairs (e.g. Dutch Parallel Corpus for Dutch-English and Dutch-French, CroCo and GECCo for German-English; see Neumann, this volume). The recent interest that contrastive linguistics and translation studies have taken in variation as well as the development of register- and genre-controlled multilingual corpora have opened up new research paths. One of them is the cross-linguistic analysis of genre- and register-sensitive discourse features. Biber et al.’s (1998: 106) statement that “we […] know surprisingly little about discourse similarities or differences across texts and registers” still holds today, especially in contrastive and translation studies. The present volume aims to make a contribution towards filling this gap in cross-linguistic research (see also Taboada et al., 2012). doi 10.1075/bct.87.01int 2016 © John Benjamins Publishing Company

2

Marie-Aude Lefer and Svetlana Vogeleer

The six chapters included in this volume were all originally presented at the international conference on Genre- and Register-related Text and Discourse Features in Multilingual Corpora jointly organized by the Linguistic Society of Belgium and the Marie Haps School for Translators and Interpreters (now the Louvain School of Translation and Interpreting, Université catholique de Louvain) in January 2013. The present volume was originally published as a special issue of Languages in Contrast. In the following paragraphs, we introduce each contribution in turn. In his chapter, Doug Biber presents the multi-dimensional analysis (MDA) approach (cf. Biber, 1988; 1995) and offers a detailed survey of MDA studies in a range of languages (English, Czech, Korean, Portuguese, Spanish, Taiwanese, among others) and discourse domains (e.g. university spoken and written registers, job interviews, written legal registers, blogs). MDA is based on the bottom-up statistical analysis of co-occurring linguistic features, such as nouns, long words, prepositions, type/token ratio and attributive adjectives, with a view to identifying linguistic dimensions that are characteristic of a particular language or domain. Biber’s underlying assumption is that linguistic features co-occur in texts because they serve related communicative functions. In addition to highlighting a number of dimensions that are peculiar to a given language (e.g. honorification in Korean) or domain (e.g. evaluation of possible explanations in biology research articles), Biber’s survey of MDA research makes it possible to uncover two strong candidates for universal dimensions of register variation: (i) a fundamental opposition between clausal (oral) discourse and phrasal (literate) discourse (functionally, this corresponds to the distinction between a personal, involved focus and an informational focus) and (ii) the opposition between narrative and non-narrative discourse, which indicates that narration plays a central role in human communication. Biber shows that narration contrasts sharply with the other rhetorical modes (such as exposition, description, argumentation), which do not emerge as candidates for universal dimensions of register variation in the languages investigated in MDA research to date. Cross-linguistic register studies are also at the heart of Stella Neumann’s contribution, which addresses the methodological and theoretical issues that arise when conducting register-oriented contrastive and translation studies. Drawing on a top-down, systemic functional register theory, Neumann distinguishes between three different approaches to cross-linguistic investigations involving register: (i) studies that control the corpus used for register (i.e. the corpus under investigation contains only texts from a single register), thereby acknowledging register as a potential confounding factor in cross-linguistic research, (ii) studies that contrast given linguistic features in several registers, and (iii) studies that take register as their main object of research. In the third approach, the linguistic features analysed are used to characterize registers contrastively (across languages



Introduction

or across language varieties, e.g. original vs. translated language). Approaches (ii) and (iii) are illustrated with the help of English-German case studies based on the CroCo corpus (Hansen-Schirra et al., 2012), which is made up of original and translated texts from eight registers. In these sample analyses, Neumann relies on a new weighting procedure where register frequencies for a particular linguistic feature are seen relative to the frequency of the feature in a register-neutral reference corpus (so-called “relative register values”; cf. Neumann, 2013), which allows for a direct cross-linguistic comparison of the results. The following three contributions can be said to belong to Neumann’s second approach: they all examine a particular linguistic feature (viz. lexical bundles, clause-initial adverbials and tag questions) across different genres or registers, with the aim of disentangling language-specific and genre- or register-sensitive discourse features. They all compare original, i.e. non-translated, languages. Situated in the field of contrastive phraseology, Sylviane Granger’s chapter focuses on lexical bundles, i.e. “sequences of word forms that commonly go together in natural discourse” (Biber et al., 1999: 990ff), such as ‘this is why’, ‘with this in mind’, ‘I don’t want to’, ‘it is possible to’. Two languages (English and French) and two genres (newspaper editorials and parliamentary debates) are examined, with a particular focus on metadiscursive bundles (discourse organizers and stance markers), and especially stems (sequences that contain a subject and a verb; cf. Altenberg, 1998). The contrastive study of lexical bundles poses a number of methodological challenges, namely the selection of bundle lengths and minimum frequency thresholds that ensure the cross-linguistic comparability of the data. Granger advocates including a wide range of bundle lengths (e.g. 3- to 7-word sequences) and selecting a proportionally similar number of bundle types for each bundle length in the languages and genres investigated. The corpus-based quantitative and qualitative investigations reveal clear cross-genre and cross-linguistic variation patterns. For example, in newspaper editorials, French makes use of a higher number of bundles in general, and stems in particular, than English. However, this cross-linguistic trend is not confirmed in parliamentary debates, where English relies more heavily on stems than French. As regards cross-genre variation, the data suggest that parliamentary debates in both languages display an overall higher degree of formulaicity than editorials and that they make more use of stems containing 1st person pronouns (e.g. ‘I have no doubt that’, ‘we have to make sure that’). Editorials, on the other hand, are found to favour impersonal and existential structures (e.g. ‘it may well be that’, ‘there is no point in’). These findings show that the lexical bundle approach is a powerful tool to investigate pragmatics and rhetoric across languages and genres.

3

4

Marie-Aude Lefer and Svetlana Vogeleer

Hilde Hasselgård’s study looks into the discourse-structuring functions of clause-initial adverbial adjuncts in English and Norwegian. Two genres are compared: news reportage and contemporary fiction. Based on Hasselgård’s (2010) classification framework of adverbials, the study reveals that initial adverbials (and adverbials in general) are more frequent in Norwegian than in English (this trend is especially salient in fiction). Also, initial adjuncts in Norwegian are found to be less likely to convey new information than in English. These contrastive findings suggest that initial adjuncts in Norwegian are less marked than their English counterparts. In addition, the genre comparison brings to light that discourse linking by means of initial adjuncts is more common in fiction than in news: while in fiction initial adjuncts tend to signal cohesive relations, in news they are typically used to put less important information in the background or to frame/set the scene for what follows. María de los Ángeles Gómez González’s study accounts for the use of canonical tag questions across three languages (English, Spanish and Portuguese) in three contrasted pairs of spoken genres/registers (dialogic vs. monologic, informal vs. formal, private vs. public). Her quantitative analysis deals with four parameters: frequencies, formal features (especially polarity), distribution across spoken genres, and discursive functions. The analysis along each parameter yields a number of somewhat unexpected results. For example, it shows that while tag questions are far more frequent in Portuguese than in English (with Spanish occupying a middle position), English displays the highest proportion of tags in sentencemedial position, where they have a focusing function. The study also shows that in English tag questions are used almost exclusively in dialogic, informal, private discourse. By contrast, in Spanish and especially in Portuguese, they are also common in monologic public discourse. The quantitative functional analysis leads to the conclusion that in Spanish and Portuguese tag questions perform a greater variety of discursive functions (e.g. information structuring) than in English. The last chapter, situated in the field of opinion mining, is by Maite Taboada, Marta Carretero and Jennifer Hinnell. Their quantitative research focuses on two Appraisal categories, namely Attitude and Graduation, in non-professional online movie reviews across three languages (English, German and Spanish). As highlighted by the authors, one of the most difficult issues in opinion mining is to decide what and how to annotate. The present article raises a number of valuable insights into these questions insofar as the criteria applied could be applied in other studies in the field. The Attitude analysis is based on the polarity of the reviews (positive vs. negative) and accounts for three Attitude types (Affect, Judgement and Appreciation). As for the Graduation analysis, it accounts for intensifiers and downtoners, both related to either Force or Quantity. The results of the cross-linguistic analysis show that some of the differences between the languages under



Introduction

discussion are due to sociocultural factors. For example, the German dataset is characterized by a high percentage of Appreciation (expressions related to aesthetic qualities: ‘remarquable’, ‘innovative’, etc.) and a low percentage of Affect (the speaker’s or somebody else’s emotional response, e.g. happiness, fear), while Spanish positive reviews are characterized by the highest percentage of Affect. The English negative reviews typically follow the ‘positive first’ pattern, that is, giving some positive evaluation before expressing an overall negative attitude. This pattern is not found in the other two languages, which can only be explained by means of sociocultural factors. By contrast, some other cross-linguistic differences are found to be truly linguistically-driven. For instance, Graduation (intensifiers and downtoners) in German appears to be more nuanced than in English and Spanish, the reason being that German has many discourse particles to express Graduation in a more fine-grained way. This study sheds new light on dissimilarities in how Appraisal is expressed in different languages, and offers a model of how to objectivize and delimit the sociocultural and linguistic grounds of these dissimilarities.

References Altenberg, B. 1998. “On the Phraseology of Spoken English”. In Phraseology: Theory, Analysis, and Applications, A.P. Cowie (ed), 101–122. Oxford: Oxford University Press. Biber, D. 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511621024. Biber, D. 1995. Dimensions of Register Variation: a Cross-linguistic Perspective. Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511519871. Biber, D., Conrad, S. and Reppen, R. 1998. Corpus Linguistics: Investigating Language Structure and Use. Cambridge: Cambridge University Press. Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. 1999. Longman Grammar of Spoken and Written English. Harlow: Pearson Education Ltd.  doi: 10.1017/CBO9780511804489. Hansen-Schirra, S., Neumann, S. and Steiner, E. 2012. Cross-linguistic Corpora for the Study of Translations – Insights from the Language Pair English-German. Berlin: de Gruyter Mouton. Hasselgård, H. 2010. Adjunct Adverbials in English. Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511676253. Johansson, S. 2007. Seeing through Multilingual Corpora. On the Use of Corpora in Contrastive Studies. Amsterdam: John Benjamins. Neumann, S. 2013. Contrastive Register Variation. A Quantitative Approach to the Comparison of English and German. Berlin: de Gruyter Mouton. Olohan, M. 2004. Introducing Corpora in Translation Studies. London: Routledge. Taboada, M., Doval, S. and González, E. (eds). 2012. Contrastive Discourse Analysis: Functional and Corpus Perspectives. Linguistics and the Human Sciences 6(1–3).

5

Using multi-dimensional analysis to explore cross-linguistic universals of register variation Douglas Biber

Northern Arizona University, USA

Multi-dimensional analyses have been conducted for many different discourse domains and many different languages. Using bottom-up statistical analyses, these studies have investigated specific patterns of register variation in several different discourse domains of English, as well as the more general patterns of register variation in many different languages. Each study identifies linguistic dimensions that are peculiar to that particular language/domain. However, the more theoretically interesting finding is that linguistically similar dimensions emerge in nearly all of these studies. Two of these dimensions are especially robust, making them strong candidates for universal dimensions of register variation: (1) a fundamental opposition between clausal/‘oral’ discourse vs. phrasal/‘literate’ discourse, and (2) the opposition between narrative vs. nonnarrative discourse. The present paper introduces the methodology of multidimensional analysis and surveys the research studies carried out to date, with an emphasis on these potentially universal patterns of register variation. Keywords: register variation, multi-dimensional analysis, oral discourse, literate discourse, cross-linguistic universals

1. Introduction One major focus of previous corpus-based research has been to describe the ways in which linguistic features vary across registers. Such research can be carried out with a cross-linguistic focus, investigating whether registers vary along similar linguistic parameters, and differ to similar extents, across languages. However, there are two major methodological challenges that must be addressed when attempting to conduct quantitative corpus-based research from a cross-linguistic perspective: – What registers should be compared? – What linguistic features should be compared? doi 10.1075/bct.87.02bib 2016 © John Benjamins Publishing Company

8

Douglas Biber

Regarding the first challenge, there are two major considerations: some registers are simply not found in some languages, and the ‘same’ register might not actually be the same across languages. For example, there are oral cultures that lack written registers altogether. At the same time, such cultures might have registers like ‘oral histories’ that are not well developed in more literate cultures. Comparisons of the ‘same’ register can also be problematic. Even within a culture, there are important differences among sub-registers. For example, we would not want to conduct a historical investigation of academic research writing in English by comparing history books from 1850 to biochemistry journal articles from 2000; any historical differences would be confounded with the linguistic differences among academic disciplines. In fact, the ‘same’ register can change historically. Thus, science research writing in the 18th century was much more personal and narrative than science research writing at the end of the 20th century (see e.g. Atkinson, 1999). Comparisons of the ‘same’ register across cultures confront the same issues, but they are even more complicated, because cultures make different distinctions among sub-registers. For example, Biber and Conrad (2009: 34–35) discuss differences among specific types of public speeches in a Samoan fono (e.g. laauga vs. tali; see Duranti, 1994), and among different types of sermons in Somali (wacdi vs. tafsiir). Determining whether linguistic features are comparable across languages is at least as challenging as the register comparisons. One obvious problem is that some linguistic features are restricted to particular languages. For example, some native American languages have grammatical markers for evidentiality; some Asian languages have honorifics; some African languages have particles to mark topic or focus. A more pervasive problem is that supposedly equivalent linguistic features do not serve the same range of syntactic functions across languages. For example, Somali has case particles (u, ka, ku, la) that occur in pre-verbal position; these particles are translated as prepositions in English (e.g. ‘in’, ‘on’, ‘for’, ‘with’). However, the Somali particles are used only to identify noun phrases with an adverbial relationship to the verb, and thus they are not at all equivalent to English prepositions, which can also be used for nominal modification. The lack of direct one-to-one structural equivalences across languages can lead to misleading quantitative comparisons.1 For example, Biber (1995: 76) shows that finite relative clause constructions are almost ten times more common in Somali than in English. This finding is difficult to interpret until we consider the role of 1.  An additional problem is that words are not comparable across languages. Thus, a single word in a highly agglutinative language like Turkish or Finnish corresponds to many words in isolating languages like English or Chinese. This difference affects quantitative comparisons, which norm counts to a rate per a fixed amount of text (e.g. rate per 1,000 words), because the basis for this norming is not comparable (see Biber, 1995: 63–66).



Using MD analysis to explore cross-linguistic universals of register variation

relative clauses in relation to the entire grammatical systems of the two languages. Relative clauses are extremely common in Somali because they are essentially the only structural device for noun modification in that language. (Attributive adjectives also occur in Somali, but they can be regarded as a special type of relative clause.) In contrast, English has many different clausal and phrasal devices used for noun modification, including non-finite postnominal clauses, attributive adjectives, pre-modifying nouns, post-modifying prepositional phrases, and appositive noun phrases. When this full inventory of structural devices is taken into account, nominal modifiers are actually more frequent in English registers than in Somali registers. The preceding paragraphs have briefly surveyed a range of difficulties arising in the attempt to conduct cross-linguistic investigations of register variation. These include problems in determining which registers are comparable across languages/cultures, and problems in determining which linguistic characteristics are comparable across languages. It would be possible to address each of these problems, proposing specific methods for dealing with the issues. However, that is not the goal of the present paper. Rather, the goal here is to introduce a completely different approach — multi-dimensional (MD) analysis — which bypasses most of these difficulties by analysing each language on its own terms. This approach can then be used for cross-linguistic analysis by comparing the patterns of register variation across languages: linguistically (the composition of each dimension), functionally (the functional correlates of each dimension), and situationally (the register distributions along each dimension). MD analyses have been conducted for many different discourse domains and many different languages. Using bottom-up statistical analyses, these studies have investigated specific patterns of register variation in several different discourse domains of English, as well as the more general patterns of register variation in many different languages. Each study identifies linguistic dimensions that are peculiar to that particular language/domain. However, the more theoretically interesting finding is that linguistically similar dimensions emerge in nearly all of these studies. Two of these dimensions are especially robust, making them strong candidates for universal dimensions of register variation: 1) a fundamental opposition between clausal/‘oral’ discourse vs. phrasal/‘literate’ discourse, and 2) the opposition between narrative vs. non-narrative discourse. In the sections below, I first introduce the methodology of multi-dimensional analysis, and then survey the research studies carried out to date, with an emphasis on these potentially universal patterns of register variation.

9

10

Douglas Biber

2. Overview of methods for MD analysis The multi-dimensional (MD) analytical approach was originally developed to investigate the linguistic patterns of variation among spoken and written registers (see e.g. Biber, 1986; 1988; 1995). Studies in this research tradition have used large corpora of naturally-occurring texts to represent the range of spoken and written registers in a language. These registers are compared with respect to ‘dimensions’ of variation (identified through a statistical factor analysis), comprising constellations of linguistic features that typically co-occur in texts. Each dimension is distinctive in three respects: – It is defined by a distinct set of co-occurring linguistic features; – It is associated with particular communicative functions; – There are different patterns of register variation associated with each dimension. The MD approach uses statistical factor analysis to reduce a large number of linguistic variables to a few basic parameters of linguistic variation: the ‘dimensions’. In MD analyses, the distribution of individual linguistic features is analysed in a corpus of texts. Factor analysis is then used to identify the systematic co-occurrence patterns among those linguistic features — the ‘dimensions’ — and then texts and registers are compared along each dimension. Each dimension comprises a group of linguistic features that usually co-occur in texts (e.g. nouns, attributive adjectives, prepositional phrases); these co-occurrence patterns are identified statistically using factor analysis. The dimensions are then interpreted to assess their underlying functional associations. Critics have sometimes misunderstood the empirical basis of these dimensions, claiming incorrectly that MD analysis begins with the functional interpretations; that the analyst groups linguistic features together based on his/her preconceived notions about those functions; and that the entire procedure is therefore circular. These criticisms reflect a fundamental misunderstanding of MD methodology. First, the dimensions represent the groupings of linguistic features that statistically co-occur in texts — not clusters of features that the analyst groups together because he/she thought they all served the same function. The order of analysis is exactly the opposite of that claimed by these critics: the empirical/statistical analysis occurs first, to identify the statistical co-occurrence patterns; the functional interpretation follows, to try to offer an explanation for why that particular set of linguistic features tends to co-occur in texts. There is no circularity here. The analysis shows empirically that these sets of linguistic features tend to co-occur in texts, regardless of the functional interpretations.



Using MD analysis to explore cross-linguistic universals of register variation

Early MD analyses investigated the relations among general spoken and written registers in English, based on analysis of the LOB Corpus (15 written registers) and the London-Lund Corpus (6 spoken registers). 67 different linguistic features were analysed computationally in each text of the corpus. Then, the co-occurrence patterns among those linguistic features were analysed using factor analysis, identifying the underlying parameters of variation: the factors or ‘dimensions’. In the 1988 MD analysis, the 67 linguistic features were reduced to 7 underlying dimensions. (The technical details of the factor analysis are given in Biber (1988, Chapters 4–5) — see also Biber (1995, Chapter 5).

Table 1.  Summary of the major linguistic features co-occurring on Dimensions 1–5 from the 1988 MD analysis of register variation. Dimension 1 Involved vs. Informational Production

Positive features mental (private) verbs, ‘that’ complementizer deletion, contractions, present tense verbs, ‘wh’-questions, 1st and 2nd person pronouns, pronoun ‘it’, indefinite pronouns, ‘do’ as pro-verb, demonstrative pronouns, emphatics, hedges, amplifiers, discourse particles, causative subordination, sentence relatives, WH-clauses Negative features nouns, long words, prepositions, type/token ratio, attributive adjectives

Dimension 2 Narrative vs. Non-narrative Discourse

Positive features past tense verbs, 3rd person pronouns, perfect aspect verbs, communication verbs Negative features present tense verbs, attributive adjectives

Dimension 3 Situation-dependent vs. Elaborated Reference

Positive features time adverbials, place adverbials, other adverbs

Dimension 4 Overt Expression of Argumentation

Positive features prediction modals, necessity modals, possibility modals, suasive verbs, conditional subordination, split auxiliaries

Dimension 5 Abstract/Impersonal Style

Positive features conjuncts, agentless passives, ‘by’-passives, past participial adverbial clauses, past participial postnominal clauses, other adverbial subordinators

Negative features WH-relative clauses (+subject gaps, object gaps), phrasal coordination, nominalizations

11

12

Douglas Biber

After the statistical analysis is completed, dimensions are interpreted functionally, based on the assumption that linguistic co-occurrence reflects underlying communicative functions. That is, linguistic features occur together in texts because they serve related communicative functions. Table 1 summarizes the factor analysis from the first major MD analysis of English spoken and written registers (Biber, 1988), including both a list of the most important linguistic features comprising each dimension and the functional label assigned to each one. Each dimension can have positive and negative features. Rather than reflecting importance, positive and negative signs identify two groupings of features that occur in a complementary pattern as part of the same dimension. That is, when the positive features occur together frequently in a text, the negative features are markedly less frequent in that text, and vice versa. On Dimension 1, the interpretation of the negative features is relatively straightforward. Nouns, word length, prepositional phrases, type/token ratio, and attributive adjectives all reflect an informational focus, a careful integration of information in a text, and precise lexical choice. Text Sample 1 provides a short example of these linguistic characteristics co-occurring in an academic article. (1) Text sample 1: Technical academic prose The need to study early development in the network of early family relationships, rather than just in the context of the mother-child dyad, has been increasingly recognized (Belsky 1984; Parke and Buriel 1998). The literature on children’s early cooperation with both parents, however, remains very limited. Volling, McEl-wain, Notaro, and Herrera (2002) found that the predictors of one form of toddler cooperation, committed compliance, differed from other child- and father-child dyads. In particular, little is known about cross-relationship effects in socialization.

This text sample is typical of written expository prose in its dense integration of information: frequent nouns and long words, with most nouns being modified by an attributive adjective, a pre-modifying noun, or a prepositional phrase (e.g. “The literature on children’s early cooperation with both parents”; “the predictors of one form of toddler cooperation”; “cross-relationship effects in socialization”). The set of positive features on Dimension 1 is more complex, although all of these features have been associated with interpersonal interaction, a focus on personal stance, and real-time production circumstances. For example, 1st and 2nd person pronouns, WH-questions, emphatics, amplifiers, and sentence relatives can all be interpreted as reflecting interpersonal interaction and the involved expression of personal stance (feelings and attitudes). Other positive features are associated with the constraints of real-time production, resulting in a reduced surface form, a generalized or uncertain presentation of information, and a generally



Using MD analysis to explore cross-linguistic universals of register variation

fragmented production of text; these include ‘that’-deletions, contractions, proverb ‘do’, the pronominal forms, and final (stranded) prepositions. Text sample 2 illustrates the use of positive Dimension 1 features in a workplace conversation: Text sample 2: Conversation at a reception at work Sabrina: I’m dying of thirst. Suzanna: Mm, hmm. Do you need some M & Ms? Sabrina: Desperately. Ooh, thank you. Ooh, you’re so generous. Suzanna: Hey I try. Sabrina: Let me have my Snapple first. Is that cold — cold ? Suzanna: I don’t know but there should be ice on uh, . Sabrina: I don’t want to seem like I don’t want to work and I don’t want to seem like a stuffed shirt or whatever but I think this is really boring. Suzanna: I know. Sabrina: I would like to leave here as early as possible today, go to our rooms and pick up this thing at eight o’clock in the morning. Suzanna: Mm, hmm. (2)

Overall, Factor 1 represents a dimension marking interactional, stance-focused, and generalized content (the positive features on Table 1) vs. high informational density and precise word choice (the negative features). Two separate communicative parameters seem to be represented here: the primary purpose of the writer/speaker (involved vs. informational), and the production circumstances (those restricted by real-time constraints vs. those enabling careful editing possibilities). Reflecting both of these parameters, the interpretive label ‘Involved vs. Informational Production’ was proposed for the dimension underlying this factor. A second major step in interpreting a dimension is to consider the similarities and differences among registers with respect to the set of co-occurring linguistic features. To achieve this, dimension scores are computed for each text, by summing the individual scores of the features that co-occur on a dimension (see Biber, 1988: 93–97). For example, the Dimension 1 score for each text is computed by adding together the frequencies of private verbs, ‘that’-deletions, contractions, present tense verbs, etc. — the features with positive loadings (from Table 1) — and then subtracting the frequencies of nouns, word length, prepositions, etc. — the features with negative loadings. Once a dimension score is computed for each text, the mean dimension score for each register can be computed. Plots of these mean dimension scores allow linguistic characterization of any given register, comparison of the relations between any two registers, and a fuller functional interpretation of the underlying dimension.

13

14

Douglas Biber      INVOLVED INVOLVED

| | TELEPHONE CONVERSATIONS | 35 + FACE-TO-FACE CONVERSATIONS | | | 30 + | | | 25 + | | | 20 + Personal letters | PUBLIC CONVERSATIONS, SPONTANEOUS SPEECHES | INTERVIEWS | 15 + | | | 10 + | | | 5 + | Romance fiction | PREPARED SPEECHES | 0 + Mystery and adventure fiction | General fiction | Professional letters | BROADCASTS -5 + | Science fiction | Religion | Humor -10 + Popular lore, editorials, hobbies | | Biographies | Press reviews -15 + Academic prose, Press reportage | | Official documents | INFORMATIONAL

Figure 1. Figure Mean1.scores registers alongalong Dimension 1: Involved vs. vs. Informational Mean of scores of registers Dimension 1: Involved Informational Production Production Written registers are in italics; spoken registers are in CAPS Written registers are in p italics; spoken in CAPS 84.3%)are [adapted from Figure 7.1 in Biber 1988]. (F = 111.9, < .0001, r2 =registers 1 (F = 111.9, p  “well, now we have said everything we want to about that. And I say you were jealous of me, weren’t you?”



As in English, SCTQs may also show reversed or constant polarity patterns depending on whether or not the negative or affirmative polarity expressed in anchors and tags are kept constant or are reversed, as illustrated in examples 9 and 10, respectively. (9) a. b.

No creo, ¿no? Declarative anchor: Negative — Negative Polarity (−/−) “I don’t think so, do you?” Eh, Juan, ¿me estás escuchando, verdad?2 Interrogative anchor: Affirmative — Affirmative Polarity (+/+) “Eh, Juan, are you listening to me, are you?”

(10) a. a′. b.

Échate más, ¿no? Imperative anchor: Affirmative — Negative Polarity (+/−) “Help yourself to more, will you?”3 ¡Hombre, eso está claro! ¿no? Exclamative anchor: Affirmative — Negative Polarity (+/−) “Man, that is clear, isn’t it?” No va a dejar de ir a Milán, ¿verdad? Declarative anchor: Negative — Affirmative Polarity (−/+) “She is not going to stop going to Milan, is she?”

The Spanish counterparts of non-canonical English TQs, that is those consisting of anchors and such invariant tags as e.g. ¿eh? “eh?”, ¿de acuerdo? “right?”, ¿verdad que no? “isn’t that so?”, ¿qué? “what?”, ¿sí? “yes?”, as well as disjunctive tags (¿o no? “or not?”, ¿o qué? “or what?”) fall outside the scope of this analysis; three illustrations of these are presented in example 11. 2.  From just the written form, example 9b could just as well be a declarative anchor. What would decide the case is intonation, i.e. whether the anchor has rising intonation or not. 3.  Alternatively, the QT in example 10a could also be translated as “won’t you?”. English has a meaningful choice here, whereas Spanish doesn’t.

100 María de los Ángeles Gómez González

JAV: Un bar, un bar, te cuesta treinta, o trein, veinte o treinta millones. IDO: (Son) Veinte millones, ¿eh? “JAV: A bar, a bar costs you thirty or thir, twenty or thirty million. IDO: It’s twenty million, eh?” Le habrán tocado después, pero cuando era pequeña no… ¿verdad que no?

“She must have had them later on, but when she was little she didn’t… isn’t that so?” c. Te compras la gafa, te vas a una óptica… la compró, ¿o no? “You buy the glasses, you go to an optician … She bought them, or didn’t she?”

(11) a. b.

Turning to the Portuguese counterparts of ECTQs (PCTQs), our attention is centred on two constructional patterns: – TQs that have grammatically-dependent tags with the structure não SER?/ SER? “not BE”/“BE”, such as those in example 12, which unlike ECTQs only consist of the auxiliary verb SER agreeing in (primary) tense and person with the verb of the anchor and accordingly constitute a variant of grammaticallydependent tags; – TQs that have invariant question tags with either pois não? “so not” or não?/ não é “not”/“not is”, which, as shown in example 13, are rendered as grammatically dependent tags in ECTQs and would correspond to tag constructions with ¿verdad? and ¿no? in Spanish: (12) a. O que é isso do gene, ah, que por vezes, pode, condicionar toda uma nossa vida, não é? (CORALR pmedsc01) Interrogative anchor: Affirmative — Negative Polarity (+/+) “What is it that gene, ah, that sometimes can condition our life, can’t it?” b. Disse-me que também esteve em, Joanesburgo, foi? (CORALR pubdl07) Declarative anchor: Affirmative — Affirmative Polarity (+/+) “You say you’ve also been to Johannesburg, have you?” (13) a. b.

E, não há dúvidas, pois não? (CORALR pfammn01) Declarative anchor: Negative — Negative Polarity (−/−) “And / there are no doubts, are there?” As criadas diziam vossa excelência. E agora já não , não é? (CORALR pfamcv03) Declarative anchor: Negative — Negative Polarity (−/−) “Maids used to say your highness, and now they don’t, do they?”

Again, the Portuguese counterparts of disjunctive and non-canonical English TQs have been discarded from the present analysis: the former have disjunctive tags



Canonical tag questions in English, Spanish and Portuguese 101

with the pattern disjunctor + não “not” (+ V)/disjunctor + V, represented in example 14, whereas the latter consist of anchors and such invariant tags as correcto? “right?”, não já? “yeah?”, hã? “see?”, hum?, eh?, okay?, as well as such fixed tag expressions as não é verdade? (“isn’t that true?”), não te parece? (“don’t you think (so)?”), or não achas? (“wouldn’t you say?”), samples of which are provided in example 15. Similarly, Portuguese TQs that have lexically dependent variant tags have been excluded from this study because, as already noted, their tags are not grammatically dependent on their anchors like those of ECTQs, but rather they involve the repetition of the lexical verb and usually have the pattern não “not” + V (≠ SER “BE”)? / V?, as shown in example 16. (14) E isso é cá em Lisboa, ou não? (CORALR pfamdl16) “And that is here in Lisbon, or isn’t it?” (15) a. b.

Eles precisam, de uma boa imagem, não é verdade? (CORALR pfamdl23) “They need (to make) a good impression, isn’t that right?” Suponho que já toda a gente fez esta parte, não já? (CORALR pnatte02) “I suppose everybody has done that part already, right?”

(16) Mas uma pessoa pode estar apresentável sem gravata, não pode? (CORALR pfamcv02) “But a person may look presentable without a tie, mightn’t they?”

To close this section, let us now focus on the position of tags in CTQs. In the examples provided so far we have seen that question tags are appended to their anchor clauses, and accordingly they are produced in two stages only, i.e. anchor presentation + tag, which are generally packed into two different intonation units, although there may be more. However, there are also instances of ECTQs, SCTQs and PCTQs in which the tag is parenthetically inserted within the anchor clauses, as reproduced in examples 17–19. (17) I think it actually goes through Brussels doesn’t it? the motorway (18) Pues bueno, lo primero tu nombre, ¿verdad? era Enrique “So well first your name was it? it was Enrique.”4

4.  The English translation here preserves the word order of the Spanish example and for that reason it doesn’t sound at all normal for English — and again there is the problem of deciding whether a positive or a negative tag is more appropriate.

102 María de los Ángeles Gómez González

(19) Mas foste tu que escolheste não foste? a zona. “But it was you who chose wasn’t it? the zone.”5

In examples 17–19, as part of the anchor follows the tag, medial question tags are realized in three stages here dubbed ‘pre-tag anchor presentation’ + ‘tag’ + ‘posttag anchor resolution’, which are normally mapped onto three intonation groups or more. Besides final and medial position, tags can frequently occupy yet another location in the three languages and that is between conjoined clauses in complex paratactic or hypotactic sentences, as shown in examples 20 and 21, extracted from the English and Spanish sub-corpora, respectively. These have been classified as instances of final CTQs on the grounds that they are placed at the end of either the first or the second conjoin.6 (20) Uh we’d all made our separate decisions hadn’t we? and obviously had similar problems in deciding how we were going to go about this (21) Se ha convertido en una ideología más desdibujada, ¿no? y está amenazada por las críticas de otras ideologías “It’s turned into a rather blurred ideology, hasn’t it? and it is threatened by the criticisms of other ideologies”.

2. The data and method The corpora analysed are, for English, the spoken component of the International Corpus of English-Great Britain (ICE-GB) (Nelson et al., 2002) consisting of 637,562 words and 300 texts, in addition to two sub-corpora of the Integrated Reference Corpora for Spoken Romance Languages (C-ORAL-ROM) for Peninsular Spanish (333,482 words and 210 texts) and for European Portuguese (317,916 words and 152 texts), which amount to a total of 651,398 words and 362 texts (Cresti and Moneglia, 2005). Tables 17 and 2 display the structure, text types and codes of both corpora conforming a sample of 1,288,960 words and 662 texts. The data were automatically retrieved using the search utility programs available in each corpus, namely ICE-CUP III (ICE-GB) and Contextes (C-ORAL-ROM), 5.  The literal translation of this example would be “But it was you who chose weren’t you? the zone”. 6.  For a discussion of two-staged and three-staged CTQs in English and Spanish, see Gómez González (2012). 7.  In Table 1 the number of words (600,000) is presented as an approximation.

Canonical tag questions in English, Spanish and Portuguese 103



Table 1.  Size and structure of ICE-GB. -

-

which allowed us to run tagged searches for ECTQs and lexical checks of the (string of) words contained in the tags of both SCTQs and PCTQs. The tokens retrieved were then examined and pruned taking into account the criteria for inclusion and exclusion explained so far. Finally, possible interactions among variables were statistically investigated using (Pearson) chi square tests as implemented by SPSS, which tests interactions between just two nominal (categorical) variables (Field, 2009: 702–721). 3. Discussion of findings This section offers a mainly quantitative analysis of the ECTQs, SCTQs and PCTQs found in our corpus, focusing on four parameters of contrast and comparison: (i) frequencies (Section 3.1), (ii) formal features (Section 3.2), (iii) distribution across genres (Section 3.3), and (iv) functional features (Section 3.4). 3.1 Frequencies This investigation is based on the analysis of 2,473 tag questions, of which 718 are ECTQs, 1,020 correspond to PCTQs, and the remaining 735 tokens are SCTQs. Figure 2 gives the relative frequency of the three data sets per 10,000 words, henceforth pttw. As appears from Figure 2, ECTQs (11.26 pttw) are three times less frequent than their analogous constructions in Portuguese (32.08 pttw) and twice less frequent than their Spanish counterparts (22.04 pttw). In Section 3.3, the

104 María de los Ángeles Gómez González

Table 2.  Size and structure of C-ORAL-ROM. Section

Context domain

Informal Family/Private

Spanish words

Portuguese Text subtypes words

 198,475

 189,758

 160,663

 175,209

Text codes

Monologue

  42,082

  45,937

dialogue

famcv

Dialogue

  88,974

  86,950

monologue

fammn

Telephone

  29,607

  24,322

conversation

telef

Public

  37,812

  32,549

conversation

Monologue

   6,116

   7,696

dialogue

pubcv

Dialogue

  31,696

  24,853

monologue

pubmn

 135,007

 128,158

  72,268

  66,140

business dialogue business monologue conference monologue law conversation law monologue political debate prof. explanation conversation prof. explanation monologue preaching monologue preaching conversation

natbucv natbumn

interviews meteo news reportages science sport talk show

medin medmt mednw emedrp medsc medsp medts

Formal Natural context

Media

TOTAL OF WORDS AND TEXTS

  62,739

  62,018

333,482 210 TEXTS

317,916 152 TEXTS

natcomn natlacv natlamn natpd natpecv natpemn natpremn natprecv

Canonical tag questions in English, Spanish and Portuguese 105

35

32.08

30 25

22.04

20

ECTQs SCTQs PCTQs

15 11.26

10 5 0

Figure 2.  Normalized frequencies of ECTQ, SCTQs and PCTQs (pttw).

higher frequency of PCTQs and SCTQs is attributed to their higher flexibility as they are used in wider contexts and serve a wider variety of purposes than ECTQs. 3.2 Formal features This section first explains the percentages and relative frequencies of mood choices in the anchors of ECTQs, SCTQs and PCTQs. This is followed by a description of the positional tendencies of tags, as well as of the polarity of the constructions across the three languages. Starting with the first task, the findings displayed in Table 3 report a (highly) significant association, overall, between language and illocutionary type of the anchor clause (χ2 (4) = 32.26, p  DO > HAVE > WOULD > CAN > OTHER. This scale confirms the findings of previous investigations (Nelson, 2004; Quirk et al., 1985; Biber et al., 1999; Tottie and Hoffmann, 2006) and shows the preference of English tags for either the three primary auxiliary/full verbs (BE, DO, HAVE) or for anchor clauses with perfect tenses (Elsness, 1997). The situation regarding the correlation between the Subject pronouns in the tags, on the one hand, and the mood choices in the anchors, on the other, is reflected in Table 5, which reports the following frequency scale of Subject pronouns in QTs: IT > YOU > THEY > HE > THERE > WE > I/SHE > ONE/THAT, where ‘there’, ‘one’ and ‘that’ are the only three words other than personal pronouns. This finding is in accordance with Tottie and Hoffmann’s (2006) data, although they only found ‘there’ as an alternative to personal pronominal Subjects. In addition, Table 4 appears to suggest that Subject selection in the tags is determined by the mood choices in the anchors. For whilst declarative anchors allow all possible Subject choices in the tags, interrogative, exclamative and especially imperative anchors seem to impose more constraints. Imperative anchors offer two tag choices Table 4.  Tag operators in ECTQs. Tag operators English

BE

DO

HAVE

WOULD

CAN

OTHER

Total

Count

391

184

48

35

17

43

718

%

54.4

25.6

6.7

4.9

2.4

6

(highly) significant association between language and illocutionary type based on the differences between observed and expected frequencies in each cell of the table produced by SPSS.

Canonical tag questions in English, Spanish and Portuguese 107



Table 5.  Variation in the anchors and tags of ECTQs. Mood choices in anchors declarative

it

you they he

there we

I

she one that

153

43

30

24

21

20

47.3 22.2 10.1 6.3

4.4

3.5

3.1 2.9 0.1

0.1

0

0

1

2

0

0

  20

5

10

0

0

0

0

0

   8

1

0

0

0

0

   3

22

1

1

718

Count 325 %

exclamative

Count 13

4

%

20

65

interrogative Count 5 % imperative Total

Total

Subject pronouns in the tags

1

0 1

0 1

0

1

1

687

62.5 12.5 12.5 12.5

Count 0

2

%

66.7

Count 343

160

%

69

0

0

0

33.3 70

44

30

25

22

47.8 22.3 9.7

6.1

4.2

3.5

3.1 3.1 0.1

0.1

only, i.e. ‘you’ and ‘we’, whereas both exclamative and interrogative anchors display four possibilities: ‘it’, ‘you’, ‘I’ and ‘she’ in the former, and ‘it’, ‘you’, ‘they’ and ‘he’ in the latter. It follows from this that if the anchor is marked in terms of mood, that is, if an option other than declarative (the most frequent of all) is chosen, then the tag is more constrained in terms of Subject choices. However, given the very low frequencies for interrogative and imperative anchors, no really solid conclusions can be drawn on the basis of these data. When considering the operator pronoun combinations in the tags of ECTQs, 123 variations were found, the ten top ranks of which are shown in Table 6,10 in contrast with the 200 combinations reported by Tottie and Hoffmann (2006), but remarkably the first four ranks are exactly the same in both studies. Note that the most frequent pattern is ‘isn’t it?’ and this is three times as frequent as ‘is it?’, second in rank, which in turn is twice as common as ‘aren’t they?‘ and ‘don’t you?’, ranked third and fourth, respectively. Let us move on to the Spanish and Portuguese data, reproduced in Tables 7 and 8, respectively. In Spanish, we can see that ¿no? (698 tokens, 94.6%) is 18.8 times more frequent than ¿verdad? (37 items, 5.4%), and it can be combined with all possible mood choices of anchors, whereas ¿verdad? is not used with imperatives, although both tags mostly appear with indicative anchors, a claim that holds true for the three languages under analysis, as already noted at the beginning of this 10.  The three cases of CTQs found in the sample with an imperative anchor display two patterns that are not among the first 13 ranks listed in Table 6, namely: ‘shall we?’ (1 token), ‘will you?’ (2 hits).

108 María de los Ángeles Gómez González

Table 6.  The tag patterns of ECTs. Tag

Count % 718

Rank Declarative Exclamative Interrogative Rank Tottie & Hoffmann

isn’t it?

159

22.1   1

156

2

1

1 isn’t it?

is it?

  64

  8.9   2

  59

3

2

2 is it?

aren’t they?

  30

  4.1   3

  29

1

3 aren’t they?

don’t you?

  30

  4.1   4

  27

2

1

4 don’t you?

doesn’t it?

  28

  3.8   5

  27

1

5 do you?

wasn’t it?

  25

  3.4   6

  23

2

6 don’t they?

aren’t you?

  18

  2.5   7

  18

7 aren’t you?

do you?

  18

  2.5   7

  18

8 wasn’t it?

did you?

  17

  2.3   8

  16

don’t they?

  16

  2.2   9

  16

9 haven’t you?

1

10 are you?

isn’t he?

  14

  1.9 10

  14

11 weren’t it?

didn’t you?

  13

  1.8 11

  13

12 didn’t you?

haven’t you?   12

  1.6 12

  12

13 isn’t he?

can you?

  10

  1.3 13

  10

14 didn’t it?

does it?

  10

  1.3 13

   8

15 doesn’t it?

1

Table 7.  Variation in the anchors and tags of SCTQs. Tags

Mood choices in anchor clauses ¿no? ¿verdad?

Total

declarative

exclamative

interrogative

imperative

Total

Count

673

18

6

1

698

%

96.4

2.6

0.9

0.1

Count

32

3

2

0

  37 735

%

86.5

8.1

5.4

Count

705

21

8

1

%

95.9

2.9

1.1

0.1

section. In Portuguese, the first eight ranks of question tag patterns are: não é? > não? > não foi? > é? > não era? > foi? > pois não? > era? > não foste? > não eram?. This scale shows a preference for negative tags and it indicates that the tag with the highest rank, não é?, is by far the most frequent (almost 31 times as frequent as the second and 78 times as frequent as the third). Considering now the position of question tags in CTQs, a statistically significant association has been found to exist between this variable and the variable of

Canonical tag questions in English, Spanish and Portuguese 109



Table 8.  The tag patterns of PCTs. TAG

Count 1,020

%

Rank

Declarative

Interrogative

não é?

937

91.9%

1

933

4

não?

  29

  2.9%

2

  28

1

não foi?

  12

  1.4%

3

  12

é?

  10

  0.9%

4

  10

não era?

  10

  0.9%

4

  10

foi?

   9

  0.8%

5

   9

pois não?

   8

  0.7%

6

   8

era?

   2

  0.2%

7

   2

não foste?

   2

  0.2%

7

   2

não eram?

   1

  0.1%

8

   1

language (χ2 (2) = 44.54, p  (−/−) (14%)> (+/+) (2.1%) > (−/+) (0.1%), and (+/−) (86.3%) > (−/−) (8.7%) > (+/+) (4.4%) > (−/+) (0.6%), respectively. It could be claimed that the two Romance languages share an overwhelming preference for negative tags probably due to: – the negative semantics of ¿no?, which is much more frequent than its affirmative counterpart ¿verdad?; – the negative meaning of pois não?, a polarity-dependent tag that can only occur after negative anchors; – the fact that both não é?/não? similarly encode negative meaning and, although they can be used after affirmative or negative anchors, they are biased towards negative anchors.

112 María de los Ángeles Gómez González 100 86.3%

90

83.8%

80 70

69.9%

60

ECTQs SCTQs PCTQs

50 40 30 20

14% 8.7%

10 0

0.4%

Positive-Negative 80.5%

Negative-Negative 8.5%

16%

13.7% 4.4%

2.1%

Positive-Positive 6.1%

0.6% 0.1%

Negative-Positive 4.9%

Figure 5.  Polarity types across ECTQs, SCTQs and SCTQs.

This preference is less skewed in English, because there is a higher proportion of constant positive polarity TQs which come third in the scale, to the detriment of constant negative polarity TQs which are the rarest in the sample, while the two reversed types are the most frequent: (+/−) (69.1%) > (−/+) (16%) > (+/+) (13.7%) > (−/−) (0.4%). The same polarity scale is reported in Tottie and Hoffmann’s (2006) analysis of English TQs, and they also observe that (+/+) TQs are more frequent in British than in American English. Lastly, concerning constant negative polarity TQs, in spite of the facts that some English speakers consider them to be ungrammatical and some scholars demand usage evidence of this pattern (Quirk et al., 1985: 813), the existence of constant negative polarity TQs is confirmed by the three instances attested in our sample, one of which has been reproduced in example 4d. The structural tendencies described so far allow us to put forward a polarity/ tag-based paradigm for TQs that is presented in Table 10, according to which: – Portuguese is situated at the most variant end of the cline as, while admitting TQs with both invariant and variant question tags, it is the latter that display the highest degree of variability in kind of the three languages; – Spanish is placed at the opposite extreme using exclusively invariant QTs; – English occupies an in-between position, because it allows for both invariant and variant tags, but the latter are more restricted in nature than their Portuguese equivalents. In Section 3.4 the types of CTQs displayed in Table 10 will be examined in connection with the functions they serve in discourse, but before this can be undertaken, an analysis of their distribution across genres is called for, which is offered in Section 3.3.

Canonical tag questions in English, Spanish and Portuguese 113



Table 10.  A ‘tag-/polarity-based’ paradigm for TQs in English, Portuguese and Spanish.  Language

Question Tags (QTs)

Portuguese

Variant Lexically-dependent QTs (outside this study) Grammatically-dependent QTs não SER SER Invariant pois não?

English



não é?/não?



Other (outside this study)

Variant Grammatically dependent QTs

Invariant (outside this study) Spanish

Invariant

¿no? ¿verdad? Other (outside this study)

+ structural tag variance

Polarity

Polarity-neutral, but polarity-biased towards (+/−) TQs although the (−/−) type is also relatively frequent Negative polarity (−) Affirmative polarity (+) Polarity-dependent, only found in (−/−) TQs Polarity-neutral, but polarity-biased towards (+/−) TQs although the (−/−) type is also relatively frequent

Polarity neutral, but polarity-biased towards (+/−) TQs although the (−/+) and (+/+) types are also relatively frequent Polarity-neutral, but polarity-biased towards (+/−) TQs although the (−/−) type is also relatively frequent Negative polarity (−) Affirmative polarity (+) 

− structural tag variance

3.3 Distribution across genres As the text-type classifications for ICE-GB and C-ORAL-ROM are not identical (see Section 2) and often TQs in each text yielded numbers too low to allow for reliable quantification, text types were regrouped in the three languages along just three parameters (monologic/dialogic, formal/informal, and private/public), thereby obtaining just six text categories: private/informal dialogues, public/formal dialogues, public/informal dialogues, private/informal monologues, public/formal monologues, public/informal monologues. The results displayed

114 María de los Ángeles Gómez González

in Figure 611 confirm the dialogic nature of CTQs, as they are much more frequently used in conversations than in monologues across the three languages. The difference is dramatic in the case of English (98.7% in dialogues vs. 1.3% in monologues), while Spanish and Portuguese show similar percentages with dialogic CTQs (74.6% and 77.2%) being over twice as common as the monologic ones (25.4% and 22.8%).

Dialogues Count

%

Monologues Count

ECTQ SCTQs

709 548

98.7 74.6

ECTQs SCTQs

9 176

1.3 25.4

PCTQs

788

77.2

PCTQs

232

22.8

%

Private/Informal % Count 65.6 ECTQ 471 29.2 SCTQs 215 45.9 PCTQs 468

Private/Informal Count % ECTQ 8 1.2 SCTQs 50 6.8 PCTQs 156 15.3

Public/Formal Count % ECTQ 238 33.1

Public/Formal Count % ECTQ 1 0.1

SCTQs PCTQs

SCTQs PCTQs

223 200

30.3 19.6

Public/Informal Count % ECTQ SCTQs PCTQs

103 22

14 21.2

Public/Informal Count % ECTQ

110 120

14.9 11.7

SCTQs PCTQs

23 54

3.2 5.3

Figure 6.  Distribution of CTQs across genres in English, Spanish and Portuguese. 11.  Only the results observed in Spanish and Portuguese can be compared and tested for statistical significance in the six text-subtypes recognized, private/informal dialogues, public/formal dialogues, public/informal dialogues, private/informal monologues, public/formal monologues, public/informal monologues, because the public/informal category is not present in the English data. But the three languages can be compared along the remaining four text-types, private/informal dialogues, public/formal dialogues, public/informal dialogues, private/informal monologues, on the basis that the public material in ICE-GB is similar to the public/formal combination in C-ORAL-ROM.

Canonical tag questions in English, Spanish and Portuguese 115



Furthermore, ECTQs tend to be used more in private/informal settings both in dialogues (65.6%) and monologues (1.2%) than in public/formal situations whether they be dialogic (33.1%) or monologic (0.1%). The situation in the two Romance languages is rather different. The findings in Spanish report a statistically significant association (χ2 (1) = 17.10, p  focusing, the first three being also reported as most frequent by Tottie and Hoffmann (2006) though in a different order (confirmatory > facilitating > attitudinal); – informational > attitudinal > regulatory-delaying > focusing; – regulatory-delaying > informational > attitudinal > focusing. 12.  The hortatory and phatic categories were omitted from the chi square analysis because their inclusion would have led to too many expected frequencies being less than 5, and for that reason the numbers of degrees of freedom are only 10 (5 x 2), not 14. 13.  In Table 12, to avoid the profusion of cells with low numbers the instances of hybrid focusing TQs (e.g. focusing-informational, focusing-challenging) have been excluded from the focusing category, and have been assigned to their corresponding co-functional category (e.g. informational, challenging).



Canonical tag questions in English, Spanish and Portuguese 121

These findings lead us to conclude that English, Spanish and Portuguese speakers mostly use CTQs to perform a variety of interpersonal functions: seek information or confirmation from their addressees or, to a lesser degree, express their feelings or attitudes with respect to the proposition encoded in the anchor, or even in fewer cases boost the anchor’s negative force. As a consequence, TQs profile themselves as discourse strategies well suited for the establishment, maintenance or threatening of good social relations in dialogic discourse (Recsky, 2006), generally in the private/informal domain in the case of English, and in public and informal contexts in Spanish and Portuguese, respectively. Distinctively frequent in English are facilitative tags which are used to invite the addressee to do something out of positive politeness. The scarcity of the facilitative and hortatory types in Spanish and their absence from the Portuguese data possibly uncover cultural differences between these two Romance cultures, on the one hand, and English, on the other, or otherwise they suggest that these two functions are performed by other markers in Spanish and Portuguese. In addition, as to exchange-centred TQs, the only use that occurs in the three languages is the focusing type, which, as already remarked, registers the highest percentages in English although it comes fourth in the frequency rank across the three languages. Curiously enough, the regulatory-delaying subtype is pervasive in Portuguese (45.4%), but it is much less common in Spanish (23.4%) and does not appear at all in the English corpus. As already noted, exchange-centred tags mostly pursue a variety of textual effects which make them particularly well suited for (public) monologic discourse. Often in Spanish but particularly in Portuguese, monologues as well as narrative and explanatory interventions in conversations provide ample evidence of the usage of focusing and regulatory tags. One last conclusion to be drawn from the findings seems to transcend the language-based differences just mentioned. For one thing, while exchange-centred TQs have been found to mostly occur in formal/public monologic discourse, addresseecentred and speaker-centred TQs cluster around informal/private dialogic discourse. 4. Conclusions The results reported here reveal five major findings. First, CTQs are far more common in the two Romance languages, Portuguese and Spanish, particularly in the former, than in English. Second, the central type of CTQs consists of a declarative anchor followed by a tag with the pattern ‘isn’t it?’ in English, ¿no? in Spanish, and não é? in Portuguese, in addition to displaying a polarity of the reversed (+/−) type across the three languages. Third, taking tag-variance and polarity as reference for a comparison, a working paradigm for CTQs has been proposed according to which:

122 María de los Ángeles Gómez González

– Portuguese is situated at the most variant end of the scale, displaying both lexically and grammatically dependent QTs, as well as invariant QTs; – Spanish is situated at the opposite extreme, using exclusively invariant QTs; – English occupies an in-between position, because it allows for both invariant and variant QTs, but the latter are restricted to the grammatically-dependent type. Fourth, according to our data, the most frequent functions fulfilled by TQs are informational > regulatory > attitudinal > focusing (in decreasing frequency). However, ECTQs show less functional flexibility than their Spanish and Portuguese counterparts on the grounds that there are no instances of the regulatory and phatic types in the English corpus, and so the resulting scale is informational > attitudinal > facilitative > focusing. Besides, ECTQs are almost confined to dialogic conversations and mostly serve an interpersonal (addressee-centred or speaker-centred) purpose (information/confirmation-seeking, attitudinal or facilitative). Portuguese and Spanish show a greater array of functional choices and report a greater percentage of regulatory TQs, which are the predominant ones in Portuguese, especially in monologic discourse, although quite frequently SCTQs and PCTQs are also employed to seek information/confirmation, most notably in public and informal contexts, respectively. And fifth, medial tags perform a focusing function across the three languages, but they turn out to be more common in English than in Spanish and Portuguese. This is also the case of facilitative TQs, which are very rare in Spanish and apparently not used in Portuguese. In conclusion, I hope to have shown that the quantitative, trilingual approach adopted in this study can contribute to both the contrastive study of CTQs and to a better understanding of the meaning and use of their correspondences in discourse across languages. Further research is needed to complete the picture based on reliable comparative statistics so as to be able, for instance, to: (i) account for all possible TQ patterns including the canonical and the non-canonical types (with variant or invariant tags) in English, Spanish and Portugese; (ii) uncover the intonation patterns of the constructions; and (iii) reveal the sociolinguistic factors that trigger their use. This chapter is only a modest step towards that aim, but hopefully it can indicate a viable direction in which to proceed.

Acknowledgements The support of the Spanish Ministry of Economy and Competitiveness (MINECO) is gratefully acknowledged (FFI2010-19380). I would also like to thank Alba Dias and Milagros Torrado for their help in compiling the data. My heartfelt gratitude also goes to Susana Doval, Elsa

Canonical tag questions in English, Spanish and Portuguese 123



González and Chris Butler, who generously sacrificed some of their precious time over the 2012 Christmas holidays to assist me with the statistical analysis of my data. I am also grateful to Lachlan Mackenzie and again to Chris Butler for their untiring encouragement and their perspicuous comments on earlier versions of this manuscript. Last, but not least, my gratitude also goes to the anonymous referees and the editors of this volume, who have helped me greatly by pointing out weaknesses in presentation and argumentation. Responsibility for any remaining flaws rests with me.

References Aijmer, K. 1979. “The Function of Tag Questions in English”. In Papers from the Fifth Scandinavian Conference of Linguistics, T. Pettersson (ed), 9–17. Lund: Acta Universitatis Lundensis and Stockholm: Almqvist and Wiksell. Algeo, J. 1988. “The Tag Question in British English: It’s Different i’n’it?” English World-Wide 9:171–191.  doi: 10.1075/eww.9.2.03alg. Algeo, J. 1990. “It’s a Myth, innit? Politeness and the English Tag Question”. In The State of the Language, C. Ricks and L. Michaels (eds), 443–450. Berkeley: University of California Press. Algeo, J. 2006. British or American English? Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511607240.

Allerton, D. 2009. “Tag Questions”. In One Language, Two Grammars: Differences between British and American English, G. Rohdenburg and J. Schlüter (eds), 306–323. Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511551970.017. Axelsson, K. 2006. “Tag Questions in English Translations from Swedish and Norwegian – Are There Differences?” In Svenska som källspråk och målspråk. Aspekter på översättningsvetenskap: artiklar från en forskarkurs vid Göteborgs universitet höstterminen, B. Englund Dimitrova and H. Landqvist (eds), 4–21. Göteborgs universiteit: Översättarutbildningen vid Humanistiska fakulteten. Axelsson, K. 2011a. Tag Questions in Fiction Dialogues. PhD Dissertation, University of Gothenburg. Axelsson, K. 2011b.“A Cross-linguistic Study of Grammatically Dependent Question Tags: Data and Theoretical Implications”. Studies in Language 35(4):793–851.  doi: 10.1075/sl.​35.​4.​02axe.

Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. 1999. Longman Grammar of Spoken and Written English. London: Longman. Briz, A. 1998. El español coloquial en la conversación. Barcelona: Ariel. Briz, A., Pons Bordería, S. and Portolés, J. 2008. Diccionario de partículas discursivas del español [on line]. Available at www.dpde.es/. Brown, P. and Enfield, N. 2006. “Epistemic Dimensions of Polar Questions: Declarative Questions and Tag Questions in Comparative Perspective”. In Questions and their Responses [on line], N. Enfield, J.P. de Ruiter. Max Planck Institute for Psycholinguistics, March 23–March 25. Available at www.mpi.nl/events/PastEvents/Questions. [Last accessed 27 March 2006]. Cameron, D., McAlinden, F. and O’Leary, K. 1989. “Lakoff in Context: The Social and Linguistic Functions of Tag Questions”. In Women in their Speech Communities: New Perspectives on Language and Sex, J. Coates and D. Cameron (eds), 74–93. London: Longman.

124 María de los Ángeles Gómez González Carter, R. and McCarthy, M. 2006. A Comprehensive Guide Spoken and Writen English Grammar and Usage. Cambridge: Cambridge University Press. Christl, J. 1996. “Muletillas en el español hablado”. In El español hablado y la cultura oral en España e Hispanoamérica, T. Kotschi, W. Oesterreicher and K. Zimmermann (eds), 117– 143. Madrid: Vervuert. Coates, J. 1989. “Gossip Revisited: Language in All-Female Groups”. In Women in their Speech Communities, J. Coates and D. Cameron (eds), 94–122. London: Longman. Coates, J. 1996. Women Talk. Cambridge, MA: Blackwell. Cortés, L. and Camacho, M. 2005. Unidades de segmentación y marcadores del discurso. Madrid: Arco Libros. Cresti, E. and Moneglia, M. 2005. CORALROM. Integrated Reference Corpora for Spoken Romance Languages. DVD+vol. Amsterdam: John Benjamins. Cruz Ferreira, M. 1981. “Tag Questions in Portuguese: Grammar and Intonation”. Phonetica 38:341–352.  doi: 10.1159/000260037. Cuenca, M.J. 1997. “Form-Use Mappings for Tag Questions”. In Discourse and Perspective in Cognitive Linguistics, W. Liebert, G. Redeker and L. Waugh (eds), 3–19. Amsterdam: John Benjamins. Domínguez Mujica, C.L. 2005. “Marcadores de (in)conclusión en el español de MéridaVenezuela”. Boletín de Lingüística 23:3–23. Elsness, J. 1997. The Perfect and the Preterite in Contemporary and Earlier English. Berlin: Mouton de Gruyter.  doi: 10.1515/9783110810264. Fernández Sanmartín, A. 2006. “La expresión de la modalidad epistémica en español conversacional”. Unpublished manuscript. Universidade de Santiago de Compostela. Field, A. 2009. Discovering Statistics Using SPSS (3rd edition). Los Angeles, London, New Delhi, Singapore and Washington: Sage. García Vizcaíno, M.J. 2005. “El uso de los apéndices modalizadores ‘no?’ y ‘eh?’ en español peninsular”. In Selected Proceedings of the Second Workshop on Spanish Sociolinguistics, L. Sayahi and M. Westmoreland (eds), 89–101. MA: Cascadilla Proceedings Project. Gili Gaya, S. 1971. Curso superior de sintaxis española. Barcelona: Gredos. Gómez González, M.Á. 2012. “The Question of Tag Questions in English and Spanish”. In Encoding the Past, Decoding the Future: Corpora in the 21st Century, I. Moskowich and B. Crespo (eds), 59–97. Newcastle upon Tyne: Cambridge Scholarly Press. Gonçalves, F. 1995. Negação frásica em Português: Caracterização com referência ao processo de aquisição. Dissertação de Mestrado, Lisboa: FLUL. Hagemeijer, T. and Santos, A.L. 2004. “Elementos polares na periferia direita”. Actas do XIX Encontro Nacional da Associação Portuguesa de Linguística, C. Nunes Correia and A. Gonçalves (eds), 465–476. Lisboa: Associação Portuguesa de Linguística. Hoffmann, S. and Mukherjee, J. 2011. “Canonical Tag Questions in Asian Englishes: Forms, Functions and Frequencies in Hongkong English, Indian English and Singapore English”. Paper presented at the 32nd Conference of the International Computer Archive of Modern and Medieval English (ICAME 32). Oslo, Norway, 1–5 June 2011. Holmes, J. 1983. “The Functions of Tag Questions”. English Language Research Journal 3:40–65. Holmes, J. 1984. “Hedging your Bets and Sitting on the Fence: Some Evidence for Hedges as Support Structures”. Te Reo 27:47–62. Holmes, J. 1986. “Functions of ‘You Know’ in Women’s and Men’s Speech”. Language in Society 15:1–21.  doi: 10.1017/S0047404500011623. Holmes, J. 1995. Women, Men and Politeness. White Plains, NY: Longman.

Canonical tag questions in English, Spanish and Portuguese 125



Huddleston, R. 1970. “Two Approaches to the Analysis of Tags”. Journal of Linguistics 6:215– 222.  doi: 10.1017/S0022226700002619. Huddleston, R. and Pullum, G.K. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hudson, R. 1975. “The Meaning of Questions”. Language 51:1–31.  doi: 10.2307/413148. Kay, P. 2002. “English Subjectless Tagged Sentences”. Language 78:453–81.  doi: 10.1353/lan.​2002.​0161.

Kimps, D. 2007. “Declarative Constant Polarity Tag Questions: A Data-Driven Analysis of their Form, Meaning and Attitudinal Uses”. Journal of Pragmatics 39(2):270–291.  doi: 10.1016/j.pragma.2006.08.003.

Ladd, D.R. 1981. “A First Look at the Semantics and Pragmatics of Negative Questions and Tag Questions”. Proceedings of CLS 17.164–171. Lakoff, R. 1975. Language and Woman’s Place. New York: Harper and Row. Martín Zorraquino, M.A. and Portolés, J. 1999. “Los marcadores del discurso”. In Gramática descriptiva de la lengua española, vol. 3, I. Bosque and V. Demonte (coords.), 4051–4213. Madrid: Espasa Calpe. Martins, A. 2006. “Emphatic Affirmation and Polarity: Contrasting European Portuguese with Brazilian Portuguese, Spanish, Catalan and Galician”. In Romance Languages and Linguistic Theory. Selected papers from Going Romance 2004, J. Doetjes and P. González (eds), 197– 223. Amsterdam: John Benjamins. Moniz, M., Batista, F., Trancoso, I. and Mata, A.I. 2011. “Analysis of Interrogatives in Different Domains”. In Proceedings of the Third COST 2102 International Training School Conference on Toward Autonomous, Adaptive, and Context-aware Multimodal Interfaces: Theoretical and Practical Issues, A. Esposito et al. (eds), 134–146. Berlin: Springer Verlag. Moravcsik, E.A. 1971. “Some Cross-linguistic Generalizations about Yes-No Questions and their Answers”. Stanford Working Papers on Linguistic Universals 7:45–193. Nässlin, S. 1984. The English Tag Question: A Study of Sentences Containing Tags of the Type ‘Isn’t it?, Is it?’. Stockholm: Almqvist and Wiksell International. Nelson, G. 2004. “Negation of Lexical ‘Have’ in Conversational English”. World Englishes 23: 299–308.  doi: 10.1111/j.0883-2919.2004.00353.x. Nelson, G., Wallis, S. and Aarts, B. 2002. Exploring Natural Language: The British Component of the International Corpus of English. Amsterdam: John Benjamins. O’Connor, J.D. 1955. “The Intonation of Tag Questions in English”. English Studies 36:97– 105.  doi: 10.1080/00138385508596942. Ortega Olivares, J. 1985. “Apendices modalizadores en español: los ‘comprobativos’ ”. In Estudios románicos dedicados al profesor Andres Soria Ortega, A. Soria et al. (eds), 239–255. Granada: Universidad de Granada. Pons Bordería, S. 1998. Conexión y conectores. Estudio de su relación en el registro informal de la lengua. Anejo XXVII de Cuadernos de Filologia. Valencia: Universitat de Valencia. Portolés, J. 1998. Marcadores del discurso. Barcelona: Ariel. Quirk, R., Greenbaum, S, Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Recsky, L. 2006. “Epistemic Modality and Spoken Discourse: An English-Portuguese Crosslinguistic Investigation”. Linguagem e Ensino 9(1):159–185. Reese, B. 2007. Bias in Questions. PhD Dissertation, University of Texas at Austin. Rigney, A.C. 1999. “Questioning in Interpreted Testimony”. Forensic Linguistics 6(1):83– 108.  doi: 10.1558/sll.1999.6.1.83.

126 María de los Ángeles Gómez González Rodríguez Muñoz, F.J. 2009. “Estudio sobre las funciones pragmadiscursivas de ‘¿eh?’ ‘¿no?’ en el español hablado”. Revista de lingüística teórica y aplicada 47(1):83–101. Roesle, A. 2001. Tag Questions in British and American English. M.A. Thesis, University of Zurich. Stenström, A.-B. 1997. “Tags in Teenage Talk”. In Studies in English Corpus Linguistics, U. Fries, V. Müller and P. Schneider (eds), 139–147. Amsterdam: Rodopi. Stenström, A.-B. 2005. “Teenagers’ Tags in London and Madrid”. In Contexts -Historical, Social, Linguistic. Studies in Celebration of Toril Swan, K. McCafferty, T. Bull and K. Killie (eds), 279–291. Bern: Peter Lang. Tottie, G. and Hoffmann, S. 2006. “Tag Questions in British and American English”. Journal of English Linguistics 34(4):1–29.  doi: 10.1177/0075424206294369. Tottie, G. and Hoffmann, S. 2009. “Tag Questions in English: the First Century”. Journal of English Linguistics 37(2):130–161.  doi: 10.1177/0075424209332962. Ultan, R. 1969. “Some General Characteristics of Interrogative Systems”. Working Papers on Language Universals 1:41–63. Ultan, R. 1978. “Some General Characteristics of Interrogative Systems”. In Universals of Human Language. Volume 4, Syntax, J.H. Greenberg (ed), 211–248. Stanford: Stanford University Press.

Loving and hating the movies in English, German and Spanish Maite Taboada, Marta Carretero and Jennifer Hinnell

Simon Fraser University (Canada) / Universidad Complutense de Madrid (Spain) / University of Alberta (Canada)

We present a quantitative analysis of evaluative language in a genre in which it is particularly prominent, that of movie reviews. The data chosen are non-professional consumer-generated reviews written in English, German and Spanish. The reviews are analysed in terms of the categories of Attitude and Graduation within the Appraisal framework (Martin and White, 2005). A number of similarities in the distribution of the Appraisal subcategories were found across the three languages, such as the high frequency of Appreciation and the narrow relationship between the global polarity of the reviews and the individual polarity of the spans. More importantly, the analysis uncovers a number of cross-linguistic distributional differences, which may be explained in terms of a wide array of factors, such as lexicogrammar, word order, argumentative style or sociocultural reasons. Keywords: Appraisal, evaluation, genre, consumer reviews, English/German/ Spanish

1. Introduction After expressions of desires and needs, the linguistic manifestation of emotion and evaluation is probably one of the most basic functions of language (cf. personal function in Halliday, 1975). One of the aspects of language that we make use of every day is the expression of our own emotions, and the evaluation of persons and objects around us. This expression of emotions and evaluations is studied under different umbrella terms in linguistics and other social sciences. Studies of affect (Batson et al., 1992), subjectivity and point of view (Banfield, 1982; Langacker, 1990; Traugott, 1995; 2010), evidentiality (Chafe and Nichols, 1986; Aikhenvald, 2004), attitudinal stance (Biber and Finegan, 1988; 1989), modality (Palmer, 1986; Bybee and Fleischman, 1995; Portner, 2009) and appraisal (Martin and White,

doi 10.1075/bct.87.07tab 2016 © John Benjamins Publishing Company

128 Maite Taboada, Marta Carretero and Jennifer Hinnell

2005), to mention just a few in each area, all aim at explaining how we use language to convey emotions, evaluation and subjective expressions. As this body of work shows, researchers have been interested for a long time in how we use language to express evaluation and subjectivity, but even more so recently, spurred in part by interest in the automatic extraction of opinions found online (for surveys, see Pang and Lee, 2008; Liu, 2015). At the same time, research in Systemic Functional Linguistics (Halliday, 1985; Halliday and Matthiessen, 2004) has resulted in a now mature proposal for classifying and studying evaluation, the Appraisal framework (Martin and White, 2005). With some notable exceptions, most of that research is carried out in English and, when other languages are studied, they tend to be approached in isolation, without comparison across languages. In this chapter, we present a corpus study of evaluation in a common online genre, that of movie reviews, carried out contrastively across three languages: English, German and Spanish. We collected data from similar sources and with similar characteristics, and annotated it using the categories proposed within Appraisal. Our analysis shows that, although there are differences across the languages, the common genre seems to lead to a similar breakdown of Appraisal categories. Differences across languages are centered, on the one hand, around the lower levels of language (different word order configurations) and, on the other hand, around the higher-level discourse organization. In Section 2 we provide a summary of the Appraisal framework, and its place in Systemic Functional Linguistics. Section 3 is devoted to explaining our methodology for annotation, including a description of the corpus used. Results for each of the languages, and a discussion of these results, are provided in Section 4, followed by a general discussion and conclusions in Section 5. 2. Appraisal in English and other languages The Appraisal framework belongs in the systemic functional tradition started by Halliday (Halliday, 1985; Halliday and Matthiessen, 2004), and has been developed mostly in Australia by Jim Martin, Peter White and colleagues (Martin, 2000; White, 2003b; Martin and White, 2005). Martin (2000) characterizes Appraisal as the set of resources used to negotiate emotions, judgements, and valuations, alongside resources for amplifying and engaging with those evaluations. The Appraisal framework is an approach to evaluative language in line with systemic functional linguistics, in the sense that meaning-making is conceived as a choice among different terms in a system. Martin considers that Appraisal resources form a system of their own within language, and divides the Appraisal system into three distinct sub-systems Attitude, Engagement

Loving and hating the movies in English, German and Spanish 129



and Graduation, which will be described below. Since Martin’s approach is lexically rather than grammatically based, he is primarily concerned with those words and semantic categories of words that allow a speaker to express different types of evaluations. Figure 1 summarizes the Appraisal network, with some example realizations, showing prototypical cases within the respective sub-systems. Let us now briefly discuss the main sub-systems of Appraisal. Attitude has three subsystems: Affect is used to construe emotional responses about the speaker or somebody else’s reactions (e.g. ‘happiness’, ‘sadness’, ‘fear’); Judgement conveys moral evaluations of character about persons or, less commonly, non-human entities (e.g. ‘ethical’, ‘deceptive’, ‘brave’); and Appreciation captures aesthetic qualities, most often of objects and natural phenomena (‘remarkable’, ‘desirable’, ‘harmonious’, ‘elegant’, ‘innovative’). These categories convey different options among which a speaker/writer can choose in order to communicate an evaluation: the speaker/ writer may express his/her feelings (‘I love this movie’), praise the work of one or more persons involved in the film (‘The director and the cast have done a really good job’) or praise the movie itself (‘This is a great movie’). In earlier work (Taboada and Grieve, 2004), we characterized each sub-system as appearing in prototypical sentences: Affect may be conveyed by adjectives that appear in sentences such as ‘I was X’ (‘I was sad’, ‘I was scared’). Judgement uses Affect sad, cheerful, anxious Attitude

Judgement lucky, tragic, powerful Appreciation engaging, lovely, dull Monoglossic

Contract

Engagement Appraisal

Proclaim naturally, admittedly... but, indeed, X shows/demonstrates that Y

Heteroglossic

Expand

Graduation

Force a little, somewhat, very Focus true, effectively, kind of

Figure 1.  Appraisal system.

Disclaim no, didn’t, although, amazingly

Entertain Perhaps, it seems to me, expository questions Attribute X argues Y, many people believe X, X claims, X states

130 Maite Taboada, Marta Carretero and Jennifer Hinnell

‘He was X’ (‘He was brave’, ‘He was a coward’), whereas Appreciation is seen in the ‘It was X’ pattern (‘It was interesting’, ‘It was beautiful’). Martin (2003) proposed the frames ‘I feel (very) X’; ‘It was X of him to do that’; and ‘I consider it X’ for Affect, Judgement and Appreciation, respectively. Those are, naturally, simplified prototypical patterns, and all correspond to inscribed instances, those that are explicitly expressed in the text. Instances that are not inscribed are considered to be evoked, in which “an evaluative response is projected by reference to events or states which are conventionally prized” (Hunston and Thompson, 2000: 142). Thus, ‘a bright kid’ or ‘a vicious kid’ are inscribed. On the other hand, ‘a kid who reads a lot’ or ‘a kid who tears the wings off butterflies’ present evoked Appraisal. The Engagement system refers to the distinction between heteroglossic and monoglossic expressions, following proposals by Bakhtin (1981). In a heteroglossic expression, inter-subjective positioning is open, because utterances invoke, acknowledge, respond to, anticipate, revise or challenge a range of convergent and divergent alternative utterances (White, 2003b; 2003a; Martin and White, 2005). The alternative is Monoglossia, where no alternative view, or openness to accepting one is present. Monoglossic utterances are presented as facts. Within Heteroglossia, the two possibilities are Contract and Expand, depending on whether possibilities for different opinions are either limited or open. Contract can, in turn, take the form of Disclaim (position at odds with or rejecting some contrary position) or Proclaim, where a speaker or writer sets themselves against, suppresses or rules out alternative positions. Examples of Disclaim are negative statements; Proclaim can be expressed through adverbials such as ‘naturally’ or ‘admittedly’, as seen in Figure 1. When expansion is possible, the two possibilities are Entertain, where the speaker represents the position as one of a number of possible positions, and Attribute, where the proposition is presented as externally grounded, in the words of another speaker. Finally, the Graduation system allows modulation of the evaluation by emphasizing or downtoning other expressions. Within the Graduation system, the options are Focus and Force. Focus graduates according to prototypicality; it is divided into Sharpen, which enhances prototypicality (‘a true friend’) and Soften, which downtones it (‘a kind of friend’). The expressions graduated with Focus are usually non-gradable. Force, which modulates gradable expressions, is divided into Intensification and Quantification, both of which can be emphasized or downtoned. Examples of emphasized and downtoned Intensification are ‘very interesting’ and ‘a little bit sad’, respectively; examples of emphasized and downtoned Quantification are ‘many advantages’ and ‘few good scenes’, respectively. In this chapter, we are focusing on the two systems of Attitude and Graduation: the analysis will be based on the spans of Attitude and how they are emphasized or downtoned by spans of Graduation. Engagement is undoubtedly important,



Loving and hating the movies in English, German and Spanish 131

but, given its complexity, it falls outside the scope of our current annotation work. We are investigating, in related work, how Engagement overlaps with other areas of the linguistic system, such as modality (Carretero and Taboada, 2015) and the expression of non-veridicality (Trnavac and Taboada, 2012). Although much ground remains to be covered, Appraisal is quite well understood in English, with a wide range of studies dealing with different genres, from political discourse/news stories (White, 1998; Coffin and O’Halloran, 2006) and different types of narratives (Macken-Horarik, 2003; Page, 2003), including those produced by school children (Martin, 1996; Coffin, 1997) and by children in the process of language acquisition (Painter, 2003) to discussions of literary texts (Love, 2006) and casual conversation (Eggins and Slade, 1997). A few studies for other languages exist. For Spanish, Kaplan (2007) studied television news (including both sound and images) and how the news editors mark their own point of view, and Achugar (2008) tracked the construction of memory during Uruguay’s military dictatorship in part through Appraisal analyses of historical documents. In German, Becker (2009) studied English-German political interviews, focusing on the expression of Engagement. In general, the three languages in this chapter have been studied as pairs, with no work covering all three at the same time. House has extensively studied GermanEnglish differences, and has observed differences in communicative style, such as more directness and explicitness in German as opposed to English (House, 2006). The other main work we should mention here is Johansson’s study of ‘love’ and ‘hate’ in English and their cognates in Norwegian, which inspired the title of this chapter (Johansson, 1998). Johansson found that the verbs ‘love’ and ‘hate’ are used much more frequently in Norwegian translations of English than in original Norwegian texts, because of their higher frequency in English. We will see that ‘love’ and ‘hate’ and their equivalents in German and Spanish are actually quite infrequent in our corpus, because they express Affect, which, as we will show, is not very common in our corpus, in contrast to Appreciation. 3. Corpus and methodology 3.1 Tri-lingual comparable corpus We collected a three-way corpus of movie reviews in English, German and Spanish. The English corpus comes from the website Epinions; the Spanish one from Ciao; and the German corpus from the German version of Ciao.1 Most of the texts are 1.  The websites are, respectively: www.epinions.com, www.ciao.es and www.ciao.de.

132 Maite Taboada, Marta Carretero and Jennifer Hinnell

part of the SFU Review Corpus (Taboada, 2008), a larger collection of reviews of books, movies and other consumer products, portions of which are annotated with Appraisal labels and with Rhetorical Structure Theory relations (Mann and Thompson, 1988). The texts belong to the online review genre, texts written by non-experts and posted online, with the purpose of being informative to other potential viewers. This genre differs, in formality and structural characteristics, from that of critical reviews written by professional movie critics and printed in newspapers (Taboada, 2011). For this chapter, we selected 50 reviews for each language, equally divided between favourable reviews (positive reviews) and unfavourable reviews (negative reviews), as indicated by the reviewer’s label of “recommended” or “not recommended”. Because the texts were collected at different times, they review many different movies, typically those recently released at the time. The English corpus, collected in 2004, reviews the following movies: “Bad Santa”, “Calendar Girls”, “The Cat in the Hat”, “Elf ”, “Gothika”, “The Haunted Mansion”, “The Last of the Samurai” and “Mona Lisa Smile”. The Spanish corpus, collected in 2008, contains reviews of a number of Hollywood movies (“27 Dresses”, “Bridge to Terabithia”, “Indiana Jones and the Kingdom of the Crystal Skull”, “Juno”, “Ratatouille”, “Sex and the City”, “Shrek 3”, “The Dark Knight” and “Wall-E” among them), as well as some Spanish ones (“El orfanato”, “Los crímenes de Oxford”). Finally, the German corpus was collected in 2011, and it reviews primarily 2010–2011 Hollywood hits in their German translations, among them “True Grit”, “Freundschaft Plus” (“No Strings Attached”), “127 Hours”, “Tron Legacy”, “The American”, “Little Fockers”, “Gullivers Reisen” (“Gulliver’s Travels”), “Unstoppable-Ausser Kontrolle” (“Unstoppable”), “The Road”, “The Green Hornet”, and “Avatar 3D”, as well as original German films, such as “Der letzte schöne Herbsttag”. The English corpus has been analysed elsewhere (Taboada, 2011) as an instance of a particular genre, with a detailed description of the stages it contains. Upon examination (but not full analysis) of the texts in the other two languages, we can state that the generic structure is quite similar to that of English. This similarity is to be expected, since the corpus is comparable, in that the genre is the same across languages; the medium, the writers and the audience are assumed to be similar; and the texts all deal with the common experience of watching a film and commenting on it. Although there are potential differences in the three languages and cultures, we found that the generic structure was very similar, with Description and Evaluation stages present in all texts. These stages can be sub-classified as description or evaluation of plot, characters, or other aspects of the movie. In addition, some reviews contain optional Background and Subject Matter stages, where additional details or a general summary of the movie’s content are provided. For analysis of the English stages, see Taboada (2011). In terms of register (Halliday,



Loving and hating the movies in English, German and Spanish 133

1989; Eggins and Martin, 1997), they all share similar properties of field, tenor and mode, with the field being discussion and evaluation of a movie, the tenor being informal (but with an unknown audience), and the mode written. Table 1 shows a summary of the number of words and sentences for each language. The sentence count is approximate, and it is derived from the UAM CorpusTool,2 the system used to annotate the data (see Section 3.2 for more detail). 3.2 Annotation methodology The annotation was carried out by the three authors of the chapter. To ensure that it was consistent across all annotators, we designed a procedure that outlined clearly the segments to be annotated, and provided general guidelines for the annotation. We found early on that the most difficult task was not so much how to annotate particular segments, but to decide what to annotate, that is, what constituted a span or segment. In order to help with those decisions, and to keep annotation consistent across languages and annotators, we created an annotation manual, and have reported preliminary analyses of the English and Spanish data (Taboada and Carretero, 2012). The main aspects are summarized in the remainder of this section. First of all, we decided that ‘markables’ (i.e. units to be annotated) are those aspects that refer to the movie itself, not to aspects external to the movie, such as opinions of the reviewer about anything other than the movie, reasons why they went to the theatre on a particular day, or references to other movies involving the same professionals (director, actors, scriptwriters, etc.). Similarly, evaluative aspects about the characters or atmosphere of the movie are excluded. For instance, the description of a character as “this mean, controlling mother” in example 13 is not part of the artistic achievement of the movie. The description of the performances as clichéd, or of Mitchell’s character as “a little too much” is, however, the territory of the scriptwriter and the director, and therefore part of the artistic merit of the movie. We identify markables with underlining.

(1) While the smaller performances of Juliet Stevenson, Marian Seldes, and Donna Mitchell as Dunst’s mother were used to a good degree despite their clichés while [sic] Mitchell’s character as this mean, controlling mother was a little too much. [no1] (no1=English corpus, negative review #1)

2.  www.wagsoft.com/CorpusTool/ 3.  Examples are reproduced verbatim, including any typos, grammatical errors and punctuation errors (such as lack of accentuation in Spanish). Because some of the spans that we reproduce are incomplete, examples do not always have final punctuation.

134 Maite Taboada, Marta Carretero and Jennifer Hinnell

Table 1.  Corpus statistics. English

German

Spanish

Words

33,387

47,128

52,845

Words, favourable reviews

17,219

25,401

36,071

Words, unfavourable reviews

16,168

21,727

16,774

Sentences

  1,766

  2,608

  2,593

The reason to exclude aspects not intrinsic to the movie is that the annotation work is part of a larger project that aims at extracting opinion automatically, by processing texts and providing a score that corresponds to the text’s sentiment about the movie, book or product being reviewed (Taboada et al., 2011). Because opinions about aspects other than the movie are not relevant in such a task, we decided not to annotate them. In terms of the categories to be annotated, we focused on the Attitude and Graduation systems, leaving Engagement for future work. In our initial annotations, we realized that Engagement is a much more complex system, not only because of the different types of parts of speech involved (verbs, adverbs, etc.), but also because it often entails annotating large chunks of the text; Engagement tends to be expressed through longer phrases or even entire sentences. Despite our decision to leave Engagement out for now, in some cases Engagement was close to Attitude or Graduation. For instance, in example 2, “definitely” expresses Engagement, but its effect in conjunction with the adjective “long” is very close to that of intensifiers. Another case is presented in example  3, where there are two downtoners, “I guess” and “if you are into those types of things”. The second one we have annotated as part of the expression of Attitude, but we believe that “I guess” expresses mainly Engagement, and should not be part of the markable.

(2) Definitely too long [no17]



(3) neat, I guess, if you are into those types of things [no19]

Within Graduation, annotation was complicated in cases where there were markables containing two types of Graduation. Due to limitations in CorpusTool, it is not possible to annotate both instances without separating the span. In these cases the element with wider scope took precedence. For instance, in example 4, einfach (“simply”) takes scope over absolut, and therefore it is annotated (as Focus), and the intensification of absolut (normally Force) goes unannotated. (4) …einfach absolut unlustig [nein1] “…simply altogether not funny”

Loving and hating the movies in English, German and Spanish 135



The main criterion to include a span as a markable was Attitude. Graduation was only annotated if it was associated to a span of Attitude. In addition, the Attitude span had to be clearly polar, that is, either positive or negative. The final system we used for annotation, excluding Engagement, but with an expanded Graduation system, is presented in Figure 2. affect ATTITUDE attitude-type Appraisal focus GRADUATION

ATTITUDE TYPE POLARITY FOCUS TYPE

judgement appreciation positive negative

sharpen

soften int-emphasizer INTENSIFICATION intensification TYPE int-downtoner FORCE force TYPE quantification QUANTIFICATION quant-emphasizer TYPE quant-downtoner

Figure 2.  Appraisal system used for the annotation.

Once the topics or aspects to be annotated have been decided, a number of other decisions affect the scope of the constituents, e.g. how much of the context of a particular word or expression should be included in the annotation. In general, we include only evaluative words and those words that are part of the evaluative content. For adjective+noun combinations, only the adjective is included, unless the noun also conveys evaluation. In ‘a good girl’, then, ‘good’ is the only markable. A few exceptions to this general rule are detailed below. All of the expressions of Attitude were classified as either positive or negative. In fact, the main criterion for deciding whether a unit was a markable (an expression of Attitude) or not was whether we could attribute polarity to it. In very few cases we felt that there was Attitude but no clear polarity. The decision in those cases was to either discard them as markables or to assign the most likely polarity. Below we outline the main areas where we had to make decisions as to what to annotate, or how to annotate it. 1. Coordinated elements with ‘and’ are part of the same span, but elements coordinated with commas or other punctuation are separate. This is because coordinated spans with ‘and’ tend to convey the same Attitude, whereas punctuation may indicate contrast rather than addition. Coordination with ‘and’ tends to refer to nouns and adjectives, but some cases merit the inclusion of two separate clauses in one span. In example 5, two clauses (“took the rights they had” and “ran with them”) are coordinated, with subject ellipsis in the second one. The Attitude (negative Judgement in this case) is similar. Moreover,

136 Maite Taboada, Marta Carretero and Jennifer Hinnell

the example is evocative of the idiomatic expression ‘take the money and run’, which would be treated as a single span.



(5) In THE CAT IN THE HAT, the producers and writers took the rights they had and ran with them, [no15]

This generalization also covers cases where the two coordinated elements are both modified by the same adjective that precedes the first element, as in example 6, since both adjectives are in the scope of the modifier.4

(6) zu zurückgezogen und introvertiert [nein10] “too withdrawn and introverted”



However, it can also be the case that two adjectives in a coordinated span each have different modifiers. In these cases the spans are annotated separately. A further exception to annotating coordinated elements as one span is in strings of more than two adjectives as in ‘X, Y and Z’, where the first pair of items are separated by a comma and the second pair are separated by the conjunction ‘and’. In these cases the adjectives are annotated individually to yield three separate spans, since there is less likelihood of all three conveying the same Attitude. 2. Comparatives and superlatives as well as discontinuous modifiers of intensity are included completely, because the term of the comparison or the part of the modifier that follows the head word is also part of the evaluative content. In example 7, the main evaluation is sosa (“dull”), but the discontinuous modifier as a whole contributes to understanding the intensity of that evaluation and is thus included. Similarly in example 8, where the main evaluation is dünn (“thin”), but the comparison “as thin as Lindsay Lohan’s hair extensions” identifies the intensity of the evaluation and is therefore included.

(7) …esta vez la trama es tan sosa que permanezco impasiva toda la película. [no1–3] “… this time the plot is so dull that I remain impassive throughout the movie.”

(8) Die Story ist so dünn wie Lindsay Lohan’s Haarverlängerungen [nein4] “The story is as thin as Lindsay Lohan’s hair extensions”

3. Modal auxiliaries to an evaluative verb are included in the markable. The modal verb often affects the Graduation conveyed by the verb.

(9) … it is fantastically watchable and jam-packed with witty one-liners and should appeal to a wide ranging audience. [yes3]

4.  Translations into English of the German and Spanish examples are our own.





Loving and hating the movies in English, German and Spanish 137

In the German example in 10, the modal möchte in möchte empfehlen (“would like to recommend”) downtones the recommendation.

(10) Ich möchte den Film empfehlen (er ist aber beileibe nichts für einen gemütlichen Familien Kinoabend) [ja12] “I would like to recommend the film (though it is by no means something for a cozy family movie night)”

4. An entire idiomatic expression is a markable, since it cannot usually be decomposed into constituent parts. In example 11, we see an idiomatic expression in Spanish, tirar por la borda (lit. “to throw something overboard”, i.e. “to not make good use of something”). (11) una historia que puede dar tanto de sí y que se tira por la borda [no2–19] “a story with so much potential and that is thrown overboard”



The entire idiomatic expression is considered as a markable, even if its constituents are separated by other constituents. That is to say, if the word order had been se tira una historia por la borda (lit. “is thrown a story overboard”), the span would have included la historia. An alternative would have been to consider se tira and por la borda as separate constituents. This procedure, however, would lead to an artificially high number of markables, since those two items are properly part of the same expression.

5. Repetitions are sometimes used for emphasis. They are considered as a single span, and the repetition is annotated as Graduation:Force:Intensification. (12) una película sosa sosa [no1–1] “a dull dull movie”

6. Conditionals and similar constructions recommending a movie are included in their entirety, since it is the combination of protasis and apodosis that conveys the evaluation. In example 13, it is the combination of the two clauses that results in a negative evaluation of the movie under review, “The Cat in the Hat”. (13) A dog even pees on someones food. If this is what Hollywood thinks is quality childrens entertainment… they are mistaken [no16]



This includes imperative-like conditionals, which occur frequently in the German corpus, as in example 14.

(14) Spart euch das Geld und die Zeit und investiert es sinnvoller [nein9] “Save yourselves the money and time and invest it more wisely”

138 Maite Taboada, Marta Carretero and Jennifer Hinnell



Also included as full spans are recommendations (or non-recommendations as in the case below), where the sentiment results from the main clause containing a setup, followed by a subordinate (here comparative) clause, as in example 15.

(15) Diese Aneinanderreihung von Banalitäten interessieren genauso viel wie die Tatsache, dass nach dem Kinobesuch mein Schuhband aufgegangen war. [nein1] “This sequence of banalities is about as interesting as the fact that my shoelace came undone after the movie.”

7. Causality. Causation contains two parts, one for the causer and another for the effect. The word expressing cause, if it is present, is excluded. But the division should be such that none of the spans is discontinuous. If the expression of Affect is modified by a comparison, it is an expression of Graduation under its span. In example 16, “this confusion” is the causer, and therefore not included. The span here is “bigger laughs”, the effect (Affect, with Force:Intensification). (16) However this confusion was the cause of some of the film’s bigger laughs [yes1]

8. Word order. In German spans are often quite long, due to aspects of morphosyntax such as the presence of separable prefix verbs, or the verb-second constraint. In the case of separable prefix verbs, the span has to include the verb stem since the prefix is not really intelligible without the verb, as in example 17, where the infinitive aussehen is separated to sieht … aus in its finite form. The example includes the separable prefix aus in the middle of the span, but also the verb sieht at the beginning, as the prefix without the verb would make no sense. Compare this with the translation of this span into English, where the verb ‘looks’ would not be included in the markable if this were an original English sentence. (17) Steven Seagal ist ganz schön fett geworden und sieht so schmierig aus wie ein Liter Motoröl [nein4] “Steven Seagal has gotten really fat and looks as slimy as a litre of motor oil.”



The verb-second constraint in German also created some overly long spans. For instance, in example 18, the inflected auxiliary (hätte; here inflected for tense and subjunctive mood) occurs in second position in the sentence, and the rest of the verb phrase, including a modal verb (dürfen “let”), is in sentence final position. This would normally mean that the full sentence is annotated; however, in order to avoid including unnecessary information in the markable, the auxiliary was omitted from the span, as the overall modality of the

Loving and hating the movies in English, German and Spanish 139



sentence was also present in the span through the modal. Note that in the English translation, the German double infinitive construction is conveyed only by the modal ‘should’. (18) Hier hätte man die Figuren durchaus stärker beleuchten dürfen [nein10] “Here one could have illuminated the characters more.”

9. Information structure. In German it became important to mark spans within spans. German tends to present more information in a single clause; English by contrast favours subordinate clauses for this purpose. In English, the sentence in example 19 would have been more authentically written as example 20. In general, German tends to be accommodating of long passages, so double-layered annotations such as the one in example 19 were needed more frequently. (19) Barbara Hershey als Ninas Mutter […] ist ebenso überzeugend wie der großartige Vincent Cassel als Mischung aus gnadenlosem Intendanten und erotischem Lehrer. [ja15] “Barbara Hershey as Nina’s mother is just as convincing as the amazing Vincent Cassel as a mix of merciless director and erotic teacher” (20) Barbara Hershey as Nina’s mother is just as convincing as Vincent Cassel, who was amazing as a mix of merciless director and erotic teacher.

We completed several iterations of the annotation, comparing first the markables, to check that the spans selected were comparable, and then the annotations themselves, to make sure we were consistent in our annotation procedure. We conducted two experiments with three reviews each, in addition to previous experiments for an English-Spanish comparison (Carretero and Taboada, 2014). We also created an annotation manual, to record our general principles. Once we were satisfied with the level of comparability across annotators, we proceeded with the annotation of the full corpus. Then, each annotator sent half the annotation that she had carried out to each of the other two annotators for revision; in this way, all the reviews were annotated by an annotator and revised by another. The results of this final annotation are described in the next section. 4. Results Table 2 summarizes the number of spans annotated for each language, and the breakdown in terms of the three main categories annotated (Affect, Judgement and Appreciation within the Attitude category). The analysis was also broken down into reviews that provided a favourable assessment of the movie (‘positive’) and those that provided an unfavourable one (‘negative’).

140 Maite Taboada, Marta Carretero and Jennifer Hinnell

Table 2.  Summary of Appraisal realization in the three languages. English n

German %

n

Spanish %

  1,882

n

%

Spans (Attitude)

  1,312

Affect

   203

15.47

   121

  6.43

   389

19.95

Judgement

   400

30.49

   447

23.75

   481

24.67

Appreciation

   709

54.04

  1,314

69.82

  1,080

55.38

Graduation

   553

42.14

  1,030

54.73

   965

49.49

Total number of words

33,387

47,128

  1,950

52,845

The reviews, overall, are dominated by spans of Appreciation. This is to be expected, as the reviews are about objects of art, the prime example of Appreciation. Nevertheless, Affect and Judgement contribute to the total, with a somewhat surprising higher number of spans of Judgement. One may expect that Affect would be second after Appreciation, since the reviews are likely to convey the emotions that the movies triggered in the authors. However, Judgement about the cast and director, and to a certain extent about the characters, takes precedence, and for German the Judgement proportion more than triples that of Affect. In other words, the reviewers focus on the object to be reviewed, including cast and director (and sometimes criticism of audience members), over their own emotions when watching the movies. We could say that the reviews are not about loving and hating the movies, but about the artistic merit of the movies as works of art. The rest of this section summarizes results for each language. 4.1 English The reviews in English are, as in the other languages, dominated by spans of Appreciation. Table 3 shows that, in most cases, the distribution of the three subcategories is very similar in positive and negative reviews. In both, Appreciation still accounts for the majority of Attitude spans, followed by Judgement, with Affect a distant third. Unlike in Spanish, the numbers of spans in positive and negative reviews are quite similar, as is the number of words in each (17,034 in positive reviews; 15,971 for negative). Naturally, the proportions of positive and negative spans are reversed in each type of review. However, positive reviews still contain a fair amount of negative evaluation (over 21%), and, conversely, negative reviews contain about 25% of positive spans. The reasons for this are twofold. First of all, even though the overall



Loving and hating the movies in English, German and Spanish 141

Table 3.  Appraisal in positive and negative reviews (English). Positive n

Negative %

n

%

Spans (Attitude)

649

663

Affect

109

16.80

  94

14.18

Judgement

195

30.05

205

30.92

Appreciation

345

53.16

364

54.90

Polarity

649

Positive

509

663 78.43

167

25.19

Negative

140

21.57

496

74.81

Spans w/o Graduation

362

55.78

397

59.88

Spans w. Graduation

287

44.22

266

40.12

Focus

  38

13.24

  48

18.05

Force

249

86.76

218

81.95

Sharpen

  32

84.21

  32

66.67

Soften

   6

15.79

  16

33.33

Intensification

173

69.48

136

62.39

Quantification

  76

30.52

  82

37.61

Int. emphasizer

113

65.32

106

77.94

Int. downtoner

  60

34.68

  30

22.06

Qu. emphasizer

  50

65.79

  72

87.80

Qu. downtoner

  26

34.21

  10

12.20

Focus type

Force type

Intensification type

Quantification type

evaluation may be, for instance, negative, there are often positive aspects in the movie that the authors point out. The second related reason is that the reviews (in particular negative reviews) typically follow a rhetorical pattern where positive aspects are mentioned first, and then the writer moves on to negative ones. This is a common pattern in (North American) English discourse, in both written and spoken informal discourse. A negative evaluation by itself is perceived as too blunt, and is often softened by a few initial words of praise. A short example of this pattern is presented in example 21, part of a negative review. This is the second paragraph in the review, which the author uses to mention “the good stuff ”. The

142 Maite Taboada, Marta Carretero and Jennifer Hinnell

rest of the review contains a negative evaluation of the plot, the length and the movie overall. Even within this paragraph, the pattern is repeated, with a positive evaluation first and a criticism second. (21) The good stuff….the visual production itself with its ultra-stylized appearance. It looks nice, but did the budget of a third world nation need to be spent to create this film? [no17]

This type of pattern tends to contain expressions of Attitude but, as a whole, may also be seen as part of the Engagement system in Appraisal. Martin and White (2005: 124–126) discuss pairings of concurring opinions, followed by a counter move, and they state that these options can be characterized as concessions.5 In their view, the concurring move (the positive evaluation) is provided reluctantly, as a rhetorical move where ground is initially given, only to be taken away. In our examples, reluctance does not seem to be present; rather the positive evaluation seems genuine. We argue that the significance lies in the way in which the combination of evaluations is expressed, with the positive evaluation first: there seems to be a need to present that before negative evaluation can be introduced. Carretero and Zamorano (2013) discuss this connection between modal adverbs (part of the Engagement system) and concession. Itakura and Tsui (2011), in a study of English and Japanese book reviews, also found that praise often preceded criticism in the English reviews. This positive-first, negative-mostly pattern does not seem to be as widespread in its reverse form when the overall evaluation is meant to be positive, but we still find some examples with a similar structure. In example 22 we can read the first few sentences of a very positive review. It starts with the author’s misgivings about the film before he or she saw it, but soon the tone changes to one of positive evaluation. (22) I’ve got to admit that the first time I saw the trailer for Calendar Girls I wasn’t exactly charged with anticipation. The trailer did manage to make the film look fun, but the concept strongly reminded me of The Full Monty, which is a film I personally found a little overrated. If I wanted to add further concerns I could tell you about how I decided to search Epinions for some reviews on the film, and my search revealed nothing but a small selection of more, adult titles. Nevertheless I recognized that not only did the trailer look fun, but I had been pleasantly surprised by another British comedy that didn’t appeal to me, Bend It Like Beckham. With those thoughts running through my mind I decided to throw caution to the wind and go down to the local cinema to see Calendar Girls. I’m glad I went to see it now, because the film has remained just that. Fun! … [yes1] 5.  Thanks to Cristina Boccia for pointing this out, in response to a query in the Appraisal list.



Loving and hating the movies in English, German and Spanish 143

Expressions of Attitude were realized by different parts of speech, with adjectives being most frequent (‘convincing’, ‘worth seeing’, ‘pathetic’, ‘irritating’), followed by verbs, including verbs that denote mental processes (‘really enjoy’), behavioural processes (‘hadn’t laughed that hard or that consistently at just a preview in a very very long time’) and other processes (‘groaned’, ‘hailed’). Nouns (‘masterpiece’, ‘clichés’, ‘scene-stealer’) and adverbials (‘feebly’, ‘willy-nilly’) are also frequently used. In some cases, an entire sentence conveys the opinion and is therefore a span (‘Go see this movie’). There are frequent ready-made phrases (‘two thumbs way up’), but also interesting new expressions (‘[Jon must] think that if he chops up a story up into a bunch of little skits, inserting a ton of physical comedy, then a movie will magically rise from it all’). 4.2 Graduation in English About 42% of the expressions of Attitude in the English reviews contained some type of Graduation. We devote this section to the complexities of Graduation. The overwhelming majority of expressions of Graduation convey Force, rather than Focus, in both the positive and negative reviews. Force has two different aspects, Intensification and Quantification, with the former being much more frequent, in similar proportions for the negative and positive reviews. It seems to us that this may be a general phenomenon in English. Furthermore, emphasizing intensifiers are more common than downtoning intensifiers. Intensifiers are typically adverbs of Graduation, such as ‘very’, ‘incredibly’, ‘extremely’ or ‘somewhat’, ‘kind of ’/‘kinda’. In some cases, more creativity is deployed, as in example 23, where an expression of attitude (“hilariously”) becomes an expression of Graduation and intensifies the adjective “funny”. In this example, and all that follow, underlining indicates the entire markable, and bold the part that conveys Graduation. In example 24, the concession expressed through “still” serves as a downtoner. (23) That’s what makes Bad Santa such a hilariously funny movie… [yes21] (24) It doesn’t really deserve many nominations, much less the wins it’s somewhat likely to receive. Nevertheless, it’s still above-average in my opinion, and though it has a lot going against it, it has a lot going for it as well. [yes6]

Most expressions specialize and act as either intensifiers or downtoners, but ‘few’ and ‘a few’ appear in both cases. For instance, ‘a few mistakes’ is a downtoned expression, whereas ‘a few brilliantly crafted scenes’ seems to use ‘a few’ as an intensifier, since it suggests that the existence of several brilliantly crafted scenes is appreciated, due to the scarcity of such high quality scenes.

144 Maite Taboada, Marta Carretero and Jennifer Hinnell

As for Quantification, the most common realization is in superlatives (‘one of the best movies out’), expressions with ‘too’ (‘too many’, ‘too much’), and expressions of quantity (‘exactly one funny moment’). Example 25 quantifies by referring to audience age. (25) I know that, even at 11, I would have been disturbed by the sexual jokes in it, had I understood them. [no15]

In both Intensification and Quantification, emphasizers are more numerous than downtoners. This seems to be a result of the straightforward nature of the reviews, which tend to lack subtlety. When downtoning is used, it tends to be an instance of litotes, the negation of the opposite opinion, such as ‘not really good’, where ‘really’, usually an intensifier, becomes a downtoner because it is part of the negation. Example 26 shows Quantification realized as a conditional clause. The positive assessment of “good” is quantified, and downtoned, by the conditional clause “only if you know when to stop”. Similarly, in example 27, the endorsement in “suggest” is downtoned by the conditional. (26) OK, I’ll give credit here, although I really didn’t like the path they took to get there, the story line was basically the same. Fun is good, but only if you know when to stop. [no14] (27) As for home viewing, I would suggest it if you have a good bass backing. [yes20]

We have found that conditional and concessive relations have an important role to play in the interpretation of evaluation, but that clearly determining how the relation affects the evaluation is quite complex. See Trnavac and Taboada (2012) for a study of such relations in this corpus. Within Focus, most of the instances correspond to Sharpen. Interestingly, negative reviews contain a few more expressions of Soften than positive ones. We believe that this is because writers are more likely to soften a negative opinion than a positive one. In the following examples, we see softening expressions. An interesting case is example 28, where a positive evaluation is softened by “depending on what film she does”, but preceded by a strong endorsement, “without a doubt” (which is not part of the markable because it conveys Engagement). (28) Julia Roberts without a doubt, is a good actress depending on what film she does. [no1] (29) The so-called ‘humor,’ as mentioned above, is so pandering, so foul, and so UN-SEUSS, [no10] (30) The setting wasn’t exactly what I had envisioned either. [no13]



Loving and hating the movies in English, German and Spanish 145

Most of the cases of Graduation involved one of the two systems, either Force or Focus. In some cases, however, both types were present. As the CorpusTool annotation program does not allow for double annotation of items within the same system (see Section 3.2), we had to choose the main label under which to classify these cases. We made a note of them, and found in total about 20 instances in English. In example 31, we find a Graduation of “original” both in terms of Force:Quantification (“wholly”) and Force:Intensification (“at all”). We gave priority to Force:Quantification for “wholly” because it has scope over the entire expression. (31) It may not be wholly original at all [yes8]

4.3 German The German reviews yielded almost 1,900 spans. Like English (and Spanish), the German reviews are dominated by Appreciation spans. Notably, however, Appreciation accounts for almost 70%, whereas for the other two languages Appreciation spans account for around 55%. The higher percentage of Appreciation spans in contrast to the other languages is balanced in the lower percentage of Affect spans in German, at 6.43%. Table 4 shows the distribution of spans across positive and negative reviews. With 214 additional spans in the negative reviews, there is a greater difference in the number of spans in the positive and negative contexts than in English (almost equal) and Spanish (88 more spans in the positive reviews). Word count differences are more pronounced than they are in the English corpus, but closer than in the Spanish reviews, with 25,401 words in the positive reviews, and 21,727 in the negative. In German, there is only a small difference in the occurrence of negative comments in positive reviews and vice versa, with marginally more positive comments in the negative reviews (21.37% vs. 19.80%). Thus the positive-first pattern for negative comments described above for English does not hold to the same degree. Rather, it seems that Germans tend to balance their negative comments in positive reviews just as much as their positive comments in negative reviews. Whereas positive reviews contain more Appreciation, negative reviews contain a higher percentage of both Judgement and Affect. In fact, there are two negative reviews with more Judgement than Appreciation. This means that rather than appraising the film, the idea, the acting, etc., the writers of negative reviews direct more comments towards the creators of these elements, i.e. the director, the scriptwriter, the actors. The following example demonstrates the positive/negative alternation mentioned above. It also shows verbs frequently used to express Judgement

146 Maite Taboada, Marta Carretero and Jennifer Hinnell

Table 4.  Appraisal in positive and negative reviews (German). Positive n Spans (Attitude)

834

Affect

  47

Negative %

n

%

1,048   5.60

   74

  7.06

Judgement

170

20.40

  277

26.43

Appreciation

617

74.00

  697

66.51

Polarity

834

Positive

669

1,048 80.20

  224

21.37

Negative

165

19.80

  824

78.63

Spans w/o Graduation

374

44.84

  477

45.52

Spans w. Graduation

460

55.16

  571

54.48

Focus

  97

21.10

   93

16.29

78.90

  478

83.71

Force

363

Focus type

  97

Sharpen

  79

81.40

   72

77.42

Soften

  18

18.60

   21

22.58

   93

Force type

363

Intensification

220

60.60

  281

  478 58.79

Quantification

143

39.40

  197

41.21

Intensification type

220

Int. emphasizer

205

93.20   6.80

  281   245

87.19

   36

12.81

Int. downtoner

  15

Quantification type

143

Qu. emphasizer

104

72.70

  118

59.90

Qu. downtoner

  39

27.30

   79

40.10

  197

in the German reviews, namely gelingen (“succeed”) (or gelingt nicht, “not succeed”), and scheitern (“to fail”). (32) Grundsätzlich ist die Idee der Story ja wirklich sehr interessant und könnte auch wahrscheinlich richtig gut auf die Leinwand gebracht werden. Aber Regisseur und Drehbuchautor George Nolfi gelingt dies leider nicht. [nein14] “The idea for the story is fundamentally really very interesting and could probably come across really well on screen. But director and script-writer George Nolfi unfortunately doesn’t achieve this.”



Loving and hating the movies in English, German and Spanish 147

Adjectives can also be used to express Judgement, as in example 33, where the comment is about the director’s achievement. (33) Der gelungene Umgang mit der Hauptfigur… [nein10] “The successful manipulation of the main character…”

Schauspielern (“to act”) is similarly used to express Judgement, as in example 34, and here the polarity is added by the modal kann (“can”). The positive comment preceding the negative evaluation highlights the contrast between the author’s expectation and her evaluation of the actor’s performance. (34) Guillaume Canet kann eigentlich schauspielern. Warum er in diesem Fall so zurückhaltend ist, so unglaublich falsch und deplatziert wirkt und vor allem so verklemmt, dass weiß kein Mensch. [nein21] “Guillaume Canet can act. Why in this case he is so withdrawn, comes across as so unbelievably badly cast and out of place, and above all so uptight, no one knows.”

Modals and full sentences are also used to express Judgement, as in examples 35 and 36, respectively. (35) Es bleibt einfach langweiliger Schrott, den man nicht filmen sollte. [nein21] “It is simply boring junk, that no one should film.” (36) Das hätten auch noch Erstsemester an einer Filmakademie mit Laiendarsteller geschafft [nein21] “This could also have been achieved by first-semester students at a film school with amateur actors.”

In German, Affect was frequently marked by the very colloquial expression schade, (“too bad”, “what a shame”, “bummer”), often standing on its own, or incorporated at the beginning of a sentence as in example 37, where the annotation is for negative Affect, and Focus-Sharpen. (37) Schade eigentlich, ich hatte mir echt mehr versprochen. [nein10] “Too bad, really. I had expected a lot more.”

4.4 Graduation in German Annotating for Graduation in German proved to be very complex. First, we examine the distribution presented in Table 4. Then we will elucidate some of the complexities particular to German. As Table 4 shows, there were 460 instances of Graduation annotated in the positive reviews, and 571 in the negative reviews. Of these, the overwhelming majority were Force (over 75% in both cases). Focus

148 Maite Taboada, Marta Carretero and Jennifer Hinnell

accounted for the remainder, 21.10% in the positive and 16.29% in the negative reviews. As for Force type, German, too, showed that Intensification was more prevalent in both the positive and negative reviews, though the difference between the frequency of Intensification and Quantification achieve was not as great as in English. With regard to emphasizing and downtoning Force, emphasizers are, as in English, more frequent than downtoners. This pattern is more marked for Intensification than for Quantification, where intensifiers were downtoned only 6.8% and 12.81% in the positive and negative reviews respectively, and quantifiers 27.3% and 40.1% respectively. Finally, Focus was more often sharpened than softened, and in comparable frequencies in both sets of reviews. In the German corpus, Intensification emphasizers were typically adverbs such as deutlich (“clearly”), wirklich (“really”), comparatives and superlatives, temporal adverbial phrases such as zu keiner Zeit (“at no time”) and either/or expressions (weder/noch) and other phrases such as alles andere als (“anything but”). Nie (“never”) was a frequent emphasizer, as it is more emphatic than the regular negation with nicht (“not”), as is the commonly used überhaupt nicht (“not at all”). There were also more creative emphasizers, as in the Intensification mit Müh und Not in example 38, which emphasizes the negative evaluation of the review. (38) Mit Müh und Not einen von fünf Felsblöcken, die alles unter sich begraben, was diesen Film gut hätte werden lassen können. [nein2] “Barely [lit., “with effort and hardship”] one out of five stones [stars], which bury everything that could have made this film a good one.”

Common downtoners included modals, as in example 39. (39) Ich möchte den Film empfehlen (er ist aber beileibe nichts für einen gemütlichen Familien Kinoabend) [ja12] “I would like to recommend the film (though it is by no means something for a cozy family movie night)”

The German particle zwar sometimes functions as a downtoner, since it adds a sense of hedging to the statement, as in example 40. (40) Der Handlungsverlauf ist zwar vorhersehbar [ja10] “While the plot development is predictable”

Quantification included adjuncts relating to an amount or quantity, broadly construed, such as mehr oder weniger (“more or less”), mehr und mehr (“more and more”), zeitweise (“occasionally”) or a whole, such as völlig, komplett, ganz, and gesamt, all of which can be loosely translated as “entirely”, “completely” or “absolutely”. Temporal adverbs such as ab und an/ab und zu (“now and then”) also downtoned Quantification. In negative contexts Quantification also included gar



Loving and hating the movies in English, German and Spanish 149

nicht (“not at all”) and an keiner Stelle (“at no point”), which is similar to the absolute negative nie (“never”), and kaum (“hardly”). Phrases such as auf ganzer Linie (“in its entirety”) that refer to a quantity were also common.6 The German negation particle kein proved complex to annotate. Depending on context it can operate as a simple negator with no Graduation, where it negates a noun (kein gutes Schauspielen is simply “not good acting”, where the evaluation is a straightforward negative polarity). At other times kein evokes Quantification, as in example 41. Here, a simple negation would have negated the sentence using nicht, as in er hat nicht das Gefühl (“he didn’t have the feeling”), rather than “at no point did he have the feeling”. (41) Man hat an keiner Stelle das Gefühl, auf die beiden würde ein lohnenwertes Ziel warten [nein12] “One has at no point the feeling that a worthwhile goal awaited the two [characters]”

Kein was also involved in many cases of Intensification as a downtoner, as shown in example 42. (42) …keine wirklich grosse Überraschung [nein15] “…not a really big surprise”

Litotes, the practice of expressing an affirmative by negating its opposite, was a tricky phenomenon to annotate in German. We decided, for all languages, not to annotate litotes on its own, as it is too difficult to determine what the alternative to the negative statement is. However, very often litotes expressions included Quantification by using words like viele (“many”). When these quantity expressions were negated, they served as downtoners and required annotation, as in example 43. (43) Ich erwarte eigentich nicht viel [nein1] “I actually don’t expect much”

This could also have been expressed in the positive, as either ‘I expect little’ or ‘I expect something’. As mentioned, the affirmative is hard to deduce precisely from the negative. However, here it is annotated as negative/downtoned due to the quantity word nicht viel (“not much”).

6.  Note that one of the complexities in annotating Graduation in German was whether to consider emphasized negation using nichts and the phrases gar nicht and überhaupt nicht as Quantification or Intensification. The judgements in these cases depended on whether the noun or verb that was being negated could be construed to have quantity in any sense.

150 Maite Taboada, Marta Carretero and Jennifer Hinnell

It was difficult to assess litotes involving kein. As mentioned previously, kein can indicate Quantification emphasis (as in English “none”), or a simple negation (“no”). It does serve in litotes, as shown in example 44. However, where the sentence negation occurs before an intensifier or Focus-sharpening adverb, as in example 45, kein serves as a simple negator. (44) kein so schlecht gewählter Titel [nein16] “not such a badly chosen title” (45) Nolfi (direktor) verfolgt in meinen Augen kein richtiges Ziel fuer seine Geschichte [nein14] “In my eyes, Nolfi (the director) doesn’t follow a real goal for his story”

For the first of the pair, the alternative would have been to say “this is a well-chosen title”, which results in the example being annotated as positive Appreciation with no Graduation. The second, however, is not litotes and is simply negative Judgement with a sharpened Focus resulting from richtiges (“real”). Concerning the two categories of Focus, Focus:Soften was marked in German primarily by particles eher (“rather”), etwas (“somewhat”), eine Art (“a kind of ”). Focus:Sharpen was characterized by the adverbs einfach (“simply”), often in combination with nur, as in einfach nur (“simply/just”), eigentlich (“actually”), ziemlich (“quite”), and ganz (“quite”) as in ganz gut, “quite good”. Note that this use of ganz differs from the meaning listed above under Quantification in which it refers to a whole. Finally, there were frequently two types of Graduation in a single phrase. As agreed, only the type with the widest scope was annotated. In example 46 below, in manchen Stellen (“in some places”) has wider scope than the intensified expression zu zurückgezogen und introvertiert (“too withdrawn and introverted”). Thus in example 46 the span is marked for Force:Quantification:Downtoner of in manchen Stellen, and not Force:Intensification:Emphasizer. (46) …in manchen Szenen in meinen Augen zu zurückgezogen und introvertiert [nein10] “…in some scenes in my opinion too withdrawn and introverted”

In sum, the German reviews were characterized by a higher proportion of Appreciation spans than English and Spanish; however, negative reviews favour Judgement and Affect more than positive reviews do. Furthermore, Graduation in German proved to be complex and very dependent on context, rather than on particular phrases. Unique to German in particular were the nuances described above in annotating negation with the particles nicht and kein.



Loving and hating the movies in English, German and Spanish 151

4.5 Spanish The results of the annotation of the Spanish movies are displayed in Table 5. The number of spans for positive reviews (1,019) is higher than for negative reviews (931); however, we must remember that the total number of words of the positive reviews is much higher: the ratio of words per span is 35.40 in the positive reviews and 18.02 in the negative reviews. This difference in density is largely due to the greater length of the description stages in the positive reviews, in which reviewers, often driven by their enthusiasm, write lengthy paragraphs that contain few or no evaluative spans about the plot or technical details. Table 5.  Appraisal in positive and negative reviews (Spanish). Positive

Negative

n

%

n

%

Spans (Attitude)

1,019

52.26

931

47.74

Affect

  235

23.06

154

16.54

Judgement

  202

19.82

279

29.27

Appreciation

  582

57.11

498

53.49

Polarity

1,019

Positive

  843

931 82.73

204

21.91

Negative

  176

17.17

727

78.09

Spans w/o Graduation

  494

48.48

479

51.45

Spans w. Graduation

  513

50.34

452

48.55

Focus

   50

  9.75

  47

10.40

90.25

405

89.60

Force

  463

Focus type

   50

Sharpen

   26

52.00

  24

51.06

Soften

   24

48.00

  23

48.94

Intensification

  284

61.34

227

56.05

Quantification

  179

38.66

178

43.95

Int. emphasizer

  208

73.24

165

72.69

Int. downtoner

   76

26.76

  62

27.31

  47

Force type

Intensification type

Quantification type Qu. emphasizer

  139

77.65

150

84.27

Qu. downtoner

   40

22.35

  28

15.73

152 Maite Taboada, Marta Carretero and Jennifer Hinnell

With regard to the kind of spans, Appreciation amounts to more than half (57.11% in the positive reviews and 53.49% in the negative), which is not surprising, since the main object of the reviews are non-human entities (the movies). The next most frequent spans are those of Judgement, which, summing up the two types of reviews, total 481, while those of Affect add up to 389. The percentage of Appreciation is 3.62% higher in the positive reviews. The difference is larger in the case of Affect, whose percentage is 6.52% higher in the positive reviews: writers of positive reviews tend to express their positive feelings (and, less commonly, those of other viewers) by several kinds of devices: verbs of mental processes of emotion such as gustar (“like”), encantar (“love”), disfrutar (“enjoy”), alegrar(se) (“make glad”/“be glad”), emocionarse (“get excited”), sorprender (“surprise”) or denials of negative verbs such as no defraudar (“not dissapoint”), adjectives such as contento (“glad”), identificado (“identified”), satisfecho (“satisfied”), fascinado (“fascinated”) or atento (“attentive”); nouns such as sonrisa (“smile”), risa (“laugh”) or encanto (“charm”), or idiomatic expressions such as partirse de risa (“burst out laughing”) or saltarse las lágrimas (“shed tears”). In contrast to Affect and Appreciation, Judgement is over 10% more frequent in the negative reviews. Reviewers tend to evaluate the unsatisfactory performance of the director or the cast with a number of common devices: adjectives such as pobre (“poor”), pésimo (“very bad”) or absurdo (“absurd”); nouns or nominal expressions such as mediocridad (“mediocrity”), poca credibilidad (“little credibility”) or verbs referring to actions concerning a better original book or script, such as destrozar (“spoil”) or destruir (“destroy”). Some expressions refer to moral issues, such as sabe embaucar (“knows how to fool”) or auténtico timo (“real swindle”). In certain cases, the spans occupy whole sentences, as in Yo con mi cámara super 8 y un muñeco de plastilina lo hubiera hecho mejor (“I would have done a better job with my super 8 camera and a play-doh puppet”). Concerning polarity, positive spans outnumber negative ones. In the positive reviews, positive spans sum up more than 80% of the cases for the three subtypes of Attitude, while in the negative reviews negative spans amount to 78% of the cases. The percentage of spans with contrary polarity to that of the overall review is slightly smaller than for the English reviews, both in the positive and the negative reviews. The rhetorical pattern of initial spans with contrary polarity for the English reviews, mentioned in Section 4.1, was not found to be frequent in the Spanish reviews. Rather, these spans occur in no fixed position, and mostly refer to: individual performances of actors or actresses, as in example 47, which was found in a negative review; concrete aspects or parts of the movie that do not match its overall quality, such as example 48, which again was in a negative review; or reported opinions by other people than the reviewer, as in example 49.



Loving and hating the movies in English, German and Spanish 153

(47) Beth, interpretada por Julie Cox, me sorprendió por su excelente interpretación [no2–12] “Beth, interpreted by Julie Cox, surprised me with her excellent performance.” (48) muy bien a nivel de efectos especiales y música [no2–20] “very well with regard to special effects and music” (49) La película dura 2 horas y 20 minutos. Hay gente que dice que se le hace pesado porque están acostumbrados/-as a ver un solo capítulo. [yes5–5] “The movie lasts 2 hours and 20 minutes. There are people who say that it feels tedious because they are used to seeing one chapter only.”

4.6 Graduation in Spanish With regard to Graduation inscribed in Attitude, about half the spans are graduated (50.34% for positive reviews and 48.48% for negative reviews). Within Graduation, roughly 10% of the spans display Focus and 90% display Force for both kinds of reviews. Further quantitative research on the cases of Focus showed that, out of the total 97, 65 of them (67.01%) correspond to spans of Appreciation. That is, Focus occurs mostly in spans that evaluate the movie itself or other nonhuman entities, or else aesthetic features of the movies. Examples of these spans of Appreciation graduated with Focus are auténtico disparate (“real nonsense”), patética por un lado (“pathetic on the one hand”), sencillamente magistral (“simply masterful”), perfectas en todos los aspectos (“perfect in all aspects”). With regard to the subtypes of Focus, cases of Sharpen occur with roughly the same frequency as those of Soften. Although the quantity of Focus spans in Judgement is small, totalling 16 occurrences, it is worth pointing out that 12 of them (75%) are of Sharpen. That is to say, the spans concerning the director of the cast of the movie, as well as ethical issues, are more often emphasized than downtoned in terms of prototypicality (including completeness). Some examples of Judgement spans graduated with Focus:Sharpen are completamente prohibidas (“completely forbidden”), totalmente plana (“totally flat”) or claro ejemplo de amistad, de superación y de optimismo (“clear example of friendship, of will and of optimism”). Concerning Force, Intensification predominates over Quantification in both categories, especially in the positive reviews. This predominance is particularly strong for spans of Judgement in the positive reviews (71.76%). These often display favourable evaluative spans about the cast consisting of adjectives or adverbs modified by intensifying adverbs, mostly muy (“very”) or tan (“so, such”). Other common devices are the span with the superlative adjective el mejor (“the best”) and nouns modified by the adjective gran. Examples of these spans are muy creíble

154 Maite Taboada, Marta Carretero and Jennifer Hinnell

(“very believable”), muy, muy simpático (“very, very friendly”), tan bonito (“so pretty”), tan bien (“so well”), gran magnetismo (“great magnetism”) or gran estrella (“great star”). Affect differs from the other subcategories in that it displays a predominance of Quantification over Intensification (52.00% in the positive reviews and 60.94% in the negative ones). The reason is that Affect is commonly realized by verbs of emotion modified by expressions of quantity. Instances of spans of this kind are gustó mucho (“pleased a lot”), disfruté mucho muchísmo (“I enjoyed a lot really a lot”), esperaba bastante (“I expected quite a lot”), tanto me defraudó (“disappointed me so much”). With regard to the subtypes of Intensification, emphasizers are more common than downtoners. That is to say, reviewers tend to reinforce their evaluations rather than to downtone them. The global percentages are similar in the positive and negative reviews. Research on subcategories of Attitude shows that this predominance is lowest in Affect, especially in the positive reviews (the number of downtoners is 18, totalling 37.5%), due to the frequent cases of litotes, i.e., downtoning by means of negation, such as la sensación final no fue la que deseaba (“the final impression was not what I wanted”), no me ha desagradado (“I haven’t disliked it”), or no es un film que te asombra por su excelencia (“it is not a movie that astonishes because of its excellence”). The proportion of emphasizers is highest in Judgement, especially in positive reviews (51 cases, 83.61%), due to the frequency of intensified spans such as those mentioned in the previous paragraph. From a qualitative point of view, it must be noted that the partitive turns an evaluative span with a superlative expression as a downtoner, as in example 50. (50) De lo que más me ha gustado son las conversaciones que tiene con Kevin [no1–17] “Among the things that I liked best are the conversations that she has with Kevin”

Concerning Quantification, emphasizers predominate over downtoners, which is consistent with the reviewers’ tendency to reinforce their evaluations. This predominance is almost the same as that found in Intensification in the positive reviews, but it is markedly higher in the negative reviews. The reason is that quantifying emphasizers are often used in negative spans so as to create a humorous effect, as in todo decaía (“everything was decaying”), se pasan la mayor parte del tiempo a grito limpio (“they spend most of the time shouting like mad”) or nada ni nadie en la película pasa de ser plano como el encefalograma de un muerto (“nobody or nothing in the movie is anything but flat like a corpse’s encephalogram”). As a global comment, we may state that the most striking differences in the distribution of subcategories were found between spans of Affect and those of Judgement, Appreciation being in a middle position in many respects.



Loving and hating the movies in English, German and Spanish 155

5. Discussion and conclusions In all three languages, the distribution of Appraisal categories was found to be similar, with Appreciation being the most frequent type of Attitude, followed by Judgement and then Affect. In other words, the reviews are not actually about ‘loving and hating’ movies, but rather about movies as works of art. Expressions of love and hate, which would be categorized as Affect, are actually quite infrequent. In Spanish,7 positive reviews may be considered as an exception to this tendency. As we discussed in Sections 4.5 and 4.6, expressions of positive feelings did occur with some frequency. However, there is a caveat: the two most common positive verbs of Affect, gustar (“like”) and encantar (“love”) have a different syntactic pattern from that of their English equivalents. Their pattern resembles that of the English verb ‘please’, in that the subject is the entity that provokes the feeling and the object is the entity that feels. No occurrences were found of the verb amar, whose syntax is the same as that of its English equivalent ‘love’. In this way, the role of the movie (or an entity related to it) as initiator of the emotion, is emphasized, while viewers play a receptive role as undergoers of these emotions. Thus, in spite of the relatively frequent expressions of Affect, this does not mean that Spanish positive reviews are mainly about loving or hating movies; rather, writers recall being pleased, impressed or moved by them. This perspective is also visible in the Spanish negative reviews, in which the overall percentage of Affect was lower. No occurrences were found of odiar, the most direct correlate of the English verb ‘hate’, nor of any close synonym such as detestar (“detest”) or aborrecer (“loathe”). Instead, negative Affect was very commonly expressed by negating gustar; to a lesser extent, other expressions were also used, such as defraudar (“disappoint”), aburrirse (“get bored”) or enfadado (“angry”). It might be suggested that the prevalence of the gustar pattern in the expression of Affect in Spanish could possibly result in an overall higher frequency in comparison to other languages: cross-linguistic research of Appraisal in different languages could well confirm this hypothesis. The genre that all the texts belong to seems to drive the types of Attitude being expressed. Similarly, the high level of Graduation (over 40% for the three languages), is a reflection of the evaluative nature of this genre, in which reviewers feel the need to graduate the strength or the applicability of the evaluative expressions that they use. In the three languages, Intensification is more common than 7.  In this section, reference to the languages is a shortcut indicating the language, as shown in our corpus. That is, when we say “Spanish”, we mean “the Spanish data in our corpus”. We believe that we cannot make any generalizations from this sample of this particular genre to the three languages/cultures as a whole.

156 Maite Taboada, Marta Carretero and Jennifer Hinnell

Quantification, and this predominance is stronger in the positive than in the negative reviews. This suggests that many negative spans express a semantic element of ‘not anything’ (quantifying emphasizer) or ‘not much’ (quantifying downtoner). As well as these cross-linguistic similarities, we also found some interesting differences across the three languages. In particular, there are a number of characteristics of the German reviews that vary from their English and Spanish counterparts. Differences in span length resulted from the particulars of German word order, in addition to the fact that, generally, German sentences are more accommodating of long passages. In some instances, double layers of annotations were required. In comparison to the other languages, the distribution of Evaluation in German was also different. Appreciation spans account for almost 70%, where for the other two languages the spans account for around 55%. Affect spans are markedly less frequent than in the other two languages. Positive reviews contained more Appreciation and less Judgement than negative reviews. The remarkably high percentage of Judgement in negative reviews, as compared to positive reviews, is shared by the Spanish part of the corpus. This suggests that, in these two languages, criticism is expressed with regard to the (lack of) achievement or performance of individuals such as the director or the cast members, while expressions of praise focus on the object of success, such as the camera work, script, or scenery (rather than the person). No clear lexicogrammatical factors were found to account for this difference, which seems then to be due to sociocultural reasons. Another similarity between the Spanish and German negative reviews is the almost identical distribution of positive comments in negative reviews (though the Spanish positive reviews have fewer negative comments than the German ones). The English data showed a higher frequency of positives in negative reviews than the other way around. Thus the positive-first feature described for English is less striking in German and Spanish. As discussed earlier, Graduation in German proved complex. In fact, over 54% of the spans in the German reviews contained Graduation as compared to about 42% for English and approximately 49% for Spanish. Though in all languages the majority of cases were Force (in which Intensification was more frequent than Quantification), there was more Focus (18.7%) in the German reviews than in English (15.6%), and the difference is even larger with Spanish (under 10%). The highest percentage of Graduation in German may be explained by the existence in this language of many particles that serve to sharpen and soften Focus, such as eher (“rather”), etwas (“somewhat”), einfach (“simply”), einfach nur (“simply/ just”), eigentlich (“actually”), ziemlich (“quite”). German also differs from English and Spanish in the distribution of its downtoners: the intensifying downtoners are remarkably less frequent (roughly 13%, as compared to 22 % for English and 27%



Loving and hating the movies in English, German and Spanish 157

for Spanish), while the quantifying downtoners are considerably more common (about 40%, as compared to 12% for English and 16% for Spanish). In other words, at the lexicogrammatical level, German is different from English and Spanish, in word order, and frequency and type of Graduation. On the other hand, the overall organization and distribution of positive and negative evaluations showed that English is the outlier, with a recurring structure in reviews where positive aspects are presented first, before a negative evaluation is presented. We have, elsewhere, referred to this pattern as a form of vernacular argumentation (Taboada and Gómez-González, 2012). Concerning Spanish, the higher frequency of expression of Affect in positive reviews in comparison to the other two languages, together with a plausible explanation for this frequency in terms of lexicogrammar, were already mentioned above in this section. Another remarkable feature lies in the category of Focus, in which the number and percentage of sharpeners and softeners is almost the same, while English and German displayed a considerably higher number of sharpeners. Examples of common Spanish softeners that play a role in the balance are en parte (“in part”), más o menos (“more or less”) or por así decirlo (“so to speak”). The findings reported above suggest a number of possible extensions for the research presented here. For example, a contrastive analysis of English, German and Spanish movie reviews of this kind might also be applied to the category of Engagement; the analysis of Attitude could include more languages, so that a wider cross-linguistic perspective of this category in movie reviews would be obtained; and finally, the analysis could be extended to reviews of different products, which would reveal the ways in which the expression of Attitude is influenced by the kind of item reviewed. At a more general level, contrastive analysis across genres would uncover pervasive dissimilarities in the expression of Appraisal in different languages; in their turn, research on the reasons for these differences would also be worthwhile: some differences might well be due to grammar and lexis (for example, the frequency of concrete grammatical patterns, words or items from a lexical field), and others might be accounted for by social factors, when the expression of given Appraisal options is more common in one culture or another. Observations of this kind would lead to a more accurate view of Appraisal as a whole.

Acknowledgements We would like to thank Nicola Bergen and Patrick Larrivée-Woods for initial annotation of the English data. This work was funded by the Natural Sciences and Engineering Research Council of Canada (Discovery Grant 261104–2008; PI: Maite Taboada), and by the Spanish Ministry of Science and Innovation, under the project Creación y validación de descripciones contrastivas

158 Maite Taboada, Marta Carretero and Jennifer Hinnell (inglés-español) a través del análisis y la anotación de corpus: Aspectos lingüísticos, metodológicos y computacionales (Creation and validation of contrastive descriptions (English-Spanish) through the analysis and annotation of corpora: Linguistic, methodological and computational aspects), with reference number FFI2008–03384 (PI: Julia Lavid, Universidad Complutense de Madrid).

References Achugar, M. 2008. What We Remember: the Construction of Memory in Military Discourse. Amsterdam: John Benjamins. Aikhenvald, A. 2004. Evidentiality. Oxford: Oxford University Press. Bakhtin, M. 1981. “Discourse in the Novel” [Translated by C. Emerson and M. Holquist]. In The Dialogic Imagination: Four Essays by M.M. Bakhtin, M. Holquist (ed), 259–422. Austin: University of Texas Press. Banfield, A. 1982. Unspeakable Sentences: Narration and Representation in the Language of Fiction. Boston: Routledge and Kegan Paul. Batson, C., Shaw, L. and Oleson, K. 1992. “Differentiating Affect, Mood, and Emotion: Toward Functionally Based Conceptual Distinctions”. In Emotion. Review of Personality and Social Psychology, M. Clark (ed), 294–326. Newbury Park, CA: Sage. Becker, A. 2009. “Modality and Engagement in British and German Political Interviews”. Languages in Contrast 9(1):5–22.  doi: 10.1075/lic.9.1.02bec. Biber, D. and Finegan, E. 1988. “Adverbial Stance Types in English.” Discourse Processes 11(1):1– 34.  doi: 10.1080/01638538809544689. Biber, D. and Finegan, E. 1989. “Styles of Stance in English: Lexical and Grammatical Marking of Evidentiality and Affect”. Text 9(1):93–124. Bybee, J. and Fleischman, S. (eds). 1995. Modality in Grammar and Discourse. Amsterdam: John Benjamins. Carretero, M. and Taboada, M. 2014. “Appraisal in English and Spanish Consumer Reviews of Books and Movies: a Contrastive Study of the Expression of Graduation within the Scope of Attitude”. In Evaluation in Context, G. Thompson and L. Alba-Juez (eds). Amsterdam: John Benjamins. 221–239. Carretero, M. and Taboada, M. 2015. “The Annotation of Appraisal: How Attitude and Epistemic Modality Overlap in English and Spanish Consumer Reviews”. In Thinking Modally: English and Contrastive Studies on Modality, J.R. Zamorano (ed). Newcastle: Cambridge Scholars. 249–270. Carretero, M. and Zamorano, J.R. 2013. “Annotating English Adverbials for the Categories of Epistemic Modality and Evidentiality”. In English Modality: Core, Periphery and Evidentiality, J. Marín-Arrese, M. Carretero, J. Arús and J. Van der Auwera (eds), 317–356. Berlin: Mouton de Gruyter. Chafe, W. and Nichols, J. 1986. Evidentiality: the Linguistic Coding of Epistemology. Norwood, NJ: Ablex. Coffin, C. 1997. “Constructing and Giving Value to the Past: an Investigation into Secondary School History”. In Genre and Institutions: Social Processes in the Workplace and School, F. Christie and J.R. Martin (eds), 196–230. London: Cassell.

Loving and hating the movies in English, German and Spanish 159



Coffin, C. and O’Halloran, K. 2006. “The Role of Appraisal and Corpora in Detecting Covert Evaluation”. Functions of Language 13(1):77–110.  doi: 10.1075/fol.13.1.04cof. Eggins, S. and Martin, J.R. 1997. “Genres and Registers of Discourse”. In Discourse as Structure and Process. Discourse Studies: a Multidisciplinary Introduction, T.A. van Dijk (ed), 230– 256. London: Sage.  doi: 10.4135/9781446221884.n9. Eggins, S. and Slade, D. 1997. Analysing Casual Conversation. London: Cassell. Halliday, M.A.K. 1975. Learning How to Mean. London: Edward Arnold. Halliday, M.A.K. 1985. An Introduction to Functional Grammar (1st edition). London: Arnold. Halliday, M.A.K. 1989. “Context of Situation”. In Language, Context and Text, M.A.K. Halliday and R. Hasan (eds), 3–14. Oxford: Oxford University Press. Halliday, M.A.K. and Matthiessen, Ch. 2004. An Introduction to Functional Grammar (3rd edition). London: Arnold. House, J. 2006. “Communicative Styles in English and German”. European Journal of English Studies 10(3):249–267.  doi: 10.1080/13825570600967721. Hunston, S. and Thompson, G. (eds). 2000. Evaluation in Text: Authorial Distance and the Construction of Discourse. Oxford: Oxford University Press. Itakura, H. and Tsui, A.B. 2011. “Evaluation in Academic Discourse: Managing Criticism in Japanese and English Book Reviews”. Journal of Pragmatics 43(5):1366–1379.  doi: 10.1016/j.pragma.2010.10.023.

Johansson, S. 1998. “Loving and Hating in English and Norwegian: a Corpus-based Contrastive Study”. In Perspectives on Foreign and Second Language Pedagogy. Essays Presented to Kirsten Haastrup on the Occasion of her Sixtieth Birthday, D. Albrechtsen, B. Henriksen, I.M. Mees and E. Poulsen (eds), 93–103. Odense: Odense University Press. Kaplan, N. 2007. La construcción discursiva del evento conflictivo en las noticias por televisión. PhD Dissertation, Universidad Central de Venezuela, Caracas. Langacker, R. 1990. “Subjectification”. Cognitive Linguistics 1(1):5–38.  doi: 10.1515/cogl.1990.1.1.5. Liu, B. 2015. Sentiment Analysis : Mining Opinions, Sentiments, and Emotions. Cambridge: Cambridge University Press. Love, K. 2006. “APPRAISAL in Online Discussions of Literary Texts”. Text and Talk 26(2):217– 244.  doi: 10.1515/TEXT.2006.010. Macken-Horarik, M. 2003. “APPRAISAL and the Special Instructiveness of Narrative”. Text 23(2):285–312. Mann, W. and Thompson, S. 1988. “Rhetorical Structure Theory: Toward a Functional Theory of Text Organization”. Text 8(3):243–281.  doi: 10.1515/text.1.1988.8.3.243. Martin, J.R. 1996. “Evaluating Disruption: Symbolising Theme in Junior Secondary Narrative”. In Literacy in Society, R. Hasan and G. Williams(eds), 124–171. London: Longman. Martin, J.R. 2000. “Beyond Exchange: Appraisal Systems in English”. In Evaluation in Text: Authorial Distance and the Construction of Discourse, S. Hunston and G. Thompson (eds), 142–175. Oxford: Oxford University Press. Martin, J.R. 2003. “Introduction: Special Issue on Appraisal”. Text 23(2):171–181.  doi: 10.1515/text.2003.007.

Martin, J.R. and White, P. 2005. The Language of Evaluation. New York: Palgrave. Page, R. 2003. “An Analysis of APPRAISAL in Childbirth Narratives with Special Consideration of Gender and Storytelling Style”. Text 23(2):211–237.  doi: 10.1515/text.2003.009. Painter, C. 2003. “Developing Attitude: An Ontogenetic Perspective on APPRAISAL”. Text 23(2):183–209.  doi: 10.1515/text.2003.008. Palmer, F. 1986. Mood and Modality. Cambridge: Cambridge University Press.

160 Maite Taboada, Marta Carretero and Jennifer Hinnell Pang, B. and Lee, L. 2008. “Opinion Mining and Sentiment Analysis”. Foundations and Trends in Information Retrieval 2(1–2):1–135.  doi: 10.1561/1500000011. Portner, P. 2009. Modality. Oxford: Oxford University Press. Taboada, M. 2008. “SFU Review Corpus [Corpus]”. Vancouver: Simon Fraser University. Available at: www.sfu.ca/~mtaboada/research/SFU_Review_Corpus.html. Taboada, M. 2011. “Stages in an Online Review Genre”. Text and Talk 31(2):247–269.  doi: 10.1515/text.2011.011.

Taboada, M., Brooke, J., Tofiloski, M., Voll, K. and Stede, M. 2011. “Lexicon-based Methods for Sentiment Analysis”. Computational Linguistics 37(2):267–307.  doi: 10.1162/COLI_a_00049. Taboada, M. and Carretero, M. 2012. “Contrastive Analyses of Evaluation in Text: Key Issues in the Design of an Annotation System for Attitude Applicable to Consumer Reviews in English and Spanish”. Linguistics and the Human Sciences, 6(1–3):275–295.  doi: 10.1558/lhs.v6i1-3.275.

Taboada, M. and Gómez-González, M. de los Á. 2012. “Discourse Markers and Coherence Relations: Comparison across Markers, Languages and Modalities”. Linguistics and the Human Sciences 6(1–3):17–41.  doi: 10.1558/lhs.v6i1-3.17. Taboada, M. and Grieve, J. 2004. “Analyzing Appraisal Automatically”. In Proceedings of AAAI Spring Symposium on Exploring Attitude and Affect in Text (AAAI Technical Report SS-04– 07), 158–161. Stanford University, CA: AAAI Press. Traugott, E. 1995. “Subjectification in Grammaticalisation”. In Subjectivity and Subjectivisation: Linguistic Perspectives, D. Stein and S. Wright (eds), 1–54. Cambridge: Cambridge University Press.  doi: 10.1017/CBO9780511554469.003. Traugott, E. 2010. “(Inter)subjectivity and (Inter)subjectification: a Reassessment”. In Subjectification, Intersubjectification and Grammaticalization, K. Davidse, L. Vandelanotte and H. Cuyckens (eds), 29–74. Berlin: De Gruyter Mouton.  doi: 10.1515/​9783110226102.1.29. Trnavac, R. and Taboada, M. 2012. “The Contribution of Nonveridical Rhetorical Relations to Evaluation in Discourse”. Language Sciences 34(3):301–318.  doi: 10.1016/j.langsci.​2011.​10.005.

White, P. 1998. Telling Media Tales: the News Story as Rhetoric. Ph.D. Dissertation, University of Sydney. White, P. 2003a.“Beyond Modality and Hedging: a Dialogic View of the Language of Intersubjective Stance”. Text 23(2):259–284.  doi: 10.1515/text.2003.011. White, P. 2003b. “An Introductory Course in Appraisal Analysis”. Available at Appraisal website www.grammatics.com/appraisal [Retrieved 16 March 2009].

Index

A adjunct  74–75 contingency adjunct  82–83, 85 initial adjunct  74–76 manner adjunct  75 space adjunct  82, 85 time adjunct  81, 83–84 Affect  129–131, 140, 145, 152, 154–155 see also Attitude affirmative anchor, see positive anchor affirmative polarity  97, 99, 113 see also negative polarity anchor  95–98, 101–102 see also tag question construction, question tag clause  96, 101 annotation  133–135, 139 see also markable Appraisal  129, 135, 140, 146, 151 categories  128, 155 framework  128 system  128–129, 135 Appreciation  129–130, 140, 145, 150, 152, 155–156 see also Attitude span  145, 150, 156 argumentation  27, 49 see also exposition, narration vernacular argumentation  157 Attitude  129, 135, 143, 155 see also Affect, Appreciation, Judgement marker  60

B backgrounding  90 see also contrastive focus, discourse linking, thematic development bleaching  53 bottom-up statistical analysis, see statistical factor analysis

see also backgrounding, contrastive focus, thematic development downtoner  135, 143–144, 148–149, 154 see also Force, Graduation downtoning intensifier, see downtoner

C canonical tag question (CTQ)  94, 96 see also invariant tag, noncanonical tag question, variant tag cohesive device  62 contingency adjunct, see adjunct contrastive focus  85, 88, 90 see also backgrounding, discourse linking, thematic development co-occurring linguistic features, see linguistic features CroCo Corpus  44 see also fiction, instruction manuals, letters to shareholders

E editorials  63, 65–66, 68 see also Mult-Ed corpus emphasizer  135 see also Force, Graduation emphasizing intensifier, see emphasizer Engagement  128–131 Europarl corpus  63 see also parliamentary debates evaluation  127–129 see also negative evaluation, positive evaluation exposition  27, 49 see also argumentation, narration expression of Attitude, see Attitude

D dimensions  9–13, 15, 37 see also oral/literate dimensions, narrative dimensions, specialized dimensions, subdimensions of variation  10, 15, 21, 25, 31 discourse linking  90

F factor analysis, see statistical factor analysis fiction  15, 44, 76, 90 field (of discourse)  38–39, 133 see also mode (of discourse), tenor (of discourse) final tag  109–110 see also medial tag Focus  130, 143–144, 148, 150, 153, 156–157

162 Index see also Graduation, sharpener, softener Force  130, 143, 145, 148, 153, 156 see also downtoner, emphasizer, Graduation, Intensification, Quantification formulaic sequences  69–70 formulaicity  65 frequency threshold  62, 64 G goal orientation  38–40, 50 Graduation  130, 134–135, 143, 145, 147, 150, 153, 156 see also Focus, Force H Halliday  36–38, 128 Halliday and Matthiessen  50–51, 74 I indicator  37–40 negative indicator  40, 50 positive indicator  40, 50 initial adjunct, see adjunct instruction manuals  44–46 see also CroCo Corpus Intensification  130, 141, 144, 148, 153–154 see also Force, Quantification invariant tag  98–99, 101, 122 see also canonical tag question, non-canonical tag question, variant tag J Johansson  35, 42, 131 Judgement  135, 140, 145, 147, 153, 155–156 see also Attitude L letters to shareholders  44, 53 see also CroCo Corpus lexical bundle  60–63 metadiscursive bundle  62 lexical variation  47–48 linguistic dimensions, see dimensions

linguistic features  7, 8, 10–11, 13, 21, 25, 27–29, 36, 41 see also negative features, positive features M manner adjunct, see adjunct markable  133, 135, 137 see also annotation marked theme, see theme Martin and White  128–129, 130, 142 Matthiessen, see Halliday MD analysis/multidimensional analysis  9, 10–12, 15, 21, 31–32 MD studies  16, 21, 25, 27–29, 31–32 see also register variation of register variation  21, 31 medial tag  109–110, 122 see also final tag metadiscourse  60 see also lexical bundle mode (of discourse)  38–39, 133 see also field (of discourse), tenor (of discourse) mood  37, 45–47, 105–110 Mult-Ed corpus  63 see also editorials N n-grams  60–62, 64 narration  25, 27, 49–50 see also argumentation, exposition narrative dimensions  25–27, 31 see also dimension, oral/ literate dimensions, specialized dimensions, subdimensions negative anchor  98, 111 see also anchor, positive anchor negative evaluation  140–142, 147, 157 see also evaluation, positive evaluation negative features  11–13

see also linguistic features, positive features negative indicator, see indicator negative polarity  97–100, 112 see also affirmative polarity news  76–77, 90 non-canonical tag question (TQ)  96, 98 see also canonical tag question, invariant tag, variant tag O oral/literate dimensions  16, 21–22, 25, 32 see also dimensions, narrative dimensions, specialized dimensions, subdimensions oral-literate opposition, see oral/literate dimensions orientation  78 see also goal orientation P parliamentary debates  63, 65–67 see also Europarl corpus phraseology  59 polarity-biased question tag  98, 113 polarity-dependent question tag  98, 111 positive anchor  98 see also anchor, negative anchor positive evaluation  142, 148 see also evaluation, negative evaluation positive features  11–13 see also linguistic features, negative features positive indicator, see indicator Q Quantification  130, 141, 144, 148, 153–154 see also Force, Intensification question tag (QT)  94–95, 98 see also anchor, tag question construction

Index 163

R reference corpus  43, 46–47 register variation  9–10, 21 see also MD studies of register variation S sharpener  157 see also Focus, Graduation, softener situational context  36, 38, 41 softener  157 see also Focus, Graduation, sharpener space adjunct, see adjunct specialized dimensions  28–31 statistical factor analysis  10, 21 stem  60, 67–68 subdimensions  38, 40, 47, 49

systemic functional linguistics  37, 128 T tag, see question tag tag question (TQ)  94–96, 115–116 see also anchor, question tag construction  95–96 tenor (of discourse)  38–39, 133 see also field (of discourse), mode (of discourse) tertium comparationis  42, 77 thematic development  90 see also backgrounding, contrastive focus, discourse linking, theme theme  73–74 see also thematic development

marked theme  74, 77 theory-driven approach  37 see also top-down approach time adjunct, see adjunct top-down approach  37 see also theory-driven approach type/token ratio  48, 66 V variant tag  96, 98, 101, 112 see also canonical tag question, invariant tag, non-canonical tag question W White, see Martin word order  109, 138–139 WordSmith Tools  64, 66

This volume contributes to illing a gap in corpus-based research by investigating the ways in which linguistic features vary across genres/registers cross-linguistically. It brings together insightful chapters by leading scholars in the ield, fruitfully exploiting genre- or register-controlled multilingual parallel and comparable corpora to: (i) problematize cross-register variation in a multilingual perspective, (ii) address methodological and theoretical issues raised by registeroriented contrastive and translation studies, (iii) investigate the crosslinguistic and cross-genre variation of speciic linguistic features, such as lexical bundles, sentence-initial adverbials and tag questions, (iv) identify cross-cultural and cross-linguistic dissimilarities in expressing a functional category, viz. Appraisal, in the ield of opinion mining. The book ofers new cutting-edge research that should be of interest to specialists in contrastive linguistics, translation studies and cross-cultural studies. Originally published as a special issue of Languages in Contrast 14:1 (2014).

ISBN

978 90 272 4275 4

John Benjamins Publishing Company