Article body

1. Introduction

Motion and speech are basic abilities through which human beings express themselves. Among other stylistic features, written narrative texts abound in expressions indicating how characters move from one place to another (both literally and figuratively) (e.g. He scrolled across the room to the door, Slobin 1996: 211) and interact through language (e.g. “Trouble,” he chuckled, Caballero 2015: 1408). Interestingly, some languages even allow their users to integrate these two domains to produce colourful expressions (e.g. Jerry stormed through the hermitage from cell to cell […], Vergaro, Sandford, et al. 2014: 418). Language is an integral part of cognition (Lakoff 1987: 40), and thus, while the mother tongue of authors influences the way they think, construe meaning and write, that of translators has considerable impact on how they process and report information. In other words, translators, as intercultural mediators, are responsible for transferring the mental image produced by reading the source text into an appropriate mental rendering in the target language (Slobin 2000: 123). The study of the cognitive processes that underlie the translation of motion and speaking events is an interesting research issue that is informed by both Cognitive Linguistics and Descriptive Translation Studies. Cognitive Linguistics provides key concepts and tools (e.g. Talmy’s theory of lexicalization patterns 1985: 63; 2000: 117; Slobin’s Thinking-for-translating hypothesis 1996: 209; 2000: 123) that allow translation scholars to analyse the translation process (and product) in a more rigorous way (on the potentials of Cognitive Linguistics in addressing translational issues, see Rojo and Ibarretxe-Antuñano 2013; Alves and Jakobsen 2020). As Ibarretxe-Antuñano and Filipović (2013) stress, the importance of Talmy and Slobin’s theories for translation does not rest on grammaticality, but rather on rhetorical preferences: “they show us that a correct, loyal, and perfect translation is not always the best choice, but that languages, and therefore speakers, favour some constructions over others […] on the basis of their rhetorical style” (2013: 253). As far as the study of rhetorical style in intertypological translation scenarios is concerned, one of the most fruitful areas of research has proved to be manner-of-motion, since users of different language typologies devote divergent amounts of attention to manner,[1] which has consequences for the translation product of motion events (for instance, loss and gain of information have been identified in the literature). While the translation of manner-of-motion has often been examined through the lens of the “Thinking-for-Translating” hypothesis, and a considerable amount of data is available in the literature,[2] fewer studies have dealt with the translation of manner-of-speaking within this cognitive framework. My contribution thus aims to investigate manner-of-speaking in a German>Spanish translation scenario. To the best of my knowledge, no studies have so far examined this aspect of the German language, either as source language or target language (previous studies have dealt with the following translation scenarios: English>Italian/Spanish/Hungarian and Chinese/Hungarian>English – see Section 2). This paper aims to fill this gap in the literature by examining the translation of reporting verbs – and, in particular, a selection of verbs incorporating manner (e.g. brummeln [mumble], haspeln [gabble]) in a section of the Parallel Corpus German/Spanish (PaGeS), an online corpus of contemporary narrative texts compiled at the Universidade de Santiago de Compostela.[3] My findings are compared with data from previous studies dealing with the communication and motion frames. The first section of this paper begins by outlining the “Thinking-for-Translating” framework and its precedents, and then reviews previous studies dealing with the translation of manner-of-speaking. The second section details the materials and methods used, as well as the research questions pursued. The third section presents the analysis and resulting data. Finally, the fourth section offers conclusions and suggestions for further research.

2. Theoretical background issues

Native speakers of German and Spanish characteristically pay different degrees of attention to the semantic component of manner (Talmy 1985: 63; 2000: 117). This is due to the fact that these languages belong to different typologies. Generally speaking, in the context of the linguistic expression of motion, Talmy classifies languages depending on how they encode the semantic component of path. Satellite-framed languages (including Germanic languages) usually encode manner and motion in the verb, and path through a “satellite”[4]:

By contrast, verb-framed languages (including Romance languages) tend to lexicalise motion and path in the verb, and manner, if relevant, through separate elements functioning as adjuncts (gerunds, adverbial clauses, prepositional phrases):

In these two examples, while in German the verb conflates motion and manner (flog [flew]) and path is mapped onto a satellite (weg [away]), in Spanish the verb expresses information about path (salió [exited]) and manner is lexicalised through a gerund (volando [flying]).

Inspired by the work of Talmy and by other cognitive-based frameworks such as Frame Semantics (Fillmore 1982), Slobin (1996; 1997) has pioneered several studies of the translation of motion verbs. He observes that, in narrative translations into Spanish (from English), manner is only transferred on 51% of the occasions (They ran downstairs>Corrieron escaleras abajo [They ran the stairs downwards]); in the remaining cases, it is omitted (I ran out of the kitchen door>Salí por la puerta de la cocina [I exited through the door of the kitchen]) or neutralised (scramble up>trepar [climb]). Meanwhile, in the inverse typological combination (verb-framed language>satellite-framed language) manner is added on 25% of the occasions in narrative translations (lorsque le comte de Buondelmonte entra dans sa chambre [when the Count of Buondelmonte entered his room]>when the Count of Buondelmonte stepped into his room).

In an attempt to make Talmy’s theory of lexicalisation patterns more flexible, so that it captures different degrees of expression of manner in languages belonging to the same typology, Slobin (2004: 251) suggests a manner-salience cline, according to which speakers of high-manner-salient languages (like German) regularly give information about manner, whereas low-manner-salient languages (like Spanish) provide information about manner only when it is relevant. Slobin (2006: 64-68) justifies this manner-salience cline mainly in terms of lexical availability, diversity of the manner-verb lexicon, and cognitive effort. Regarding lexical availability, Slobin argues that the more accessible and codifiable a semantic component of manner in a language is, the more frequently it will tend to express information on manner. He also observes that manner is more codifiable in a language when it is expressed mainly through a personal verbal form (as opposed to an impersonal verbal form), a unique morpheme (as opposed to a clause) or a frequently used lexical item (as opposed to an uncommon one). Slobin (1997: 459) proposes two tiers of lexicons with respect to the encoding of manner: a) verbs that include basic motion verbs (for instance: walk); and b) verbs that denote expressive information (for instance: dash) (cf. Vergaro, Sandford, et al. 2014: 411, where reference is made to two tiers of verbal lexicons in the case of manner-of-speaking). Slobin argues that satellite-framed languages include a greater diversity of second-tier verbs, and thus a larger diversity of manner-verbs, than do verb-framed languages. He also concludes that users of verb-framed languages have to make a greater effort when expressing manner, since in these languages the verb usually expresses path, while manner has to be encoded through other mechanisms (such as gerunds, adverbial clauses, prepositional phrases). By contrast, in satellite-framed languages, path is encoded through a satellite, and manner is simply lexicalised in the verb.

Drawing on the results of his studies, and with the aim of explaining the discursive consequences of typological differences in the process of translation, Slobin (1996: 209; 2000: 123) proposes the “Thinking-for-Translating” hypothesis, according to which translators codify different pieces of information depending on the mechanisms and limitations of the target language (usually their mother tongue). For the last decade, numerous research studies have confirmed this hypothesis in the context of the expression of motion in narrative translations, and have identified techniques and strategies[5] adapted to specific translation problems both in inter- and intratypological scenarios (e.g. Filipović 1999: 20; 2008: 32; Ibarretxe-Antuñano 2003: 154; Cifuentes-Férez 2006: 88; 2013: 10; Lewandowski and Mateu 2016: 194; Alonso Alonso 2018: 365). One of the translation problems most commonly analysed in the literature has been manner-of-motion, in which the great majority of studies have included the English language as source or target language. The data so far indicate that, in translation from a satellite-framed language into a verb-framed language, omission of manner-of-motion is usually found (e.g. in Molés-Cases 2018, analysing a German>Spanish corpus of children’s and young adult literature, this loss takes place in 26.27% of instances). Meanwhile, in the inverse typological combination, there is a tendency for this component to be added (e.g. in Molés-Cases 2019 study, this is found to occur in 23.39% of instances in the Spanish>German subcomponent of the PaGeS corpus). Intratypological variation as to manner has also been identified in satellite-framed languages (cf. Slobin 2004: 225).

As far as the methodologies applied in previous research on thinking-for-translating motion in narrative texts are concerned, two broad types of procedure should be highlighted. The first is represented by research studies that have made use of random manual searches for motion events in source and target texts, as pioneered by Slobin (1996).

My procedure was to open a book at random and read until finding a motion event […] As a first impressionistic observation, I found that I often had to open a Spanish book several times to find a page with a motion event, whereas this was hardly ever the case for the English novels.

Slobin 1996: 207

The contributions of Oh (2009) and Vergaro (2011; 2012) are examples of studies applying this method. Within this group, two prominent trends can be observed: studies examining Chapter 6 of The Hobbit (Tolkien 1937) (e.g. Ibarretxe-Antuñano 2003; Slobin 2005; Sugiyama 2005; Lewandowski and Mateu 2016) and works focusing on the Harry Potter book series (Rowling 1997-2007) (e.g. Cifuentes-Férez 2006; 2013; Garczarczyk 2012). The number of motion events examined using this method ranges from 9 to around 400 (including both source and target events).

The second group is characterised by research studies diverging from those in the first group in terms of corpus size, corpus analysis tools and unit of analysis. Firstly, this includes works based on large electronic corpora (and hence including texts written by a greater diversity of authors and translators). Corpus-based tools, or corpora including such tools, are thus frequently used in this second methodological approach. In addition, the starting points of the analyses are generally lists of verbs classified in previous reference studies. Examples of this group include Iacobini and Vergaro (2012) and Cappelle (2012). Iacobini and Vergaro (2012) present a quantitative and qualitative examination of an English>Italian corpus composed of 5 novels and one collection of short stories,[6] as well as part of the Open Parallel Corpus (OPUS),[7] limited to the first 50 occurrences. Searches for 40 motion verbs in this study (drawing on Slobin 2008) yielded an empirical basis of around 300 motion events. The contribution of Cappelle (2012), which delves into translation universals and has a quantitative scope based on 273 motion verbs (following Levin 1993), makes use of a French>English and a German>English parallel corpus (several subsections of the Translational English Corpus, including between 400,000 and 550,000 tokens)[8] and a comparable English corpus (a subsection of nearly 19,500,000 tokens from the British National Corpus).[9]

The phenomenon of the translation of manner has also been approached in the literature through the communicative frame, and more specifically through the study of manner-of-speaking verbs (Zwicky 1971), with a special focus on those introducing direct speech. Manner-of-speaking refers to the different ways humans interact through language. For the purposes of this paper, it will be defined here in Talmy’s (1991: 484) terms, as an external factor of the speaking event which “fill[s] in, elaborate[s], add[s] to or motivate[s]” this framing event.

Contreras (1988) conducted the first exploratory study of the translation of this phenomenon in a corpus of English>Spanish narrative texts. Contreras presents a bilingual lexicon of verba dicendi. The resulting data allow him to conclude that dialogues translated into Spanish are far more varied and richer in this type of verb than their English counterparts (here 635 verbs were observed in Spanish, compared to only 215 verbs in English).

Inspired by the cognitive frameworks mentioned above, and by Slobin’s observation of satellite-framed languages having a larger and more diverse lexicon of manner-verbs than verb-framed languages, Rojo and Valenzuela (2001) analyse the translation of verbs of saying in a corpus composed of four narrative works in English and their corresponding translations into Spanish.[10] The authors examine a total of 400 English verbs (100 verbs randomly selected from each novel) and their corresponding translations into Spanish. The results are classified into two broad groups: a) general verbs of saying (GEN) – that is, verbs including the two main frame elements in a communicative event, the Speaker and the Addressee (for instance: say, speak); and b) specific verbs of saying (SPEC) – that is, verbs including reference to any other element in the communicative event (for instance: manner, intention – murmur, insist). In their analysis, Rojo and Valenzuela identify four translation strategies:

  • GEN → GEN (55%): a general verb from the ST (source text) is translated by a general verb in the TT (target text) (e.g. say>decir [say]);

  • GEN → SPEC (20.75%): a general verb from the ST is translated by a specific verb in the TT (e.g. talk>explicar [explain]);

  • SPEC → SPEC (22.25%): a specific verb from the ST is translated by a specific verb in the TT (e.g. confess>confesar [confess]);

  • SPEC → GEN (2%): a specific verb from the ST is translated by a general verb in the TT (e.g. discuss>hablar [speak]).

Contrary to what was expected according to Slobin’s results regarding the translation of motion in the typological combination satellite-framed language>verb-framed language, in the case of specific verbs of saying (Slobin 1996; 1997), the most common strategy observed in this study is SPEC → SPEC (i.e. transference), while SPEC → GEN (i.e. generalisation) is quite uncommon. Moreover, the GEN → SPEC strategy (i.e. specification) also appears quite frequently. As we have seen, in the translation of motion in the typological combination satellite-framed language>verb-framed language, Slobin (1996: 212; 1997: 459) identified a high degree of omission of manner – what, in Rojo and Valenzuela’s (2001: 470) terms, would be comparable to SPEC → GEN. Rojo and Valenzuela explain that the data in their study indicate a different behaviour with regard to the translation of verbs of saying than to the translation of motion verbs.[11] According to the authors, translators probably consider that describing a way of saying is more relevant than describing a way of moving. In addition to considering the elements of Speaker and Addressee, the translation also includes other information, such as Manner and Time.

[I]t could be that translators feel that the “way” of saying in a dialogue is somehow more relevant than the “way” of motion in the description of a scene of movement. If you lose manner in a verb of movement, you just lose detail in the description; however, losing information in a specific verb of saying could hinder the correct interpretation of the utterance within a given context.

Rojo and Valenzuela 2001: 473

Rojo and Valenzuela also identify a larger diversity of specific verbs of saying in the Spanish translations than in the original English texts (56 Spanish verbs versus 46 English verbs). Thus, these data do not confirm the greater diversity of the manner-verb lexicon of satellite-framed languages in the case of verbs of saying. Indeed, as the authors explain, while in dialogs the tendency in English seems to be to resort to the verb say, in Spanish it is more common to use other forms to avoid repetition, and additional information is frequently specified for stylistic reasons.

In a study dealing with the translation of reporting verbs in literary texts translated from English into Hungarian and vice versa, Klaudy and Károly (2005: 14) propose the Asymmetry Hypothesis. According to this hypothesis, explicitations in the L1→L2 direction are not always counterbalanced by implicitations in the L2→L1 direction, which they hold to be due to the fact that translators, when possible, prefer to resort to explicitation strategies. The corpus examined by Klaudy and Károly is composed of the English novel 1984 (Orwell 1948) and its translation into Hungarian, and the Hungarian novels Édes Anna (Kosztolányi 1926) and Szent Péter esernyője (Mikszáth 1895) and their translations into English (Anna Édes, St. Peter’s Umbrella). The authors randomly analyse 100 reporting verbs in each of the original and translated versions, making a total of 600 reporting verbs. The resulting data allow the authors to confirm the Asymmetry Hypothesis, since explicitation in the English>Hungarian translation scenario is not counterbalanced by implicitation in the inverse linguistic combination. In other words, the translations into English (from Hungarian) do not tend towards simplification, but rather towards preserving the large variety of reporting verbs in the Hungarian originals. Although this is not mentioned by the authors, variation as to manner was not observed in this intratypological study, since both English and Hungarian belong to the satellite-framed group.

Another piece of work stemming from that of Rojo and Valenzuela (2001) is Shi (2008), which investigates the Chinese>English translation of verbs of saying that introduce direct speech. Shi analyses a corpus composed of Chapters 29 and 77 of the Chinese classic work Hong Lou Meng and its translation into English by Yang Hsien-yi and Gladys Yang (A Dream of Red Mansions). This study identifies a total of 211 fragments including verbs of saying. Analysis of the translated fragments in comparison with the originals indicates that, while type of communication is omitted in 46 cases, in 149 fragments general communication expressions are present, and 16 fragments include specific communication expressions. Overall, the data suggest that a larger diversity of general manner verbs of saying is present in the Chinese original, while in the English translation there is a larger variety of specific verbs of saying (here four elements are identified: Manner, Content, Answer and Aspect). Specifically regarding manner-of-speaking, the translation includes 78 fragments incorporating this information and 32 that omit it. The importance of this study is that it confirms that, although both Chinese and English belong to the satellite-framed group, these languages present divergences in relation to how they encode communication expressions, both general and specific, and that this has consequences for translation (cf. Slobin 2004: 251, on the manner salience cline). Shi points out the need to make use of studies encompassing a variety of authors, translators and periods of time, with the aim of further examining the degree to which a writer’s individual style might be conditioned by aspects of the languages she is working in (Shi 2008: 205).

Also following Rojo and Valenzuela’s (2001) research study, Mastrofini (2015) investigates the domain of manner-of-speaking in an English>Italian corpus composed of four contemporary novels and their translations.[12] Her study initially searched a list of 176 verbs (based on Vergaro, Sandford, et al. 2013) – from which 83 verbs were finally encountered in the corpus – and analysed 776 occurrences. The study’s conclusions can be summarised in two main insights: more manner-of-speaking verbs were identified for Italian than for English (148 verbs and multi-word expressions in Italian versus 83 verbs in English), and the two languages present similar ways of conflating manner-of-speaking (i.e. through the verb root). In other words, the translation of manner-of-motion differs from that of manner-of-speaking in intratypological translation scenarios (in this case, English>Italian), since loss of manner-of-speaking in Italian was seldom observed, contrary to the tendency in the case of manner-of-motion (cf. Iacobini and Vergaro 2014, which gives a frequency of almost 40%).

Another contribution focusing on translations of narrative texts from English into Spanish is the exhaustive study presented in Caballero (2015). Caballero examines verbs introducing direct speech in a corpus composed of twelve fiction narratives originally written in English, and their corresponding translations into Spanish (with 1,585,120 and 1,713,986 tokens, respectively). The aim of the investigation is twofold: to compare the use of verbs of communication with non-speech patterns in these languages and to explore whether the typological differences between English and Spanish described in the previous literature on motion are validated in the case of speech events. Caballero mainly observed the following types of verb in both English and Spanish texts: illocutionary verbs (acknowledge/reconocer), manner verbs (stammer/tartamudear), turn-taking verbs (conclude/concluir), verbs related to gestures (frown/fruncir el ceño) and verbs related to physical actions and states (clear throat/carraspear). She concludes that the differences between English and Spanish regarding the way direct speech is introduced are fundamentally discursive (i.e. while English authors prefer to resort to manner verbs, the preference of Spanish translators seems to be for illocutionary and turn-taking verbs). She illustrates this metaphorically in the following terms: “The differences between the two languages could, then, be summarized as showing (English) versus telling (Spanish), that is, while Spanish provides readers with a script, English offers the film” (2015: 1427). The results of Caballero’s study also indicate that the typological divergences between these languages observed in the case of motion expressions are not sustained in the case of speaking events (i.e. Spanish is not lexically poorer in verba dicendi than English).

A similar study dealing with literary texts is Ruano San Segundo (2017), which analyses the translation of reporting verbs into four Spanish translations of Charles Dickens’s 1854 novel Hard Times (by J. Ribera 1972; Lázaro Ros 1992; Víctor Pozanco 1995; Ángel Melendo 2005). The starting point of the study is that the translation of reporting verbs is a phenomenon that contributes significantly to the characterisation of Dickens’s characters, as explained by the author. Specifically, Ruano San Segundo analyses reporting verbs (with the exception of say, for its neutrality). While in the original, 42 reporting verbs are found, the numbers of such verbs in the four translations are as follows: 25 (Ribera), 34 (Melendo), 39 (Ros), and 48 (Pozanco). It is also important to note that the version translated by Pozanco is the translation in which the general verb decir [say] is found most frequently (50 occurrences), usually translated from a specific verb in the original. According to Ruano San Segundo, the data show that none of the translations preserves in its entirety the value of the original with regard to the specific verbs of reporting, and that this fact has a stylistic effect on translation, modifying the impressions that Dickens’s readers tend to form about characters. At first sight, these results diverge from the findings of previous studies on manner-of-speaking (probably because more diverse corpora with a wider range of authors and translators are needed, cf. Shi 2008), but they are very precise regarding Dickens’s style and any loss or gain of information in his translations into Spanish. Ruano San Segundo’s study stands out for its corpus-based methodology, which makes use of concordance software such as WordSmith Tools.

As we have seen, it is not only manner-of-motion that has been explored in the literature; manner-of-speaking has been also examined – albeit to lesser extent, and only in very specific linguistic combinations that mostly include English as source language. This paper aims to shed some light on the study of manner-of-speaking within the “Thinking-for-Translating” framework, through a new linguistic combination (German>Spanish), making use of a large parallel corpus (PaGeS) while producing both qualitative and quantitative analyses. It is intended to enrich the empirical data in this line of research, confirming the (dis)similarities between the communicative and motion frames in translated narratives.

3. Materials and methods

3.1. Unit of analysis: reporting verbs

The study makes use of Rojo and Valenzuela’s (2001) method. The verbs examined are divided into two categories: the German general verb sagen [say],[13] and a series of manner-of-speaking verbs. German verbs have received much attention in the literature, and numerous semantic classifications have been proposed in recent decades (e.g. Helbig and Schenkel 1975; Snell-Hornby 1983; Schröder 1993; Baker, Fillmore, et al. 2003; Harras, Proost, et al. 2007). Regarding the manner-of-speaking verbs examined here, this paper draws on Snell-Hornby’s (1983: 170-173) proposal, and more specifically on her category “talking and uttering” (Table 1) – with the exception of the subcategory “animal and bird sounds,” which is not relevant for the study of reporting verbs.

Table 1

Verbs of talking and uttering (Snell-Hornby 1983)

Verbs of talking and uttering (Snell-Hornby 1983)

-> See the list of tables

In total, since the verbs murmeln [mutter] and plappern [chatter] are included in two subcategories, the unit of analysis of this paper is composed of 33 reporting verbs in German (32 manner-of-speaking verbs and one general verb of saying) (cf. Iacobini and Vergaro 2012).

3.2. Corpus and method

This study makes use of the online corpus PaGeS, a bilingual narrative corpus compiled at the Universidade de Santiago de Compostela (Spain) by the research group SpatiAlEs.[14] PaGeS, which at the time of this investigation amounts to around 25,000,000 tokens, includes not only original and translated texts in German and Spanish, but also a small percentage of translations into these languages from a third language.[15] The corpus is integrated using an advanced search tool that enables filtering according to the following categories: language (German, Spanish), text (original, translation, translation from third language), part-of-speech, subgenre (fiction, non-fiction), year, author and ID-work. In this study, the German originals (5,081,806 tokens) and their corresponding translations into Spanish (5,311,191 tokens) constituted my empirical raw material.

The procedure used was first to search the 33 reporting verbs by language (German), part-of-speech (finite verb) and text (original texts). Two problems were identified at the initial stage of the research study: there was a significant disparity in the results between the different manner-of-speaking verbs (for instance, while 1,385 cases of flüstern [whisper] and 1,663 of schreien [yell] were observed, only six cases of krakeelen [shout] were identified, and one occurrence of haspeln [gabble]); at the same time, the PaGeS corpus does not enable thinning of query results. For these reasons, in an attempt to clean the data and maintain proportionality, I considered a random 10% of the total results for each manner-of-speaking verb in my analysis (cf. Slobin 1996: 207; Iacobini and Vergaro 2012: 78). Regarding the analysis of the general verb sagen, since PaGeS includes 30,452 occurrences of this verb and 10% of these was still too much for a manual qualitative analysis, I made a random selection of 300 occurrences of this verb as well (cf. Slobin 1996: 207). I established that each case included in the final sample introduced direct speech. It is important to note that, despite the methodological restrictions outlined above, the quantity of data analysed was significant (919 original motion events in German and 919 motion events translated into Spanish – see Table 2) (cf. Slobin 1996; Cifuentes-Férez 2006; 2013; Oh 2009; Iacobini and Vergaro 2012; Mastrofini 2015), and certainly sufficient for the purposes of the study, since it deploys not only a quantitative but also a qualitative approach, which allowed me to explore the phenomenon of the translation of manner-of-speaking in considerable depth (i.e. How often is manner-of-speaking missing, transferred or added in translations into Spanish? How is it specifically rendered? What factors underlie the translation process?) (see Sandelowsky 2003; Dörnyei 2007; Meister 2018, on the benefits of mixed-methods research). The specific source and target texts from which the examined pairs of fragments (ST-TT) derive are included in the Appendix (here both literature for a general audience and children’s and young adult literature can be found: 38 and 15 pairs of original and translated novels respectively).

3.3. Data analysed

As Table 2 indicates, I examined a total of 919 pairs of ST-TT fragments. In particular, my analysis was based on 619 pairs of fragments originally including a manner-of-speaking verb and 300 pairs of fragments originally including the general verb sagen [say].

Table 2

ST-TT Pairs of fragments analysed

ST-TT Pairs of fragments analysed

Table 2 (continuation)

ST-TT Pairs of fragments analysed

-> See the list of tables

As Table 2 shows, for five manner-of-speaking verbs (johlen [yell], klönen [natter], quasseln [babble], ratschen [chat], schwafeln [twaddle]), no cases of direct speech were encountered in the corpus and these were ultimately disregarded. Thus, the final total of reporting verbs considered in the study was 28 (including the general verb sagen [say]).

3.4. Research questions

As noted above, in this paper I examine a series of German reporting verbs in a subsection of the online narrative corpus PaGeS, composed of original texts in German and their corresponding translations into Spanish. Following works based on the “Thinking-for-Translating” hypothesis, mindful of the consequences that the typological divergences have for the translation of motion events (i.e. mostly loss or gain of manner), and by employing a linguistic combination for now unexplored in the literature, this paper sets out to investigate further the possible differences arising in the ways translators deal with the motion and communication frames in narrative texts. The research questions this paper aims to answer are the following:

Research Question 1: Is manner-of-speaking lost in the translation into Spanish (from German) of specific reporting verbs in narrative texts? This question is motivated by the results of comparable studies on manner-of-motion in the typological combination satellite-framed>verb-framed language (for instance: Ibarretxe-Antuñano 2003: 154, for English>Basque/Spanish; Cifuentes-Férez 2013: 10, for English>Spanish; Alonso Alonso 2018: 365, for English>Galician; Molés-Cases 2018: 172, for German>Spanish), which have confirmed that a frequent translation technique is loss of manner. Given that German native speakers tend to focus their attention on manner-of-motion, while Spanish native speakers usually devote more attention to other information (such as path), sometimes at the expense of manner, my interest is to examine whether loss of manner-of-speaking is also often encountered when comparing novels translated into Spanish with their corresponding German originals.

Research Question 2: Is manner-of-speaking (or any other information) gained in the translation into Spanish of the German general reporting verb sagen [say] in narrative texts? This question arises out of the stylistic tendency observed by Rojo and Valenzuela (2001: 473) and Caballero (2015: 1423), in translations from English into Spanish, to avoid the repetition of the general verb decir [say]. Considering that, for stylistic reasons, the use of synonyms is in general much more common in Spanish than in German (cf. Nord 2018: 71), here I examine whether the general verb sagen [say] presents a wide range of solutions in novels translated from German into Spanish – and if so, whether those solutions enrich the target texts with additional information (e.g. manner, intention).

Research Question 3: In a satellite-framed>verb-framed language translation scenario, do translators show the same behaviour when dealing with motion and communication frames in narrative texts? This question emerges naturally from the first two questions. According to Rojo and Valenzuela’s (2001: 475-476) English>Spanish translation study, Spanish translators seem to devote more attention to the way of saying than to the way of moving. Furthermore, as noted above, these authors observe that manner-of-speaking is not lost in novels translated from English into Spanish, while manner-of-speaking (and other information, including time, intention and topic length) is gained when translating the general verb say into Spanish. While these phenomena have been empirically confirmed for the English>Spanish combination (see also Caballero 2015: 1423), this paper attempts to explore them in a German>Spanish translation scenario. The comparison of the data of this study with results of previous studies drawing on the motion and communication frames will contribute to an in-depth exploration of factors underlying the satellite-framed>verb-framed language translation process, and to a better understanding of the behaviour of translators dealing with these two frames.

4. Analysis and results

This section details and discusses the results of the study in both qualitative and quantitative terms, particularly in relation to translation techniques, translation solutions and verbal lexicon. Section 4.1. presents the results regarding the translation of manner-of-speaking verbs, while Section 4.2. details the data on the translation of the general verb sagen [say]. Section 4.3. offers a discussion of the results.

4.1. Analysis of manner-of-speaking verbs

This section presents the data resulting from my analysis of the translation of manner-of-speaking verbs, which is related to the first research question. In what follows, I will explore whether manner-of-speaking is lost in the translation from German into Spanish of a series of manner-of-speaking verbs that introduce direct speech in narrative texts (see Section 3.1). Firstly, I will introduce the translation techniques observed in my analysis, providing examples and some quantification. Secondly, I will itemise the translation solutions in Spanish for the manner-of-speaking verbs searched for in German.

I identified three techniques in this study: transference, omission of manner and omission of speaking event. The first two techniques are comparable with Rojo and Valenzuela’s (2001: 472) SPEC→SPEC and SPEC→GEN strategies. The third is an adaptation of the original translation strategy of omission of motion event proposed by Cifuentes-Férez (2006: 88; 2013: 10). I then provide several examples.

In Examples 3 and 4 (transference), the original information regarding heulen [howl] and haspeln [gabble] is present in the translation. Thus, both the original and the translated versions include specific contexts and the same information is provided to readers of both the source and target texts.

In Examples 5 and 6 (omission of manner), the specific information contained in schreien [shout] and wispern [whisper] is generalised in the translation by using the verb decir [say]. These nuances of schreien and wispern are not expressed either in the context. In other words, here manner-of-speaking is missing in the translation and must be inferred by the reader. In addition, here the translation does not contribute to characterisation (cf. Ruano San Segundo 2017) or to scene-setting.

In Examples 7 and 8 (omission of speaking event), any lexical reference regarding reporting is not incorporated in the translation. More specifically, the information introduced by murmeln [mutter] and stocken [falter] is not included in the target text and again must be inferred by the reader.

Table 3 records the frequency of each of the translation techniques encountered.

Table 3

Frequency of the translation techniques encountered for manner-of-speaking verbs

Frequency of the translation techniques encountered for manner-of-speaking verbs

-> See the list of tables

As Table 3 shows, the most predominant technique is transference (97.90%), followed by omission of speaking event (1.13%) and then omission of manner (0.97%). Since the techniques of omission of manner and omission of speaking event are so uncommon, no significant correlations can be inferred between the various subcategories defined by Snell-Hornby (1983) and translation techniques. In general terms, manner-of-speaking is almost always maintained in translation (the cases of omission of manner and omission of speaking event are considered here isolated cases). This provides an answer to my first research question, which was whether manner-of-speaking would be lost in the translation from German into Spanish of specific reporting verbs in narrative texts. In terms of the possible impact of the nature of the target audience of the works examined, no differences were observed between children’s and young adult literature, on one hand, and literature for a general audience, on the other. This seems unremarkable, given the low percentages of the techniques of omission of manner and of speaking event.

Table 4 provides the translation solutions encountered for each German manner-of-speaking verb.

Table 4

Translations into Spanish of the German manner-of-speaking verbs

Translations into Spanish of the German manner-of-speaking verbs

Table 4 (continuation)

Translations into Spanish of the German manner-of-speaking verbs

-> See the list of tables

As Table 4 shows, the manner-of-speaking verbs with more diverse translation solutions in Spanish are flüstern [whisper], schreien [yell], stocken [falter] and zischen [hiss]. While the verbs flüstern and schreien were among the most predominant verbs in the original corpus (see Table 2), and represent prototypical verbs in their respective subcategories, it is not surprising that the verbs stocken [falter] and zischen [hiss] produce a wide array of translation solutions (in some cases including not just equivalent verbs, but also multi-word expressions – cf. Mastrofini 2015: 88), given the highly specific nuances they convey. For instance, besides denoting a sibilant sound, the verb zischen [hiss] also indicates a kind of annoyance or irritation, which can be denoted by expressions such as echando chispas, enfadado, and furioso, included in Table 4. All in all, the 27 manner-of-speaking verbs in the German originals become 84 speaking verbs or expressions in the various Spanish translations (including very few instances of general verbs, e.g. decir [say], hablar [talk]).

4.2. Analysis of the general verb sagen

Next, I will detail the results from the translation analysis of the general verb sagen [say]. This is related to the second research question: Is manner-of-speaking (or any other information) gained in the translation into Spanish of the German general reporting verb sagen [say] in narrative texts? Here I will firstly present the translation techniques identified in the analysis, with examples and quantification. Then the focus will be transferred to the verbal lexicon of Spanish target texts.

The translation techniques identified in this case are transference, specification and omission of speaking event. The first two are comparable with Rojo and Valenzuela’s (2001: 472) GEN→GEN and GEN→SPEC strategies. Again, the technique of omission of speaking event is based on previous literature on motion events (cf. Cifuentes-Férez 2006: 88; 2013:10). Examples of each of these techniques will be given below.

In Examples 9 and 10 (transference), both the original and the translation convey general information on communication through the two general verbs of saying in each language: sagen [say] and decir [say].

In Examples 11 and 12 (specification), the communication event is specified in the translation (i.e. manner), since the original general verb sagen is translated as the specific verbs murmurar [murmur] and exclamar [exclaim]. Given that the nuances gained are not present in the original context in any way, the target versions are thus here richer in rhetorical resources, as seems conventional in the Spanish language in relation to the reporting of speech (cf. Rojo and Valenzuela 2001; Caballero 2015).

In Examples 13 and 14 (omission of speaking event), the general verb sagen has been substituted in both translations by a hyphen indicating a dialog. This could be related to the typographical differences between German and Spanish for introducing speech scenes (cf. Caballero 2015: 1397, on the impact of typography in such cases).

In relation to frequency, as Table 5 shows, the results are more divergent.

Table 5

Frequency of translation techniques encountered for the general verb sagen [say]

Frequency of translation techniques encountered for the general verb sagen [say]

-> See the list of tables

The general information regarding sagen is retained in 68% of cases, transformed through specific information in 29.67% of cases and omitted in very few (2.33%). These data provide an answer to my second research question, confirming that manner-of-speaking (and other information) is gained in the translation into Spanish of the German general reporting verb sagen in narrative texts. As for the possible influence on the results of the study of the nature of the target audience, no important differences were observed between the two subgenres examined here (30.33% of the data for specification, and 28.58% of the data for omission of speaking event, derives from translation of children’s and young adult literature). These data are not surprising, since, in the corpus examined, novels aimed at a general audience significantly outnumber novels aimed at a younger audience.

In order to explore what other information is gained in the translation, I will focus in what follows on the technique of specification, which captures nearly 30% of those cases where the original included the general verb sagen [say] (see Table 5). As Table 6 illustrates, a total of 39 specific reporting verbs in Spanish were encountered in the analysis of this technique. These are classified in terms of the following frame elements defined by Rojo and Valenzuela (2001: 474), who draw on the Speech-Act framework used by Fillmore in FrameNet: Intention, Manner, Time and Topic length (I have also followed the observations in Caballero 2015: 1399, regarding verb categorisation). It should be noted that the classification of some verbs within specific frame elements is a complex matter, since some may be placed in more than one category simultaneously. For this reason, some verbs have been included under more than one frame element (cf. Rojo and Valenzuela 2001: 474).

Table 6

The verbal lexicon of the Spanish translations for the cases of specification

The verbal lexicon of the Spanish translations for the cases of specification

-> See the list of tables

Some illustrative examples for each frame element include the following. In Example 15, manner is incorporated into the translation: the original sagen becomes balbucear [babble] in the target text. The inclusion of this verb in the target text clearly contributes to the creation of a specific personality for the character Boston (e.g. a shy, insecure or scared person).

In Example 16, it can be seen that the translator chooses the verb pedir [request], which incorporates Intention (pragmatic force).

In Example 17, the translator uses the verb añadir [add], which adds further explanation of a point referred to previously in the conversation (frame element: Time).

Finally, in Example 18, information about the length of the message is incorporated by the verb comentar [comment], which indicates the introduction of new information to the conversation[16].

As Table 6 makes clear, five translation solutions refer to manner-of-speaking: balbucear, exclamar, murmurar, soltar and suspirar. These are present in a total of ten translated fragments. In other words, on the one hand, the original contexts including the general verb sagen are specified in 29.67% of the Spanish translations (see Table 5); on the other hand, 10.41% of those specifications refer to manner-of-speaking. The most frequently added frame element is clearly Intention (60.41%), followed by Time (25%) (see Table 6). These results are consistent with Caballero’s 2015 findings regarding the Spanish preference for illocutionary and turn-taking verbs.

4.3. Discussion

As we have seen, the first research question in this study was whether manner-of-speaking is lost in the translation from German into Spanish of specific reporting verbs in narrative texts. As has been confirmed, loss of manner is rarely observed, being represented by only 0.97% of cases (echoing the results of comparable previous studies, such as Mastrofini 2015, for an English>Italian translation scenario). Transference is thus the most predominant technique observed in my analysis of manner-of-speaking verbs. These data indicate that the main results of the “Thinking-for-Translating” hypothesis in the specific combination satellite-framed>verb-framed language confirmed for the motion frame – i.e. loss or omission of manner (e.g. around 30% specifically for German>Spanish, Molés-Cases 2018: 174; around 50% for English>Spanish, Slobin 1996: 212; 1997: 459) – do not apply in the case of the communication frame.

The second research question was whether there was any gain of manner-of-speaking (or any other information) in the translation from German into Spanish of the general reporting verb sagen in narrative texts. Replicating the results obtained by previous studies focusing on the communication frame (Rojo and Valenzuela 2001; Caballero 2015), the translators into Spanish tended to avoid the repetition of the general verb decir [say] here too, probably for stylistic reasons (i.e. to comply with the rhetorical conventions of the Spanish language). In consequence, additional frame elements were included in the translation in 29.67% of the cases: Intention, Manner, Time and Topic length. Among these, Intention and Time were clearly the frame elements most frequently added to the translation (in 60.41% and 25% of cases, respectively) in the corpus studied (cf. Caballero 2015). Meanwhile, manner was specified in 10.41% of cases. These results are in fact more in line with the usual translation tendencies of motion events in the inverse translation combination (verb-framed language>satellite-framed language) (see Molés-Cases 2019: 162, which confirms that manner-of-motion is added in around 23% of cases in a Spanish>German corpus; and Slobin 1996: 212, 1997: 459, where a rate of around 25% is established for the addition of manner in a Spanish>English corpus).

The third and final research question asked whether, in translating from a satellite-framed to a verb-framed language, translators show the same behaviour in dealing with the communication and motion frames in narrative texts. Drawing on the results of previous studies focusing on the communication frame (Rojo and Valenzuela 2001: 474; Caballero 2015: 1423) and the motion frame (e.g. Slobin 1997: 209; Cifuentes-Férez 2013: 10; Alonso Alonso 2018: 365), and on the answers to the two previous research questions, this study confirms that translators act differently when translating within these two frames, since manner-of-motion is often omitted in translations from German into Spanish (see Molés-Cases 2018: 174), while manner-of-speaking and other information are usually transferred, or even added, to the Spanish translation. This fact clearly indicates that the two frames are not considered equally important by translators, who tend to devote more attention to the communication frame than to the motion frame, probably because manner-of-speaking affects key factors such as authorial intention, characterisation (cf. Ruano San Segundo 2017) and scene-setting.

Although the corpus examined includes texts aimed at a broad, diverse audience (both a general audience and children and young adults), this does not seem to have had an overall impact on the results of the analysis. As I have noted, however, it should be recalled that the empirical basis examined here consists mostly of novels aimed at a general audience (38 novels for a general audience versus 15 novels for children and young adults) and that this kind of comparison would be better addressed through the examination of balanced and comparable corpora in future research. It is also important to note here that, while Rojo and Valenzuela (2001) consider their four translation strategies in a single study (55% GEN → GEN, 20.75% GEN → SPEC, 22.25% SPEC → SPEC, 2% SPEC → GEN), in this paper two different analyses are proposed, the first focused on manner-of-speaking verbs and the second on the general verb sagen. Moreover, while the relevant percentages are not themselves comparable, overall tendencies (maintenance of manner in the analysis of manner-of-speaking verbs – Research Question 1; specification of the communication event in the analysis of the general verb sagen – Research Question 2) are quite similar. Finally, in contrast with the motion frame, in which verbs of satellite-framed languages usually outnumber the verbs of verb-framed languages, in the Spanish translations studied here a high number of manner-of-speaking verbs were encountered. In my examination of translators’ treatment of 27 German manner-of-speaking verbs, the Spanish translations included a total of 84 verbs and multi-word expressions. Similarly, in my exploration of translators’ treatment of the general verb sagen, Spanish translators used 39 speaking verbs. These data suggest that the Spanish verb lexicon has greater granularity than the German verb lexicon specifically in the communication domain. In other words, the Spanish language seems to have a richer range of resources for the description of speaking events than the German language.

5. Concluding remarks

This study has examined the translation of the phenomenon of manner-of-speaking in a narrative corpus containing German originals and translations into Spanish, a satellite-framed>verb-framed combination that has not previously been explored through the lens of the “Thinking-for-Translating” hypothesis as it relates to speaking events. The main purpose of the investigation was to explore whether the consequences for translation encountered in the line of research exploring motion events from the perspective of the “Thinking-for-Translating” hypothesis (i.e. loss and gain of information) were sustained in the case of speaking events. Several conclusions can be drawn from my analysis.

First, manner-of-speaking is mostly translated in the corpus examined – in other words, transference is the most predominant translation technique observed. This is the first evidence presented in this article regarding the dissimilarities between the translation of motion events and speaking events. Second, the German general verb sagen [say] is specified in Spanish in around 30% of cases (the additional frame elements included being Intention, Manner, Time and Topic length), probably for stylistic reasons (i.e. to comply with conventions of narrative texts in Spanish). In Spanish there is a tendency to avoid the repeated use of the general verb decir [say] in narrative texts and alternative forms appear frequently in translations (e.g. balbucear [babble], contestar [answer], insistir [insist], and the like). As I have indicated, the use of specific reporting-verbs can contribute to the definition of characters’ personalities and the setting of scenes. These data also point to a divergent treatment of motion and speaking events by Spanish translators. All in all, building on the findings outlined above, this paper provides further empirical evidence for the claim that the motion and communication frames in narrative texts are usually treated differently by Spanish translators; moreover, it does so based on a large online German>Spanish corpus (including a diversity of authors and translators, cf. Shi 2008: 205) and significant additional empirical data. I therefore argue that typological differences between source and target language do not seem so important – i.e. are not so consequential – in the case of manner-of-speaking as in that of manner-of-motion. It seems that translators into Spanish afford more importance to the description of ways of speaking than to the characterisation of motion events. The causes of these differences are stylistic in nature (i.e. any alteration of a communication event has potential consequences in terms of characterisations, scene setting and the author’s personal style). The resulting data also allow us to confirm that Spanish is not lexically poorer in verba dicendi than German, contrary to what has been confirmed for motion verbs in the literature (cf. Slobin 1997: 459). All in all, thinking-for-translating manner-of-speaking remains an under-unexplored issue, and interesting areas for further investigation would include comparable studies conducted with other linguistic/typological combinations (also considering diatopic variation – Berthele 2004, for motion events), as well as other frames (e.g manner-of-vision, manner-of-eating/drinking) and other genres (e.g. comics, drama, stories). The differences and similarities between literature for a general audience and for children and young adults is another potentially rich area of inquiry and could be pursued most effectively through the use of balanced and comparable corpora. Another potentially fruitful area of future research would be further examination of the asymmetry hypothesis (Klaudy and Károly 2005: 14) and of the potential explanatory power of translation universals applied to speaking events (cf. Cappelle 2012, for motion events).