The editor’s invisibility Analysing editorial intervention in translation

Most corpus-based studies of translation use published texts as the basis for their corpus. This overlooks interventions by other agents involved in translation such as editors, who may have significant influence on the translated text. In order to study editors’ influence on the translation product, this paper presents a comparative analysis of manuscript and published translations, which allows a differentiation of actual translated language and edited translated language. Based on a tripartite parallel corpus of English business articles and their translations into German, I analyse translators’ and editors’ influence on grammatical metaphoricity of the text, specifically on the use of nominalisations. One finding is that a significant amount of nominalisation is re-verbalised by editors. The results show that translated language may often be the result of significant editorial intervention. Thus, by just considering source text and published translation, our picture of what translators actually do may be significantly distorted.


Towards a holistic view of the translation workflow
Research on translated language usually limits translation to the act of translating. That eclipses the many other agents that are active in the translation workflow in the sense of a document production process as described by Gouadec (2007, Chapter 3). Along with Muñoz Martín (2010), I consider the translation workflow as "the period commencing from the moment the client contacts the translator and ending when the translation reaches the addressee, or when the translator is paid" (Muñoz Martín 2010, 179). We may also call this understanding the "agency of the translation as event […] which may indeed be the product of a fractured and multiple type of human agency" (Harvey 2003, 69).
Corpus studies of translation and multilingual discourse production have so far neglected to use corpora that reflect the production process of a translation, as "editing […] frequently remains invisible in conventional corpus-based studies comparing translated and non-translated language" (Kruger 2012, 354). A step towards such a holistic study of the translation workflow is the study of the effect of editing by a comparison of manuscript and published translations. This may "lead to improvements in the ecological validity of experimental settings" (Muñoz Martín 2010, 179; see also Saldanha and O'Brien 2013, 110) and "provide exciting opportunities for analysis of the language of translation" (Utka 2004, 223). The influence of editors should be taken into account whenever we discuss findings from corpus-based analyses of translations, especially where journalistic texts are concerned, as is done by Delaere (2015, 128). To promote the use of manuscript corpora, this paper reports on an analysis of grammatical metaphor in translated texts before and after editing. Specifically, I will investigate how translators and editors differently influence the nominal or verbal style of a text.

Nominal and verbal style in English -German translation
German is generally considered to prefer a nominal style and consequently a higher information density (Göpferich 1995, 420-421;Fabricius-Hansen 1999, 203;Krein-Kühle 2003, 160;Hansen-Schirra et al. 2009, 110). In spite of this, several studies have observed that translations to German often have a verbal style. This may be because translators often seem to turn source text nominal constructions into verbal ones, which has been taken as evidence for explicitation (Konšalová 2007). Alternatively, there may be shining through of the source language "as many verbal structures are translated literally" (Hansen-Schirra 2011, 147).
This apparently paradoxical situation leads to what Hansen-Schirra (2011, 136) calls "hybridisation" of verbal and nominal style. Following Teich (2003), English structural conventions should shine through in the German translations, resulting in a more verbal style than would be exhibited by non-translations. On the other hand, translators often adapt the translation to meet the norms of the target language, which has been termed "normalisation" by Baker (1996).
In their analysis of English and German translated texts, Hansen-Schirra et al. (2009, 112-113) find a noticeably large number of modified nominal phrases, from which they conclude that the translations are modified to fit the patterns of the target language. In terms of the preference of nominal and verbal styles, they also find that English translations have a more nominal style than German ones (117).
But how do we know that it is really translators who are responsible for this? The hybridisation paradox may be addressed by investigating how translators and editors engage differently in nominalising or verbalising source text constructions. A differentiation of translators' and editors' actions may shed light on which actors are responsible to what extent for the nominal or verbal style of the text.
In Section 2, I propose a possible structure of the translation workflow in the holistic sense and discuss editing in the translation context. Section 3 introduces the notion of grammatical metaphor and discusses its applicability to the study of translation, as well as the effects of using grammatical metaphorisation in translated text. In Section 4, I outline the method used in this paper to study the phenomena of nominalisation and verbalisation in the tripartite corpus. Section 5 gives examples of the most commonly found metaphorisation patterns and presents the quantitative analysis of the frequency at which metaphorisation patterns occur.

Studying editing in translation through corpora
'Editing' is defined as an activity that consists in comprehending and evaluating a text written by a given author and in making modifications to this text in accordance with the assignment or mandate given by a client. Such modifications may target aspects of information, organization, or form with a view to improving the quality of the (Bisaillon 2007, 296) text and enhancing its communicational effectiveness. While Bisaillon's definition only refers to texts "written by a given author, " it is surely applicable also to translations. Mossop (2014, 1) defines editing and revising as "reading […] in order to spot problematic passages, and making any needed corrections or improvements, " but restricts editing to texts that are not translations. I will use the term 'editing' in this paper irrespective of whether the texts are translations or non-translations.
A translation workflow may consist of the translation and editing stage as depicted in Table 1. I adopt Jakobsen's (1999) scheme of the translation act (see also Norberg 2003), according to which the reviser works with the text after the translator, but possibly still within the translation company that has been assigned the job (Rasmussen and Scholdager 2011). Editors then work with the text at the editing stage in the publishing company (see Table 1). Mossop (2014) defines four types of editing: stylistic editing, copyediting, structural editing and content editing. The latter two are to do with the organisation and order of the text and with additions or subtractions from the content  Ko 2011, 124). Copyediting is defined as bringing the text into conformity with grammar rules, "good usage" or a house style, and stylistic editing refers to the creation of a "readable" text by "making sentences more concise" or removing ambiguity (Mossop 2014, 30-31). This study draws on articles translated for the Harvard Business Manager (see Section 4). At this magazine, all the above types of editing are done by one person who constantly refers to the source text, according to Britta Domke, one of the magazine's editors (personal communication): Wenn wir mit der Redigatur eines übersetzten Textes beginnen, legen wir uns in der Regel den Originaltext aus der HBR daneben und vergleichen beides Satz für Satz, sowohl sprachlich als auch inhaltlich. Sprachlich arbeiten wir zum Teil noch stark an den Übersetzungen, je nach Qualität und Engagement des jeweiligen Übersetzers. So zerhacken wir Bandwurmsätze in leichter verständliche Einzelteile, formulieren Substantivierungen und Passivkonstruktionen um und streichen überflüssige Hilfsverben.
[When we start editing the translated text, we usually place next to us the original text from the HBR and compare both texts sentence by sentence, with regard to both language and content. We also edit the language of the translations, at times significantly, depending on the quality and effort of each translator. Thus, we split convoluted sentences into more comprehensible pieces, reformulate nominalisations and passive constructions and remove superfluous auxiliary verbs.] Thus, the editors work according to the house style guidelines for translators, which, in the section on language and style, ask translators to write comprehensively and lively and to avoid the nominal style, the passive and impersonal constructions, among other things.

Nominalisation and verbalisation as grammatical metaphor
This section will introduce the theoretical background to the study of nominalisation as grammatical metaphor. I will briefly introduce the notion of grammatical metaphor and how it has been applied to translation studies and contrastive analysis. I will then discuss the effects of grammatical metaphor in translated language and possible motivations for using it.
Nominalisation can be subsumed under the notion of 'ideational grammatical metaphor' (for an overview of the origins of this notion, see Taverniers 2003). Grammatical metaphorisation is defined as an action by which "processes and qualities are construed as if they were entities" (Halliday and Matthiessen 2004, 637). In languages such as English and German, where events, actions or states are typically realised as verbs, they would thus be realised as nouns by grammatical metaphor (Halliday and Martin 1993, 141).
In this process, there can be 'congruent realisations' , which can be considered the unmarked or "typical ways of saying things" (Halliday 1994, 321), as well as 'metaphorical realisations' , which are in some sense "different from that which would be arrived at by the shortest route" (Halliday and Matthiessen 2004, 658). While these definitions are somewhat vague, the term "deverbal nominalisation" (Heyvaert 2003, 66) is more precisely defined as relating to cases where processes, which would congruently be expressed as verbs, are in fact expressed metaphorically as nouns. They then function as Thing in the nominal group rather than as Process in the clause (Halliday and Matthiessen 2004, 656).
Metaphorisation has been argued to be intrinsically related to translating. Steiner (2001, 11) argues that de-metaphorisation is part of the process of understanding in the translation act, followed by re-metaphorisation as the translator drafts the target language text, which has been supported by empirical studies such as Hansen (2003, 150) for translations from German and French into English. Steiner argues that the question is then to what extent translators metaphorise their texts: [H]ere the process of re-metaphorisation is cut short below the degree to which it might otherwise go. The reasons could be language-specific (i.e. because of typological-contrastive properties of the languages involved), they could be registerspecific (i.e. in cases where the target language and context suggest a lower degree of metaphoricity), and/or they could have to do with a lack of effort or ability on the part of the translator -or, indeed, with some as yet unknown factor. (Steiner 2001, 15) In any case, Steiner (2001, 11) expects that the frequency of grammatical metaphorisation is lower in translations than in non-translated texts in the same language and also lower than in their source texts. That claim cannot be tested by the method employed in the present study, and, as Steiner says, it is hard to test anyway "as all the typological factors play a major role" (ibid.). Steiner's claim, however, seems to refer primarily to cases of re-metaphorisation, as he says that translators "will often not go all the way back up the steps of grammatical metaphorisation" (ibid.). Because of this focus on re-metaphorisation, it is not clear what role cases of "new" metaphorisation play in his account, that is, metaphorisations that the translator introduced, such as nominalisations of source text verbs. To explore the full extent of metaphorisation in translated text, it is necessary to arrive at a definition of grammatical metaphor that applies to contrastive analysis and includes all cases of metaphorisation.

Grammatical metaphor in contrastive analysis
Most existing studies of grammatical metaphor apply the notion in a monolingual context. Studies that apply the concept to the investigation of parallel translation corpora have been conducted on the CroCo corpus (Alves et al. 2010;Hansen and Hansen-Schirra 2012). Alves et al. (2010, 115) argue that through such studies, "grammatical differences and translation shifts across languages can be categorised more precisely and exhaustively than traditional approaches have proposed. " In their analysis of a parallel corpus of corporate communication, they show, among other things, that deverbal nominalisations are the most frequent part-of-speech shift in the English -German translation direction, which they partly attribute to the finding that English has more verbal forms than German, but also to the preference for nominal style in German (116). Analysing the same corpus, Hansen and Hansen-Schirra (2012, 145) find that "in most cases the translations adhere to the typical norms of the target language. " Also using the CroCo corpus, Rüth (2012) studies the development of metaphoricity between 1975-1985 and 2005-2011 in professional German discourse in order to find out whether popular science writing is becoming more technical. Among other things, she finds that the ratio between words belonging to nominal-type phrases (nouns, adjectives and prepositions) and those belonging to verbal-type phrases (verbs, adverbs, conjunctions; the differentiation is based on Steiner 2001, 26) has decreased over the time span she analyses, from 4.8 nominal-type phrases per verbal-type phrase in the earlier corpus to 4.3 in the later corpus (Rüth 2012, 78). She argues that this change towards a more verbal style simplifies cognitive processing of the text and thus increases comprehensibility of the texts for the targeted readers (82).
Apart from the above, few scholars have considered grammatical metaphorisation from a cross-linguistic point of view. Translation decisions may be analysed as expressing a source text form congruently or metaphorically, "as seen from the perspective of the source language to the target language" (Steiner 2004, 149). 1 In the field of translation studies, such a perspective is almost inherently taken by using terms such as 'source' and 'target' language. Thus, we talk about metaphorisation if "the target language expression is more metaphorical than the source language expression" (Alves et al. 2010, 115), and about de-metaphorisation if the opposite is the case.
Indeed, the strength of using the notion of grammatical metaphor in a contrastive analysis of translated texts is argued for by Steiner (2004, 159), who identifies as the strongest interpretation of "congruence" that where it "is taken to mean primary in some sense. " Halliday and Matthiessen (1999, 235) define "primary" in a "logogenetic" sense, that is, as something that "comes earlier in the text, " in "the unfolding of the act of meaning itself " (18). Along with the phylogenetic and the ontogenetic time frame, they go on to call them "the three major processes of semohistory, by which meanings are continually created, transmitted, recreated, extended and changed" (ibid.). I would argue that there is nothing to keep us from extending the logogenetic sense to include the act of translating a text.
Defining grammatical metaphor in such a cross-linguistic sense can help shed light on the hybridisation paradox mentioned above. In his account on the use of the notion of grammatical metaphor for contrastive analysis, Steiner (2004, 145) argues that "the mapping of semantic roles and grammatical functions in German seems to be more constrained than in English, " so that English has a higher tolerance of personalisation than German. This means that there may be cross-linguistic differences in metaphoricity and congruence even though the constructions are structurally identical (145). The semantic restriction of Agents to "animate and conscious participants" (Kunz 2010, 166) and consequent lower tolerance in German for personalisation is one reason why a nominal style is encountered more commonly in German writing than a verbal style. Steiner (2004, 155) further shows that increases in metaphoricity on one level can sometimes lead to decreases on another. Thus, if a clause complex is translated as a clause simplex, as in Example (1), taken from Doherty (1991, 155), metaphoricity is usually increased (Steiner 2004, 155). But, in accordance with his caveat about cross-linguistic differences in metaphoricity mentioned above, Steiner points out that a congruent translation in this case would actually lead to a "highly metaphorical personification" (ibid.).
[For the solution of this problem, two mechanisms have evolved in plants by which photosynthesis becomes overlaid/superimposed.] 2 That is an important observation for the popular science genre, which has been the basis of the majority of the previous research in this field and where the agents are often non-human beings. The prevalence of the phenomenon of personification, however, may differ in other genres. In the business genre, for instance, agents are usually human (managers, leaders, staff) or entities for which personification is acceptable (board, company). Example (2) from my corpus is highly similar to (1), but here the personification of Firmen ('firms') is acceptable, so that the editors chose to de-metaphorise the nominalisation.
(2) To get at that, some firms create markets for new customer information in (HBR 1/10,94) which employees rate the value of contributions. [To solve this problem, some firms put their trust in a system in which employees rate the customer information added by their colleagues.]

Effects of grammatical metaphor on the translated text
The use of grammatical metaphor in translation has several effects. In terms of the textual metafunction, metaphorical realisation allows the author to draw on the "Given/New" organisation of the information unit (Halliday and Matthiessen 2004, 642), as well as to treat an expression realised as a nominal group textually as a discourse referent (644). In the interpersonal metafunction, the effect of a metaphorical realisation can be that the expression is not given the interpersonal status of a proposition or a proposal, thus "making it inarguable" or presenting it as established (645). This happens when meanings are realised metaphorically by a group or a phrase, which, unlike propositionalised sequences, cannot be modalised, doubted or argued.
Example (3) shows a case of deverbal nominalisation in the draft translation that is reverbalised in the published version. Notice how the modalisation may even in the English source text is turned into often want in the translation. The intermediate version at the manuscript stage uses sometimes and thus represents a middle point between the source text and the published version. There seems to be a continuum in the loss of modality in this particular construction whereby a tentative suggestion becomes a strong claim as the text passes through the stages of the translation workflow. That loss of modality is at least partly due to the use of grammatical metaphor (see Alves et al. 2010, 117 for a similar example where nominalisation leads to the loss of modality).
(3) People may even shun the development of new resources in order to preserve (HBR 7/10,102) existing values and retain power. Another effect of grammatical metaphor relates to the comprehensibility of the translated text. The use of grammatical metaphor may increase lexical density and thus make translations more difficult to process. Halliday and Martin (1993, 86) argue that "a great deal of semantic information is lost when clausal expressions are replaced by nominal ones. " While the general impression is that nominalisations are more difficult to understand than active verbs (Ravelli 1988, 144-145;Taverniers 2003, 25-26), most scholars that study the phenomenon do not go further in the interpretation of their data than to speculate that high metaphoricity may make a text less comprehensible (Coleman 1964;Lassen 2003, 24, 164-165;Müller-Feldmeth et al. 2015, 250). There is even some evidence that a higher degree of grammatical metaphoricity does not make a difference to comprehension (Palumbo 2008). Müller-Feldmeth et al. (2015, 241) show that popular science authors use nominal style "to introduce scientists and their work, " and verbal style "to explain scientific ideas. "

The manuscript corpus and method
This study draws on a tripartite parallel corpus "in which two or more components are aligned, that is, are subdivided into compositional and sequential units (of differing extent and nature) which are linked and can thus be retrieved as pairs (or triplets, etc.)" (Fantinuoli and Zanettin 2015, 4). A traditional translation corpus of 26 English source texts from 2006 to 2011 and their German translations is complemented by a corpus of the manuscripts of those translations. The manuscripts were sent to the publisher by the translation company Rheinschrift 3 and thus represent translated language before it underwent the editing process. Five different translators have translated the texts at that company, and six different editors have worked on the text at the Harvard Business Manager. The overall size of the corpus is 315,955 words (see Table 2). Up to fairly recently, instances of grammatical metaphor have been considered "hard to annotate and extract automatically" (Teich 2003, 157) so that most analyses have been conducted manually (see Rüth 2012, 56 for an overview). Steiner (2001) suggested some ways of automatically analysing grammatical metaphoricity of texts by measuring lexical density, average sentence length, typetoken-ratio or the relation between prepositions and conjunctions, but those measurements only allow statements about the text as a whole and not about specific instances of grammatical metaphor.
A further step towards the corpus-based study of individual instances of grammatical metaphor was taken by Rüth (2012), who uses the Canoo Unknown Word Analyser for a derivational analysis. This allows her to register shifts in part of speech such as nominalisations. The most advanced method to date is used by the CroCo project, whose corpus features a multi-layered alignment and annotation, which allows the identification of shifts in part of speech and thus instances of grammatical metaphor (Alves et al. 2010, 117; see also Hansen and Hansen-Schirra 2012). As the present corpus does not feature such a complex alignment, I have conducted a semi-manual analysis adopting parts of the strategies employed by Steiner (2001) and Rüth (2012).
The three subcorpora were first part-of-speech tagged using TreeTagger (Schmid 1995), which uses the Stuttgart-Tübingen-Tagset for German (Schiller et al. 1999) and the Penn Treebank tagset for English (Marcus, Santorini, and Marcinkiewicz 1993). The three subcorpora were then sentence-aligned and manually analysed for shifts between noun and verb in order to discover discrepancies between translation manuscripts and the published translations.
The following strings indicating nominalisation and verbalisation were analysed in the corpus. The -ung/-ungen suffix, shown in Example (4), is the most productive nominalising suffix in German, at least with regards to deverbal nominalisation (Demske 2000;Shin 2006).
(4) It's easy to surmise that many of them felt unsure of their readiness for the (HBR 1/09,54) role -we think of them as "much to prove" leaders.

Es drängt sich die Vermutung auf, dass sich viele von ihnen nicht sicher waren, der Aufgabe gewachsen zu sein. Wir stellen sie uns als Führungskräfte vor, die viel zu beweisen haben. (manuscript)
Da drängt sich die Vermutung auf, dass viele von ihnen zweifelten, ob sie der Aufgabe gewachsen sind. Wir nennen sie "Führungskräfte auf Bewährung". (HBM 3/09,10) [This begs the assumption that many of them were not sure whether they were up to the task. We think of them as leaders who have much to prove.] [This begs the assumption that many of them doubted whether they were up to the task. We call them "leaders on probation".] A common lexeme formation pattern for verbalisation is -ier-(see Neef 1999, 216), as in Example (5).
(5) In the process, such leaders raise the level of emotional engagement that (HBR 6/12,76) employees bring to company life in general. Other search strings are the preposition-article combinations beim, zum, bei der and zum and the English gerund marker -ing, which also tends to trigger nominalisation in its translation (see Rumpeltes, in preparation). Table 3 gives a detailed overview of all the metaphorisation patterns that the analysis has brought up, given in absolute figures and percentages of the total. The metaphorisation patterns are listed in three-part abbreviations that represent the three steps from source text via the manuscript to the published text. To allow for a more detailed view in this initial step, I have separated verbal gerunds (Vg) and nominal gerunds (Ng) in English, and verbal infinitives (Vi) and nominal infinitives (Ni) in German. For instance, VgNNi means that the source text has a verbal gerund (Vg), the manuscript has a full nominal form (N) and the published version has a nominal infinitive (Ni).

Quantitative findings
The vast majority of nominalisations used the -ung suffix. Nominalisations that use nominalised infinitives (Ni, 40 occurrences) are much less frequent. The majority of them (31 occurrences) occur in patterns where the published text has retained a nominal form (VNiNi, VgNiNi, VgNiN). They occur much less frequently (only 9 times) in patterns where there is a verbal form in the published document (VNiV, VgNiV). Editors thus seem much less likely to reverbalise a nominalisation if it occurs as a nominal infinitive than if it occurs in a -ung form. As my investigation focusses on nominalisation and verbalisation, those patterns involving adjectives (11 instances) at any point in the translation workflow are excluded from the analysis. Patterns where the form was omitted at some point (also 11 instances), either because the editors deleted content or because the translator has added it, are also excluded from further analysis; only occurrences where all three texts had either a nominal or a verbal form have been listed.
Assorting gerunds and infinitives to their respective verbal or nominal classes, we can divide the data into six metaphorisation patterns, which are given in Table 4. Editors maintain a large part of translators' nominalisations (yielding the VNN pattern). Given their statement that they regularly consult the source text (see Section 2), they will be aware that these are nominalisations and thus seem to agree with the translators' decision to nominalise. The second most common pattern (VNV), however, is the editors' "unpacking" of the grammatical metaphor (Halliday and Matthiessen 1999, 255) when they reverbalise a significant amount of the translators' nominalisations. In addition, the third most commonly observed pattern consists of cases where editors verbalise nominal forms that the translators maintained (NNV). At the same time, nominalisations of verbal forms that the translators maintained from the source text (VVN) are rather rare.
Thus, around 15% of the editorial changes are new metaphorisations, that is, metaphorisations of a form that the translator maintained from the source text (VVN and NNV). Some 30% of metaphorisations are cases where the editors changed the manuscript translation to revert a metaphorisation that the translator introduced (VNV and NVN). 5 Regarding the grammatical metaphoricity under analysis, both manuscript and editing stages affect the text to about the same extent. At the manuscript stage, the most common metaphorisation pattern is deverbal nominalisation (VNN and VNV), which happens a lot more frequently than denominal verbalisation. The editing stage, on the other hand, affects the text mainly by verbalisation (VNV and NNV). Editors are much more likely to verbalise nominal structures than to nominalise verbal structures, especially when the source text also has a verbal form. This difference is summarised in Table 5.

Discussion of the findings
Example (7) demonstrates the metaphorisation patterns VgNV (underlined) and VNN (in bold). The gerundive Increasing has been translated as the noun Die Steigerung, and then changed into the verb erhöhen by the editor. The verb to build was also translated as the noun Steigerung, which was kept by the editor, though in accusative case instead of genitive.

5.
Because the translation workflow is viewed in the logogenetic sense (see Section 3.2), I still consider these cases to be metaphorisations, because a shift occurs in comparison to the previous step in the creation of the text.
(7) Increasing brand equity is best seen as a means to an end, one way to build (HBR 1/10,94) customer equity.  (HBM 3/10,86) [The increase of a brand's value is best seen as a method to achieve the more important goal, the increase of the customer's value.] [To augment the value of a brand can at best serve to achieve a more important goal: the increase of customer value.] For an example of the NVV pattern, see (5) above. Example (8) demonstrates the metaphorisation pattern NNV. The source text noun phrase improvements in profitability has been translated as the equivalent but somewhat clunky noun (Rentabilitätssteigerung). That clunkiness may have been the reason why the editors have changed the noun into the more readable verb phrase steigerten die Rentabilität.
(8) The communications plans determined by our model resulted in sharp (HBR 3/06,131) improvements in profitability. The metaphorisation pattern VVN is exemplified in (9), where the verb phrase to build a national presence was translated as the verb phrase eine landesweite Präsenz zu schaffen. The editors then condensed that phrase into the noun Expansionsbemühungen.
(9) (HBR 9/08,82) But it got reckless in its attempts to build a national presence.

Doch in seinen Versuchen, eine landesweite Präsenz zu schaffen, wurde das Unternehmen leichtsinnig. (manuscript)
Doch bei seinen Expansionsbemühungen wurde das Unternehmen leichtsinnig. (HBM 12/08,96) [But in its attempts to build a national presence, the company got reckless.] [But in its attempts of expansion, the company got reckless.] Gerunds, which share attributes of both nouns and verbs (Mackenzie 1996, 326), are commonly differentiated into verbal and nominal gerunds (Abney 1987;Houston 1989;Alexiadou, Iordăchioaia, and Schäfer 2011, 159). Thus, a gerundive form is considered to be nominal if its object is a prepositional phrase and it is modified by an adjective, and it has a verbal character if it licenses an accusative object and is modified by an adverb (Alexiadou, Iordăchioaia, and Schäfer 2011, 28), as is the case in Example (10).
(10) Developing new resources internally is faster and more effective than acquir-(HBR 7/10,102) ing them from external parties […] Die interne Entwicklung neuer Ressourcen ist schneller und effektiver als der Kauf bei externen Anbietern […] (manuscript) Selbst zu entwickeln ist schneller und effektiver, als etwas von externen Anbietern zu kaufen. (HBM 2/11,84) [The internal development of resources is faster and more effective than their acquisition from external parties.] [Developing [something] yourself is faster and more effective than to buy something from external parties.] A similar distinction applies to nominal and verbal infinitives in German. Analogous to Mackenzie's (1996, 326) continuum from fully verbal to fully nominal expressions, nominal infinitives may be argued to be closer to the verbal spectrum than -ung nominalisations because they maintain more characteristics of verbs such as the same form as the verbal infinitive and the lack of a plural form.
Example (11) shows the phrase das räumliche Zusammenrücken der Mitarbeiter (literally 'the spatial moving together of staff ') as a translation of bringing people closer to one another, where the -ung nominalisation Beseitigung in the manuscript translation has been turned into a nominal infinitive, das Beseitigen, in the published translation.
(11) There's some evidence that removing physical barriers and bringing people (HBR 7/11,102) closer to one another does promote casual interactions. Finally, the published translation in Example (12) shows a verbal infinitive in German (Kopieren, 'copying') while the manuscript translation has a nominal infinitive (das Kopieren, 'the copying').
(12) The sense of permission was strengthened by the fact that copying is perceived (HBR 7/11,102) as work. [Since copying is seen as work, this strengthened the feeling that these conversations were allowed.] Of course the method applied in this study has some limitations. It is difficult to elicit, for instance, how metaphorical individual nominalisations are. The main example for this issue in the present corpus is the word Wertschöpfung ('value creation' , 'added value'). In the source texts, authors readily use verb phrases such as create value or add value. In German, I would argue that Wertschöpfung does not really have a verbal equivalent. While there are two instances of such a verbal form in this corpus (see Example (13)), German newspaper reference corpora show very few instances for the lemma Wert schöpfen. 6 The existing uses of this verb phrase may be said to exist as calques from English.
(13) So if there is some chance that a deal between a buyer and a seller can create (HBR 12/09,101) extra value, it's better to negotiate than to hold an auction. [If there is a possibility that the closing of a deal between a buyer and a seller will lead to an extra value creation, negotiations rather than auctions should take place.] [If there is a possibility that the closing of a deal between a buyer and a seller will create extra value in the case of negotiations, you should negotiate instead of conducting an auction.] The question is, by nominalising the verb phrase create extra value, how much does the translator metaphorise, if there is no verbal equivalent? And how "con-6. See dwds.de/?view=3&qu="Wert schöpfen". gruent" can we call the editor's use of Wert schöpfen, if the phrase is unconventional in German? These issues can only be tackled by a qualitative analysis following a corpus-based identification of metaphorisations (see Bisiada 2017 for such an analysis).
To conclude, the VNV metaphorisation pattern is the most interesting one because it shows the value of the analysis of the manuscript stage in the translation workflow. The 202 forms that make up the VNV pattern would be considered literal (congruent) translations in conventional analyses, when in fact in the actual translated text they are nominalisations. Thus, in an analysis that ignores the effect of the editing stage, almost a third of metaphorisation may go unnoticed, which shows that editing can have a significant effect on the translated language.
The diverse influences on the text discovered in this study may provide some clues to the explanation of the hybridisation paradox (Hansen-Schirra 2011, 136) mentioned at the outset. At least in the business article register, translated text may be a product of the hybrid nature of the translation workflow where many agents intervene, not just of the act of translation itself. Similar evidence for the register of newspaper texts is provided by Delaere (2015, 161), who finds that "genres which go through an editorial process are more likely to contain standard language" than genres which do not.
For the business genre, there is evidence that editors split sentences to a significant extent in order to increase the readability of the text (Bisiada 2016), an aim that is perhaps not at the top of translators' aims when producing their translation. Andújar (2016) has provided evidence for the literary register in support of the claim that editors' influence standardises the language in the text, eliminating, for instance, translators' attempts to recreate orality in their target texts (Andújar 2016, 417).
The studies cited above form part of a recent movement to inform corpusbased contrastive and translation studies by providing a more fine-grained view of how the translation product is created. Research into the influence of editors on the translated text demonstrates that contrastive and translation studies cannot rely solely on published texts to inform their statements about translated language. Scholars should, where possible, take into account the fact that interventions have taken place on the translated text, which, depending on the genre, may be significant enough to warrant dedicated study.
The findings support the idea that translation production is a multi-agent activity, and should be studied as such. If the published target text, the final product that the vast majority of recipients see, can be shown to be rather different to, say, the text that the translator originally produced, then the study of translated language should also take a multi-layered view, with a firm account of whether the language being studied is that encountered in the final product or in any intermediate stage that preceded it.