The translation of food-related culture-specific items in the Valencian Corpus of Translated Literature (COVALT) corpus: a study of techniques and factors

ABSTRACT This article aims to analyse the translation of food-related culture-specific items (CSI) in the English–Catalan subcorpus of the Valencian Corpus of Translated Literature (COVALT). This general aim can be broken down into two specific aims: to find out what techniques prevail in the translation of these cultural items, and to determine what factors influence the choice of specific techniques. Corpus analysis is carried out by means of the Corpus Query Processor. The theoretical framework deals with the definition and scope of the concept of CSI, the classifications of techniques put forward in the literature for the translation of CSI, and the position of food- and drink-related elements within the broader category of CSI. Analysis of the results yielded by the corpus shows that neutralising techniques prevail over foreignising and domesticating ones, with the latter coming last in descending order. The most prominent factors identified are non-existence of the source text (ST) item in the target culture, different degrees of institutionalisation, the ST item having been imported into the target culture, and different degrees of granularity. Correlations between techniques and factors are never very strong, but some are strong enough to deserve further attention.


Introduction
This article aims to find out how food-related culture-specific items (CSI) fare in literary translation through an analysis of the English-Catalan section of the Valencian Corpus of Translated Literature (COVALT) corpus. Food-related items are often culture-specific and can therefore be regarded as a potential source of translation problems. In Section 2, a brief overview will be provided of the notion of CSI, followed by an account of classifications of translation techniques and of the place occupied by food in cultural approaches to translation. In Section 3, the main aims of the study, as well as the methodology employed, will be spelled out. In Section 4, the results of the analysis will be presented and discussed, with an emphasis on the relative frequency of translation techniques and the correlation 1 between techniques and factors impinging on translators' decisions. In Section 5, some conclusions will be drawn from the data.
As the literature review in Section 2 will make abundantly clear, many studies are already available on the translation of CSIs, including food-related ones. Why, then, another? The main contribution of the present article lies in two of its defining features: use of corpus-based methodology and consideration of the factors influencing translators' decisions. Most previous studies are case studies focusing on a single literary or audiovisual work or, at best, on a limited number of such works. Extremely valuable as such studies are, their results have limited validity as they cannot be held to be representative of a more general translation field or context. The results of the present study, on the contrary, are representative of one such field: the translation of English narrative works into Catalan in the region of Valencia between 1990 and 2000since those are the parameters defining the English-Catalan subcorpus of COVALT. As to factors, they are viewed as explanatory concepts that allow the analyst to go beyond description.
2. The translation of food-related culture-specific items 2.1. Culture-specific items Even though culture permeates the whole text a translator is faced with, it makes itself particularly felt at certain points by means of references to objects, situations, ideas, beliefs, values, etc. that belong to the community in which the text has been produced. In translation studies, these references have been variously referred to as cultural references (or referents), cultural elements, culture-specific items, realia or culturemes. All these terms intend somehow to point to a concept that can be more or less intuitively grasped, at least by translators and translation scholars. But the concept itself is beset by two problems, which need to be briefly addressed.
The first problem is one of definition: must the concept include the element of difference or not? Most translation scholars regard this element as fundamentalsee, for instance, Bödeker (1991, p. 65), Franco Aixelá (1996, Davies (2003, p. 69) or Olk (2013, p. 346). In the present paper, the element of difference in the definition of CSI will be taken as a pre-requisite, and accordingly only those food-or drink-related items in our corpus that refer to realities that either do not exist as such in the target culture or reveal significant mismatches across cultures will be regarded as culture-specific.
The second problem concerning the concept of CSI is scope. Given the comprehensive definition of culture advocated by anthropologists and scholars from other disciplines, it must be rather difficult to provide a thorough classification of cultural fields, or domains, because their sum total must equal everything, i.e. all it takes for a human being to be able to live in a given community. Even so, translation scholars have made several attempts at such a classification. A review of those attempts should include at least Nida (1945), Newmark (1988), Florin (1993), Katan (1999), Molina Martínez (2006 and Tymoczko (2007). In this respect, Section 2.3 will explore how food and drink fit into the big picture of CSIs.

Translation techniques for culture-specific items
Most studies on the translation of CSI make use of some kind of typology of techniques, or procedures, to account for the type of relationship that obtains between source text (ST) cultural items and their matching target text (TT) segments. Even though it is far beyond the scope of the present paper to provide an exhaustive account of the typologies proposed, in the next paragraph I will briefly refer to a number of them, and discuss some of their similarities and divergences. Newmark (1988) put forward a very thorough classification, whose main weakness might lie in its redundant nature and the fact that categories are discrete and not arranged according to any principle (see Marco, 2004Marco, , 2007. Franco Aixelá (1996), in a very comprehensive study, did arrange translation procedures according to their degree of intercultural manipulation: some tend towards conservation (of the ST cultural item), others towards substitution (by a target culture item). Davies (2003), while praising several aspects of Franco Aixelá's contribution, was sceptical of the possibility of ranking procedures on a scale according to their degree of adaptation. Therefore, she provided a list of strategies with no underlying principle. Valdeón (2008) looked at proper nouns and consumer goods from the point of view of the contribution they make to irony in an American sitcom. Translation techniques for these items include (2008, p. 216) preservation, and substitution with a different source-culture item, with an international item, with a target-culture noun, with corrupted forms of target-culture items, with a superordinate or with a target-culture related item. Frank provided not a full-fledged classification like the above, but a sort of minimalist threefold distinction (2009, p. 9): 'Faced with cultural markers, the translator has the choice to leave them intact, to give equivalents, or to provide neutral terms. ' Olk (2013) provided a classification including seven categories: transference, transference + explicitation, transference + explanation, target-language expression referring to the source culture, neutral explanation, omission, and substitution of a cultural reference by a cultural equivalent. These categories were ranked on an exoticising/naturalising scale. Orozco Jutorán (2014) distinguished between cases in which a given concept is available in both the source and target cultures (and therefore conceptual equivalence is possible) and cases in which the concept in question does not exist in the target culture (equivalence being then possible only on the linguistic, not the conceptual, plane). Conceptual and linguistic equivalence are realised through a number of different techniques. Finally, De Marco (2015), in an account of how New Zealand food is translated into Italian in guidebooks, identified the following strategies: neutralisation and toning down; chunking and generalisation; naming and translating; and clarification and supplementing.
Davies suggested that there is 'considerable overlap' (2003, p. 70) between the translation procedures identified by such authors as Newmark (1988), Hervey andHiggins (1992), Franco Aixelá (1996) and Katan (1999), and the same seems to be true of most classifications outlined in the previous paragraph. But there are also divergences.
My own approach to the classification of translation techniques for CSI (Marco, 2004(Marco, , 2007 partly overlaps with several of the classifications above, and is particularly close to Olk's in that it merges the two criteria of foreignisation/domestication and translator intervention. Translation techniques are thus aligned on a cline with maximal distance from the target reader at one end and minimal distance at the opposite end. The actual techniques identified are the following: (a) borrowing of the ST item, which may be pure or naturalised (i.e. adapted to the spelling and morphology of the target language); (b) literal translation; (c) neutralisation (defined as replacement of the ST item with a more or less lengthy or detailed explanation of its import), which may take the form of description or generalisation/particularisation (i.e. replacement with a more general or more particular item, even though the latter option is not frequent). At any rate, neutralisation entails deletion of the culture-related item as such (as remarked by Olk, 2013); (d) amplification/compression: a certain amount of information is added or omitted in the TT when compared to the source (even though compression will probably be rare); (e) intracultural adaptation, in which the ST culture-related item is replaced by another item also belonging to the source culture but more familiar to TT readers. It could be argued that this technique should be placed closer to the foreignising end of the scale, as it implies using a source CSI. That is undoubtedly a sound argument. However, it is placed near the opposite end for two reasons: because it involves a high degree of translator intervention and also of distance from the ST; and because it is an attempt to bring the text closer to the target reader by making it less foreign; (f) intercultural adaptation, in which a target culture item is substituted for the ST item; (g) omission, which is included among domesticating techniques because it involves removing the traces of source culture specificity from the TT.
Techniques a and b may be said to stay close to the ST and not to bridge the distance separating it from the target reader; techniques e, f and g bring the text closer to the target reader; and techniques c and d do make an attempt to bridge the cultural gap but keep, at the same time, a safe distance from the target pole. These techniques will be used below in my discussion of the data yielded by the corpus. 2

Food as culture in translation studies
As far as my particular purposes in this paper are concerned, it should be stressed that food cuts across most (if not all) of the categories distinguished in the classifications mentioned in Section 2.1. Newmark's (1988), for instance, is based on a five-fold distinction: ecology, material culture, social culture, organisations, and gestures and habits. Food and drink are explicitly mentioned by the author under the heading 'material culture', but most raw materials in food come from natural sources and belong therefore to the domain of ecology; meals often become social occasions, both in work settings and in our leisure time; they can certainly transcend the private sphere and play an important role in the life of political or religious organisations; and they definitely constitute habits, often punctuated with gestures (e.g. table manners) and rituals (e.g. table prayers and blessings). Montanari (2004, pp. xi-xii) claimed that food is culture when it is produced, prepared and consumed; and later on added (2004, p. 73) that 'the organ of taste is not the tongue but the braina culturally (and therefore historically) determined organ through which value criteria are learned and transmitted '. 3 No wonder, then, that food should feature prominently in several empirical studies on the translation of CSI. The results of these studies seem to suggest that the treatment of CSI in translation, unsurprisingly, is sensitive to such factors as genre, type of readership, function of cultural items in the ST, cultural distance between the two languages involved, etc. Davies (2003, p. 92) summarised the treatment of food-related items in translations of the Harry Potter books by saying that '[t]he overall impression is very much of a haphazard treatment, where each reference seems to be dealt with in an ad hoc fashion without any clear underlying strategy. ' Inggs (2003, p. 288), as regards Russian translations of one of C.S. Lewis's Narnia books, claimed that 'both translations […] adopt strategies of simplification or clarification', which often involve the use of a more general word. In the case of the Harry Potter books, the translators 'have retained intact both the cultural backdrop and the moral values put forward in the works ' (2003, p. 295). Franco Aixelá (2009, p. 10) pointed to 'general neutralisation of culturally specific aspects of the original work' as the prevailing tendency in French translations of the quintessentially British Paddington Bear stories. Similarly, Mussche and Willems (2010, p. 491) concluded that 'attempts at domesticating the text are rare in the Arabic translation of Harry Potter', the prevailing tendency being neutralisation. Omission also plays a major role in those translations.
The Translator's 2015 special issue, guest-edited by Delia Chiaro and Linda Rossato, is an important step ahead in the study of the interface between food and translation because of its monographic nature. Several contributors drew analogies between both domains. Cronin (2015) looked at the implications of slow translation, the alleged counterpart of slow food; Desjardins, Cooke, and Charron (2015) explored the relationship between food studies and translation studies in Canada, which is governed by the asymmetrical position of the former in the English-and French-speaking communities. Rossato (2015) presented the extremely interesting case of British chef Jamie Oliver's particular journey to Italy, the TV series and cookbook ensuing from it, and their Italian translations. This article reminds us in a very illuminating fashion of the fact that food translation is inextricably bound up with issues of power and, most notably, identity. The same fact emerges in De Marco's (2015) contribution. Finally, Gaspari's article (2015) is unique in this volume in that it used corpus-based methodology to analyse food-related phraseology in the English and Italian versions of the descriptions of national pavilions at the universal exhibition Expo Milano 2015. Results, based on a comparallel corpus, uncovered mismatches not strictly related to typological differences between the two languages.
Special attention should be paid to Oster and Molés Cases's (2016) study, which focused on three groups of food-and drink-related items in the German ST component of the COVALT corpus (282,739 words) and how they fare in Catalan and Spanish translation. The three sets of items are culture-specific foodstuffs and drinks from Germany or Austria, actions carried out while eating and drinking (such as chewing or sipping), and metaphorical expressions in which food is the source domain (such as bitterness). For the first set (the other two are not so relevant for my present purposes), the authors found out that the prevailing techniques were generalisation and description, both of them neutralising techniques. Intercultural adaptation also features prominently in the data. From the perspective of the foreignisation/domestication scale, this means that translators endeavour to bring the text closer to the target reader by reducing source-culture specificity; from the perspective of culturicity, or cultural markedness, this means that over 50% of the translation solutions are not CSIs in themselves. Since there are obvious parallelisms between Oster and Molés-Cases's work and my own (both retrieve data from COVALT, and the methodology is very similar), their results will be used as a touchstone for mine in the conclusions.

Aims and method
As stated in the introduction, the present article is a study on how food-related CSI are dealt with in the English-Catalan subcorpus of COVALTa multilingual corpus made up of the translations into Catalan of narrative works originally written in English, French and German published in the autonomous region of Valencia from 1990 to 2000, together with their corresponding STs. 4 The English-Catalan subcorpus comprises 36 English STs, amounting to 1,201,757 words, and their corresponding TTs in Catalan (1,343,631 words). Table 1 includes the titles of both STs and TTs, as well as the names of authors and translators.
The main aims of the study can be stated as follows: (a) to find out what translation techniques prevail in the translation of food-related CSI; (b) to determine what factors (if any) may be said to condition the choice of specific techniques.  The method employed in the present study unfolded in the following seven steps: (1) A word list was extracted from the ST component of the English-Catalan subcorpus of COVALT by means of CQPweb, 'a web-based corpus analysis system' (Hardie, 2012) that is a web adaptation of the original Corpus Query Processor (CQP). CQP is the central component of the IMS Open Corpus Workbench, a set of tools for corpus management developed at the Institut für Maschinelle Sprachverarbeitung of the University of Stuttgart.
(2) All food-related lexical items on that word list were (manually) identified.
(3) The choice of search words was narrowed down. Since the outcome of step 2 was a list of extremely heterogeneous items (including not only foodstuffs proper but also ways of cooking, adjectives describing tastes or textures, eating places, names of meals, eating and drinking utensils, etc.), items to be entered later as queries were restricted to three kinds: foodstuffs proper, ways of cooking (e.g. baking, frying, scrambled) and names of meals (e.g. breakfast, dinner, supper). Incidentally, items from other groups were included if they looked particularly promising. 5 (4) Bilingual concordances for the search words chosen in step 3 were extracted by means of CQPweb. (5) Concordances were analysed in order to identify food-related words or word strings that may be regarded as CSIs in that they contain that element of difference which is essential in the definition of CSI subscribed to in this article. (6) Results (ST CSI plus their matching segments in the translation) were copied onto an Excel spreadsheet in order to classify segment pairs by translation technique and make the relevant counts. (7) Segment pairs were classified by factor (on the basis of a data-driven list of factors impinging on the choice of techniques) and the relevant counts were made, using the same tools as in step 6.

Results and discussion
The word list yielded via step 1 consisted of 37,573 word types. These were not lemmas but forms, as, even if the corpus is tagged for lemma, the word list utility provides lists of words, not lemmas. After a first manual scan of this list, a second list was obtained, which included 'all' food-related lexical items. The inverted commas suggest that it is impossible to guarantee a 100% degree of recall, because a word list displays lexical items in isolation and the analyst is therefore bound to judge intuitively, with no help from context. However, it seems reasonable to assume that this method enables the analyst to retrieve most food-related items in the corpus, as only very opaque items would fail to be detected. (For instance, 'lady fingers' was only detected because it occurred in the vicinity of a search word, 'mayonnaise'. Otherwise it would have remained hidden, since neither lady nor fingers would have been regarded as likely to yield relevant, foodrelated results.) This second list comprised 1,212 items. As explained in the previous section, it was further narrowed down according to three specific criteria in order to reach a manageable number of search words, which turned out to be 459 (see Table 2 for a list of the top 25 word forms on the frequency list). These words were inserted as queries in the CQPweb query box. The concordancing process was somewhat simplified by the fact that searches were made by lemma, so that a single query (say, [lemma = 'potato']) might yield results for more than one of the 459 search words (in my example, for both 'potato' and 'potatoes'see Figure 1).
Step 5 was the final one in the identification of culture-specific food-related items. The four previous steps might be said to be stepping-stones in the process of gradually retrieving useful material from the mass of thousands of word types in the corpus. Some of those word types occurred hundreds of times (e.g. 'meat', 220 times), whereas others were hapax legomena (i.e. they featured just one occurrence), with many intermediate possibilities.  Tables 3 and 4. Table 3 offers a list of the most productive queries, i.e. of search words yielding ≥5 relevant results. Table 4, by way of illustration, includes the food-related CSIs retrieved for the search word pudding, one of the top items on list 3, together with their corresponding TT segments and classification by translation technique.) That is the end-point of the whole sifting processthe total number of food-related CSIs found in the corpus. 6 The results of step 6, classifying segment pairs by translation technique and making the relevant counts, are displayed in Table 5 and Figure 2. If translation techniques are considered individually, the most frequent one is literal translation (21.12%), followed by generalisation (19.92%), description (19.52%), intercultural adaptation (13.55%), pure borrowing (7.97), intracultural adaptation (7.57%) and naturalised borrowing (6.37%). The remaining techniques are used only marginally or not at all (amplification). If we group techniques together according to where they stand on the foreignisation/ where things are explained in such a way that the explanation cannot be said to belong to either the source or the target cultural milieus. When they do choose between one of the two ends, they incline to the source more often than to the target culture (35.46% vs. 23.11%). In step 7, a qualitative and quantitative analysis was carried out with a view to setting up links between translation techniques used and possible factors influencing the choice of  techniques. The qualitative part of this analysis aimed at identifying relevant factors, and the following were found: 8 (a) non-existence: the ST foodstuff simply does not exist in the target culture (e.g. shepherd's pie, pancake, toad-in-the-hole, plum-duff);   (b) degree of institutionalisation: the ST foodstuff, or something similar, may be known in the target culture, but its degree of institutionalisation (i.e. its cultural markedness) is much higher in the source than in the target culture (e.g. fish and chips, packed lunch, baked beans, cornflakes, Girl Guide cookies[); (c) imported CSI: the ST foodstuff has been imported into the target culture, but it is still perceived as foreign, as evinced by the fact that the word used to refer to it is a loan word (e.g. pudding, curry, cheddar cheese, beef steak, doughnut, toffee); (d) degree of granularity: a given concept, or lexical field, is more highly developed in the source than in the target culture, thus showing a higher degree of granularity in the former than in the latter. In other words, where the concept in question is more highly developed, more distinctions are made, and they are more fine-grained (e.g. compare the beer field in English, which finds expression in the corpus in such varied terms as lager, ale, stout or hop-bitters, and in Catalan, where the general word cervesa is always used, minor qualifications being introduced by means of adjectives or prepositional phrases); (e) false friendship: the two cultural systems have functional equivalents for a given concept, but those equivalents either are only apparent (e.g. the words in Easter egg and ou de Pasqua have exactly the same meaning, but the foodstuffs they refer to are not exactly the same) or have differing levels of currency (e.g. pint, as a measure of capacity, is in full use in many English-speaking countries, whereas its Catalan equivalent pinta is regarded as obsolete); (f) figurative use: the foodstuff in question is used figuratively, either in free combinations ('a face like mashed potatoes') or in idioms ('nutty as a fruitcake'); (g) proper noun: the ST foodstuff is a proper noun used to refer to a character ('Ginger Nut' in Melville's Bartleby); (h) third culture: the ST foodstuff is neutral as far as the ST and TT cultures are concerned, since it belongs to a third culture (e.g. crêpe au chocolat, Rhein-wein, pemican); (i) trademark: a brand name, or trademark, is mentioned, instead of the generic term (e.g. Yorkie, Jamieson, Horlicks, Marmite); (j) polysemous CSI: a given word may designate several types of food, thus giving rise to potential ambiguities and misunderstandings (roll in such foodstuffs as white rolls, jam rolls or cabbage rolls); (k) invented CSI: the foodstuff in question does not exist in the real world but has been invented for fictional purposes (e.g. glimigrim, Fian bread, peya-roots).
The quantitative part of step 7 involved analysing the correlation between translation techniques and conditioning factors. In what follows, the results of that analysis will be reviewed, starting from the foreignising end.

Pure borrowing
Pure borrowing is used 20 times. Unsurprisingly, it occurs three times in cases where the ST item has been imported into the target culture ('cheddar cheese' → 'formatge cheddar'), three times with invented CSI ('Fian bread' → 'pa fian') and three times when a third culture is involved ('crêpe au chocolat' → 'crêpe au chocolat'). These factors seem fully compatible with preservation of source CSI. It is perhaps more surprising, though, that pure borrowing should also occur under less favourable circumstances, such as non-existence of the CSI in the target culture (once, in 'arrowroot biscuits' → 'pastissets d'arrowroot' ['little cakes of arrow-root']a food item that hardly any target reader would be familiar with), use of a trademark (four times, as in 'Marmite' → 'Marmite') or a widely different degree of granularity (six times, 'stout' → 'stout', 'ale' → 'ale'). The latter cases would seem to call for more decided intervention on the translator's part, but, as it is, translations show (arguably) undigested chunks of source culture in their make-up. Moreover, there is one case of pure borrowing + amplification.

Naturalised borrowing
If pure borrowing spans a relatively wide range of factors, naturalised borrowing shows a perfect correlation with a single factoran imported CSI. What that means is that naturalised borrowing cannot occur unless the ST CSI has been imported into the target system. However, this works in just one direction, as it does not imply that whenever the 'imported CSI' factor is present, naturalised borrowing inevitably follows. In other words, there are 38 cases in the corpus where the ST CSI was available in the target culture in the form of a loan word, but the loan word was used in only 16 out of those 38 (e.g. 'pudding' → 'púding', 'curry' → 'curri'). In the remaining 22 cases, a technique other than naturalised borrowing was preferred, such as generalisation or intercultural adaptation. Furthermore, there is one case of naturalised borrowing + literal translation and one of naturalised borrowing + amplification.

Literal translation
The third (and last) foreignising technique is literal translationone of the major techniques, as it features 53 times in the data (21.12% of the cases). It correlates with as many as nine factors, but is not evenly distributed across them. In 23 cases it correlates with 'different degree of institutionalisation': e.g. 'fish and chips' → 'creïlles fregides i peix' ('fried potatoes and fish'). Furthermore, in 13 cases literal translation correlates with CSI that do not exist in the target culture. Within this group, a distinction should be made between those cases in which literalness does not stand in the way of communication, as the translation solution is self-explanatory (e.g. 'buckwheat cakes' → 'pastissets de fajol o blat negre' ['little cakes of buckwheat, or black wheat']) and those in which the TT segment is opaque and cannot be referred by an average reader to any known foodstuff in their cultural universe (e.g. 'shepherd's pie' → 'empanada de pastor', which is a wordfor-word rendering of the ST CSI). Literal translation correlates five times with false friendship and five times with a polysemous ST CSI. The remaining correlations between literal translation and other factors show a rather low profile (three cases or less).

Neutralisation: generalisation/particularisation
Let us now move on to neutralising techniques. The major neutralising techniques in my corpus are generalisation and description. Generalisation is used 50 times (19.92% of the cases), and it tends to correlate with three factors: different degree of granularity (21), nonexistence (11) and different degree of institutionalisation (nine). The most conspicuous examples of generalisation being used as a solution to different degrees of granularity across linguistic and cultural systems are beer, commented on above, and berries, which show a wider variety in the source culture: e.g. 'muskeg berries' → 'baies' ('berries'). Generalisation correlates with non-existence in such cases as 'corned beef' → 'carn curada' ('cured meat'). As to the correlation between generalisation and different degrees of institutionalisation, a good example could be 'maple syrup' → 'xarop' ('syrup'), this kind of syrup being well-known, for instance, in Canada and very marginal in the Catalan-speaking area. The intersection between generalisation and other factors provides rather few examples. As to the reverse of generalisation, i.e. particularisation, it is used only twice in the corpus. This is hardly surprising, as translation problems stemming from cultural specificity are seldom solved through an even higher dose of specificity.

Neutralisation: description
Description is the second major neutralising technique, as it features in 49 cases (19.52% of the total). Like generalisation, it tends to correlate with non-existence (18) and with a different degree of institutionalisation (17); but, unlike generalisation, it seldom co-occurs with a different degree of granularity (only two). At the intersection of description and non-existence we find, for instance, 'a scotch egg' → 'un ou bullit amb una llonganissa' ('a boiled egg with a sausage'). An illustrative example of description co-occurring with a different degree of institutionalisation would be 'Girl Guide cookies' → 'els pastissets aquells que fan les exploradores' ('those little cakes made by explorers'), where it is the association between cookies and the Girl Guides of Canada that achieves cultural status and must be negotiated somehow. The third factor correlating with description is figurative use of source CSI, as in 'bread and butter' when used to refer to common, everyday things or situations. 'Bread and butter' is translated as 'la quotidianitat' ('the everyday', 'the quality of something being everyday') and 'la vida corrent' ('normal life'), both solutions emphasising the tenor of the metaphor and leaving aside its vehicle, i.e. its food-relatedness. Other factors have little or no bearing upon the use of description as a technique.

Amplification
The last neutralising technique is amplification. Rather surprisingly, amplification is not used at all in the corpus by itself, and only twice in combination with other techniques (once with pure borrowing and once with naturalised borrowing), as seen above. The line separating description from amplification is admittedly thin; some of the cases here classified under 'Description' do entail some kind of amplification, when the information contained in the TT segment is compared to that in the ST segment; but since the cultural markedness of the CSI has disappeared in the translation (i.e. the TT segment could not be regarded as a CSI), they have been regarded as instances of neutralisation proper under the heading 'Description', rather than as instances of amplification.

Intracultural adaptation
The last group of techniques are domesticating or familiarising techniques. Intracultural adaptation is not very prominently used in my data -19 cases, 7.57% of the total.
Furthermore, 11 out of those 19 concern the same item, 'Ginger Nut' → 'Closca de Gingebre' ('Ginger Shell'), as applied to one of the characters in Melville's Bartleby. There is a strong correlation, then, between this technique and its use when the source CSI is a proper noun. Since the Bartleby character is named after a popular cake which is not even known in the target culture, the translator must have thought it suitable to steer clear of literal translation while retaining the reference to ginger, so as not to supply a target culture referent. Intracultural adaptation occasionally correlates with other factors, such as figurative use (three times) and source CSI being trademarks (twice).

Intercultural adaptation
The major domesticating technique is undoubtedly intercultural adaptation, which features 34 times in my data and accounts for 13.55% of the cases. It does not correlate strongly with any factor, spanning six of them. Even so, the strongest correlation occurs with non-existence, with 11 cases, as in 'griddle-cake syrup' → 'llet condensada' ('condensed milk'). More surprisingly, the second strongest correlation is with 'imported CSI'e.g. six times in 'doughnuts' → 'bunyols' (a Valencian kind of sweet, or small cake, not at all like a doughnut). Intercultural adaptation correlates five times with different degrees of granularity and five times with different levels of institutionalisation. An interesting example of the latter is 'cold pie' → 'embotit'embotit being a generic term to refer to many different kinds of cold meat, most of them shaped like big sausages rather than pies. Correlations of intercultural adaptation with other factors (false friendship, figurative use) are rather weak.

Omission
Finally, omission is used very sparingly in my data (five times, 1.99% of the total). Moreover, it spans several factors, from non-existence (quart as a unit of capacity) to imported CSI (grog is once omitted even though the very same item is available in the Catalan lexicon).

Strength of the correlations between techniques and factors
Correlations between techniques and factors are shown in Table 6. Those same correlations are displayed diagrammatically in Figure 3. If Figure 3 could be viewed in three dimensions, it would offer a landscape of peaks and valleys alternating along both the technique and the factor axes. That picture is somewhat blurred in the above two-dimensional representation. The fact that there are no very strong correlations between particular techniques and particular factors finds diagrammatic expression in the moderate height of peaks. Particular junctures are certainly more productive than others. As seen above, existence of imported CSI seems to be a pre-requisite for naturalised borrowing (16); literal translation is most likely to occur when the ST CSI are more highly institutionalised in the source than in the target culture (23) or when it does not exist in the latter (13); generalisation is most strongly associated with different degrees of granularity (21), non-existence coming way behind (11); description is most often used when the ST CSI do not exist in the target culture (18) or when they are more highly institutionalised in the source culture (17); intracultural adaptation correlates rather strongly with ST CSI being proper nouns (11, although this figure might be biased by frequency of occurrence of a single culture-specific item, 'Ginger Nut'); and intercultural adaptation is used relatively frequently (in 13 out of 34 cases) when the ST CSI do not exist in the target culture. These data account for all boxes in Table 6 with double figures. Peaks, however moderate, are indicative of some sort of causality (certain factors favour certain techniques); but it is weak causality, as most techniques span several factors and most factors help to account for several techniques (though not to the same degree). Table 6 and Figure 3 provide an account of the correlations between techniques and factors in terms of raw frequency. Therefore, a relatively high figure at a given juncture might be due either to the relative strength of the correlation or to the sheer frequency of the technique or the factor in question. In order to neutralise the latter variable, table figures could be expressed in percentages. But we need a double percentage, since every figure stands both in a column and a row. The value 5, for instance, at the intersection of 'Literal translation' and 'False friendship', stands in relation both to the total number of occurrences of the 'Literal translation' technique (53) and of the 'False friendship' factor (10), and would then be assigned a different percentage for each of these relations, or axes. A method has been devised, therefore, to measure the strength of a correlation in terms of the two percentages that can be assigned to every figure in the table. These two percentages are multiplied together and then divided by 100so as to keep figures within a manageable range. The logic of this is as follows. If the correlation between technique x and factor y were perfect, the percentages assigned to their intersection would be 100% for the technique row and 100% for the factor column. If these two values are multiplied together and then divided by 100, the result is 100. The highest possible value for a correlation, then, would be 100. That would entail that technique x is used whenever factor y is present, and that whenever technique x is used, factor y is present. The farther a given value strays from this ideal value, the lower it will be and the weaker the correlation. The results for the strength of correlations thus expressed, in relative terms, are displayed in Table 7 and Figure 4.
The landscape presented by Figure 4 is remarkably flatter than that in Figure 3. What that means is that the relative correlation values obtained by taking into account percentages rather than raw frequencies signal fewer peaks, which consequently stand out more clearly in a comparatively flatter landscape. The strongest correlation is that between intracultural adaptation and proper noun; but we must bear in mind that all tokens of this correlation belong to a single segment pair typea hardly representative instance in its being unique in the corpus. The second strongest correlation is that between naturalised borrowing and imported CSI: this is not surprising at all, because, as remarked above, availability of an imported CSI seems to be a logical requirement for its linguistic representation to be naturalised in the target culture. It is the highest-ranking correlations after the first two that are indeed worthy of attention, as they concern frequent techniques and factors. Generalisation and degree of granularity yields a score of 24.5; literal translation and degree of institutionalisation scores 17.82; and description and non-existent CSI (in the target culture) yields a value of 11.6. These values are, admittedly, still not  very high when seen against the backdrop of 100 as the highest correlation value; causality is still weak; but less weak than in most other correlations.

Conclusions
To sum up, two main conclusions will be highlighted, which address the two aims of the study as formulated in Section 3.
The results of the quantitative analysis of corpus data as regards the relative frequency of translation techniques show remarkable similarities to the results of previous studies, but also some differences. As explained in Section 4, foreignising techniques account for 35.46% of the cases under scrutiny, neutralising techniques for 40.24% and domesticating techniques for 23.11% (leaving aside combinations of techniques, which are very few). In Oster and Molés-Cases's study (2016) referred to above, neutralising techniques (generalisation and description taken together) account for 33 out of 56 cases, i.e. almost 60%. The other three techniques reported in their data (intra-and intercultural adaptation, as well as omission), which in my alignment are regarded as domesticating techniques, account for over 40%. No cases are reported of borrowing or literal translation (foreignising techniques). Oster and Molés-Cases's final balance, then, is different from the one presented here, as neutralising and domesticating techniques are better represented in their data, whereas foreignisation occurs more often in mine. That difference might be partly accounted for by the internal composition of both corpora, as the German-Catalan subcorpus of COVALT is predominantly made up of translations of children's and young adult literature, whereas in the English-Catalan subcorpus that component plays a less central role. The German-Catalan subcorpus is considerably smaller than the English-Catalan one (the ST component of the former comprises 282,739 words, whereas that of the latter is made up of 1,201,757 words). However, it is difficult to gauge the impact of corpus size on the results yielded by each subcorpus. At any rate, neutralising techniques are more often employed than the other groups of techniques in both studies. This seems to tally with the results of most studies reported on in Section 2.3, even if they did not always aim at quantification (cf. Frank, 2009, p. 10;Inggs, 2003, p. 288;Mussche and Willems, 2010, p. 491). Significantly, these four studies dealt with literary texts, which to a certain extent warrants comparability. Other studies may yield different results. Valdeón, for instance, pointed to 'a tendency towards domestication of the ST' (2008, p. 229) in the Spanish translation of the American sitcom Will & Grace; but comparison between the translation of literary and audiovisual products would perhaps be out of focus, as norms governing different activities may differ widely. As far as literary translation is concerned, translators often neutralise and very often avoid domestication.
My second conclusion concerns the correlation between techniques and factors. The main results as regards this correlation have been summarised at the end of Section 4 and need not be repeated here. Correlations seem to point to weak causality, as most techniques span several factors and most factors span several techniques. However, the peaks in Figures 3 and 4 (especially the latter) might be worth exploring, as the correlations signalled by them are not so weak. It might be worth finding out whether these correlations tend to occur in other sets of data. Let us not forget that the whole point of empirical