A Look at What is Lost: Combining Bibliographic and Corpus Data to Study Clichés of Translation

1 Metaphors, Clichés, and Perceptions of Translation

Abstract features of human experience are commonly understood with reference to more concrete, physical objects and relationships. We tend to think of time, for instance, as a valuable resource, something that you can lose, or waste, or run out of. Conceptual metaphor theory systematically studies how humans map and employ such correspondences linguistically as well as cognitively (Kövecses 2010; Lakoff and Johnson 1980). There are usually multiple pathways available for discussing the same abstract sensation or activity: thinking, for instance, can be conceived of as a process of digestion or of construction, meaning that a palatable argument tends to be solid as well. Sometimes, potential mappings conflict. Expressing the expectation that something will go downhill, for instance, can be interpreted in positive fashion: given the laws of gravity, downward movement tends to be easier than its opposite. Nonetheless, what is low tends to be valued less than what is high, as commonly symbolised in shared expectations about heaven and hell, or in the idea that a story can either be uplifting or depressing. Despite the flexibility and dynamism conceptual metaphors afford, some mappings may achieve a fair level of dominance or stability within a particular field of experience. The idea of communication, for instance, is often framed in terms of what is commonly called the conduit metaphor, according to which words act as containers for mental contents (Guldin 2010; Reddy 1979). If a writer puts their ideas into words successfully, a reader should have no trouble unpacking the message.

The conduit metaphor operates on the basis of a clear distinction between meaning and form, and represents communication as a transparent act of transportation. This framing also influences the conceptualization of more specific communicative activities such as translation. A common Renaissance figure of speech, for instance, conceives of translation as akin to “pouring something from one vessel into another” (Hermans 1985, 121). A good translation, from this point of view, is one that meticulously preserves the qualities of its source material, regardless of a change in context. Not only should as much as possible of the transmitted substance be preserved, it should also be shielded from contamination or corruption. Indeed, up until today some language service providers expect their translators not to “inject any creativity” when handling a client’s content (e.g., Stephanou 2020). The translator’s task, in this respect, is restricted to running errands, careful not to leave fingerprints on the customer’s valued package. The implications of this view have been elaborately explored by Venuti (1995, 2019 in terms of instrumentalism and invisibility: translation, it seems, is expected to be mere reproduction, and any human agency involved in the process is supposed to escape detection. Those expectations are not without consequence, as they affect the status of the professional translator, whose role has for centuries been associated with characteristics such as dependence and servility (Simeoni 1998).

Admittedly, in certain academic circles translation is today celebrated as a highly creative and transformative activity, to the extent that one can speak of “translatophilia,” or the “fetishisation of translation in hypercorrection of its perceived marginalisation” (Lee 2022, 543). The cultural value of translation is also foregrounded by major institutional bodies such as the European Union (EU). A recent report produced as part of the EU Work Plan for Culture is entitled Translators on the Cover (European Commission 2022), echoing similar calls for visibility by loose-knit social media campaigns and more formal statements by organisations such as the British Society of Authors (2021). Nevertheless, the heightened appreciation of translation work is generally restricted to the literary sector, which occupies only a very small segment of the industry. Most translators continue to toil in relative anonymity, subject to severe economic pressures and embedded in digital labour platforms that limit personal agency in business relations and which render visibility dependent on competitive rating systems (Fırat 2021). Thus, while the demand for translation has grown considerably in recent years, this has not resulted in better working conditions, increased remuneration, or a revaluation of translators’ professional status (Moorkens 2017).

The limited value accorded to translation work today can partly be ascribed to developments in the field of machine translation (MT), which serves as a constant reminder of potential automation (Vieira 2020a). Recent innovations in generative artificial intelligence serve as an additional source of professional unease. Successes in translation automation are often accompanied by assurances from specialists that human translators will not become superfluous (Way 2020, 326). Nevertheless, the nature of much translation work is fundamentally changing, and post-editing machine-generated content, for instance, has already become a widespread practice. Post-editing guidelines often differentiate between full and light edits, with the latter only aiming for comprehensible output, which does not have to be perfect in terms of style or grammar (Hu and Cadwell 2016). The distinction signals the continued hold of the conduit metaphor on perceptions about what translation activity entails: MT output is seen as containing the essence of a linguistic message, which the human translator is then invited to repackage to a variable degree.

Venuti (2019, 84) argues that such restrictive, “long-entrenched assumptions about translation” are encoded in clichéd proverbs such as traduttore traditore and poetry is what gets lost in translation. Proverbs are generally characterised by a degree of lexical invariance as well as broad pragmatic applicability: you can encourage someone to strike while the iron is hot in settings related to business, politics, sports, romance, or even blacksmithing. Venuti acknowledges that many formulaic expressions convey a multiplicity of meanings in accordance with their varied context of use, but argues that common phrases involving translation are remarkably inflexible. Purportedly they only function to convey instances of failure in translation practice (Venuti 2019, 88). Little evidence is provided by Venuti for this claim, however. He complements his assertion with an insightful genealogy that historically and rhetorically contextualises common formulaic expressions about translation, but there is no reason to assume that his scholarly examples are representative of the range of discursive functions such expressions have potentially come to fulfil.

Therefore, in this article we seek to illustrate the usefulness of a corpus-assisted approach to the study of clichés related to translation. Corpora are large, electronically stored collections of language data, often accessed via a specialised computer interface. Gu (2023) and Gómez Castro (2021) provide useful overviews of the varied applications of corpus tools in the discipline of translation studies. In general, corpora facilitate the identification and study of particular language features on a large scale, and tend to challenge people’s intuitions about the repertoire of linguistic patterns we rely on to communicate. In this article, we acknowledge that there is likely to be a connection between clichés invoking translation, common conceptualizations of translation practice, and the status of professional translators. We argue, however, that any assessment of such clichés is incomplete, and potentially misguided, if their attested usage in natural language data is not taken into account. In our study, we focus on one particularly common phrase, namely lost in translation, and investigate its usage in a number of discursive environments, as detailed in the following section.

2 Data Sources and Approaches to Linguistic Patterning

Chesterman (1997) argues that theoretical thinking about the nature of translation tends to revolve around a set of recurrent concepts and oppositions, such as the idea of equivalence or the distinction between source and target. As Zanettin et al. (2015) illustrate, bibliographic studies can shed light on the discursive organisation of the discipline of translation studies, and thus on its internal conceptualisation of translation. Discussions of translation, however, are not restricted to the domain of academic writing, and a more comprehensive account of how translation is commonly conceptualised, by means of proverbs or otherwise, should thus take alternative sources into account. In a study mostly concerned with the public image of simultaneous interpreting, Diriker (2009) lists various sources that, in addition to scholarly works, can be consulted to obtain a clearer picture of how translation activities are discursively represented, such as dictionaries, statements by professional organisations, and features in televised or print media. Ben-Ari (2010) draws attention to fictionalised portrayals of translators in film, theatre, and literature. Kotze et al. (2021, 150) capture readerly “perceptions of and responses to translation” by analysing a corpus of Goodreads reviews. Vieira (2020b) samples a database of newspaper articles in order to study the discursive framing of developments in machine translation. More broadly, newspapers can serve as a good starting point to study “how translation is perceived as a general phenomenon” (Fawcett 2000, 295).

Newspapers are currently under considerable pressure, partly because of competition in the information marketsphere with online social networks. For many news outlets, digital transformation has thus become an essential prerequisite for survival. While degrees of success, as well as modes of access and engagement, vary widely for individual outlets, visiting news websites is a very common activity globally, meaning that the overall reach of web-based news is considerable (Newman et al. 2022). Therefore, the first data source queried in this study is the News on the Web Corpus (NOW), a continually growing collection of articles from anglophone web-based newspapers and magazines, available through a dedicated interface as part of the English Corpora collection and containing material from 2010 onwards. At the time of writing (April 2023), the corpus comprises close to 18 billion words. Due to its large size, and its inclusion of sources from numerous countries (such as the United Kingdom (UK), the United States (US), Bangladesh and Jamaica) the corpus serves not only as a database of news discourse, but also as a “robust monitor corpus” for tracking trends in the English language as a whole (Davies 2016).

A preliminary exploration of the corpus revealed that the word translation occurs close to 150,000 times. The interface’s collocation tool indicates that translation is mentioned remarkably often in the vicinity of the following lexical items: English, distribution, commercial, sale, widespread, any, content, reproduction, unauthorised, and lost. This list of the 10 most frequent collocates largely points towards a peculiarity of the data gathering process: many websites contain a blanket copyright statement that advises against unauthorised reproduction and dissemination practices, including translation. While the explicit connection between intellectual property and translation in online publishing environments is interesting in itself, the collocates are not very revealing about how translation is discussed in particular news items. However, one of the collocates in the list, namely lost, is indicative of common proverbial usage, and is therefore of immediate interest to the present study. Across the corpus as a whole, the phrase lost in translation occurs more than 6000 times, and for each year since 2010 attests to several hundred instances. In order to gain a snapshot of recent usage, we focus on the year 2021, the last year that was completely covered in the corpus when our data analysis started.

The corpus currently lists 708 occurrences of lost in translation for the year 2021, and this number is fairly stable, although the NOW interface signals that duplicate removal is still ongoing. We extracted all instances from the online corpus environment in order to manually check the context surrounding the phrase of interest. During this process, we took note of and removed duplicates, articles published before 2021, articles that were no longer available, data that could not be categorised as news (e.g. forum comments), and instances for which potentially relevant metadata was lacking (e.g. place of publication). When an article contained more than one instance of the phrase, but without a meaningful difference in usage (e.g. multiple mentions of the movie Lost in Translation in a single news item about the Oscars), we also removed redundant occurrences. After this procedure, 423 instances remained, and they constitute the data from NOW analysed in the following section. To obtain information about general usage patterns in public discourse, we considered it important to begin the analysis in a realm unconstrained by academic sensibilities, such as that of online news publications. We do, however, acknowledge that academic writing partakes and often aims to consciously intervene in the conceptualization of translation, and for this reason we complemented our findings from NOW with occurrences of the phrase lost in translation found in two scholarly databases, namely BITRA and SCOPUS, in search of significant similarities or differences in usage.

BITRA, the Bibliography of Interpreting and Translation, is an open-access online database, currently providing information such as abstracts, titles, and publication dates for more than 92,000 publications on translation-related subjects, including books, journal articles and PhD theses (Franco Aixelá 2022). Although its scope is global and the database is regularly updated, its main curator acknowledges that the entries sketch a fairly Eurocentric picture of the disciplinary landscape (Franco Aixelá 2013). A recent report (Franco Aixelá 2021) indicates that more than half of documented entries (52.2 %) describe research communicated in the English language. The Spanish (18.4 %) and French (12.0 %) languages are reasonably well-represented. German, Portuguese, Italian, Chinese and Catalan each account for between 1 and 8 % of research included in BITRA. For studies produced in other languages, an English title tends to be provided, meaning that our study, while focused on an English expression, is in principle not restricted to anglophone content, as will be illustrated in the analysis below. We queried BITRA by searching for the phrase lost in translation in the Title field, as we are primarily interested in highly visible scholarly uses of what the newspaper corpus indicates to be a commonly employed catchphrase. In response to the query, BITRA returned 134 entries across a variety of document types, with journal articles accounting for 63 % of occurrences. As journal articles constitute only 46.8 % of overall items in the database (Franco Aixelá 2021), one could infer that this document type seems especially susceptible to the use of stock phrases in publication titles. BITRA assigns subjects to its entries, and a primary exploration of the data indicated that multiple entries employing variations on the phrase lost in translation relate to specific knowledge areas such as the legal or the medical domain. We therefore supplemented our analysis of BITRA with data from SCOPUS, an abstract and citation database covering a large number of subject areas including the physical as well as social sciences. Entries from SCOPUS were selected by searching for occurrences of lost in translation in the article title, abstract, or keywords of documented publications.

In sum, the analysis presented in the next section considers material drawn from three distinct but complementary sources. It is guided by a fundamental principle of corpus research, namely “the search for – and belief in the importance of – recurring patterns” (Partington, Duguid, and Taylor 2013, 8). The core pattern, namely the common phrase lost in translation, was selected on the basis of a preliminary exploration of the corpus data, in accordance with the study’s scope of interest. Using a corpus to study common phrases can indicate the range of communicative functions they fulfil, and can also reveal how people conceptualise and evaluate particular relations, situations and activities (Stubbs 2009, 2010). Thus, we query the newspaper corpus as well as the bibliographical databases in order to determine, in response to Venuti’s argument, whether it is indeed the case that formulaic expressions related to translation overwhelmingly focus on failures of translation practice. In order to categorise different uses of the phrase, awareness of co-text and context is essential, meaning that our approach required us to read longer stretches of text and take metadata into consideration for each documented instance.

Even though, in the analysis presented below, comparative frequencies and percentages are provided where informative, and collocational observations are dependent on statistical measures, our study mainly offers qualitative information. Over the past decade, corpus linguistics has witnessed a tendency towards elaborate statistical reporting, while the degree of linguistic description, for instance in the form of text excerpts, has noticeably diminished (Larsson, Egbert, and Biber 2022). A similar trend characterises recent developments in corpus-based translation studies (Buts and Jones 2021). In this study, however, we will illustrate our claims by means of extended examples, as discussing nuanced differences in usage patterns constitutes a simultaneous act of categorization and interpretation, and is therefore in need of linguistic illustration. Focusing on text rather than numbers also allows us to compare observations across databases directly, without mathematical mediation: BITRA and NOW are resources of a fundamentally different scope, structure, and composition, but they do contain examples of similar phrasing, which is the phenomenon we aim to assess.

3 The Circulation and Functions of the Lost Trope in Online News and Academic Discourse

4 Concluding Remarks

In the first section of this article, we discussed the perceived connection between formulaic expressions about translation and the public image of translation practice. We then set out to test Venuti’s (2019, 88) hypothesis that proverbs and clichés involving translation are not only formally inflexible, but also restricted in their capacity to express anything beyond the practical impossibility of translation. As the analysis of lost in translation presented in the previous section illustrates, Venuti’s argument does not reflect usage patterns in web-based newspapers, publications in translation studies, or scholarly publications in the medical field. The general trends observed indicate that it is unlikely the argument consistently holds for any discursive domain.

In support of Venuti’s position, however, one could argue that, once the expression lost in translation detaches itself from linguistic concerns, as partially signalled by the scarcity of the classic reference to poetry in our dataset, it is no longer relevant to discussions of intercultural communication. The data from SCOPUS on cellular translation processes as well as medical knowledge translation, for instance, can arguably be easily distinguished from the core concerns of professional translators and scholars of translation studies. However, the vocabulary of translation used in the medical sciences often shows remarkable parallels with that employed in the study of cultural mediation: the translation of mRNA into polypeptide is described as “faithful” when felicitous, is based on encoding and decoding procedures, and may involve errors and “mistranslations” (Mohler and Ibba 2017). Furthermore, scholars concerned with the conceptualization of medical knowledge translation, which mediates between basic research insights and clinical treatment, have explicitly drawn on insights from theoretical work in translation studies (Engebretsen, Sandset, and Ødemark 2017). Consequently, a strict conceptual separation between interlinguistic, intercultural, and other concepts of translation is artificial, and reflects neither the linguistic nor the scholarly status quaestionis.

However, acknowledging the variety of related functions that the concept of translation fulfils across discursive domains naturally introduces a problem of scope. In this paper, only one cliché of translation was elaborately addressed, though a very common one, and only data from web-based newspapers and scholarly databases was taken into account. Other data would reveal additional usage patterns that further complement translation’s proverbial complexity. Kotze et al.’s (2021, 167) study of Goodreads reviews, for instance, also identifies lost in translation as a commonly employed trope, but given the literary nature of the platform, it is logical that many of the instances found there adhere more closely to the perspective that Venuti identifies and critiques. Thus, given the sheer range of applicability characteristic of expressions relating to translation, it is obvious that large datasets such as bibliographic and corpus resources provide an important – yet never completely comprehensive – vantage point for research into public perceptions of translation activity, as well as for the further development of theoretical and conceptual discussions. Furthermore, as argued in this article, combining particular datasets can function to reveal relevant interdisciplinary connections, as well as interactions between scholarly discourse, popular culture and news media.

Our analysis indicates that, across the resources examined, the main referent of lost in translation can be a particular communicative feature or object characteristic, but can just as well be an agent subjected to a state of profound confusion. This usage challenges the primacy of the conduit metaphor in our thinking about translation. The image of a package to be carefully carried across a linguistic boundary is complemented with, and to some extent supplanted by, an image of a perplexed traveller attempting to navigate an unfamiliar environment. In short, the underlying conceptual metaphor is one of orientation rather than transportation. These domains are not mutually exclusive, and they are both derivative of the fundamental human perception that bodies, be it of flesh or of knowledge, enact movements within particular spatial configurations. Thus, the metaphorical reference to orientation is neither surprising nor novel – translators have been told for decades that they should be able to navigate the cultural landscape in which the material they are working with is embedded, but the fact that a commonly used stock phrase can partly shift its metaphorical affiliation is nevertheless significant. Furthermore, it is notable that recent proposals for more accurate and sophisticated metaphors to capture the nature of translation activity include a conceptualization of the translator as a maze-walker (Mossop 2021). The introduction of novel, creative metaphors thus operates in line with subtle changes in conceptual metaphors, which are not consciously crafted. It seems that in translation today, like in a maze, you are more likely to get lost than to lose something.

The notion of translation as an encompassing state or site of potential confusion, which humans continually attempt to navigate, also ties in with theoretical approaches that privilege process over product, and trajectory over target. Marais (2019, 6), for instance, argues that translation processes take “a plethora of forms, depending on medium, technology, culture, time, and space, among others”. Similarly recognising the formal flexibility of the idea of translation as such, Cronin (2017, 171) argues for the study of the entire “tradosphere,” namely “the sum of sum of all translation systems on the planet, all the ways in which information circulates between living and non-living organisms.” On the one hand, such theoretical expansions of translation studies’ disciplinary purview can be characterised as overly ambitious, vague, and challenging to render operational. On the other hand, the expansions seem to tie in well with the broad understanding of translation that is already pervasive in a variety of discursive environments, including popular culture and news media. Questions of mapping and orientation will thus continue to be of the utmost relevance not only to the concept of translation itself, but also to how we theoretically approach it. A central concern, in this regard, is the fact that the relationship between the visibility of translation and the visibility of translators seems to be one of inverse correlation: as the concept of translation expands to new domains and sites of application, it typically becomes harder to situate human agency. If only there were a cliché to capture this process concisely.