Towards an Empirical Verification of the Gravitational Pull Hypothesis

Ulrike Oster

Introduction

One of the focal points of Translation Studies over the last 30 years has been the identification and empirical testing of a number of properties that are thought to characterise translated language. Baker’s work (e.g., 1993) is often cited as the initial driving force behind this trend, but the idea that translated language might be a kind of third code (Frawley 1984), governed by laws of possibly universal character (Toury 1977, 1991, 1995), originated much earlier. Baker took stock of previous research and proposed a number of translation universals including explicitation, simplification, standardisation and convergence. These presumed properties of translated texts then became hypotheses to be tested empirically, and were (and to some extent still are) at the heart of corpus-based translation studies.

However, the idea of translation universals has been subject to criticism almost since its inception. Some authors have rejected the concept of universals altogether (Tymoczko 1998), while others have contributed to a clearer understanding. The latter case is exemplified by Malmkjær (2008), who does not deny the possibility of translation universals but formulates conditions for their existence. Malmkjær highlights a certain tension between the concept of norm, which is socio-cultural and therefore variable in nature, and that of universal, which would not allow for any kind of variation. For this author, any feature to be considered universal must have a cognitive basis; on the other hand, a trait that cannot be explained on cognitive grounds can only attain the status of norm because norms are the result of specific configurations of socio-cultural factors. This would be the case of the presumed universals identified by Baker: all of them would constitute translation norms (Malmkjær 2008, 55) subject to spatio-temporal conditioning factors. On the other hand, Malmkjær does assign universal status to a hypothesis formulated by Tirkkonen-Condit (2002, 2004), the so-called Unique Items Hypothesis (UIH). Malmkjær (2008, 57) concludes that ‘if the concept of the universal is to retain any theoretical bite in our discipline, we would do well to reserve it for use in connection with phenomena such as this [the UIH], for which it makes sense to produce a cognitively based explanation’.

The UIH postulates that typical elements of the target language (TL) tend to be underrepresented in translated language when their use is not triggered by the existence of a formal equivalent in the source language (SL). Tirkkonen-Condit refers to these elements without a formal equivalent in the SL as unique items. This hypothesis, however, contradicts one of the hypotheses proposed by Baker in her seminal 1993 paper: the overrepresentation of typical TL elements in translated language. This feature would be consistent with what Toury (1995) called the law of growing standardisation, which has subsequently been called normalisation (e.g., Kenny 2001). The underlying idea is that translators tend to favour those elements and structures of a language that are well established by norm and usage, thus avoiding the risks associated with more creative, unusual formulations.

So how can both statements be true? On the face of it, the validity of one would negate that of the other, but both have been defended and both have found some empirical support. The overrepresentation hypothesis has been confirmed in studies such as Kenny (2001) (albeit partially), Bernardini (2007) and Corpas (2008); the UIH, for example, in Eskola (2004) and Martínez Vilinsky (2012). Thus, assuming that they are not mutually exclusive and that they both have explanatory validity, a theory of translation should postulate under which conditions overrepresentation will occur and under which other conditions underrepresentation will prevail.

This is precisely what Sandra Halverson’s Gravitational Pull Hypothesis (GPH) aims to do. The GPH is based on concepts and categories specific to cognitive linguistics and bilingualism theory and represents an attempt to reconcile the seemingly contradictory claims of over- and underrepresentation of typical TL elements. It has been formulated and reviewed by Halverson in several papers (e.g., 2003, 2010, 2017). However, despite its enormous descriptive and, above all, explanatory potential, the GPH remains under-researched. Halverson (2017) herself carries out an empirical study with the semantic network formed by the different senses of the English verb get and its two most common Norwegian equivalents. Hareide (2014, 2017a, 2017b) uses one of the methods suggested by Halverson to test the GPH: working with two language pairs with an identical target language and focussing on an item that can be considered unique in one of the two pairs but not in the other. Data-Bukowska (2021) applies the GPH to literal translation of sentence-initial noun phrases in a Norwegian-German and German-Norwegian context. Lefer and De Sutter (2022) use a multifactorial research design to study concatenated nouns in two modes of interlingual mediation (EN-FR simultaneous interpreting and written translation). The reader is referred to Halverson’s chapter in this volume for a more thorough overview of GPH-based research.

COVALT’S research background in Corpus-based Translation Studies (CTS), combined with a strong interest in the cognitive aspects of translation, led our research group to the GPH as a theoretical and descriptive framework for our work. Since 2001, COVALT has been working on the compilation and analysis of the eponymous corpus (Valencian Corpus of Translated Literature), a set of parallel and comparable corpora, all of which consist exclusively of narrative fiction. The parallel corpora have German, French and English as source languages and Catalan and Spanish as target languages. The comparable corpora consist of the translated texts (into Catalan or Spanish) and a component of narrative texts originally written in Catalan or Spanish. The study of the COVALT corpus had so far focused mainly on the analysis of two aspects: (a) stylistic or textual features that often pose translation problems; and (b) some of the presumed properties of translated language (the aforementioned translation universals or norms).

The group therefore set itself an ambitious goal for its latest research project:¹ To use the GPH to advance knowledge about the cognitive basis of the properties of translated language. The project pursues two objectives: to test the GPH empirically and to refine it. To this end – that is, to gather evidence that confirms or refutes the hypothesis – a set of lexical and grammatical items has been selected from the source and target languages, whose properties are likely to cause translation effects. The analysis is based on a mixed-methods approach that combines different types of data: textual data from COVALT (in its different language combinations) and experimental data from elicitation and translation tasks.

This book is a compilation of the results of this research project and is structured as follows. In the first chapter, ‘The gravitational pull hypothesis: Status and prospects’, Sandra Halverson sets the scene for the book by tracing the origins and development of the GPH, describing how it fits into current translation and interpreting research, and anticipating where it might or should lead. The chapter begins with a clarification of important theoretical issues underlying the GPH: a terminological disentanglement of ‘salience’ and ‘entrenchment’, an answer to the question of why entrenchment effects should be particularly visible in translation, and a reflection on the convenience of analysing gravitational effects in both semasiological and onomasiological categories as well as on the influence of frequency and different usages on SL and TL configurations and connectivity. This is followed by an overview of empirical studies relevant and related to the GPH, namely machine learning studies on so-called ‘translationese’ and corpus-based translation studies, leading to a summary of the current status of the hypothesis. Finally, the chapter situates the sociocognitive approach of the GPH within a broader, usage-based theory of translation and identifies questions that remain and need to be addressed in future work.