The Impact of the Source Language in Spanish Translations: A Survey on English Counterfactuals ‘Should have’

  • Isabel Repiso

…plus d’informations

Couverture de Volume 63, numéro 1, avril 2018, p. 1-276, Meta

Corps de l’article

1. Introduction

Just as we can talk about situations which we believe to be the case, we can talk about situations that might have been performed but were not. Typically, this is done by counterfactual sentences, such as The police should have intervened. The present study analyzes the Spanish translations of the construction Should have + Past participle in counterfactual contexts. Previous studies have shown crosslinguistic differences in the way languages encode counterfactuality (Van linden & Verstraete 2008). These authors have shown that there is no one single dedicated marker to encode counterfactuality across languages but rather a combination of elements that have other functions in other contexts (i.e., past-tense markers, aspectual markers, modal markers). In Spanish, the use of the perfect subjunctive is significantly higher compared to the combination of the past conditional and a modal marker (Repiso 2014). If the grammatical categories to express counterfactuality change from one language to another, the question arises whether translators replicate the source construction or use prominent constructions of the target language. Does English, as a source language, have an impact on the Spanish translations either structurally or semantically? The present survey describes the translators’ output constructions and sets them against the usage-based constructions of native Spanish.

Our objective is twofold. First, we aim to describe the linguistic features of the different constructions provided by translators in Spanish. Special attention will be paid to the description of (i) the tenses and moods more frequently used by translators, (ii) the word order of the output constructions in Spanish, and (iii) the Spanish modal markers related to Should. Second, we aim to test whether the constructions provided by translations are coherent with the language use of native Spaniards. This will be done by searching equivalent items in the Spanish reference corpus CREA.[1] The results of our survey should contribute to elucidating the impact of the source construction Should have + Past participle in Spanish counterfactual readings, from a structural and a semantic viewpoint. Since word-by-word translation would give a non-deviant form in Spanish from a grammatical perspective (i.e., Debería haber + Past participle), we hypothesize that this will be a prominent solution for translators. Because compounded by three elements – a modal marker, an auxiliary verb and a lexical verb-, Should have + Past participle is likely to be translated by what Newmark (1986: 141) labels ‘micro’-procedures (transcriptions, literal translations) subject to syntactical constraints and lexical systems. If our hypothesis proves true, then one might wonder whether modal markers are equally used by English and Spanish native speakers. Significant differences have been demonstrated between Spanish and French on evaluative markers suggesting that native speakers of French tend to ‘modalize’ their discourse, whereas Spanish speakers rarely use modal verbs in counterfactual contexts (Repiso 2013). This typological difference has semantic implications and thus, denoting evaluative judgments by means of modal verbs seems to be a specificity of French that Spanish somehow lacks. The analysis of Should have + Past participle addresses this question with special emphasis.

Our article is organized as follows. In the first section, we highlight some typological differences between languages that have been at the origins of important translation studies. Subsequently we introduce the semantic domain of counterfactuality to which the construction Should have + Past participle is frequently associated. In the third section we present the corpus of written texts and translations that we have reunited, as well as the methods used (i) to elicit the Should have occurrences and (ii) to run statistical analyses. Results are presented from section 4.1 to 4.3 (English corpus, Spanish translations and reference corpus CREA, respectively). In the fourth section we discuss our findings and relate them to previous studies. Conclusions are drawn in the sixth section.

2. Contrastive linguistics and translation studies

Crosslinguistic studies have shown that different languages might have different salient categories to express time, space, movement or irreality (Carroll et al. 2008; Van linden and Verstraete 2008). In general, these differences lead L2 learners to explicitly replicate the salient categories of their L1 in their oral productions (Slobin 1996; Carroll et al. 2012; Hung 2012; Repiso 2014). As avowed connoisseurs of metalinguistic issues, translators avoid overt transfers. Previous surveys at the crossroad of crosslinguistic and translation studies have given evidence on the attention paid by translators to the salient categories of the target language (Fabricius-Hansen 1999; Doherty 2004[2]; Slobin 2005). Slobin (2005) has shown that the French, Spanish and Portuguese translations of The Hobbit (Tolkien 1937) are less specific regarding the expressions of path and manner compared to the German and Dutch translations. Slobin (2005) attributes this result to the grammatical devices available in Germanic languages which typically encode motion by satellite verbs (go in/out/down/up), whereas Romance languages generally encode motion by lexical verbs (entrar, salir, subir, bajar).[3] Satellite verbs allow German languages to be more specific in the description of motion events concerning manner, path and trajectory compared to Romance languages. In the aim of being coherent with their native language, the Romance translators of The Hobbit frequently omit some spatial details present in the source language.

Fabricius-Hansen (1999) has approached typological differences from a macro-structural perspective comparing German and Norwegian. Since German texts present a relatively higher degree of subordination and Norwegian texts are characterized by a more paratactic style, her survey focuses on how Norwegian translators deal with the problem of informational density. The way in which it is solved is by ‘sententialization’ (one source sentence is translated by a sequence of independent sentences) or by ‘clausalization’ (rendering non-clausal source constituents as subordinate clauses without changing sentence boundaries).

Doherty (2004) explores the constraints of English and German regarding the word order of attributes and appositions at a phrasal level. These typological differences frequently lead German translators to reorganize dependencies in their output texts. Doherty (2004) claims that not to do so could be seen as a violation of the target language expectations or as presenting information in a discourse-inappropriate way. The above studies suggest that being aware of the typological differences between languages is a critical ingredient of translation’s good practices.

A specific grammatical property of English is its high level of composition (i.e., when two or more radical elements form a single word). This linguistic feature was described by Sapir (1921), along with affixation and word-order, as a linguistic process making it possible to pair languages within a formal taxonomy. The word-formation variable has been analyzed by Lefer and Cartoni (2013) in order to test whether the English – less and – un negative affixes have an effect in translated texts. Much of the contribution of linguistics to descriptive translation studies lies in providing reliable variables – e.g., language pair, formal properties – that may work as explanatory factors.

The present study focuses on the translations of Should have + Past participle into Spanish. Our objective is to test whether the word by word equivalent Debería haber + Past participle is privileged by translators or, conversely, whether they use other constructions not so formally similar to the source one but, nonetheless, used in Spanish. This question actually relates to the notion of covert transfer or ‘positive transfer’ accordingly to Toury (1995).[4] We will henceforth talk about covert transfer as a subtle influence which leads to a form that is possible in the source language but does not correspond to what the native speakers normally do in such a context. Covert transfers are hardly recognizable to the reader because the output construction is still grammatical. Nonetheless, covert transfers mean a shift at the level of frequency ratios between more or less frequent constructions pertaining to a specific semantic domain.

3. The notion of counterfactuality

The present study investigates the translation of the English construction Should have + Past participle, which is generally associated with counterfactual meanings. Pietrandrea (2010) defines counterfactuality as a subdomain of a larger semantic category that she calls irreality. From a semantic viewpoint, counterfactuality combines the apparently contradictory features of potentiality and non-actualization (Versatrete 2005). Counterfactual sentences express something that was desired but did not happen, or an action that was intended but not carried out in the end, as in example (1).

The speaker’s state-of-knowledge at the moment of the topic time plays a crucial role in processing a meaning as potential or counterfactual. Potentiality implies the speaker’s uncertainty about the actualization of the propositional content, whereas counterfactuality implies the speaker’s certainty about the non-actualization of the propositional content. One is more likely to be certain about present or past states-of-affairs rather than future states-of-affairs. Which explains why some authors insist on placing counterfactuality in the non-future time (Iatridou 2000; Vetters 1994). The temporal span to which counterfactual sentences refer is actually a future within the past (Grevisse and Goosse 2008).[5] Thus, in example (1) the construction Should have + Past participle expresses a temporal value of non-factual posteriority rooted in the past.

Figure 1

Graphical representation of other possible worlds

Graphical representation of other possible worlds

-> Voir la liste des figures

Past tense morphology has been mentioned as a component of counterfactual morphology (Iatridou 2000), along with modal markers (Van linden and Verstraete 2008). Typological differences between Spanish and French have been observed in the oral production of counterfactual scenarios (Repiso 2014). In a mutation task consisting of providing alternative scenarios to factual states-of-affairs native Spanish speakers significantly used the perfect subjunctive with no modal markers (e.g., la policía hubiera intervenido), whereas native French speakers privileged the combination of the past conditional and a modal marker (e.g., la police aurait dû intervenir). The frequencies of these grammatical devices proved to be a major difference between French and Spaniards.

In Spanish, counterfactual readings can be conveyed by different tenses and moods. According to grammarians, the most recurrent tenses are the pluperfect subjunctive (2 and 3), the past infinitive within iterative sentences (4), and the past conditional within if- clauses’ main sentences (5). From a semantic viewpoint, example (2) usually denotes the speaker’s dismissing the topic or theme conversation. The pluperfect subjunctive is used in a concessive manner to reject an eventuality that the speaker considers false (i.e., In that case they had warned me). Within negative sentences (3), the pluperfect subjunctive denotes a polarity and, at the same time, asserts the lexical content of the verbal segment (i.e., You was actually late).

Beyond flexional features like tenses, the notion of probability can also be encoded in Spanish by lexical means, like the periphrastic segments Tener que + Infinitive and Deber + Infinitive. The former is more frequent than the latter and conveys a modal message that is presented by the speaker as something inevitable (RAE 2009: 2144). Unlikely, the periphrastic Deber + Infinitive conveys a message socially desirable or accordingly to the norms of a given community. Both of these formulae are used to denote epistemic values in a wide range of indicative tenses such as the present indicative (6), the Spanish imperfecto (7), the past simple (8) and the present conditional (9). Thus, the periphrases Tener que or Deber + Infinitive can be seen as different points within a scale of intensity or evaluative language in Spanish. This raises the question whether the construction “Should have” is more frequently used in English with a binding connotation or a desirable connotation. It may well be that in some contexts the semantic equivalence of Should have is the Spanish Tener que, whereas in other contexts Deber que is more accurate. Anyway, the translator seems to have some degree of choice here.

Other periphrases expressing some degree of probability are Haber que + Infinitive (10) and Poder + Infinitive (11). We will discuss some of the examples above in section 3.2 during the presentation of the Spanish translations.

4. Corpus and methods

Our study is based on 1.7 million-word Social Sciences corpus covering 8 essays, 4 political biographies and 2 dystopian novels. Our purpose was to collect the same sample of texts both in English and Spanish, so that we could work on a parallel corpus. We reunited texts from political activists Naomi Klein and Noam Chomsky (2 each), political leaders Nelson Mandela, Barack Obama and Malala Yousafzai, psychologist Daniel Kahneman and feminists Audre Lorde, Silvia Federici and Judith Butler. We also included Virginia Woolf’s essay A room of one’s own, Ray Bradbury’s Fahrenheit 451 and George Orwell’s 1984. The nature of these 14 texts is diverse. Among the essays, Kahneman’s is devoted to the cognitive mechanisms guiding decision-making processes. The rest of them are critical with hegemonic representations such as capitalism – Klein’s and Chomsky’s writings – or patriarchal institutions – Federici’s, Butler’s and Woolf’s texts – . We targeted these authors because some of their books made part of our personal library in Spanish. The decision to extend our corpus to biographical and dystopian literature was based on the assumption that these types of texts would frequently consider ‘what could have been but was not.’ This assumption turned out to be particularly true for Mandela’s autobiography and Obama’s biography on his father. Young activist Malala’s and poet Audre Lorde’s autobiographies enriched our corpus together with Bradbury’s and Orwell’s novels.

We used the freeware toolkit AntConc to identify the contextualized occurrences of Should have + Past participle. We elicited 95 sentences containing this item. We stored every sentence carrying the targeted item within a table in a Word file. For the sake of transparency, every occurrence was given a code composed by the initials of the author plus a figure (e.g., VW#01, BO#95). This code appears in the Results section to quote or comment some of our examples. Working with AntConc allowed us to quantify the occurrences of Should have + Past participle per manuscript and thus, to have a picture of the authors who used it more and less frequently (Figures 2 and 3).

In general, we proceed to identify the equivalent translations in Spanish by using AntConc although a small part was done manually.[7] Every translated sentence was stored right after the English source sentence in a Word table to allow a segmented overview. Then we analyzed the grammatical features of the Spanish segment corresponding to Should have + Past participle (e.g., tenses and verbal moods of the 95 translations, presence or absence of modal markers, word order) and placed the data in an Excel file. Finally, the Spanish translations were crossed with a reference corpus in Spanish containing 154 million words (CREA), which is open and accessible online.

Table 1

Corpus’ source texts and translations

Corpus’ source texts and translationsCorpus’ source texts and translations

-> Voir la liste des tableaux

4.1 The English corpus

The quantification of each manuscript’s words and the quantification of the Should have items gave us an approximate volume about the role of counterfactuality in conceptualizations. The volume of the irrealis readings expressed by Should have + Past participle is 0.005% in our English sample. Of course counterfactuality can be expressed by other means not analyzed here, like if- clauses or nominalizations. But still the message is clear: discourses referring to counterfactual scenarios are only a small part out of what we communicate. Biographies turned out to be one of the most prolific types of text to obtain Should have items (18 occurrences in Obama’s Dreams from My Father and 12 in Mandela’s Long Walk to Freedom). This seems to be logic with the motivation of writing such a text and the intellectual exercise of looking back to the past with some degree of criticism.

Interestingly, the use of Should have + Past participle was not correlated to the use of Should. Kahneman’s Thinking fast, thinking slow presented the higher ratio of Should (259 occurrences) but only 3 of them were succeeded by have plus a past participle. This dropped Kahneman’s ratio to 0.001% in what concerns the number of Should have + Past participle per word. The same ratio was observed in Federici’s and Butler’s texts. As said above, the sample’s mean of Should have + Past participle per words was 0.005%, and this is precisely the value found in Malala’s biography. The highest ratio of Should have per words was observed in Woolf’s essay (0.04%) followed by Obama’s and Bradbury’s texts (0.01% respectively). This huge gap between the English author and the rest of the sample’s writers can be observed in Figure 2. The 12 occurrences found in Woolf’s text represented a particular case compared to the rest of our corpus’ essays, which contained an average of 3.5 occurrences per author (i.e., 25 occurrences from 7 manuscripts). This mean is lower than the average of occurrences found in the biographical type of texts, for which we obtained a mean of 10.25 occurrences per manuscript (i.e., 41 occurrences from 4 authors). The dystopian novels from Ray Bradbury and George Orwell revealed not to be so prolific texts regarding the production of Should have + Past participle, accounting for a mean of 6 occurrences per manuscript. As said at the beginning of Section 3, the nature of the 14 texts compounding our corpus is diverse. We can distinguish 6 narrative texts – i.e., 4 biographies and 2 novels – and argumentative texts represented by 8 essays, on the other hand. As for the production of Should have + Past participle, we can therefore conclude that narrative texts proved to be more prolific, especially biographies.

Figure 2

Ratio of Should and Should have + Past participle per author

Ratio of Should and Should have + Past participle per author

-> Voir la liste des figures

Figure 2 highlights the difference between Woolf’s and the rest of the group regarding the use of Should in relative terms. Woolf’s text contains more occurrences of Should per words than any other (0.17%). The second text in which Should is more used per words is Kahneman’s Thinking fast, thinking slow (0.13%). This means that A room of one’s own is a text highly modalized in itself. In Section 3.2 we will see whether this stylistic feature has an impact on the Spanish translation. Conversely, the lowest ratio of Should per words comes from Butler’s and Lorde’s texts (0.01% and 0.02%, respectively). The mean’s group of Should per words is 0.07%. Figure 2 shows, for every manuscript, the relationship between the occurrences of Should and Should have + Past participle per words. A summary of the numerical data for each author is available in the Appendices (Appendix 1). The group’s standard deviation regarding the frequencies of Should is 0.042, whereas the group’s standard deviation regarding Should have + Past participle is 0.01. This means that the data points regarding the amount of Should have + Past participle are clustered more closely around the mean compared to the Should values, which are spread slightly further from the mean. Therefore, Figure 2 illustrates a more compact group regarding the frequencies of Should have + Past participle and a more eclectic group regarding the frequencies of Should. Figure 3 represents graphically the number of occurrences of Should have + Past participle per words, for every manuscript. Mandela’s and Obama’s biographies appear in the top area of Figure 3, suggesting that this type of text is fertile in conceptualizations evoking counterfactual states-of-affairs.

Figure 3

Frequency of Should have + Past Participle per author

Frequency of Should have + Past Participle per author

-> Voir la liste des figures

4.2 The Spanish translations

Our results shows that Should have + Past participle is translated in Spanish by a repertoire of verbal tenses and periphrastic solutions. For example:

From a formal viewpoint, translations (12) to (14) share the word order of the source construction (i.e., Modal marker + Auxiliary verb + Past participle). Both translations (15) and (16) carry the Spanish subjunctive but the difference between them is that number (16) lacks a modal verb. From these general observations, we can distinguish different factors allowing us to classify the Spanish translations: (i) word order,[8] (ii) the presence or absence of a verbal form related to Deber (Should), and (iii) the main verb’s mood. These features were accurately described one by one for each of the 95 occurrences in an Excel table (Appendix 2). The detailed description of these variables allowed us to run statistical analyses using the software R. The point of doing so was to know whether the presence of a modal verb was correlated to a certain word-order or to a certain length across the translated constructions.

The most frequent pattern used by translators is Debería haber + Past participle (29.5%). This construction was followed by the Debía haber + Past participle (13.7%), and by the pluperfect subjunctive with no modal verb (12.6%). The construction formed by the pluperfect subjunctive carrying the modal Deber in the past participle position was used in 9.5% of the translations. In other words, the preferred option by translators was example (12), which is a word-by-word translation of the English source construction. The second translation more frequently used corresponded to example (14), which consists of the Spanish imperfect. Note that both of these translations replicate the word-order of the English source construction (i.e., a conjugated modal verb followed by the auxiliary verb’s infinitive followed by the construction’s main verb in the past participle position). This is to say that the two most frequent translations share the same semantic and word-order patterns, which appear in Appendix 2 under the acronyms of MAV and CIP, respectively. Together, these translations represent 43.1% of the total.[9] This result suggests that the English source construction has a notable effect in the word-order and the semantics of the Spanish translations. The third and the fourth translations most frequently used corresponded to examples (16) and (15), respectively, which both contain subjunctive verbal forms. Together they represent 22.1% of the total translations but, unlikely examples (12) and (14), their semantic and word-order patterns do not converge. The four most frequent translations discussed above – examples (12), (14), (16), (15) – represented 65.2% of the total. In Table 2, they have been highlighted in grey to distinguish them from other translations relatively less used.

Table 2

Translations of Should have + Past participle (n = 95)

Translations of Should have + Past participle (n = 95)

-> Voir la liste des tableaux

Table 2 shows a great variability in the Spanish constructions provided by translators. This is because of some constructions from which only one or two occurrences were elicited, like the progressive periphrasis Ir a + Infinitive (iba a ser), the modal verb Poder (podía) or some forms from the pluperfect conditional carrying Haber as a main verb (i.e., habrían transferido; habrían tenido que ser). In addition, we observed the use of an infinitive as well as two verbal forms within adverbial constructions (i.e., en lugar de ir a parar; como debía; de lo que debía).

The main verb’s mood more frequently used was the conditional, with a frequency of 35 out of 95 (36.8%), followed by the indicative (32.6%) and the subjunctive (27.3%). The comparison of these frequencies revealed no significant differences, (3, N = 95) = 1.9582, p = 0.37. This result means that the subjunctive, the indicative and the conditional are similarly used across the translated constructions in Spanish.

A factor that revealed significant differences was the presence of the modal verb Deber within the translated constructions. Independently of the tense or the main verb’s mood, Deber appeared in 71 out of 95 translations. Chi-squared analysis revealed that this feature is a salient pattern of the output constructions, (2, N = 95) = 52.725, p <0.001. The high frequency of Deber contrasted with the low frequency of the modal expression Tener que, from which we obtained 6 occurrences.[10] In five of them Tener que appeared as the past participle of a pluperfect subjunctive (e.g., Hubieses tenido que verla). Only in one occurrence Tener que appeared as the past participle of a past conditional (e.g., Habrían tenido que ser idénticos).

Figure 4

Verbal mood within Spanish translations (n = 95)

Verbal mood within Spanish translations (n = 95)

P-value resulting from the t-test analysis

-> Voir la liste des figures

The comparison between the translated constructions in Spanish revealed critical differences between some translators. Concerning the tense variable, we found a significant difference between Laura Pujol’s Una habitaciónpropia on the one hand, and Antonio Resines and Herminia Bavia’s El largo camino hacia la libertad as well as Fernando Miranda and Evaristo Páez Rasmussen’s Los sueños de mi padre on the other hand, t (95) = 4.2426, p = 0.0006 (Figure 4). The pluperfect subjunctive with no modal markers was more frequently used by the former (e.g., hubiera hecho), whereas the latter privileged the past conditional marked by the modal Deber (e.g., debería haber hecho). Figure 4 shows that Alfredo Crespo’s Fahrenheit 451 is the only translator to join Laura Pujol’s preference for subjunctive tenses. But unlikely Pujol, Crespo tends to mark his translations with the modal Deber in the past participle position (e.g., hubiera debido). Figure 5 shows that keeping a modal marker is a generalized pattern in the translations of María Durante’s Zami, Julia Fernandez’s Yo soy Malala, Fernando Miranda and Evaristo Páez Rasmussen’s Los sueños de mi padre and Antonio Resines and Herminia Bavia’s El largo camino hacia la libertad.

Figure 5

Presence of modal markers related to “should” within Spanish translations (n = 95)

Presence of modal markers related to “should” within Spanish translations (n = 95)

P-value resulting from the t-test analysis

-> Voir la liste des figures

Concerning the maintenance of a modal marker in the Spanish translations, Pearson’s correlation test confirmed a significant relationship between the word order ‘Modal marker + Auxiliary verb + Past participle’ and the use of tenses from the indicative mood, r = 7.4341, n = 95, p = 5.064e-11. In other words, keeping the word order of the source language generally implies for the translator to exclude the subjunctive as a way to express counterfactuality in Spanish. Conversely, the use of the Spanish subjunctive allows disengaging from modal markers more likely compared to indicative predicates, as in example (18). Another factor that prevents the transferring of modal markers in Spanish is the length of the output construction. Those compounded in Spanish by one or two units are less likely to contain Should equivalents, as in example (19).

The Spanish translations composed by two verbal units or even just one represented 35.7% over the total (34 out of 95). Within this subgroup, verbal forms from the subjunctive were the most used (17 occurrences), followed by verbal forms from indicative tenses other than the conditional (10 occurrences) and by verbal forms from conditional tenses (7 occurrences). This means that the use of non-compounded tenses – as the present subjunctive in example (19) – prevents translators from including a modal verb in the output construction.

Figure 6 illustrates the distribution of two factors across the translations: (i) the presence of a modal marker, and (ii) the mood of the verbal construction. Figure 6 shows that the combination of indicative tenses and a modal marker is the most frequent way of translating Should have + Past participle.[11] Conversely, the presence of modal markers is less frequent in those translations carrying subjunctive tenses.

Figure 6

Mood and modal markers in the Spanish translations

Mood and modal markers in the Spanish translations

-> Voir la liste des figures

4.3 The reference Spanish corpus CREA

The results from our corpus were crossed with the Spanish reference corpus CREA, which contains 154 million words coming from 140.000 written documents published between 1975 and 2004. About 50% of the CREA’s documents come from different types of manuscripts and the other 50% from press articles. Different varieties of Spanish are represented, with 50% of documents coming from diverse American varieties (Antilles, Chili, Mexico, Argentina, USA, among others), and 50% from Spain. The quantity of texts reunited in this corpus and the diversity of their sources made it appropriate in our mind in order to cross the translation outcomes from our Social Sciences parallel corpus.

Table 3

CREA’s frequencies corresponding to the Spanish translations[12]

CREA’s frequencies corresponding to the Spanish translations12

-> Voir la liste des tableaux

Interestingly, the Spanish output constructions more frequently elicited in our corpus represented 0.004% of the CREA’s volume. This suggests that from a semantic viewpoint counterfactual readings are similarly represented in the conceptualizations of our English corpus (0.005%) and in the CREA Spanish corpus.

Searches throughout the CREA corpus were made on-line by collocations of two verbal units. In Spanish, the collocations carrying a verbal form from Deber plus the auxiliary infinitive Haber (e.g., debería haber) are in general followed by a past participle. Because of this grammatical regularity, searches concerning this type of construction were rigorous and the frequencies obtained exact. As for the searches involving pluperfect subjunctives, we proceed by collocations made of the auxiliary hubiera or hubiese plus either a modal past participle – debido or podido – or a past participle coming from verbs that we assumed as being pretty much used in Spanish. For this reason the results of Table 3 involving subjunctive tenses are approximate and we assume that the CREA corpus contains other occurrences that we have not accounted for. The results based on the CREA corpus revealed three features accounting for the expression of counterfactuality in Spanish. These features are the following:

  • The predominant use of the pluperfect subjunctive as a prompting tense for counterfactual readings. Table 3 shows a higher use of the pluperfect subjunctive compared to the word-by-word translations Debía / Debería / Debimos haber + Past participle (5,316 occurrences versus 1,071, respectively). This ratio represents a great shift compared to the frequencies found in the Spanish translations. The pluperfect subjunctive was used in 22% of the output constructions (cf. Table 2). However, in the CREA corpus it represents 83%. Interestingly, the frequency of modal verbs in the past participle position represented 38% within the pluperfect subjunctive – i.e., Hubiera or hubiese podido / debido –, which is similar to the amount found in our corpus (Figure 6). This suggests that translators are aware about the frequencies of constructions combining a modal verb and a subjunctive tense in Spanish.

  • The non-iconicity of modal verbs to express the speaker’s subjectivity in Spanish. The amount of occurrences marked by modal verbs in Table 3 represented less than the total’s half. Modal verbs Deber and Poder were elicited in 3,091 constructions against 3,296 modal-free constructions (48% versus 52%, respectively). This suggests that the role of modal verbs is less important than the use of subjunctive tenses in order to convey counterfactual readings in Spanish. Moreover, the low frequency of the verb Deber within pluperfect subjunctive occurrences (2.3%) raises the question whether it is perceived as a redundant combination for Spanish users.

  • The use of modal verbs within verbal predicates in written Spanish, which generally appear in the past participle position (e.g., No hubiese debido tener libros). Table 3 suggests that modal verbs are more likely to be used in the past participle position within pluperfect subjunctives rather than as indicative tenses introducing a periphrasis (2,020 occurrences versus 1,071, respectively). In other words, Spanish privileges verbal sequences like hubiera debido / podido rather than debería / debía haber. This preference challenges the overuse of debería / debía haber by translators compared to hubiera debido. If we apply this observation to the translations of section 3.2, we conclude that the construction No hubiese debido tener libros (RB#58) is likely to sound more familiar to native Spanish readers than Todo debería haber sido diferente en Rosario (NK#23). Still both translations contain the modal verb Deber but whereas the latter replicates the word-order of the source language by means of the conditional tense, the former carries the pluperfect subjunctive.

5. Discussion

In the present article Should have + Past participle has been considered and analyzed as a unit of translation. Among his translation rules, Newmark (1986: 129) states that a translator should not translate an SL item into a TL item which would normally and naturally have another equivalent in the SL. The importance of producing output texts according to the language use of native speakers and readers is crucial, especially when involving a major language like English. Resistance to transfers differs from one linguistic community to another and tolerance tends to increase when translation is carried out from a highly prestigious language/culture (Toury 1995: 278). Positive transfers are less noticeable than negative ones, but they still can make the difference between sourcier-driven and cibliste-driven translators, to put it in Ladmiral’s terms (2014).

Attempts to produce translations which would represent their sources ‘as closely as possible’ are in constant tension with the attempts of establish in a target culture and language native-like texts. (Toury 1995: 276)

Our data confirmed the incidence of literal translation – fairly straightforward word-by-word processes (Pym 2014: 13)- in Spanish related to the English construction Should have and thus, the impact of the source language at the morphological, compositional level. Previous results in this domain have shown that English translations involving the suffix – less and the prefix un – were markedly differently depending on whether the source languages were Romance or Germanic (Lefer and Cartoni 2013). Our results suggest that translators practice is not coherent with the language use of native Spaniards in counterfactual contexts. We base this claim on the asymmetrical frequencies found in the CREA corpus and our sample concerning the construction Debería haber + Infinitive. Accordingly to grammarians, the access to non-factual interpretations in Spanish is genuinely prompted by the use of subjunctive tenses.[13]

El modo subjuntivo aparece de forma destacada en una serie de contextos modales o modalizados que suelen llamarse no factuales (es decir, no reales, no verificados o ni experimentados), lo que sugiere que la flexión del subjuntivo podría considerarse la manifestación en forma gramatical de las nociones abstractas que corresponden a estos contextos.

RAE 2009: 1866

The subjunctive mood is used remarkably in certain modal or modalized contexts which are called non- factual (i.e., unreal, non-verified or non-experimented). This suggests that the flexional features of the subjunctive could be seen as the grammatical way to encode abstract notions that pertain to suchcontexts. (our translation)

However, the translations analyzed proved to disregard the role of the Spanish subjunctive and favored in general the combination of conditional or indicative tenses with a modal marker. At a lexical level, translators preferred the modal marker Deber over Tener que – 71 versus 6 occurrences, respectively –. The translators’ overuse of Deber over Tener que is not coherent with the frequencies of both modal markers in Spanish, where the latter appears to be more frequent than the former (RAE 2009: 2144).

Beyond the subjunctive mood, Spanish displays a variety of indicative tenses that may refer to counterfactual contexts. Among them, the so-called auxiliares modales de obligación fill the functions of the subjunctive mood by means of the Spanish imperfecto (RAE 2009: 1871). This is a feature that Spanish translators seem to be aware of and which they applied in 13.7% of the output translations. This said, the past conditional also plays an important role and is considered a direct concurrent of the pluperfect subjunctive, although the latter is extensively more used in the American varieties of Spanish.

Alternan habría cantado y hubiera cantado en la mayor parte de los contextos, como en Yo lo {habría ~ hubiera} hecho[14] de otro modo, si bien se ha observado que en el español americano es más común la segunda opción que la primera.

RAE 2009: 1795

In the majority of contexts ‘habría cantado’andhubiera cantado’are used randomly – e.g., I {would have ~ have had}made it differently – . However, it has been observed that the latter is more frequent than the former in the American variety of Spanish. (our translation)

Thus, we can conclude that the past conditional translations of Should have + Past participle are coherent with the Spanish prominent tenses in counterfactual contexts. But what factors explain its higher frequency over the pluperfect subjunctive translations? In Section 3, evidence has been given about the influence of the word-order and the length of translations in using the indicative or the subjunctive mood. We consider the construction Debería haber + Past participle a word-by-word translation that might result from both automatized mechanisms and linearity in the translator practices. Translating Should have by Debería haber has economic advantages in terms of time and cognitive processes for the translator. However, it has an effect in the readability of the output text by Spanish users, who assist in a reversal in the frequencies of the prominent constructions related to counterfactual contexts. The question arises whether the preference by Debería haber has to do with the translators’ approach to the English construction Should have. To which extent do they approach and process it as one entity or as an organization of lower-level linguistic entities? Are translators who privilege the Spanish subjunctive with no modal verbs more influenced by a communicative approach? Are translators who privilege the Spanish conditional plus a modal verb more influenced by the task’s linearity? These are questions that further research must clarify.

6. Conclusions

We hope to have drawn attention to the crucial role that linguistic factors – e.g., word-order, lexical units and length – play in translation and, more specifically, in covert or positive transfers. The degree of awareness of the target language’s uses and its prominent categories makes the difference among translators. In the present article, we have shown that the most frequent way of translating Should have in Spanish is the word-by-word translation Debería haber. This preference is not coherent with: (i) the marginal role of modal verbs to express the speaker’s subjectivity in Spanish; (ii) the use of modal verbs within verbal predicates in written Spanish, which generally appear in the past participle position (e.g., No hubiese debido tener libros); nor (iii) the predominant use of the pluperfect subjunctive as a prompting tense for counterfactual readings.

Translating the counterfactual construction Should have + Past participle by Debería haber can be considered an idiomatic error in Spanish because it implies a reversal in the frequencies of the Spanish counterfactual constructions. The higher frequency of the pluperfect subjunctive over Debería haber in the reference corpus CREA suggests that Newmark’s principle of frequency (1986: 145) works as a reliable cue to prevent “translationese” (the influence of properties of the source language in a translated text or construction). This author claims that equivalent frequency of usage in source and target language applied to grammatical structures and lexis is particularly useful as an additional method of verifying a translation.

Hansen (2010) makes the distinction between genuine translation errors and idiomatic errors. The former are errors based on the relationship between a source text and a target text, whereas the latter must be regarded as ‘errors in translation’ or usual problems of language text production. Nonetheless, idiomatic errors can have a considerable impact on the quality of the target text (Hansen 2010: 386). Assigning equivalence values between the English Should have and the Spanish Debería haber may be the result of a covert translation’s error. House (2001) claims that the function of a covert translation is to reproduce in the target text the function the original has in its frame and discourse world, so that the translator must attempt to recreate an equivalent speech event. The frequency of the pluperfect Spanish Hubiera / Hubiese in counterfactual contexts makes it a more idiomatic candidate than Debería haber. Being sensitive to the salient categories of the target language seems to be an effective strategy to prevent word-by-word translations and thus, to enhance the readability of the output text.

The findings presented in our survey point out some empirical hints for translators and translation students working from English to Spanish. First, the way in which we extracted and processed data via AntConc may be replicated in the classroom as a way for students to explore corpora and to test hypothesis quantitatively. Second, our survey provides reliable cues to prevent the word-by-word translation Debería haber + Past participle. Some useful strategies to do so are (i) avoiding infinitive periphrastic constructions and modal markers related to deber and tener que (Should), as well as (ii) favoring subjunctive mood’s tenses and output constructions carrying less than three constituents (i.e., two-word constructions or one-word constructions). At a descriptive level, linguistic variables such as semantic elements and length proved to have an impact in the target language. Raising awareness on these features seems crucial not only to encourage critical analyses among translation students but also to stimulate their sensitivity to the output texts’ readability.

Parties annexes