Counterexpectational Translations: The Dutch Markers Toch and Eigenlijk Contrasted with Their English Correspondents

The Dutch particle toch and modal adverb eigenlijk both belong to the broad domain of ‘contrast’: their prototypical functions have been described as ‘concession’ and ‘counterexpectation’, respectively, and both have been attributed emphatic discourse uses (e.g. toch as a common ground reminder and eigenlijk as a reinforcement marker). This study analyses the multifunctional scope of toch and eigenlijk in Dutch and maps their translation equivalents in English through an in-depth analysis of the Dutch-English component of the Europarl corpus. For eigenlijk the most common correspondents (in fact, really, actually) belong to the core of the semantic domain of (counter)expectation, whereas none of the English correspondents of toch is able to capture the marker’s duality as both a concessive and a common ground marker. The high degree of zero correspondences in English source texts suggests the translators’ strong awareness of the markers’ pragmalinguistic contribution.


Introduction
Especially since the 1980s particles have enjoyed scholarly interest in the study of Dutch (see e.g. Van der Auwera and Vandeweghe, 1984;Hulshof, 1987) Contrastive PragmaticS 4 (2023)  in the same period when discourse markers started to attract attention in anglophone regions (see e.g. Schiffrin, 1987). Taxonomic, terminological and definitional issues have ever since been the focus of many debates (Degand et al., 2013), testifying to the heterogeneous nature of these items. In English they are often taken to form a class of pragmatic items belonging to different word classes (e.g. adverbs, interjections), whereas in Dutch (as well as in e.g. German and Swedish) they are typically considered a distinct word class of items carrying pragmatic force (referred to as particles, e.g. Van der Wouden and Caspers, 2014). Interestingly, descriptions of (modal) particles in Dutch (or German) (e.g. Abraham, 1991) and those of discourse markers in English (e.g. Schourup, 1999) have been remarkably similar, in that both types of items are attributed high frequencies, have been shown to be multifunctional, have undergone grammaticalisation processes, do not carry propositional content and are hence often considered semantically and syntactically optional. Especially their functional spectrums -i.e. the range of functions they can fulfil -would appear to demonstrate considerable overlap, while their syntactic behaviour displays diverging preferences (Degand et al., 2013). In a Dutch-English cross-linguistic perspective, roughly three types of correspondence can be discerned. First, there are English markers with Dutch counterparts that are functionally and formally similar, such as look and kijk in (1), which are both derived from a verb of looking, occur in the same position and perform comparable functions. The Dutch particle zelfs, on the other hand, can easily be expressed by the adverb even in English (as in (2)) but has a different origin and the syntactic constraints differ (even occurring in front of the main verb and zelfs in front of the phrase it holds focus over). The modal particle maar, finally, does not have a ready equivalent in English, which instead offers different solutions depending on the context, such as an emphatic DO construction in (3).
'KIJK, this way we'll never find a solution.' Look, this way we'll never find a solution.
(3) Doe de deur maar dicht. 'Close the door MAAR.' Do close the door. Buysse Contrastive PragmaticS 4 (2023) 178-212 As Aijmer and Simon-Vandenbergen (2003) point out, translation approaches to the study of discourse markers can, on the one hand, deepen our insight into the functional and semantic potential of specific markers in one language by verifying how the same meaning or function can be conveyed in another language in the same context, and on the other hand, enhance our knowledge of discourse markers in general.
This study will zoom in on two particles that occur with a high frequency in spoken Dutch but have rarely (if at all) been subjected to corpus analyses: the concessive marker toch and the (counter)expectation marker eigenlijk. Our aim is to identify their correspondents in English and through this interlingual comparison foster insight into these markers' functional potential. To reach this aim we will address the following research question: how are the Dutch markers toch and eigenlijk translated into English, and do different functions of toch and eigenlijk also yield different translation correspondents? Toch has clear cognate forms in other Germanic languages such as German (doch) yet not in English, whereas eigenlijk has an intuitively obvious correspondent in English (actually) even though these are not cognates. It would, therefore, stand to reason that the translation profiles will look markedly different.
In Section 2 toch and eigenlijk will be introduced through a brief literature overview. After the data and methodology have been presented in Section 3, the results will be discussed in Sections 4 and 5, both in functional and quantitative terms.

Toch and Eigenlijk
By and large, three core functions can be posited for toch, as outlined by Hogeweg et al. (2011). First, it serves as a contrastive marker of concession. In such cases, toch responds to an inference belonging to the common ground, as in (4) where one might infer from Bill's illness that he is unable to come to an event but toch indicates awareness of this default inference while introducing a proposition that contradicts it. Second, toch may indicate inconsistency with common ground, in which case toch is stressed. In (5) toch marks that B's attempt to get a piece of equipment to work is inconsistent with the established common ground that it does not work. Third, when toch is unstressed it can act as a reminder of common ground, as in (6), where it reminds the coparticipant of the speaker's ability to perform a task. Zeevat and Karagjosova (2009) explain this apparent paradox in the uses of toch -expressing both "a correction" and "a reminder of some common ground fact" (2009: 136)through grammaticalisation processes: toch originates from a form that was Contrastive PragmaticS 4 (2023) 178-212 used in an interrogative context in which "the speaker seeks reconfirmation of a hearer opinion" (2009: 152); in a declarative context, toch can then evoke both confirmation and its opposite, the distinction being made clear by intonation and sentential position.
(4) Bill is ziek maar toch wil hij komen. 'Bill is ill but TOCH he wants to come.' ("still")  Foolen (2006) analyses toch in its concessive use as an "adversative connective" (2006: 64), and when it marks inconsistency with common ground he considers it a "modal particle" (2006: 66). In the latter case toch can be found in declaratives (the speaker checks with the hearer whether they still share the same attitude), imperatives (marking a request), exclamatives (indicating surprise), and wh-questions. Van Bergen et al. (2011) describe eigenlijk as "among the 50 most frequent tokens in the Corpus of Spoken Dutch" and yet they observe that there are "surprisingly few" (2011: 3878) analyses of the marker. Its core function is that of an "expectation management device" (van Bergen and Bosker, 2018: 191), which prepares co-participants for what the speaker assumes is an unexpected proposition. As van Bergen et al. (2011) explain, "[b]y using eigenlijk the speaker acknowledges that there is an alternative expectation or interpretation of reality that is more likely from the perspective of the hearer (as estimated by the speaker)" (2011: 3881). Eigenlijk may take scope over a proposition, as in (7), where it indicates the speaker's understanding that the proposition that he is a man may run counter to expectations. It may also function at a metalevel to show awareness of the unexpectedness of the utterance itself, e.g. in (8) the speaker indicates that they should know the answer to their own question but don't.
(9) Wanneer de last niet dichter bij het lichaam kan gebracht worden, moet je trachten het lichaam, eigenlijk het lichaamszwaartepunt, dichter bij de last te brengen. 'When the burden cannot be brought any closer to the body, one should try to bring the body, EIGENLIJK the bodily center of gravity, closer to the burden.' (Mortier and Degand, 2009: 356) Both toch and eigenlijk have been included in some contrastive studies, but this has mostly involved comparisons with German rather than English. This is particularly the case for toch, which has been likened to German doch (e.g. Foolen, 2006;Hogeweg et al., 2011;Zeevat and Karagjosova, 2009). In a corpus of translated fiction, Buysse (2017) identifies toch as one of the most common translation equivalents of English question tags in Dutch, occurring both as an invariant tag in clause-final position (signalling hearer involvement) and as a modal adverb in medial position (prompting co-participants to retrieve common ground information). Eigenlijk appears in Aijmer and Simon-Vandenbergen's (2004) mapping of the semantic field of 'expectation' based on (translation) data from English, Swedish and Dutch. Together with in feite ('in fact') it turned out to be "most multifunctional " (2004: 1795) in Dutch, and actually, really and in fact were said to "belong to the core of the lexical field" (2004: 1797) in English. The nature of the observed translations led Aijmer and Simon-Vandenbergen to suggest that eigenlijk is a close "match" (2004: 1794) of actually and really (and to a lesser extent of in fact).

Data and Method
The approach adopted for this study is that of a contrastive analysis based on a parallel bilingual corpus, i.e. a corpus that "consists of texts in two different languages, with the texts in one language being the originals of the translations in the other language" (Vandevoorde, 2020: 11). Translation data enable forms in one language to be mapped against corresponding forms in another Contrastive PragmaticS 4 (2023) 178-212 language in a one-on-one comparison. As Aijmer and Altenberg (2002) contend, "a bidirectional translation corpus offers a fruitful basis for studying areas where there is a lack of equivalence between languages" (2002: 38), which would appear to be the case between Dutch and English for toch and, possibly to a lesser degree also for eigenlijk.
The source text (ST) and target text (TT) belong to the same context, with two exceptions: they have different target audiences belonging to different cultures, and the translator is not in full command of the content of the text. A translator, rather, autonomously interprets a ST and subsequently renders a TT that conveys the ST author's intentions in the best way possible. Johansson (2006) observes that "[w]hat we are studying is the result of this interpretation (and recreation) process" (2006: 17), which can be likened to co-participants' interpretation processes in interaction.
Parallel corpora of spoken data are quite rare, certainly those including Dutch. An interesting exception is Europarl (Koehn, 2005), a sentence-aligned corpus extracted from the proceedings of the European Parliament and available for 21 languages. Europarl7 was released in 2012, contains texts from the period 1996-2011, and can be accessed and searched through Sketch Engine (Kilgarriff et al., 2014). The proceedings are verbatim reports of plenary sessions of the European Parliament, yet they are not linguistic transcriptions, in that filled pauses and truncations have been removed and transcriptions do not contain any prosodic information.
The setting is clearly institutional: turn changes are managed by the meeting chair, speech times are restricted, the topics are pre-determined, etc. Since toch and eigenlijk have not been attributed turn management roles, this context is unlikely to impact our findings. Even though some, if not most, of the discourse in the European Parliament may be scripted (as speakers will have prepared their interventions), many Members of the European Parliament (MEPs) clearly divert from their scripts (e.g. in explicitly picking up on colleagues' turns).
The Dutch-English section of Europarl contains 59,756,704 tokens. Toch was found 26,711 times in this corpus and eigenlijk 11,426 times. The maximum number allowed (10,000 tokens) was downloaded. As tokens are not tagged to distinguish between those that appear in Dutch STs and those in Dutch TTs, in order to obtain the former, only those attributed to native speakers of Dutch were selected. This resulted in 1,970 tokens of toch and 2,038 of eigenlijk appearing in Dutch STs. Similarly, to single out toch and eigenlijk in Dutch TTs, tokens uttered by a native speaker of English giving rise to these in the Dutch translations were selected, yielding 2,040 tokens of toch and 719 of eigenlijk. In order to ensure the feasibility of this study, all selections were further narrowed down Buysse Contrastive PragmaticS 4 (2023) 178-212 through random selection of 200 tokens for each marker and each translation direction (resulting in a total of 800 tokens). Randomisation was conducted manually: the relevant tokens were placed in order of appearance in the corpus, and subsequently the first ten in each group of 100 tokens were selected, except for eigenlijk in Dutch TTs (for which more were selected in each group of 100 -e.g. tokens 1-10, 41-50 and 80-86 in the first group of 100 -to arrive at 200 tokens in total, distributed more or less evenly across the corpus). The selected tokens were produced by 41 unique speakers for toch and 43 for eigenlijk in the Dutch STs, and the English utterances that gave rise to toch and eigenlijk in Dutch TTs were produced by 54 and 85 speakers, respectively. The higher number of speakers in the latter can be explained by the higher number of English-speaking than Dutch-speaking MEPs in the European Parliament at the time of data collection.
Each of the 800 tokens was submitted to a functional analysis by the re searcher, who drew inspiration from previous analyses of the markers. The classification was submitted to a process of intra-coder reliability testing (i.e. the researcher repeated the classification process about one month after the first round), which resulted in a 99% correspondence rate. Subsequently, a second coder was introduced to the functional classifications of both markers and independently coded the tokens, yielding a correspondence rate of 89%. The two coders, finally, discussed the tokens lacking inter-coder agreement to reach consensus.
All translation correspondents were identified and counted. In line with Johansson's "framework of correspondence" (2007: 25), correspondents are looked at from both translation directions (in STs as well as TTs), and both those that are overtly expressed (i.e. with a linguistic form in the text in the other language) and those displaying zero correspondence (i.e. with no linguistic form in the text in the other language) are identified.

4
Analysis of Toch

4.1
Functions of Toch Ten functions of toch were attested in the corpus (Table 1). Among these are the two core functions as they come forward in the literature: a marker of 'concession' , and a reminder of common ground. The other functions are to varying degrees related to either or both of these. 'Counterexpectation' is in the case of toch an implicit form of 'concession,' whereas 'justification of the relevance of upcoming talk' functions at a metadiscursive level, where it concessively invokes conceivable reservations against the prospective talk. Similarly, Contrastive PragmaticS 4 (2023) 178-212 in 'seeking support for an assertion or opinion' toch tacitly instructs coparticipants to rely on assumed common ground. Other functions, still, appear to merge elements of both core functions. When expressing 'insistence,' toch invokes that a question or request was already known to co-participants (common ground) but although there may have been good reasons not to act upon these, the speaker insists to do so (concession). As a hedge, it demonstrates awareness that the speaker's assertion defies the expectation of alignment with co-participants' opinions (concession) but also that the speaker seeks understanding for their position (establishing common ground). In its 'causal' function, toch concessively dismisses conceivable objections to a proposition by relying on assumed common ground. When it can be paraphrased by 'at least,' toch narrows the scope of a general proposition to a specific context (concession), again by referring to assumed shared knowledge. Thus, the functional picture that emerges is that of a polysemous marker that has two core functions, each of which are related to less central functions, and since the two main functions are wholly compatible, their meanings also combine in a range of minor interpersonal uses.
The functional analysis in the following sub-sections will illustrate the functions and toch's translation correspondents. Table 1 Functions of toch in the present corpus

Functions of toch
Mainly concessive -Concession -Counterexpectation -Justify the relevance of upcoming talk -Disbelief Mainly common ground -Reminder of common ground -Seek support for an assertion or opinion Concessive & common ground -Insistence -Hedge -Causal -'at least' (narrow the scope of a general proposition to a specific context)

Buysse
Contrastive PragmaticS 4 (2023) 178-212 4.1.1 Concession Toch's core 'concessive' function (cf. Foolen, 2006: 64-66;Hogeweg et al., 2011: 54) accounts for almost half of all analysed tokens (N = 94 or 47%) in the Dutch STs and even considerably more in the texts translated from English (N = 124 or 63%). In such instances the speaker indicates that, in spite of the prior co-text, the proposition marked by toch should be considered after all. In excerpt (10) the MEP has just noted that more spaces are being made smokefree, and now claims that in spite of these efforts, smoke prevention should not be neglected either.
Nevertheless social pressure must be stepped up to a point where young people especially do not start smoking, and women to not take up smoking either.
(#5586080, Maes1) With 23 corresponding tokens, the preferred translation correspondent of 'concessive' toch is nevertheless and its twin form nonetheless, along other concessive/contrastive forms, notably still (15), despite this (13), yet (4), and however (3). In 27.7 percent of cases, toch received zero correspondence (N = 26), i.e. the TT did not contain an overt correspondent. In nearly all of these instances another concessive or contrastive element was also present in the ST and is indeed conveyed in the TT. The presence of ondanks ('despite') in the previous phrase in (11) reduces the need to add a second concessive marker in the same clause in the translation, just as the combination of maar ('but') and toch in (12) can be rendered with a single word in English (however), although this may make the TT less expressive and less emphatic.
In practice, women are often discriminated against during this period, despite rights on paper.
(#7467872, Buitenweg) Interestingly, the proportion of zero correspondence is even higher when going from English STs to Dutch TTs. Toch is added in the Dutch TT without a specific corresponding form in the English ST in 38.7% (N = 48) of concessive cases.
Here too the English ST often contains another contrastive or concessive element but so does the TT. In (13) a concessive relation is marked in the English ST (whilst), which receives a corresponding form in the subclause in the Dutch TT (hoewel, 'while'), as well as toch in the main clause, which amounts to a double concessive marking.
(13) Whilst I understand the situation, for example, in Austria, where they have gone further, in the interests of the single market we should maintain the legal base.
(#17130070, Goodwill) Another noteworthy observation is that in the TTs toch is rarely sparked by nevertheless and nonetheless (N = 9, 7.2%) but all the more so by yet, which accounts for one in four concessive tokens of toch (N = 31) and is virtually absent as a translation correspondent of toch in the reverse direction (N = 4, 4.3%).

4.1.2
Counterexpectation As it appears in the present corpus, toch expressing 'counterexpectation' is a specific type of concession that relies on an implied proposition which had raised certain expectations that should now be revised (cf. Elffers, 1992: 67). In (14) the speaker implies that the Conference of Presidents had been expected to postpone a debate, which in the end turned out not to be the case.  Six out of eleven tokens of this function of toch are translated with after all, three have zero correspondence, and one is translated with the prototypical marker of counterexpectation (actually) and another with at any rate. All eight instances in the Dutch translations had no English ST correspondent.

4.1.3
Reminder of Common Ground Another prototypical function of toch is when it acts as a reminder of common ground (cf. Foolen, 2006: 66;Zeevat and Karagjosova, 2009: 136;Hogeweg et al., 2011: 57), as in (15), where the speaker voices the assumption that all MEPs are familiar with one of the founding documents of the European Union.
Surely we are all familiar with the Treaty of Rome. (#196145, Bolkestein) Toch's correspondents are restricted here to four markers that each stress the self-evident nature of the marked proposition: surely (N = 5), after all (N = 2), of course (N = 2), and the modal auxiliary must (N = 1). Half of the tokens with this function are not translated (N = 11). Translators may have considered the available English correspondents too outspoken as compared to toch, which is more multifunctional and hence vaguer and subtler in reminding coparticipants that they should already be aware of a proposition. In the TTs nine out of eleven have zero correspondence, the only overt correspondence being after all (N = 2).

4.1.4
Seeking Support for an Assertion or Opinion A strongly related context is when toch seeks support for an assertion or opinion by appealing to assumed common ground. In this regard, rather than reminding co-participants of common ground, the speaker appears to "check" (Foolen, 2006: 67) whether there is sufficient common ground to conclude that co-participants share the same attitude or opinion as the speaker. In excerpt (16), the MEP assumes a general consensus for his assertion that a football championship does not warrant severe policing measures. Again, surely is the most common correspondent in the English translations (N = 15), joined by of course, at least, definitely and although (N = 1 each).
Only 7 out of 26 tokens (27%) are not translated. This function resembles the 'reminder of common ground' function in its choice of correspondents, yet not in its proportion of zero correspondence. The latter could be related to the speaker's rhetorical goals: in seeking support for an opinion, overt calls on the self-evident nature of an assertion can be helpful in convincing an audience, whereas reminding someone of something could be more face-threatening. Over half of Dutch translations did not have an overt ST correspondent (N = 18, or 56%), but as with the reverse translation direction, surely was by far the most frequent English correspondent for this function (N = 11, or 34%). Others were definitely, so and the modal auxiliary must (N = 1 each).

4.1.5
Justify the Relevance of Upcoming Talk Speakers may feel the need to justify the relevance of upcoming talk, and toch can help them to do so subtly: it invokes the concessive reasoning that in spite of conceivable reservations against bringing up a point, it is still worthwhile doing so. In (17) the need to justify the upcoming talk arises from the observation that it has already been mentioned several times before.
It has been said in this House a number of times, and it is not unimportant. insists and persists. The speaker in (18) insists on receiving an answer to a question she had raised earlier, which in the English translation is phrased with even greater force (at long last). 4.1.7 Disbelief In a wh-question, toch can betray a sense of disbelief (Foolen, 2006: 68), indicating that in spite of its feasibility or conceivability a specific state of affairs has not yet materialised, which the speaker finds surprising, as illustrated in (19) In such instances, toch is most often not translated (N = 11), and alternatively with an emphatic DO structure (N = 2), and with really, the fact is that and to say the least (N = 1 each). Also in the other translation direction zero correspondence rules, as it applies to all nine instances of this function.
4.1.9 Causal Foolen (2006) reports a 'causal' context for toch, indicating that "now that what is stated in the toch-sentence is the case, action X can be carried out as well" (2006: 69). The MEP in (21) asserts that the decisions of the European Parliament's Bureau can readily be made public, since these become public after a while anyway.
That cannot be such a big problem because these Minutes become public after a certain time anyway.
(#7523048, Maij-Weggen) This function appears three times in the Dutch STs, and each of these tokens is translated with anyway; the nine tokens in the Dutch TTs had zero correspondence in all but two cases (which had anyway and seeing as as their respective English sources). (2006) dubs "Flemish toch" (2006: 63), because he considers this particular to Belgian Dutch (as spoken in Flanders, the northern region of Belgium). It can be paraphrased as 'at least': the speaker narrows the scope of a general proposition to a specific context. In a parenthetical marked by toch, the strong sense of urgency mentioned in (22) is narrowed to the outcome of a summit in the Dutch city of Scheveningen: this sense had at least become apparent there. Note that toch here appears in clause-medial position although its more typical position in this function would be clause-final. Foolen (2006), however, also provides such an example, elaborating that "from the context, it is clear that the intended use is the one we have been discussing in this section" (2006: 64), i.e. it passes the paraphrase test. There are only two such instances in the corpus in each of both translation directions, none of which had a correspondent in English.

Toch in Numbers
The unchallenged dominant function of toch is that of expressing 'concession' , accounting for 47 percent in the Dutch STs and 62 percent of those in the Dutch TTs ( Figure 1). It is followed by the support-seeking function in 12.5 and 16 percent of tokens, respectively, and is a reminder of common ground in 10.5 and 5.5 percent of cases. All other functions are quantitatively rather more marginal. Table 2 lists all correspondents retrieved for toch in both the English TTs and the English STs, ranked in descending order of overall frequency. In almost four in ten cases toch was not translated by an English correspondent at all, and in more than half of the cases it was added by the translator without an explicit trigger in the English ST. Unsurprisingly, contrastive and concessive markers (nevertheless, nonetheless, despite, however, yet, although) account for  28 percent of all translation choices in English TTs (N = 56), which even rises to 46 percent if zero correspondences are disregarded. A similar picture emerges for the English STs, but the ranking looks significantly different, with nevertheless/nonetheless commonly featuring in translations but not in sources, whereas the reverse holds for yet (as for however, albeit to a lesser extent).
Other types of items particularly mark a proposition as self-evident (e.g. surely, after all). If correspondents are mapped against toch's functions (see Appendix 1 for a full overview), some patterns emerge reflecting toch's functional diversity. Concessive and contrastive markers (nevertheless, nonetheless, still, despite this, however, yet, but, but still and having said this/that) are exclusively used to express 'concession' and anyway is restricted to the 'causal' function, whereas surely and of course are shared by 'seeking support' and 'reminder of common ground' , and after all by the latter and 'counterexpectation.' English does not seem to have any marker that reconciles the various functions that toch does. As a result, translators did not provide a corresponding form for 39 percent of tokens (N = 78) in English TTs. For 'concession' and 'seek support' (the two largest functional categories), as well as for 'counterexpectation' and 'causal,' this rate drops below 30 percent, but for all other categories it ranges between 52 and 100 percent. Even more interestingly, translators added toch in Dutch TTs in 52 percent of cases (N = 104) without an explicit trigger in the English ST. Again, this rate is smaller for the large 'concessive' function (38%) but considerably higher for all other functions.

5
Analysis of Eigenlijk

Functions of Eigenlijk
The nine functions eigenlijk fulfils in the corpus picture the marker as one that combines contrast with reality and factuality (Table 3). When it expresses a(n explicit or implicit) contrast, it emphasises that certain beliefs are at odds with (the speaker's view of) reality, which can be extended to a juxtaposition of a current state of affairs with a more desired, or improved version of reality. There are also uses where eigenlijk functions more in the textual domain of elaboration, while preserving some sort of contrastive element, e.g. by contrasting how an issue is presented with what the speaker believes it boils down to ('marking the essential'), adding a comment marking the proposition in which it is embedded as more precise ('noteworthy comment'), reformulating a proposition, etc.

5.1.1
Explicit Contrast According to van Bergen et al. (2011), eigenlijk always requires "a prior context as well as a contrastive environment to be used felicitously" (2011: 3890). The opposition can hold between a proposition explicitly present in the prior co-text (Mortier and Degand, 2009: 354) and the eigenlijk-marked proposition that the speaker presents as closer to the facts. In (23) the proposition marked by eigenlijk asserts that women are victims of human trafficking rather than culprits as suggested in the prior proposition. Interestingly, the English  Mortier and Degand (2009: 354) also distinguish an implicit type of opposition, when eigenlijk on its own marks a proposition as more real or truthful than Table 3 Functions of eigenlijk in the present corpus

Functions of eigenlijk
Contrast -Explicit contrast (between two explicit propositions) -Implicit contrast (between a proposition and an inference) -Improving a state of affairs (contrast of current with envisaged state of affairs) Elaboration -Marking the essential -Mark a noteworthy comment -Metadiscourse -Reformulation -Ask for clarification -Reinforcement

Buysse
Contrastive PragmaticS 4 (2023) 178-212 an inference underlying the prior proposition (rather than that proposition itself). In (24) the proposition marked by eigenlijk (and by actually in the TT) responds to the speaker's inferred assumption that immigration services in their turn assume that the situation in Somalia is safe (and can hence send refugees back to this country).
Maybe we ought to ask the Council to stop sending refugees back to Somalia since the situation there is actually not at all safe for a large proportion of the people.
In English TTs in fact and actually are the most common correspondents for this 'factual' function of eigenlijk, with 8 and 6 tokens respectively, followed by other markers emphasising 'factuality' , i.e. really (N = 5), in actual fact, in reality, the fact is that (N = 1 each), and by after all and strictly speaking (N = 1 each). Nine tokens received zero correspondence. The situation is different in the Dutch TTs, for which actually (N = 12) is the most common trigger, followed by really (N = 7) and a host of markers occurring only once (in fact, rather, just, even, in reality, it is fair to say, so and somehow). Especially the marginal position of in fact is striking here. 45 percent of eigenlijk tokens (N = 22) did not have a ST correspondent.

5.1.3
Improving a State of Affairs The conveyed contrast can also be targeted at improving the current state of affairs when the proposition that eigenlijk marks pitches an implicit or explicit situation or action against one the speaker considers more desirable but that has not yet materialised. In excerpt (25) the speaker has just critically evaluated a colleague's report and now suggests that something be added to it in a next phase. The translator has retained the contrastive element indicated by Dutch echter ('however') as well as the conditional deontic modality expressed by zou moeten ('should' or 'would have to'), but leaves out the additional subjective perspective that eigenlijk brings to the ST utterance. In six out of fifteen instances eigenlijk, aimed at improving a state of affairs, was not translated explicitly. Its correspondents in the English TTs are: in fact (N = 4), really (N = 3), actually and in actual fact (N = 1 each). The Dutch translations were only triggered by ST correspondents in four instances, involving four markers (really, in fact, rather and even though), leaving eleven out of fifteen tokens without a correspondent.

5.1.4
Marking the Essential Mortier and Degand (2009) observe a 'reformulation' function of eigenlijk when it "emphasize[s] something which is deemed important" (2009: 357), shifting focus to more important informational elements of the discourse. In (26) the MEP has been discussing the virtues of a proposed EU directive on pensions, which culminates in his assessment that this amounts to a constitution for occupational retirement schemes.

(26) Eigenlijk kun je zeggen dat dit een soort grondwet voor bedrijfspen sioenvoorzieningen is in Europa.
You can actually describe this as a sort of constitution for occupational retirement provision in Europe.
(#12744167, Pronk) Although this category would not appear to belong to the core functions of eigenlijk, it is the largest in the corpus, accounting for 40.5 percent (N = 81) of this marker's tokens. This is probably due to the setting of the European Parliament, where MEPs continuously proffer assessments of proposals and situations, which often includes pointing out that which they believe is essential or what a proposal or situation boils down to. Again, the most frequent TT correspondents are actually (N = 23) and in fact (N = 21), followed at a long distance by in actual fact (N = 5), really (N = 4), surely (N = 2), and a host of markers conveying 'essentiality,' i.e. at the end of the day, after all, essentially, if truth be told, in a way, in effect, the truth is that and ultimately (N = 1 each). In 18 instances eigenlijk was met with zero correspondence in the TT. Also for this function the picture looks somewhat different in the other direction. Really is beyond any doubt the preferred trigger for eigenlijk (N = 14), whereas in fact is entirely absent in the correspondent list, containing basically (N = 5), indeed (N = 2), rather, just, essentially, perhaps, to put it simply and virtually (N = 1 each). As with the previous functions, the proportion of zero correspondence is higher (N = 20, or 41%). Noteworthy Comment Eigenlijk can also be used in an elaborative context when it highlights a noteworthy comment. This use appears to be absent in previous analyses of eigenlijk, but has been pointed out for the English markers actually and in fact (Aijmer andSimon-Vandenbergen, 2004: 1797;Aijmer, 2013: 85,113;Buysse, 2020: 34). In a parenthetical, the speaker in (27) comments that a proposal had already been mentioned by the rapporteur of the committee that had discussed the directive under debate.
(27) Dat houdt wat onze fractie betreft wel in, en dan spreek ik namens de hele fractie, dat een aantal punten er niet in hoeft, dat had de rapporteur eigenlijk al impliciet vermeld, laten we dat opschonen en niet in de richtlijn laten staan. This does imply for our Group -and I am speaking on behalf of the entire group -that a number of points are redundant, something which, in fact, the rapporteur has already hinted at.
(#1812203, Oostlander) By far the most common English TT correspondent for eigenlijk in the 'comment' function is in fact (N = 12), which is joined by actually, in actual fact, albeit and as a matter of interest (N = 1 each), apart from five zero correspondences. Also in English STs in fact (N = 2) occurs as well as really (N = 1) and three zero correspondences.
5.1.6 Metadiscourse In 13 instances eigenlijk is used for a metadiscursive utterance, indicating an element in the discourse that the speaker feels the hearer may find unexpected. The speaker in excerpt (28) announces that he will divert from common parliamentary practice by posing only one question in the debate. Actually is eigenlijk's translation correspondent in 4 TT instances, really in 3, and in fact and just in 1 each. Four tokens were not translated explicitly. The only overt marker used in English STs is really (N = 5); the remaining nine were zero correspondences.
Contrastive PragmaticS 4 (2023) 178-212 5.1.7 Reformulation As Mortier and Degand (2009) point out, eigenlijk can be used for reformulation to express "a (slight) deviation from an original proposal " (2009: 356). In (29) the MEP first situates a decision on the day of the parliamentary debate (or perhaps simply in the present, which can also be expressed by vandaag 'today'), and then reformulates this to refer to the exact moment when the decision is due to be taken (viz. morgen, 'tomorrow').
Today -or rather tomorrow, in fact -we face a very tricky decision. (#22245968, Corbey) All ST tokens of eigenlijk in the 'reformulation' category were translated with an overt English correspondent, four with in fact, and one each with really and rather. The Dutch translations had indeed (N = 3) and rather (N = 1) as their overt sources, in addition to one zero correspondence.

5.1.8
Ask for Clarification In interrogative sentences, eigenlijk can mark questions for clarification, at which point it expresses some level of annoyance. In (30) the speaker sees plenty of reason for the EU to take a stand against Algerian authorities on human rights issues, and raises the question why nothing is being done in this respect. Eigenlijk highlights that the speaker considers the lack of action unexpected.
Maar waarom gebeurt dat eigenlijk niet? Plenty of reason, therefore, to take the Algerian authorities to task about this. But why, in fact, is this not being done? (#4404758, Van den Bos) In three instances eigenlijk is translated into English with actually and in one each with in fact and just, whereas four tokens remain without a translation correspondent. Dutch translations were triggered by in the first place (N = 5), actually (N = 3), really and in fact (N = 2 each), and just (N = 1), whereas ten instances had no overt ST correspondent.

Buysse
Contrastive PragmaticS 4 (2023) 178-212 5.1.9 Reinforcement Finally, five instances were identified of what can be seen as reinforcement of a prior proposition (Buysse, 2020). This is similar to what Mortier and Degand (2009: 357) call the "enhancement" function of eigenlijk, and it closely resembles a use of in fact (yet not of eigenlijk) described as a "strong commitment to the proposition" (Schwenter and Traugott, 2000: 12). Interestingly, this function was only attested in the Dutch TTs, and all of these were sparked by the same English marker: in fact. In (31) the speaker first negatively assesses a specific proposal regarding VAT, and then moves on -marked by in fact (and by eigenlijk in the TT) -to radically denounce the VAT system altogether.
(31) I do not see the introduction of a common system of VAT as a priority and I do not agree that the introduction of a common system is long overdue. In fact, nothing would please me, my party and the bulk of business people in the United Kingdom more than the complete abolition of VAT.

Eigenlijk in Numbers
In the corpus sample studied here eigenlijk is especially used to 'mark the essential' , which is certainly the case in the Dutch STs, where it accounts for 40.5% of the tokens (Figure 2). In Dutch TTs this use shares the first place with the 'factual' function (24% each), whereas the 'contrastive' function is third in both translation directions (11% in Dutch STs, 17% in TTs). The remaining functions are considerably less common, and one ('reinforcement') only occurs in Dutch TTs.
As Table 4 shows, by far the most common translation correspondents for eigenlijk in the English TTs are in fact and actually, which together account for just under half of all tokens of eigenlijk (N = 98 or 49%). Most other forms also highlight 'factuality' (e.g. really, in actual fact, if truth be told, the fact/truth is that). One in four tokens is not matched with an English TT form, which is considerably lower than for toch (viz. 39%).
The top three correspondents is the same for STs, but the order is completely reversed: really sparks one in five eigenlijk tokens, followed by actually Contrastive PragmaticS 4 (2023) 178-212 (11.5%) and in fact (6.5%). If zero correspondences are removed from the equation, the dominant positions of really and in fact in sources and translations, respectively, become even clearer: really holds a 35.1% share of all corresponding forms in English STs, and in fact a 37.6% share in English TTs.    In fact is used at least once in every function that was attested in the corpus (Appendix 2), and actually and really appear in all but two functions ('reformulation' and 'reinforcement'). Due to the dominance of these three markers, there is a long list of, typically more specific, markers occurring only once or twice. By and large, these express 'factuality' (e.g. the fact/truth is that, in reality, if truth be told) or 'essentiality' (e.g. basically, essentially, to put it simply, virtually, at the end of the day, ultimately). In the English STs, in the first place stands out as the most common correspondent in the 'asking for clarification' category, which manages to capture some of the irritation eigenlijk conveys in such interrogatives.
English correspondents of toch are largely exclusive to one of either of the marker's main uses ('concession' and 'common ground'), suggesting that no single marker in English captures the functional diversity of toch. This, in its turn, can be taken as evidence that toch ranks highly in the list of semantically underspecified markers. The highest correspondence rates for toch were noted with nevertheless/nonetheless and yet in 'concessive' functions, and for surely -"a positioning device for redefining common ground between individuals" (Downing, 2006: 39) -in functions that relate to common ground.
Particularly striking are the zero correspondence rates. Given the semantic and syntactic optionality of discourse markers, they could be regarded as ready candidates for omission in translations, especially when they have no formal correspondent in the other language (Aijmer and Altenberg, 2002: 29). Toch is indeed not explicitly translated in 39 percent of cases. Although some of these instances seem related to the presence of a similar device in the TT (e.g. another concessive marker), there is also a functional difference, in that concessive relations receive an overt translation considerably more often than others. Possibly, functions that mostly contribute to the coherence of the discourse are less prone to omission than those that contribute to interpersonal issues. What is even more noteworthy is the very high number of toch tokens -across functions -that was included in Dutch TTs without an explicit correspondent in the ST. For eigenlijk, too, translators have clearly elected to provide correspondents in a large majority of cases in English TTs, and added the marker in Dutch TTs not forced by an ST correspondent.
In this light, eigenlijk and, particularly, toch do not appear to fit the picture of traditional explanatory frameworks within Translation Studies. On the one hand, the high degree of TT tokens not triggered by a ST marker appears to confirm the tendency that TTs are more explicit in their language use (cf. Blum-Kulka's (2000) "explicitation hypothesis"). On the other hand, other tendencies, such as the Unique Items Hypothesis (Tirkkonen-Condit, 2004), would suggest that if the TL contains an item for which the SL does not have a ready equivalent, this item will occur less frequently in translated texts in the TL than in this language in general. In the current sample, this does not hold for toch: even though it lacks a clear English equivalent, toch occurs even more frequently in the Dutch translations than in the Dutch source texts. In general, it would appear that translators are strongly conscious of implicit pragmatic signals and of the pragmalinguistic potential of toch (as they are of eigenlijk) to convey these in Dutch. Such observations call for further studies at the interface between Contrastive Pragmatics and Translation Studies that also give greater prominence to the latter than was feasible within the current study.
Contrastive PragmaticS 4 (2023)  Although the period in which the data were collected and the size of the corpus should guarantee that no single translator could have had a dominant hand in the translations, translator preferences cannot be ruled out. Especially in an institutional setting, the community of practice within which translations are carried out could have a bias towards certain linguistic phenomena, such as a preference for more formal discourse markers like nonetheless. The many additions of eigenlijk, and particularly toch, on the other hand, suggest that EP translators also try to capture oral discourse features.
In the present study the Dutch markers formed the focal point, but further research could investigate the mutual correspondence (cf. Altenberg, 1999) of some of the markers that have surfaced. An in-depth corpus analysis of, on the one hand, the Dutch correspondents of actually, in fact and really and on the other hand, the English correspondents of eigenlijk, in feite and werkelijk could be revealing in this respect. The scope could also be extended to other languages, such as German, for which toch's cognate doch has already received quite some scholarly attention but a systematic contrastive corpus analysis between the Dutch and German marker is still lacking. Finally, other modes or registers including e.g. written (fictional) texts, or output from interpreting could be included.