Conditional clauses in novice academic English: A comparison of Norwegian learners and native speakers

This study concerns the syntactic, semantic and pragmatic features of conditional clauses in novice academic English. The material comes from the VESPA corpus, representing Norwegian advanced learners of English, and the BAWE corpus, representing English L1 students in British universities. The learners are shown to overuse conditionals in general, but to mostly master their syntactic and semantic features. The overuse may be associated with the interpersonal functions of conditionals. The epistemic use in argument building is more apparent among the native speakers.


Introduction
This article studies conditional clauses introduced by if and unless in two corpora of novice academic English.The investigation includes a comparison between English as a first and a second language.The conditionals are analysed for syntactic, semantic and pragmatic features, i.e. syntactic position of the conditional clause, verb forms used in the conditional and its matrix, and whether the conditional is open, hypothetical or pragmatic (see below for definitions).The aim is to survey how conditionals are used and whether there any differences between native and non-native usage of conditional clauses in (novice) academic texts.As a corollary, I will also discuss any learner problems appearing in the material.
According to Declerck & Reed (2001: 9), conditional constructions are two-clause structures "in which one of the clauses is introduced by if [...] or by a word or phrase that has a meaning similar to if".The construction relates two situations to each other in such a way that one clause contains a condition for some aspect of the situation referred to by the other clause.Conditional constructions can be paratactic (e.g.Do that and I'll beat you), but are typically hypotactic (ibid: 8).In a hypotactic construction the conditional can be headed by a conjunction, as in (1), or it can be marked by subject-operator inversion, as in (2).The present study includes only the conjunction-headed type, mostly due to the difficulty of retrieving the inverted type in an untagged corpus. (1) If the viewer loses interest in the character, the story itself will be pointless.
(VESPA)1 (2) I do not feel for example, that the poem would have been as effective had she initially applauded her work.(BAWE) Gabrielatos (2013) argues that not all types of conditionals are well explained in coursebooks and learner grammars, which prompts him to hypothesize over-or under-representation of certain patterns in learner English.His investigation, based on the International Corpus of Learner English compared to the British National Corpus (BNC), produced inconclusive results, and Gabrielatos (ibid.: 164) recommends that future studies should take the learners' L1 background into account, and that the L1 reference corpus should be more similar to the learner corpus.This study follows both of his recommendations.

Material and method
The material for the investigation comes from two sources: the Varieties of English for Specific Purposes dAtabase (VESPA), which contains discipline-specific academic texts written by advanced learners of English, and the British Academic Written English corpus (BAWE), which similarly contains discipline-specific academic texts written by British university students.From both corpora, I selected the English literature discipline and limited the material to texts written by students whose first language was Norwegian or English, respectively.Both corpora include BA-and MA-level texts (with a predominance of undergraduate texts), but this distinction was not counted as a variable in the present study.The selections differ in size, with 100,073 words from VESPA and 292,021 from BAWE.
VESPA was accessed by means of WordSmith Tools 6 (Scott 2012), and BAWE via SketchEngine (https://the.sketchengine.co.uk).Searches were made for the items if and unless.The resulting concordances were reviewed manually to exclude the following: if introducing a nominal clause, if as part of a complex conjunction (as if, even if), if as a misprint for of, if introducing a phrase (e.g. if any), and subordinate clauses not attached to a matrix clause.Furthermore, I excluded quoted material from literary texts and secondary sources.The VESPA texts have been tagged in such a way that WordSmith searches can ignore quoted material (see Paquot et al 2013: 379), and for BAWE this was done manually by checking the context.
In the spirit of Contrastive Interlanguage Analysis (Granger 1996, Hasselgård & Johansson 2011) this study involves comparison between first-and second-language use.The native speaker corpus (BAWE) is not seen as a target norm-the BAWE writers too are novice academic writers-but as a background against which it is instructive to view the writing of the advanced learners of English represented in VESPA.Dancygier & Sweetser (2005: 4) state that "there is something about if which engages the curiosity of the analyst".Indeed, conditionals have been thoroughly examined from a variety of angles, including logic, semantics, syntax and pragmatics.In this short article, I cannot give a fair representation of previous research, but see Traugott et al. (1986), Athanasiadou & Dirven (1997), Dancygier & Sweetser (2005) and Gabrielatos (2010) for relevant overviews.This section will outline the features of conditionals that are analysed below, with frequent reference to previous studies.

Syntax
A conditional clause functions as an adverbial in its matrix clause, in which it can occur at the beginning or end, and occasionally in the middle.The positions are referred to as initial (before the matrix), medial (within the matrix; i.e. between its subject and a last obligatory argument) and end position (after the matrix), following the analysis of Biber et al. (1999) and Hasselgård (2010).Examples are given in (3)-( 5).The same positions are used in Norwegian (Hasselgård 2014), which may be helpful to Norwegian learners. (3) If you have a book with detailed descriptions, and endless events, you cannot include everything in a two-three hour movie.(VESPA) (4) The individual, if Foucault is right, is left to act alone in a bleak world that has become incommunicative, de-stabilised and seemingly meaningless… (BAWE) (5) Clearly she would never have been able to marry John if she had come forward.(VESPA) In contrast to other types of adverbial clauses, conditional clauses prefer initial position (cf.Diessel 2005: 454;Hasselgård 2010: 89).According to Diesel (2005: 461) this is because conditionals "describe a contrastive situation that establishes a specific framework-a specific semantic constellation-for the discourse that follows".Similarly, Ford argues that "the use of if-clauses in initial position has to do with the fact that their meaning lends itself inherently to discourse organizational work.Ifclauses, at the content level alone, function primarily to limit the framework of interpretation for an associated main clause" (1993: 133).Diessel suggests that the use of end position for conditionals requires special motivation, e.g. if it is a kind of afterthought or for some reason "does not affect the semantic interpretation of the main clause" (2005: 463).Hasselgård (2014: 193) found a certain association between placement and semantic type of conditional: while initial position is generally the most frequent option, a higher percentage of hypothetical than of open conditionals occurs in end position.Conditional clauses may be attached to different types of matrix clause.Examples (3)-(5) all have a declarative main clause as the matrix.Example (2) above shows a that-clause containing a conditional, and the present material also contains examples of interrogative and imperative main clauses, as well as several types of finite and non-finite subordinate clauses as matrixes for conditionals.See further Section 4.1.

Verb forms in conditionals
The prototypical combinations of verb forms in a conditional construction, often found in pedagogical grammars, are the following (cf.Römer 2007: 357): Type 1: if + simple present à will + inf.; Type 2: if + simple past à would + inf.; and Type 3: if + past perfect à would + perfect inf.Römer (ibid.)describes these combinations as the "teaching norm".To this set is sometimes added a "Zero conditional" with the combination if + simple present à simple present (Gabrielatos 2013: 156).Carter & McCarthy (2006: 747) note that modals other than will/would are used, as well as different future-referring expressions, including the simple present.
In general, past tense forms in the if-clause are associated with a greater degree of hypotheticality (Comrie 1986: 94), e.g.If he comes, I'll run away vs.If he came, I would run away.It has also been noted that the Zero conditional is used for habitual situations (ibid.)and for generic meanings (Dancygier & Sweetser 2005: 95), as illustrated by ( 6), from Dancygier & Sweetser (ibid.).( 6) In fact if you bring any reading matter they confiscate it.
Non-corpus-based accounts of conditionals are often restricted to the prototypical verb-form combinations, while usage-based accounts display a much higher degree of variation (e.g.Römer 2007).Reporting on a study of conditionals in the BNC, Römer states that "the usage norm allows for a variety of tense form sequences in if-clauses […] with combinations of simple present + simple present being the most frequent type, followed by Type 1 conditionals" (2007: 359).Studying tense forms in conditionals in two corpora of EFL writing by German advanced learners, Römer discovers no clear mirroring of the teaching norm, but finds that the learners, despite being advanced, "seem to be confused about the choice of appropriate tense-form sequences in conditionals" (ibid: 360).

Semantic types of conditionals
Conditional constructions are usually described in terms of a protasis (the conditional adjunct) and an apodosis (the matrix clause); see e.g.Comrie (1986), Huddleston and Pullum (2002: 738).Conditionals are generally divided into open and hypothetical, depending on whether the condition in the protasis can be fulfilled or not (e.g.Biber et al. 1999: 819).2According to Huddleston and Pullum, open if-conditionals ('If P, then Q') exclude the possibility of a true P and a false Q, and also imply that Q is a consequence of P (2002: 739).This cause-effect relationship (ibid.)can be direct or inferred.A hypothetical conditional, on the other hand, "differs from the open in that it entertains the condition as being satisfied in a world which is potentially different from the actual world" (ibid: 748).Open conditionals are exemplified by ( 1), ( 3) and ( 4) above, while ( 2) and ( 5) illustrate hypothetical conditionals.
As pointed out by a number of scholars (e.g.Comrie 1986;Sweetser 1990;1996), not all conditional adjuncts fit neatly into this distinction.The protasis may "specify the condition under which the speaker makes the utterance" (Quirk et al. 1985(Quirk et al. : 1096)); i.e. it does not specify a condition under which the proposition in the matrix clause is true, but rather a condition under which the proposition or proposal in the matrix clause applies.These were called "proposition-related" in Hasselgård (2010: 229) and correspond roughly to Sweetser's (1990: 128) epistemic conditionals.An example is given in ( 7), where the protasis specifies a necessary condition, but no cause, for the validity and relevance of the proposition.Another type specifies the condition for making a speech act, as illustrated by ( 8).These are called speech-act conditionals following Comrie (1986: 81) and Sweetser (e.g.1990: 123). 3Epistemic and speech-act conditionals are similar in that both types express "conditions on the relevance of the main proposition" (Haegeman 1984: 500).Gabrielatos (2013) uses the term "indirect conditionals" for both types, following Quirk et al. (1985Quirk et al. ( : 1089)) If you look up the word in the Oxford Dictionary of English, its meaning will be defined as having or showing compassion or benevolence.(VESPA)

Pragmatic functions of conditionals
The pragmatic functions of conditionals have already been touched upon in the above discussion of semantic types.However, some studies of conditionals, e.g.Ford (1996), Carter-Thomas (2007) and Warchal (2010) make further distinctions of pragmatic functions in specialized contexts (conversation, medical discourse and research articles, respectively).Of particular interest are the interpersonal functions that can be ascribed to epistemic and speech-act conditionals.Warchal (2010: 142) argues that conditionals "can have a potential interpersonal function aimed at establishing agreement between the writer and the reader of an academic text".Examples ( 8) and ( 9) illustrate a type of "rhetorical" conditional which allows "the writer to share with the receiver the responsibility for the claim made in the main clause" (ibid.: 147).That is, the reader is included or addressed in the conditional and thus invited (or expected) to concur with the proposition in the matrix clause.Carter-Thomas (2007: 150) discusses the potential of conditional constructions to act as "polyphonic operators".She argues that the "nonassertive value of the operator, if […] leaves a space open for the reader to question or reject" the proposition of the Q-clause (ibid.: 159).An example is given in (10).The label "refocusing conditionals" is assigned to "those occurrences which have a marked argumentative function" (ibid.: 160), as in (11).The present study will not include a detailed analysis of pragmatic functions, but will use insights from the cited studies in the discussion of individual examples.

Corpus investigation
This section presents the corpus analysis of conditionals, discussing both quantaitive and qualitative features while comparing learner and native speaker usage.

Frequencies and structural properties of conditionals
Table 1 shows that the two corpora contain similar numbers of conditionals.However, relative to corpus size (cf.Section 2), the conditional clauses under study are more than three times as frequent in VESPA as in BAWE, with 124.9 and 39.4 occurrences per 100,000 words, respectively.The learners' overuse of conditionals may be related to the pragmatic functions of the constructions, but may also reflect an influence from conversational English, where conditionals are more frequent than in academic prose (Biber et al. 1999: 820).The proportional distribution of conjunctions is similar between the corpora: unless-clauses are marginal, and if-clauses dominate.The use of unless is illustrated in (12).Due to their low frequency, they will not be given particular attention in this study.If-clauses co-occur with a then-clause more frequently in BAWE than in VESPA.As (13) suggests, the if-then construction tends to convey explicit argumentation.The less frequent use of if-then clauses in VESPA is surprising, both because of its explicitness, which should in principle appeal to learners, and because it should be familiar to them as Norwegian has an equivalent combination (hvis-så).
The majority of conditionals occur with main clause matrixes, as shown in Table 2.There is little difference between the corpora.A slightly higher percentage of the matrixes in BAWE are subordinate clauses, which might indicate a higher degree of syntactic complexity.However, the difference is not statistically significant according to a chi square test comparing subordinate and main clauses (p=0.5922).

Positions of conditionals
As Table 3 shows, the placement of the conditionals is similar between the corpora, with initial position being most frequent, and medial position least frequent. 4While initial position is dominant for conditionals in general, it is the only available position in the if-then structure.The unless-clauses in the material are all in end position, as in ( 12).Although such clauses in initial position are not hard to come by, a spot check in the BNC suggests that unless-clauses are more apt than ifclauses to take end position.Positional freedom seems to correlate with matrix clause type.There is practically no difference between the corpora in this respect.It is only in main declarative matrixes that all three positions are used for conditionals, and initial position is by far most common, with 70% in BAWE and 78% in VESPA.In interrogatives, all but one conditional (in VESPA) are initial.In subordinate matrixes, however, only that-clauses and adverbial clauses seem to offer initial position as a real alternative, though even here, end position is used slightly more often than initial position.In all other subordinate matrix clauses, end position predominates.Medial position is generally rare for long adverbials (Hasselgård 2010: 107), and occurs only with parenthetically inserted conditionals, as in (4) above and ( 16) below.Ford & Thompson (1986: 359) found that relative matrixes favoured end position of the conditional.While they tried to find discursive motivations for this, I believe the reason is purely syntactic.In this clause type initial position is is not available for the conditional because it is obligatorily taken by the relative pronoun (even in its zero realization).An example is given in (15).Medial conditionals, however, are found in relative clauses, in which case the conditional is bracketed by punctuation, as in ( 16).
(15) We have to ask ourselves if child murder is something we all are capable of if we are pushed as far as Sethe (or Margaret)?(VESPA) (16) Once we look at Merry we might conclude that the Swede failed.Which, if we look at the society during the 1960s, was something many Americans felt they did.(VESPA) Non-finite matrixes invariably have the conditional in end position in the present material, thus supporting Ford & Thompson's observation (ibid.) that infinitive clauses favour this pattern.Again the main reason may be the rather restricted word order flexibility of most types of subordinate clauses cf.Givón (1979: 83).

Semantic types of conditionals
The analysis of semantic types of conditionals involves some degree of subjective interpretation, thus other analysts might have analysed some of the clauses differently.While I was unable to provide a second opinion, the analysis should be indicative of the distribution of the semantic types in the two corpora.As Table 4 shows, the two writer groups have roughly equal proportions of open and speech-act conditionals, but the learners have more hypothetical conditionals, and the native speakers have more epistemic conditionals.5It is not immediately clear why this should be so.Hasselgård (2014: 193) found a lower proportion of both hypothetical and epistemic conditionals in Norwegian than in English, so the learner behaviour is not likely to be L1-induced.6Note that the material has been categorised on the basis of meaning and context, not formal features.
Thus, examples ( 17) and ( 18) have both been analysed as hypothetical, despite not having the prototypical verb forms (past perfect + would have -en).The events reported in (17) clearly did not take place "in the present day", thus the construction must be hypothetical.Similarly the then-clause in ( 18) is counterfactual; hence the construction must be hypothetical.
(17) The victim, however, would in the present day have the law on her side, if she found the strenght to fight and leave her violent relationship.(VESPA) (18) In effect, if Foucault's theory were true, then we would not be able to read his own essay and understand it so coherently.(BAWE) (19) If we accept stereotypical definitions of male speech, this masculine core may be why her discourse is characterised by eloquence.(BAWE) Epistemic conditionals often appear as part of argument building as in ( 19).The conditional gives the premise for the argument in the matrix clause.This type of conditional also occurs relatively often with interrogative matrixes in examples such as ( 14) above.Example ( 19) has a first-person pronoun in the conditional, which is another common feature of the epistemic conditionals; cf.Section 4.5.Figure 1 correlates the semantic types of conditionals with placement.Speech-act conditionals have been left out of the figure since they are so few.The columns show the proportional distribution across positions for each semantic type in both corpora.It must be recalled that the numbers are low in some of the categories, cf.Table 4.Still, Figure 1 indicates that the Norwegian learners apply the same principles for placement regardless of the semantic type of conditional.The BAWE writers appear more likely to use end position for hypothetical conditionals, though this is one the categories where numbers are low and percentages may be deceptive.Medial position contains only epistemic conditionals in BAWE in contrast to VESPA, but again the numbers are too low to suggest a trend.

Verb forms in conditional constructions
Figure 2 shows that the use of verb forms in conditional clauses does not differ much between the learners and the native speakers.In both corpora the present tense dominates, followed by the past tense.Note that both categories contain cases of present and past with perfect and progressive aspect.Modalised verb phrases are rare in the conditional clauses, as are nonfinite verb forms.Figure 3 shows the use of verb forms in the matrix clauses of the conditionals, revealing slightly greater differences between the corpora than was the case for conditional clauses in Figure 2. The most striking difference is the more frequent use of will/would in VESPA, which might be related to the "teaching norm" (Römer 2007), but also to the greater frequency of hypothetical conditionals in VESPA.Among the "other modals", can is most frequent in both corpora: twice as frequent as the second most frequent, which is could in BAWE and might in VESPA.
The high frequency of the simple present in both conditionals and matrixes reflects a high number of zero conditionals (present + present combinations): 41 (32.8%) in VESPA and 40 (34.8%) in BAWE, which is high compared to Römer's (2007: 358) frequencies for any of the corpora she investigated.The zero conditionals are open, epistemic or speech-act conditionals, i.e. they fulfil all the semantic functions except hypothetical.The pattern is similar between the corpora.

Pragmatic functions of conditional constructions
As noted in Section 4.1, a major difference between the corpora is what seems to be a massive overuse of conditionals among the learners compared to the native speakers.It is reasonable to link the overuse to the interpersonal functions of conditionals (Carter-Thomas 2007;Warchal 2010).Previous studies of other interpersonal features, e.g.Fossan (2011), Paquot et al. (2013), Hasselgård (2009;2015) have shown that Norwegian learners are highly visible authors and favour an interactive style of writing.(In this light it may be surprising that they do not overuse epistemic and speech-act conditionals; however, the learners' frequencies of pragmatic conditionals mirrors a difference between English and Norwegian reported in Hasselgård (2014: 192).)Space does not permit an exhaustive analysis of the pragmatic functions of conditional clauses.However, I will briefly comment on some features of reader reference in the conditional.First, interrogatives with conditionals are more frequent in VESPA than in BAWE, which may be related to a generally higher occurrence of questions (Paquot et al. 2013: 384).
(20) If the main argument of the book is to be mindful of the slippery slope, why does the post-pandemic scenario seem preferable to the pre-pandemic scenario?(VESPA) However, the readers of these texts are addressed not only by means of questions, but also by pronominal reference, cf. ( 16) and ( 19).The learners and the native speakers use we in conditionals with almost the same frequency.But in addition the learners frequently use if one (rare in BAWE) as well as if you (absent from BAWE), see ( 21) and ( 22).Although the reference of you and one is generic rather than personal, the pronouns, like the inclusive we illustrated in (19), draw in the readers and allow them to share "the responsibility for the claim made in the main clause" (Warchal 2010: 147), and also to take part in-and agree with-the argument.
A less interactive use of conditionals, the if-then structure, is more popular with the native speakers (cf.Table 1).This is the refocusing structure used in conditional constructions with a "marked argumentative structure" (Carter-Thomas 2007: 160).As noted in section 4.3, the higher frequency of epistemic conditionals among native speakers can probably also be related to argumentative uses.

Conditionals and learner problems
There are more similarities than differences between the corpora, indicating that the learners have few problems with the use of conditionals.However, one particular area of difficulty was identified during the analysis: the use of verb forms in hypothetical conditionals.In ( 23), the matrix seems to require a perfective would have changed to match the past perfect in the conditional.Example (24) presents a more serious problem with verb forms.Though not widespread, the problem of using the hypothetical would in the conditional was found in several texts.The problem in ( 25) is that the present tense forms conceal the fact-evident in the context-that the situation described is counterfactual.Thus, the use of verb forms in hypothetical conditionals remains a problem for some of the learners in VESPA, even at this advanced level of proficiency.

Concluding remarks
The study has shown that Norwegian learners are not only prolific, but also quite proficient, users of conditional clauses.In general, there are more similarities than differences between the two writer groups, with the major exception of the frequency with which conditionals are used.Section 4 has shown that both syntactic and semantic features of conditionals have similar proportional distributrions across the material.The learners appear to master the expression of all semantic types of conditionals as well as their placement and their integration in a range of matrix clause types.Section 4.3, however, noted a certain overrepresentation of hypothetical conditionals with a corresponding underrepresentation of epistemic ones.Both groups use conditionals for expressing arguments, but this is more widespread in BAWE.
The learner data indicated problems with verb forms in hypothetical conditionals.It might be useful for learners to study hypothetical contexts in order to learn not only how to use verb forms correctly but also to find other means of expressing hypothetical meaning.The learners' overuse of conditionals is believed to be related to their interpersonal uses.A closer look at the pragmatic functions of the constructions may shed more light on this issue.
A limitation of this study is its restriction to conditionals marked by if and unless.Other studies report that alternative realizations of conditionals are rare, but the possibility remains that they might alter the picture of how conditionals are used in novice academic writing.More importantly, the inclusion of causal and concessive clauses would complement the account of how these writers construct their arguments.

( 10 )
It makes sense then, if the modern present is so different from the past, to use a different quality of form to express our experience of it.(BAWE) (11) If the many characters Neo is helped by along his journey are constructs of the psyche, then it is possible that they represent aspects of the self.(BAWE)

( 12 )
But when the original is a great classic, it might prove prudent not to add to the discourse unless you consider yourself of the same calibre as the original author.(VESPA) (13) If Mrs Alving represents modernism, then Pastor Manders is the voice of Victorian values.(BAWE)

Figure 3 .
Figure 3. Verb form in matrix clauses of conditionals (percentages)

( 21 )
One might fail to see all of this if one is not also critical of the way the Swede lives, … (VESPA) (22) That means that if you are a man you will act and dress the way we are taught from birth men are supposed to do.(VESPA) 7

( 23 )
We cannot know this, but there is a possibility that Dr. Luce would change his diagnosis if Cal had been honest about his sexual desires.(VESPA) (24) In addition, she joined the organization which would have risked her marriage if Jack would know about it.(VESPA) (25) If such changes are made, it can be difficult to give the viewer the same sense of realism which the novel provided.(VESPA)

Table 1 .
Frequencies of conditional clauses and distribution of conjunctions in VESPA and BAWE (raw frequencies and percentages).

Table 2 .
Matrix clause types occurring with conditionals.
Most of the examples given above have a main declarative as matrix, but(2) has a that-clause and (12) has an infinitive clause.Example (14) shows the typical use of conditionals with an interrogative matrix: the conditional appears in initial position and establishes a relevant framework for asking (and interpreting) the question.(14)If African American experience is like a jazz record, what is being done when it hits scratches?(VESPA)

Table 3 .
Positions of conditionals

Table 4 .
Semantic types of conditionals