Suppose you want to tell someone that you believe that the first edition of The Illogician will be a success, provided that the articles in it are well written. You could convey this message in the usual way, by uttering a sentence such as (1).

(1) If the articles are well written, The Illogician will be a success.

However, it would not be strange to instead say (or hear) exactly the same sentence, but in a different order:

(2) The Illogician will be a success, if the articles are well written.

We thus see variation in the clause order of conditional sentences. The canonical order is visible in (1), where the antecedent precedes the consequent, whereas the conditional in (2) is ’non-canonically’ ordered: here, the consequent precedes the antecedent.

The existence of this variation is, I think, quite interesting. After all, nothing changes to the literal meaning of (1) if you for some reason decide to formulate it as (2). Why, then, do we sometimes decide to utter conditionals in a different order, or do we at least find this decision acceptable? What makes us prefer, in certain situations, the non-canonical order over the canonical one?

Connectedness

As with many linguistic phenomena, it is very likely that the answer to this intriguing question is not one-sided. More factors are probably in play, influencing us to prefer one order over the other. One such factor, for instance, could have to do with the different social functions conditionals can carry out. For example, a conditional could be used to make a prediction. We saw an example of such a predictive conditional in (1) and (2): in the consequent, it is mentioned what would happen (The Illogician becoming a success) given the event described in the antecedent takes place (the articles being well written).

A completely different way of using conditionals is illustrated in sentence (3) – an example taken from Austin (1970, 212). What seems to be the case in these often called ‘biscuit conditionals’ is that the antecedent is used to make a comment on the speech act performed in the consequent, in the case of (3) the offering of biscuits.

(3) There are biscuits on the sideboard if you want them.

While predictive and biscuit conditionals are just two instances of the many social functions a conditional can fulfil, they illustrate an important way in which these functions can diverge. Whereas predictive conditionals such as (1) imply a clear causal or temporal connection between their antecedent and consequent, biscuit conditionals such as (3) do not – the biscuits are on the sideboard, whether you want them or not. Let us call this, following Reuneker (2022), a difference in their connectedness.

So, what then would be the prediction regarding clause order preference based on this ‘connectedness’ feature conditionals may or may not possess? When there is a strong connection present between the clauses of a conditional, it is conceivable that this connection incites a desire to place the clauses in a specific order. Take predictive conditionals: since the event described in the antecedent precedes the event of the consequent in a temporal and causal way, the most natural thing to do as a speaker seems to be to order the sentential clauses correspondingly. In fact, research shows that a mismatch between the temporal and clause order of a conditional often results in a response time penalty for listeners (Evans and Newstead 1977).

Conversely, when a conditional lacks connectedness, a preference for one specific clause order becomes much less obvious, if not completely absent. We saw biscuit conditionals as an example of non-connected conditionals: the antecedent merely comments on what is being said in the consequent. One can imagine how it would not matter much where or when this commenting takes place. For instance, (4) and (5) seem equally natural:

(4) If you’re thirsty, there is beer in the fridge. (5) There is beer in the fridge, if you’re thirsty.

Thus, according to the hypothesis described above, conditionals high in connectedness receive a clear preference for one of the clause orders – the canonical in the case of predictive conditionals – while those low in connectedness do not.

Given and new

Connectedness is, however, only one possible factor influencing our clause order preferences. A very different one discussed in the literature looks at the information structure of the discourse in which the conditional is uttered, and more specifically at the so-called given-new principle. According to this principle, we prefer a sentence to first present information that is given in or inferable from the context, if it includes any. Only after this would we like the sentence to introduce information that is completely new. For example, if we are talking about how I struggled for the exam of Modal Logic, and I would like to say to you that 1) I almost failed the exam but that 2) Mary easily passed, the preferred thing for me to do is to utter those two propositions in that order, not the other way around. The given-new principle has been very successful in explaining how discourses are structured and has received a lot of empirical support (see e.g. Clark and Haviland 1977; Haviland and Clark 1974).

It is hopefully easy to see how this principle can be applied to the topic of discussion. If a conditional is uttered in a discourse and the information in the antecedent relates much more directly to what has been said before than the information in the consequent, the given-new principle makes the prediction that we would prefer the canonical order. If the roles are reversed and the consequent, not the antecedent, contains the information that connects the conditional to the previous context, we would expect to observe a stronger preference for the non-canonical order.

Of course, there are many more factors one could think of that possibly influence our preferences for conditional clause ordering. For instance, the order seems to depend not only on the relative amount of words present in the antecedent or the consequent (see e.g. Diessel 2005; Ford and Thompson 1986), but also on the mode (spoken vs written) in which the conditional appears (Reuneker 2020).

An experimental investigation

In an experimental study I carried out with Lotte Hogeweg (Radboud University), we put both the connectedness and the given-new hypothesis to the test (Van der Leer and Hogeweg 2024). We presented Dutch predictive and biscuit conditionals embedded in a small context to native speakers, asking them to indicate their preference for clause order on a slider. Interestingly, we found confirmation for both hypotheses. Even more interestingly though, we also found a general preference for the canonical order across all conditions. Since the non-canonical order does occur in natural language, it is likely that there are factors next to connectedness and information structure that language users take into account when deciding on a preference for clause ordering.

Bibliography

Austin, J. L. 1970. “Ifs and Cans.” In Philosophical Papers, 205–32. London: Oxford University Press.

Clark, Herbert H., and Susan E. Haviland. 1977. “Comprehension and the Given-New Contract.” In Discourse Production and Comprehension, edited by Roy O. Freedle, 1–40. Hillsdale, NJ: Erlbaum.

Diessel, Holger. 2005. “Competing Motivations for the Ordering of Main and Adverbial Clauses.” Linguistics 43 (3): 449–70. https://doi.org/10.1515/ling.2005.43.3.449 .

Evans, Jonathan St. B. T., and Stephen E. Newstead. 1977. “Language and Reasoning: A Study of Temporal Factors.” Cognition 5 (3): 265–83. https://doi.org/10.1016/0010-0277(77)90005-1 .

Ford, Cecilia E., and Sandra A. Thompson. 1986. “Conditionals in Discourse: A Text-Based Study from English.” In On Conditionals, edited by Alice Ter Meulen, Charles A. Ferguson, Elizabeth C. Traugott, and Judy S. Reilly, 353–72. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511753466.019 .

Haviland, Susan E., and Herbert H. Clark. 1974. “What’s New? Acquiring New Information as a Process in Comprehension.” Journal of Verbal Learning and Verbal Behavior 13 (5): 512–21. https://doi.org/10.1016/S0022-5371(74)80003-4 .

Reuneker, Alex. 2020. “Clause Order and Syntactic Integration Patterns in Dutch Conditionals.” Linguistics in the Netherlands 37 (1): 119–34. https://doi.org/10.1075/avt.00041.reu .

———. 2022. “Connecting Conditionals: A Corpus-Based Approach to Conditional Constructions in Dutch.” PhD thesis, LOT. https://doi.org/10.48273/LOT0610 .

Van der Leer, Thomas J. C., and Lotte Hogeweg. 2024. “Als Dan or Dan Als?: Preferences in Clause Order for Dutch Conditionals.” Nota Bene 1 (2): 276–92. https://doi.org/10.1075/nb.00017.van .