In the domain of clause combining, the degree of syntactic integration may icon- ically correspond to the degree of semantic integration (see Cristofaro 2003). Starting from the layered structure of the clause in which arguments and satellites are attached at various levels from the predicate up to the clause (Foley and Van Valin 1984; Hengeveld 1989; Cinque 1999), it can be shown that lower-level satellites such as time adverbial clauses tend to be better integrated than higher- level relations such as concessives (see Lehmann 1988; Hengeveld 1998; Croft 2001; and, specifically for Dutch, Smessaert et al. 2005). The question is of course how to measure “integration”. For Dutch, we are in the fortunate position that its verb-second main clauses offer a clue to integration: if the subordinate clause triggers inversion in the main clause, it is integrated, as it occupies the first position, “pushing” the subject to post-verbal position. If, on the other hand, the subordinate clause is followed by subject-verb word order (no inversion), then it is in left-detached position. By way of illustration, let us take a look at two different types of syntactic constructions:
(8) Toen hij thuis kwam, begon het te regenen.
when he home came began it to rain
‘When he arrived home, it started to rain.’
(9) *Toen hij thuis kwam, het begon te regenen.
when he home came it began to rain
(10) Al is hij ziek, hij komt toch.
even_if is he ill he comes anyway
‘Even if he is ill, he is still coming.’
(11) *Al is hij ziek, komt hij toch.
even_if is he ill comes he anyway
The temporal subordinate clause in (8) obligatorily triggers inversion in the main clause, whereas the concessive clause in (10) is obligatorily followed by subject-verb word order. This is not a coincidence, but iconically reflects the higher level of attachment of concessive clauses. If the only indication of the level of attachment would be the presence or absence of inversion in the main clause, this would of course be a circular argument, but fortunately, there are other semantic and formal considerations: temporal adverbial clauses trigger tense agreement (“consecutio temporum”), but concessive clauses can under certain circumstances bail out, see (12) vs. (13).
(12) Toen hij thuis kwam/*komt, begon het te regenen.
when he home came/comes began it to rain
‘When he arrived/*arrives home, it started to rain.’
(13) Al staat/stond hij niet als erg betrouwbaar bekend,
even_if stands/stood he not as very reliable known
hij sprak toen de waarheid.
he spoke then the truth
‘Even if he is/was not known to be very reliable, he nevertheless spoke the truth back then.’
This can be seen as an indication that the (ook) al-concessive clauses are less integrated in the main clause, the reason being that they are separately asserted. This also explains why they behave differently with regard to clefting: concessives cannot (easily) be clefted (see Smessaert et al. 2005), and why con- cessives more easily allow speech act adverbs, such as eerlijk gezegd (‘honestly said’).
What examples (8)—(11) show is that integration vs. non-integration as signaled by the presence or absence of inversion of the main clause verb is a cue for the language user to partly assess the semantics of the subordinate clause. In principle, one could argue that the integration of the subordinate clause is lexically projected from the conjunction, but this is not the case. Just as with argument realisation, a constructional approach is superior over a projectionist approach, as the template (integration/non-integration) carries meaning of its own, although some degree of convention always limits the possible variation. In the case of integration/non-integration, the semantic contribution of the constructional template can be seen at work in the use of several conjunctions which allow both construals. As shown by Smessaert et al. (2005), many a Dutch conjunction occurs both in integrated and non-integrated patterns (e.g. aangezien ‘since’, vooraleer ‘before’, als ‘if’, terwijl ‘while’, tenzij ‘unless’), but the two construals yield different readings. Take for instance the conditional conjunction als. If the conditional clause modifies the state-of-affairs, the conditional is integrated and triggers inversion in the main clause. If the conditional clause modifies the speech act, it is attached at a higher level. There is a vast literature on the topic (Davison 1981; van der Auwera 1986; Sweetser 1990, among others), but the difference can easily be illustrated with the set of examples under (14)-(15).
(14) Als het regent, word je nat.
if it rains get you wet
‘If it rains, you’ll get wet.’
(15) Als je het nog niet gezien hebt, het regent buiten.
if you it yet not seen have it rains outside
‘In case you haven’t noticed, it’s raining outside.’
There are other conditionals introduced by als that fail to trigger inversion in the main clause, but the non-integrated nature is motivated here as well: either they border on concessive meaning, as in (16) (from Konig and van der Auwera 1988: 112), or they are subjunctive (or counterfactual), as in (17) (from same time integrated and non-integrated. The picture is rather complicated for Dutch (see Konig and Van der Auwera 1988 and Van der Horst 2008 for details). The reason this pattern is ignored is that it only rarely occurs as the only possible pattern in Present-day Dutch (see Smessaert et al. 2005, appendix A). Furthermore, I will concentrate on sentence-initial subordinate clauses only here. Also, the presence of the subordinator dat is not taken into consideration here. In some subordinate adverbial clauses, it is obligatory (e.g. omdat), in other subordinate adverbial clauses, it is optional (be it substandard) (e.g. toen dat, hoewel dat) and in still other contexts it is excluded (e.g. als *dat).
Konig and van der Auwera 1988: 114). Note that in the latter case, integrative construal with inversion is also possible, unlike conditionals like (16).
(16) Als ik als schilder slecht was, als metselaar was
if I as painter bad was as mason was
ik echter een katastrofe/succes.
I however a catastrophe success
‘If I was bad as a painter, as a mason however I was a catastrophe/success.’
(17) Als ik in jouw plaats was, ik zou hem aanklagen.
if I in your place was I would him sue
‘If I were in your position, I would sue him.’
To summarise, integrated vs. non-integrated patterns of subordinate clauses are horizontally related in their constructional network. It is only in contrast to integration that non-integration is semiotically meaningful. There is crosslinguistic variation in the absolute degree of syntactic integration of time adverbial clauses, and it is only by comparing the different subordinate constructions that the motivation becomes clear, so that whatever degree of integration a temporal subordinate clause has in a language, it will not be less integrated than a concessive clause. The precise formal features that define the network are language-specific: obviously, the criterion of inversion only works in a V2 language (see also Verstraete 2003).
-  I assume the reader has a basic idea of the word order principles in Dutch clauses. If not,the reader can be referred to Haeseryn et al. (1997:1221-1400), Verstraete (2003) or Zwart (2011).
-  Note that (ook) al concessives are different in another respect as well: they do not featureverb-final syntax of typical subordinate clauses, but rather have verb-initial (in fact V2) syntaxin the subordinate clause. I will come back to this issue below.
-  Space limitations prevent an in-depth illustration of all possible types of subordinate adverbial clauses in Dutch. I will ignore the fact that there is a third pattern: semi-integration, inwhich the subordinate clause is in left-detached position, but is resumed by a correlative element in sentence-initial position in the main clause, so that the subordinate clause is at the