2. DECLARATIVE V2 IS LIMITED TO ASSERTED PROPOSITIONS

The illocutionary act normally associated with declaratives is that of asserting, roughly the act of uttering a sentence with the intention of making the addressee accept the content of it and take it as part of the common ground. Given that V2 word order in Swedish and other languages is excluded in the complement of the factive predicate ångra (‘regret’), see (3a), where the content of the clause is presupposed (already part of the common ground), we seem to have evidence for a relation between V2 word order and assertion or the corresponding assertoric force.Footnote ¹ We can at least conclude that a fully presupposed proposition cannot be expressed by using V2 word order.

1. (3)

Looking at semi-factives like upptäcka ‘discover’, we see that factivity is irrelevant to V2, but whether the content of the embedded clause is presupposed by both the speaker and the hearer or only by the speaker appears to be relevant for the possibility of verb second (see Wiklund et al. Reference Wiklund, Bentzen, Hrafnbjargarson and Hróarsdóttir2009 for details).

1. (4)

In (4), the proposition expressed by the embedded clause has to be presupposed by the speaker but need not be presupposed by the hearer, in which case the illocution of the clause potentially qualifies as an assertion; the sentence may be uttered with the intention of informing the addressee about the fact that Lena had already left. To the extent that we want to accept calling something that is presupposed asserted in this context, which is a matter of terminology, the data confirm the old assumption that the less presupposed the content of the clause is, the more compatible it is with verb second. (4) contrasts with (3) in precisely this way. In (3), the content of the embedded clause has to be part of the common ground (presupposed by both speaker and hearer).Footnote ² Therefore, it does not tolerate the V2 word order, cf. (3a).

Another point to note is that there is compatibility between verb second and matrix negation in Scandinavian under semi-factive verbs, as in (4).Footnote ³ This is in fact also possible with säga ‘say’ as long as negation is an illocutionary negation (in the sense of Searle & Vanderveken Reference Searle and Vanderveken1985). The negation of a non-assertive verb like tvivla ‘doubt’ shows the same (marginal) compatibility, see (5a), despite the fact that this verb is otherwise not compatible with V2 word order, cf. (5b).

1. (5)

These facts show us that there is a relation between V2 word order in declaratives and assertion in the sense that V2 word order is ruled out in cases where the speaker is not undertaking some commitment to the proposition expressed, as in (5b), and in cases where there is no wish to update the common ground (to inform the hearer of something), as in cases where the embedded proposition is already part of the common ground, (3a). In what follows, I will show that even though V2 is restricted to assertive environments in the above sense, the bi-conditional version of the illocution hypothesis as formulated in (2) above – quasi-formalized in (6) – does not hold.

1. (6) Strong illocution hypothesis of V2 (declaratives)

   V2 word order ↔ Illocutionary force

3. V-IN-SITU DECLARATIVES

One immediate problem for the strong version of the illocution hypothesis arises from the fact that non-V2 word order is the default word order of Swedish embedded clauses and it is not clear in what sense clauses with non-V2 word order under sa, as in (1a), and upptäckte ‘discover’ in the relevant sense, as in (7), are not also assertions; at least (7) yields the interpretation of a commitment to the embedded proposition on the part of the speaker and on the parenthetical use (see below) we can also infer a wish that the embedded proposition is added to the common ground (i.e. it also becomes known to the addressee).

1. (7)

The sentences in both (4) and (7) seem to have parenthetical uses, just like (1a) and (1b), in the sense that there seems to be no way to distinguish the two word orders in terms of notions like main assertion (Hooper & Thompson Reference Hooper and Thompson1973). In Wiklund et al. (Reference Wiklund, Bentzen, Hrafnbjargarson and Hróarsdóttir2009), a detailed investigation of Icelandic, Norwegian, and Swedish is conducted, using the diagnostics presented in Simons (Reference Simons2007) in search for differences between the two word orders regarding the status of the clause in terms of main assertion (or Main Point of Utterance – MPU in the terminology of Simons Reference Simons2007). Whenever the content of an embedded clause alone can constitute the answer to a question, the embedded clause has the possibility of being the MPU. Wiklund et al. (Reference Wiklund, Bentzen, Hrafnbjargarson and Hróarsdóttir2009) shows that V2 word order is neither necessary for a clause to represent the main assertion nor is it unambiguously the main assertion of the sentence. As an illustration, consider the example sentences in (9) and (10), which are all possible answers to the question in (8). In (9), the embedded clause expresses the main assertion (he did not come to the party because he did not have time), and in (10) the whole sentence represents the main assertion (he did not come to the party because someone said something that offended him). In both uses, non-V2 word order (a-sentences) and V2 word order (b-sentences) are both possible.

1. (8)

2. (9)

3. (10)

MPU-readings, just like V2 word order, are not possible when the content is already part of the common ground or when it is under non-assertive predicates (e.g. tvivlar). Under assertive predicates and semi-factives, however, MPU-readings are possible, just as V2 word order is; see Wiklund et al. (Reference Wiklund, Bentzen, Hrafnbjargarson and Hróarsdóttir2009) for details. Importantly though, non-V2 word order is also possible.

4. IDENTIFYING FORCE

Suppose that relevant clauses with non-V2 word order nevertheless lack the force of an assertion. The question is how to identify that force. Assertive particles, speech act adverbials, and swear words would seem to constitute candidate indicators. However, these elements appear perfectly fine also with non-V2 word order, see (11).Footnote ⁴

1. (11)

According to Julien (Reference Julien2008, Reference Julien2009), V2 word order is preferred with the swear word för fasen, which she claims has a closer affiliation with (certain types of) illocutionary force than the discourse-oriented expressions exemplified above. However, I have not been able to find a speaker of Swedish who systematically prefers V2 word order when faced with examples like the following (see Wiklund Reference Wiklund2009b):

1. (12)

2. (13)

3. (14)

In fact, the non-V2 word order is judged to be preferred by some speakers, which may be taken as evidence that assertoric force is not the difference at issue between the word orders. Examples can also be retrieved from the internet, see (15) below found on Google (2010–01-18).

1. (15)

It is not unreasonable to think that the presence of a swear word has the effect of producing a force of some kind. Importantly, however, such a force is also perceived with the V-in-situ word order in the presence of a swear word and, to me, also in the presence of the other discourse-related elements exemplified above.

Although this fact does not preclude an interpretative difference between V2 and non-V2 word order in the absence of the above elements, V2 does not appear to be obligatory in the presence of the purported illocutionary force features. Note also that the above elements are impossible in precisely those environments where MPU readings are unavailable and where V2 word order is impossible in Swedish:

1. (16)

If assertoric force is possible in the absence of V2 word order, this weakens the connection between V2 and illocution considerably. The important observation is that the three phenomena – V2, MPU readings, and discourse oriented expressions (including modifiers of assertoric force) – are restricted to identical environments (asserted clauses) but may occur independently from one another.Footnote ⁵ Unless we have evidence to the contrary, the availability of these phenomena may be derived from whatever is responsible for the ‘root’ status of a clause, which in dependent clauses may be, for example, the presence of a certain layer of the C-domain in combination with other configurational properties. Thus, for example, V2-clauses appear to contrast with clauses displaying non-V2 word order in not being topicalizable, as in (17a) vs. (17b). This is also true of the embedded clauses in (11) that lack V2 word order but involve discourse elements, cf. (17c).Footnote ⁶

1. (17)

From the above facts, we may conclude that only a weaker version of the illocution hypothesis of V2 makes the correct predictions for Swedish, see (18).

1. (18) Weak illocution hypothesis of V2 (declaratives)

   V2 word order → asserted clause

Anyone who wishes to link V2 to assertoric force needs an inclusive definition of assertion in the sense that it has to cover also semi-factive verbs like upptäcka ‘discover’. For the present purpose, we may assume (19) below.

1. (19) Assertion_def: The speaker intends

   1. a. to undertake a commitment to the proposition expressed and

   2. b. to update the common ground.

The first part of this definition, (19a), captures the difference between members of minimal pairs like those in (5) above. The second part, (19b), captures the impossibility of having the V2 word order in a clause with no update potential, for example, clauses embedded under ångra ‘regret’ in Swedish.Footnote ⁷ An additional piece of evidence against the strong version of the illocution hypothesis comes from appositives of the kind shown in (20).

1. (20)

The force of (20) must be interrogative. However, the appositive clause is not part of the question but instead constitues an assertion, given something like (19). The appositive displays non-V2 word order and, as shown in (21), V2 word order is in fact not even possible.

1. (21)

Note that V-in-situ is possible also in root clauses – in exclamatives of the kind shown in (22a). In these, a force is arguably present (exclamative) but V2 word order is impossible, as (22b) shows.

1. (22)

Although this fact says nothing about the relation between V2 and the specific illocutionary force of assertion (see Truckenbrodt Reference Truckenbrodt2006 on the difference between exclamatives and epistemic speech acts), it is another case where V-in-situ appears to come with illocutionary force. Examples of what appears to be the reverse situation – V2 without illocutionary force – can be found in Icelandic A (see Jónsson Reference Jónsson1996, Gärtner Reference Gärtner, Delsing, Falk, Josefsson and Sigurðsson2003, and Hrafnbjargarson & Wiklund Reference Hrafnbjargarson and Wiklund2009), where V2 word order is possible also under non-assertive and factive predicates.Footnote ⁸

5. LOOSE ENDS

Summing up what we have seen so far, verb second does not always yield force and force does not appear to require verb second, not even the forces associated with epistemic speech acts. The contribution of the V2 word order to the perceived difference between sentences like (1a) and (1b) (in the absence of modifiers) remains unclear. A natural next step is to pin down what is meant by force here, now that we have seen that assertion alone does not always distinguish V2 from non-V2 word order. One immediate possibility that comes to mind is that (1a) and (1b) differ from each other in point of view-ness (see Speas & Tenny Reference Speas, Tenny and Di Sciullo2003 and references cited therein; cf. Wiklund Reference Wiklund2009b). The importance of the discourse-oriented expressions qua evaluative elements mentioned above indicates that sentience/evidentiality may be relevant. Expressions of that kind yield an interpretation that, to me, appears identical to that produced by V2 word order. To me, the point of view in (1a) seems associated with either the matrix subject or with the speaker (being responsible for evaluating the truth of the embedded assertion), whereas the point of view in (1b) appears associated with only the speaker (being responsible for evaluating the truth of the embedded assertion). Swear words that prima facie seem to include reference to the speaker, such as fan i mig ‘devil in me’ are thus, as predicted, compatible with both word orders in my variety of Swedish:

1. (23)

Although further research is needed to gauge the validity of these intuitions, these facts indicate that verb second has to do with the ‘very fine’ structure of the left periphery and encodes sentience/evidentiality, at least in Swedish. In embedded clauses, V2 word order disambiguates cases where the evidential role (the sentient being who is responsible for evaluating the truth of a proposition) coincides with the speaker and cases where this role can also be assigned to the matrix subject. I expect there to be micro-variation in this respect. If I am correct, the quasi-marked (parenthetical) status of embedded verb second (vis-à-vis the canonical non-V2 word order) is derivable from the interference of the speaker's point of view in environments where the subject of a verb of speakingor propositional attitude would be the most natural binder of the relevant role from a semantic point of view but where it is prevented from establishing this relation due to the V2 clause being an island for such binding. A similar difference appears to be present in Icelandic (or varieties of Icelandic) between the subject-initial and non-subject-initial V2 word orders in embedded clauses (Gunnar Hrafn Hrafnbjargarson, p.c.). The subject-initial word order is ambiguous between the two readings, whereas the non-subject-initial word order yields speaker-anchored assertoric force.