Íslenskt mál og almenn málfræði - 2020, Side 300

tion. Or this was the point. Moreover, the frequency of Adv−Vfin is also some-
what higher in interrogatives than in declaratives in those modern-like or
adjunction-based grammars that are insensitive to assertion features.
Question 7: Indeed, I do not expect there to be language-internal (here: morpho-
logical, morphosyntactic) factors that prevent Icelandic syntax from morphing
into Danish syntax — or the other way round. Given the scope of the variation
we see in 19th century Icelandic, I find it hard to accept the consequences of the
opposite view. This would entail that the grammar that was successfully inter-
nalised by some speakers in my data is in some sense ruled out by the system;
that it is impossible, ruled out by UG and thus unlearnable. The conclusion
would then be that the textual record is riddled with data that could never be a
part of natural human languages. I can certainly understand how the idea appeals
to theoretical linguists, but I find this view of the attested variation rather hard
to swallow.
I do not feel committed to the grammar competition approach although I
adopted it for convenience. The analysis provided by Heycock and Wallenberg
(2013) indeed suggests that all other things being equal, the fitness of V-to-T is
inferior to V-to-C. This lack of fitness means that, given enough time, Icelandic
should evolve into Danish in terms of verb-adverb placement because learners
would eventually always settle on a representation without V-to-I movement.
First of all, there are additional syntactic factors that may provide evidence
for a Split IP and these factors may give rise to obligatory V-to-I movement (cf.
Bobaljik and Thráinsson 1998). An example of this is the transitive expletive
construction (TEC), which has been argued to correlate with obligatory V-to-I.
TECs are attested in Icelandic, also in my 19th century data. If TECs require a
Split IP or, alternatively, a high and a low position for the finite verb above the
vP/VP level, e.g. in IP and/or a Split CP, they arguably contribute to the fitness
of V-to-I even in the absence of morphological cues. These and presumably vari-
ous other syntactic cues that signal or provide evidence for the structural posi-
tion of the verb, perhaps even (Full NP) Object Shift which interacts with verb
movement (Holmberg’s Generalisation), would increase the fitness of V-to-I.
More generally, what effect do factors such as expletives, null subjects and pres-
ence vs. absence of Stylistic Fronting have on the fitness of V-to-I? Does it per-
haps increase? Conversely, what does the more widespread availability of
embedded topicalisation in Icelandic (=V-to-C) mean in terms of fitness (cf.
Gärtner 2016). Is it then reduced? These would be relevant questions to deter-
mine whether or not V-to-C is expected to win out over V-to-I over time if syn-
tax and morphology are independent from one another, as I assume (see my dis-
sertation §3.7.4).
However, I would also like to voice my doubt that we should be looking for
explanations for the stability of V-to-I. After all, there is evidence that V-to-I
Heimir Freyr van der Feest Viðarsson300