Íslenskt mál og almenn málfræði - 2020, Side 297

argued to produce more non-standard features than students in their final year
and high-achieving students will tend to produce more standard texts than low-
achieving students. In a way, these are partly also similar in effect to viewing
usage through the lens of social ranks.
I agree that covariation among (sociolinguistically) related variables would be
very interesting to explore further, perhaps on a subset with “data-dense” indi-
viduals and by removing noise from the sá/hinn variable to be able to use that
data set completely. As Caroline Heycock also points out, this may, of course,
also be attempted by hand, by comparing data-dense individuals across the three
studies. Apparently, there are hints of this already in my text, although I did not
attempt to compare the variables systematically in this way.
Question 4: It is true that the subsequent developments in Danish and Icelandic
are quite different, but the scenario that I assume is also different in these two
languages. Sundquist (2003) argues that Stylistic Fronting (SF) could provide a
model for verb raising that involved only V-to-C, with no independent V-to-I,
which resulted in raising only in potential V-to-C contexts. I suggest that what-
ever mechanism derives SF is also responsible for generating Adv−Vfin, not the
type we find in Old Icelandic or the later targeted change,1 but of the type we
find towards the end of the period under study and in Modern Icelandic. The
question whether there are two stages of reanalysis rather than just one is partly
dependent on not only one’s analysis of Adv−Vfin but also on the analysis of SF
(for a recent overview, see Ásgrímur Angantýsson 2017). Whether this Modern
Icelandic-like Adv−Vfin variant and SF are best accounted for in terms of IP
adjunction or perhaps some other focus-related operation at a Split-CP level is
something I did not want to commit to. If SF is analysed in terms of IP adjunc-
tion, for instance, as has been proposed, the stages of reanalysis are brought
down to one instead of two. However, rather than offering an analysis of SF, I
wanted to emphasise relevant environments where Adv−Vfin could be argued to
be modelled after subjectless structures where we find similar configurations
featuring the same order but with a subject gap (however this is best represent-
ed).
Adv−Vfin and SF have been argued to be most permissible in a similar set of
contexts, typically occurring in “modifying” sentences such as relative clauses (cf.
Ásgrímur Angantýsson 2011). In an attempt to make sense of (the origin of) this
Replies to Caroline Heycock’s questions 297
1 I assume that Adv−Vfin was originally possible only in cases where some (light/
pronominal) of the elements of the construction were somehow ‘extracyclic’ in the sense of
Zwart (2005 and later work), and where the non-adjacency of S−Vfin thus did not violate
the V2 constraint. This stage was followed later by targeted change during the Early Modern
period where Adv-Vfin is always an option in embedded contexts and correlates with the
availability of V-to-C.