Íslenskt mál og almenn málfræði - 01.01.2023, Side 183

I furthermore show, by testing transitivity and case morphology within the same
experiment, that case morphology can be as salient as the number of arguments
in specific contexts. Second, the data call for a model of rule formation which
accounts for nested non-default productivity and does not hinge on frequency to
determine gradient overgeneralization, since patterns which are not statistically
dominant (the dative) can become productive and multiple rules can apply to a
context simultaneously. I argue this is most compatible with rule-based accounts
which assume that rules are derived from distributional information in the input
(Yang 2002, Albright and Hayes 2003, O’Donnell 2015 and Yang 2016). Third, I
show how such an account, a combination of Yang’s Tolerance Principle (2016)
and his Variational Model of Language Acquisition (2002), can also account for
the acquisition of the Icelandic case marking system in well-known scenarios of
language variation and change, and that a case should be made for stochastic excep-
tions in addition to stochastic rules. I argue that the underlying mechanisms in lan-
guage acquisition, namely a constant search for productive rules while tracking
exceptions, and both convergence with and divergence from the language environ-
ment, shape the directionality and dynamics of language variation and change.
I will now provide an overview of the dissertation, focusing on the methods
employed, representative results and their implications.

1. Icelandic datives
As Joan Maling (2002:31) and others (e.g. Barðdal 1999, 2001 and 2008, Svenonius
2002, Thráinsson 2007, Barðdal 2008, Jónsson 2013, Wood 2015) have observed,
“one of the striking features of Icelandic syntax is the frequency with which
verbs seem to govern the dative case”. But how do children approach this striking
feature and learn to use it productively? Before turning to the question of how
children approach Icelandic datives, it is necessary to understand how linguists
have done so. I start Chapter 1 with a brief descriptive glance at the expression
of the Icelandic case marking system on the surface before discussing relevant
aspects of previous research on the distribution, structural status and semantics
of the dative. Following this overview, the status of exceptions within different
theoretical accounts is discussed. I conclude with an overview of previous work
on case marking in language acquisition and language variation and change. In
this section, the focus is on case marking in canonical contexts and I therefore
exclude e.g. dative subjects in the passive. This is based on evidence that children
‘filter’ out movement to some extent in the acquisition of argument structure
(Perkins 2019).
When taking a brief glance at the Icelandic case marking system, a few very
simple patterns emerge: The default (also referred to as structural or unmarked)
patterns consist of nominative for subjects and accusative for objects (and possi-
bly dative for indirect objects). These patterns have a broad semantic scope, but
agents are always nominative and prototypical patients are usually accusative.
Project rationale and core ideas 183