Íslenskt mál og almenn málfræði - 01.01.2023, Side 194

ciations between case and meaning, using case as a cue for meaning and produc-
ing non-default datives more with experiencers, goals and motion themes than
agents and patients. These associations seem to change through development,
with the dative-goal association emerging before the dative-motion theme one
for example. Furthermore, the form-meaning mappings present differently in
comprehension (or word learning) and production. Thus, dative subjects are a
more salient cue than dative objects in comprehension, but the pattern is inverted
in production, where the dative is produced more on objects than subjects. But
how do these associations between form and meaning emerge and how is their
productivity determined? Can they be derived from distributional information in
the input?
Investigating datives in the acquisition of Icelandic does not only entail
addressing questions about (morpho)syntactic bootstrapping (discussed in
Chapter 2) and rule formation and productivity (outlined in Chapter 3). It also
involves questions about the derivability of linguistic knowledge. How much can
children derive from the input? What is not derivable and therefore potentially
innate? This is the topic of Chapter 4. In a recent paper on the derivability of
linking rules (UTAH and rUTAH8), Pearl and Sprouse (2021:295) discuss the
importance of investigating derivability from realistic child input in general.
They argue that in language acquisition theories, “the general direction of inves-
tigation is to minimize or eliminate innate scaffolding proposals when (empiri-
cally) possible” and that “results that show us which representational components
can be derived in principle—and which can’t—contribute to that effort”. I fully
agree with the importance of maximizing derivability approaches, and that innate
scaffolding should be considered a proposal of last resort. Of course, this is clear-
er in the case of overt morphological case marking than the positional licensing
of linking rules, where arguments for universality and innateness are more
robust, with quite a few approaches to linking rules and syntactic bootstrapping
assuming innate mappings between form and meaning (e.g. Lidz et al. 2003 and
Fisher et al. 2010). Going back to the results of Chapter 2 which show how case,
just as argument number, can be used as a cue for verb meaning, it is interesting
to ask what the theoretical implications are for syntactic bootstrapping if mor-
phosyntactic bootstrapping is based on derived mappings between form and
meaning.
To investigate this, I conducted an incremental analysis of the arguments of
verbs appearing with a frequency >5 (N = 410) in an approximately 750,000
word child language corpus composed of interactions between children (aged
from six months old to 13 years old) and adults. The corpus consists of data from
GJEUM (the Einarsdóttir et al. 2019 language sample corpus), the Sigur jóns -
dóttir (2007) longitudinal corpora and the longitudinal Icelandic data available in
CHILDES (Strömqvist et al. 1995). The arguments were syntactically and the-
Iris Edda Nowenstein194
8 Uniformity of Theta Assignment Hypothesis (UTAH) and the relativized Uni -
formity of Theta Assignment Hypothesis (rUTAH).