Íslenskt mál og almenn málfræði - 01.01.2023, Side 190

ing for a syntactic bootstrapping approach to language acquisition where univer-
sal cues are privileged (Lidz et al. 2003). Indeed, despite the fact that the univer-
sality of argument structure cues has been called into question (Bowerman and
Brown 2008), support for the syntactic bootstrapping account has even been
found in work with argument-drop languages such as Japanese and Turkish,
where children use syntactic frames as cues – in addition to e.g. case morphology
(Göksun et al. 2008 and Matsuo et al. 2012). In this respect, one hypothesis
(Leischner et al. 2016) is that children rely less on the number of arguments and
more on case when the syntax is less reliable (word order less rigid). The results
in Chapter 2 suggest that this does not need to be the case. Under certain condi-
tions, case and argument number can be equally valuable as cues.
The results also contribute to an ongoing debate about the resources children
use to map form to meaning, particularly in the context of psych verbs (Harts -
horne et al. 2015 and 2016, White et al. 2017, Hacquard and Lidz 2018, Harrigan
et al. 2019, Shablack et al. 2020), i.e. predicates describing psychological states,
and their experiencers.7 I argue that data from Icelandic, a language with very
limited argument-drop and a relatively rigid word order (like English), but rich
case morphology (like Turkish), can provide important insights in this context.
The results from the experiments in Chapter 2 provide qualified support for a
morphosyntactic bootstrapping account that does not exclusively rely on univer-
sal cues, and I therefore follow e.g. Göksun et al. (2008), Matsuo et al. (2012) and
Leischner et al. (2016), in building on the insights from the syntactic bootstrap-
ping literature without limiting the learning possibilities to argument number
and order. The basic theory is the same: Form drives the acquisition of meaning
(Fisher et al. 2020) and verbs can be grouped into semantic categories based on
their structural behavior (see also Pinker 1989, Levin 1993 and Yang 2016). But
other properties than word order and argument number can be relevant, and in
languages like Icelandic, morphosyntactic bootstrapping, based on case, is a strong
candidate. Since a learning model detects the available systematic mappings of
form and meaning (Yang 2016), cue universality is not crucial to cue availability.
In the context of formal case theories, the results show that learners do associate
case with specific semantics despite the non-exact nature of the mappings.
Still, it is important to bear in mind that the use of case cues varies across con-
texts in the experimental results of Chapter 2. Dative on subjects, as opposed to
objects, seems to be more salient. This could be interpreted in various ways, one
of them being that the low frequency of non-nominative subjects in general,
paired with the generalization that non-nominative corresponds to non-agents,
provides a more accessible contrast to default case marking in the language.
Another interpretation could be related to effects of parsing on the availability of
cues (Trueswell et al. 2012). It could then be argued that the different positions
Iris Edda Nowenstein190
7 Note that terminology varies when it comes to psych/perception/attitude/mental
state verbs. Here I use the term psych verb in the broadest sense, including all the categories
just mentioned.