Event
Ava Irani will be defending her dissertation proposal on Friday, March 16th at noon. The
defense will be held in the Linguistics Library (3rd
floor, 3401-C Walnut Street Suite 300, C Wing).
The link to her proposal can be found here
<http://avairani.net/files/Irani_proposal_learning_from_positive_evidence.pdf>.
The abstract is below.
*(Working) Title:*Learning from Positive Evidence: The Case
of Verb Argument Structure and the Problem of Overgeneralization
*Advisor:*Julie Anne Legate
*Proposal Committee:*Charles Yang, Kathryn Schuler, Robin Clark
This dissertation examines the way in which children learn
verb argument structure through positive evidence in the
input. We examine the acquisition of raising and control
constructions, the acquisition of causatives, and the
acquisition of passives. Throughout this dissertation, I
evaluate theories of language acquisition that rely on
indirect negative evidence (Ambridge et al. 2008; Becker
2014; Bowerman and Croft 2008; i.a.), and show that these
approaches cannot adequately account for the learner's
developmental timeline. I also discuss the problem of
overgeneralization (Baker 1979), which refers to instances
where children overgeneralize a rule, creating a superset
grammar. Adopting Yang's (2016) Sufficiency Principle, I
show that children generalize rules in their grammar when
there is sufficient evidence in the input, and they retreat
from their generalization when there is insufficient
evidence in the primary linguistic data.
First, I address the acquisition of two kinds of verbs:
raising verbs that do not introduce an external argument,
and control verbs that do. I show that children can
differentiate these verbs using positive evidence in the
input in the form of non-referential expletive subjects. I
argue against indirect negative evidence approaches (e.g.
Becker 2014) that claim for the acquisition of control
predicates through the absence of inanimate subjects with
control verbs. I also address the use of indirect negative
evidence in accounting for the overgeneralization of the
causative alternation rule (Ambridge et al. 2008; Bowerman &
Croft 2008; Pinker 1989). I show that there is no effect of
statistical preemption in the unlearning of the causative
rule. Furthermore, I provide evidence against entrenchment
accounts, which claim that if a verb is heard in an
intransitive frame enough times, the learner will assume
that it can only occur as an intransitive. I argue against
this approach by showing that verb frequency alone cannot
account for the overgeneralization of the causative rule,
and the retreat from overgeneralization. Finally, I discuss
the acquisition of passives, which have been argued to be
acquired late in English (e.g. Wexler 2004). I argue that
children's knowledge and use of passives is directly
reflected in the input they receive. I present novel child
data of early passive use in English, and show that the low
frequency of their occurrence is due to the low frequency of
English passives in the input. Using the Sufficiency
Principle, I also present an account of when the passive in
English is acquired by the learner. The aforementioned cases
together show that children's mastery of verb argument
structure is reflective of the evidence available to them in
the primary linguistic data.