UC Berkeley

Grammatical argument structure constructions (ASCs) in English interact with the verbs that act as the lexical head of the clause. This interaction results in some arguments being instantiated while others are omitted. One broad type of omission is known as null complement anaphora (NCA), also commonly called null instantiation. In NCA, core semantic participants are not instantiated as overt arguments, but are nevertheless understood in context. The omitted elements may be the direct objects of verbs, or may even be other constituents such as clausal complements and adjuncts. Some of the types of omissions examined are illustrated in (1) – (5).

(1) The hat doesn’t match ø[Goal, e.g. ‘my outfit’].

(2) They arrived ø[Goal, e.g. ‘in DC’] safely.

(3) I joined ø[Group, e.g. ‘the society’] yesterday.

(4) Did you apply ø[Position, e.g. ‘to that job’]?

(5) He wrote a great speech and delivered it ø[Goal, ‘to the audience’] last night.

Using large semantically annotated and corpus data sets, primarily from the FrameNet Annotation Database, this dissertation presents two important results with respect to lexical and constructional regularities in omission patterns. One main finding is that the omissible element usually qualifies, at an image schematic level, as the ground in a figure-ground relation. Some of the physical verbs discussed include verbs of motion (move, arrive, approach, depart, chase, float), object manipulation (give, implant, provide, join, manacle), perception (listen, glance, peer, gaze), and those verbs that combine motion and object manipulation (splatter, spray, propel, throw, transfer). In all of these cases, either the goal, or the source, or the location is omissible. These are frame elements that tend to be construed as the ground in a figure-ground configuration. Omissions in which the figure-ground rule is observed constitute 68% of a sample (n=2,005) of the annotated sentence data. I provide a Construction Grammar model to account for these figure-ground asymmetries, and show how they consistently result in the omission of elements that end up in the ground. I claim that this generalization holds at a high image schematic level for all of these types of verbs, and therefore this constitutes a lexical frame-based generalization.

The second observation from the data is that metaphor located in the grammar of the argument structure construction itself plays a role in licensing NCA. So, where at first arrive and cajole do not seem to share any commonalities with respect to their semantics more generally, we can at least deduce why their core semantic roles are omissible in sentences such as We arrived ø and He cajoled her ø. It is because in the former, the goal of arrival is the physical ground relative to which the figure is moving, while in the latter the goal of cajoling is metaphorically construed as the ground relative to which he is metaphorically propelling her. The Action frame element (that which he is cajoling her to do) would be instantiated metaphorically via an into-PP: e.g., He cajoled her into marrying him. I propose a model of grammar that incorporates metaphor as part of the argument linking pattern of clausal argument structure constructions. I provide data showing how metaphor is used to structure the domains of Communication, Thinking and Action, and propose a classification of metaphoric argument structure constructions. The main dimension along which metaphoric ASCs are classified is whether the verb in the clause is evoking the target domain or the source domain of the metaphor (I arrived at the conclusion (verb evokes source) vs. He cajoled her into marrying him (verb evokes target)).

I also provide a methodological innovation in the way argument omission is studied. Namely, I suggest that to understand omission, we have to look at equivalent sentences in which those same frame elements are in fact instantiated. We do this in order to gauge the syntactic diversity with which the instantiation is possible. The same frame element could be instantiated using any number of complement types, e.g., The Goal frame element in I arrived home (NP) vs. I arrived at the airport (PP)). I assume that the range of syntactic strategies for instantiating the same frame element can influence whether that frame element is able to be null instantiated at all. Therefore, in order to understand why a particular frame element was candidate for omission in the first place, we must understand the possible ways in which it could be syntactically overt. Most saliently, I find that in many cases one or more of those syntactic strategies available to the instantiation of a frame element are metaphoric in nature, construing that element by use of prepositional phrases with in, into, out, out of, from, against, etc.