The Pages of The Gray Wizard

Predicate Arguments

1.1 predicates and arguments

The traditional view of grammar divides a sentence into two parts, subject, and predicate. Thus in the sentence

Galdor slew the dragon

Galdor is the subject and slew the dragon is the predicate.

An alternative view (and the one used throughout this grammar), holds that the sentence consists of a predicate and one or more arguments. In the sentence above, the predicate is slew and there are two arguments, Galdor and the dragon. Notionally, the predicate expresses the relationship (here the act of slaying) between the arguments (here Galdor and the dragon). In this view, a basic assumption is that the arguments both

1. differ in their semantic relationships to the predicate; and
2. are clearly distinguished from one another through grammatical marking. 

In ámman îar these distinctions are made through 

1. semantically motivated predicate inflections; and 
2. syntactically motivated case markings on the head nouns of the arguments. 

Thus:

In this sentence, the semantic relationships are identified by predicate inflections (i galdranne... ernurgoraen — Galdor as agent; an i feng — the dragon as patient) (see below for a discussion of semantic roles and predicate inflections). The syntactic distinction between the arguments is shown by case inflections (ergative for Galdor and absolutive for the dragon). Reversing these inflections would alter the semantic relationship of the arguments to the predicate and produce a very different sentence.

Ergative (e) and Absolutive (-) case inflections mark syntactic relations of the predicate arguments that are discussed below. The position of the auxiliary verb eleth and the predicate inflections (en which marks the patientive argument and er- which indicates the presence of an agentive argument) mark semantic roles which are discussed below.

This grammar distinguishes between the basic means by which arguments of a predicate are expressed, and more specialized means that have definite semantic restrictions. Here I refer to the former as core arguments of the predicate and the latter as oblique arguments.

ámman îar is a very predicate-centric language in that much of its grammar is driven by the valency of predicates, i.e. by the number and type of noun phrase arguments that a particular predicate can take. The arguments of a predicate can be described in at least two ways:

1. Reference to semantic roles; and
2. Reference to syntactic relations

1.1.1 semantic roles

The notion of semantic roles (also called participant, deep case, thematic, or theta roles) is that arguments of a predication play some semantically identifiable part or role in that predication. Consider the following sentences:

1. Galdor slew the dragon with his sword.

2. The sword slew the dragon.

3. The dragon was slain.

Merely noting that the "subjects" of these sentences are Galdor, The sword and The dragon respectively, fails to recognize that the semantic role of these "subjects" is different in each case. In (1) Galdor is a semantic Agent (i.e. he was the instigator of the event); in (2) The sword is a semantic instrument (i.e. it was used to accomplish the event); and in (3) The dragon is a semantic Patient (i.e. something happened to it). Conversely, simply to note that the dragon is the "direct object" in (1) and (2) and the "subject" in (3) fails to recognize that its semantic role is invariant in all three sentences.

In purely notional terms, it is possible to identify a large number of roles that are played by the arguments of a predicate. The number of such semantic roles is not only potentially large (including Agent, Force, Patient, Instrument, Manipulator, Target, Donor, Gift, Recipient, Speaker, Addressee, Experiencer, Perceiver, Impression, Benefactor, Malefactor, Location, Direction, Setting, Purpose, Time, Manner, et. al.), but there seems to be no general agreement among linguists as to what they are or even how many there are.

There are three problems with such notional roles.

1. Like all such notional features, they cannot be defined in any precise way and therefore it is not possible to apply them unambiguously. 

2. It is always possible to suggest more distinctions so that there is, in principle, no limit to the number of possible roles. 

3. They are often partly based on syntactic distinctions and so are not truly notional.

1.1.2 syntactic relations

There exist purely syntactic relations contracted between an argument and its predicate. Although these relations may closely correlate with semantic roles , they cannot be identified with them. The following syntactic relations are relevant to a discussion of ámman îar predicates:

The single argument of an intransitive predicate, which we shall symbolize as S. Arguments in this relation with the predicate will be said to be in S-function.

For the prototypical transitive predicate, the roles played by its arguments are a semantic agent and a semantic patient. Therefore,

These labels are clearly mnemonic for subject, agent, and patient respectively. Although, the subject status of most single-argument intransitive constructions is non-controversial, the same cannot be said for the correlation of semantic roles and syntactic relations of transitive predicates. The advantage of having arbitrary labels for the syntactic functions A and P rather than using the semantic roles agent and patient is that we can continue to use the arbitrary symbols even when we pass beyond the prototypical transitive constructions to other constructions that have similar morphology and syntax. The mnemonic reference of S to "subject" is, however unfortunate in my view. As will be discussed below under Grammatical Relations, subject as a relation is not a useful concept for ámman îar. However, since this terminology is the generally accepted one, I will continue to use it here.

1.1.3 grammatical relations

Languages that use an accusative system for marking the syntactic relations of a predicate"s core arguments traditionally define the relation S=A as the Subject Relation, i.e. NPs in S- or A- function are in the Subject Relation to the predicate. This relation is for various syntactic as well as semantic reasons considered to be primary while NPs in P-function (i.e in object relation to the predicate) are considered secondary. As will be seen, ámman îar uses an ergative system of marking the syntactic relations of a predicate"s core arguments and imposes ergative syntactic constraints on complex sentence structures. In this system, the relation S=P is the primary relation and the A-function argument is considered secondary. In this grammar, I will refer to this primary relation as the Pivot Relation and the secondary relation as the Agentive Relation. The rationale for this terminology will be discussed below.

1.1.4 mapping semantic roles to syntactic relations

Semantic roles are conceptual notions whereas syntactic relations are morphosyntactic categories. While semantic roles influence the morphosyntax, they are not morphosyntactic categories. Languages realize the potentially large number of semantic roles by imposing discrete categories on them, mapping clusters of them in a many-to-one relationship to a limited number of syntactic relations (S,A, P) in terms of prototypes. Referents that are "close enough to the prototype are expressed by noun phrases in the same syntactic relation as are more prototypical referents. In ámman îar these mappings are expressed by the case inflections on the head nouns of the argument NPs.

In addition to case inflections that express the syntactic relations of the predicate arguments, ámman îar identifies a small subset of semantic roles. (Agent, Patient, Theme) using several syntactic devices called Predicate Inflections. Thus, ámman îar expresses both the syntactic relations and the semantic roles of predicate arguments. These two types of argument marking must be understood to be completely independent of and orthogonal to each other. The former are based on prototypical relations of NPs to the predicate while the latter are based on the semantics of the particular utterance.

1.1.5 predicate inflection

In the sentence below, one argument can be identified as agent (notionally the one that performs the action) and the other as patient (notionally the one that undergoes or is affected by the action). In ámman îar this distinction is made using two semi-redundant devices, 

1) Semantic marking of the predicate arguments, 

a) immediately preceding the patientive argument if present and 

b) prefixed to the lexical verb in the presence of an agentive argument;

2) Pronominal cross-referencing on the lexical verb which specifies the semantic roles of its arguments. 

This distribution of semantic markings is referred to as Predicate Inflection in this grammar. Thus

Here, the Patientive argument, (en i feng, the dragon, is preceded by a patientive particle (en); the lexical verb, enurgoro, is preceded by the agentive inflection (er-) indicating the presence of an agentive argument, galdranne, Galdor. Redundantly, the lexical verb is inflected for Agent/Patient valence (-ae-), indicating that the predicate takes both an agent and a patient argument.