1. Nuclear predication

The structure of the predication is determined by the combinatorial possibilities of the predicate, as defined in the predicate frame. All predicates are lexical items of the language. In principle FG does not recognize “abstract predicates”. This means that predicates can normally occur in actual linguistic expressions of the language.

The meaning of the most basic predicates is such that it can be analyzed in terms of combinations of the meanings of semantically simpler predicates. Thus, the English predicates ‘die’, ‘kill’ and ‘murder’ are basic predicates, since there is no rule in English by means of which they could be formed; they therefore belong to the lexicon of English. Their meanings are structured in the sense that they can be analyzed in terms of more elementary predicates.

In order to account for the semantic properties of predicates and the semantic relations between them, each predicate frame in the lexicon is provided with the number of meaning postulates. Meaning postulates specify what predicates are entailed by given predicate frame.

1. Predicate frames – the general format of predicate can be illustrated with the predicate frame for the English verb ‘give’:

Give_v(x₁:<anim>(x₁))_Ag(x₂)_Go(x₃:<anim>(x₃))_Rec

The predicate frame specifies the following types of information concerning the predicate:

– the form of the predicate – this form is coded in some standard type of phonology representation. By convention, we represent the form of the predicate by the written form of the infinitive. For many languages we will need the stem or root rather than the infinitive or any other inflected form of the predicate. The predicate may have irregular forms. All forms which cannot be productively derived must be contained in the lexicon, these irregular forms must be stored with the predicate frame in the lexicon. This may lead to an interpretation of ‘the form of the predicate’ as ‘the set of all unpredictable forms of the predicate’.

– the syntactic category – alongside verbal predicates, we recognize at least adjectival (A) and nominal (N) predicates. Many of the subcategorization and selectional properties of the predicate, however, need not be explicitly coded in the predicate frame. For example, that ‘give’ is ditransitive is clear from the fact that it takes 3 arguments, that it is agentive is clear from the semantic function on the first argument position.

– the quantitative valency – the number of arguments that the predicate takes to form predications. The argument position are symbolized by the variables x1, x2,…,xn which mark the argument slots. We distinguish one-place (monovalent), two-place (bivalent), three-place (trivalent), and in general n-place predicates. In natural languages the maximum quantitative valency of basic predicates seems to be 3, and that of derived predicates – 4.

– the qualitative valency – the types of argument that predicate takes, as specified by the semantic functions of the argument, and the selection restriction imposed them. In case of predicate frame, they tell us that the argument of ‘give’ play the role of Agent, Goal and Recipient. The selection restriction such as <animate>, inform us about the semantic type of the terms which can be inserted into the argument position if one is to arrive at a ‘normal’, non-metaphorical type of predication.

The predicate frame codes a considerable amount of information concerning the semantic and syntactic combinabilities of the predicate. The predicate thus provides a “blueprint” for the types of predication. When two predicate frames differ in any of the features describes above they are, by definition, two different predicate frames.

The predicate frames themselves define the kinds of structures in which the can be used: we get a predication when the argument slots of the predicate frame are filled with them.

The order in which the predicate and the argument are given in the predicate frame has no direct or necessary relation to the linear order in which these constituents will finally be expressed. Predicate frames are order-free structures, the constituent of which will finally be linearized by the placement rules. Languages with quite different constituent ordering patterns can nevertheless be describes in terms of the same format of predicate frames. Since the constituent of the predicate frame have no linear order to begin with, they cannot be reordered or moved one position to another. The numbering of the argument positions x1, x2, x3,… reflects a priority hierarchy defined over the semantic functions, in the sense that Agent arguments are central to the predication than Goals, and these more central than Recipients. We shall have reason to distinguish “1^st argument” (such as Agent), 2^nd argument (such as Goal), and 3^rd argument (such as Recipient).

Predicate frames can be called ‘open predications’ – they provide structures from which predications can be formed through the insertion of term structures. The term ‘open predication’ can be used for any predicate frame which has at least one term position which has not been filled with a term structure. When all term positions of predicate frame have been filled by term, we speak of a ‘closed predication’. There are also open predications which result from partially filling in the argument position of the predicate frame. Of special importance are open predications in which all argument positions except one have been filled in with terms.

1. Arguments vs Satellites – we distinguish two types of term positions: argument and satellite positions. Arguments are those terms which are required by some predicate in order to form a complete nuclear predication. They are essential to the integrity of the SoA (States of Affairs) designated by the predicate frame. If we leave them out, the property/relation designated by the predicate is not fulfilled or satisfied. Satellites are not in this sense required by the predicate; they give optional further information pertaining to additional features of the SoA (Level 1), the location of the SoA (Level 2), the speaker’s attitude towards or evaluation of the propositional content (Level 3), or the character of the speech act (Level 4). Arguments relate to the predicate. Satellites relate to the predication clause. Arguments have a more central position in the clause, satellites – a more peripheral one. Both arguments and satellites are terms, i.e. expressions which can be used to refer to entities. The satellite ‘in the car’ clearly contains a term referring to a particular car; and ‘in the afternoon’ can be used to refer to that moment of time that, by convention, we call “afternoon”.

Arguments are obligatory, satellites are optional constituent of the clause, and: arguments are characterized by one set of semantic functions, satellites by another set, and these two sets do not overlap. In certain cases it may be difficult to decide whether a given term is an argument or a satellite.

There is a certain degree of overlap between argument and satellite semantic functions. In other words, certain semantic functions may sometimes mark a satellite, sometimes an argument.

1. Mark sold his car in Berlin
2. Mark lives in Berlin

In each of these constructions the constituent ‘Berlin’ has the semantic functions of Location. In the 1^st example, however, it has the status of a Level 2 satellite, which locates the whole SoA of Mark’s selling a car in the spatial dimension, while in the 2^nd example it is an essential argument of the predicate ‘live’. ‘Live’ in this sense designates a relation between an animate being and a Location, and the SoA is not complete if the Location term is left out.

A satellite can be left out without affecting the grammaticality or the meaning of the remaining construction, whereas leaving out an argument will either render the remainder ungrammatical or change its semantics. Arguments may be left unspecified in certain settings. Predicates can be put to so-called ‘absolute’ uses.

1. Mark was eating an apple.
2. Mark was eating.
3. Mark eats.

In the 1^st example ‘eating’ is used as a two place predicate specifying a relation between some animate being and some beverage. In the other examples, the second argument is not specified. In the 3^rd example can only be interpreted in the pregnant sense of ‘being a habitual consumer of food’: it comes close to ascribing a property to Mark, rather than describing an action in which he is involved. In the 2^nd example seems to ambiguous between the sense of 1^st example, with a contextually retrievable second argument left unspecified, and the pregnant sense which we find in the 3^rd example.

1. Selection restrictions –

Give_v(x₁:<anim>(x₁))_Ag(x₂)_Go(x₃:<anim>(x₃))_Rec

The selection restriction itself has the form of a predicate frame: it designates a property which imposes a condition on the types of terms which may be inserted into the argument position.

In the example: Mark was drinking an apple – both position recognize that there is something strange about this example, but they differ on how this strangeness should be accounted for. Position I holds that the strangeness is a function of the linguistic properties of ‘drink’ and ‘an apple’, and that these properties must somehow be coded in the linguistic description. Position II says that the strangeness can be attributes to our knowledge of the world: we know that people do not normally drink apple and that is why the example is strange. Linguistically speaking, there is nothing wrong with it.

Learning a language means at least learning the predicates of the language. It is learning the predicate frame, which includes the combinatorial possibilities of the predicate. In part, the combinatorial possibilities of a predicate are determined by semantic factors. This is especially clear in the case of predicates which can only combine with a very restricted class of arguments. Take a predicate such as ‘blond’. It is impossible to explain the most current usage of this predicate without mentioning ‘hair’. The restriction must thus be a fact about the predicate, to be captured in the predicate frame.