A clause designates a situation. The center of a situation is a situation core. In other words, a clause designates a situation core enriched or specified by further components, among them importantly its participants. For example, designates a situation in which Mary employs Peter, in other words, a situation whose core is an action of employment, specified by Mary as the actor and Peter as the undergoer.

Orientation in morphology

Deverbal nouns

Nominalization of a clause may leave this semantic structure intact. This is what happens in .a. This expression still designates the same situation, witness the synonymy of .a and #b.

Things are different with .b and #c. These are still nominalizations of . However, .b does not designate a situation, but instead one of the participants of the underlying situation, viz. the one coded as subject. (In the universe of discourse at hand, it refers to Mary). Neither does .c refer to the situation of employment, but instead to that one of its participants which is coded as direct object of the verb. (In the present universe of discourse, it may refer to Peter.)

In these latter nominalizations, the underlying proposition is oriented towards one of its participants.¹ Oriented propositions are typically, though not necessarily, open propositions. A nominalization which leaves the nature of the designatum of the base intact, i.e. which designates the same situation (core) as the base, is non-oriented or plain. A nominalization which designates or modifies one of the participants of the underlying situation is oriented towards the participant in question.²

The nominalizations in the example set are deverbal nouns. These are categorized by the following terms in the descriptive tradition:

• A deverbal noun designating a situation core (like .a) is a nomen actionis (‘action noun’) (or abstractum).
• A deverbal noun designating the participant coded as subject of the verbal base (like .b) is a nomen agentis (‘agent noun’).
• A deverbal noun designating the participant coded as direct object of the verbal base (like .c) is a nomen patientis (‘patient noun’).
• In the same sense, there are nomina instrumenti (like mixer ‘what one mixes with’), nomina loci (e.g. Spanish nouns in -dero like atracadero “lugar donde se atraca”, Engl. mooring ‘place where one moors’) and some more.

Beside this system, there is also the concept of the nomen acti ‘noun of the done’, which designates the result of an action that produces a result. The nouns cut in and huuch' in are examples.

Cut in is neither a nomen actionis, as it does not designate the act of cutting, nor a nomen patientis, as the underlying object of the cutting here is the arm; instead, it designates the visible result of the act. Likewise, Yuc. huuch' is a nomen acti derived from the verb huch' ‘grind’. It neither designates the action of grinding nor the object ground (which is corn), and instead the result produced. Nomina acti are frequent in linguistic terminology. For instance, at the start of the section on nominalization, where it says “the NP the shooting of the hunters is a nominalization”, the noun nominalization is a nomen acti, as it designates the result of the operation called by the same term.

The differently oriented deverbal nouns are seldom all distinct in linguistic structure. Often, the same process derives nomina actionis and nomina agentis, which leads to wide-spread polysemy of deverbal nouns. This may be seen in English management, nominalization of manage, which as a nomen actionis means the action of managing, and as a nomen agentis, means the set of managers. It is then the context which decides which kind of orientation is present in the given case. Consequently, a test to ascertain the orientation of a deverbal noun is provided by its selection restrictions. For instance, if the task is to ascertain whether the derivations of employ shown in are oriented and if so, which participant they designate, one puts them in a context like I invited __ for dinner. The result is that the sentence becomes ungrammatical with .a, but grammatical with #b and #c. Consequently, #a cannot designate a human being.

Participles

Up to here, we have seen orientation in nouns s.s. (substantives). It is also operative in deverbal adjectives. In fact, it appears that all deverbal adjectives are oriented. Of special interest in this connection are participles. A participle oriented towards the subject slot of the verb is an active participle, as in ; a participle oriented towards the object slot of the verb is a passive participle, as in #b.

Latin, too, uses the active () and the passive () participles as heads of oriented nominalizations which, allowing for verbal dependents, retain some clausal properties.

Voice

Such deverbal adjectives and expanded adjectivals may be used in different syntactic functions. Both English and Latin use participial constructions both as attributes accompanying nominals, as in and its translation, and as appositions accompanying noun phrases, as in the other examples. However, a concept may also function as a predicate, and so can a participial construction. Since it is a nominal predicate, languages which require a copula with nominal predicates combine it with a copula. In such a construction, the copula becomes an auxiliary which, together with the participle, forms a periphrastic verb form, as in .

Orientation is an operation accompanying nominalization. As may be gathered from , the combination of these two operations provides the structural basis of verbal voice. In a dynamic perspective – more specifically, a grammaticalization perspective –, voices start out in periphrastic constructions based on oriented verbal nouns. It is only by further grammaticalization that a language may acquire voice in simple finite verbs.

The theory of orientation proposed up to now provides for orientation of a proposition towards such variables contained in it which may be coded as actants of its verb. However, to the extent that the formation of participles is done in the grammar rather than in word formation, it may extend to such clause components which are not governed by the verb. This is what happens in Philippine languages. These use oriented verbal nominals as (copulaless) predicates and thus signal the semantic function of that nominal component of the clause which is topicalized (or in traditional Philippinist parlance, which is the focus).

shows the active, passive and indirect-passive voice of a trivalent verb. In .b, an instrument is promoted to topic function. Finally in .b, the same happens with a beneficiary. This voice system allows for any clause component to get into topic position. Further syntactic processes which, in other languages, apply to a large range of syntactic functions, like relativization, may in such a language be restricted to the topic function without any loss in functionality, since any component may be brought into topic function.

Givon, Talmy 1997, "Grammatical relations: an introduction." Givon, Talmy (ed.), Grammatical relations. A functionalist perspective on structure. Amsterdam & Philadelphia: J. Benjamins; 1-84; p. 19. Die interlinearen Glossen waren gegenüber der Vorlage zu korrigieren.
Abkürzungen: ACC accusative, AGNR agent nominalizer, BEN benefactive, BENNR benefactive nominalizer, DAT dative, ERG ergative, INSTNR instrument nominalizer, INSTR instrumental, PATNR patient nominalizer, RECNR recipient nominalizer, TOP topic

In the case of Bikol, verbal predicates are actually nominal in character. However, voice is also found with finite synthetic verb forms. In other words, orientation may take the form of a conjugation category. Latin is a case in point:

Orientation of concepts

This means that orientation is primarily a property of a proposition. It may be coded on its verbal predicate. And it may persist there if the proposition is nominalized. The following table provides a cross-classification of the two fundamental categories of concepts with the alternative of being or not oriented.

The three unproblematic kinds figuring in the table have been discussed above. It remains to be verified whether a finite polyvalent verb of an independent sentence is necessarily oriented towards its subject (or its absolutive actant, in the case of a syntactically ergative language). If so, then a simple independent sentence, not only a sentence with a nominal predicate, but also a sentence with a verbal predicate, would have the bipartite structure of an Aristotelian categorial judgement. By the same token, orientation would then not be restricted to nominalization.

Orientation in syntax

The syntactic constructions that code orientation of a clause are known as relative clauses of different kinds:

1. An oriented clause which functions as an adjectival modifying a head nominal is a plain relative clause.
2. An oriented clause may function as a nominal. There are two varieties of this construction:
1. The clause may be oriented towards one of its lexically coded nominal components. This is an internal-head relative clause.
2. The clause may be oriented towards a syntactic position which is not occupied by a lexical NP. This is a so-called free relative clause. Its subtypes are the headless relative clause and the light-headed relative clause.

Orientation in an adjectival clause

Orientation of the plain attributive relative clause is here illustrated by a postnominal relative clause introduced by a relative pronoun. shows the subject as the function of the relativized position, the direct object and a local complement.

As a multifunctional formative, the Latin relative pronoun not only subordinates the relative clause and signals its attributive relation to the head. It also occupies the relativized position, thus signalling the orientation of the subordinate clause and identifying the syntactic function of the position towards which the clause is oriented.

Orientation in a substantival clause

Orientation in an internal-head relative clause

The mechanism will here be illustrated by the circumnominal relative clause. shows an object clause – thus, a substantive clause – subordinated by an all-purpose nominalizer.

In , the same construction is interpreted as a circumnominal relative clause.

At the semantic level, the following operations take place in :

1. The dog is identified as the semantic core of the concept anchoring to take place.
2. The rest of the proposition is oriented towards the position occupied by the nucleus.
3. The oriented proposition is interpreted as a specifier of the nucleus.

While this takes place at the semantic level, there is nothing in linguistic structure to code it. It is only by the selection restrictions of the matrix verb that we know that the subject of must be an animal, while the direct object of can be anything perceptible. Since a situation is perceptible, no further operations are necessary to interpret , and the interpretation of the subordinate clause as a substantive clause becomes the default. Its interpretation as relative construction is, however, not excluded.

Orientation in a free relative clause

Where the only available relative-clause strategy involves a relative pronoun, the only free relative clause is a light-headed relative clause. In Latin, it is formed by simply omitting the head.

The relative clause in is oriented towards the subject position, the one in is oriented towards the direct object position. The relative pronoun has the same function as in the adjectival clause, except that its gender and number here have nothing to agree with and instead provide the meaning of the “light head”.

On the other hand, the relativized position may be occupied by nothing, the relative clause being oriented by the nominalizer. Such a headless relative clause may function as a noun by itself. and show headless relative clauses oriented towards the direct object and the place actant, respectively.

In and , it is the nominalizing suffix that indicates the orientation. The category of the resulting construction is ‘nominal’.

Coding of orientation

The operation of orientation has a semantic side, which relates to the kind of object designated by the construction: the non-oriented base designates a situation; orientation converts the proposition into a complex concept. Just like other operations such as nominalization and attribution, this operation may or may not be coded in linguistic structure. If it is not, the clause is just subordinated and nominalized to some degree. It may then be formally indistinguishable from a general subordinate clause, as in the initial examples of the section on positional types of relative clauses, or from a substantive clause () or from an action noun like management (s. above). In , the dependent clause shows no internal symptoms of nominalization. It denotes a concrete concept and, in combination with the following head noun, functions as a prenominal modifier. However, such clauses may either be oriented towards an internal head or towards an empty place. Either of the two nominals contained may be the internal head, which produces readings #a and #b. Moreover, it may be oriented towards an implicit place participant, which produces reading #c.

Given an entity of a certain category and its semantic role in a proposition, there is some degree of interdependence between category and function. To the extent that these two aspects of a referential expression do not vary independently of each other, some redundancy is built into the system; and consequently, some parsimony in coding is possible. Especially for local, temporal and manner clauses, it is often sufficient to either code the orientation or the semantic function of the clause. This is treated in a separate section.

More generally, it should be kept in mind that a complex construction involves a set of operations to bring it about. It is not necessary that each of these operations be coded, nor is there, in every case, a hierarchy of codeworthiness in the sense that coding a certain elementary operation is more necessary than coding another one. What matters is that there be some coding. The rest can (more or less safely) be left to inferences. In the present case, speakers usually rely on selection restrictions to determine the nature of what is designated by a subordinate clause or a nominalization.

¹ The concept and the term of 'orientation' go back to Kaznelson 1974:221-228.- The German equivalent of the English term ‘orientation’ is Ausrichtung. Unfortunately, this is also a frequent translation of English alignment (German Ausfluchtung), with the consequence that the German term Ausrichtung is homonymous between two syntactic concepts having to do with argument relations.

² Occasionally (e.g. Álvarez González 2016, based on the literature cited there), non-oriented nominalization is called ‘action/state nominalization’ or ‘process nominalization’, and oriented nominalization is called ‘argument nominalization’ or ‘participant nominalization’. All of these terms except the last one are too narrow. A non-oriented nominalization designates a situation, no matter what the degree of its dynamicity (from property via state and process to event) and control. An oriented nominalization may be oriented towards an argument of the underlying proposition or to some participant not occupying an argument position of the predicate, like a nomen instrumenti or nomen loci, e.g. in .