In this chapter I will first discuss a number of difficulties that arise if one wishes to establish the valency (or valencies) of a certain predicate and the fact that more than one predicate frame is associated with many predicates. Next, I will deal with the syntactic functions of arguments, that is the obligatory elements that together with the predicate constitute the nuclear predication, and with the semantic functions these arguments may fulfil. This chapter will end with a number of examples of predicate frames.
In a modern language the valency of a predicate can quite often be determined by using the linguistic intuition of speakers of that language. This is, of course, excluded in the case of Latin. Another way to establish the valency of a predicate is to study the occurrences of that predicate in a `corpus' of adequate size and variety and to decide on statistical grounds what its valency is. Latin does not differ from other languages in this respect, apart from the fact that the corpus we have contains a relatively high number of stylistically marked texts (many are literary texts). Furthermore, our corpus consists solely of written documents. Written language differs from spoken language in several respects. These limitations, however, are to a large extent outweighed by the availability of lexicons for very many authors, the work done at the Thesaurus Linguae Latinae (TLL) in Munich, the presence of computer corpora and, recently, the compact disc corpus of the Packard Humanities Institute. (b) The role of context
If in Latin some entity is known or supposed to be known on the basis of contextual or situational information, a writer or speaker need not refer to that entity in an explicit way. This general rule also holds for entities
that must be regarded as arguments in relation to a specific predicate. An example is:
(1) orabo ut mihi pallam reddat quam dudum dedi (sc. ei) (`I'll beg her to give me back the mantle I gave her a while ago', Pl. Men. 672)Examples like (1) are especially frequent in certain context types: 
(2) (C. Piso … orator) minime … tardus in excogitando (`(the orator Gaius Piso …) by no means slow in invention', Cic. Brut. 239; cf. defensionem/causam excogitare)
(3) ne quis te invitum polliceri … cogat (`that no one shall compel you to promise against your will', Cic. Ver. 2. 148)
(4) nullo hoste prohibente aut iter demorante (`no enemy to hinder him or delay his march', Caes. Gal. 3.6.5.)
(5) at enim veteranis suspectum nomen est M. Bruti? … equidem non arbitror (`but the name of Marcus Brutus is suspected by the veterans? … I myself think not', Cic. Phil. 10.15)
(6) suscepi causam, Torquate, suscepi (`I have undertaken the case, Torquatus, I have undertaken it', Cic. Sul. 20)
In all these examples the context provides the information the reader needs in order to understand which elements are associated with the predicate. In our grammars and dictionaries the predicate in these cases is often labelled `absolute' or `used absolutely'. For details concerning the omissibility of elements in a continuous text, see also chapter 12.(c) More predicate frames, more valencies
Many predicates have more than one meaning. As a consequence they also have more than one predicate frame which may also differ in quantitative valency. Sometimes a predicate has several predicate frames even though the semantic differences involved are only small. I will discuss this problem in some detail.
(i) Some predicates have more than one predicate frame without having different valencies. A predicate's having more than one predicate frame often follows from its having more than one meaning, as in the case of reddere:
(7) orabo ut mihi pallam reddat quam dudum dedi (`I'll beg her to give me back the mantle I gave her a while ago', Pl. Men. 672)
(8) haec (fama) itinera infesta reddiderat (`this rumour had made the routes full of danger', Caes. Civ. 3.79.4)
In (7) reddere means `to give back'. It is quite obvious that reddere in this case, just like all verbs of `giving', is a three-place predicate. In (8), however, reddere
means `to render'. As such it can be compared to facere and creare, where in addition to a Subject constituent and an Object constituent, a third argument is required which agrees with the Object in case (a so-called Object Complement).  In other cases of predicates with more than one predicate frame, the differences are much more subtle. In the examples (9) and (10) the difference in construction does not correlate with a difference in meaning. However, we will see below that they are not simply exchangeable in their contexts (see crosssection 5.2.4. (b), p. 50, and crosssection 12.3.3., p. 255):
(9) arma umeris circumdare (`to throw the armour around the shoulders') (cf. Verg. A. 2.510)
(10) filo collum circumdare (`to circle the neck with a riband') (cf. Catul. 64.377)
(ii) Many predicates have more than one valency and more than one predicate frame which correlate with a difference in meaning. A good example of this is the verb dicere. With this verb we may distinguish two-place dicere (= `to speak')  and three-place dicere (= `to say, tell'):
(11) est oratoris proprium apte, distincte, ornate dicere (`it is the orator's peculiar ability to speak with propriety, clearness, elegance', Cic. Off. 1.2)
(12) tum mihi Roscius … alia multa … dixit (`then Roscius told me many other things', Cic. Quinct. 78)However, distinguishing two predicate frames for the verb dicere is not sufficient. There is another three-place dicere, when it means `to appoint someone as'. There is even another two-place dicere, which, again, differs from dicere in (11):
(13) prooemium dicere (`to pronounce') In view of such variation one might object that it does not make much sense to speak about the valency of a predicate. One might also ask how the frames are related to each other and how the meanings are related to each other.
One reason for establishing the frame or frames of a certain predicate (say, dicere) is that we may use it or them for describing properties of other predicates or classes of predicates. In (11), for example, dicere can be compared to loqui. In (13), however, dicere is comparable to pronuntiare. Such a paradigmatic comparison enables us to distinguish various predicate frames and various valencies for a verb, while at the same time giving us criteria to describe the frames of other verbs that do not show the same variety of frames. One more example will illustrate this point. Consider examples (14a) and (14b):
(14a) John was eating his hamburger
(14b) John was eating
If we simply compare the two sentences in (14), we might conclude that to eat is a one-place verb, since the (b) sentence is grammatical, and therefore his hamburger is an optional constituent. However, eating in the sense of eating up resembles other verbs which indicate `taking through the mouth and oesophagus into the stomach' like to swallow and to devour, which are strictly two-place predicates. Eat in (14b), however, resembles verbs like to dine and to feast, which are strictly one-place predicates. John was dining resembles (14b) in this respect. Combining these two things it seems better to assume that the verb to eat has two valencies. Other verbs of `processing food' are either one-place or two-place.
If a predicate has more than one frame and valency, these frames and valencies can quite often be shown to be related to each other in a systematic way. Many verbs, for example, can occur both as a `neutral/unspecified' activity and as a specific activity, such as sing/sing the national anthem. The unspecified activity is one-place, while the specific one is a two-place predicate. A related example is the verb to paint, which is two-place when used in a specific sense, but one-place when denoting professional activity: paint a wall/paint (i.e. `be a painter').(d) The relationship between active and passive (valency reduction)
In most languages that have an active/passive distinction the Agent or Cause of the action indicated by the verb is absent in passive sentences. Furthermore, it is often the case that the Agent or Cause cannot be retrieved from the context. Recent studies on the use of the passive in English and German texts have shown that in c. 80% of sentences with a passive verb the Agent or Cause is not indicated. In Latin texts too, the Agent or Cause is most often not expressed. On the basis of these observations one might assign different predicate frames to the active and passive forms of verbs. One might state this in a different way saying that passivization is above all valency reduction. It has been shown, however, that at least in English there are also cases where the Agent or Cause must be expressed. Since no research on this particular aspect of Latin has been carried out, a few English examples will serve to illustrate the point: 
(A) (15) An irresistible desire to run away possessed me
(16) *I was possessed
(17) I was possessed by an irresistible desire to run away
The Cause constituent cannot be left unexpressed with the verb to possess.
(B) (18) Somebody followed me
(19) I was followed
(20) Curse one followed curse two
(21) *Curse two was followed
With the verb to follow the inanimate Cause has to be expressed. The animate Agent, however, can be omitted.
(C) (22) His parents brought him up
(23) He was brought up by his parents
(24) *He was brought up
(25) He was brought up in Cambridge
The Agent can be omitted only if some Adjunct is present in the sentence.
(D) (26) This sonnet was written by Shakespeare
(27) *This sonnet was written
The fact that a sonnet was written is so trivial, that the information makes no sense without further specification. I will return to the active/passive relation further on ( crosssection 2.4. on p. 19).(e) The cognate object (valency extension) 
In Latin certain one-place predicates can be extended into two-place predicates. An example is mirum somnium somniare (`to dream a wonderful dream'). In standard grammars mirum somnium is often said to be an `internal' or `cognate' Object. The objecthood of the constituent appears from the fact that it may be passivized. The noun in such noun phrases has a meaning that is semantically related to the meaning of the verb. Usually there is also a close morphological relation between the nominal and verbal expressions. The noun is normally accompanied by an adjective or another expression as its Attribute. It is usually the Attribute which contains the most important information. This appears among other things from its position with respect to the Head noun: it normally precedes, as is the rule for Attributes carrying Focus function (see crosssection 9.4. on p. 186). Examples can be found in K.–St. I.274–8.
In chapter 1 `arguments' have been defined in the following way: arguments are those constituents related to a specific predicate that cannot be omitted without making the remaining sentence ungrammatical. This definition of the notion `omissibility' has a number of unsatisfactory consequences. Example (14) illustrates the fact that the predicate to eat must be regarded as a monovalent predicate if the omission test is used as the only criterion. However, there seems to be a semantic distinction involved (`non-specific' vs. `specific' activity). Paradigmatic comparison also indicates that it is feasible to distinguish two frames, both a monovalent and a bivalent to eat. Therefore, a
more precise definition of `omissible' is in order: a constituent is considered omissible if (a) the remaining utterance is grammatical, (b) the lexical meaning of the remaining constituents does not change, and (c) the semantic relations between the remaining constituents are not changed.
In (14a) the constituent his hamburger is taken as non-omissible, because if it were to be eliminated was eating would be interpreted in a different way. There is another way to demonstrate that his hamburger in (14a) stands in a closer relationship to was eating than a constituent like heartily in (28):
(14a) John was eating his hamburger
(28) John was eating heartilyIf in examples (14a) and (28) the constituents his hamburger and heartily are omitted, the remaining utterance (John was eating) is still grammatical. Both constituents (his hamburger and heartily) therefore seem to be equally omissible. However, a difference between the two constituents becomes apparent if we apply the so-called do so test as developed in early transformational generative grammar. Note the result if we replace the predicate eat by the `pro-verb' do, as, for example, in (29):
(29) John was eating and he did so heartilyIn (14a), however, the constituent his hamburger cannot be isolated from was eating in a similar way:
(30) *John was eating and he did so his hamburgerCompare also:
(31) What was John doing heartily? # eating
(32) *What was John doing his hamburger # eatingOn the basis of this do so test one might wish to claim that his hamburger has a different, in fact closer, relationship to was eating than does heartily. I believe that the do so test is a useful test in Latin syntactic research, although the attested instances of facere referring to a preceding action are rare. [7a] The number of instances in which facere is used in that way in conjunction with other –omissible –constituents is even smaller. However, instances of this type do occur, like (33):
(33) istam rem inquisitam certum est non amittere:: edepol me lubente facies (`this matter shall not escape investigation, I am resolved on that :: dear me, sir, do investigate, and welcome', Pl. Am. 848)In this example facies refers to istam rem inquisitam … amittere. As a consequence, the constituent me lubente that is added to it must be facultative. Similarly te lubente in (33a) must be regarded as a facultative constituent:
(33a) istam rem te lubente non amittam inquisitam
In the following examples we also encounter combinations of the pro-verb facere and various satellites:
(34) ego perfodi parietem … et sene sciente hoc feci (`I dug a hole through the wall and this with the old gentleman's knowledge', Pl. Mil. 142–4)
(35) exora, blandire, palpa :: faciam sedulo (`entreat, coax, cajole:: I'll do my best', Pl. Poen. 357
(36) amat a lenone hic :: facere sapienter puto (`he is in love with a pimp's girl :: sensible of him, I judge', Pl. Poen. 1092)
(37) dotem dare te ei dicas, facere id eius ob amicitiam patris (`you might say you are giving her a dowry, doing so out of friendship for her father', Pl. Trin. 737)
(38) suscepi causam … et feci libenter (`I have undertaken the cause and done so gladly', Cic. Sul. 20The examples (34)–(38) prove that the following ones are possible as well:
(34a) ego perfodi parietem sene sciente
(35a) exora sedulo
(36a) sapienter a lenone hic amat
(37a) dotem dare te ei dicas eius ob amicitiam patris
(38a) suscepi causam libenterThe above examples have been chosen in order to show that in Latin the proverb facere may substitute for the nuclear predication and allow the addition of satellites such as Circumstance (so-called ablative absolute construction) in (33) and (34); Manner in (35), (36) and (38); and Motive in (37). However, arguments do not occur with facere in this way. We do not find examples like:
(38b) *… et feci causamThe facere test may help us in specific situations to determine the valency of a predicate. However, the valency of a predicate will be established much more frequently on the basis of statistical observations with regard to the occurrence of that predicate in a given corpus of texts: if a predicate occurs frequently with certain types of constituents, or if such constituents are frequently to be understood, and if these constituents are not likely to be satellites (that is, might in principle be added to any predicate), then the constituents probably should be regarded as arguments.
In this section I will discuss the syntactic functions arguments may fulfil. The semantic functions of arguments will be taken up in the next section. I will not
consider the so-called impersonal verbs and impersonally used verb forms like pluit, licet, miseret and concurritur, parcitur, respectively. For these I refer the reader to table 2.4 on p. 23.
(i) Subject. With monovalent predicates like ambulare the single argument occurring in an active sentence fulfils the function `Subject'. The finite verb form agrees with this constituent in Number and Person, for example ambulat with pater in pater ambulat. With bivalent and trivalent predicates the function `Subject' is fulfilled by the constituent that determines agreement of the finite verb form in Number, Person and sometimes also Gender (in complex verb forms, for example laudaturus est).
N.B. 1. Finite verb forms in Latin also manifest agreement with persons or things which are to be understood as Subject in a given situation or context, for example laudat (laudaturus est). In such cases we may speak of an `internal Subject'. N.B. 2. In the Accusative and Infinitive construction the expression of agreement is restricted (only Number and Gender in periphrastic forms).
N.B. 1. Finite verb forms in Latin also manifest agreement with persons or things which are to be understood as Subject in a given situation or context, for example laudat (laudaturus est). In such cases we may speak of an `internal Subject'.
N.B. 2. In the Accusative and Infinitive construction the expression of agreement is restricted (only Number and Gender in periphrastic forms).
(ii) Object. The constituent that becomes Subject in passive sentences with two-or three-place predicates will be called Object.
(iii) Indirect Object. In active sentences three-place predicates will usually be accompanied by a Subject and an Object argument. Three-place predicates meaning `transfer', in the sense both of `communication' and of `handing over', normally also require a third argument (the `Addressee' or `Recipient'). This argument, which typically is expressed by the dative case, will be called the `Indirect Object'.
(iv) Complement. This label will be used to refer to those arguments that cannot be defined as Object or Indirect Object in the sense indicated above. As a consequence it is used both for third arguments (for example, filo (`riband') in filo collum circumdare (`to circle the neck with a riband')) and for second arguments that cannot become Subject in passive sentences, for example the argument expressed by a dative case form as with favere (`to show favour to').
(v) Finally I will also distinguish the functions Subject Complement (see ch. 1, n. 3) and Object Complement (infesta in ex. (8) on p. 7).
There is no consensus among linguists about the terminology to be used for the syntactic functions fulfilled by arguments. In fact, only the notion Subject is used more or less consistently.  The notions Object and Complement are also used to refer to any non-Subject argument of a predicate. Hence one finds expressions like `Dative object', `Ablative object', etc. The label `Indirect Object' is used quite often to denote simply any argument constituent marked by the dative case, so for example the argument of a two-place predicate
marked by the dative, which we defined as `Complement' above, as in favere alicui. In this book I have chosen formal criteria to define the syntactic functions of arguments. Agreement is such a criterion, used to define the Subject (however, this criterion is not always sufficient as we noted above). Passivization is another criterion which can be used to define the Object and to distinguish it from the Complement. However, this criterion has certain defects that we will discuss shortly. Finally, the Indirect Object is not without problems either. I will discuss the matter in some detail below.
(a) If we take the possibility of passivization as a decisive criterion for identifying an argument as an Object, the italicized items in (39) cannot be considered Objects. As a consequence they will be regarded as Complements.
(39) nimis plebem amplecti videbatur (`he seemed to support the people too much')
pater gladio utitur (`father is using a sword')
pater litteris studet (`father devotes himself to literature')
pater filii miseretur (`father feels pity for his son')
me paenitet imperii nostri (`I feel ashamed of our empire')Conversely, the italicized constituent in (40) is considered an Object, since passivization is possible:
(40) hostes in custodia habebant Romani (`the Romans kept the enemies in prison')
(40a) hostes in custodia habebantur a Romanis (`the enemies were kept in prison by the Romans')If, however, habere does not mean `to keep', but rather `to own' (as in (41)), the constituent in the accusative will be described as a Complement, because there is apparently no passive like (41a):
(41) librum habeo (`I have the book')
(41a) ?liber habetur (in the sense of: ?`the book is owned by me')The criterion of passivization thus forces us to assume different syntactic functions for the second argument depending on the meaning of the predicate. But the problem is even more difficult. It is not self-evident which verb forms are to be considered passive verb forms. In our final example in (39) me paenitet imperii we took imperii as a Complement. However, expressions like res paenitenda are completely viable. If we do not regard the gerundive as a passive form, as many people would prefer,  only finite passive verb forms can be used as a criterion for establishing the function of a second argument.
(b) With three-place predicates I made a distinction above between third arguments in the `Indirect Object' function and other third arguments, by
definition occurring in the function Complement. However, in the definition a semantic element has been introduced which is lacking elsewhere. The label was said to apply to arguments of predicates with the meaning of `transfer' expressed in the dative case. The notion would apply, therefore, to arguments of verbs like dare (`to give'), adimere (`to take away'), dicere (`to tell'), that is, verbs of `giving' and `saying', as well as to third arguments with verbs like necem moliri (`to meditate murder'), copiam facere (`to create an opportunity'). However, the notion does not apply to the third argument, expressed by the ablative, with verbs like circumdare (`to circle'). My reason to use this half semantic/half syntactic definition is that it is more or less in line with the terminology in Latin grammars and in itself is not ambiguous.
In the previous sections I discussed the syntactic functions arguments may fulfil. In the following sections I will pay attention to the semantic functions or semantic roles of arguments. In this book the following roles will be identified:
Agent: the entity that controls a certain action or position. 
(42) egredere ex urbe, Catilina, libera rem publicam metu; in exsilium … proficiscere (`leave the city, Catilina, free the commonwealth from fear, go into exile', Cic. Catil. 1.20) 
(43) rem tene, verba sequentur (`concentrate on the content, the words will follow', Cato fr. 80.2 J.)
(44) Romani hostes in custodia habebant (`the Romans kept the enemies in prison')
Patient: the entity to which a certain action is applied or to which something happens.
(45) see rem publicam in (42)
(46) veteres leges aut ipsa sua vetustate consenuisse aut novis legibus esse sublatas (`the ancient statutes have either sunk into the decrepitude of their old age, or been repealed by modern legislation', Cic. de Orat. 1.247)Other semantic functions of arguments are the following:
Cause (for example: the wind opened the door)
Recipient (with dare)
Addressee (with dicere)
Direction (with se conferre)
Place (with versari, habitare) 
Following Dik (1978), still another semantic function (`Zero') may be recognized, for example Alexander in Alexander erat rex Macedonum. It
should be pointed out that contemporary linguistic theories offer different opinions both on the number and on the definitions of semantic functions. 
There is no one-to-one relationship between syntactic and semantic functions. The examples given above illustrate the fact that both arguments with the semantic function Agent and arguments with the semantic function Patient may fulfil the syntactic function Subject: not every Subject is an Agent nor is every Agent a Subject. The latter point may be illustrated with example (47): the Agent of pultando is identical with that of confregi:
(47) pultando … confregi fores (`I smashed the panels, pounding', Pl. Mos. 456)
With the action pultando an Agent is implied. However, no overt Agent is expressed. Therefore (47) has no constituent that can be labelled `Subject' of pultando.
The content of a nuclear predication derives from the meaning of the predicate and the meaning(s) of the arguments belonging to the predicate which fulfil a particular semantic function with respect to that predicate. The predicate and its argument(s) together denote a certain situation or event. From now on I will use the label `state of affairs' instead of `situation' or `event' etc. The content above all depends on the meaning of the predicate. Depending on its meaning (or meanings), each predicate has one or more `predicate frames' determining the number of arguments and the semantic functions they may fulfil. In order for lexemes to fulfil a specific semantic function with respect to a predicate they must have specific semantic features. To put it differently, with each predicate and predicate frame certain selection restrictions are associated that specify which classes of lexemes can occur as argument(s) with that predicate. Thus the semantic function Agent can only be fulfilled by lexemes having the semantic feature `human being'. 
The interrelatedness of the meaning of the predicate and the arguments can be illustrated with the following examples:
(48) puer mature dentiit (`the boy cut teeth early')
(49) puer sibilabat (`the boy was hissing')
(50) ventus sibilabat (`the wind was hissing')Dentire differs from sibilare in so far as the boy's getting or not getting teeth does not depend on any active involvement of the person referred to by puer: puer has no `control' over its taking place. With sibilare, however, puer has control as to (or can control) whether the action takes place or not. As a consequence, `the boy' in (48) cannot be interpreted as an Agent whereas this is perfectly possible in (49). In (50), again, ventus cannot be interpreted as the
Agent, because the wind cannot control somebody or something. The semantic function Agent is reserved for human beings. Ventus can better be interpreted as Cause. The predicate sibilare allows both an Agent and a Cause, apparently.
We now turn to the various types of predications that can be distinguished. Following Dik (1989: ch. 4) I distinguish four types, based on the presence or absence of two properties, viz. `control' and `dynamism'. I will now discuss these two properties in some detail. A more detailed classification will be offered in chapter 11. 
The discussion of examples (48) and (49) in the previous section has shown that some actions or situations are called `controllable', if a person is capable of bringing about or not bringing about the action or situation. The predicate sibilare (`to hiss'), for example, is controllable, dentire (`to cut teeth') is not. It follows that someone can be asked or ordered to do something he or she is capable of bringing about. Someone can also undertake or promise to do something or not to do something. When someone has control over a certain action or situation, an imperative verb form can be used to ask that person to take care of it, given the fact that he or she is capable of doing it. The imperative verb form, however, is not feasible, if it cannot reasonably be expected that a person is capable of bringing about the event or situation desired. Consider the following examples:
(51) John, come here!
(52) *John, fall asleep A similar restriction holds for predicates governed by specific predicate types (see crosssection 7.1.2. on p. 100). Latin examples which illustrate this point are:
(53) Is (Orgetorix) sibi legationem ad civitates suscepit. In eo itinere persuadet Castico … ut regnum in civitate sua occuparet quod pater ante habuerat (`he took upon himself an embassy to the community. In the course of his travels he persuaded Casticus to seize in his own state the kingship which his father had held before', Caes. Gal. 1.3.3)
(54) *persuadet Castico ut regnum haberet (`he persuades Casticus to possess the kingship')Examples (49) and (50) show that the fact alone that the predicate sibilare is controlled does not imply that the nuclear predication as a whole is controlled as well.  In order for this to be the case, an Agent is required who really controls the state of affairs. The distinction between `controllable' and `non-controllable' states of affairs can be established quite easily with the help of the imperative verb form test or a test by which a predication is made dependent on a predicate of `ordering' (that is, by creating an embedded clause with a
predicate of ordering, or checking whether instances of this are attested). The distinction between controlled and non-controlled nuclear predications appears to be useful from another point of view as well, viz. to assess the possibility of combining such predications with specific satellites. The so-called Dativus commodi, for example, occurs only with controlled nuclear predications. We will turn to this in chapter 3.(b) Dynamic/non-dynamic
Example (48) contains the description of a change in the boy's physical condition:
(48) puer mature dentiit (`the boy cut teeth early')The boy originally had no teeth. Dentiit indicates that this toothless stage gradually came to an end and was followed by one in which the boy had a set of teeth. In (55):
(55) puer mature dentes habebat (`the boy had his teeth early')it is not the process of teething that is described, but rather the final result of this process, that is, the state a person is in who has a set of teeth. I will call dentire `dynamic' (a process under way), dentes habere non-dynamic (a state). Just as in the case of control, this distinction is important for the types of satellites that may be added to a nuclear predication. Non-dynamic nuclear predications may be expanded with satellites that denote duration (e.g. (56)), while with dynamic nuclear predications this is not always possible, e.g. (57):
(56) duodequadraginta annos tyrannus Syracusanorum fuit Dionysius, cum … (`for thirty-eight years Dionysius was tyrant of Syracuse', Cic. Tusc. 5.57)
(57) *tres menses opus perfecit (*`he finished the work for three months')Dynamic nuclear predications, on the other hand, may be expanded with satellites indicating the time consumed in bringing about the change (`time within which'), as in (58):
(58) ipse … Tarraconem paucis diebus pervenit (`he arrives at Tarraco in a few days', Caes. Civ. 2.21.4)
(59) tribus mensibus opus perfecit (`he finished the work in three months')More will be said about the distinction `dynamic'/`non-dynamic' in crosssection 11.1.1. on p. 214.
Along the two parameters of control and dynamism nuclear predications can be organized into four types of states of affairs, as is shown in figure 2.1. Latin examples which have already been noted are: for Action (42), Position (44), Process (46) and State (56).
Figure 2.1: Typology of states of affairs
Many languages possess morpho-semantic means by which the same predicate can be used in more than one state of affairs. Voice (in Latin: active/passive variation) is such a means. Quite often passive sentences are regarded as no more than counterparts of the active sentences in which basically the same semantic relations exist between the arguments and the predicate:
(60) pater filium laudat (`the father is praising his son')
(61) filius a patre laudatur (`the son is being praised by his father')In examples like these the `counterpart' description seems adequate. Both sentences describe the action of praising, though each from a different perspective. However, such examples only illustrate part of the problem at hand. We have already seen (p. 9) that more often than not a constituent denoting the Agent is absent. Quite frequently no Agent is presupposed either, as in (62):
(62) Gallia est omnis divisa in partes tres (`Gaul is a whole divided into three parts', Caes. Gal. 1.1.1)The sentence describes the state Gaul is in and no action by some Agent is implied. (Some people might even prefer to regard divisa as a (stative) adjective.) Furthermore, the passive also occurs with verbs that have no Object, like itur (`they go'), acriter pugnatum est (`they fought fiercely'), the so-called impersonal use of the passive. The passive can be considered a means to leave the Agent out of account, or, more generally speaking, a means to reduce the number of arguments of a predicate. 
We have seen so far that nuclear predications can be described in various ways:
– quantitatively, according to the number of arguments that are required with a certain predicate. We have seen that one predicate can have more than one valency;
– qualitatively, according to what kind of arguments can co-occur with a certain predicate, that is, according to predicate frames. We have already noted that many predicates have more than one predicate frame. Such a description might be in terms of the syntactic functions of the arguments (for example misereri (`to feel pity') is a two-place predicate with a predicate frame containing a SUBJECT + COMPLEMENT: pater filii miseretur – `the father feels pity for his son'). It might also be in terms of the semantic functions of the arguments.In tables 2.1–2.3 I give a number of examples of nuclear predications. The examples provide information about the quantitative valency, the predicate frame and the syntactic function of each argument. Complications like passivization are left out of account.  Following Scherer (1975), I make a distinction between three groups of predications:
(i) `normal' nuclear predications with verbs that denote Actions, Processes, Positions and States.
(ii) nuclear predications with a copula and with verbs like putare (`to consider') and appellare (`to call').
(iii) so-called impersonal expressions.A complete description of nuclear predications should contain more than the number of arguments and their syntactic functions. We might also expect information about the syntactic category of the arguments in their respective syntactic function (for example, that the Object constituent with dicere may be a noun, a pronoun, a clause (Accusative and Infinitive), etc. – see ch. 7). A third requirement relates to the semantic functions of the arguments. The fourth refers to the semantic category of the constituents (for example: `animate'/`inanimate').
N.B. The type of information mentioned above should also be available in a well-organized dictionary. However, this is not enough. Further important information would relate to:
– which satellites may be added to a nuclear predication (see crosssection 3.1. on p. 25)
– the relative frequency of possible meaning differences and predicate frames
– the historical development
– stylistic restrictions on the use of certain predicates or of certain predicate frames.
It is relatively easy to prove that mono- and bivalent predicates are indeed mono- and bivalent. 19a In the case of three-place predicates this is much more difficult. Subject and Object have a special position among arguments in that they are involved in passivization.  In our texts the predicates that have been classified as three-place are often accompanied by only two arguments, because the third argument can be retrieved from the preceding context. The argument marked by the accusative seems to be omitted less frequently than dative or ablative constituents.
|sunt||bis septem nymphae||mihia|
|iussit||Caesar||milites||suum adventum adspectarec|
|hortatur||Caesar||suos||ne animo deficiant|
[added 12-08: Garcia Hernandez (1992a,b) draws attention to the parallelism between aliquid est alicui and (ego) do aliquid alicui.
The dativus possessivus + esse + subject construction cannnot be exchanged for the habeo + acc. construction. In the classical period abstract subjects predominate in the dat. poss. construction, which are relatively infrequent in the other expression. In Plautine Latin, concrete subjects are normal. See Löfstedt (1963). The same goes for the replacement by ad-expressions.
The adnominal use of the possessive dative (instead of a genitive) is typical of Merovingian and Danubean Latin (Iliescu-Macarie 1964). It is especially present in Western Balkan Latin (Herman 1965). Its frequency can be explained bij the incertainty of the formation of genitive forms of local proper names in -o and -r. In the African documents in Diehl there seem to be no instances, neither of the use of the adnominal ad-expression (Gaeng 1992: 120, 123).
The dativus possessivus is directly related to the use of the dative with trivalent verbs, according to Herslund (1988: 293). He also relates the dativus auctoris to these two usages, instead of connecting this to the dativus commodi, as Ernout-Thomas do: - dat mihi librum - mihi est liber - mihi consilium captum iamdiu est]
b A passive counterpart can be found in Ov. Pont. 4. 12. 16 ridear (`I should be laughed').
c For the construction with iubere see also crosssection 7.4.3. (on p. 128).
d In this class of verbs governing a so-called double accusative it is not always easy to determine which constituent is omissible and which one becomes Subject in a passive sentence. The reason is that apart from a three-place predicate frame such verbs often also have a two-place predicate frame in which either the person or the thing may occur as Object. In the three-place frame of some verbs (for example docere `to teach') only the person can become Subject in the passive construction. With other verbs, for example flagitare (`to demand'), this is not entirely unambiguous:
– Petreius atque Afranius cum stipendium ab legionibus … flagitarentur (`when the legions were demanding their pay from P. and A.', Caes. Civ. 1.87.3; mss. flagitaretur)
– Consules … ut referrent flagitati sunt (`the consuls were appealed to to promote a measure', Cic. Red. Pop. 11)
– flagitabatur ab eis cotidie cum querelis bonorum omnium, tum etiam precibus senatus, ut meam causam susciperent (`they (the consuls) were daily importuned by the complaints of all patriotic men, even by the entreaties of the Senate, to take up my cause', Cic. Sest. 25).
Tables 2.2 and 2.3 contain verbs that belong to different classes, for example intransitive verbs like esse (copula), fieri and manere, transitive ones like habere and putare, and also causative verbs like efficere. Scherer calls
sentences in which these verbs occur `Bestimmungssätzen' (`specifying sentences'). In such sentences the Subject constituent with esse etc. or the Object constituent are specified by a noun, an adjective or something comparable. The nuclear predications denote various states of affairs. In table 2.2 (esse, fieri, manere, etc.) we find nuclear predications which following our classification on p. 19 (see fig. 2.1) we would call Position (contentus esse), Process (fieri) and State (carus esse). The nuclear predications containing habere and putare denote Positions, those with facere and constituere Actions.
|Verbal Part||Subject Complement||Subject||Complement|
a Describing in integro and sic as a Subject Complement is not unproblematic. One might ask whether esse in this type of expression is a copula (see chapter 1, note 4). My description is supported by instances of coordination of adverbs with unambiguous Subject Complements, as in:
– quicquid natura tradit, et aequale omnibus est et statim (`whatever Nature communicates belongs equally to everyone and comes immediately', Sen. Ep. 121.20).
|cognoverant||()||eum||magni animi/parato animo|
a [modified 12-08] The English edition has as an example: facit Homerus Laertem agrum colentem. Parallels for this use of facere (= fingere) can be found in TLL s.v. 117.40 ff. Lambertz (1982: 375) regards it as a causative construction, to be compared with the infinitive construction in example (33e) on p. 111. He refers to Cic. Tusc. 5.115 Polyphemum Homerus cum immanem … finxisset, cum ariete … colloquentem facit eiusque laudare fortunas (`but Homer, having represented Polyphemus as a savage monster, depicts him also as conversing with a ram and congratulating it on its good fortune'). It seems better not to regard this use of facere as causative.
|licet||me isto tanto bono uti|
The so-called impersonal expressions (table 2.4) constitute a very heterogeneous class. 20a It contains both subjectless verbs like pluit (`it is raining') and concurritur (`they are running to the same spot') and verbs that only occur in the third person singular. Pluit, tonat and other meteorological verbs are treated as zero-place verbs in spite of poetical and popular etymological  expressions [add like Iove tonante (`when Jupiter is thundering', Cic. Phil. 5.7). I have two reasons for doing so: (a) we do not find obligatory Attributes with derived verbal nouns (nominalizations), as we do in, for example, amor patris (`the love of the father'); (b) we do not find this type of expression in the Accusative and Infinitive construction with molestum est (`it is annoying') and comparable expressions. This is always possible with one- and more-place verbs. 
We do not have sufficient information about the relative frequency of the various types of nuclear predications. From the data collected for the chapter on the case system (chapter 5) it appears that two-place predicates are much more frequent than three-place ones. They are also more frequent than one-place predicates.
As to the lexical category of the arguments, it is clear that animate entities frequently appear in the syntactic function Subject, in particular in the case of two-place and three-place predicates. This can be interpreted as a reflex of the tendency of men to talk about how people are and what they do. 
A thorough investigation of the valencies of Latin verbs, and the ways to determine valency of verbs in general can be found in Happ (1976). More or less severe criticism is expressed by Bolkestein (1977a), Guiraud (1978), Serbat (1978) and Vester (1981) in their reviews of Happ's monumental study. For the notion `omissibility' see Pinkster (1972c: 76); on `expanded valency' see Korhonen (1977: 194–6). The do so test is discussed by Happ (1976: 401–10). Data relevant to this test can be found in Lodge's Lexicon Plautinum, s.v. facere, p. 590–2; TLL, s.v. facere 107.31 ff.; Thesleff (1960: 20–1). Context types in which arguments can be omitted more easily are mentioned by Happ (1976: 239–61). A statistical approach to the valency problem is provided by Greule (1982: 206–19). Valency dictionaries have been composed for German by Helbig & Schenkel (1969) and Sommerfeldt & Schreiber (1977; 1980). Favarin (1979) has a proposal for Latin.
For the notion `state of affairs' I refer the reader to Dik (1989). Lists of nuclear predications in Latin are given by Happ (1976: 548–80) and Scherer (1975: 126–45). For zero-place verbs see Rosén (1983: 195).
The semantic functions distinguished in this book are discussed in detail in Dik (1978) (he uses the label `Goal' for my term `Patient'). Recent discussion on the passive can be found in Bolkestein & Risselada (1987), Flobert (1975: 534–65) and Pinkster (1984a; 1985a).
Pinkster, Harm (1942-) , Latin Syntax and Semantics [info], xii, 320 p.: ill.; 24 cm. [word count] [Pinkster].