Chapter Action Nominal Constructions

by Maria Koptjevskaja-Tamm

1. Introduction

Map 62A shows the distribution across the world’s languages of constructions corresponding to the English John’s running and the enemy’s destruction of the city. These are action nominal constructions (ANCs), i.e. constructions which have an action nominal (AN, running, destruction) as their head and contain a reference to the participants in the situation designated by that AN. ANs themselves are either nouns or at least occur in typical nominal positions and show inflectional properties and/or combinability with adpositions typical of nouns. They are, however, in some reasonably productive way derived from verbs, either derivationally or inflectionally, and refer to events and/or facts, i.e. not just to actions, as the name might imply.  The issue at stake is whether words like running and destruction exist in a language at all and if so, how they combine with their arguments – such as John  (the S argument), the enemy  (the A argument) and the city  (the P argument). (See also Chapter 102 for the notions S, A and P.)

In the examples above, for instance, the ANCs are very similar to possessive noun phrases in that the S and the A appear as ‘s-possessors (cf. Peter’s house) and the P appears as of-possessor (cf. the house of my parents). The whole ANC the enemy’s destruction of the city  looks very similar to the NP the enemy’s map of the city, headed by an ordinary underived noun. On the other hand, these two ANCs differ significantly from the corresponding finite clauses John runs  and The enemy destroyed the city, in which neither S nor A nor P needs any marker for signalling their syntactic functions. Such similarities between ANCs and possessive NPs, as opposed to the corresponding clauses, are by no means universal, and even in English there are other types of ANCs, such as Peter’s singing the Marseillaise  and the destruction of the city by the enemy, where the argument marking is to a certain extent analogous to that in finite clauses.

Comparing the internal structure of ANCs with that of NPs and finite clauses has been a hot issue in the last three to four decades, especially in the formalist tradition, and the literature is too vast even to be listed here. The main cross-linguistic studies of ANCs are Comrie (1976a), Comrie and Thompson (1985) and Koptjevskaja-Tamm (1993, 2003); the latter two works provide numerous cross-linguistic generalizations relating to ANCs.

There is a long tradition of distinguishing between derivational action nominals (like conquest, refusal  and arrival ) and inflectional action nominals, like English gerunds (Chomsky 1970). This distinction will not be made here because it is precisely action nominals that in many languages pose serious problems for a clear-cut distinction between derivational and inflectional forms, in that the various criteria suggested for distinguishing inflection and derivation clash when applied to them (Comrie 1976a, Koptjevskaja-Tamm 1993: 263-266, and Haspelmath 1996). There is on the whole relatively little consensus concerning the terminology to be used for what is called action nominals  and action nomional constructions  in the present chapter. Thus, ”verbal nouns”, “gerunds”, ”nominalizations”, ”masdars”, ”infinitives” often, though not necessarily, refer to ANs (Koptjevskaja-Tamm 1993: 22-43). Grimshaw (1990) uses the term "complex event nouns".

ANCs contrast with clausal nominalizations, which are not considered in this chapter. In ANCs, by definition, it is the head itself that has been turned into a noun, or nominalized, and thus represents a kind of lexical nominalization. In clausal nominalizations, on the other hand, it is the entire finite clause that gets nominal inflectional features, so that its head cannot be considered a noun (Comrie and Thompson 1985: 392), as in (1) from Imbabura Quechua.

(1) Imbabura Quechua (Cole 1982: 43) 

Pedro

ya-n

[ñuka

Agatu-pi

kawsa-ni]-ta.

Pedro

think-3

I

Agato-in

live-1- acc

'Pedro thinks that I live in Agato.'

More specifically, the verb in such nominalized clauses normally retains all its verbal characteristics (such as person and number inflection) and does not occur with nominal derivational affixes, although the nominalized clause as a whole may to some extent be treated like an NP. Clauses may be nominalized in different ways, for instance by taking an article or by attaching nominal inflectional suffixes to finite verbs, such as the accusative case in (1). In the latter case, the nominalized verb inflects in the same way as non-derived nominals, but there may still be arguments against considering it as an action nominal. In a few cases, however, the distinction between ANCs and clausal nominalizations is difficult to draw.

Most often ANCs serve as complements to predicates and refer to propositions, facts or events (to use the semantic taxo-nomy launched by Vendler (1967)). Verbs such as to assert  and to believe  combine with ANCs referring to propositions, the verbs to know  and to regret  take ANCs referring to facts, while the verbs to hear  and to continue  take ANCs referring to events. The exact range of the meanings and functions typical for the ANCs, as well as their frequency in actual discourse, varies considerably across languages and can only be understood fully in the context of the overall system of clausal complementation in the given language. Accordingly, finding examples of ANCs may be more or less easy in different languages.

2. ANCs vs. clauses vs. noun phrases

ANCs show considerable variation in their internal structure, i.e. in the way ANs combine with their arguments. And there is a clear reason for variation here, both within a language and cross-linguistically: ANs are intermediate between typical verbs and nouns, and ANCs are intermediate between clauses and NPs. 

Thus, verbs typically refer to events which involve various participants, they normally take arguments and function as predicates in clauses. Clauses can typically occur on their own, assert the occurrence (or non-occurrence) of events and locate them temporally. Nouns typically refer to persons, objects, places and other more or less concrete entities. They function as heads in noun phrases, which are typically used as arguments to predicates (e.g., as subjects and objects). Nouns do not normally take arguments (with the possible exception of kin terms and other relational nouns), but often take various attributes. Possessive NPs, like Peter’s house, cover numerous relations and constitute, in a sense, an unmarked, or default, option for combining a noun with one dependent nominal at a time.

Now, ANCs are semantically very close to finite clauses: thus, both John’s running  and John runs  can describe one and the same event involving the same participant, and the same goes for the enemy’s destruction of the city  and the enemy destroyed the city. However, ANCs, as opposed to normal finite clauses, merely name the event and occur in functions typical of NPs, for instance as arguments to predicates (as in The enemy’s destruction of the city led to a complete collapse of the country’s economy). In these functions ANs themselves inflect or combine with adpositions in the same way as normal nouns do, while they may (and typically do) lack verbal characteristics such as mood and tense markers. There is, however, often a need to refer to the participants in the situation described by an AN, i.e. to combine the AN with one or more arguments - a feature that clearly distinguishes them from normal nouns, which typically lack arguments. This closeness of ANs both to verbs and to nouns manifests itself in the structure of ANCs: in some languages ANCs look more or less like other NPs, while in others they show clear similarities to finite clauses.

Thus, cross-linguistically ANCs may be classified in accordance with their sentence-likeness and/or NP-likeness, i.e. with the degree to which the relations between an AN and its arguments are signalled in the same way as verb-argument relations in finite clauses (“sentential marking”) or as head-dependent relations in NPs. The typology of ANCs assumed here is based on the marking of the arguments themselves (by means of case-marking and adpositions) as a primary parameter, in a few cases supplemented by head-marking as an additional parameter (see Koptjevskaja-Tamm 1993: 58-60 for extensive discussion). The main cross-linguistic generalization here is as follows: 

ANC Universal: ANCs as parasites 
In all languages where ANCs exist, they use marking (both dependent-marking and head-marking) which is also used in either finite clauses or possessive NPs.

In other words, ANCs, in a sense, do not exist as autonomous, independent constructions, but always "borrow" syntactic means from finite clauses and/or NPs. 

3. Defining the values

Values of Map 62A. Action Nominal Constructions
Go to map
 ValueRepresentation
Sentential: dependent-marking of the finite clause is retained for S, A and P 25
Possessive-Accusative: S/A treated as possessors, P retains sentential marking 29
Ergative-Possessive: S/P treated as possessors, A treated differently 21
Double-Possessive: All major arguments treated as possessors 7
Other: Minor patterns 6
Mixed: Several patterns in the same language 14
Not both A and P in the same construction 24
No action nominals 42
Total: 168

All the values will be illustrated by transitive ANCs, i.e. those involving A and P. The first value (the Sentential ANC type) is exemplified by Godoberi (Daghestanian; eastern Caucasus). Example (2) shows a (di-)transitive ANC where the major syntactic functions are dependent-marked according to the same ergative-absolutive pattern as would be involved in a corresponding finite clause: S and P are in the nominative, A is in the ergative, and the recipient is in the dative case.

(2) Godoberi (Kazenin 1996: 160) 

aHmadi-di

maHamadi-łi

rec'i

ik̄-ir

 

Ahmad-erg

Mahamad-dat

bread.abs

give-an

 

'Ahmad's giving bread to Mahamad.'

Example (3) from Meadow Mari (Uralic; Russia) illustrates value 2 (the Possessive-Accusative type), whereby the A (and the S) are treated as possessors, while the P retains its sentential form. In (3), the A takes the genitive case and the AN attaches the corresponding possessive suffix, according to the general pattern of possessive NPs in the language, while the P takes the accusative case used for direct objects in finite clauses.

(3) Meadow Mari (Natal’ja Serdobol’skaja, p.c.) 

möj-ön

pis’ma-m

voz-öm-em

I-gen

letter-acc

write-an-1sg.poss

’my writing of the letter’

Example (4) from Russian illustrates value 3 (the Ergative-Possessive type), whereby the P (and the S) are treated as possessors, here appearing in the genitive case, while the A is treated differently, as a more peripheral NP (for instance, here as the agent in passive sentences, but not necessarily so).

(4) Russian

ispoln-enij-e

sonat-y

pianist-om

perform-an-nom

sonata-gen

pianist-instr

’the performance of the sonata by the pianist’ 

Value 4 (the Double-Possessive type), whereby all the S, A and P in ANCs are treated as possessors, is illustrated by example (5) from Tukang Besi (Austronesian; Sulawesi, Indonesia).

(5) Tukang Besi (Mark Donohue, p.c.) 

te

basa-'a

nu

La

Petrus

nu

boku

core

read-an

poss

prop.m

Peter

poss

book

’Peter's reading of the book.’

Value 5 encompasses several minor patterns, each having a very restricted cross-linguistic distribution. Thus, in (6) from Chamorro  (Austronesian; Micronesia), the A ('Juan') is treated as a possessor, while the P (‘tuba’) appears with the preposition ni, otherwise used for various kinds of obliques, such as instruments and passive agents.

(6) Chamorro (Topping 1973: 221, Sandra Chung, p.c.) 

ginimen 

Juan 

ni 

tuba 

def.art

drink.an

Juan

obl

tuba

‘Juan's drinking of the tuba (= a traditional drink)’ 

Value 6 is ascribed to languages that have several different types of ANCs. For instance, English has ANCs of the Possessive-Accusative type (Peter’s singing the Marseillaise), of the Ergative-Possessive type (the destruction of the city by the enemy) and of the Double-Possessive type (the enemy’s destruction of the city).

A number of languages have action nominals, but these can never combine with both A and P within one and the same construction – these languages are shown with value 7 on the map. Sometimes there are strategies for expressing both A and P at the same time without, however, making both of them syntactically dependent on the AN. Thus, the English ANC X’s buying of the dress  may be rendered by constructions like “X’s dress-buying” (where the P and the action nominal form a compound), or “the buying of the dress that X did” or “the buying that X bought the dress”. The latter option is illustrated by (7) from Yoruba.

(7) Yoruba (Ayo Bamgbose, p.c.) 

rírà 

tí 

Olú 

ra 

aso 

buy.an

rel

Olu

buy

dress

‘Olu’s buying of the dress’ 

Finally, some languages completely lack action nominals as a class – these have the value 8. Their closest correspondents to ANCs vary: examples are finite subordinate clauses in Chalcatongo Mixtec, clausal nominalizations in Slave and many other North American languages, and verbal nouns with defective paradigms in Gagauz (Turkic; eastern Balkan). 

4. Geographical distribution

The most striking and clearest areal tendency concerns the existence of ANs themselves. Australian languages seem to lack ANs, while often having other deverbal derivatives. North America is another area where ANs, and in particular ANs combining with arguments, are a rare phenomenon. ANs are common in both Europe and Africa, but in a number of African languages they cannot combine with A and P at the same time.

5. Acknowledgement

The work presented here was supported by grant (2002-1478) from the Swedish Research Council.