Jussi Ylikoski
Defining Non-finites: Action Nominals,
Converbs and Infinitives
Abstract
In the recent typological literature on non-finite verb forms—converbs, infinitives, action
nominals and participles—these forms have been defined both in terms of (i) their wordclasses and (ii) their syntactic functions, often without differentiating between the two
points of view. On the basis of data from Uralic and Indo-European languages of Europe,
this paper is intended to clarify and refine the definitions of action nominals, converbs and
infinitives. It appears that action nominals can be defined quite simply as verbal nouns (and
participles as verbal adjectives), whereas infinitives and converbs are better defined with
reference to their complementary functions, the difference between the two categories lying
in their relative obligatoriness vs. optionality in a sentence. Furthermore, it is argued that
the mutual relations of various non-finites are best understood by examining them from
both synchronic and diachronic perspectives simultaneously, as converbs and infinitives
often have their origins in case-marked action nominals.
1.
Introduction
In this paper, I aim to examine and develop definitions of certain types of verb
forms that in traditional terminology are characterized as non-finite.
Apparently, the most common subcategories of non-finites include forms that
have been labeled infinitives, participles, converbs and action nominals (verbal
nouns or masdars). The view that these subcategories make up a more or less
organized system of non-finites seems not to have gained much theoretical
interest until the typological studies of non-finites—especially converbs—in
the 1990's. Consequently, much of what will be said in this paper about
different kinds of non-finites and their mutual relations will be centered on
recent typological studies of converbs and the relatively scanty overall views of
non-finites there.
SKY Journal of Linguistics 16 (2003), 185–237
186
JUSSI YLIKOSKI
Even though this paper aims to contribute to the general typology of nonfinites, it must be admitted that the typological perspective adopted here is in
fact very narrow, not extending far from the confines of what has been
understood as Standard Average European (see e.g. van der Auwera 1998a:
814ff.; Haspelmath 2001). The focus is almost entirely on synthetic, suffixal
languages spoken in Europe, and for the present purposes—but by no means
universally—the notion of non-finite is largely taken for granted and
understood in its traditional sense; i.e., in contrast to finite forms, non-finites
are not usually marked for such categories as tense, mood, aspect, person or
number, and they do not function as only predicates of independent sentences
(cf. Koptjevskaja-Tamm 1999: 146). Another working hypothesis that makes
non-finites look much simpler than they actually are is the assumption that
most non-finites can easily be analyzed as belonging to one of the four
subcategories discussed here. However, it seems unquestionable that the
categories infinitive, participle, converb and action nominal are, to quote
Haspelmath (1995a: 1) on converbs, “universally applicable or crosslinguistically valid in the sense that they are found in various languages
irrespective of their genetic and areal connections, and must be seen as
belonging in some way or other to universal grammar.” The main focus of this
paper is on defining converbs, infinitives and action nominals and their mutual
relations; participles will be discussed to a lesser extent. These forms are not
examined from a synchronic point of view only, but from a diachronic
perspective as well.
The structure of the paper is as follows: Section 2 summarizes two main
approaches to defining and classifying most typical non-finites, i.e. the
attempts to define non-finites by their word-classes (“the word-class
approach”) or by their syntactic functions (“the functional approach”). In
Section 3, I discuss the problematic asymmetry between the two approaches
and try to point out some terminological and conceptual inaccuracies that seem
to lead to such asymmetry. In Section 4, I reconsider the definitions and
interrelations of the main subcategories of non-finites with reference to certain
Uralic and Indo-European languages of Europe; the functional approach to
non-finites is favored as it can also take into account diachronic facts and less
common types of non-finites. The usefulness of the word-class approach is reexamined in Section 5, where it is shown that lexicalization and
grammaticalization of non-finite verb forms appear to support the word-classbased definitions of only some non-finites, whereas others are still better
defined in terms of their syntactic functions. Section 6 presents a summary of
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
187
different (synchronic and diachronic, function-based and word-class-based)
views on non-finites, their definitions and mutual relations.
2.
Main types of non-finites and their definitions
The subtypes of non-finite verb forms that are here called the main types of
non-finites include the infinitive, the participle, the converb and the action
nominal. This is not to say that these forms are prototypical (“non-combined,”
“canonical” or “strict”; see e.g. V.P. Nedjalkov and I.V. Nedjalkov 1987: 75;
V.P. Nedjalkov 1995: 97; I.V. Nedjalkov 1998: 422, 425) in the sense of exact
one-to-one correspondence between non-finite forms and their functions, as it
may well be the case that non-finites are most typically used in more than one
syntactic function (cf. van der Auwera 1998b: 275). Rather, the abstract
concepts behind the terms infinitive, participle, converb and action nominal are
idealizations in many ways; in other words, one might call these forms ideal
non-finites. Moreover, it seems that much of the recent typological discussion
on converbs and other non-finites rests on tacit assumptions about what these
forms ideally are.
It is remarkable that it is not easy to find many languages where the
inventory of non-finites really matches the ideal system of non-finites that may
be inferred from the recent typology of non-finites. However, Hungarian may
tentatively be considered a good representative; for the time being, (1a–d)
serve to illustrate the main types of non-finites, i.e. the infinitive (1a), the
participle (1b), the converb (1c) and the action nominal (1d):
(1)
Hungarian
a.
A
lány sír-ni akar-t ~
kezd-ett.
the girl cry-INF want-PAST.3SG begin-PAST.3SG
‘The girl wanted ~ began to cry.’
b.
Egy sír-ó
lány be-jö-tt
a
szobá-ba.
a
cry-PTCP.PRES
girl in-come-PAST.3SG the room-ILL
‘A crying girl entered the room.’
c.
A
lány sír-va
jö-tt
be a
szobá-ba.
the girl cry-CONV come-PAST.3SG in
the room-ILL
‘The girl entered the room crying.’
d.
A
lány sír-ás-a
ingerel
engem.
the girl cry-AN-3SG irritate.3SG I.ACC
‘The girl’s crying irritates me.’
188
JUSSI YLIKOSKI
2.1 Non-finites as non-verbs
It is important to note that of the four non-finites that illustrate the main types
of non-finites, the last one, the action nominal in -ás/-és (1d), is not considered
an inflectional verb form in traditional Hungarian grammar, but a derived
deverbal noun instead. The same holds true for many action nominals in
western European languages: even though they can often be formed from all
verbs in an entirely regular and productive manner, and the semantic relation
between verb stems and action nominals always remains the same, they are still
considered derived nouns as they function as heads of NPs whose functions are
similar to NPs headed by underived nouns. In the descriptions of many Turkic
and Caucasian languages, however, action nominals are often treated as
paradigmatic verb forms to the extent that they are even used as citation forms
of verbs in dictionaries (e.g. in Lezgian, [Haspelmath 1996: 47]). Crosslinguistically, action nominals form a long continuum between fully productive
forms with many verb-like syntactic properties and less productive derived
nouns with various morphological, syntactic and semantic idiosyncrasies
(Comrie and Thompson 1985: 358–391; Koptjevskaja-Tamm 1993). However,
it is good to bear in mind that action nominals are granted a status of a verb
form (gerund) in traditional descriptions of Latin and English as well. There
are also many other types of (de)verbal nouns (denoting agents, results,
instruments etc.; see e.g. Comrie and Thompson 1985: 349–358; KoptjevskajaTamm 1993: 18–21; Muysken 1999: 248–250), but it seems that only action
nominals have been considered non-finite verb forms every now and then.
Participles, in turn, have been defined as verbal adjectives. At the
beginning of European linguistic tradition, grammarians such as Dionysius
Thrax and Marcus Terentius Varro viewed Greek and Latin participles as
word-classes of their own (Itkonen 1991: 193, 199). Greek
‘sharing,
partaking’ and its Latin calque participium were used to refer to the view that
participles have morphosyntactic properties of both verbs and nouns. Since the
rise of the notion of adjective in the Middle Ages, participles have traditionally
been defined as verbal adjectives, and this tradition appears to continue
unquestioned: “Participles are defined as adjectival verb forms” (KoptjevskajaTamm 1993: 42), “Participles are best defined as verbal adjectives,” “The
definition of participle (‘verb-derived adjective within a verbal paradigm’) . . .”
(Haspelmath 1994: 152).
The view that non-finites are not only verbs but also verb-derived
members of some other word-classes has been emphasized along with the
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
189
advent of the notion of converb in linguistic typology. The term converb has
been adopted to typological studies from the Altaicist tradition in order to
avoid such obscure and complex terms as adverbial participle, conjunctive
participle, gerund or gerundive (French gérondif) when speaking of non-finites
that are said to resemble adverbs to the extent that they can be defined as
verbal adverbs (especially Haspelmath 1994: 153; 1995a: 3–4; 1996: 50; van
der Auwera 1998b: 276). At least in the descriptions of Uralic languages, such
non-finites have been labeled as verbal adverbs (German Verbaladverb)
already in the 19th century (e.g. Wiedemann 1884: 176–179).
It appears that consistent attempts to define non-finites by their wordclasses have not taken place before the 1990's and Martin Haspelmath in
particular. After having published papers on participles (1994) and converbs
(1995a, 1995b), Haspelmath has continued defining them uniformly within a
more theoretical framework. In his 1996 paper, he presents the notion of wordclass-changing inflection as a partial answer to the problematic dichotomy
between the traditional ideas of inflection and derivation. In contrast to the
present consensus, he argues that inflection, too, can be word-class-changing
(or transpositional), a view already present in the writings of Charles Bally and
Lucien Tesnière in the first half of the 20th century (Haspelmath 1996: 50).
In short, action nominals, participles and converbs are seen as inflectional
verb forms that simultaneously belong to the word-classes of nouns, adjectives
and adverbs, respectively. They are considered inflectional on the basis that
their formation is (nearly) completely regular, general and productive
(Haspelmath 1996: 47); however, they have morphological and syntactic
properties of word-classes other than verbs, and in this sense they can be
analyzed as having acquired a new word-class membership. The reason to
regard them as verb forms at the same time is that they preserve the “lexeme
word-class” which determines the internal syntax of the phrase (or clause)
headed by a non-finite; at the same time, however, the external syntax—the
syntactic status of the non-finite outside its phrase—depends on its new “wordform word-class” (p. 52). In example (2) from Lezgian, the action nominal has
the internal syntax of a verb (i.e., it governs the subject wun and the adverbial
modifier fad), but it is a noun by its external syntax, which can be seen from
the ergative case suffix required by the main predicate:
Lezgian (Haspelmath 1996: 44)
[you.ABS early get.up-AN-ERG] we.ABS surprise do-IMPF
‘That you are getting up surprises us.’
190
JUSSI YLIKOSKI
In Haspelmath’s view, the phenomenon of word-class-changing inflection is
not limited to the formation of action nominals, participles and converbs, but
nouns can be inflected into adjectives, adjectives into adverbs and so on. In
other words, if converbs are seen as verbal adverbs, suffixes such as English
-ly may be seen as devices to form adjectival adverbs (beautifulA-lyADV), and
Upper Sorbian possessive adjectives, for instance, are examples of nominal
adjectives that preserve an internal syntax typical of nouns, e.g. attributive
modifiers (Haspelmath 1996: 52). At the end of his article, Haspelmath (1996:
58–62) acknowledges that the boundaries between inflection and derivation
and those between preservation and non-preservation of internal syntax are
vague; furthermore, he demonstrates that the degree of inflectionality
(regularity, productivity and generality) as opposed to derivationality correlates
with the degree of preservation of internal syntax (for exceptions from this
plausible tendency, see Koptjevskaja-Tamm 1993: 263–266). It should be
noted that while action nominals are traditionally considered deverbal
derivations, participles are still seen as part of the verbal paradigm; one
important reason for this is apparently that in European languages at least,
participles preserve the internal syntax of verbs (e.g. accusative objects) better
than action nominals.
It might already be said at this point that it appears doubtful whether
Haspelmath’s ideas about word-class-changing inflection are equally
applicable for defining action nominals, participles and converbs. Furthermore,
it is remarkable that when defining these types of non-finites as verbal nouns,
adjectives and adverbs, he does not attempt to define infinitives or their
relation to the other main types in any way.
2.2 Non-finites by their syntactic functions
It was mentioned in the introduction that the traditional definition of (non-)
finiteness includes the observation that unlike finite forms, non-finites do not
usually function as only predicates of independent sentences. Conversely, this
means that non-finites usually have other syntactic functions that might be
characterized as untypical of (finite) verbs. Therefore, it is quite
understandable that such non-predicative verb forms have been further
subdivided with reference to the various non-predicative functions they have.
One of the surprisingly few authors who define more than one
subcategory of non-finites solely in terms of their syntactic functions is Igor’
V. Nedjalkov (1998). He distinguishes only three main types of non-finites: (i)
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
191
the participle, a non-finite used in the attributive function, (ii) the converb,
used in the adverbial function and (iii) the infinitive, used in the object
function in complement clauses (I.V. Nedjalkov 1998: 421–422; see also V.P.
Nedjalkov 1995: 97 ). It is noteworthy that this division is much in line with
the common practice of subdividing finite subordinate clauses into relative
clauses, adverbial clauses and complement clauses (e.g. in Shopen 1985b and
the Croom Helm and Routledge descriptive grammars). Neither is it
uncommon to regard participial, converbal and infinitival constructions as
corresponding non-finite clauses. Table 1 provisionally describes functional
interrelations between these three types of non-finites and their finite
counterparts. It is not supposed to provide new information but only to
explicate what seem to be some of the common but often implicit assumptions
about their functions:
Non-finite verb form:
(Finite) subordinate
clause:
Syntactic function:
participle
converb
relative clause adverbial
clause
attribute
adverbial
infinitive
complement
clause
object
Table 1. The main functions of participles, converbs and infinitives (according to
Nedjalkov 1998: 421–422) and their finite counterparts (to be revised in Tables 3 and 4)
At first sight, there does not appear to be much difference whether participles
are defined as verbal adjectives or as attributes, i.e. as non-finites used in
similar noun-modifying functions as underived adjectives. In the same vein, it
might seem obvious that a verbal adverb means more or less the same as a verb
form that adverbially modifies a verb or a whole clause. However, in
comparison to the word-class approach to non-finites, in the functional
approach it is not infinitives but action nominals that are left outside the
otherwise neat division in Table 1.
In the sections that follow, I will attempt to find solutions to the partial
incompatibility between these two approaches. In other words, I am trying to
find a point of view from which all four main types—and some less typical
non-finites in addition—can be seen as forming a more or less coherent system
of non-finites. As will become clear in the next section, a great deal of
confusion has resulted from mixing word-class-based definitions with
functional (syntactic) approaches to non-finites, although it is clear that these
two approaches are interconnected in many respects.
192
3.
JUSSI YLIKOSKI
Problems and inaccuracies resulting from mixing the two
approaches to non-finites
In the preceding section, two different approaches to define non-finites were
shortly described, and it was seen that both the word-class approach and the
functional approach leave out one of the four main types, i.e. infinitives and
action nominals, respectively. The asymmetry between these two approaches
appears to have remained unnoticed in earlier literature. However, there have
been attempts to apply both approaches simultaneously; in fact, it seems that in
quite a few definitions of non-finites found in recent typological studies, these
approaches are to some extent mixed. In my view, this has led to inaccuracies
that hinder us from seeing some quite systematic interrelations between these
main types of non-finites. It will be argued below that there is too strong a
tendency to think that the word-class of a given non-finite can be deduced
solely from its syntactic functions; or vice versa, conclusions about the
functions of particular non-finites are sometimes drawn from their having
already been defined in terms of their new “word-form word-class.”
It was mentioned in the previous section that Haspelmath (1994, 1995a, 1996)
defines action nominals as verbal nouns, participles as verbal adjectives, and
converbs as verbal adverbs. To be precise, it must be added that he actually
defines these categories by their syntactic functions as well. According to
Haspelmath (1995a: 3), “Table [2] shows the parallels between the three types
of derived verb forms that are used when the verb is used in a non-prototypical
syntactic function”:
Word class:
Derived verb
form:
Noun
masdar
(= verbal noun)
Syntactic
function:
argument
Adjective
participle
(= verbal
adjective)
adnominal
modifier
Adverb
converb
(= verbal adverb)
adverbial
modifier
Table 2. Derived verb forms with different word class status (Haspelmath 1995a: 4; to
be revised in Tables 3 and 4)
In the following, I do not intend to go into details of various problems of
identifying and defining word-classes either language-internally or universally.
Word-classes may be defined by various (phonological, morphological,
syntactic, semantic, pragmatic or ontological) criteria, often by combining two
or more of them to characterize prototypical representatives of different word-
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
193
classes. In the functional-typological framework, it is customary to highlight
the importance of morphosyntactic and semantic criteria (e.g. Sasse 1993: 647–
651; Pajunen 1998: 60–61). As I will confine my remarks to predominantly
synthetic, suffixal Uralic and Indo-European languages spoken in Europe, I
presuppose—in accordance with the traditional view—that these languages
have separate word-classes of at least verbs, nouns, adjectives and adverbs,
which is a prerequisite to assuming that there might be verbal nouns,
adjectives and adverbs in a language (cf. Haspelmath 1994: 152).
As was noted earlier, I will concentrate on examining the definitions and
interrelations of action nominals, converbs and infinitives. Participles will be
discussed less extensively in Sections 4.5, 5 and 6. For now, it is enough to say
that I agree with Haspelmath (1994, 1995a and 1996), Koptjevskaja-Tamm
(1993: 42) and many others that participles are used mainly in a nounmodifying, attributive function and that they have additional features that make
it plausible to characterize them as verbal adjectives (e.g. word order position
and case/number/gender agreement with the head noun). It is, however,
doubtful whether an attributive use of a non-finite alone is a sufficient reason
to label it a verbal adjective; I will return to this in Section 6.
3.1 Action nominals
Action nominals (action nominalizations, masdars, nomina actionis; gerunds
in the Latin and English sense) are, by definition, verbal nouns, i.e.
nominalized verbs that denote actions or processes. To continue speaking of
idealized main types of non-finites, I mean by action nominals such (nearly)
fully productive and regular forms that have basically all the morphological
and syntactic properties of prototypical nouns. Admittedly, there are different
kinds of deviations from this ideal, such as the Latin gerund in -nd-, which
does not have a nominative form, or the Korean and Mongolian action
nominals, which lack genitive forms. The non-existence of some case forms
(e.g. genitive) or number marking of action nominals may sometimes result
from the fact that they are semantically impossible or inapplicable
(Koptjevskaja-Tamm 1993: 37–38). Furthermore, I wish to emphasize that in
contrast to authors like Koptjevskaja-Tamm (1993), I do not use the term
action nominal to refer to such more or less idiosyncratically derived deverbal
nouns as the English destruction, collapse or discovery (cf. KoptjevskajaTamm 1993: 3–6); in this paper, the term refers to forms such as the English
gerund in -ing and its equivalents in other languages.
194
JUSSI YLIKOSKI
One could imagine that it is not difficult to describe the syntactic
functions of prototypical or ideal (and often actual) action nominals. However,
it seems to be a widespread practice that verbal nouns are described as
functioning as arguments only. One of the most explicit statements of this view
is made by Haspelmath when explaining why his definition of converb
includes the notion of adverbial (cf. Table 2):
The definitional criterion ‘adverbial (subordination)’ is primarily intended to exclude
masdars/verbal nouns (nonfinite verb forms specialized for argument
subordination, or complementation) and participles (nonfinite verb forms
specialized for adnominal subordination). (Haspelmath 1995a: 7; emphasis mine.)
Similarly, when defining converbs, van der Auwera (1998b: 278) refers to the
feature [±argumental] that separates action nominals from non-argumental
converbs. V.P. Nedjalkov (1995: 97), in turn, asserts that an action nominal
(“gerund,” “a deverbal noun that is part of the system of verb forms”) occupies
the positions “of a nominal actant,” i.e. subject and object positions. Similar
ideas can be inferred also from Noonan (1985: 60–62, 65), as he describes the
use of action nominals as complements only.
It should naturally be obvious that nouns function not only as arguments
but in other positions as well. Perhaps the most important “additional” function
of action nominals is that they can be used as free adverbial modifiers, because
they usually inflect for all cases and function as complements of all
adpositions. The action nominals in (1d) and (2) occur in argument (subject)
positions. The action nominal constructions in (3–4) serve to demonstrate the
less emphasized functions of action nominals:
Lezgian (Haspelmath 1993: 389, 391–392; 1995a: 39–40)
[self-GEN duty-PL
conscientiousness-SRDIR fulfill-AN-DAT because]
!
village-GEN people-GEN among big respect
chairman-DAT
"
be.in-PAST
‘Since he fulfilled his duties conscientiously, the chairman enjoyed great respect
among the villagers.’
#
#
$
%
$
"
he-DAT [[Ali stay-AN
for] what do-FUT-COND] know-IMPF-NEG-PAST
‘He didn’t know what to do in order for Ali to stay.’
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
195
What makes the above definitions even more curious is that it is well known
that adverbial action nominal constructions such as in (3–4) exist and that it is
widely recognized that converb forms—adverbial by definition—tend to
develop from adverbially used action nominals (Koptjevskaja-Tamm 1993: 44;
Haspelmath 1995a: 49; 1999: 114; Tikkanen 2001: 1121). In the light of (3–4),
it ought to be clear that the bottom line in Table 2 should be revised so that
action nominals are indicated to have syntactic functions of not only arguments
but of adverbial modifiers as well. This, of course, renders the parallels
(between action nominals, participles and converbs) that Table 2 is intended to
demonstrate somewhat dubious. Moreover, the practice of describing action
nominals as “argumental” non-finites has had problematic effects on attempts
to understand the interrelations between infinitives and other non-finites.
3.2 Converbs
The definitions of converbs abound with terms such as verbal adverb,
adverbial participle, adverbial verb form, adverbial modifier, adverbial
subordination and adverbial functions, and there are some quite
straightforward statements that the word-class status of a non-finite can be
inferred from its syntactic functions:
Care should be taken to distinguish participles (= inflectional verbal adjectives) . . .
from verb forms used for adverbial subordination, i.e. verbal adverbs (Haspelmath
1994: 153; emphasis mine.)
A converb is defined here as a nonfinite verb form whose main function is to mark
adverbial subordination. Another way of putting it is that converbs are verbal
adverbs, just like participles are verbal adjectives. (Haspelmath 1995a: 3; emphasis in
bold mine.)
It should be clear that an adverbial modifying function alone does not result in
a word-class status of an adverb, although definitions of adverbs are usually
based on their syntactic functions to a much greater degree than definitions of
verbs, nouns or adjectives (Sasse 1993: 664). The so-called adverbial positions
in a sentence may be occupied by nouns in adverbial case forms, adpositional
phrases and finite adverbial clauses as well, and their semantic functions (as
modifiers of time, manner and place etc.) are approximately the same as those
of true, normally inflexible adverbs (here, now, yesterday, well etc.), which in
turn are often labeled adverbs only for lack of reasons to regard them as
members of any other word-class (cf. Sasse 1993: 664; van der Auwera 1999:
196
JUSSI YLIKOSKI
8). It is regrettable that many of the misstatements concerning the notion of
adverbial seem to result from its multiple meanings, referring to a syntactic
function (comparable to e.g. subject or attribute) on the one hand, and to
similarity or identity with the word-class of adverbs (cf. nominal or adjectival)
on the other.
Even though action nominals and participles can be considered verb
forms that have an external syntax typical of nouns and adjectives,
respectively, it appears more difficult to defend the view that converbs possess
properties allowing them to be characterized as verbal adverbs. After all, it is
not clear whether adverbs as a word-class have any specific morphological or
syntactic properties (besides the lack or scarcity of inflection) that distinguish
them from other constituents in adverbial positions. The most important reason
to view converbs as verbal adverbs appears to be the desire to see them as
analogues to verbal nouns and verbal adjectives, which are much better
established (see Section 6). Furthermore, the definitions of adverbs (or
adverbials, for that matter) represent a paradigm example of a definition whose
circularity is widely acknowledged and still accepted: adverbs are repeatedly
said to modify “non-nouns,” i.e. verbs, entire clauses, adjectives and adverbs
(e.g. Schachter 1985: 20; Sasse 1993: 663; Ramat and Ricca 1994: 290, 307;
van der Auwera 1999: 9). Thus, the term verbal adverb does not suffice to
specify that the non-finites in question are practically modifiers of verbs, VPs
or entire clauses, but not of adjectives or adverbs. (In the following, adverbial
is used to denote “ad-verbial” and “ad-sentential” syntactic functions only.)
3.3 Infinitives
In spite of the traditional idea of infinitives as part of the so-called non-finite or
nominal verb forms, they have not been labeled verbal nouns in recent
typological literature. This appears to be a correct decision as the “verbal noun
slot” (as in Table 2) is, in a sense, better reserved for action nominals, which
have essentially all morphological and syntactic properties of nouns, whereas
infinitives generally lack such properties (e.g. case inflection; see also
Koptjevskaja-Tamm 1993: 36–37). It is, however, somewhat dissatisfying to
think that the infinitive is the only main type of non-finites that does not have a
new word-form word-class and that, accordingly, one should be led to
conclude that only infinitives must be classified as verbs and verbs only
(besides Haspelmath 1995a, 1996, see Noonan 1985: 65). When it comes to
more functional approaches to non-finites, the definitions of infinitives and
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
197
action nominals look very much alike, to a degree that rather obscure
statements have arisen about the mutual relations of action nominals, converbs
and infinitives. — These problems will be discussed more thoroughly in the
following section.
It was mentioned above that I.V. Nedjalkov (1998: 421–422) considers
the infinitive one of the three main types of non-finites, and according to his
functional definition infinitives are used as objects, i.e. in complement clauses.
Other authors hold very similar views: In his cross-linguistic study of
infinitives, Haspelmath (1989) does not present an exact definition of the
infinitive, but in his view infinitives tend to originate from purposive
(adverbial) verb forms that are gradually used in different kinds of complement
clauses, as complements of manipulative verbs (‘order’, ‘cause’), desiderative
verbs (‘want’, ‘prefer’), modal predicates (‘be able’, ‘have to’), evaluative
predicates (‘interesting’, ‘funny’), and later in the grammaticalization process
as complements of verbs of thinking (‘seem’, ‘believe’), utterance (‘say’,
‘claim’) and cognition (‘know’, ‘realize’) (Haspelmath 1989: 298–299). In the
same vein, Koptjevskaja-Tamm (1993: 44) states that cross-linguistically
infinitives occur as complements of manipulative and desiderative verbs, and
according to V.P. Nedjalkov (1995: 97) infinitives typically function as
“clausal actants” of such verbs as begin, order, etc. Thus, the Hungarian
infinitive, occurring as a complement of the verbs akar ‘want’ and kezd ‘begin’
in (1a), can be considered a typical infinitive.
Complement clauses—both finite and non-finite—are often understood as
objects, but it is also common to acknowledge that there are other argumental
positions where very similar clauses occur. When discussing sentential
complementation, Noonan (1985: 42) states that complementation is
the syntactic situation that arises when a notional sentence or predication is an
argument of a predicate. For our purposes, a predication can be viewed as an
argument of a predicate if it functions as the subject or object of that predicate.
However, it should be noted that many adverbials, too, can be regarded as
arguments in the sense that they are obligatory dependents without which
sentences may be considered ungrammatical or at least elliptical, e.g. ?John
went vs. John went home ~ to sleep (cf. Itkonen 2001: 320–322). It will be
demonstrated in Section 4.3 that infinitives are best defined as non-finites that
are used in various argumental functions, including obligatory adverbials. It is
also notable that when converbs are defined as adverbial, the term is constantly
—though implicitly—used to refer to free adverbial modifiers, i.e. adjuncts.
198
JUSSI YLIKOSKI
Although the boundaries between arguments and adjuncts are by no means
clear-cut, it appears that it is obligatoriness (or van der Auwera’s [1998b: 278]
feature [±argumental]) that best characterizes the difference between the
traditional notion of infinitive and the more recent notion of converb. Adopting
this view, I wish to discard the completely unrelated view that the difference
between infinitives and converbs could be described in terms of word-class
membership (i.e. verbs vs. verbal adverbs).1 Furthermore, it appears that the
functional approach is much more applicable than the word-class approach also
when the mutual relations of infinitives, converbs and other non-finites are
examined from a more diachronic perspective.
To conclude this section, I present a preliminary revision and combination
of Tables 1 and 2. Table 3 takes into account all four main types of non-finites.
According to what has been argued above, it is the infinitives and not the
action nominals that are presented as “argumental” non-finites. It would be
superfluous to specify functions of action nominals as they are completely
deducible from the fact that action nominals are verbal nouns and, accordingly,
have essentially all the functions of underived nouns. On the other hand, both
the infinitive and the converb are defined only in terms of their complementary
syntactic functions, and they are left without designation of new, non-verbal
word-classes:
Non-finite verb
form:
Syntactic function:
“New word-class”:
infinitive
converb
argument
(= subject,
object,
obligatory
adverbial)
—
(free)
attribute
adverbial
(= adjunct)
—
participle action nominal
—
adjective noun
Table 3. The four main types of non-finite verb forms, their syntactic functions and
“new word-classes” (revision of Tables 1 and 2, to be further revised in Table 4)
The contents of Table 3 will be scrutinized in more detail in the sections that
follow. Diachronic development of various non-finites in Uralic and Indo1
Although it appears that no one has expressed such a view explicitly, see e.g. Noonan
(1985: 65) and Koptjevskaja-Tamm (1993: 25).
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
199
European languages of Europe will be used to demonstrate that there is a crosslinguistic tendency for case-marked verbal nouns to develop into various new
non-finites that usually fit the more or less established concepts of infinitives
and converbs. Deviations from these main types are easily defined with
reference to their syntactic functions.
4.
On definitions and interrelations of action nominals, converbs and
infinitives with reference to Uralic and Indo-European languages
of Europe
4.1 The functions of action nominals distribute over infinitives and
converbs
In this section, I continue to comment, clarify and refine some of the recent
statements concerning the mutual relations of action nominals, infinitives and
converbs. The following quotations will help to understand my argumentation:
. . . the verb forms called infinitive in most European and many other languages do
have a specific form and a specific meaning (Haspelmath 1989). Infinitives are
generally used (a) in complement clauses with (roughly) irrealis meaning and (b) in
purpose clauses. . . . One important function of infinitives is to mark (purposive)
adverbial subordination. . . . Thus, should we say that an infinitive is a kind of
converb? Probably not. The best-known infinitives, those of European languages,
lack one crucial converb property: these infinitives are not used primarily for
adverbial subordination, but their primary use is in complement clauses.
Evidently, we are dealing here with a continuum of grammaticalization: erstwhile
purposive forms are increasingly used in a nonadverbial complement function. The
more a purposive form moves away from its original adverbial function, the less it
can be regarded as a converb. (Haspelmath 1995a: 28; emphasis mine.)
V.P. Nedjalkov and I.V. Nedjalkov [19872] say explicitly that a converb is not an
infinitive. Haspelmath (1995: 28) would agree, but his claim that the category of
infinitive is not on a par with masdar, participle and converb is convincing: the
functions of infinitives distribute over masdars and converbs. (van der Auwera
1998b: 275; emphasis mine.)
2
Nedjalkov and Nedjalkov (1987: 75) argue that a prototypical converb, among other
things, “does not occur in the position . . . of the predicate actant,” which is the canonical
position of a prototypical infinitive (cf. V.P. Nedjalkov 1995: 97; van der Auwera 1998b:
274).
200
JUSSI YLIKOSKI
In other words, Haspelmath (1995a: 28) states that (i) converbs function as
adverbials and (ii) infinitives function mainly as complements and, to a lesser
extent, as adverbials expressing purpose. Furthermore, it is this latter function
of a purposive verb form (or a purposive converb) where true, complemental
infinitives often originate. Haspelmath does not really discuss action nominals
in this connection, but van der Auwera (1998b: 275) seems to interpret him as
if he defined converbs as adverbials, action nominals as complements, and
infinitives as both complements and adverbials. — Recall that van der Auwera
(p. 278) defines action nominals as having the feature [+argumental], thus
accepting Haspelmath’s (1995a: 7) view that they are non-finites “specialized
for argument subordination, or complementation” (see Section 3.1). One could
suspect whether Haspelmath (1995a) is really claiming that infinitives are “not
on a par” with other non-finites as van der Auwera sees it. However, in a later
paper on converbs he indeed appears to hold this opinion:
The converb is best defined here as ‘a non-finite verb form whose main function is to
mark adverbial subordination’ . . . In being an adverbial verb form, it contrasts with
three other main kinds of non-finite verb forms: participles (i.e. adjectival verb forms
used in relative clauses), verbal nouns (nominal verb forms used in complement
clauses or noun clauses3), and infinitives (which are typically intermediate
between verbal nouns and converbs in that they occur both in complement
clauses and adverbial clauses of purpose, cf. Haspelmath 1989). (Haspelmath
1999: 111; emphasis mine.)
In the following, I aim to argue for a nearly opposite view on the interrelations
of these forms. More specifically, I wish to demonstrate that, contrary to van
der Auwera’s (1998b: 275) claim that “the functions of infinitives distribute
over masdars [= action nominals] and converbs,” it is more reasonable to say
that the functions of action nominals distribute over infinitives and
converbs, and furthermore, at least from a panchronic perspective to nonfinites, it is the category of action nominals that appears not to be on a par with
infinitives and converbs.
Although nominal functions of action nominals have already been
presented in examples (1d) and (2–4) from Hungarian and Lezgian, sentences
3
The mention of noun clauses here might be taken as referring to the use of NP-like action
nominal constructions in non-argument (= adverbial) functions, too. However, Haspelmath
does not in any way spell out this possibility, and it would clearly contradict his earlier
statements about the functions of action nominals (cf. Section 3).
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
201
(5–6) exemplify the use of the Komi action nominal in -öm, a descendant of
the supposedly Proto-Uralic action nominal in *-mA:
(5)
(6)
Komi (KomiLuke 7:45)
sijö menam lokt-öm-s'an'-öj
ez
na
s/he [I.GEN come-AN-EGR-1SG] NEG.PAST.3SG yet
kok-ös
okal-öm-ys'.
[feet-ACC.1SG
kiss-AN-ELA]
‘She, since I came in, has not ceased to kiss my feet.’
dugdyv
cease.CONNEG
Komi (KomiJohn 11:31)
Mar'ja-lys'
termas'-ömön mödödts'-öm-sö
addz-öm böryn
[[Mary-ABL hurry-CONV leave-AN-ACC.3SG] see-AN
after]
jevr'ej-jas tšötš
pet-i-sny
börs'a-ys.
jew-PL
immediately go-PAST-3PL after-3SG
‘After seeing Mary leave hurriedly, the Jews followed her immediately.’
lit. “After seeing Mary’s leaving hurrying, ...”
In Komi, as in many Uralic and Turkic languages of easternmost Europe, the
action nominal constructions constitute an important part of clausal
subordination. The action nominal in -öm is a fully productive verbal noun that
preserves the internal syntax of verbs to the extent that it takes accusative
objects (kokös [5], mödödts'ömsö [6]) and adverbial modifiers (termas'ömön
[6]), even though the “subject” is usually marked with the genitive or ablative
case or a possessive suffix on the action nominal itself (menam loktöms'an'öj
[5], Mar'jalys' . . . mödödts'ömsö [6]).4 The most important thing to note about
these examples is that action nominal constructions function as objects
(accusative mödödts'ömsö), as obligatory adverbials (elative okalömys') and as
optional, free adverbials (egressive loktöms'an'öj, adpositional addzöm böryn).
As can be inferred from the English translation of (5), kokös okalömys' may be
equated with infinitives, i.e. non-finites that are used primarily for
complementation. If the governing clauses sijö . . . ez na dugdyv or She has
4
Nominative “subjects” of the -öm form are also possible, e.g. ves'kydlun verm-öm-ödz
[righteousness win-AN-TERM] ‘until the victory of righteousness’ = ‘until righteousness
wins’ and petuk-ys kytsas'-öm-ödz [rooster-NOM.3SG crow-AN-TERM] ‘until the rooster
crows’ (7a; Ylikoski 2001: 211, 221–222 n. 11). The ablative case (Mar'jalys' in [6]) is
used when the head noun (accusative-marked mödödts'ömsö) is the object of the main
predicate; this kind of complementarity resembles ordinary NPs in object positions, where
the ablative case replaces the otherwise genitive-marked possessor (e.g. Bartens 2000: 93–
94).
202
JUSSI YLIKOSKI
not ceased were left without their non-finite complements, they would remain
more or less elliptical.
4.2 Converbs
The optional action nominal constructions menam loktöms'an'öj in (5) and
Mar'jalys' termas'ömön mödödts'ömsö addzöm böryn in (6) functionally
resemble converbal constructions where non-finite heads are more or less
opaque forms committed to particular adverbial subordinating functions. It is,
in principle, possible to replace addzöm böryn with an obsolete and dialectal
converb in -mys't; addzöm böryn and the converbal addzymys't have identical
meanings of anteriority (‘after seeing’). The egressive action nominal form
loktöms'an'öj expresses the interpropositional relation labeled as ‘since’Anteriority (I.V. Nedjalkov 1998) or Terminus a quo (Kortmann 1997; 1998).
Its semantic counterparts, Terminus ad quem (‘until’) and Posteriority
(‘before’), can be expressed either by transparently case-marked action
nominals (7a) or by opaque converbs (7b):5
(7)
Komi Permyak (IO p. 55, KomiPMatthew 26:75)
a.
petuk-ys
kytsas'-öm-ödz te
kuim-is' ötkazitts'-an
[rooster-NOM.3SG crow-AN-TERM] you 3-ELA renounce-FUT.2SG
me dynis'.
I
from
b.
petuk kytsas'-tödz kuim-is' te
me dynis' sus'kis'-an.
[rooster crow-CONV] 3-ELA you I
from
renounce-FUT.2SG
‘Before the rooster crows, you will deny me three times.’
Furthermore, termas'ömön in (6) could in principle be analyzed as an action
nominal (-öm) in the instrumental case (-ön), although the formation in -ömön
has otherwise acquired additional semantic functions that exceed the meaning
of plain nouns in the instrumental case, and consequently, forms like
termas'ömön can be considered converbs as well (Fokos-Fuchs 1958: 284–
287; Ylikoski 2001: 206–207).
5
Here, as elsewhere in this paper, the term opaque refers to the opacity of a non-finite
marker in itself, even though its relation to the verb stem may be quite straightforward.
Likewise, a non-finite construction is said to be transparent when the combination of an
action nominal marker and a case suffix attached to it is morphologically and semantically
transparent.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
203
The following Bible verse is a particularly illustrative example of the fact
that the syntactic functions of converbs are exact equivalents of adverbially
used verbal nouns and underived nouns (such as proper names) alike:
(8)
Komi (KomiMatthew 1:17; also Ylikoski 2001: 212)
tadzi Övram-s'an' David-ödz
dzon'nas
das n'ol' ts'uzanvuz.
so Abraham-EGR David-TERM totally
14
generation
Vavilon
mu-ö vötly-tödz
David-s'an' das n'ol' ts'uzanvuz.
[Babylon land-ILL exile-CONV] David-EGR
14
generation
Vavilon
mu-ö vötl-öm-s'an' Kristos-ödz bara
das n'ol'
[Babylon land-ILL exile-AN-EGR] Christ-TERM again 14
ts'uzanvuz.
generation
‘So from Abraham to David there are fourteen generations; and from David until the
carrying away into Babylon there are fourteen generations; and from the carrying
away into Babylon to Christ there are fourteen generations.’
Again, the converb in -tödz (vötlytödz) is interchangeable with an action
nominal with the terminative case suffix (cf. 7a–b; see Ylikoski 2001: 212, 222
n. 13). In fact, the formation in -tödz itself consists of an earlier verbal noun in
*-t followed by the terminative suffix -ödz (see e.g. Fokos-Fuchs 1958: 295–
299). However, as *-t is not a productive suffix any longer, -tödz must be
analyzed as an opaque, indivisible converb marker.
Morphological, syntactic and semantic similarities between case-suffixed
nouns, action nominals and converbs abound in many Uralic languages as new
converbal forms seem to be constantly developing from nominally inflected
action nominals. As a result, many of the converbs retain traces of their origin
so that the boundaries between action nominal constructions and converbs
remain vague. The most common types of Finnish converbal forms can be seen
in the following “minimal sextet”:
(9)
Finnish (Nikanne 1997: 338)
Pekka tek-i
rikokse-n ...
Pekka make-PAST.3SG crime-GEN
‘Pekka committed a crime ... ’
204
JUSSI YLIKOSKI
a.
b.
c.
d.
e.
f.
juo-malla
olut-ta.
[drink-conv(“3INF.ADE”)
beer-PART]
‘Pekka committed a crime by drinking beer.’
juo-matta
olut-ta.
[drink-conv(“3INF.ABE”)
beer-PART]
‘Pekka committed a crime without drinking beer.’
juo-dessa-an
olut-ta.
[drink-CONV(“2INF.INE”)-3SG beer-PART]
‘Pekka committed a crime while drinking beer.’
juo-den
olut-ta.
[drink-CONV(“2INF.INS”)
beer-PART]
‘Pekka committed a crime drinking beer.’
juo-dakse-en
olut-ta.
[drink-CONV(“1INF.TRA”)-3SG beer-PART]
‘Pekka committed a crime in order to drink beer.’
juo-tua-an
olut-ta.
[drink-CONV(“PTCP.PASS.PAST&PART”)-3SG
beer-PART]
‘Pekka committed a crime after drinking beer.’
To begin with, it should be noted that the term infinitive in traditional Finnish
grammar refers to certain historical-morphological groupings of various nonfinites, not merely to the complemental non-finites, i.e. infinitives in the
generally accepted sense of the word. The so-called adessive and abessive
forms of the third infinitive (9a–b) are not perceived as instances of action
nominals, although it is evident that morphologically they consist of the
component -mA- followed by nominal case endings. Unlike the true action
nominal in -minen, the third infinitive “inflects” only for five or six of more
than a dozen cases in Finnish.6 The so-called second infinitive forms (9c–d) are
6
In addition to the adessive (-mAllA, 9a) and abessive (-mAttA, 9b) forms of the third
infinitive, the illative (-mAAn), elative (-mAstA) and inessive (-mAssA) forms will be
discussed in Sections 4.3 and 4.4. The so-called instructive form of the third infinitive
occurs only as a complement for the verb pitää ‘must, have to’ in the obsolete/dialectal
construction type exemplified by (i):
(i)
Sinu-n
ei
pidä
juo-man
olut-ta.
you-GEN NEG.3SG must.CONNEG drink-“3INF.INS” beer-PART
‘Thou shalt not drink beer.’
The instructive case in Finnish is a productive category in plural only; the non-finite in
-mAn, however, is historically an instructive singular form. The existence of the -mAn form
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
205
a bit more opaque, as there are only two case endings which are attached to a
morphophonologically less salient element -te- (~ -de- ~ -e- ~ -re- ~ -le- ~
-ne-), and roughly the same applies to the purposive converb or “the first
infinitive translative” in (9e). Moreover, linguistically untrained speakers
scarcely analyze the -n in (9d) as the instructive singular case suffix, an
otherwise unproductive inflectional category in the language. The anterior
converb in (9f) can be considered a completely opaque converb. Even though it
can be said to consist of the passive past participle (-tU) followed by the
partitive case suffix (-A), its meaning is hardly related to those of passive
participles or the partitive case (Nikanne 1997: 345–346). Like kytsas'tödz and
kytsas'ömödz in (7) or vötlytödz and vötlöms'an' in (8), the Finnish converbs in
(9a–f) modify the main clause in a way that resembles both ordinary NPs and
finite adverbial clauses; they, too, can express time (9c, f), means or manner
(9a, d), purpose (9e) or lack of certain circumstances (9b).
To return to the claim that converbs are verbal adverbs, it must be
admitted that in languages like Komi and Finnish, there are plenty of true
adverbs that could be used in place of the converbal constructions in (7–9).
Nevertheless, compared to verbs, nouns or adjectives, adverbs are a
heterogeneous and less open word-class in both languages and they do not
appear to have special morphological, syntactic or semantic properties that
would give reason to say that converbs are verbal adverbs. In fact, if converbs
should be labeled as verbal adverbs only because they can be said to function
as adverbial modifiers, there would not be many reasons for not labeling
ordinary case-inflected nouns like Övrams'an' ‘since Abraham’ or Davidödz
‘until David’ in (8) as “nominal adverbs,” a solution that would not make much
sense (cf. Ramat and Ricca 1994: 301–303). Similarly, the converb form
sahaamalla ‘by sawing’ in (10) is hardly more of an adverb than sahalla ‘with
a saw’, the adessive form of the noun saha ‘saw’:
(10) Pekka pieni
halo-t
sahaa-malla ~ saha-lla.
Pekka make.small.PAST.3SG firewood-PL saw[V]-CONV
saw[N]-ADE
‘Pekka cut the firewood by sawing them ~ with a saw.’
—and its passive variant -tA-mAn—is a further reason to consider the forms in -mA separate
from verbal nouns.
206
JUSSI YLIKOSKI
4.3 Infinitives
In Section 3.3, especially in Table 3, infinitives were defined as non-finites
used as arguments, whereas converbs were said to be optional adverbial
modifiers. It was emphasized that many so-called adverbials, too, are
obligatory arguments without which a sentence would remain ungrammatical,
and kokös okalömys' [feet.ACC.1SG kiss.AN.ELA] ‘from kissing my feet’ in (5)
was seen as an instance of the obligatory action nominal constructions that
correspond to opaque infinitives in other languages. It ought to be evident that
in actual language use, obligatory and optional adverbials cannot be
distinguished in absolute terms; various contextual and pragmatic factors
together with world knowledge make it possible to produce, understand and
accept highly elliptical utterances. Nevertheless, it appears intuitively obvious
that sentences like ?Pekka began, ?Pekka wanted, ?Pekka ceased, and ?Pekka
went are less complete than Pekka committed a crime or Pekka cut the
firewood.
It has already been mentioned that Haspelmath (1995a: 28; 1999: 111)
defines infinitives as having typically two separate functions: in addition to
their primary use as complements, they are often used as adverbial modifiers to
express purpose. Van der Auwera (1998b: 275) seems to approve of this view,
and due to their purposive functions, Haspelmath and van der Auwera regard
infinitives as “distributing over” or “intermediate between” action nominals
and converbs. Haspelmath (1995a: 28; 1999: 111) refers to his 1989 paper
where he shows that cross-linguistically, primarily complemental infinitives
tend to develop from purposive non-finites. According to Haspelmath (1989:
289), the first step in the grammaticalization process is that the “local allative
meaning” of a non-finite construction is extended so that the non-finite can be
analyzed as having a purposive meaning as well:
(11)
(Haspelmath 1989: 289)
a. Mary went to Sabina’s apartment.
b. Mary went to take photos of Sabina.
c. Mary bought a camera to take photos of Sabina.
Haspelmath acknowledges that to take photos of Sabina in (11b) is partly
locative in meaning, expressing the direction of motion (comparable to the tophrase in 11a) whereas in (11c), the non-finite construction cannot be thought
of as a directional, but merely as a purposive modifier. Having presented these
examples, however, Haspelmath does not refer to the differences between
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
207
directional-purposive and purely purposive non-finites at all. More specifically,
he does not pay any attention to the fact that even though both instances of to
take photos of Sabina in (11b, c) may be labeled adverbial or purposive, there
is an obvious difference in the well-formedness of these sentences if the
infinitival clauses are omitted:
(11')
b. ?Mary went.
c. Mary bought a camera.
Without going into the details of the corresponding clauses in individual
languages, I would like to point out that cross-linguistically, infinitives—i.e.
opaque non-finites used primarily as complements of manipulative,
desiderative and other “modal” verbs—are more likely to occur as somewhat
obligatory directional-purposive adverbials (11b) than as clearly optional, nondirectional purposives (11c). Haspelmath (1989: 302–303; 1995a: 28) does
note that while acquiring more and more complemental functions, the
infinitives-to-be tend to need reinforcement in order to express purpose, i.e. to
be used in their original functions. This is what has happened to the zuinfinitive in German, for instance, which has been reinforced by um.
Haspelmath does not, however, remark that purposive um zu-constructions are
used almost exclusively in sentences like (11c), not as (partly directional)
obligatory arguments for verbs of motion where it is more natural to use
unreinforced infinitives (with or without zu). Likewise, in English it is much
more natural to add the words in order to reinforce the purpose clause in (11c)
than in (11b). Note, however, that it is not uncommon to have a reinforced in
order to-clause in sentences like (11''a) where the PP to Sabina’s apartment, in
a sense, already fills the place of the directional argument of a motion verb:
(11'')
a. Mary went to Sabina’s apartment (in order to take photos of her).
b. ?Mary went in order to take photos of Sabina.
c. Mary bought a camera in order to take photos of Sabina.
It appears that it is precisely the optional non-directional purposive use of
infinitives (as in 11c) that tends to be reinforced across languages. In addition
to in order to and um zu (~ Dutch om te), Swedish för att, French pour,
Spanish para and Russian
&, for instance, are used in nearly identical
syntactic-semantic environments. In Uralic languages, similar reinforcements
can be found in e.g. Estonian (et + infinitive in -da), Mari (manyn + infinitive
in -aš), Komi (med(ym) + infinitive in -ny) and Udmurt (šuysa + infinitive in
208
JUSSI YLIKOSKI
-ny). It must be admitted that the actual boundary between reinforced and
unreinforced infinitives or their obligatoriness/optionality remains vague.
Nevertheless, it is important to note that the differentiation between
directional-purposives and (non-directional) purposives has significance in
other languages, too. In Finnish, a special converb form in -tAkse- (9e) is used
in the purposive function, whereas another form, the infinitive in -mAAn, is
used with verbs of motion as well as with many other types of verbs that
typically take infinitives as their complements (see 14a, c below). In
Hungarian, the infinitive in -ni (cf. 1a) is also used in sentences like (11b), but
in sentences corresponding to (11c), it is much more appropriate to use a finite
adverbial clause with the conjunction hogy ‘(in order) that’ instead.
Interestingly, an analogous phenomenon can be observed in Modern
Greek, a language with no infinitives. Haspelmath (1989: 305–308) presents
Greek as an example of a language where the reinforcement of complementlike purposive finite clauses resembles the reinforcement of purposive
infinitives described above. Modern Greek uses subjunctive ná-clauses in
functions that correspond to infinitival clauses in other European languages. ná
is a grammaticalized and reduced remnant of the earlier purposive marker hína,
and according to Haspelmath, in order to express the original purposive
meaning of ná-clauses, ná must be reinforced by the preposition yá (já) ‘for’.
In this context, it is intriguing to note that when describing purposive já naclauses, Joseph and Philippaki-Warburton (1987: 31) add that the “particle na,
by itself, can express purpose, especially after verbs of motion” (emphasis
mine):
Modern Greek (Joseph and Philippaki-Warburton 1987: 31)
(12) já
na nikís-i
o
jánis, prép-i
na min pés-i
[for PRT win-3SG the.M John] must-3SG [PRT NEG fall-3SG]
‘In order that John win, he must not fall.’
θ
(13) ír -a
na se
come.AOR-1SG [PRT you.ACC
‘I came (in order) to help you.’
θ
voi ís-o
help-1SG]
Examples (12–13) nicely support the view that the “purposive” clauses—finite
and non-finite alike—occurring with verbs of motion formally align with
expressions that are regarded as complements to verbs of various modal
meanings (e.g. ‘must’ + na-clause in 12, ‘come’ + na-clause in 13). The
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
209
purposive expressions that may need reinforcement are more likely to be nondirectional modifiers of non-motion verbs (já na nikísi o jánis in 12).
Similar ideas can be found in Joseph’s (1983) study of the loss of
infinitives in the Balkan languages; in defining the object of his study, he
points out that cross-linguistically the forms called infinitives are often used (i)
as complements of verbs (to volitional verbs in particular), (ii) as complements
of adjectives (e.g. French jolie à regarder, English pretty to look at), and (iii)
in expressions of purpose (Joseph 1983: 31–32). This view comes fairly close
to that of Haspelmath’s since according to him, typical functions of infinitives
include being used as complements to modal and evaluative predicates; such
predicates include both verbs and adjectives (e.g. have to, [be]
possible/able/necessary/interesting/funny) (Haspelmath 1989: 298–299).
Therefore, one can agree with both Joseph and Haspelmath and regard
infinitives as non-finites that occur (i) as complements (to verbs and adjectives
alike) and (ii) as purposive verb forms. I wish to define infinitives in even
more abstract terms, i.e. as non-finites that function as complements, in the
sense that complement covers obligatory or argumental adverbials as well.
After having explicated the functions that infinitives typically have,
Joseph (1983: 31–32) argues that the Latin supine “may best be regarded
simply as a variant form of the infinitive,” the accusative supine ( ' ( ) being
used to express purpose and the ablative ( ' ( ) being used as complements to
adjectives. Again, it is fascinating to note that the accusative supine is used
with motion verbs only, i.e. in a more or less locative meaning determined by
the main verb. Purely purposive clauses without restrictions concerning their
main clauses must be expressed by other means, e.g. finite ut-clauses. Old
Church Slavonic presents a related situation where, according to Joseph (1983:
103), the supine in
may be regarded as an “allo-form” of the infinitive in -ti,
a view already held by Meillet (1934: 242): “(le supin) n’est plus déjà qu’un
doublet de l’infinitif employé après les verbes de mouvement.” What Meillet
and Joseph seem to have in mind is that various verb forms together can
constitute a single category of infinitive. Interestingly, this suggests that there
underlies a cross-linguistically valid, albeit quite abstract, syntactic-semantic
concept of infinitive, which in turn is realized in individual languages either as
a single form or several alloforms. This view might be useful in the analyses of
various sets of non-finites in Uralic languages, too; I will return to this below.
It has happened both in Latin and Slavic that infinitives and supines have
merged together by way of loss of the supines, and the infinitives taking over
their functions. This appears quite natural. In addition to most Romance and
210
JUSSI YLIKOSKI
Slavic languages, there are plenty of languages that possess only one opaque
non-finite that is specialized to all complemental functions in question,
including the directional-purposive complements of verbs of motion. (My
purpose here is not to give an exact account of how common it is for a nonfinite to cover the functions of, say, the Latin non-finites in -re, -tum and .)
Although Haspelmath (1989: 288) opposes the traditional view that the
infinitive is in itself quite a meaningless verb form, the semantic content of
infinitives still remains vague. As complements, infinitives can be said to
express various “modalities,” but actually the semantic functions of infinitives
are largely determined by their main verbs. Considering this, it is
understandable that there is no real need for separate infinitive forms to express
modalities like irrealis-directive (e.g. want to drink), irrealis-potential (e.g.
[be] able to drink) or realis-non-factive (e.g. seem to drink) mentioned by
Haspelmath (1989: 298). Furthermore, the same appears to apply to the socalled purposive infinitives that function as complements of verbs of motion;
the relation between the main verb and the infinitive (e.g. go to drink) does not
really need to be explicated (cf. German Ich gehe (zu) trinken where the
“directional” marker zu is often omitted).
This brings us back to the major difference between (obligatory,
complemental or argumental) infinitives and (optional, adverbial or
adjunctival) converbs. Even though converbs, too, may have quite vague
meanings (see e.g. König 1995; V.P. Nedjalkov 1995: 106–109; I.V.
Nedjalkov 1998: 424, 432–439), it is clear that as they are supposed to express
diverse interpropositional relations, one and the same form can hardly be used
in too many functions. The Finnish purposive converb in (9e) is a case in point:
(9)
e. Pekka tek-i
rikokse-n juo-dakse-en
Pekka make-PAST.3SG crime-GEN [drink-CONV-3SG
‘Pekka committed a crime in order to drink beer.’
olut-ta.
beer-PART]
The converb form juodakseen is needed in order to explicate that the relation
between the two propositions is that of purpose and not of means, simultaneity
or anteriority, for instance. It is also understandable that a special form in
-tAkse- —and none of the non-finites specialized for complemental
functions—is used to spell out the purposive relation between the two
propositions conjoined. Put concretely, it may be redundant to underline the
exact relation between going and drinking beer (not to speak of wanting and
drinking beer) but much less so between committing a crime and drinking beer
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
211
where the purposive relation is unforeseeable (cf. 9a–f). This also explains why
the only infinitive of a language often welcomes reinforcement in the purely
purposive expressions (cf. in order to in the English translation of 9e), but not
necessarily in the directional-purposive expressions.
Why, then, are there situations like those in Latin and Old Church
Slavonic where the (macro-)category of the infinitive includes “alloforms” like
the supines -tum,
(Latin) and -t (OCS)? In a way reminiscent of
Haspelmath (1989: 288), I believe that the nature of these forms and their
mutual relations “can best be understood if the infinitive is approached from a
diachronic perspective.” All these forms are considered to originate in case
forms of ancient verbal nouns: like the Latin -tum, the OCS supine in -t
probably represents the former accusative form of a verbal noun; the Latin
derives from the ablative case and the infinitive endings seem to stem from the
locative (Latin -re) and from the dative (OCS -ti) forms of earlier verbal nouns
(Vineis 1998: 307, 312; Lunt 2001: 247). It appears that the functions of these
forms have been similar to the corresponding case forms of ordinary nouns in
the beginning, but after they have been analyzed as independent verb forms,
there have presumably been few reasons to have two or three opaque
complementary alloforms of semantically somewhat blank infinitives.7
In many Uralic languages, there is a richness of non-finites that are
diachronically intermediate between transparent case forms of action nominals
and fully opaque infinitives or converbs. It was shown in (5–8) that the Komi
action nominal in -öm resembles ordinary nouns in that it can be inflected for
all cases, and the case forms of action nominals function as nouns usually do
(cf. the egressive in 5 and 8, the elative in 5, the accusative in 6 and the
terminative in 7a). The Komi action nominal has formal and functional
equivalents in many related languages, but in Finnish and the other Finnic
languages, the Proto-Uralic action nominals in *-mA have survived only as
some completely lexicalized deverbal nouns such as synty-mä ‘birth’ (< ‘be
born’), elä-mä ‘life’ (< ‘live’) and juo-ma ‘drink’ (< ‘drink’ [verb]). However,
in addition to lexicalized items, the same -mA can be seen in at least five nonfinites that have been labeled the third infinitive in traditional Finnish
grammar. It was mentioned in Section 4.2 that “the adessive and abessive
forms of the third infinitive” (juomalla [9a], sahaamalla [10] and juomatta
7
Joseph (1983: 261 n. 33) adds that the Latin infinitive is, in fact, sometimes used to
express purpose; again, it can be specified that the infinitive is used in directional-purposive
functions only (see e.g. Palmer 1954: 319–320; Woodcock 1959: 18–19).
212
JUSSI YLIKOSKI
[9b]) are better seen as converbs rather than action nominals, let alone
infinitives. Two other forms of the third infinitive—the illative (14a) and the
elative (14b)—are better labeled as infinitives, however:
(14) a. Pekka rupes-i
~
pysty-i
~
tul-i
Pekka begin-PAST.3SG manage-PAST.3SG come-PAST.3SG
juo-maan
olut-ta.
[drink-“3INF.ILL” beer-PART]
‘Pekka began ~ managed ~ came to drink beer.’
b. Pekka lakkas-i
~
kieltäyty-i
Pekka cease-PAST.3SG refuse-PAST.3SG
juo-masta
olut-ta.
[drink-“3INF.ELA” beer-PART]
‘Pekka ceased ~ refused to drink beer.’
The origin of the suffix -mAAn is quite compatible with Haspelmath’s (1989)
view that infinitive forms tend to arise from purposive action nominal
constructions, although there seem to be no signs of a non-directional
purposive use of this form (cf. 9e). The infinitive in -mAstA, however, has
developed from a nearly opposite construction (“from the act of V-ing”). In
addition to Finnic languages, very similar non-finites exist in Sámi and
Mordvin languages. In many Uralic languages, certain verbs implying not to
do something (e.g. ‘cease’, ‘refuse’, ‘forbid’) require their complements in a
case with an ablative (separative) meaning ‘from’; remember example (5),
where the Komi verb dugdyny ‘cease’ takes its complements in the elative case
in a way that closely resembles English cease + from (i.e. ez na dugdyv kokös
okalömys' translates literally as ‘[she] has not yet ceased from kissing my
feet’).
Komi okalömys' is still a fully transparent action nominal in the elative
case but the corresponding forms in e.g. Finnish (suutele-masta < action
nominal + elative), North Sámi (cummástalla-mis < action nominal + locativeelative) and Erzya Mordvin (pals'e-med'e < action nominal + ablative) have
lost their noun-like transparency and/or acquired verb-like syntactic properties
to the extent that it is feasible to consider them more or less independent non-
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
213
finites (see e.g. Bartens 1979: 51–54; 1999: 150–151; Ylikoski 2002: 77–82).8
It is important to note that these new types of verb forms have been called
infinitives in descriptions of Sámi and Mordvin as well: in addition to the
historicizing label“action locative,” the Sámi non-finite in -mis ~ -mes has at
times been called the second infinitive (example 15 below; see Ylikoski 2002:
77 and references therein), and in the descriptions of Mordvin, it is customary
to speak of the third infinitive, the infinitive in -mado ~ -modo ~ -med'e ~
-mda, or the ablative infinitive (e.g. GMJa 1980: 270–271, 275–276; Bartens
1999: 150–151).
In other words, in descriptions of languages like Finnish, North Sámi and
Erzya Mordvin, there is more than one non-finite that has been considered an
infinitive. Reasons for this are often left implicit, but there appear to be indices
that make it understandable and approvable to think that there are indeed
several distinct infinitives in these languages, or—bearing in mind the views of
Meillet (1934: 242) and Joseph (1983: 103)—“alloforms” or “doublets” of a
single category of infinitive. It was noted above that in Finnish grammatical
tradition, the term infinitive is used to refer to various non-finites, some of
which could be better called converbs (see 9–10). The infinitives in
descriptions of the Sámi and Mordvin languages, however, can be thought of
as instances of infinitives in the more typological sense advocated here, i.e. as
more or less opaque non-finites that are specialized for complement functions.
On this account, there are two infinitives (-t and -mis ~ -mes) in North Sámi
and three infinitives (-ms, -mo ~ -me and -mado etc.) in Erzya. In the same
vein, one could say that there are possibly three true infinitives in Finnish,
namely those in -mAAn (14a, c), -mAstA (14b) and -tA (14c below), the latter
originating from a lative form of an ancient verbal noun and now an opaque
infinitive form (“the shorter form of the first infinitive,” held as the “basic
form” of the verb). What is common to all of these non-finites is that they are
used almost exclusively as obligatory complements of various verbs whose
semantic equivalents tend to take “infinitives” as their complements
worldwide, i.e. those listed by Haspelmath (1989: 298–299) among others (cf.
above). As was discussed in connection with Latin and OCS supines, it appears
that it is somewhat superfluous to have more than one infinitive in a language.
Since the infinitives are quite abstract in meaning, minimal pairs with distinct
8
There have, however, been attempts to analyze Komi forms like okalömys' in (5) as
instances of the so-called m-infinitives in accordance with the third infinitive in Finnish
grammatical tradition (see Ludykova 1984; Cypanov 1997: 33–34).
214
JUSSI YLIKOSKI
semantic contents are not easy to find; instead, it is not uncommon that two
infinitives can occur in free variation (for similar examples in Mordvin, see
Bartens 1979: 53; 1999: 150):
(14) c. Pekka alko-i ~
ehti
Pekka begin-PAST.3SG have.time.PAST.3SG
juo-maan
~
juo-da
olut-ta.
[drink-“3INF.ILL” drink-“1INF(.LAT)” beer-PART]
‘Pekka began ~ had time to drink beer.’
(15)
North Sámi
Máret vajáldahti-i
lohka-t
~ lohka-mis dan
girjji.
Máret forget-PAST.3SG [read-1INF
read-2INF that.GA book.GA]
‘Máret forgot to read that book.’
The major difference between infinitives and converbs is clearly visible in
grammatical descriptions: since infinitives are obligatory arguments required
by certain verbs but not by others, it is customary—and indeed relevant—to
present lists of verbs that take infinitives as their complements. This can be
seen even in Finnish grammars where different usages of the infinitives in -tA,
-mAAn and -mAstA are characterized by presenting their respective main verbs
but—despite the label infinitive—converbs such as those in (9a–e) are most
rationally described by referring to their meanings, with no attempt to
enumerate the infinite number of possible main verbs (see e.g. Karlsson 1999:
183–192).9
In discussing common grammaticalization paths of infinitives,
Haspelmath (1989: 301) refers to parallel developments in the area of nominal
case markers where it is not unusual that grammatical cases stem from cases
with less grammatical functions. An often-mentioned example of such
9
The differences between ways to describe the use of obligatory infinitives and the use of
optional converbs are reminiscent of standard descriptions of the infinitival constructions in
Romance languages such as French: The plain (prepositionless) infinitives are simply said
to occur as complements to verbs such as laisser ‘let’, pouvoir ‘can, be able’, vouloir
‘want’ and aller ‘go’ (!); de + infinitive is used with cesser ‘cease’, refuser ‘refuse’, tâcher
‘try’ etc., and à + infinitive with chercher ‘attempt’, apprendre ‘learn’, commencer ‘begin’
etc. Adverbial constructions such as après + infinitive (‘after V-ing’), avant de + infinitive
(‘before V-ing’), pour + infinitive (‘in order to V’) and sans + infinitive (‘without V-ing’),
in turn, are efficiently described by referring to their adverbial meanings only. (For more
about the untypical nature of the French infinitive, see Note 11.)
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
215
development is Spanish a, which has developed from a directional preposition
(Latin ad) to mark both indirect and (specific, animate) direct objects; in the
latter function, a is nowadays considered an accusative marker (see e.g.
Haspelmath 1989: 301; Blake 1994: 173; Lehmann 1995: 110). To continue
with parallels between case markers and non-finite verb forms, it is good to
note that grammatical (or syntactic) cases such as nominatives, accusatives,
genitives and datives do not carry their own meanings to the same extent as the
so-called semantic (or concrete) cases. Moreover, just as there is little need for
multiple infinitives in a language, there is, in principle, no need for a great
number of “alloforms of accusatives,” i.e. cases that function as second
arguments of verbs. Semantic cases, on the other hand, resemble converbs in
that they usually occur as adverbial modifiers and there may be a number of
them, with quite specific meanings. In languages with extensive case systems,
there is always a multitude of local cases (Blake 1994: 153–155). It is also
noteworthy that while both infinitives and accusatives are generally described
as evolving from directional expressions (Haspelmath 1989: 301; Blake 1994:
173; Lehmann 1995: 110–112), some object markers—and in this sense
“accusatives”—in Finnic, Sámi and Mordvin have developed from the Uralic
ablative in *-tA (see e.g. Itkonen 1972; Harris and Campbell 1995: 362–363).
This, of course, closely resembles the development of elative- and ablativemarked infinitives in the same languages (14b and 15).
A further parallel to interrelations between infinitives and converbs is
probably worth mentioning: Infinitives are often described as non-finite
counterparts of finite complement clauses and converbs as equivalents to finite
adverbial clauses (cf. Table 1 in Section 2.2). Without pursuing this matter any
further, I wish to point out that it is not uncommon to refer to finite
complement clauses as that-clauses. By comparison, none of the terms such as
while-clause, after-clause, if-clause or in order that-clause is used in the
general sense of “adverbial clause.” Needless to say, there is a need to maintain
a distinction between various adverbial conjunctions separate from each other,
whereas the majority of complement clauses may be introduced by a
semantically void all-purpose complementizer.
Finally, one must remember that in a natural language everything is in
flux. The dichotomies between infinitives and converbs, between obligatory
and optional, between arguments and adjuncts, between grammatical cases and
semantic cases, and between that-clauses and adverbial clauses are far from
clear-cut. What I have been proposing is only that the mutual relations—both
synchronic and diachronic—between infinitives and converbs resemble those
216
JUSSI YLIKOSKI
of different case markers on one hand and those of different types of finite
dependent clauses on the other, and that these relations might be best described
in terms of relative obligatoriness vs. optionality. It still remains a fact that in
many languages, expressions of purpose do formally coincide with
complements (cf. 11c); possible reasons for this may be that both the adverbial
purposives and the irrealis complements of manipulative and desiderative verbs
such as ‘order’ and ‘want’ refer to goals to be achieved in the future.
Haspelmath (1989: 299) points out that the difference between the two types is
that in the latter cases the purpose element is expressed in the lexical meanings
of the main verbs instead of their complements (cf. Bybee et al. 1994: 228–
230). Again, it might be thought that complex clauses of the type “verb of
motion + infinitive” represent an intermediate type where the directionalpurposive relation between ‘come’ and ‘drink beer’, for instance, can be
inferred from the meaning of the main verb as well as from the appearance of
its complement.
4.4 Less prototypical non-finites developed from action nominals
An attempt to present a uniform view of action nominals, infinitives, and
converbs and their interrelations should also be able to take into account
certain “less prototypical” non-finite constructions that also have their origins
in case-suffixed action nominals. Even though it was demonstrated by
examples (14c) and (15) that different infinitives—or alloforms of a single
infinitive—can occur in free variation, the Finnish infinitives in -mAAn
(illative) and -mAstA (elative) are not interchangeable. Instead, they can even
form a minimal pair when used in connection with verbs of motion: the
sentence Pekka tuli juomaan olutta (14a) ‘Pekka came to drink beer’ clearly
contrasts with Pekka tuli juomasta olutta ‘Pekka came from drinking beer’.
(Apparently, the origins of the third infinitive forms are most transparent when
the main verb is a verb of motion.)
Another less typical non-finite that etymologically parallels the infinitives
in -mAAn and -mAstA is the form in -mAssA or “the inessive form of the third
infinitive,” which is mainly used to form a periphrastic progressive with the
verb olla ‘be’ as well as to express actions that are concomitant with the states
or actions expressed by governing posture or motion verbs, e.g. Pekka on/istuu
juomassa olutta [P. be/sit.3SG drink.“3INF.INE” beer.PART] ‘Pekka is/sits
drinking beer’. These functions are a quite understandable outcome of an
earlier action nominal in a locative case (i.e. “Pekka is/sits in the act of
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
217
drinking beer”), and it appears that the relative obligatoriness/optionality of
these forms often depends on the presence of locative modifiers in a manner
reminiscent of directional-purposive PPs and infinitives in (11a–b, 11'a and
11''a) (see Hyvärinen 1982: 74–75). If infinitives and converbs are defined
only in terms of their syntactic functions (and not in terms of word-classes), it
is relatively simple to define and describe forms like -mAssA in comparable
terms.
The functional approach to non-finites also enables a more comprehensive
description of forms like -mAttA, which can be used not only as a negative
converb (a free modifier of, in principle, any well-formed sentence; see 9b),
but also as a complement of verbs like jäädä ‘stay, remain’ and jättää ‘leave’
(e.g. Pekka jätti oluen juomatta [P. leave.PAST.3SG beer.GEN
drink.“3INF.ABE”]‘Pekka left the beer undrunk’, “Pekka left the beer without
drinking”) as well as with the verb olla ‘be’ to form a kind of periphrastic
negation (Pekka on tänään juomatta olutta [P. be.3SG today drink.“3INF.ABE”
beer.PART] ‘Pekka does not drink beer today’, “Pekka is without drinking beer
today”).
It is fascinating to see that most of the relatively transparent forms of the
Finnish third infinitive have quite opaque equivalents in North Sámi. Even
though the two infinitives in -t and -mis occur in free variation with verbs such
as vajáldahttit ‘forget’ (15), they can form minimal pairs when governed by
verbs of motion (16a–b; cf. Finnish -mAAn and -mAstA above). There is also
an opaque non-finite in -min ~ -me(n) whose use closely corresponds to that of
Finnish -mAssA; it can, however, also replace the second infinitive in certain
dialects and in the literary language as well (for the origins of the suffixes -mis
and -min, and their relation to the action nominal in -n ~ -(p)mi, see e.g.
Korhonen 1974 and Ylikoski 2002: 75–82):
North Sámi (16b from Sammallahti 2001)
(16) a. )*
+
"
Máhtte come-PAST.3SG pick.berries-1INF
‘Máhtte came to pick berries.’
b. )*
+
,
"
Máhtte come-PAST.3SG pick.berries-2INF pick.berries-MIN
‘Máhtte came back from picking berries.’
As the non-finite in -min ~ -me(n) can also appear as a complement of the verb
vajáldahttit ‘forget’, there are, in principle, as many as three different non-
218
JUSSI YLIKOSKI
finites that might be considered alloforms of the ideal of a single infinitive in a
language:
(15')
Máret vajáldahtii lohka-t ~ lohka-mis ~ lohka-me dan girjji.
‘Máret forgot to read that book.’
All these forms may reasonably be analyzed as instances of infinitives in the
sense of “opaque non-finites used for argument functions”; all of them function
as objects, i.e. non-finite complements of vajáldahttit that can be replaced by
(genitive-)accusative marked nouns or pronouns as well as by finite
complement clauses introduced by the general complementizer ahte ‘that’. As I
have already proposed in connection with (14c) and (15), an important reason
for such a high degree of “infinitival allomorphy” may be that the semantic
relations between verbs like ‘forget’ and their complements (e.g. ‘reading that
book’) are quite unambiguous, irrespective of the exact form of the non-finite
in question.
4.5 From action nominals to converbs and infinitives: verbalization,
adverbalization or denominalization?
Once again, I take a look at the problems of defining non-finites by their wordclasses. It has been shown that case forms of action nominals often tend to
develop into new non-finites that may be called infinitives and converbs.
Converbs are sometimes understood as verbal adverbs, which appears to some
extent analogous to adverbial case forms of underived nouns being lexicalized
into indeclinable adverbs. However, it is not easy to fit infinitives into this
framework; they do not behave like nouns, nor do they correspond to adverbs
or members of any other word-class either.
A solution to the problem of determining the “word-form word-class” of
infinitives might be found in Comrie and Thompson’s (1985: 369–370)
expression the verbalization of nominal forms. They acknowledge that there
are many stages in the verbalization of action nominals into non-finites like
the Slavic infinitive or some of the Finnish converbs discussed above.
According to Comrie and Thompson, modern Slavic infinitives have become
members of the verbal paradigms as they no longer have nominal categories
such as cases, but rather possess “virtually all of the typically verbal categories
(apart from person and number, like most non-finite forms),” which
presumably refers to categories such as aspect and reflexivity.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
219
The verbalization of action nominals seems to lead not only to new
infinitives but to new converbs and other, less typical non-finites as well.
Previously (in Ylikoski 2002: 101–116), I have attempted to demonstrate that
in North Sámi a new converb in -miin has developed from the action nominal
in the comitative case. However, it seems that the only morphosyntactic feature
that clearly differentiates the miin-form from the action nominal is the
possibility of the VO word order: In accordance with a possibly universal
feature of word order in action nominal constructions (Koptjevskaja-Tamm
1993: 59, 185–187, 255), the patient marking of the Sámi action nominals
resembles that of possessors in ordinary NPs, and it seems that the VO order
has emerged after the reanalysis of genitival attributes of verbal nouns (GN) as
objects of converbs (OV); the case-marking of the attribute/object—the
genitive-accusative case—has not been subject to change. These kinds of
explanations are possible only if action nominals are viewed as nouns and not
verbs.
Keeping in mind that action nominals are, by definition, verbal nouns
with respect to their external syntax while their internal syntax can in some
languages be highly verb-like (the “sentential type” of action nominal
constructions in Koptjevskaja-Tamm’s [1993] typology), it is plausible to think
of action nominals—rather than indeclinable infinitives and converbs—as
prima facie instances of the word-class-changing inflection in the sense of
Haspelmath (1996). When an action nominal category as a whole gradually
loses its noun-like inflection (cf. the Finnish “third infinitive”) or only some of
its case forms branch off the otherwise noun-like action nominal (e.g. the
North Sámi converb in -miin), one might say that the action nominal is being
verbalized, or rather that the action nominal is simply losing its noun-like
external syntax, i.e., the nominalization of a verb is being denominalized. In
any case, it seems untenable to claim that grammaticalization chains of the type
action nominal → infinitive fundamentally differ from the development action
nominal → converb.
In this context, it is worth noting that another major source of new
converb forms is participles losing their adjectival inflection when used in nonattributive adverbial or copredicative functions (Haspelmath 1995a: 17–20).
Correspondingly, this process could probably be described as
deadjectivalization of verbal adjectives. — It should go without saying that a
clear dividing line between participles and converbs cannot be drawn either.
220
5.
JUSSI YLIKOSKI
Lexicalized and grammaticalized non-finites as evidence for and
against alleged new word-classes
So far, I have been defining and describing non-finites with regard to their
synchronic functions on the one hand, and to their origins on the other. In
addition to their past and present, it is enlightening to take a look at the future
of these forms in order to evaluate claims about their word-class status. In the
following, I will briefly outline the paths of lexicalization and
grammaticalization that turn individual non-finite forms into new lexemes.
It is almost banal to point out that individual, productively and regularly
formed action nominals (i.e. forms that normally denote actions) may be
lexicalized into nouns with quite specific meanings, e.g. English painting and
wedding; Finnish sanominen ‘quarrel’ (< ‘say’); Komi $ - ‘face’ (< ‘be
born’). (Cf. also the relics of the action nominal in *-mA in Finnish, mentioned
in Section 4.3.) Likewise, it is well known that participles tend to develop into
fully lexical adjectives, e.g. English following, Finnish seuraa-va [followPTCP.ACT.PRES] and Russian
š
[follow-PTCP.ACT.PRES-M] ‘id.’;
Finnish tunne-ttu [know-PTCP.PASS.PAST] ‘well-known’, tu-ttu (< archaic
[know-PTCP.PASS.PAST]) ‘familiar’ or—in a way similar to underived
adjectives—into nouns, e.g. Finnish tuttu and tuttava (< archaic [knowPTCP.PASS.PRES]) ‘acquaintance’; juopu-nut [get.drunk-PTCP.ACT.PAST] ‘drunk
person’, English drunk both ‘one who is drunk’ and ‘drunkard’ etc. These
developments are probably best regarded as instances of zero derivation (cf.
Scalise 1988: 565–566).
5.1 Converbs
As might be expected, converbs or “verbal adverbs” are often lexicalized into
adverbs: examples of this tendency include Finnish tieten-kin (< archaic
[know.CONV-also]) ‘of course’, verra-ten [compare-CONV] ‘relatively,
comparatively’, lakkaa-matta [stop-CONV] ‘incessantly’, odotta-matta [expectCONV] ‘unexpectedly’; Finnish elä-issä-än [live-CONV-3SG], Estonian ela-des
[live-CONV] ‘ever’; Russian
[be.silent-CONV] ‘silently’ and Hungarian
fordít-va [turn-CONV] ‘vice versa’.
However, individual converb forms develop not only into adverbs but
also into adpositions and conjunctions. In other words, converb forms may be
lexicalized into new members of an open or lexical word-class (i.e. adverbs) or
they may be grammaticalized and become members of closed or grammatical
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
221
word-classes (i.e. adpositions and conjunctions). Developments of the latter
type are also widespread: in addition to numerous examples from English (e.g.
concerning, considering, during, excepting, excluding, regarding), Haspelmath
(1995a: 38) mentions among others German entsprechend ‘according to’ (<
‘correspond’) and Russian spustj-a [let.down-CONV] ‘after’. Haspelmath notes
(p. 37) that when a converb form is grammaticalized, the (often implicit)
subject argument disappears and the object of the converb becomes the
complement of the new adposition. However, its seems that converb forms of
even certain intransitive verbs may become adpositions, and the complements
of such adpositions may originate from the subjects of those converbs, e.g.
Finnish viiko-n kuluessa [week-GEN within] ‘within a week’ < [week-GEN
pass.CONV] ‘a week passing’. Also adverbials may become complements: e.g.
huomiseen mennessä [tomorrow.ILL by] ‘by tomorrow’, < [tomorrow.ILL
go.CONV] “when going to tomorrow” and huomise-sta lähtien [tomorrow-ELA
since] ‘since tomorrow, from tomorrow on’ < [tomorrow-ELA go.away.CONV]
or Hungarian holnap-tól fogva [tomorrow-ABL since] ‘id.’ < [tomorrow-ABL
hold.CONV]. — Note that many of the deverbal prepositions in Germanic and
Romance languages have developed from the adverbial (i.e. converbal)
functions of participles (cf. the English and German examples above); for indepth studies focused on this topic, see Kortmann (1992); Kortmann and
König (1992).
The grammaticalization of a converb into an adverbial conjunction is
closely related to the development of de-converbal adpositions. It is not
unusual for a de-converbal conjunction actually to be a conjunctional
expression composed of an adposition-like converb and a general
complementizer; e.g. English considering that; Russian
.
‘although’, “not looking at the fact that”; Finnish huolimatta siitä, että
‘although’, “without worrying about the fact that.” In addition to the
development of adverbial conjunctions, there is a cross-linguistically common
path of development by which a converb form meaning ‘saying’ is first used as
a quotative marker and later as a more general complementizer that marks
many if not all complement clauses (see Haspelmath 1995a: 40–41 and
references therein). (In discussing the grammaticalization of converbal
constructions, Haspelmath [1995a: 41–45] also describes how converb forms
may—not unlike other non-finites—become main verbs of periphrastic
aspecto-temporal categories as well as applicative markers; for the present
purposes, however, I will pay attention only to those instances of
grammaticalization that lead to the development of grammatical words.)
222
JUSSI YLIKOSKI
In the light of the examples above, it appears that if defined with
reference to their lexicalization and grammaticalization (or simply
lexicalization in a broad sense that also includes the development of
grammatical lexemes; cf. Kortmann 1992: 431), converbs are not only “verbal
adverbs” but also—at least latently—“verbal adpositions” and even “verbal
conjunctions.” Furthermore, it is interesting to note that there are some adverbs
that have been lexicalized from entire converb constructions, i.e. from converb
forms together with their own subjects, objects or adverbial modifiers; e.g.
Finnish silmin nähden (sometimes spelled as one word, silminnähden) ‘visibly,
clearly’ (< silm-i-n näh-den [eye-PL-INS see-CONV] ‘seeing with one’s eyes’)
and totta puhuen (tottapuhuen) ‘in fact, to tell the truth’ (< tot-ta puhu-en
[truth-PART speak-CONV] ‘telling the truth’), and Hungarian rangrejtve
‘incognito’ (< rang-rejt-ve [social.class-conceal-CONV] “concealing one’s
social class”).
It is intriguing to note that the common denominator for converbs,
adverbs, adpositions and adverbial conjunctions alike is their adverbial nature:
Adpositions and adverbial conjunctions differ from adverbs in that they do not
function as adverbial modifiers by themselves, but rather as heads of
(adpositional) phrases and (adverbial) clauses that are adverbial modifiers as a
whole exactly like converbal constructions and adverb phrases. In other words,
the adverbial modifying functions of converbal constructions remain basically
the same in the course of lexicalization or grammaticalization, irrespective of
the resulting syntactic reanalyses. For instance, the Finnish converb in -(t)essA
preserves its function as a temporal modifier (see 9c) both after being analyzed
as a temporal adverb eläissään ‘ever’ and as temporal postpositions kuluessa
‘within’ and mennessä ‘by’. It appears that even the “word-class approach” to
the diachrony of converbs further supports the claim that instead of being
labeled as verbal adverbs, converbs are best defined as adverbial non-finites,
i.e. as non-finites that function as adverbial modifiers of verbs and clauses.
5.2 Infinitives
Finally, it is essential to try to find out what becomes of individual infinitive
forms once they are detached from verbal paradigms. Interestingly enough, it
appears that lexicalization and grammaticalization of infinitives is relatively
different from the other non-finites. As I already partly suggested in Section 3,
it seems that one of the reasons not to define infinitives as instances of wordclass-changing inflection is that the “verbal noun slot,” for instance, is better
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
223
reserved for action nominals. Furthermore, it seems that one could even argue
that the slots for “verbal adverbs,” “verbal adpositions” and “verbal
conjunctions” are filled by adverbial converbs.
Compared to the development of action nominals, participles and
converbs, it is strikingly difficult to find examples of words that originate in
infinitives. Also, the lexicalization or grammaticalization of infinitives has
received hardly any attention in previous studies. Even in Joseph’s (1983)
thorough study of the Balkan infinitive loss, there are only a couple of
mentions of lexical remnants of the lost infinitives. Similar expressions are in
fact familiar from other languages. In the Tosk dialect of Albanian, the
infinitive is preserved only in idioms like do me thënë ‘that is to say’, literally
“it wants to say” (Joseph 1983: 95–96), i.e. the exact equivalent of the French
ça veut dire. Another example of a lone survivor of the infinitive loss are the
Macedonian
‘maybe’ and Greek (Otranto dialect of Italic Greek)
telèste (or selèste, << thélei ést(h)ai) ‘id.’ (Joseph 1983: 73, 110). The
development of these forms closely corresponds to the more or less lexicalized
expressions maybe, French peut-être and Russian
& $. Otherwise, the
Balkan infinitives have developed—with verbs meaning ‘want’ as their main
verbs—into future tenses, not unlike the will-future in English (Joseph 1983:
41, 108, 163 et passim).
The lexical remnants of Balkan infinitives do not essentially differ from
the few crystallized infinitival expressions in other European languages. What
is of particular interest is that even though words and idioms like French peutêtre, ça veut dire and their equivalents in other languages may be characterized
as adverbs, they differ from converb-derived adverbs in that they consist of
non-finites along with their main verbs, as if to further underline the mutual
interdependence of infinitives and their main predicates. I am aware of only
one quite clear instance of a word that was originally a plain infinitive form,
namely the Finnish maata (or maate) and its cognates in some closely related
languages. Having developed from the adverbial functions of the so-called
first infinitive of the verb meaning ‘lie, sleep’, it is now an adverb whose
meaning and use sometimes correspond to the English to bed, e.g. Pekka meni
maata ‘Pekka went to bed’ (cf. Saukkonen 1965: 19–21, 61–62). The Erzya
Mordvin postposition sams ‘until’ is homonymous with the -ms infinitive form
of the verb ‘come’, but its origin is better understood in the light of the
situation in Moksha Mordvin, where the form in -ms also occurs as a converb
of posteriority (Bartens 1979: 44–47; 1999: 155; cf. Komi converb in -tödz in
7b and 8). Thus, even the Erzyan postposition may originate from a possible
224
JUSSI YLIKOSKI
earlier stage where the non-finite in -ms has had converbal functions; e.g $
$
[this.day-GEN until] ‘until today’ < [this.day-GEN come.CONV] ‘until
this day came’.10
There is one common feature of infinitives that deserves special attention.
The category of infinitive as a whole has a relatively strong tendency to be
homonymous with the action nominal. To quote Disterheft (1980: 198), “the
oblique case marking [of former action nominals] has generally become so
disassociated from any paradigm that these former oblique abstracts are
capable, paradoxically enough, of reentering the nominal system.” As a result,
in Ancient Greek, the action nominal consisted of the infinitive plus the neuter
definite article, not unlike the situation in Modern German (e.g. Greek einai :
to einai, German sein : das Sein; cf. Koptjevskaja-Tamm 1993: 42, 301 n. 7).
Disterheft regards Latin infinitival subjects (e.g.
) as early
signs of a similar development, which later led to case-inflected infinitives
(e.g. accusative
‘be permitted’). In modern Romance languages such as
Spanish, the forms in -r function as infinitives and action nominals alike. In
“Balkanized” Romanian, the Latin infinitive ending -re has completely turned
into an action nominal formative (Joseph 1983: 167), “thus completing a full
cycle of noun-to-infinitive-to-noun” (Disterheft 1980: 198).11 Homonymy
between infinitives (i.e. indeclinable and thus opaque forms in complement
positions) and action nominals (i.e. declinable verbal nouns) can be found in
many languages outside Europe (e.g. Hindi and Swahili) as well.
In Finno-Ugric languages, there are no clear examples of similar
development, but the history of infinitival markers shows that even the outward
10
Similarly, the Mari postposition šumeš(ke) ‘until’ is transparently the posterior converb
form of the verb šuaš ‘come’. The dual nature of the Mokshan non-finite in -ms is a good
example of two rather different non-finites that can still be regarded as natural outcomes of
an action nominal in a directional (illative) case. Note that the infinitival and converbal
functions of the -ms are clearly separate from each other (Bartens 1979: 31–51), i.e., there
appears to be no gray area comparable to the obligatory~optional directional-purposive nonfinites discussed in Section 4.3.
11
In French, the infinitives in -r(e) do not function as verbal nouns (excepting some fully
lexicalized nouns, e.g. devoir ‘duty’, dîner ‘dinner’ and pouvoir ‘power’). It was already
seen in Note 9 that certain verbs require their infinitival complements to be preceded by
prepositions such as de or à. Interestingly, as some prepositional infinitival constructions
are used in adverbial functions, the infinitive forms become reminiscent of action nominals
in adverbial adpositional phrases; compare the English translations of après + infinitive
(‘after V-ing’) or sans + infinitive (‘without V-ing’) mentioned in Note 9.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
225
appearance of an infinitive may begin to resemble the original action nominal.
Even though most Finno-Ugric infinitives are generally assumed to originate
from various combinations of verbal nouns and directional case suffixes, the
original case suffixes have often been lost, and as a result, present-day
infinitive markers look more or less the same as the supposed action nominal
suffixes on which the infinitives were originally based; e.g. the Finnic
infinitive in -tA ~ -dA (and Sámi -t) << verbal noun in *-tA + lative *-k, and
the Estonian infinitive in -ma (and Livonian -m(õ)) << verbal noun in *-mA +
illative -hAn (> Finnish -mAAn in 14a, c); approximately the same has
happened to the Komi and Udmurt infinitive in -ny and to the Hungarian
infinitive in -ni. It is important to note that it is usually infinitives and not
converbs that have completely lost their former case endings and even turned
into action nominals. A partial explanation may be found in the fact that
infinitives (as obligatory complements) do not carry very specific meanings in
themselves, and as a result their appearance is easily subject to phonological
reduction.—Consequently, in order to continue to express the relation of
purpose, these forms tend to be reinforced (cf. Section 4.3).
Once again, it is instructive to compare infinitives to accusatives. Crosslinguistically, the accusative case appears to be a sort of cul-de-sac in the
grammaticalization chain where accusative cases commonly develop from
datives, which in their turn have developed from directional and benefactive
cases. According to Lehmann (1995: 110), the only theoretically possible
function to which accusatives could be further generalized is that of an
absolutive case, but this type of grammaticalization has not been attested
because absolutives are generally unmarked. The development of adverbial
(directional-purposive) action nominal constructions or converb forms into
accusative-like infinitives and finally into (nominative forms of) new action
nominals might possibly be considered as a loose analogue of the hypothetical
directional >>> absolutive chain.
In sum, it can be concluded that even though the view that action
nominals are verbal nouns and participles are verbal adjectives is supported by
a multitude of lexicalized deverbal nouns and adjectives, developments of
individual converb and infinitive forms hardly evidence specific new wordclasses, but rather once again highlight the centrality of their syntactic
functions, as well as the importance of the obligatory/optional distinction in
differentiating between infinitives and converbs.
226
6.
JUSSI YLIKOSKI
Discussion
In this section, I summarize the main points of the preceding argumentation.
To get a more comprehensive picture of the system of non-finite verb forms, it
seems necessary briefly to discuss the definitions of participle and the position
of participles in relation to other (idealizations of) non-finites. Against the
background of what has been said about action nominals, converbs and
infinitives in the preceding sections, participles—though not the main topic of
this paper—may tentatively be described as follows: It was already mentioned
that participles are usually defined as verbal adjectives and/or as non-finites
functioning as attributes, i.e. modifiers of nouns. In Section 3, I agreed that the
non-finites called participles generally have morphosyntactic features of
adjectives that may be seen as indices of verb-class-changing inflection.
Morphologically, participles usually agree with their head nouns as underived
adjectives do, and their positions with respect to their heads resemble those of
adjectives (Haspelmath 1996: 44, 49). Functionally, they are similar to
adjectives in that their primary function is to modify nouns. Furthermore, they
also have secondary functions identical to those of adjectives; they are used as
(adjectival) predicates, which appears to have resulted in the development of a
multitude of periphrastic aspecto-temporal categories in various languages. In
addition to this, it was mentioned above that participles, when lexicalized,
usually turn into adjectives and—in a manner identical to adjectives—into
nouns. Yet again, it was briefly mentioned at the end of Section 4.5 that
participles may be “deadjectivalized” and become converbs; such copredicative
functions of participles directly correspond to similar use of true adjectives, too
(see Haspelmath 1995a: 17–20).
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
227
It is important to note that just as adverbial functions do not equal the
word-class of adverbs, not all non-finites that are used attributively are to be
labeled as verbal adjectives; compare such fairly common uses of infinitives
like will to learn or its Finnish equivalent halu oppi-a [will learn-1INF]. These
infinitives are attributes in the strictest sense of the term (‘modifiers of nouns’),
but they do not appear to have any explicitly adjectival properties, and they
cannot be replaced with adjectives as easily as participles can. Furthermore, it
is only expectable that action nominals—as verbal nouns—are also used as
(genitival) attributes, e.g. Latin ars amandi and its English and Finnish
translations the art of loving and rakasta-mise-n taito [love-AN-GEN art] (see
Itkonen 2001: 331, 350).—Apparently, as these latter types of attributival nonfinites appear never to have been called participles, the term participle is best
reserved for its traditional use as a designation for the most adjective-like nonfinites. However, it appears that the epithet verbal adjective does not refer to as
thorough a process of word-class-changing inflection as verbal noun;
participles still lack such adjective-like properties as comparative and
superlative degrees.12
12
According to Haspelmath (1996: 63 n. 6), the lack of comparative and superlative
degrees of (German) participles is “due to purely semantic factors.” However, it seems to
me that the semantic functions of comparative and superlative degrees do not differ
remarkably from the adverbial modifiers ‘more’ and ‘most’, cf. the most interesting book
and the book that interests (people) most. It appears that in relation to comparation, a
“verbal adjective” is either a true verb form—whereupon it may have an object and
adverbial modifiers (including ‘most’; see ii.a)—or it is a lexicalized adjective, able to take
the superlative form but hardly an object (ii.b) (see also Zucchi [1993: 219ff.] for analogous
examples of the dual nature of the Italian infinito sostantivato):
(ii)
Finnish
a. (itse-ä-ni)
(eniten) kiinnosta-va
self-PART-1SG most
interest-PTCP.ACT.PRES
‘the book that interests (me) (most)’
kirja
book
b. (??/*itse-ä-ni) kiinnosta-v-in
kirja
self-PART-1SG interest-PTCP.ACT.PRES-SUP book
‘the most interesting (*me) book’
Note also that (adjectivalized) participles can sometimes be turned into adjectival adverbs
like English interest-ing-ly, surpris-ing-ly or Finnish kiinnosta-va-sti, yllättä-vä-sti ‘id.’.
The internal syntax of such “verbal adverbs” is more that of adjectives than of verbs:
228
JUSSI YLIKOSKI
Drawing the threads together, I present Table 4, intended to capture my
central arguments about the main functions of the four main types of nonfinites as well as my views on the relevance of defining these categories as
instances of word-class-changing inflection (or word-class-changing word
formation in general). Converbs, for instance, are considered “verbal adverbs”
only in a diachronic perspective where it could probably be equally plausible to
label them as “verbal adpositions,” or better still, as individual instances of
deverbal adverbs and adpositions:
Non-finite verb
form:
Syntactic function:
infinitive
converb
participle
argument
(= subject,
object,
obligatory
adverbial)
—
(free)
adverbial
(= adjunct)
attribute
(+
adjectival
predicate)
—
adjective
noun
adverb,
adposition,
conjunction
adjective
( noun)
noun
“New wordclass”:
noun,
Direction of
adverb
lexicalization
(in the broad sense
that comprises the
development of
grammatical
words):
action
nominal
—
(those of
nouns)
Table 4. The four main types of non-finite verb forms, their syntactic functions and
“new word-classes” (revision of Table 3)
It is evident from Table 4 that the asymmetry between the two approaches still
remains: infinitives and converbs are best defined in terms of their syntactic
functions, whereas it would be superfluous to define action nominals as nonfinites with argumental, adverbial and attributive functions. Participles, by
comparison, seem definable both by their (predominantly) attributive functions
kiinnostava-mmin [interesting-COMP.ADV] ‘more interestingly’, yllättävä-mmin [surprisingCOMP.ADV] ‘more surprisingly’ instead of ??/*enemmän kiinnostavasti ‘interestingly
*more’, ??/*enemmän yllättävästi ‘surprisingly *more’.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
229
and by their (not completely) adjectival morphosyntax. It must, however, be
remembered that action nominals are often left outside the class of non-finite
verb forms; reasons for this may include the fact that the internal syntax of
action nominals is often less verb-like than that of participles. Furthermore, the
formation processes of action nominals appear to be more idiosyncratic than
those of other non-finites; on the continuum between inflection and derivation,
action nominals are probably more derivational than participles.
To return to the claims that infinitives are not “on a par” with other nonfinites, but rather “distribute over” or “are typically intermediate between”
action nominals and converbs (see Section 4.1), I hope that I have been able to
demonstrate that it is more reasonable to say that (idealized) infinitives and
converbs are in complementary distribution (barring the problematic boundary
between purposive and directional-purposive non-finites). In addition, it can be
seen from Table 4 that participles are, in a sense, also on a par with infinitives
and converbs, whereas the use of action nominals covers—although in quite an
abstract manner—the functions of all these other forms. If action nominals (as
the clearest example of word-class-changing inflection) were categorically left
outside non-finite verb forms, the rest of the non-finites discussed here could
be defined fairly uniformly by reference to their syntactic functions only. — In
fact, this appears to be approximately the way I.V. Nedjalkov (1998: 421–422)
defines infinitives, converbs and participles (see Section 2.2). However, it must
be admitted that the adjective-like nature of participles clearly separates them
from infinitives and converbs.
Once again, I feel compelled to defend my view that there are few reasons
to label converbs as verbal adverbs. When evaluating Haspelmath’s (1995a: 3–
4) definition of converb, Bickel (1998: 383) states that due to the
(morphological) property non-finite, it
has the advantage [over V.P. Nedjalkov’s (1995) conception of converbs13] that the
definition of converb (“nonfinite adverbial verb form”) is conceptually parallel . . . to
the traditional definition of participles (“nonfinite adjectival verb form”) and masdars
or verbal nouns (“nonfinite nominal verb form”).
Ricca (1997: 188), in turn, interprets Haspelmath as putting
13
Nedjalkov (1995: 97) defines a converb as “a verb form which depends syntactically on
another verb form, but is not its syntactic actant,” with no reference to finiteness nor to the
word-class of adverbs.
230
JUSSI YLIKOSKI
more stress on the functional role of converbs, which beautifully completes the
paradigm with ‘verbal adverbs’ alongside ‘verbal adjectives’ (i.e. participles) and
verbal nouns.
I do not intend to deny that the notion of verbal adverb would indeed
“beautifully complete” the picture of non-finites, but apparently, that could
happen only by ignoring infinitives not only as prima facie non-finites, but also
as daughters of action nominal constructions and as sisters of many converbs
(see Section 4.3). The alternative I am proposing makes converbs conceptually
parallel to infinitives and participles, and at least in a diachronic perspective, to
action nominals as well.
Again, if converbs are simply defined as free adverbial verb forms (i.e.
verbal adjuncts) in the syntactic sense (parallel to argumental infinitives and
attributive participles), the definition could possibly be considered to include
the so-called narrative converbs as well. Such “cosubordinate,” “copulative,”
“non-modifying” or “propositionally nonrestrictive” non-finites, found in many
Indo-Aryan, Dravidian, and “Altaic” languages of Asia as well as in the Turkic
and some Uralic languages of eastern Europe are generally viewed as nonfinite counterparts to the coordination of finite clauses; the same forms are
often used in modifying (“adverbial”) functions, too. Although the problems of
differentiating between the modifying and (non-modifying) conjoining
functions remain, they might be best seen as belonging to the domain of
semantics rather than syntax (see e.g. Johanson 1995: 321–322, 327–330).
(Ultimately, this would lead to labeling even the most exotic types of medial
verbs and clause chaining as converbal constructions; cf. Haspelmath 1995a:
20–27; van der Auwera 1998b; Tikkanen 2001: 1115–1116.)
Finally, it must be admitted that the reality behind the generalizations
presented in Table 4 is much more complex. As in earlier typological attempts
to define one or more types of non-finites, the discussion in the previous
sections has centered on idealizations of non-finites. It was noted at the
beginning of this paper (in Section 2) that good examples of “ideal” systems of
non-finites are actually rather difficult to find, and Hungarian was presented as
a plausible candidate to represent such an ideal, as each of the non-finites in
(1a–d) is used in quite specific functions; compare the use of the English -ing
form in the translations of the Hungarian participles, converbs and action
nominals. In practice, non-finites often have functions of more than one of the
four main subcategories discussed here. Many so-called participles of
Germanic and Romance languages are used both as attributes (participles) and
as adverbial modifiers (converbs). Common homonymy between infinitives
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
231
and action nominals was mentioned in the preceding section. In many Turkic
and Uralic languages, a non-finite may share the functions of participles and
action nominals, and the Moksha Mordvin form in -ms functions both as a
typical infinitive and as a converb of posteriority (Section 5.2). (See also V.P.
Nedjalkov 1995: 104–106.) The non-finites in Hungarian represent the ideal
system also in the sense that there is only one infinitive and no “less typical”
non-finites (cf. Section 4.4).14
All that has been said in the preceding sections has centered on the
functions that non-finites have by themselves; in other words, on the nonfinites that function as relatively independent constituents within a sentence.
Koptjevskaja-Tamm (1999: 148) sees all these functions as belonging to one of
the two main groups of functions that non-finites can have. She notes that in
addition to such dependent predicates, non-finites can also be used to form
‘analytical’ or periphrastic verb forms. Such uses have briefly been mentioned
in connection with the Finnish non-finites in -mAssA and -mAttA (Section 4.4),
the development of future tenses from ‘want’ + infinitive (e.g. in the Balkan
languages and in English; see Section 5.2) as well as the grammaticalization of
converbs (Section 5.1) and participles (cf. above). Even though it may
sometimes be difficult to distinguish between periphrastic verb forms and the
“dependent predicate” use of non-finites (see e.g. Ylikoski 2002: 127–129), it
appears understandable that these functions should be kept distinct from each
other whenever possible. However, the existence of periphrastic forms once
again suggests that non-finites are usually best defined in terms of their
functions.
7.
Conclusion
In conclusion, I admit that this paper does not offer definitive answers to the
problems of defining non-finites, but rather recognizes various continua
between the idealizations of the four main types of non-finites and different
kinds of deviations from them; the continua become even more evident when
non-finites are examined from a diachronic perspective. In the preceding
sections, I have centered on commenting and refining the definitions of action
nominals, converbs and infinitives (and, to a lesser extent, participles), and it
14
In addition to the present participle in /0 1 (1b), Hungarian possesses a past participle in
-(Vt)t and a future participle in
/0
1.
232
JUSSI YLIKOSKI
appears that these main types of non-finites and the systematic nature of their
mutual relations are best understood by combining our knowledge of both the
past and present—and even the future—of these forms.
Although the ideas presented in this article are intended to have crosslinguistic applicability, I do not claim that the function-based approach to nonfinites is equally useful for describing languages other than the familiar
synthetic languages of Europe, where the morphological non-finiteness of the
verb forms in question can often be taken for granted. Problems arise when a
purely functional approach to “non-finites” is applied to morphologically finite
dependent verb forms such as those of Bantu languages, not to speak of
isolating languages where the finite/non-finite distinction is altogether dubious
(e.g. Koptjevskaja-Tamm 1999: 149; Hu, Pan and Xu 2001). Nevertheless, I
hope to have demonstrated that in the European languages, on which many of
the typological statements (and typologists’ underlying assumptions) about
non-finites are still based, the syntactic functions of action nominals generally
distribute over infinitives and converbs, the latter two categories being in
complementary distribution in terms of relative obligatoriness vs. optionality.
Abbreviations
A
ABE
ABL
ABS
ACC
ACT
ADE
ADV
AN
AOR
COMP
COND
CONNEG
CONV
DAT
EGR
ELA
ERG
FUT
G
GA
adjective
abessive
ablative
absolutive
accusative
active
adessive
adverb
action nominal
aorist
comparative
conditional
connegative
converb
dative
egressive
elative
ergative
future tense
genitive
genitive-accusative
GEN
ILL
IMPF
INE
INF
INS
LAT
M
MIN
N
NEG
NOM
PART
PASS
PAST
PL
PRES
PRT
PTCP
genitive
illative
imperfective aspect
inessive
infinitive
instructive
lative
masculine
the non-finite in -min ~ -me(n)
(“the second gerund”, “action
essive”)
noun
negation
nominative
partitive
passive
past tense
plural
present tense
particle
participle
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
SG
SRDIR
SUP
singular
superdirective
superlative
TERM
TRA
V
233
terminative
transitive
verb
References
Bartens, Raija (1979) Mordvan, tšeremissin ja votjakin konjugaation infiniittisten muotojen
syntaksi. Mémoires de la Société Finno-Ougrienne 170. Helsinki: SuomalaisUgrilainen Seura.
—— (1999) Mordvalaiskielten rakenne ja kehitys. Mémoires de la Société FinnoOugrienne 232. Helsinki: Suomalais-Ugrilainen Seura.
—— (2000) Permiläisten kielten rakenne ja kehitys. Mémoires de la Société FinnoOugrienne 238. Helsinki: Suomalais-Ugrilainen Seura.
Bickel, Balthasar (1998) Converbs in cross-linguistic perspective. [Review article on
Haspelmath & König (eds.) 1995.] Linguistic Typology 2: 381–397.
Blake, Barry J. (1994) Case. Cambridge: Cambridge University Press.
Brown, Keith & Miller, Jim (eds.) (1999) Concise Encyclopedia of Grammatical
Categories. Amsterdam: Elsevier Science.
Bybee, Joan L. & Perkins, Revere D. & Pagliuca, William (1994) The Evolution of
Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago:
University of Chicago Press.
Comrie, Bernard & Thompson, Sandra A. (1985) Lexical nominalization. In Shopen (ed.)
1985c, pp. 349–398.
Cypanov, Evgenij (1997) Pricastie v komi jazyke: istorija, semantika, distribucija.
Ekaterinburg: Rossijskaja Akademija Nauk, Ural'skoe otdelenie.
Disterheft, Dorothy (1980) The Syntactic Development of the Infinitive in Indo-European.
Columbus, Ohio: Slavica Publishers.
Fokos-Fuchs, D. R. (1958) Die Verbaladverbien der permischen Sprachen. Acta linguistica
Academiae scientiarum Hungaricae VIII: 273–342. Budapest: Akadémiai Kiadó.
GMJa (1980) = D. V. Cygankin (ed.) (1980) Grammatika mordovskih jazykov: fonetika,
grafika, orfografija, morfologija. Ucebnik dlja nacional'nyh otdelenij vuzov. Saransk:
Mordovskij gosudarstvennyj universitet imenii N. P. Ogareva.
Harris, Alice C. & Campbell, Lyle (1995) Historical Syntax in Cross-Linguistic
Perspective. Cambridge studies in linguistics 74. Cambridge: Cambridge University
Press.
Haspelmath, Martin (1989) From purposive to infinitive – a universal path of
grammaticization. Folia Linguistica Historica 10: 287–310.
—— (1993) A Grammar of Lezgian. Mouton grammar library 9. Berlin: Mouton de
Gruyter.
—— (1994) Passive participles across languages. In Barbara Fox & Paul Hopper (eds.),
Voice: Form and Function, pp. 151–177. Amsterdam: John Benjamins Publishing
Company.
234
——
JUSSI YLIKOSKI
(1995a) The converb as a cross-linguistically valid category. In Haspelmath &
König (eds.), pp. 1–56.
—— (1995b) Contextual and specialized converbs in Lezgian. In Haspelmath & König
(eds.), pp. 415–440.
—— (1996) Word-class-changing inflection and morphological theory. In Geert Booij &
Jaap van Marle (eds.), Yearbook of Morphology 1995, pp. 43–66. Dordrecht: Kluwer
Academic Publishers.
—— (1999) Converb. In Brown & Miller (eds.), pp. 110–115.
—— (2001) The European linguistic area: Standard Average European. In Haspelmath et
al. (eds.), pp. 1492–1510.
Haspelmath, Martin & König, Ekkehard (eds.) (1995) Converbs in Cross-Linguistic
Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial
Participles, Gerunds. Empirical approaches to language typology 13. Berlin: Mouton
de Gruyter.
Haspelmath, Martin & König, Ekkehard & Oesterreicher, Wulf & Raible, Wolfgang (eds.)
(2001), Language Typology and Language Universals: An International Handbook.
Volume 2. Handbücher zur Sprach- und Kommunikationswissenschaft 20.2. Berlin:
Walter de Gruyter.
Hu, Jianhua & Pan, Haihua & Xu, Liejiong (2001) Is there a finite vs. nonfinite distinction
in Chinese? Linguistics 39: 1117–1148.
Hyvärinen, Irma (1982) Suomen kolmannen infinitiivin verbisidonnaisten inessiivin,
elatiivin ja illatiivin lauseenjäsenfunktioista ja niiden saksalaisista vastineista
valenssiteorian näkökulmasta. In Lauseenjäsennyksen perusteet. Seminaari Seilissä
9–10.9.1982, pp. 59–89. Publications of The Linguistic Association of Finland 9.
Turku: Suomen kielitieteellinen yhdistys.
IO = Isuslön olan. Stockholm and Helsinki: Biblija vudzötan institut [Institute for Bible
Translation] 1993.
Itkonen, Erkki (1972) Über das Objekt in den finnisch-wolgaischen Sprachen. FinnischUgrische Forschungen 39: 153–213.
Itkonen, Esa (1991) Universal History of Linguistics: India, China, Arabia, Europe.
Amsterdam: John Benjamins Publishing Company.
—— (2001) Maailman kielten erilaisuus ja samuus. 2., uudistettu ja laajennettu painos.
Publications in General Linguistics 4. Turku: University of Turku.
Johanson, Lars (1995) On Turkic converb clauses. In Haspelmath & König (eds.), pp. 313–
347.
Joseph, Brian D. (1983) The Synchrony and Diachrony of the Balkan Infinitive: A Study in
Areal, General, and Historical Linguistics. Cambridge Studies in Linguistics,
Supplementary Series 1. Cambridge: Cambridge University Press.
Joseph, Brian D. & Philippaki-Warburton, Irene (1987) Modern Greek. London: Croom
Helm.
Karlsson, Fred (1999) Finnish: An Essential Grammar. London: Routledge.
KomiJohn = Ivan serti bur juör. Vidlana jözödöm. Stockholm & Helsinki: Institute for
Bible Translation 1998.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
235
KomiLuke = Luka serti bur juör. Vidlana jözödöm. Stockholm & Helsinki: Institute for
Bible Translation 1996.
KomiMatthew = Mat'vej serti bur juör. Vidlana jözödöm. Helsinki & Stockholm: Institute
for Bible Translation 1999.
KomiPMatthew = Mat'vej s'örti bur juör. Korö odzlan's'a vidzötöm. Helsinki &
Stockholm: Institute for Bible Translation 2001.
König, Ekkehard (1995) The meaning of converb constructions. In Haspelmath & König
(eds.), pp. 57–96.
Koptjevskaja-Tamm, Maria (1993) Nominalizations. London: Routledge.
—— (1999) Finiteness. In Brown & Miller (eds.), pp. 146–149.
Korhonen, Mikko (1974) Die Konjugation im Lappischen. Morphologisch-historische
Untersuchung, II: Die nominalen Formkategorien. Mémoires de la Société FinnoOugrienne 155. Helsinki: Suomalais-Ugrilainen Seura.
Kortmann, Bernd (1992) Reanalysis completed and in progress: participles as source of
prepositions and conjunctions. In Günter Kellermann & Michael D. Morrissey (eds.),
Diachrony within Synchrony: Language History and Cognition, pp. 429–453.
Frankfurt am Main: Peter Lang.
—— (1997) Adverbial Subordination: A Typology and History of Adverbial
Subordinators Based on European Languages. Empirical approaches to language
typology 18. Berlin: Mouton de Gruyter.
—— (1998) Adverbial subordinators in the languages of Europe. In van der Auwera & Ó
Baoill (eds.), pp. 457–561.
Kortmann, Bernd & König, Ekkehard (1992) Categorial reanalysis: the case of deverbal
prepositions. Linguistics 30: 671–697.
Lehmann, Christian (1995) Thoughts on Grammaticalization. Revised and expanded
version. LINCOM Studies in Theoretical Linguistics 01. München: Lincom Europa.
Ludykova, V. M. (1984) Skazuemoe s m-ovym infinitivom v komi jazyke. Sovetskoe finnougrovedenie 20: 173–177.
Lunt, Horace C. (2001) Old Church Slavonic Grammar. Seventh revised edition. Berlin:
Mouton de Gruyter.
Meillet, Antoine (1934) Le slave commun. Seconde édition revue et augmentée avec le
concours de A. Vaillant. Collection de manuels publiée par l’Institut d’études slaves
2. Paris: Librairie Honoré Champion.
Muysken, Pieter (1999) Nominalizations. In Brown & Miller (eds.), pp. 248–252.
Nedjalkov, Igor’ V. (1998) Converbs in the languages of Europe. In van der Auwera & Ó
Baoill (eds.), pp. 421–455.
Nedjalkov, Vladimir P. (1995) Some typological parameters of converbs. In Haspelmath &
König (eds.), pp. 97–136.
Nedjalkov, Vladimir P. & Nedjalkov, Igor’ V. (1987) On the typological characteristics of
converbs. In Toomas Help & Sirje Murumets (eds.), Symposium on Language
Universals, pp. 75–79. Tallinn: Academy of Sciences of the Estonian SSR.
Nikanne, Urpo (1997) Suomen infiniittisten adjunktien temporaalinen tulkinta. Virittäjä
101: 338–357.
Noonan, Michael (1985) Complementation. In Shopen (ed.) 1985b, pp. 42–140.
236
JUSSI YLIKOSKI
Pajunen, Anneli (1998) Pääsanaluokkien eriytymättömyydestä uralilaiskielissä. In Anneli
Pajunen (ed.), Kieliopillistumisesta, analogiasta ja typologiasta, pp. 59–109. Suomi
185. Helsinki: Suomalaisen Kirjallisuuden Seura.
Palmer, L. R. (1954) The Latin Language. London: Faber and Faber Limited.
Ramat, Paolo & Ricca, Davide (1994) Prototypical adverbs: on the scalarity/radiality of the
notion of ADVERB. Rivista di Linguistica 6: 289–326.
Ricca, Davide (1997) [Review on Haspelmath & König (eds.) 1995.] Journal of Linguistics
33: 187–192.
Sammallahti, Pekka (2001) Strict Morpheme Dependency: An Outline of North Saami
Structure. (Manuscript.)
Sasse, Hans-Jürgen (1993) Syntactic categories and subcategories. In Joachim Jacobs,
Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann (eds.), Syntax: An
International Handbook of Contemporary Research, pp. 646–686. Handbücher zur
Sprach- und Kommunikations-wissenschaft 9.1. Berlin: Walter de Gruyter.
Saukkonen, Pauli (1965) Itämerensuomalaisten kielten tulosijainfinitiivirakenteiden
historiaa, I. Johdanto. Adverbaali infinitiivi. Mémoires de la Société FinnoOugrienne 137. Helsinki: Suomalais-Ugrilainen Seura.
Scalise, Sergio (1988) Inflection and derivation. Linguistics 26: 561–581.
Schachter, Paul (1985) Parts-of-speech systems. In Shopen (ed.) 1985a, pp. 3–61.
Shopen, Timothy (ed.) (1985a) Language Typology and Syntactic Description, Volume I:
Clause Structure. Cambridge: Cambridge University Press.
—— (ed.) (1985b) Language Typology and Syntactic Description, Volume II: Complex
Constructions. Cambridge: Cambridge University Press.
—— (ed.) (1985c) Language Typology and Syntactic Description, Volume III:
Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press.
Tikkanen, Bertil (2001) Converbs. In Haspelmath et al. (eds.), pp. 1112–1123.
van der Auwera, Johan (1998a) Conclusion. In van der Auwera & Ó Baoill (eds.), pp. 813–
836.
—— (1998b) Defining converbs. In Leonid Kulikov & Heinz Vater (eds.), Typology of
Verbal Categories: Papers Presented to Vladimir Nedjalkov on the Occasion of his
70th Birthday, pp. 273–282. Linguistische Arbeiten 382. Tübingen: Max Niemeyer
Verlag.
—— (1999) Adverbs and adverbials. In Brown & Miller (eds.), pp. 8–12.
van der Auwera, Johan & Ó Baoill, Dónall P., (eds.) (1998) Adverbial Constructions in the
Languages of Europe. Empirical approaches to language typology 20-3. Berlin:
Mouton de Gruyter.
Vineis, Edoardo (1998) Latin. In Anna Giacalone Ramat & Paolo Ramat (eds.), The IndoEuropean Languages, pp. 261–321. London: Routledge.
Wiedemann, Ferdinand Johann (1884) Grammatik der syrjänischen Sprache: mit
Berücksichtigung ihrer Dialekte und des Wotjakischen. St. Petersburg: Eggers & Co,
u. J. Glasunow.
Woodcock, E. C. (1959) A New Latin Syntax. Bristol: Bristol Classical Press.
Ylikoski, Jussi (2001) Havaintoja komin konverbeistä. Journal de la Société FinnoOugrienne 89: 199–226.
DEFINING NON-FINITES: ACTION NOMINALS, CONVERBS AND INFINITIVES
237
—— (2002) Zu den abverbialen Nominalkonstruktionen im Nordlappischen. FinnischUgrische Forschungen 57: 68–166.
Zucchi, Alessandro (1993) The Language of Propositions and Events. Dordrecht: Kluwer
Academic Publishers.
Contact information:
Jussi Ylikoski
Finno-Ugric languages
Fennicum
FIN-20014 University of Turku
Finland
jussi.ylikoski@utu.fi