Melanie Seiss: On the difference between auxiliaries, serial

Download Melanie Seiss: On the difference between auxiliaries, serial

Preview text

ON THE DIFFERENCE BETWEEN AUXILIARIES, SERIAL VERBS AND LIGHT VERBS Melanie Seiss Universita¨t Konstanz Proceedings of the LFG09 Conference
Miriam Butt and Tracy Holloway King (Editors) 2009
CSLI Publications

In this paper I look at light verb, serial verb and auxiliary constructions crosslinguistically and try to set up criteria to distinguish these constructions. I argue that while a coherent set of properties can be found to distinguish light verbs from auxiliaries, it is more difficult to find crosslinguistic criteria which set serial verbs apart from light verbs and auxiliaries. This is because the class of serial verbs is not coherent, i.e. it is not clear which constructions should be considered serial verbs. Nevertheless, when looking at a specific language in detail, it can be established whether a construction may be considered a serial verb.
As a case study, I look at posture verbs in Ngan’gityemerri, a Northern Australian language. In Ngan’gityemerri, posture verbs can be used as simple verbs, in verb + coverb complexes and as clitics which attach to verb + coverb complexes. I show that while these constructions seem to be very similar at first glance, they behave differently when looked at in more detail. Thus, I argue that verb + coverb complexes are complex predicates while the encliticized posture verb should be best analyzed as an auxiliary.
1 Introduction
The study of complex predictes has received a lot of attention, both descriptive and theoretical. Butt (1995) defines a complex predicate in terms of Lexical-Functional Grammar (LFG) as follows:
(1) Definition of a Complex Predicate (Butt 1995)
• The argument structure is complex (two or more semantic heads contribute arguments).
• The grammatical functional structure (f-structure) is that of a simple predicate. It is flat: there is only a single predicate (a nuclear PRED) and a single subject.
• The phrase structure (c-structure) may be either simple or complex. It does not necessarily determine the status of a complex predicate.
Similar definitions can also be found in Mohanan (1994, 1997) or Alsina et al. (1997). Despite this clear definition, it is not easy to distinguish complex predicates from other syntactic constructions. Complex predicate constructions can be confused with coordinated or subordinated sentence constructions when the monoclausality of the constructions is not shown properly or they can be confused with monoclausal syntactic constructions like auxiliary verb constructions or serial verb constructions. In this paper I will focus on this second problem.
†Many thanks go to Rachel Nordlinger for pointing me towards the Ngan’gityemerri data and to my supervisor Miriam Butt for help with the analysis and making it financially possible for me to attend the conference.

Light verbs in complex predicate constructions, auxiliaries and serial verbs are often also very similar semantically. This makes them very hard to distinguish. As an example, consider the sentences in (2). All sentences contain an inflected form of ‘stand’ in combination with either an uninflected or inflected “main” verb and it seems that ‘stand’ in these sentences conveys mostly aspectual information. Nevertheless, the constructions have been analyzed (or at least called) differently by different researchers. Lemmens (2005) treats the Dutch example in (2a) as an auxiliary construction while Aikhenvald (1999) calls the Tariana example in (2b) a serial verb construction and Bowern (2004) uses the Turkmen sentence in (2c) as an example for a complex predicate.

(2) a. Auxiliary Construction

Ik stond te wachten. I stood to wait-INF ‘I was (standing and) waiting.’

(Dutch, Lemmens, 2005, 184))

b. Serial Verb

tuiRi-keRe na-hwa nema. bird-island 3PL-stay 3PL-stand
‘They stayed at the Bird Island for a long time.’ (Tariana, Aikhenvald, 1999, 480)

c. Complex Predicate

Ali kitabi

okuyup turdu.

Ali book-ACC read-GER ‘stand’-PST

‘Ali kept on reading a book.

(Turkmen, Bowern, 2004, 253)

I do not want to claim here that all these constructions are the same, but I argue that a careful, detailed study to decide on the status of these multi-verbal constructions is needed. Although it it clear that there is not a clear-cut difference between the constructions, I claim that differences still exist, and that criteria can be established to decide on the status of these constructions.
Distinguishing between auxiliary, light verb or serial verb constructions goes beyond the merely terminological. A unified terminology enables linguists to compare constructions crosslinguistically and to test analyses proposed for a construction in one language against the same construction in other languages. For example, Baker’s (1989) analysis of serial verbs is often criticized for only accounting for serial verbs which share their objects. Shared objecthood, however, is a defining feature for serial verbs as understood by Baker (1989). His analysis thus cannot be evaluated against serial verbs which do not share their objects.
To avoid such problems, I try to set up crosslinguistically valid criteria to distinguish between auxiliary, light verb and serial verb constructions. I briefly review the state of the art on these constructions in section 2 and propose some criteria to


distinguish the constructions. In section 3, I look at two different constructions in Ngan’gityemerri (Reid, 1990, 2000, 2002, 2003; Reid and McTaggart, 2008) as a case study. Section 4 concludes the discussion.

2 Establishing Crosslinguistic Criteria

2.1 The Problem of Serial Verbs
Serial verb constructions are an important topic in research on West African, Oceanic and Asian languages. Work on serial verbs includes among others Stewart (1963); Foley and Olson (1985); Sebba (1987); Baker (1989); Joseph and Zwicky (1990); Osam (1994); Bodomo (1996, 1997); Andrews (1997); Aikhenvald (1999, 2006); Stewart (2001); Foley (2009); Jarkey (2009); Appah (2009). In spite of this substantial body of research, still no agreed upon set of defining features of serial verbs has been established. Thus, serial verbs do not seem to be a coherent syntactic class, or, as Crowley (2002) put it, “many authors are not fully explicit about what they mean by serial verbs, with some writers simply treating any verb-verb sequence as serial verbs as long as the second verb is not obviously marked as an infinitive”. For a discussion of this problem see for example also Sebba (1987) and Lord (1993).
Most importantly, researchers differ in their views on object sharing, switchsubject constructions and shared tense, aspect and polarity features, of which the issue of object sharing is the most controversial. While some researchers, e.g. Stewart (1963), Baker (1989) or Stewart (2001), require objects to be shared, others do not require objects to be shared, e.g. Crowley (2002); Aikhenvald (1999, 2006). More precisely, most researchers agree on treating sentences like (3a) as serial verbs, because e`vba`re´ ‘food’, is the object of both verbs. Sentence (3b), on the other hand, is a combination of an intransitive and a transitive verb. Thus, the object cannot be shared and some researchers (e.g. Stewart 2001) would not treat this construction as an instance of serial verbs.

(3) a. O` zo´ de´ e`vba`re´ rhie´ ne` I`fue`ko` Ozo b˙uy food give to Ifu˙eko

‘Ozo bought the food and gave it to Ifueko.’ b. U´ y`ı h`ıa´ le´ e`vba`re´

(E` do´, Stewart, 2001) ˙

Uyi try cook food

‘Uyi managed to cook food.’ c. A`bie´!yu´wa` h`ı´ın e`rha´n kpa`a´n a`l`ımo´
Abi˙eyuwa climb tree pluck orange

(E` do´, Stewart, 2001) ˙

‘Abieyuwa climbed the tree and plucked an orange. 2001)

(E` do´, Stewart, ˙


(3c) is often called ‘covert coordination’ and is still more controversial than the examples above. Researchers who consider object sharing a defining feature clearly reject this construction as serial verbs. However, others claim that to decide whether (3c) is a serial verb construction, semantic and pragmatic features have to be taken into account. Thus, a serial verb can only be used to denote an accepted, although maybe complex, event in a culture. For example in Alamblak, an action which involves climbing a tree in order to look for insects is a reasonable event, but an action which involves climbing a tree in order to look at the moon is not (Bruce 1988, see also Durie 1997). This meaning cannot be expressed by a serial verb and (4b) is thus ungrammatical.
(4) a. m1yt ritm muh-hambray-an-m tree insects climb-search.for-1Sg-3Pl ‘I climbed the tree looking for insects.’ (Alamblak, Bruce 1988, 29)
b. *m1yt gun˜m muh-he¨ti-an-m tree stars climb-see-1Sg-3Pl
‘I climbed the tree and saw the stars.’ (Alamblak, Bruce 1988, 29)
As a result of these differences, different subgroupings have been proposed by different researchers. An early distinction along with a theoretical analysis was proposed by Foley and Olson (1985) who distinguish between nuclear and core layer serialization (see also Crowley 2002), i.e. they distinguish in principle between V and VP serialization. A distinction between covert coordination and serialization of verbs which form a complex event was proposed by Osam (1994), who calls these ‘clause chaining’ and ‘integrated serial verbs’ respectively. This distinction corresponds to what other researchers have called ‘linking type’ and ‘modifying type’ (e.g. Bamgbros.e, 1974).
Aikhenvald (1999, 2006) looks at the problem from a different angle and distinguishes serial verb constructions according to verb classes. In a symmetrical serial verb construction both verbs come from an open verb class while in an asymmetrical serial verb construction one of the verbs comes from a restricted verb class, e.g. from motion or posture verbs. Finally, Andrews and Manning (1999) propose formal analyses for very different serial verbs in Tariana and Misumalpan and discuss different understandings of serial verbs by different researchers.
Although researchers do not agree upon these differences, some properties are shared among all of them. Thus, Bowern (2008) lists the following concepts as properties of serial verbs in general:
(5) Properties of Serial Verbs (Bowern 2008)
• the clause contains two (or more) verbs under a single intonation contour
• the verbs must be full lexical verbs which can head simple predicates in their own right

• the verbs share at least one argument • the verbs behave as a single unit for tense, aspect, and polarity
While this set may be the minimal similarities of the constructions called serial verbs in the literature, it is impossible to find a proper analysis which accounts for all constructions which may fall under this definition. In the same way, these properties make it hard to distinguish serial verbs from auxiliaries and light verbs. As a consequence, serial verbs cannot be compared as a whole class to complex predicates or auxiliaries (see also Beermann and Hellan 2002). Careful languagespecific studies are needed to decide whether certain kinds of serial verbs may be auxiliaries or complex predicates, for example serial verbs which do not share their object, like causative or aspectual serial verbs, may be complex predicates or auxiliaries.
Other serial verb constructions may be distinguished from complex predicates and auxiliary constructions, for example symmetrical serial verbs in which both verbs carry their full semantic content, i.e. when they are not “light” verbs. Additionally, morphological marking for tense, person etc. can be on just one, on more or on all verbs in a serial verb construction. On the other hand, morphological marking in complex predicates is usually just on the light verb. Finally, there seems to be a difference in the semantics of many kinds of serial verbs and complex predicates. Thus, verbs in serial verb constructions denote single events which constitute a complex event together while light verbs provide more information about the event of the main verb (Butt, 1995) and auxiliaries mainly provide information about tense, aspect and mood.
To sum up, as constructions called serial verbs vary in details such as object sharing etc., they cannot be compared as a whole syntactic class to auxiliaries or light verbs. Common properties of serial verbs as proposed by Bowern (2008) or Aikhenvald (2006) are useful for a typology of serial verbs. To decide whether a given serial verb in a specific language may be a light verb or auxiliary, a detailed study of this serial verb construction is needed. In the following, I discuss some properties of auxiliaries and light verbs which may help to decide if a serial verb may be analyzed as auxiliary or light verb.
2.2 Auxiliaries and their historical development
Motion and posture verbs are common sources for auxiliaries, for example the English going-to-future or the Catalan go-past (Juge, 2006). When looking at the historical development of auxiliaries, one usually finds a consensus that auxiliaries may develop from main verbs when they acquire functional properties. There also seems to be a consensus that serial verbs can be an intermediate stage on the grammaticalization cline for auxiliaries (Anderson, 2006; Heine, 1993; Lord, 1993; Delancey, 1991). However, researchers do not agree on whether light verbs are an intermediate stage between main verbs and auxiliaries. Roberts and Roussou (2003)

discuss the development of English modal auxiliaries and state that there is some evidence for assuming that the pre-modal verbs, i.e. the verbs which developed into modals, were light verbs. However, they do not discuss this in detail and do not take a definite view on the matter. Similarly, Hopper and Traugott (1993) follow Hook (1974, 1991) in his proposal that light verbs in Hindi and other Indo-Aryan languages are an intermediate stage between main verbs and auxiliaries. However, Hopper and Traugott (2003) revise this view and state that it is not clear that auxiliaries developed from light verbs. Thus, they follow Butt’s view on light verbs. Butt and Lahiri (2002) and Butt and Geuder (2003) claim that light verbs do not develop into auxiliaries but are a dead end in the development of verb forms. They show that light verbs in Urdu have been used similarly for thousands of years. Bowern (2008) agrees with the view that light verbs are not a necessary step for the development from main verbs to auxiliaries but leaves it open if light verbs can develop into other verbal forms or inflections.
In this debate it becomes apparent that a difference in the application of the terms ‘complex predicate’ and ‘light verb’ by different researchers is the, or at least one, reason for their differing views. While for example Butt and Lahiri (2002) have a very clear, narrow definition of light verbs and complex predicates, Anderson (2006) includes various syntactic constructions, such as serial verb constructions, verb plus clausal complement sequences, clause-chained or conjunctive sequences, under the label ‘complex predicates’. No evidence to my knowledge has been presented in the literature so far that Butt and Lahiri’s (2002) kind of light verb developed into an auxiliary.
Independently of whether light verbs are an intermediate step in the development of auxiliaries, drawing a line between auxiliaries and other verb forms is complicated by the diachronic perspective. In general, we find two major terminological traditions: some researchers (e.g. Kuteva, 2001; Lemmens, 2005; Anderson, 2006) do not make a distinction as to how far a verb has been reanalysed as an aspect marker. As soon as a verb is used in this way, it is called an auxiliary. Others (e.g. Heine, 1993) acknowledge that there is a transition period where the distinction is not clear but for the constructions at the starting and end point of the historic development one can find distinguishing features. For example, in Heine’s (1993) view, an auxiliary has reached its ‘developmental end-point’ when the auxiliary can be used with its corresponding main verb, in sentences like He is going to go to the cinema.
Defining auxiliaries is further complicated by the fact that auxiliaries look very different in different languages. Thus, while most researchers agree that auxiliaries in some way position the event of the main verb in context to the speech or reference time, i.e. they convey information about tense and aspect, other properties of auxiliaries differ from language to language. Thus, in some languages auxiliaries carry all morphological information relating to a predicate such as person, number, tense/aspect/modality, negation marking etc., while in other languages auxiliaries show a reduced verbal behavior.

Connected to this question is the problem whether auxiliaries can combine with inflected main verbs or if they have to carry all inflections themselves.1 One example for a combination of an auxiliary with an inflected main verb comes from Urdu ((6)). Butt and Lahiri (2002) show that in Urdu, ‘be’ can be used as an auxiliary marking past tense in combination with main verbs which themselves can be marked in different ways.

(6) a. nadya=ko xAt



Nadya.F=Dat letter.M.Nom receive-Perf.M.Pl be.Past-M.Pl

‘Nadya had received letters.’

(Urdu, Butt and Lahiri, 2002)

b. nadya=ko xAt



Nadya.F=Dat letter.M.Nom receive-Impf-M.Pl be.Past-M.Pl

‘Nadya used to receive letters.’

(Urdu, Butt and Lahiri, 2002)

One question on which researchers also do not agree is whether the auxiliary may still carry some of its original semantic meaning. Heine (1993), however, points out that this is not a very reliable criterion as even with accepted auxiliaries such as in the English going-to-future, it is not always clear whether is going to as used in (7b) is a grammatical or verbal element.

(7) a. He is going to town. b. He is going to work. c. He is going to come.

That an auxiliary still carries some of its original meaning in certain contexts is especially common of auxiliaries which developed from posture or motion verbs. For example, Lemmens (2005) looks at aspectual posture verb constructions in Dutch which are used to convey progressive, durative or habitual meaning. Examples of such constructions are given in (8).

(8) Ik zat te lezen / ik stond te wachten / ik lag te slapen. I sat to read-INF / I stood to wait-INF / I lay to sleep-INF ‘I was (sitting and) reading /(standing and) waiting / (lying and) sleeping.’ (Lemmens, 2005, 184)

In the examples in (8) it can be argued that the meaning of the posture verbs is still important as the meaning of the main predicate fits to their meaning. However, these constructions can also be used when the agent’s posture is not an issue, or when the posture denoted by the auxiliary does not correspond to the posture of the main verb, for example as illustrated in (9).

(9) Wat zit ik hier toch rond te lopen? (pers. attestation)

what sit I here (toch) around to walk?

‘Why on earth am I walking (around) here?’

(Lemmens, 2005, 185)

1I thank Rachel Nordlinger (p.c.) for bringing up this question.


Similar examples are also discussed in Kuteva (2001). It would be very strange to call the posture verb in (9) auxiliary but exclude the posture verbs in (8) from being an auxiliary in the Dutch verbal system. Thus, in my view an auxiliary can also carry some of the original semantics of the verb it developed from.
Summing up, in my view auxiliaries developed from main verbs and can mark tense, aspect or modality. They may also carry some of their original semantic meaning and may combine with inflected main verbs. More properties can be set up to distinguish auxiliaries from light verbs, which I will discuss in the next section.

2.3 Light Verbs vs. Auxiliaries
Butt (2009) states that tests to distinguish light verbs from main verbs or auxiliaries differ from language to language. However, there are also some properties which set light verbs apart from auxiliaries crosslinguistically. Butt and Lahiri (2002) name some more properties to distinguish light verbs from auxiliaries.

(10) Properties of light verbs (Butt, 2009; Butt and Lahiri, 2002)
• light verbs are always form identical to the corresponding main verb whereas auxiliaries are usually just form identical at the initial stage of reanalysis from verb to auxiliary.
• light verbs always span the entire verbal paradigm (are not restricted to appear with just one tense or aspect form).
• light verbs do not display a defective paradigm.
• light verbs exhibit subtle lexical semantic differences in terms of combinatorial possibilities with main verbs, are thus restricted in their combinations. Auxiliaries, on the other hand, are not restricted in their combinatorical possibilities, but do not have to combine with every main verb.

When looking at complex predicates crosslinguistically, further properties of light verbs can be observed which set them apart from auxiliaries, although sometimes a very careful look is needed to distinguish the two constructions. For example, light verbs contribute semantic information about the type of event. This can sometimes include Aktionsart information, which can be confused with aspect, especially if the light verb is encoding telicity/completeness as in (11)

(11) nadya=ne


lıkh li-ya.

Nadya.F.Sg=Erg letter.M.Nom write take-Perf.M.Sg

‘Nadya wrote a letter (completely).’

(Urdu, Butt, 1995)

However, other differences also exist. Thus, light verbs can change the valency of a construction, for example in causative constructions as in (12). The light verb


faire ‘make’, adds an argument, the causer, to the construction. Auxiliaries are not able to add or reduce arguments. Passive auxiliaries, which may be considered as reducing the arguments at first glance, seem to be very different from light verbs when looked at in more detail. For example, passives do not change the basic argument structure, just its syntactic realization, and the agent can still be expressed as a adjunct.

(12) Jean a fait partir Marie. Jean has made go Marie ‘Jean made Marie go.’

(French, Rosen, 1990, 37)

Another property in which light verbs and auxiliaries differ is the ability to assign case. Light verbs may determine case assignment, e.g. in (13), the case of the subject depends on the choice for the light verb. Auxiliaries, in contrast, are usually not considered to be able to assign case, but may be sensitive to categories such as unaccusative vs. unergative.

(13) a. ilaa-ko khaanaa pasand huaa

Ila-D food-N like happen-PF

‘Ila liked the food.

(Hindi, Mohanan, 1997, 437)

b. ilaa-ne khaanaa pasand kiyaa

Ila-E food-N like do-PF

‘Ila liked the food.

(Hindi, Mohanan, 1997, 437)

Finally, light verbs may determine theta-role assignment while auxiliaries cannot. In (14), an example from Bardi, the light verbs ma ‘put’ or ga ‘carry’ result in a different theta-role-assignment when combined with the coverb abarrabarr. In (14a), there is only one theta-role, a theme. In contrast, in (14b), two theta-roles are assigned, an agent and a patient.

(14) a. abarrabarr-ma- ‘to be careless’ b. abarrabarr-ga- ‘to lead someone astray’

(Bardi, Bowern, 2004)

To sum up, light verbs and auxiliaries may differ in their combinatorical behavior, their paradigm, their ability to change the valency of a main verb and their ability to assign case or theta roles. Both may develop from main verbs, but while auxiliaries may develop further into clitics and morphological markers, light verbs seem to be a dead end. As serial verbs are a very diverse syntactic class, no claim can be made that all serial verbs are light verbs or auxiliaries on the one hand, on the other hand it cannot be claimed that no serial verb is a light verb or auxiliary either. This has to be investigated for each serial verb construction in a language in detail. In the following section, I look at two verbal constructions in Ngan’gityemerri and show that although they look very similar at first glance, one of them behaves like a light verb while the other is best analyzed as an auxiliary.


Preparing to load PDF file. please wait...

0 of 0
Melanie Seiss: On the difference between auxiliaries, serial