| AUTHOR: Anderson, Gregory D.S.
TITLE: Auxiliary Verb Constructions
SERIES: Oxford Studies in Typology and Linguistic Theory
PUBLISHER: Oxford University Press
Sanford B. Steever, unaffiliated scholar
This volume, part of the Oxford Studies in Typology and Linguistic Theory,
analyzes Auxiliary Verb Constructions (AVCs) utilizing a database of
approximately 800 languages. So vast is the topic that to make headway and say
anything interesting, the author, Gregory Anderson, limits his inquiry to the
different patterns of agreement manifested in AVCs. The resulting analysis is
not bound to any particular linguistic theory; the orientation is neither
exclusively synchronic nor diachronic, but what Anderson calls panchronic (p.
1). This allows him to include data from both the precursors of AVCs, such as
light verbs and serial verb constructions, and the remnants of AVCs entombed in
verbs as bound affixes. With the unprecedented wealth of data that he brings to
this task, Anderson is able to provide linguists with real insights into
agreement within the AVC, how it varies across languages, and how that variation
arises over time.
The languages investigated have, for the most part, some kind of fused or bound
functional elements, such as affixes or clitics (p. 3); isolating languages
seldom figure in the discussion, primarily because the central theme is the
distribution of agreement and similar markers on the auxiliary and lexical verbs
(LVs) in an AVC. The languages geographically reflect four macro-areas, Eurasia,
Africa, macro-Indo-Pacific and the New World. As understood in this book, AVCs
are ''not discrete units per se but rather mono-clausal combinations occupying a
non-discrete space on several large form-function continua that include serial
verb constructions, clause chaining, and verb plus complement clause
combinations on the one hand and tense-aspect-mood affixes on the other'' (p. 4).
The AVC is thus viewed as an intermediate stage in a historical-typological arc
that transits from syntactically complex verbal constructions through AVCs to
bound morphology, which clearly aligns the book with such studies of
grammaticalization as Heine (1993) and Kuteva (2001). The book takes an agnostic
stance on whether AVCs may be characterized by a definition independent of this
broad arc so as not to prejudice the results or exclude potentially useful data.
In keeping with this position, it eschews language-independent formal criteria
to determine whether any given element is a lexical or an auxiliary verb, as
well as notional criteria, such as marking tense, aspect, mood, etc.
Section 1.5 introduces the notions of inflection, dependency and headedness in
AVCs. While the auxiliary verb is commonly held to be the head of the AVC, how
the notion of head may be construed is subject to cross-linguistic variation.
Anderson initially distinguishes three understandings: inflectional, phrasal and
semantic head, which are responsible respectively for bearing inflections, for
locating the AVC in a larger syntactic structure and for determining such
features as valence and the semantic role between a verb and its arguments. (In
chapter 7, a fourth notion, ''prosodic'' headedness, is added which appears not to
serve in the overall classification of AVCs.) He shows that inflectional and
phrasal headedness coincide in some instances but diverge in others; in some
they are concentrated on either the auxiliary or the lexical verb, in others
they are distributed over the two. This variation underpins a typology of five
morphological patterns of agreement into which the AVCs are sorted (p. 23, ff).
The patterns are called AUX-headed, LEX-headed, doubled, split and
split/doubled, and they form the basis of chapters 2 through 5, with the split
and split/doubled patterns combined in chapter 5.
Chapter 2 treats AUX-headed AVCs in which the auxiliary is the inflectional head
and the LV occurs in one of the combining forms that a particular language makes
available; a clear example is the English ''John is singing.'' It is the most
common variant across the database and the one that most linguists would readily
recognize as involving an auxiliary. Examples within this chapter are classified
by the morphological shape the LV assumes when it combines with an auxiliary,
e.g. infinitive, gerund, participle, verbal noun, etc.
Chapter 3 treats LEX-headed AVCs in which the LV is the inflectional head as
well as the semantic head, which means that the auxiliary is a fixed or
uninflected form (p. 116). It is perhaps bound to be the most controversial
since additional criteria must be marshaled to distinguish a genuine AVC from an
ordinary sequence of verb and adverb, or verb and noun, or verb and particle, etc.
Chapter 4 treats double-headed inflection in AVCs in which the lexical verb and
the auxiliary are inflectional co-heads in that every obligatory inflectional
category imposed on the AVC is present on both the lexical and auxiliary verbs
(p. 144). This pattern is relatively rare.
Chapter 5 treats both the split and the split/doubled agreement patterns in the
AVC, in which the obligatory categories of the AVC are variously distributed
across the auxiliary and LV. In one important variant of the split pattern,
discussed further below, a morpheme marking negation is expressed on one
component, generally the LV, while the auxiliary bears the other obligatory
Chapter 6 discusses complex verb forms that appear to contain the historically
fused remnants of earlier AVCs. Anderson offers examples in which all of the
major macro-patterns in chapters 2 through 5 appear to be fossilized in simple
verb forms of many of the languages in the database. These morphologically
complicated verb forms are the end-state of the historical-typological arc in
which AVCs participate.
Chapter 7 returns to the starting point of the historical trajectory that gives
rise to AVCs. The early part of the chapter discusses the different kinds of
constructions from which AVCs are said to arise, such a serial verb
constructions (7.1) and verb-complement sequences (7.2). In short, the variation
in agreement seen in AVCs is traced to differences in the source constructions.
The remainder of the chapter (7.3) then discusses a set of recurring schemata,
centered on a limited number of lexical verbs. The verbs in these schemata are
prone to give rise over time to specific auxiliary verbs. They are tabulated at
the end of the chapter to determine whether specific verbs tend to develop into
the specific verbal categories that appear in AVCs.
Anderson succeeds in demonstrating the full range of agreement patterns that may
appear in AVCs, and in showing how they may be related to other features of the
languages in which they appear. The commentary below is meant to supplement some
of Anderson's proposals, not displace them; it draws on my own experience in
working on the descriptive and historical aspects of AVCs in the Dravidian
languages. As a matter of disclosure, Anderson's use of my work in this area is
generous, and I am pleased to have contributed in some small measure to the
underlying database. His research and conclusions will certainly impact the way
in which I approach the description and analysis of the AVC in the future.
In chapter 1, Anderson comes close to an independent characterization of the AVC
with his notion of semantic headedness. A review of his five inflectional
patterns reveals that all five exhibit the same feature ''setting'' vis-à-vis the
semantic head, viz., the lexical verb in an AVC is always the semantic head.
This opens the possibility of eliminating reference to semantic headedness and
reducing all the potential agreements patterns to just those implied by the
various settings of the inflectional and phrasal heads. Despite Anderson's
reluctance to do so, the AVC may be characterized independently of its
historical-typological linguistic arc; after all, abhorring an analytic vacuum,
a synchronic grammar is eventually going to require that such a characterization
be attempted at some point in linguistic analysis. Steever (2005) elaborates a
theory of the AVC proposed by Benveniste (1965), and applies it to the auxiliary
verb system of Tamil. In short, the AVC is a complex morphosyntactic vehicle,
generally a periphrastic verb form, that expresses those verbal categories (or
combinations of verbal categories) that are not expressed in the morphology of
the simple verb forms in the language. I believe that such a model may be
profitably applied to most of what Anderson characterizes as AVCs in this book.
This is not as a shortcoming in his approach, merely a supplementary approach
that allows us to better triangulate auxiliary verb phenomena.
In the summary of chapter 2 on p. 113, Anderson assumes that the inflections to
be borne by the AVC are finite; however, there will be many contexts, generally
subordinate constructions, in which the inflections imposed on the AVC by the
grammatical context are nonfinite. It can be interesting to see what kinds of
correlations, if any, emerge from comparisons of these finite AVCs with their
Although the text in chapter 3 (p. 142) suggests that classifying a construction
as a LEX-headed AVC requires that the auxiliary descends from a verb form or
appear in positions occupied by auxiliaries, synchronic tests may well be
available in the individual languages, e.g., the auxiliary may occupy a position
in the constituent structure that the syntax independently dedicates to verbs
(see Steever 2005: 21).
The doubled patterns discussed in chapter 4 appear in many instances to descend
historically from serial verb constructions; in rarer instances, they appear to
descend from coordinate structures (Kuteva 2001: 44 provides examples from
Danish, Norwegian and Swedish). In should be noted that since AVCs are by
hypothesis mono-clausal structures, the multiple agreement markers exist not
just to show agreement with a subject (or other argument), but to mark concord
among the component parts of the AVC, effectively marking its boundaries. There
seem to be no examples in the database in which multiple agreement markers in a
single AVC show agreement with two distinct subjects. It would be interesting to
further refine Anderson's classification by seeing if tests can be developed to
determine whether one agreement marker or both in an AVC actually mark agreement
with an argument, and if so, which one.
Chapter 5 raises but does not strenuously pursue the role that negation may play
in the distribution of verbal morphology in the split and split-doubled
agreement patterns. Indeed, ''negative spreading'' over both the auxiliary and LV
seems so uncommon that it may point to some underlying property of the AVC,
e.g., that its two components form so strong a bond that neither can be negated
independently of the other. In the split/doubled pattern, some categories go
with the LV, others with the auxiliary and still others with both. I will
comment on the micro-variation Anderson ascribes to AVCs in certain Dravidian
languages on pp. 235-36. Apart from the past negative paradigm in a handful of
South-Central Dravidian languages (see Steever 1993: Chapter 4), no single
finite verb in Dravidian can incorporate both a marker for tense/mood and one
for negation; they are in complementary distribution in simple verb forms. Thus,
the Old Telugu AVC in (166) ceppanu aytini 'I did not say' consists of ceppanu
'I did/do/will not say' and aytini 'I became'; the former encodes the negative
marker, the latter the tense marker, both have congruent allomorphs for
agreement with a first person singular subject. This pattern is contrasted with
one in South Dravidian language Old Tamil, celveem alleem 'we will not go',
which consists of celveem 'we will go' and alleem 'we did/do/will not become'.
Here the LV has the tense marker, and the auxiliary the negative marker. The
difference between the two patterns may be ascribed to a quirk of Dravidian
lexicography: South Dravidian aa-/aay- 'become' has a negative allomorph al-
'not become' while the corresponding South-Central cognate does not. The
morphemes in the two patterns are distributed between the LV and auxiliary in
accordance with these two factors. Anderson cites a third example, the Parji
present perfect, nilten medan 'I am standing, have stood up', which consists of
the juxtaposition of the past tense form of 'stand' and the nonpast form of
'be', both inflected for congruent personal endings. (On a strict reading this
is not a doubled pattern, since the TAM markers are in fact distinct, one
marking past, the other marking nonpast tense.) There is no simple morpheme for
the present perfect, so it must be expressed (if it is to be expressed at all)
through an AVC, in which the sequence of the simple past tense marker on the LV
(-t-) and the simple nonpast tense market on the auxiliary (-d-) describes a
temporal interval appropriate for representing a present perfect or present
continuous tense series (Steever 1993: Chapter 3). Finally, a minor point., I do
not claim that these three patterns in Dravidian descend from what are
characterized as serial verbs in the general literature, nor that any single
pattern is the historical source of all three. In some instances, we lack the
data to make that case; in other instances, the split/doubled AVC is
reconstructable to Proto-Dravidian; and in yet others, we have no evidence from
within the family to reconstruct a serial verb construction in
pre-Proto-Dravidian. Despite this fine-tuning, however, Anderson's taxonomy
proves its worth in embracing the synchronic diversity of AVCs as found in
Dravidian and elsewhere, and allowing for their comparison.
The remnants of AVCs described in chapter 6 are the result of fusion or
univerbation which may result in such unusual forms as circumfixes and
discontinuous personal endings (although these may have other historical
sources). To trace on the potential source of certain bound affixes to earlier
AVCs, in a few instances Anderson appears to insert morpheme boundaries where
they may no longer be synchronically motivated. In the Kamas examples in (19) on
p. 255, we may fairly wonder whether a morpheme boundary still exists between
what was historically a gerund marker on the LV and the lexical base of the
auxiliary verb. Unfortunately, since Kamas is extinct, we may never know.
Example (24) on p. 256 presents a Tamil example in which the lexical and
auxiliary verb are said to be fused, cuTappokireen 'I am going to fry' (in
Anderson's transcription the final vowel in the personal ending lacks a macron
for length). The form may be represented as the AVC cuTa.p pokireen; it is not
(yet) fused, since the main verb cuTa 'fry' can host a postclitic particle such
as =um 'and' (Steever 2005: 132-34), indicating that it remains an independent
word. Spelling conventions for Modern Tamil, still in flux and evolving, may
have contributed to this misunderstanding since the elements of tightly knit
constructions may appear graphically as one word (i.e., no internal spaces), as
commonly happens with demonstratives and the nouns they modify. Modern spoken
Tamil does, however, provide an example of what Anderson wants to show in the
AVCs with auxiliary viTa 'leave' (Steever 2005), which has fused with the LV in
the spoken language.
The historical chain from bi-clausal construction to AVC to affix, as urged in
the grammaticalization paradigm and assumed in chapter 7, is primarily
descriptive; more work is needed to coax explanatory potential out of it. It may
be that in language X an AVC developed out of a coordinate construction, but
that can rise only to the level of a necessary condition. Since it is likely
that language X has both coordinate constructions and AVCs, the development of
an AVC is the result of a split from a coordinate structure, not the total
transformation of a coordinate structure into an AVC. We could try to
distinguish the two by saying that in one structure we have a ''full'' verb and in
the other we have an auxiliary, as provided by the various schemata in chapter
7, but this still begs the question of how to characterize the AVC independent
of its historical-typological trajectory. At some point in its development,
changes in the language are targeting a structure for which it may have no
internal precedent, viz. the AVC. It may therefore be necessary to assume that
some version of universal grammar makes the AVC available to human language as a
template. At the other end of this developmental chain, an AVC is characterized
as that which develops into a (verbal) affix. It would seem a frustrating
methodological procedure to wait for an AVC to undergo univerbation before we
can say it is/was an AVC. Moreover, there are several historical outcomes for
AVCs: some may persist as such if the verbal system has reached a state of
equilibrium, others may become affixes and still others may simply be lost (as
in the case of the present tense in Tamil, Steever 1993: 181). Conversely, we
likely cannot recover AVCs from every verbal affix so univerbation cannot be a
sufficient criterion. At this stage, we cannot extract from the
historical-typological trajectory the sufficient and necessary conditions we
would like to have to define an AVC.
While Anderson's contribution provides a thorough analysis of inflectional
patterns in AVCs, it leaves untouched many aspects of AVCs, such as their
synchronic constituent structures, their behavior in larger syntactic contexts,
their interaction with other (morpho)syntactic phenomena, how they relate to
simple verb forms in the language, and the problem of compositionality in the
AVC, among others. Another area of concern in more traditional treatments of
AVCs, almost a fetish in a way, is the issue of the ordering of multiple
auxiliaries in an AVC. The vast majority of Anderson's examples have one LV and
one auxiliary; study of AVCs with two or more auxiliaries would certainly cast
more light on how the auxiliary is syntactically related to the LV as well as
the issue of what an auxiliary is combining with when it combines with both a
lexical and another auxiliary verb. Again, these are not shortcomings in what
Anderson has written about, but areas into which his research might be
It would also be helpful to have the (synchronic) word-formation rules (WFRs) of
particular languages to help us distinguish when we have an AVC and a fused
form. While this would obviously be impractical for all 800 languages in the
database, perhaps a few examples could be undertaken for illustrative purposes.
It would help solve a conundrum Anderson notes on p. 153: ''It is not a priori
clear what constitutes a bound element in a given language in every instance.''
The solution could tell us whether a given expression is an analytic
construction consisting of independent words or a synthetic verb consisting of
several (bound) morphemes. I have found an appeal to WFRs to be helpful not only
in describing and differentiating the two kinds of expressions, but also in
motivating historical change from an analytic structure (AVC) to a synthetic one
(a more complex set of bound verbal affixes, see Steever 1993).
The only real complaint I have about the book in its present form is that
Anderson replaced the ''lexical'' bases of all the auxiliaries in his examples by
the single gloss AUX; we thus cannot see whether the earlier lexical identity of
the auxiliary was 'be', 'have', 'put', 'stand', 'throw', 'spend the night',
etc., and are therefore unable to match the individual examples to the various
verb schemata in chapter 7. I would also like to have seen more statistical
ranking of the various patterns in the data set; for example, on p. 182, he
notes that in 80 percent of double AVCs, the doubled category is inflectional
marking for subject. And if the AUX-headed AVC is the most common in the
database, then just how common is it? The possession of such a vast database
would seem to invite (and enable) such analysis. Some AVCs seem not to be
treated in the text, e.g. the passive, although the passive does appear in the
final list of schemata in chapter 7. Auxiliaries that I have characterized as
marking attitude (Steever 2005) are also generally absent, but there may not
have been sufficient cross-linguistic evidence to include them.
My final set of remarks bears not upon the intellectual content or arguments of
the book but on its readability, and are intended to aid readers, not criticize
the author. The book is fairly free of typographical errors, which is truly
remarkable given the vast number of languages covered. One recurring issue,
though, is the typesetter's dropping of word spaces in the interlinear glosses
so that two adjacent glosses elide, making it difficult to determine which
morphemes go with which word. Moreover, the interlinear glosses do not regularly
align vertically with the individual words in the examples, hampering smooth
reading. The Hittite example (66) on p. 60 is particularly egregious in this
mismatch; it also lacks a translation. In examples. (69a, b), it is not clear
from the glosses that the lexical verbs are verbal nouns; the examples before
(67-68) and after (70-82) have glosses indicating a verbal noun is intended.
Pages 13-14 contain a particularly bad break between the language example and
its interlinear gloss. These typesetting gaffes, for that is what they are, may
have the unintended, but beneficial consequence of slowing readers down and
making them work harder to understand the examples. Readers should also know
that the list of abbreviations is ordered in rows rather than columns. The
abbreviation R (p.41) does not appear in the abbreviations. Unless I have missed
a new terminological development, the term ''default assignation'' (p. 44) might
better be rendered ''default assignment.'' In the gloss in (60b), the form –po in
3-child-po is not explained; perhaps ''possessive'' is intended. Ex. (107) shows a
fused form of the present perfect in Mari in which the auxiliary has been
deleted and the form of the gerund of the LV retained. But is that morpheme now
a marker of the gerund or the present perfect tense? The interlinear gloss
suggests the former; the translation, the latter. On p. 71 and in the
bibliography, the reference to Kachru (1990), should be Kachru (1987). Example
(161) on p. 84 lacks translations. Pace Anderson, in the Parji example (4) on p.
158 the main verb appears in the form of a fully finite verb, not a
(quasi)participial form, and should thus appear in section 4.3.
(Quasi)participial forms, to the extent they occur in Dravidian, do make an
appearance in Konda, Malto and Kurux. In fact, the Konda example (66) on p. 168
does use a ''participial'' form of the lexical verb, though that cannot be readily
seen from just the example given. The only misspelling I found in the running
text, on p. 214, was 'diffferent'. In footnote 1 on p. 303, numerals appear to
be missing from a list of criteria for serial verb formations.
The book succeeds admirably in presenting and analyzing a wide variety of
agreement patterns in AVCs and embedding those patterns within potential
pathways of grammaticalization. It possesses the broadest sampling of languages
in any study of auxiliaries that I know, and is a truly welcome addition to the
literature. Anderson handsomely accomplishes what he set out to do, and provides
a solid basis for continued cross-linguistic study of AVCs. No one interested in
auxiliary phenomena can afford to ignore its contents and its analyses.
Benveniste, Emile. 1965. Structure des relations d'auxiliarite. _Acta
Linguistica Hafniensia_ 9:1-15.
Heine, B. 1993. _Auxiliaries_. New York: Oxford University Press.
Kuteva, T. 2001. _Auxiliation: an enquiry into the nature of
grammaticalization_. Oxford: Oxford University Press.
Steever, Sanford. 1988. _The serial verb formation in the Dravidian languages_.
Delhi: Motilal Banarsidass.
Steever, Sanford. 1993. _Analysis to synthesis: The development of complex verb
morphology in the Dravidian languages_. Oxford: Oxford University Press.
Steever, Sanford. 2005. _The Tamil Auxiliary Verb System_. London: Routledge.
ABOUT THE REVIEWER
Sanford B. Steever is an unaffiliated scholar. He has done research and
published extensively on the Dravidian languages. His interests include the
structure and history of auxiliary verb phenomena in that family.