Aikenvald, Alexandra Y., Robert M. W. Dixon, and Masayuki Onishi,
ed. (2001) Non-canonical Marking of Subjects and Objects. John
Benjamins Publishing Company, paperback ISBN: 1-58811-044-3,
xi+364pp, $49.95, Typological Studies in Language 46 (hardback
ISBN 1-58811-043-5, $114.00)
Elena Maslova, Stanford University
This book was announced on LINGUIST at:
The book consists of a brief editors' preface, a theoretical
introduction, by Masayuki Onishi, "Non-canonically marked
subjects and objects: Parameters and properties", pp.1-52; one
area overview, by Martin Haspelmath, "Non-canonical marking of
core arguments in European languages", pp.53-85; seven language-
specific studies, by Avery D. Andrews. "Non-canonical A/S marking
in Icelandic", pp. 85-111; by Masayuki Onishi, "Non-canonically
marked A/S in Bengali" [Indo-Aryan; standard dialect of West
Bengal spoken in Calcutta, pp. 113-147; by Gabriela Hermon,
"Non-canonically marked A/S in Imbabura Quechua" [Quechua; a
dialect of Northern Highland Ecaudorian Quechua; ca. 30-40
speakers in Northern Ecuador], pp. 148-176; by Alexandra Y.
Aikhenvald, "Verb types, non-canonically marked arguments and
grammatical relations: A Tariana perspective" [North Arawak,
Amazonia, Vaupes area; ca. 100 speakers], pp. 177-199; by John
Roberts, "Impersonal constructions in Amele" [Gum, Papua New
Guinea, ca. 6000 speakers; Haija dialect], pp. 200-250; by
Kristina Sands and Lyle Campbell, "Non-canonical subjects and
objects in Finnish", pp. 251-305; and by Masayoshi Shibatani.
"Non-canonical constructions in Japanese" pp. 307-354; and three
indices (Language Index, Author Index, and Subject Index).
The approach to cross-linguistic investigation implemented in
this book (to my knowledge, pioneered by the Leningrad
Typological School in the sixties) suggests that a group of
experts in specific languages share certain basic theoretical
assumptions and approach "their" languages with similar questions
in mind. The introduction by Masayuki Onishi is intended to set
this common frame of reference. The general assumptions are
essentially those of (Dixon 1994); they are quite common in
typological literature (albeit not uncontroversial, cf., for
example, (Dryer 1997; Mithun & Chafe 1999)):
(i) All languages have transitive clauses with two core
arguments (A and O) and intransitive clauses with one argument
(S); almost all languages have 'extended transitive' clauses
with three core arguments (A, O and E); some languages have
'extended intransitive' clauses with two core arguments (S and
E). It is assumed that this list exhausts "basic" clause types.
(ii) The core categories (A, O, S, E) can be identified in any
language by their *language-specific* properties. There is,
however, a *universal* list of relevant properties; the approach
is thus basically the same as suggested by Keenan (1976), albeit
extended to Os; more or less the usual set of properties is
discussed in the introduction and used throughout the book.
(iii) It is possible to identify "canonical" case marking of
A/O/S in each specific language; some A/O/S in some languages
may receive "non-canonical marking".
The questions to be answered (for each language and, ultimately,
(a) What are the semantic factors that trigger non-canonical
marking of core arguments?
(b) Which syntactic properties of "canonical" A/O/S are
displayed by "non-canonical" A/O/S?
The major phenomena discussed in individual chapters are:
"Dative Experiencers" in "Standard Average European"
(Haspelmath). Haspelmath deals with an areal cluster of languages
("Standard Average European") in which Dative Experiencers
display virtually no subject properties. This means, strictly
speaking, that they are not "non-canonical" A/Ss, but
"canonically marked" Es.
Dative/Accusative subjects (A/S) in Icelandic (Andrews). The
Dative/Accusative A/S are not "prototypical agents", but are
quite subject-like in their syntactic behavior.
Genitive, Objective and Locative A/S in Bengali (Onishi).
Locative signals generic or indefinite reference, Genitive and
Objective, lack of control; the number of subject properties
decreases according to the following scale: Nominative >
Locative >> Genitive > Objective (where ">>" denotes a greater
distance than ">")
Accusative A/S in Imambura Quechua (Hermon): Accusative occurs
with physiological predicates and in the context of Desiderative
suffix on the verb. The latter type of non-canonical A/S retains
more subject properties than the former.
"Dative" (IO-like) S in Tariana (Aikhenvald). This encoding
occurs with verbs of physical states and shares only one
syntactic property with "canonical" subjects (same-subject
constraint in serial verb constructions). More generally, Tariana
is claimed to distinguish three Ss (A-like; O-like; and IO-like).
O-like S in Amele (Roberts). Experiencers and subjects of
Desiderative constructions are marked on the verb by a
cross-reference suffix that is generally controlled by O, but are
A variety of Finnish case marking patterns (Sands & Campbell):
Partitive and Genitive subjects; Nominative objects, etc. Basing
on a detailed description of factors which trigger different case
marking patterns, the authors conclude that virtually all of them
are "canonical" (see also the discussion below).
Double subject ("Nominative+Nominative" and "Dative+Nominative")
constructions in Japanese (Shibatani). Both NPs are shown to have
subject properties. The predicates licensing these constructions
require an additional specification of the "domain" in which its
truth can be evaluated, and this is what the "extra" subject
This brief overview can by no means do justice to the contents of
the book: each language-specific contribution gives a careful and
richly exemplified analysis of A/S/O properties in both
"canonical" and "non-canonical" versions in the given language
and a detailed description of semantic nuances and lexical
constraints (when applicable). The essays are written in a more
or less "theory-neutral" fashion, which allows the authors to
concentrate on the complexities of their specific languages and
makes these fascinating data accessible to every linguist. In my
view, this book is a must for anybody who wants to approach the
problem of universal grammatical functions (or universally
applicable theoretical constructs) in a typologically
(= empirically) responsible way.
A couple of critical remarks. There is a certain discrepancy
between the editors' preface and introduction, on the one hand,
and the book as a whole, on the other. The overall impression is
that the book was intended to be about predicate-related issues,
i.e. about (relatively small) semantic classes of verbs with
unusual valence patterns. The preface introduces the notion of
non-canonical marking as follows: "For most intransitive verbs, S
function is marked in a set way (called the canonical marking)
whereas for a small set of verbs S is marked in one or more other
ways these are referred to as non-canonical marking. Similarly
for A and O." (p. ix). The theoretical introduction is less
specific about this notion, which presumably reflects the fact
that the language-specific chapters actually cover a
significantly broader range of phenomena, including
reference-related factors, animacy, modality, etc. In other
words, the volume as a whole appears to be somewhat less focused
and homogenous than intended by the editors.
What is "non-canonical marking" to begin with? That is, how
should one distinguish between "canonical" and "non-canonical"
marking of A/S/O in a given language (see assumption (iii)
above)? The distinction seems fairly clear if a language have a
single case form for each core argument (say, Nominative for A/S
and Accusative for O -- there are no ergative languages in this
small sample anyway), and this unique marking pattern is used
consistently with almost all verbs and thus can be easily
identified as "canonical". But what if some argument(s) can take
more than one case form with virtually any verb? To give a simple
and well-known example, consider differential object marking,
whereby O takes either "Nominative" or "Accusative" form
depending on its referentiality/animacy. Are both patterns
("A=NOM & O=NOM" and "A=NOM & O=ACC") canonical? Or one is
canonical and the other is not? Interestingly, the authors appear
to assume different answers to this question: Martin Haspelmath
considers "A=NOM & O=NOM" as "canonical", and "O=ACC" as a
deviation from the canonical pattern (pp.56-57), without giving
any reasons for this analysis (why not vice versa?), whereas
Masayuki Onishi takes both options to be "canonical" in Bengali
(p.114). Neither of them addresses this discrepancy (no
cross-references, no discussion).
Essentially the same identification problem is raised in the
chapter on Finnish; Sands and Campbell's conclusion is that much
of what "seems" non-canonical in Finnish turns out to be
canonical; moreover, in descriptions of other languages, "it
would appear that at times minor exceptions and very marginal,
frozen and non-productive patterns which exhibit some
non-canonical marking (...) are not always adequately
distinguished from the main productive grammatical patterns of
the language. In some of these instances, what is at stake might
better be considered lexical properties of particular verbs
(something for dictionaries), not really significant to the
grammar." (p.297). Thus, for Sands and Campbell, "non-canonical"
appears to mean "non-productive". As already mentioned, it seems
that the notion of non-canonical marking as envisioned by the
editors is indeed close to that of non-productive (cf. the
reference to "small classes of verbs"). The problem (as I see it)
is that it is implicitly opposed to the notion of *unique*
canonical (=productive) marking; the cross-linguistic
investigation represented by the book has revealed that a lot of
phenomena fall in between, i.e. languages may have multiple
(relatively) productive case marking patterns. It remains unclear
whether one of these patterns has to be singled out as
"canonical". If yes, how? If not (i.e. if all productive patterns
are "canonical"), then what if a pattern is productive in some
contexts but lexically constrained otherwise (which seems to be
the case in Imambura Quechua and in Amele), i.e. is it still
"canonical" or not? In my view, the book would have gained a lot
if these questions had been addressed in the introduction, or, at
the very least, by means of cross-references between those
chapters which obviously differ in classifying essentially
similar (or just identical) phenomena as "canonical" or
More generally, it would be advantageous to summarize the points
where the data apparently contradict the basic assumptions
outlined in the introduction, as well as theoretical problems
raised by the investigation either in the introduction itself,
or in an "epilogue" (as it was brilliantly done, for instance, by
I. Kozinskij in the methodologically similar volume on
constructions (Nedjalkov 1988)). For example, how do "double
subject" constructions in Japanese fit in the list of basic
clause types (cf. assumption (i))? Is it theoretically plausible
to draw a strict borderline between Dative Experiencers, say, in
German and in Icelandic (insofar as they are classified as
non-core arguments in the former case and as subjects in the
latter, cf. assumption (ii))? Is it an accident that languages
picked up for interesting "non-canonical marking" phenomena are
nominative (rather than ergative)?
To conclude, I would like to stress that these criticisms should
not obscure the basic fact: the volume contains a broad range of
carefully described cross-linguistic data which pertain to the
"core" theoretical problems of linguistics, and I strongly
believe that it must and will play a major role in any further
discussions of these problems.
Dixon, R.M.W. 1994. Ergativity. Cambridge University Press.
Dryer, Matthew S. 1997. Are grammatical relations universal? In:
Joan Bybee, John Haiman and Sandra A. Thompson (eds.). Essays on
Language Function and Language Type (dedicated to T. Givon). John
Benjamins. Amsterdam/Philadelphia (115-143).
Keenan, Edward L. 1976. Toward a universal definition of subject.
In: Charles Li (ed.) Subject and topic. New York: Academic Press
Kozinskij, Isaak. 1988. Resultatives: Results and Discussion. In:
Nedjalkov 1988. (497-526)
Mithun, Marianne and Wallace Chafe. 1999. What are S, A and O?
Studies in Language 23:3 (569-596).
Nedjalkov, Vladimir (ed.) 1988. Typology of resultative
constructions. Typological Studies in Language, 12. John
ABOUT THE REVIEWER
My research interests include methodological foundations of
typology and typology of information-packaging phenomena and
valence-changing operations. As a descriptive linguist, I have
studied the Yukaghir languages of Northern Siberia.