Review of Grammatical Voice
|AUTHOR: Klaiman, M. H.
TITLE: Grammatical Voice
SERIES: Cambridge Studies in Linguistics, 59
PUBLISHER: Cambridge University Press
Andreas Jäger, unaffiliated scholar
In her book Grammatical Voice M.H. Klaiman provides a definition of voice
as a category of the verb that is cross-linguistically valid and on this
basis establishes a typology of voice systems. Voice is characterized as
the relationship between clausal predicates and nominal arguments or
argument positions with their corresponding referents or thematic roles
marked by means of verb morphology. In doing so the author tackles the
problem of delimiting voice phenomena from other categories and strategies
that likewise operate in the domain of predicate-argument relations. As
such the book is a major contribution to the study of verbal categories.
The discussion takes a formally defined concept of voice as the basis,
illustrates the range of associated functions and then goes on to identify
ontological situation structure as the key to understanding the variation
of voice systems.
The first chapter gives a brief historical overview of the study of voice
and then proceeds to define the category under investigation. Voice is
characterised as a morphological mechanism operating in the domain of
predicate-argument structure, which encompasses crucial concepts such as
transitivity, thematic roles, the distribution of core- and non-core arguments.
The author proposes a preliminary threefold typology of voice systems
encountered in the languages of the world: derived voice, basic voice and
pragmatic voice. Derived voice is introduced as a term for changes in
verbal morphology that indicate changes in the allocation of nominal
arguments to structural positions. This type involves a reconfiguration of
arguments and consequently grammatical functions. Passivization is thus
classified as derived voice, since it involves re-assignment of grammatical
functions, particularly to core arguments: The argument that is assigned
the role of patient is prototypically associated with the grammatical
function of object, while in a passive counterpart structure the very same
referent is associated with the subject. A common consequence of this is
detransitivization: The passive counterpart to an active sentence is
intransitive, is has the patient encoded as the subject and the agent
optionally encoded as a non-core argument.
In contrast to this basic voice does without a change of valence. Basic
voice systems are characterised by the observation that different voice
effects may obtain, even though subject and object retain their respective
function and/or structural position in the clause. Sentence pairs in a
given language that are marked for basic voice distinctions remain constant
in the mapping of nominal arguments and grammatical functions. Parallel to
derived voice, however, changes in the involvement of participants are
likewise expressed through basic voice systems. Verbally marked voice
contrasts in active-middle systems indicate the affectedness of the subject
by the action encoded in the verb. If there is a voice alternation, the
participant that invariably functions as the subject is thus perceived as
being affected by the verbal action in different degrees.
The third type of voice system proposed by the author involves marking of
differences in the propositional salience of arguments depending on
pragmatic factors. That is to say the relative, i.e. perceived, importance
of a participant for the proposition to be conveyed in a clausal structure
is reflected in marked voice contrasts. Voice contrasts in languages of
this type fulfil functions such as interrogativity, focalisation and
This chapter examines a number of languages that employ basic voice
systems. It is revealed that all languages of this type have in common only
a subclass of verbs allows voice marking alternations, while others can
only function in predicates marked either for active or for middle voice.
Hence an important function of the category voice is identified, namely
that of a classification parameter for lexical predicates based on their
inherent semantics. Voice alternations are shown to reflect affectedness of
the subject referent, as in the following opposition from Fula (example
(21), pp. 62-63):
(a) mi moor -ii mo.
I braid.hair-PAST.ACTIVE him/her
'I dressed his/her hair.'
(b) mi moor -ake.
I braid.hair -PAST.MIDDLE
'I got my hair dressed.'
In the (b) example the subject does not actively instigate the action, but
is affected by it.
Basic voice systems do not involve a re-mapping of nominal position and
thematic role. What is crucial is the affectedness of the participant that
is invariably encoded as the subject of a sentence. This becomes even more
apparent in the example from Tamil (example (27), p. 71):
(a) kuzantai kalai utai –kir -atu.
child.NOM leg.ACC kick -PRESENT.WEAK-Sg.NEUTER
'The child is kicking its legs (in the air).'
(b) kuzantai ennai utai-kkir -atu.
child.NOM me.ACC kick-PRESENT.STRONG -Sg.NEUTER
'The child is kicking me.'
The case marking of the subject remains constant in this opposition, but
its referent is affected by the verbal action only in (a). The author
points out that the inherent semantics of each verb entail a certain
likelihood of affectedness on the part of the subject participant and basic
voice is a means to organize the verbal lexicon accordingly.
The data presented in this chapter suggests that the basis for
understanding basic voice alternations such as active-middle, weak-strong
or active-stative as presented in chapter 2 is the nature of real-world
situations and our common expectations about their structure (i.e. the
outcome of actions and events in relation to participants, the key term
used being ontological salience). Voice alternations are shown to express
the degree of control that a participant, particularly the subject, has on
the clausal action or its consequences in a given situation. This control
construct may become apparent as a lexical property of the verb, i.e. some
verbs require the logical subject to be undergoer of its action rather than
agent. In transitive predications animacy is a relevant issue, since
animate entities are understood to have more potential for controlling a
verbally denoted process or its outcome. Cross-linguistic data is provided
to support the claim that basic voice systems reflect the control hierarchy.
This chapter presents the direct-inverse voice system, which is
characterised as functionally similar to an active-passive system, but not
at all identical. In contrast to the active-passive system of derived
voice, direct-inverse systems are characterised as pragmatic voice. The
transitivity of a verb is not altered in voice oppositions.
Marked voice contrasts in languages that employ direct-inverse systems
indicate or are sensitive to the ontological salience of nominal arguments
depending primarily on the factors referentiality, animacy and involvement
in the discourse. Speech act participants (1Sg/Pl and 2Sg/Pl) are rendered
most salient, since they are maximally referential, animate and directly
involved in the discourse. On these grounds they are prototypically encoded
as the logical subject in languages that employ direct-inverse voice
contrasts. If the referent is encoded as the logical object, this is marked
not on the nominal but on the verb as a voice contrast. This type of voice
is therefore analyzed as pragmatic in nature. Consider the following
example from Plains Cree (example (1), p. 162):
(a) ni- sekih-a -nan atim.
1- scare-theme.DIRECT-1Pl dog
'We scare the dog.'
(b) ni- sekih-iko -nan atim.
1- scare-theme.INVERSE-1Pl dog
'The dog scares us.'
Both referents involved have the potential to scare one another. It is,
however, argued that discourse participants have greater prominence for the
speaker, which is reflected in the voice contrast, i.e. first person plural
is assigned subject status directly, while 'dog' can only be assigned
subject status indirectly.
A number of languages are reanalyzed in the light of previously established
voice system characteristics. For instance it is revealed that the
intricate system of person agreement in Tanoan languages can be accounted
for in terms of a direct-inverse voice system. Also a correlation between
direct-inverse voice systems and head marking languages as well as lack of
case is attested. Here voice functions as the sole indicator of
predicate-argument relations. Klaiman reveals that pragmatic voice and
derived voice may coexist in a given language. Thus languages that mark a
direct-inverse distinction may as well allow, for instance, passivization.
This chapter is devoted to yet another kind of pragmatic voice, which
encodes information about topic status and focalisation of nominal
participants. Verbally marked contrasts therefore function as an indicator
of relevance of a nominal argument within a given discourse and convey
interrogativity, topichood or focalisation. In contrast to the type of
pragmatic voice described in chapter 4 languages that employ this type of
voice system do not adhere to hierarchies of ontological salience or
animacy when it comes to assigning statuses of subject or object. Rather
this assignment depends entirely on informational salience, as in the
following opposition from Cebuano (example (25a,b), p. 247):
(a) ni- hatag si juan sa libro sa bata.
ACTOR.voice give FOCUS Juan GOAL book DIRECTIONAL child
'JUAN gave the book to the child.'
(b) gi- hatag si juan ang libro sa bata.
GOAL.voice give FOCUS Juan FOCUS book DIRECTIONAL child
'Juan gave THE BOOK to the child.'
Data from the Mayan language family is used to illustrate the relationship
between derived voice behaviour and pragmatic function. It is shown that in
this language family derived voice, i.e. passivization, in fact functions
in the domain of discourse pragmatics. This is contrasted with Philippine
languages, where voice indices of focus and topic are the only mechanism
for the encoding of predicate-argument relations.
The author lays the foundations for a unified analysis of cross-linguistic
voice phenomena with a tentative formalization of the observations
presented in the previous chapters. The category of voice is characterized
as verbally marked alternations of assignment of arguments to positions of
superior ranking. This superiority in turn can be either relational or
informational. The formal mechanism is thus given a common functional
basis, namely that of identifying and marking a hierarchy of participants.
Accordingly the relevance of a certain hierarchy for voice marking
constitutes the proposed types. Generally speaking, derived voice is
sensitive to relations of structural positions, i.e. with logical subject
and object, basic voice is sensitive to a hierarchy of control and
affectedness and pragmatic voice is sensitive to a hierarchy of discourse
The main achievement of this study is the establishment and definition of
voice as a cross-linguistically identifiable grammatical phenomenon with
common formal and functional properties. In contrast to the somewhat
narrower views of voice (cf. Jespersen 1965, Fillmore 1968, Matthews 1974)
the book establishes a comprehensive view of the category of voice and
identifies a number of grammatical effects as voice phenomena, arriving at
a threefold typology of voice systems. It shows that in the light of
sufficient cross-linguistic data it becomes necessary to re-evaluate
longstanding assumptions about the nature and organization of voice
phenomena. The focus is on those voice phenomena that go beyond
passivization, i.e. those that particularly call for a new definition of
participant roles and predicate-argument structure. The author claims that
in these systems, basic voice and pragmatic voice (types 2 and 3), the
function of voice is to represent grammatically signalled variations in the
construal of situations. Situations in turn can be characterised by
participants being involved in events. The author further assumes a certain
likelihood of participant involvement organised in terms of hierarchies.
Certain participants are thus expected to show a certain behaviour in
relation to the event based on their position on such a hierarchy. Any
deviation from this expected behaviour results in a grammatical effect,
namely a voice contrast. Such a hierarchy is the ontological construct of
control. Participants of a situation have an inherent capacity for being
the likely controller of an action or its outcome effect. By the same token
other participants have an inherent capacity for being an affected entity.
By reference to abstract hierarchies of ontological concepts such as
potential control and actual control regulating the variety of coding
strategies of predicate-argument structure the author manages to provide a
definition of 'participant roles' in terms of expected real-world
situations that allows a wider cross-linguistic applicability and thus to
identify a unified characterization of voice as a category of the verb.
As far as the presentation of this highly complex subject matter is
concerned, I would like to add a few comments: On page 42, Klaiman states
that the emphasis is both on the 'varying functions of particular voice
systems, as well as on the category's status in the grammatical
organization of languages'. Croft (1990: 17) states that there are
basically two directions of typological investigations and both may yield
'Significant typological generalizations can be found by examining what
functions are expressed by a given form just as much as by examining what
forms are used to express a given function.'
On the whole the basis for generalization in this study is a grammatical
mechanism, or in the author's words 'grammatical behaviour', that is
identified as recurrent across languages, namely that of morphological
marking on the verb signalling some change in the predicate-argument
relation. This can be understood as a formal basis for investigation.
However, since the initially proposed three classes of voice systems are to
some extent also defined in terms of functional domains such as lexical
verb class marking or clausal pragmatics, a strict directionality is not
adhered to in the course of the study. This becomes apparent in the
three-way typology that serves as the starting point for discussion: The
concepts ''derived voice'', ''basic voice'' are defined formally in terms of
morphosyntactic behaviour. This is juxtaposed with ''pragmatic voice'', which
is defined in terms of function.
In chapter 5 Mayan mechanisms of derived voice are identified as pragmatic
in function, so there is in fact an overlap of 'formal' and 'functional
types (see p. 239). Also in chapters 2 and 3 the path of investigation
switches from a survey of functions associated with a type of voice
contrast to a functional construct, that of participant control, being
analyzed in terms of variation of morphosyntactic coding strategies. It is
only in the final chapter 6 that the common basis is identified, namely the
importance of some (ontological) hierarchy underlying and governing all
investigated voice systems, its formal expression then being subject to
cross-linguistic variation, albeit within the limits of verb morphology.
Furthermore, the author opts for an in-depth analysis of certain languages
and language families rather than discussion of relevant facts in a large
sample. It therefore provides a very thorough and detailed account of voice
phenomena in a rather small number of language families. Even though the
sample shows a considerable degree of diversity, the typological claims
should nevertheless be tested against a broader spectrum of languages in
Despite these minor points of criticism the book makes an important
contribution to typological research and most significantly the
understanding of predicate-argument structure and its organizing mechanisms
on more than one level of grammar.
Croft, W. (1990) Typology and universals. Cambridge: Cambridge University
Fillmore, C. (1968) The case for case, in Bach, E. & Harms, R. (eds.)
Universals in linguistic theory. New York: Holt.
Jespersen, O. (1965) The philosophy of grammar. New York: Norton
Matthews, P.H. (1974) Morphology – An introduction to the theory of
word-structure. Cambridge: Cambridge University Press.
| ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Andreas Jäger is interested in the grammar of verbs, verbal categories and
their cross-linguistic encoding as well as the critical evaluation and
comparison of grammatical frameworks with respect to particular phenomena
such as verb periphrasis. He has done typological research on periphrastic
constructions and speech representation.