Review of Functional Constraints in Grammar
Date: Mon, 28 Feb 2005 11:36:42 +0100
From: Wolfgang Schulze <W.Schulze@lrz.uni-muenchen.de>
Subject: Functional Constraints in Grammar: On the unergative-unaccusative
AUTHORS: Kuno, Susumu; Takami, Ken-Ichi
TITLE: Functional Constraints in Grammar
SUBTITLE: On the unergative-unaccusative distinction
SERIES: Constructional Approaches to Language 1
PUBLISHER: John Benjamins
Wolfgang Schulze (IATS, University of Munich)
The discussion of whether the assumption of so-called unergative (UE) and
unaccusative (UA) verb classes (triggering different syntactic patterns)
helps to better understand and explain distributional 'anomalies' has a
long-standing tradition. It is related to multiple suggestions to classify
the verbal lexicon of different languages (starting e.g. with Vendler
1967). In addition, the discussion (pursued especially in formal theories
of language) follows closely observations on syntactic behavioral
patterns, which stem from the typology of case alignment.
Still, it must be stressed that research on these two types of
verb 'classes' (or syntactic 'classes') concentrates on a number of
basically European languages (there are, nevertheless, some (admittedly
few) studies on unergativity and unaccusativity in non-Indo-European
languages, too -- still the impact of these studies is relatively low).
More precisely, the literature on the given issue even suggests that the
phenomena are a basically 'English' problem. The question of whether UE
and UA represent a behavioral distinction common to many more languages
often is obscured by the fact that very rarely, cross-linguistic studies
take the same diagnostic features as their point of departure. Hence, we
are confronted with a patch-work of arguments that, however, seldom sees
systematization (see Abraham 2004 for a highly illuminating example of how
such systematization can be achieved).
Ergative verbs (or constructions in the broadest sense of the word) are
conventionally defined as verbs that, when intransitive, show the 'same'
type of NP as their 'subject', that occurs as an 'object', if the verb is
used in a transitive construction. The common pattern is (English) 'the
door opened' vs. 'John opened the door' (p. 7). Here, the 'intransitive
subject' is said to stand in an analogous relationship with
the 'transitive object' (hence the term 'ergative'). As this behavior goes
against the standard 'accusative' pattern, the alternative
term 'unaccusative' is frequently used. Accusative verbs, on the other
hand, are conventionally defined as verbs that, when intransitive, show
the 'same' type of NP as their 'subject', that also occurs as a 'subject'
in corresponding transitive constructions. Here, the common pattern is
(English): 'John sang' vs. 'John sang a lullaby' (p. 7). As this behavior
now goes against the standard 'ergative' pattern, such verbs or
constructions are often called 'unergative'. The reader should not that
the term 'unergative' is somewhat misleading: As pointed out by the
authors of the book at issue, the term 'ergative' has a strong 'semantic'
connotation (Greek ergate:s 'worker').
As far as I know, the first application of the term 'ergative' has to be
ascribed to Alfredo Trombetti (1902/03) although it remains doubtful
whether it was this author who had coined the term himself or whether he
took it from P. Wilhelm Schmidt. The standard assumption that the term was
introduced by Adolf Dirr (1912:9: Tvoritel'yj (Activus, Ergativus)), as
proposed by Seely (1977) and still maintained by Dixon (1994:3) should be
If we stick to the original meaning of the term, an 'unergative'
construction would suggest a 'non-volitional, non-agentive' semantics.
However, just the opposite is true: an UE is conventionally
labeled 'agentive, controlling' etc., whereas an UA is said to encompass
the meaning 'uncontrolled, not agentive' etc. Obviously, two different
perspectives clash in the terminology: 'ergative' as a distributional
feature of constructional paradigms, and 'ergative' some kind of 'semantic
Basically, there are two ways of approaching the UE/UA phenomenology: One
the one hand, one can elaborate the diagnostics for an individual
language, neglecting the questions to which extent the UE/UA typology is
validated from a crosslinguistic perspective and whether it is based on
common, maybe universal properties of human 'linguistic cognition'. On the
other hand, one may focus on just this latter perspective, disregarding
peculiarities in the individual languages (to do both things at the same
time seems a tantalizing work). The book at issue (henceforth K&T) takes
the first perspective: It concentrates on English, suggesting basically
five diagnostic tests to validate the 'nature' of unergativity and
unaccusativity and to explore its causal background. Most importantly, K&T
do not take a monocausal perspective. Rather, they assume that "subtle
semantic and pragmatic factors are crucial to understanding the
constraints on grammatical constructions" (p. 29). They continue: "We
further propose constraints on the five English constructions [discussed
in the volume, W.S.], in which formal, functional, semantic, and pragmatic
aspects of the constructions are incorporated as parts of a complex whole,
and one dimension cannot be simply derived, or predicted, from any other
dimension." (p. 29).
It comes clear that here, the authors follow rather closely the
theoretical tenet of Construction Grammar. Else, the general perspective
seems to be directed by a critical reception of Formal Grammar traditions,
although the authors also point out that their framework "is a
continuation of a series of research conducted within the framework of
what is called Functional Syntax" (p. 28) (see for instance Kuno 1980,
1987). K&T thus aim at presenting a multicausal scenario for the
functional dimensions of UE and UA constructions in English. Still, it
must be asked from the very beginning, whether a 'single' phenomenon,
namely the distributional patterns of UE and UA constructions should be
related to a multicausal scenario. An alternative would have been -- as
has been said above -- to unveil a common motivation for all types of
constraints etc., into which these constructional patterns are involved
(see again Abraham 2004). It goes without saying that even such an
approach would not be 'monocausal' in the strict sense of the term,
because it would call for the discussion of metaphorization paths and,
most importantly, for diachronic considerations. For instance, it may well
have been that the constructional type 'the door opened' has been derived
from a middle-reflexive construction as preserved in German 'die Tür
öffnete sich', also compare:
(1) The book sells well. 'Das Buch verkauft sich gut.'
The diachronic process would have been marked for a strong
(formal) 'dereflexivization' of English, based on a constructional type
that by itself was makred for the 'anthropomorphization' of concepts
in 'subject' function of ergative verbs (see Schulze 2000 for this type
of 'promotion'). Another example would be German:
a. Paul erschreckt. 'Paul gets a fright.'
b. Eva erschreckt Paul. 'Eve frightens Paul.'
Superficially, we have to deal with the 'classical' distribution of
unaccusative (or ergative) verbs. Still, the past tense makes clear that
we have to deal with a derivational process that starts from the
a. Paul erschrak. 'Paul got a fright.'
b. Eva erschreckte Paul. 'Eve frightened Paul.'
In fact, the transitive verb ('erschrecken', tr.) is derived from the
intransitive base with the help of a jan-causative. An intermediate state
is reflected by the reflexive 'sich erschrecken' (Past: 'erschreckte
sich') 'to get a fright'. The same holds for a number of 'pairs' that
historically reflect derivational patterns. If we accept the hypothesis
that language is a historical (arte)factum, we arrive at the conclusion
that many, if not most of its constructional patterns are grounded in
processes motivated at some earlier stage of the language,
conventionalized in the habitualization processes of communicative
standards. Hence, it would be of the utmost importance first to isolate
such historical processes and motivations before turning to explanations
based on the assumption of synchronically motivated constructions and
patterns. The fact that Modern English has strongly reduced the
derivational patterns underlying the alleged UE/UA-constructions
considerably obscures the pictures. Sticking to just English data sets the
researcher at risk to be led astray.
The book under review bears a highly promising title: Functional
Constraints in Grammar. In fact, most of what the book does is to explore
such constraints that are conventionally related to UE and UA
constructions. Unfortunately, the authors do not tell in details what they
understand by 'functional'. Here, the above-given reference to the
framework of Functional Syntax (listing a great number of bibliographical
references) is nearly all the reader learns expressis verbis about this
framework. True, much of this framework lurks through the cautious
analyses later in the book; nevertheless, the reader would perhaps have
enjoyed a brief presentation of this framework in order to locate the
arguments in their theoretical frame. Instead, the authors, in
their 'introduction' (pp. 1-29) at length consider issues of UE and UA
constructions. This section is by itself highly informative, although it
must be admitted that it is marked for considerable redundancies: the
alleged nature of UE and UA constructions is summarized again and again, a
fact which renders this introduction not very stimulating. Personally, I
would have enjoyed to see the contents of this chapter being divided into
three parts: 1) The 'problem' and how the book tries to tackle it; 2) the
methodological and theoretical frame; 3) a brief overview on UE and UA
constructions together with a résumé of suggestions on how to analyze and
interpret these constructions. Unfortunately, the authors press many of
these aspects into a single section. This renders the introduction at the
same time ambitious, informative, and superficial.
In their introduction, the authors also refer to traditions to 'free' the
UE/UA-constructions from their syntactic paradigmatics; instead, UE and UA
verb classes are established based on mere semantic criteria (such as
controlhood, agentivity etc.). Not surprisingly, these verb classes are
then paralleled to the classes of intransitive 'active' and inactive'
verbs, as described in the tradition of the famous Sapirian patterns
(Sapir 1917). Still, the equation 'active' verbs = unergative, 'inactive'
verbs = unaccusative remains doubtful (p. 6). For instance, this
distinction, usually known as S-Split (see Schulze 2000 and the references
given there), is sometimes present only with certain 'persons'. Have a
closer look at two of those languages mentioned by T&K (p. 6): In Dakota,
Split-S occurs only with the first and second person (singular; the second
person plural is derived therefrom), but it is lacking in the first person
plural and in the third person. Holisky (1994: 194) summarizes the Bats
facts (East Caucasian, Nakh group) as follows: "If the intransitive
subject is third person, it will invariably be in the nominative [recte:
absolutive, W.S.] case .... If it is first or second person, however, with
some verbs it will be ergative, with others nominative [recte: absolutive,
W.S.]. The choice depends on both the semantics of the verb and the
speaker's belief about the situation in which it occurs." It comes clear
that, here, S-Split shows up as an epiphenomenon of aspects of
personality. In addition, Bats belongs much more to the Fluid-S marking
type (see Schulze 2000) than to a 'true' (lexically determined) Split-S.
In short: It is extremely dangerous to refer to S-Split strategies in
order to set up an UE/UA-typology, without elaborating the details of
these strategies in the individual languages.
It comes clear that the semantic domains described (for reference) in
the 'introduction' of K&T can hardly serve to set up a more general scheme
of UE/UA-patterns. The authors rightly emphasize that there are many
mismatches among languages with respect to the semantic classification of
verbs. They conclude: "[b]ut there is always the possibility also that the
syntactic constructions in question (used to set up UE/UA-classes, W.S.)
might not select unergative or unaccusative verbs ..., but are controlled
by the more complex interaction of verb semantics, sentence semantics, and
the discourse factors involved" (p. 17). Nevertheless, the authors decide
to use the "semantic roles of subject referents as the central criterion
for the unergative-unaccusative distinction" (p. 17). This decision may be
accepted for heuristic purposes, still it sets the authors at risk to
build their house of arguments on rather treacherous grounds. Fortunately,
the authors do not start from a mere lexical approach, that would list the
verbs at issue before testing them against given syntactic properties or
constructional patterns (such a list is offered for instance by Perlmutter
1978: 162-3). Instead, they start from five diagnostic constructions of
English, namely the there-construction (chapter 2), the why-construction
(chapter 3), the cognate object construction (chapter 4), the pseudo-
passive construction (chapter 5), and the extraposition of subject NPs
(chapter 6). K&T do not make fully clear, why they have opted for just
these constructional patterns, but it comes clear that all of them seem to
involve features of an (English-based) UE/UA-typology.
As has been said above, Chapter 2 is devoted to the English 'there-
construction' (pp. 31-65). In English, the use of the clausal initial
topic field has become considerably reduced, compare German:
(4) Gestern ging ich in die Stadt.
'Yesterday, I went to town.' / *'Yesterday went I to town.'
Instead, English has strongly functionalized the clause external focus
place, leaving the clause internal syntax unchanged. Note that for
instance in Standard German, this external slot is not (yet) available:
(5) *Gestern ich ging in die Stadt.
Naturally, the gradual 'closure' of the clause initial (internal) topic
field in English is strongly related to the loss of additional
(morphological) means to indicate grammatical relations. In addition, we
can expect that the shift from internal topic marking to external focus
marking did not happen at once. Rather, this process had been marked by
the gradual reduction of the functional scope of the topicalization
strategy. Residues of this strategy can be expected to occur in (older)
literature, memorized folk tales, and (perhaps) dialects. In fact,
the 'there-construction' discussed by K&T seems to represent just one
instance of this 'fossilization process'. Accordingly, verbs denoting
existence and appearance allow the 'there-construction', whereas other
verbs don't, compare:
a. There occurred a tragic event yesterday. (p. 31)
b. *There played three children in the playground. (p. 32).
Again note that e.g. in German, the corresponding 'da-construction' is
possible with both examples:
a. Da geschah gestern ein tragisches Ereignis.
b. Da spielten drei Kinder auf dem Spielplatz.
K&T first review the standard assumption that the constraints on
the 'there-construction' are linked to unaccusativity. Accordingly, only
typical UA-verbs denoting existence and appearance qualify for the 'there-
construction'. In section 2,3, they show, that there are unergative verbs
(such as rule, creep, crawl, amble, race, spring), which, too, can occur
in the above-mentioned construction. Likewise, they show that
the 'intransitivity constraint' does not apply either, compare:
(8) Then, all of a sudden, there reached her ear the sound of angel
voices. (p. 41)
In addition, K&T mention illustrate that "the acceptability of there-
sentences is not dependent on verbs alone, but on the position of a
locative phrase and/or on semantic and discourse factors, as well." (p.
43). This also holds for a number of UE-verbs that do not indicate
existence or appearance (such as 'swim', 'scream'). The authors carefully
analyze the basic features of the 'there-construction' from the point of
view of formal grammar and than turn a 'functional account' (chapter 2.4).
Here, K&T come to the following hypothesis: "The there construction is
acceptable to the extent that the string to the left of its logical
subject is interpretable as denoting existence or appearance" (p. 47).
This hypothesis is based on the assumption that the verbs in 'locative
there-constructions' are segmented into two layers: One expressing
existence or appearance (the 'logical subject' being linked to the
locative phrase), and another that describes the 'manner/type' of
existence/appearance. For instance, the sentence
(9) Deep in him there burned an underlying passion (p. 46)
can be paraphrased as:
(10) There was deep in him an underlying passion that burned.
In addition, K&T observe that, in English, there is a strong tendency to
restrict the locative 'there-construction' to logical subjects generally
observable for the speaker. The corresponding section (2.4.2) is extremely
interesting and stimulating, still the wealth of data does not allow me to
get into details here. The basic hypothesis goes as this: "[T]he there-
construction must be interpretable as denoting existence (or absence) or
appearance (or non-appearance) observable to the speaker (or the person
whose point of view the speaker is representing" (p. 57). From a
structural point of view, the authors assume "that there is a universal
discourse constraint to the effect that a discourse scene has to be
established first in existential and presentational sentences, and that
relevant characters are introduced into the scene" (p. 59). However, note
that refer to just six languages (four Indo-European languages, plus
Japanese and Chinese) to ground this universality claim. For other
languages, this claim does not hold: For instance, in Udi, an East
Caucasian language, the usual word order in 'there-constructions' (in
Udi 'here-constructions') is as follows ["sh" = s-hachek in original --
mia sa ash bu-ne (Field notes)
here one work exist-3sg
'There is a work (to do) (here).'
a'. *mia bune sa ash.
mia sa gala sa xinär-re bu (Shahvalad, 27)
here on place:dat one girl-3sg exist
'In a certain place, there is a girl....'
b'. *mia sa gala bune sa xinär.
If we simplify the scenario set up by K&T, we can observe that in some
languages, there is a strong preference to first indicate the Ground in
which a 'subject' exists or in which it appears, before the Figure
(or: 'subject') itself is mentioned (G < F). But there are as well
strategies that turn the matter around: Here, the Figure is mentioned
first, followed by the Ground from which it is isolated (F > G). In the
Germanic languages, G < F strategies prevail especially when a new Ground
is established. These strategies are strongly coupled with topicalization.
In English, topicalization gradually became restricted to 'existential'
sentences, most likely the prototypical core of topicalization/focusing
functions; recall that constructions of existence or identification are
frequently used to encode a focus cleft, e.g. French and Welsh:
C'est moi qui vient
it=is me who come:3sg:pres
'It is me who comes' > 'I come'
fe fydd y bechgyn yn dringo'r mynyddoedd
it be:fut:3sg art boy:pl in climbing:on mountain:pl
'It will be the boys (who are) in climbing on(to) the mountains' > 'The
boys will climb the mountains.'
K&T draw a convincing picture of how the constraints on there-construction
are motivated from a synchronic point of view. Still, it remains unclear
why English has developed these constraints. Recall that e.g. in German,
no such constraints exist:
(14) Da tanzte sie im Ballsaal. 'There she danced in the ballroom.' [UE]
(15) Da vergilbte das Papier. 'There the paper yellowed.' [UA]
A possible assumption would be to claim that the motivation described by
K&T (see above) once represented the core domain of the functional scope
of 'there constructions'. The constraints then concerned especially the
peripheral use of the construction, coupled with a gradual metaphorization
of the locative 'adverb' ('there').
Chapter 3 deals with a rather idiosyncratic constructional type,
namely "the way construction" (pp. 67-104). An example is:
(16) Mary danced her way through the park. (p. 67)
Following standard analyses, the way-construction is marked for two
aspects: First, the formula 'POSS + way' superficially occurs as
the 'object' of a nevertheless 'intransitive' verb; second, all verbs
included in this constructional type are said to be unergative
(or: 'active'). In order discuss these assumptions, K&T follow the same
methodological path as in Chapter 2: First, they give an Generative
Grammar account; then, they show to which degree the actual data go
against this account, before proposing a functional analysis. What makes
the chapter at issue special is the fact that the authors extensively
discuss alternative proposals, especially that of Construction Grammar.
After having carefully examined and tested the different way-constructions
and the alleged constraints, the authors come to the following
conclusion: "[T]he Unergative Restriction ... is seriously flawed and
untenable because it is too weak in some cases ... and too strong on
others. This shows that the way construction does not serve as evidence
for unergativity" (p. 78). In their proper analysis, K&T rightly
observe "that 'one's way' is associated with a path phrase ... and that
this path phrase expresses a physical distance through which the subject
referent moves". In fact, the lexical notion of 'one's way' makes appeal
to the well-known cognitive schema 'path' (within the source-path-goal
frame), see Lakoff (1987).
What we have at hand is a blend between the verbal semantics and the
(possessed) notion of 'way': Parts of the mental space of 'way' are
activated within the mental space represented by a given (dynamic!) verb.
The path can additionally be characterized with the help of locative
expressions, but this is not a necessary condition (pp. 80-1). The authors
offer a number of additional data to describe in more details the scope of
the way-construction, before arriving at the following conclusion: "[T]he
[way-]construction becomes acceptable to the extent that it involves a
physical, temporal, or psychological distance, the subject gradually moves
through the whole distance in an unusual manner, and the verb represents
the manner of that movement" (p. 94). Unfortunately, the authors do not
refer expressis verbis to the tradition of Cognitive Linguistics (not
necessarily Cognitive Grammar!) in order to corroborate their extremely
helpful analyses. In addition, they do contextualize the syntactic
problem, namely that 'one's way' is seen as an 'object' of nevertheless
dynamic intransitive verbs. In fact, it may be hypothesized that the way-
construction reflects a constructional type that comes close to the
prototype of Figure > Ground constructions, which often show up as
superficially 'intransitive verbs' (see Schulze 2004a, 2004b).
Accordingly, any intransitive verb is embedded into a transitive frame
(basically Referent -- Verb -- Location, to put it into simple terms),
which however, can be obscured especially with respect to the locative
domain (or its metaphorization). Thus, the way-construction resembles to
accusative-based verbs of motion, compare Latin:
Julius Romam venit
Julius Rom:acc come:pres:3sg
'Julius comes to Rome.'
This pattern, fairly well established for so-called 'accusative languages'
(better: accusatively parameterized constructions) seems to form the
syntactic base for constructional types, in which a dynamic manner
verb 'exports' its Ground to an outer NP, often in an accusative formula.
Unfortunately, the authors do not ask the question a) whether there are
other 'path-Nouns' that can be used in the same constructional type, and
b) to which extent the way-construction has its analog in other languages
(compare the German 'Weg-', the French 'chemin-'construction). Again a
more diachronic and comparative perspective would have helped to support
the yet highly elaborated and landmark analysis of K&T.
In many Indo-European as well as non-Indo-European languages, the so-
called 'cognate object construction' (COC) (or: 'figura etymologica') is a
very common phenomenon. K&T, in chapter 4 of their book, pp. 105-135),
test this construction (in English) against the hypothesis that it is
strongly correlated with unergativity. Conventionally, the COC is
interpreted as a construction that involves 'unergative' intransitive
verbs and a 'semantically / etymologically' related noun in
the 'accusative' case. An example is:
(18) The wolf howled a long howl. (p. 105).
Examples taken from other languages are:
(19) (German) Die Frau tanzte einen schönen Tanz. 'The woman danced a nice
(20) (Old Greek)
douleías douleúein oudemâs hê:tton aiskhrán
slavery:acc suffer:inf not=such few shameful:acc:f:sg
'to suffer the worst kind of slavery'
(21) (Classical Arabic)
Haaraba muHaarabata l-Gunuuni
fight:perf:3sg fight:acc art-mad=person:gen
'He fought like a madman' (lit.: 'the fight of a madman')
The Greek example already illustrates that the COC is not necessarily
restricted to UE-verbs. After giving again a Generative Grammar account,
K&T test the UE-constraints against English 'die, 'blush', 'grow', 'blow'
etc. and come to the conclusion that the UE-constraint does not hold. The
corresponding chapter (4.3) is especially helpful because it summarizes
the path of arguments related to the diagnostics of UA- and UE-verbs. In
their 'functional account' of the COC (chapter 4.4.), the authors first
maintain that COC does not necessarily involve true 'cognate' nouns, as in
(22) He slept a fitful slumber. (p. 118)
This observation is of special importance because it alludes to the
question to which extent a naïve speaker can judge upon the presence of
lexical etymological correspondence. Many COCs indeed are marked for some
kind of 'etymological rhyme', such as laugh (v/n), grin (v/n), smile
(v/n), sleep (v/n), yawn (v/n), sneeze (v/n) etc. This rhyme is even
present in a pair like die/death. In German, the stronger formal
differentiation of verb-noun marking gives even more examples for such
rhymes. e.g. gehen/Gang (go/walk), stehen/Stand (stand/stand). But there
both in English and in German (as well as e.g. in Old Greek), types of COC
that are based on purely semantic rhyming, compare English vs. German:
a. The general died the death of a hero. (p. 111)
b. Der General starb den Tod eines Helden.
An English example for phonetic/semantic rhyming is:
(24) He slept a fitful slumber (p. 118)
This type is called 'non-cognate 'cognate' objects' by K&T -- a rather
unfortunate term. It nevertheless illustrates that a COC is defined rather
by semantic or conceptual features than by true 'etymological' reasons
(which, by the way, have always to be characterized as folk-etymologies,
because the naïve speaker does not have other means to judge upon an
assumed 'cognate' relation than phonetic and semantic resemblance).
Examples of pure 'semantic' rhyming are:
(25) Van Aldin laughed a quiet little cackle of amusement. (p. 118)
(26) (German) Paul lief das Rennen seines Lebens 'Paul ran the race of his
Reviewing the given constraints on COCs, K&T arrive at the following
conclusion: "In the [COC], the cognate object (the whole NP) must
represent a specific state or event that is a subset of the possible
states/events resulting from the action represented by the verb" (p. 121).
This conclusion considers the fact, that in many languages (but not
in 'all' languages'), there is a strong preference to attributively mark
the 'cognate object', see the examples above. For instance, in Old Greek
nearly all COCs are marked by an attribute or a relative clause, rendering
unmarked COCs as collocations, such as phulakàs phuláttein 'to watch a
watch, be on guard', or phóron phérein 'pay tribute' etc.). On the other
hand, the authors observe that "in the passive construction [of COCs,
W.S.], a cognate object without a modifier is acceptable as long as
Passivization is acceptable" (p. 130). Note that e.g. in Classical Arabic,
this option does not hold, compare:
Duriba zaydun Darban shadiidan
hit:pass:perf:3sg Zayd:nom hit:acc strong:acc
'Zayd was struck violently.'
The above-given example also illustrates that one of the major features of
COCs as elaborated by K&T does not necessarily hold for more than English:
In Arabic, a COC may likewise involve a transitive verb, compare:
Daraba-huu Darban shadiidan
hit:perf:3sg:a-3sg:o hit:acc strong:acc
'He hit him hard.'
In sum, K&T have convincingly shown that "the acceptability of the [COC,
W.S.] is not simply a problem contingent upon whether the verb is
unergative or unaccusative, but a semantic, functional, and pragmatic
phenomenon in which the meaning of the verb interacts with the meaning of
the 'cognate' object, together with our knowledge based on our social
customs" (p. 135).
As for the rest of the book, lack of space does not allow me to get into
greater details. In chapter 5 (pp. 137-168), the authors turn to 'the
pseudo-passive construction and unergativity': It is a well-known feature
of English syntax that certain verbs allow some kind of 'prepositional
passive', as in the famous example:
(29) That bed was slept in by Napoleon. (p. 137)
A standard assumption is that this type of passivization is only allowed
with unergative verbs, compare the unacceptable example:
(30) *The bed was fallen on by dust. (p. 139)
The authors convincingly show that the so-called prepositional passive
(or: pseudo-passive) also works for some unaccusative verbs, as in
(31) The conclusion was arrived at late at night. (p. 146)
Obviously, there are other constraints to be accounted for in order to
explain the patterning of pseudo-passives. In their 'functional account'.
K&T elaborate a number of criteria to characterize this type of passives,
dwelling especially on features of 'involvement' and topicalization. They
conclude that the 'object' (that is the 'surface subject' must
be "involved in the actions or states represented by the verb-preposition
sequence" (p. 162). This hypothesis is of extreme importance, because it
implicitly suggests that prepositions are strongly coupled with 'their'
verb, or, to put it into other terms, that prepositions form a subtype of
verbal relations (see Schulze (in press) for some details). Hence, the
example given in (29) actually reads:
(32) That bed was slept=in by Napoleon.
Consequently, this type of passive does not differ from standard passives
such as 'that bed was made by Napoleon', to which specific constraints
apply, too. In addition, the authors argue that the construction at issue
is "acceptable only if passivization can be motivated by the Subject
Preference for Characterizational Sentences", or if is "can be justified
by the Subject-Position Preference for Topics" (p. 163). Again, the
authors stress "that the acceptability status of pseudo-passive sentences
is not a phenomenon based on the verb alone, but a semantic, functional,
and discourse phenomenon based on the meaning of the whole sentence and
its relationship to the context" (p. 168).
Undoubtedly, the chapter on English pseudo-passives is an extremely
helpful and well-done exercise in linguistic argumentation. Nevertheless,
it must be stressed that the argumentation would perhaps have been even
more persuasive, if the authors had consulted the vast literature
on 'locative passives' (or: locative focus) from a typological point of
view (see e.g. Dik 1997).
Finally, chapter 6 (pp. 169-187) turns to 'extraposition from subject NPs
and unaccusativity'. By extraposition is meant that a characterizational
NP linked to another NP can in English be moved away from its NP host,
usually to a position after the verb. An example is:
a. A man with blond hair appeared.
b. A man appeared with blond hair.
Again, it is standard to relate constraints on this constructional
variation to features of unaccusativity. In other words: Constructions
with UE-verbs are said not to qualify for this type of movement. After
having reviewed a formal approach to the problem, K&T nicely elaborate the
weak points of such an analysis. They show that certain UE verbs as well
may be involved in extraposition strategies, e.g.:
(34) An odor awakened me of something burning. (p. 175)
According to K&T, extraposition has not necessarily to do with UA-verbs.
Rather, extraposition "is allowed only if the predicate that the P
[repositional] P[hrase] crosses over represents information that is
discourse-assumed" (p. 176). After having studied a number of highly
illustrative examples, the authors modify this assumption, now
stressing "that the predicate that the PP crosses over represents
anaphorically or deictically grounded information" (p. 180). This
hypothesis is said to be based on the 'Flow-of Information Principle for
Reordering'. Accordingly, less important (given) information is placed
closer to sentence-initial position, whereas segments that represent more
important (newer) information are placed closer to sentence-final position
(p. 181). This 'Principle' comes close to what I call the 'Attention
Information Flow' (AIF, see Schulze 1998, 2004c for details). Still, it
should be kept in mind that the authors' generalization perhaps holds for
a language like English (which allows a postverbal focus field), but other
languages (such as Turkish) may reflect an alternative architecture of the
AIF. In addition, in some languages speakers seem to prefer a balanced
word order, which means that the referential domain is not loaded to much,
compare German (35) which is strongly preferred against (36):
(35) Der Mann verschwand mit wehendem Mantel. 'The man went away with a
(36) ? Der Mann mit wehendem Mantel verschwand.
In addition, the gestalt law of nearness suggests that two NPs in direct
contact inform on a rather 'inalienable' relation, whereas extraposed
constructions encode an alienable relation, compare again the German
example in (35-36) [alienable] and (37-38):
(37) Eine Frau mit Hasenscharte betrat das Geschäft. 'A woman with (a)
hare lip entered the store.'
(38) ? Eine Frau betrat mit Hasenscharte das Geschäft.
It should be noted that most of the examples given by K&T, too, represent
possessive or instrumental constructions. In other words: The question of
(in)alienability typically present with possessive/instrumental patterns
becomes apparent with the authors' examples, too. It seems that
extraposition of the type discussed by K&T is governed not just by
pragmatic features as suggested by them, but also by semantic features
related to the type of linkage between the 'host NP' and the prepositional
phrase subjected to extraposition.
The book concludes with a nice summary. Most importantly, the authors here
offer some kind of "check list for future researchers to use for
determining whether the acceptability / unacceptability contrasts they
have uncovered for a linguistic phenomenon might be due to nonsyntactic
factors" (p. 192). This list includes twenty-one parameters, most of which
are of crucial importance. Here, I cannot dwell upon the question whether
all these parameters (which mainly refer to pragmatic and semantic
features) are justified from an e.g. cognitive perspective. Still, the
reader will greatly enjoy the list because it immediately reflects the set
of arguments used by the authors to dismiss the alleged (five) tests for
unergativity / unaccusativity.
The book ends with notes (which deserve more attention than what normally
is included in such 'notes'), a rich bibliography (which however lacks a
pronounced 'typological' and 'diachronic' perspective), and two indices
(names and 'subject').
K&T's book is a extremely important and highly stimulating book not
necessarily about unergativity / unaccusativity itself, but on the way of
how alleged syntactic or semantic mechanisms should be tested against real
data in order to arrive at a more data-oriented and less formal (and less
monocausal) analysis of linguistic phenomena. The authors have thus
written a wonderful exercise in linguistic criticism, which can be
recommended for researchers in linguistics from which perspective so ever.
The fact that K&T pay much attention to the five diagnostic test, however,
render the book slightly disharmonic. The reader in vain looks for a
general criticism of the alleged unergativity / unaccusativity phenomenon.
Rather, they have to work through the book to realize that this phenomenon
does not pass the five tests.
But does this necessarily mean that the phenomenon by itself does not
exist? In my opinion, in order to answer this question, a much broader
perspective must be taken. It should include crosslinguistic, that is
massive typological evidence, the analysis of diachronic processes,
aspects of (diachronic) pragmatics, cognitive linguistics (not only
cognitive semantics!), and -- last but not least -- a robust theoretical
framework. K&T have occasionally alluded to some of these dimensions;
however, by concentrating on English, they have perpetuated the
unfortunate fact that the unergativity / unaccusativity hypothesis is
mainly based on the analysis of English. In this sense, the reader is
confronted with an empirically extremely well-founded book, which mainly
indicates the 'way' of criticizing the above-mentioned hypothesis. What it
(at least partly) lacks is the indication of and the orientation towards a
more general goal, which would help to dismiss or at least to better
ground the unergativity / unaccusativity hypothesis.
Nevertheless, it goes without saying that K&T's book ranks among the best
books on syntactic issues published in the last year. It is easy to read,
although it must be admitted that the great number of textual redundancies
may provoke the reader to skip whole passages. Doing so, (s)he will be at
risk to miss an important point. The methodological strength of the volume
renders the volume an important tool for teaching the cautious analysis of
linguistics issues. I have not found hardly any typographical or factual
error. This, too, makes the book a pleasure to read.
Abraham, Werner 2004. Ergative diagnostics: temptatio redux.
Sprachtypologie und Universalienforschung 57: 393-401.
Dik, Simon 1997. The Theory of Functional Grammar. Part I: The structure
of the Clause. 2nd revised edition, ed. by K. Hengeveld. Berlin & New
York: de Gruyter.
Dirr, Adolf 1912. Rutul'skij jayzk. Grammaticheskij ocherk, teksty,
sbornik rutul'skix slov s russkim k nemu ukazatelem. SMOMPK 42.3.
Dixon, Robert M. W. 1994. Ergativity. Cambridge: Cambridge UP.
Kuno, Susumu 1980. Functional syntax. E. Moravcsik & J. Wirth (eds.).
Syntax and Semantics 13: Current approaches to syntax, 117-135. New York:
Kuno, Susumu 1987. Functional Syntax: Anaphora, Discourse and Empathy.
Chicago: University of Chicago Press.
Lakoff, George 1987. Women, Fire, and Dangerous Things. Chicago:
University of Chicago Press.
Permutter, David 1978. The English verb (2nd edition, 1st edition 1974).
Sapir, Edward 1917. Review of 'Het Passieve Karakter van het Verbum
Transitivum of van het Verbum Actionis in Talen van Noord-Amerika' by C.
C. Uhlenbeck. International Journal of American Linguistics 1.1:82-86.
Schulze, Wolfgang 1998. Person, Klasse, Kongruenz. Vol. 1 (in two parts).
Die Grundlagen. München & Newcastle: Lincom Europa.
Schulze, Wolfgang 2000. The Accusative Ergative Continuum. General
Linguistics 37: 71-155.
Schulze, Wolfgang 2004a. Radical Experientialism (Foundations of Cognitive
Typology): Materials [Bolzano 2004 lecture; PDF] <http://www.lrz-
Schulze, Wolfgang 2004b. Kognitive Typologie [Mitteleuropa Abend / Bozen
2004: Materialien der Vorlesung; PDF] <http://www.lrz-
Schulze, Wolfgang 2004c. 2004 Pragmasyntax: Towards a cognitive typology
of the Attention Information Flow in Udi narratives. In: Augusto Soares da
Silva, Amadeu Torres, Miguel Gonçalves (eds.) Linguagem, Cultura e
Cognição: Estudos de Linguística Cognitiva, 2 vols, 545-574. Coimbra:
Almedina 2004 .
Schulze, Wolfgang (in press). Review essay on FS Frajzyngier. Studies in
Language 1, 2005.
Seely, Jonathan 1977. An Ergative Historiography. Historiographica
Trombetti, Alfredo 1902-1903. Delle relazioni delle lingue caucasiche con
le lingue camitosemitiche e con altri gruppi. Giornale della Societá
Asiatica Italiana (Firence) 15 (1902).177-201, 16 (1903).145-175.
Vendler, Zeno 1967. Linguistics in philosophy. Ithaca: Cornell UP.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Wolfgang Schulze is the Head of the Institute for General Linguistics and
Language Typology at the University of Munich. His main research topics
include Language Typology, Cognitive Typology, Historical Linguistics,
language contact, the languages of the (Eastern) Caucasus and Inner Asia,
and 'Oriental' languages. He currently works on a Functional Grammar of
Udi, on the edition of the Caucasian Albanian (Old Udi) Palimpsest from
Mt. Sinai, and on a comprehensive presentation of the framework of
a 'Grammar of Scenes and Scenarios' in terms of a 'Cognitive Typology'.