Emergent Grammar

Abstract

Proceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society (1987), pp. 139-157

Full text

Berkeley Linguistic Society, vol. 13 (1987), 139-157
Emergent Grammar
Paul Hopper
‘Fragments are the only forms I trust.’
- Donald Barthelme, See the Moon?
‘Essence is expressed by grammar.’
- Ludwig Wittgenstein, Philosophical Investigations 371
1. Emergent Grammar
1
As explorations in ‘functional grammar’ accumulate in volume and significance, it has
become a standard tactic of supporters of sentence syntax to claim that the very study of
discourse is an unreasonable agenda so long as any problems remain outstanding from the
study of sentence level syntax. This claim sometimes takes the form of challenges to
functional grammar to find discourse correlates of specific syntactic phenomena (stated
always, of course, in sentence terms). For example, we find Jerry Morgan asking the
question how such well-known phenomena as Ross’ constraints on variables could
conceivably be explained in discourse terms (Morgan 1981). Indeed, how could the
‘extraposition’ of the relative clause in such ordinary everyday sentences as:
‘The woman died in 70,000 BC who invented the wheel,’
ever be accounted for if the structure of every sentence in the language had to have a
functional explanation?
The same theme of the arbitrariness of the match between syntax and function is struck
by other linguists who have confronted discourse linguistics from the perspective of sentence
grammar; we can cite here Newmeyer’s pronouncement of a so-called ‘functionalist fallacy’
(Newmeyer 1983), the position that the lamination between structure and function will come
unglued with the slightest bubble of failure—that it takes only one syntactic fact which is
not susceptible of a functional explanation to bring down the whole precarious palace of
functionalism. Sadock (1984) has endorsed the notion of a Functionalist Fallacy.
There are, it seems to me, both superficial and more profound responses to this
argument. The simple one is to note that since both sentence grammarians and discourse
linguists agree that their work is not done, it is highly premature to speak of any ‘syntactic
facts’ which are independent of function, just as in the absence of a complete theory of

functionalism it is premature to claim that all structures have functional counterparts. It
might also be legitimately claimed that so-called syntactic facts which do not appear at first
sight to have a functional explanation may indeed have one if they are studied seriously,
that is, in real discourses. Thus in the particular example presented by Morgan, it could be
argued that extraposition of relative clauses is indeed triggered by something in discourse,
probably involving the relative salience of the main vs. the subordinate clause: extraposition
always seems to mean that the discourse importance of the relative clause outranks that of
the main clause; indeed it is the very absence of this skewing which perhaps accounts for
something I always notice when I try this sentence out on people — that it is judged to be a
very bizarre way to say what the sentence is apparently trying to say.
Critics of ‘radical pragmatics’, and ‘functional grammar’, assume that they and those
they oppose share a common view of language, that there is a pairing of autonomous (i.e.,
decontextualized) grammatical forms with ‘functions’ (whatever they might be in the
abstract), and that the only point of disagreement is whether these forms might be
eventually derivable from ‘functions’ or whether the forms must be described independently
of ‘functions’. I find a certain irony in such a use of the terms ‘function’ and ‘functionalism’,
since the very restriction of the investigation to an artificially defined level of ‘sentences’
seems to me to be quintessentially anti-functionalist. Be that as it may, I am concerned in
this paper with the more fundamental problem of the assumptions underlying the critique,
especially the assumption of an abstract, mentally represented rule system which is
somehow implemented when we speak.
It is an assumption which is very deeply entrenched in our field, and indeed is virtually
an official dogma. Consider the following. A year or so ago, the President of the LSA, Victoria
Fromkin, was asked by the editors of the Chronicle of Higher Education to submit a brief
state-of-the-art report on linguistics, to be featured in a two-page spread of similar reports
by representatives of other disciplines. Here is part of what Fromkin wrote:
‘In human speech production and comprehension, the speaker-hearer
accesses not only the mentally represented language system, but also
other cognitive systems and knowledge of the world.’ (Fromkin 1985:
13)
A whole world of unarticulated philosophical and other assumptions underlies this
statement. It is a world whose traces are glimpsed through terminological windows such as
‘access’, ‘mental’, ‘representation’, ‘language system’, and ‘cognitive system’. And that is only
the intellectual aspect of the statement, for we must not forget that it is simultaneously a
political statement also, a public inscription by the President of the Linguistic Society of the
boundaries and objectives of the field of linguistics. But I am concerned more with the basic
scenario, the one which provides for a logically prior — perhaps eventually even biologically
prior — linguistic system which is simultaneously present for all speakers and hearers, and
which is a pre-requisite for the actual use of language. It is, in other words, the scenario that
when we speak we refer to an abstract, mentally represented rule system, and that we in
some sense ‘use’ already available abstract structures and schemata.

Page 3
The assumption, in other words, is that ‘grammar’ (in the sense of the rules, constraints,
and categories of the language attributed to the speaker) must be an object apart from the
speaker and separated from the uses which the speaker may make of it. That kind of
grammar is conventionally understood to consist of sets of rules which operate on fixed
categories like nouns and verbs, specify the forms of additive categories like those of case,
tense, transitivity, etc., and restrict the possible orders in which words can occur in a
sentence. Discourse, the actual use of language, is held to be in some sense an
‘implementation’ of these structures, or the way in which the abstract mental system
possessed in its entirety by the speaker is realized in particular utterances.
Discourse linguistics has itself not always been immune to this kind of thinking. Here,
too, one frequently encounters the same assumption of a dualistic structure in discourse, the
notion that structure pre-exists discourse and that discourse is mimetically related to a
logically prior abstract organization, formulated this time in terms of paragraphs, episodes,
events, and other such macro-units. The problems of sentence grammar are not really
alleviated by treating discourses as units manifesting a consistent internal structure, in
other words effectively as extra-long sentences. We are still plagued by the problem of the ill-
ness of fit between form and function. However consistently it can be predicted that a certain
particle or aspectual form will function in a particular role in the discourse, it is rare that the
reverse is the case—that a particular form is restricted to a single specifiable discourse role.
To cut a very long story short, and thereby probably caricature the dilemma, some way out of
the vicious circle of form-to-function-to-form is needed.
This is, then, roughly the context in which the term Emergent Grammar is being
proposed. The term ‘emergent’ itself I take from an essay by the cultural anthropologist
James Clifford, but I have transferred it from its original context of ‘culture’ to that of
‘grammar’. Clifford remarks that ‘Culture is temporal, emergent, and disputed’ (Clifford
1986:19). I believe the same is true of grammar, which like speech itself must be viewed as a
real-time, social phenomenon, and therefore is temporal; its structure is always deferred,
always in a process but never arriving, and therefore emergent; and since I can only choose a
tiny fraction of data to describe, any decision I make about limiting my field of inquiry (for
example in regard to the selection of texts, or the privileging of the usage of a particular
ethnic, class, age, or gender group) is very likely to be a political decision, to be against
someone else's interests, and therefore disputed.
The notion of Emergent Grammar is meant to suggest that structure, or regularity,
comes out of discourse and is shaped by discourse as much as it shapes discourse in an on-
going process. Grammar is hence not to be understood as a pre-requisite for discourse, a
prior possession attributable in identical form to both speaker and hearer. Its forms are not
fixed templates but are negotiable in face-to-face interaction in ways that reflect the
individual speakers' past experience of these forms, and their assessment of the present
context, including especially their interlocutors, whose experiences and assessments may be
quite different. Moreover, the term Emergent Grammar points to a grammar which is not
abstractly formulated and abstractly represented, but always anchored in the specific
concrete form of an utterance.

The notion of emergence is a pregnant one. It is not intended to be a standard sense of
origins or genealogy, not a historical question of ‘how’ the grammar came to be the way it ‘is’,
but instead it takes the adjective emergent seriously as a continual movement towards
structure, a postponement or ‘deferral’ of structure, a view of structure as always
provisional, always negotiable, and in fact as epiphenomenal, that is at least as much an
effect as a cause.
The assumption of Emergent Grammar imposes on the linguist a rather radically
different view of the data base for linguistics. Although isolated, made-up clauses and
sentences will have their uses, and indeed are often indispensable short-cuts to the study of
grammar, the sources of these forms will have to be understood in a different way from that
of the fabula of abstract rules and native speaker intuitions which have become part of our
dogma. The linguist's task is in fact to study the whole range of repetition in discourse, and
in doing so to seek out those regularities which promise interest as incipient sub-systems.
Structure, then, in this view is not an overarching set of abstract principles, but more a
question of a spreading of systematicity from individual words, phrases, and small sets. I
will illustrate what I mean by considering the example of the English indefinite article a/an.
If we consider the history of this form, we find that from Indo-European times a cognate
form has meant the simple numeral ‘one’, singularity. This was still a common meaning of
án in Old English. It is seen in such examples (from Bosworth and Toller 1898, sub án) as:
God geworhte ánnan mannan, Adam, of láme
‘God created one man, Adam, out of clay’
It is also commonly used to introduce a new participant into a discourse:
án man hæfde twegen suna
A [certain] man had two sons’
Its use as a general indefinite article does not appear until later, so that in Old English án is
not found in such contexts as:
Deodric wæs Cristen
‘Theodoric was a Christian’
To take just these three functions of the predecessor of a/an in Old English, we find in
modern English not a uniform, over-all weakening of the meaning, but rather a situation in
which the weakened meanings and the older stronger meanings exist side-by side. Thus we
find, among other uses, the indefinite sense of a non-specific, classifying article:
My husband and I went to a showroom to buy a new car. After we had
test-driven one, the salesman asked us ... (from Redbook)
But the specific, new-mention sense is also found:

Page 5
They introduced me to a young woman [whose name was Ethel].
My husband and I went to a showroom to pick up a new car we had
ordered. After we had test-driven it, the salesman asked us ...
Although these senses — specific and non-specific — have usually been taken as
exhaustively dividing up the domain of the indefinite article, in fact several other uses also
exist, such as ‘one and the same’:
Birds of a feather flock together
They are all of a kind
and even ‘one’:
A stitch in time saves nine
A penny saved is a penny earned
How much is that picture-frame? — A dollar.
—where, by the way, British English would require ‘one dollar’. It is significant that these
meanings of ‘one’ and ‘the same’ are not replicable outside of the contexts — and in some
cases the specific wording — of these formulas. Thus discourses like the following seem
anomalous:
Linguists of a theory attend the same conferences.
What was left of the woods after they built the parking-lot? — A tree.
Evidently the meanings represented by the English ‘indefinite article’ are not unified under
one hyper-abstract function. Instead, an open ended set of small sub-systems has come into
being, and the membership of new occurrences of forms with the indefinite article is not
specifiable in advance, but is impromptu and negotiab1e. Even participants in the
conversation may not know whether a specific new mention or a non-specific indefinite is
intended until this has been worked out in the verbal interaction. Moreover, these subsystems
are either innovating and spreading out from an earlier more restricted usage, or are
contracting and being abandoned from an earlier wider use. We see this most clearly in the
obviously traditioned, formulaic diction of proverbs like ‘birds of a feather’, where ‘a feather’
retains not only the older sense of ‘one and the same’, but also the singular noun ‘feather’ in
the sense of plumage. The spread of the newer, indefinite-nonspecific function of a/an was
described in Hopper and Martin 1987.
The point about the retention of archaisms in proverbial language has of course often
been made. But it has less often been noted that proverbial language is only an extreme case
of repetition in discourse, at the other end of which are the morphological and syntactic
repetitions some of which are called grammar; this point is made cogently by Lambrecht
1984. In other words, real live discourse abounds in all sorts of repetitions which have
nothing to do with grammar as this is usually understood: for instance, idioms, proverbs,

clichés, formulas, specialist phrases, transitions, openings, closures, favored clause types,
and so on. There is no consistent level at which these regularities are statable. They are not
necessarily ‘sentences’, or ‘clauses’, with recurrent internal structure, but they are often used
holistically. Their boundaries may or may not coincide with the constituent boundaries of our
grammatical descriptions: subject and predicate, noun phrase, prepositional phrase.
Moreover, what is a formulaic expression in one context may not be in another; again, see
Lambrecht 1980.
It has been noted before that to a very considerable extent everyday language is built up
out of combinations of such prefabricated parts. Language is, in other words, to be viewed as
a kind of pastiche, pasted together in an improvised way out of ready-made elements.
Language is thus to be treated, in Wittgenstein's words, ‘from outside’ (cf. Wittgenstein 1958:
para. 120) — not as governed by internalized mentally represented rules, but by pre-existent
material with which discourses can be devised; cf. Staten 1984: 85-86, Smith 1978: 61-62 et
pass. Evidently an entirely parallel way of viewing language is to be attributed to Jacques
Derrida with his metaphor of language as ‘graft’: new speech acts are ‘grafted onto’ old ones
and of course serve in turn as the stock onto which further new speech acts are crafted (cf.
Culler 1982: 134-135). Becker's idea of ‘prior texts’ (e.g., Becker 1979: 244-245) is also crucial
here: previous actual utterances form the basis of new utterances. Similar observations have
been made by Bolinger, by Andrew Pawley, and others. It is this pre-patterned, pre-
fabricated aspect of speech which accounts best for the characteristic of language for which
no dualistic, double-tiered theory can provide an intuitively satisfying explanation: in
natural discourse we compose and speak simultaneously (Smith 1978: 60). There is no room
— no need — for mediation by mental structures. It is in this sense that, as Bolinger has
pointed out (Bolinger 1976), speaking is more similar to remembering procedures and things
than it is to following rules. It is a question of possessing a repertoire of strategies for
building discourses and reaching into memory in order to improvise and assemble them.
Grammar is now not to be seen as the only, or even the major, source of regularity, but
instead grammar is what results when formulas are re-arranged, or dismantled and re-
assembled, in different ways.
Looking at language this way involves a serious adjustment for the linguist, since we
have developed the habit of seeing utterances in terms of a fixed framework of rules, and
especially because we have been raised on the doctrine of the free generability of sentences,
and the privileging of novelty over prior texts. Indeed, novelty is a prized virtue in our
society altogether,
2
and we have many ways, some more subtle than others, of censuring
perceived repetitions of others’ behavior and an enormous vocabulary dealing with repetition
(copying, imitation). Yet when one examines actual specimens of speech from the formulaic
point of view the effect is a striking one, perhaps even a memorable one, in that it is then
extremely difficult to revert to the old rule-governed syntactic view of discourse. Consider
the following example from spoken English, just one of many examples from the Carterette
and Jones corpus:
Well no the problem is and this is what the psychologist has mentioned
to me. these kids wont wont show any hope like the see you take a

Page 7
normal uh the average retarded child i mean the one who doesnt have
any handicaps like blindness or deafness or something like that. he will
improve a little bit. maybe a lot. it depends on how badly disturbed he is.
but these people wont because theyre still going to no matter what
happens theyre going to be living in a fantasy world. because theyre
blind. and they have to imagine and they keep asking one question after
the other and then nothing they say makes any sense and nothing is
relevant to the situation. and it never will be because they well theres just
such a sharp line of differentiation between the normal blind and then the
emotionally disturbed blind. (Carterette and Jones 1974:422).
Even a cursory study of such passages reveals several different layers of regularity. The
formulas are easily isolated. Note just a few of them:
the problem is
has mentioned to me
these kids
you take
a little bit
maybe a lot
it depends on
no matter what happens
theyre still going to
living in a fantasy world
one question after another
nothing they say makes any sense
relevant to the situation
sharp line of differentiation
emotionally disturbed
—the last with its institutional and authoritarian subtext. It would in fact be difficult or
impossible to draw the line between a formulaic and a non-formulaic expression; Moreover,
there are single words which could themselves be said to constitute formulas in this context,
such as ‘disturbed’, ‘normal’. The stops and starts coincide with the boundaries of formulas,
which are presented and modified or withdrawn or capitalized upon in an obvious
interactional negotiation. Early in the paragraph, for example, the speaker clearly is about
to say ‘the normal retarded child’, but some way into the phrase realizes that for the
uninitiated in this context it clashes with another formula, ‘the normal child’, and launches
into a second try, ‘the average retarded child’, which also — once said — appears
incongruous (cf. ‘the average child’), and finally is forced to abandon the search for an
appropriate formula and move into a more specific level of discourse in which the properties
encapsulated in the adjective ‘normal’ are made explicit:

these kids wont wont show any hope like the see you take a normal uh
the average retarded child i mean the one who doesnt have any
handicaps like blindness or deafness or something like that.
It might be suggested that in this particular passage a sort of second hand ‘health care
professional’ jargon is manifest, in which mannerisms peculiar to a particular set of experts
intrude. (We might note, for example, the pervasive ‘will/won't’ in place of the present tense,
right out of the H.C.P.'s manual!) Yet it would be difficult to find a passage about which
some analogous remark might not be made. The point is that all discourse is in some sense
specialist discourse, molded to the speaker's personality (i.e., personal history), the situation,
and the topic. It is precisely the point about Emergent Grammar that such ‘heteroglossic’
(Bakhtin 1981:281) aspects of language necessarily become integral parts of the linguistic
description, and are not set aside as a separate agenda irrelevant to the linguistic code and
its structure.
Some of these phrases are incongruous when considered from a structural-grammar
perspective. Consider the phrase beginning ‘You take a...’, which would have to be analyzed
as a subject-verb construction. Its actual function in the monologue is quite different from
what one might predict of such a phrase on structural grounds. It is not a report ascribing an
action of taking to a second person subject. It has in fact only one function, expressed
holistically: to present a new hypothetical case into the discourse context. But this function
cannot be readily integrated into a homogeneous grammatical system whose postulates
obtain only at the level of isolated sentences and which starts with the perspective of a
solitary ideal speaker.
The systematicity which linguists have come to expect in language exists, of course,
but in a more complex way. The linguistic system is now not to be seen as something
complete and homogeneous, in which ‘exceptional’ phenomena must be set aside as
inconvenient irregularities, but as a growing together of disparate forms. This convergence
takes place through lateral associations of real utterances. Similarities spread outwards
from individual formulas, in ways that are motivated by a variety of factors, such as:
(i) phonological similarity (rhyme, assonance): he's likely to —> he’s liable to
(ii) contextual similarity: I persuaded him to —> I convinced him to,
and other kinds of resonance.
3
They do not, however, merge into the kind of uniform
grammar which would lead one to posit a uniform mental representation to subtend them.
2. Preferred Clauses
2.1 What I've been saying up to now has had the purpose of recontextualizing the notion
of grammar—not to abolish it, but rather to suspend it with a view to isolating those
regularities in discourse which we will agree to call emergent grammatical regu1arities. But
as we have seen, the doctrine of Emergent Grammar assigns an entirely different status to
grammar from what might be called A Priori Grammar:

Page 9
(1) Regularity in discourse is of many different kinds, and is, since there is continually
movement between one kind and another, moreover dynamic, not static in nature.
Consequently no principled line can be drawn between the emergent regularities designated
to be ‘grammatical’ and other regularities deemed to be ‘rhetorical’, ‘formulaic’, etc.
2) Because grammar is always emergent but never present, it could be said that it
never exists as such, but is always coming into being. There is, in other words, no ‘grammar’
but only ‘grammaticization’— movements toward structure which are often characterizable
in typical ways. It goes without saying that many phenomena which we would agree to call
grammatical are relatively stable and uniform. That is not in dispute. The point again is that
any decision to limit the domain of grammar to just those phenomena which are relatively
fixed and stable seems arbitrary.
(3) The major descriptive project of Emergent Grammar is to identify recurrent
strategies for building discourses — strategies which have intra-linguistic or inter-linguistic
generality (or both) and which move toward grammaticalization along parallel lines.
In studying discourse with a view to describing emergent regularities, it is therefore
most useful to begin by establishing frequently occurring, relatively stable clause types. A
useful concept here is that of the ‘figure’, suggested by Pete Becker. A figure is a phrase or
clause which is highly standardized in its format and which permits substitution in a few
restricted places. It has a rudimentary internal structure, but it is much closer to a formula
than to freely generated ‘sentences’. To the extent that discourse is not prefabricated, it
consists for the most part of assemblages of a small number of such figures. Knud
Lambrecht's notion of a ‘Preferred Clause Unit’ seems to be quite similar, only Becker's
concept of a ‘figure’ permits a number of such types of clause unit to be reckoned with.
Consider the following examples from Old English (Plummer 1899):
1. ond ∂a geascode he ∂one cyning ‘And then he found the king’
lytle werode ‘with a small band of men’
on wifcu∂∂e ‘a-wenching’
on Merantune, ‘in Merton’
ond hine ∂ær berad, ‘and caught up with him there’
ond ∂one bur utan be eode ‘and surrounded the hut outside’
ær hine ∂a men onfunden ‘before the men were aware of him’
∂e mid ∂am kyninge wærun; ‘who were with the king’
2 ∂a ridon hie ∂ider, ‘Then they rode up’
ond his aldormon Osric, ‘and his alderman Osric’
ond Wifer∂ his ∂egn, ‘and his thane Wiferth’
ond ∂a men ‘and the men’
∂e he be æftan him læfde ær, ‘which he left behind him earlier’
ond ∂one æ∂eling on ∂ære byrig metton, ‘and met the prince in
the villa’
∂ær se cyning ofslægen læg... "where the king lay slain"

‘Then his alderman Osric, his thane Wiferth, and the men he had left
back earlier rode up, and found the prince in the compound where the
king lay slain...’ (755 AD)
Here, a handy way of building up a discourse, such as a narrative, is to construct it by means
of a verb-initial clause, usually preceded by a temporal adverb such as ∂a ‘then’; this clause
typically elaborates a setting for an action, and may contain a number of lexical nouns
introducing circumstances and participants:
ond ∂a geascode he ∂one cyning ‘And then he found the king
lytle werode ‘with a small band of men
on wifcu∂∂e ‘a-wenching’
on Merantune, ‘in Merton’
∂a ridon hie ∂ider, ‘Then they rode up’
ond his aldormon Osric, ‘and his alderman Osric’
ond Wifer∂ his ∂egn, ‘and his thane Wiferth’
ond ∂a men ‘and the men’
It is followed by a succession of verb-final clauses, in which lexical NP's are minimally
represented. These verb final clauses are built up of a particle such as ond ‘and’, one or more
initial pronouns unrestricted as to case, perhaps a lexical noun or adverb, and then the verb:
ond hine ∂ær berad, ‘and caught up with him there’
ond ∂one bur utan be eode ‘and surrounded the hut outside’
ær hine ∂a men onfunden ‘before the men were aware of him’
∂e mid ∂am kyninge wærun; ‘who were with the king’
It is not a question of an invariant hyperform from which different clauses are derived by
processes of deletion and movement. Instead it seems that constructions spread outwards
from a small nucleus and in turn form new nuclei (something like the metastasis of
malignant cells, to co-opt a metaphor of Bolinger's), and the resultant array of clauses are in
‘family resemblance’ relationships to one another. Among the features of the figure is a
tendency to avoid phrasing which would lead to multiple lexical noun participants in the
same figure. So in (2), the multiple agents ‘they’, ‘his alderman Osric’, ‘his thane Wiferth’ and
‘the men’ are distributed over several phrases, and only one of them (hie ‘they’) is retained in
the figure. Lambrecht's work on Spoken French (e.g., Lambrecht, 1987) shows how a large
part of its grammar is invested in strategies for preserving the external form of what he calls
the Preferred Clause Unit.
3 The Malay Ergative: An Emergent Construction.
I will conclude by considering some consequences of emergent grammar for morphology.
A major postulate, or working hypothesis, of Emergent Grammar is that the more useful a
construction is, the more it will tend to become structuralized, in the sense of achieving

Page 11
cross-textual consistency, and serving as a basis for variation and extension. An elementary
example of this is ‘Watkins' Law’ (Watkins 1962: 93-96; Collinge 1985: 239-240). Calvert
Watkins has noted that the third person singular of a paradigm forms the basis for new
paradigms. The particular interest of Watkins' observation is his point that there are
asymmetries among the persons, which in fact play quite different roles in discourse, having
eventual consequences for the development of paradigms.
In written Malay texts (those used in this study were Abdullah 1932, Abdullah 1928,
both written in the 1840's), a highly frequent and favored clause type consists of a transitive
verb with an enclitic ergative pronoun, followed by a simple lexical patient. This clause type
is found in numerous types of context; the following exemplify narrative (3), and procedural
(4) discourse; the verb + clitic complex is italicized:
(3) Hata maka di-panggil-nya aku masok ka dalam bilek ‘After that
he summoned me into the room’
tempat ia menulis, ‘where he wrote’
maka di-tulis-nya sa keping surat; ‘and he wrote a letter;’
sa telah sudah, ‘when he had finished’
maka di-buka-nya peti-nya, ‘he opened his sea-chest’
di-ambil-nya tiga puloh ringgit, ‘took out thirty dollars’
di-unjokkan-nya surat serta wang itu, ‘handed [me] the letter and
the money’
(4) Maka erti salang itu, ‘And [execution by] salang means’
di-ikat-nya kaki tangan orang itu, ‘they bind the man hand and
foot’
lalu di-dudokkan-nya di-haluan perahu, ‘and put him in the bow of a
boat’
di-kayohkan-nya kapada sa buah anak sungai. ‘and row him to
one of the backwaters of the river’
The characteristic particles lalu and maka are closely comparable to the ond of Old
English; the verb has a prefix di-, which serves as an agreement prefix with third person
agents, and usually an enclitic -nya meaning ‘he, they (ergative)’, giving a transitive clause
beginning with di-V-nya ‘he, they V'ed (it)’. Discourses may now be constructed by stringing
together these transitive clauses, together with a few other quite easily characterizable
types, substituting new nouns and verbs as needed, but generally keeping the basic shape of
the figure intact.
The argument structure of these figures is along the lines of Du Bois’ Preferred
Argument Structure (Du Bois 1986). Agents are generally continuous as topics, and are
either zeroed or represented simply by the enclitic nya. Lexical nouns are for the most part
non-agents, such as patients, indirect objects, and obliques of various kinds. If there is a
lexical agent, this has a preposition oleh, ‘by’. But it will be noticed that lexical agents are
relatively few and far between; some examples of them are:

(5) maka anak-nya perempuan itu pun hendak menangkap ikan itu,
‘and his daughter tried to pick up the fish’
sa-telah di-tangkap-nya dari ekur-nya, ‘when she took hold of it by its tail’
maka di-kebaskan oleh [ERG.] ikan itu tangan-nya, "the fish jolted her
hand’
(6) maka anjing itu hendak pergi menchari ayer di-sungai itu, ‘and
the dog went down to the river to find water,’
maka tiba-tiba di-sembar oleh [ERG] buaya ‘and was suddenly snapped
up by a crocodile.’
(7) Maka oleh [ERG] Tuan Farquhar ‘And Mr. Farquhar’
di-suroh-nya ambil bangkai buaya itu, ‘had them get the crocodile's
body,’
di-gantong-nya di-pohon jawi-jawi ‘and he hung it from a fig-tree’
In example (5), there are two lexical nouns, an agent oleh ikan itu ‘by the fish’, and an
absolutive, tangan nya ‘her hand’, and the verb is di-kebaskan ‘shock, jolt [her]’. In example
(6), the agentive phrase is oleh buaya ‘by a crocodile’, and the patient is zero, being continued
from the previous clause; the verb is di-sembar ‘snap [it] up’. Transitive agents which are
lexical nouns, such as buaya ‘crocodile’ in (6) above, take the preposition oleh provided they
are specific participants in the discourse. ‘Specific’ usually means definite in the sense of
having been referred to previously in the discourse; but the noun may be new, as here, and
its individuation then depends on subsequent mentions in the discourse. Lexical agents
which are neither old nor subsequently mentioned — i.e., which do not qualify as specific
participants in the discourse — do not take oleh, as in:
(8) ada yang di-makan harimau ‘some were eaten by tigers’
where harimau ‘tiger’ is the lexical agent of di-makan ‘eat (transitive)’. These ‘indefinite,
non-specific’ nouns, then, behave like agentive pronouns in lacking the preposition and being
placed immediately adjacent to the verb stem.
Now the presence of an ergative preposition before a lexical agent is explained by the
Preferred Argument Structure. Lexical agents, being highly marked in discourse terms,
must receive a special indicator, in this case the agentive preposition oleh. But indefinite-
nonspecific lexical agents like harimau ‘tiger’ in (8) should not be exempt from case marking.
On the contrary, they are if anything even more highly marked as agents than definite
specific lexical agents. I return to this point in (iii) below.
Although most transitive agents are placed immediately after the verb in the preferred
clause unit, the lexical transitive agent may also appear outside the clause, in very much the
same way that we saw in Old English that ‘extra’ NP's are placed outside the nuclear clause.
Thus in the next example the ergative agent is placed outside the clause in front of the verb:
(9) Maka oleh [ERG] Tuan Farquhar ‘And Mr. Farquhar’

Page 13
di-suroh-nya ambil bangkai buaya itu, ‘had them get the crocodile's
body,’
di-gantong-nya di-pohon jawi-jawi ‘and he hung it from a fig tree’
There are numerous examples of this ‘extraposition’ of the ergative; it is especially found
when a single lexical agent is shared by several subsequent clauses, as here. The ergative
phrase then has a domain which extends over a number of clauses, and in fact has features
of an independent clause in its own right.
Now I want to suggest that this is exactly what is happening— that the prepositional
(lexical) ergative is emerging out of a ‘serial verb’ construction which sometimes re-appears
in its original clausal form, in much the same way that the English indefinite article
sometimes appears in contexts where its earlier specific sense is reflected. My reasons for
saying this are the following:
(i) That the ‘preposition’ oleh is verbal in origin is indisputable. Compound forms of the verb
still exist: beroleh ‘to obtain’, oleh-oleh ‘something brought back as a gift’, and the modal
boleh ‘be able, be allowed to’. They suggest a meaning like ‘acquire, achieve, manage,
accomplish’ which seems well within the typology of grammaticization of ergative
prepositions out of verbs.
(ii) The possible independence of the agentive clause from the action clause is seen nicely in
the next example (10), in which the verb is in the meng- prefixed form rather than the di-
form (‘passive’) otherwise invariably found with the ergative:
(10) Maka oleh Grandpre memberikan-lah surat itu ka tangan Enche Ha
and [name] meng:give PCLE letter the to hand Mr. [name]
‘And Grandpre handed the letter over to Mr. Ha.’
The sense that the argument ‘Grandpre’ is shared between two clauses in a serial fashion is
striking.
(iii) As previously noted, the agent of the ergative with oleh is always definite or specific. In
other words, the lexical agent with oleh retains characteristics of a topic/agent, and no doubt
reflects an original definiteness constraint on agent/topics. By contrast, lexical agents which
are non-specific were never appropriate topics of oleh, while non-lexical (i.e., pronominal)
agents were always cliticized to the main verb.
(iv) When the ergative agent is separated from the action clause, the action clause usually
also has the clitic agentive -nya, that is, the agent is referred to twice, as in example (9). This
is exactly the same as when an agent is introduced in a previous separate clause and
referred to again.
This is of course an example of grammaticization of the classical kind which has often
been noticed in the literature. What I have wanted to stress here is the need to understand
not only the formal process but the way in which that formal process emerges from a
discourse context, in other words, is anchored in particular, concrete utterances. It is this
‘prior textuality’ of the construction which explains why it has retained properties of a
separate, external clause. What we see emerging, then, is a new strategy for permitting a

lexical agent to be incorporated into a nuclear clause under certain contextual conditions,
presumably involving differences of topic continuity.
4. Conclusion
I conclude this paper with some syllogisms, extrapolated from the first couple of pages of
Radford’s textbook on transformational syntax. References, with emphasis as in the original,
are to Radford 1981:
“What is a grammar of a language? Chomsky gives an essentially mentalist answer
to this question: for him a grammar is a model (=systematic description) of those
linguistic abilities of the native speaker of a language which enable him to speak and
understand this language fluently. ... Thus a grammar of a language is a model of the
linguistic competence of the fluent native speaker of the language.” (p. 2)
“...in the case of a sentence such as:
(1) He thinks that John is wrong
it is the native speaker’s grammatical competence (his knowledge of the grammar of
his language) which tells him that he cannot be interpreted as referring to the same
person as John in (1).” (p. 3)
Some Syllogisms
[1A] A grammar of a language is a model of the linguistic competence of the fluent
native speaker of the language.
[1B] A model is a systematic description.
THEREFORE:
[1C] A grammar of a language is a systematic description of the linguistic
competence of the fluent native speaker of the language.
[2A] Grammatical competence is the native speaker's knowledge of the grammar of
his language.
[2B] (= [1C]) A grammar of a language is a systematic description of the linguistic
competence of the fluent native speaker of the language.
THEREFORE:
[2C] Grammatical competence is the native speaker's knowledge of a systematic
description of the linguistic competence of the fluent native speaker of the language.
[3A] “...in the case of a sentence such as:
(1) He thinks that John is wrong

Page 15
it is the native speaker's grammatical competence (his knowledge of the grammar of
his language) which tells him that he cannot be interpreted as referring to the same
person as John in (1).” (p. 3)
[3B] = [2C] Grammatical competence is the native speaker's knowledge of a
systematic description of the linguistic competence of the fluent native speaker of the
language.
THEREFORE:
[3C] ... in the case of a sentence such as:
(1) He thinks that John is wrong
it is the native speaker's knowledge of a systematic description of the linguistic
competence of a fluent native speaker of the language which tells him that he cannot
be interpreted as referring to the same person as John in (1).
It will be seen that ‘grammar’ begins life on page 2 in its theoretically correct style,
as a ‘model’ of the native speaker's ‘linguistic competence’. But notice that by page 3,
‘grammar’ is suddenly no longer a linguist's construct, a formal characterization of the
abilities presumed to underlie the speaker's behavior, but the knowledge itself. It has gone
from a linguist’s theory to something the speaker possesses. One would not blame Radford,
were it not that formal grammarians are quick to castigate discourse linguists for alleged
‘confusion’ over the notion of ‘grammar’, and often accuse them of not understanding this
supposedly elementary concept.
There is no question that ‘grammar’ is an infuriatingly elusive notion, and that it is
very easy to have a clear idea about what ‘grammar’ is in the sense of being able to give an
abstract definition of it, but quite another to apply that definition consistently in practice.
This asymmetry suggests that the notion of grammar is intrinsically unstable and
indeterminate, relative to the observer, to those involved in the speech situation, and to the
particular set of phenomena being focused upon. It suggests also that we need to question
the supposition of a mentally represented set of rules, and to set aside as well the idea in
Fromkin's statement which I quoted earlier, that speakers possess an abstract linguistic
system ready and waiting to be drawn upon — ‘accessed’! — in case they should ever need to
speak.
References
Abdullah ibn Abdul Qadir. 1932. Hikayat Abdullah. Singapore: Malaya Publishing House.
Malay Literature Series 4.
Abdullah ibn Abdul Qadir. 1928. Kesah Pelayaran Abdullah. Singapore: Malaya Publishing
House. Malay Literature Series 2.
Bakhtin, M. M. 1981. “Discourse in the novel.” In The Dialogic Imagination: Four Essays,
edited by Michael Holquist, translated by Carl Emerson and Michael Holquist. Austin:
University of Texas Press.

Becker, Alton L. 1979. ‘The figure a sentence makes: An interpretation of a Classical Malay
sentence.’ In Talmy Givon, ed., Discourse and Syntax, 243-260. Academic Press. (Syntax and
Semantics 12)
Bolinger, Dwight. 1976. ‘Meaning and memory’. Forum Linguisticum 1,1: 1-14.
Bosworth, Joseph, and T. Northgate Toller. 1898. An Anglo-Saxon Dictionary. Oxford:
Oxford University Press.
Clifford, James. 1986. ‘Introduction: Partial Truths.’ James Clifford and George Marcus,
eds., Writing Culture. The Poetics and Politics of Ethnography, 1-26. Berkeley: University of
California Press.
Collinge, N. E. 1985. The Laws of Indo-European. Amsterdam: John Benjamins BV. (Current
Issues in Linguistic Theory, vol. 35)
Culler, Jonathan. 1982. On Deconstruction: Theory and Criticism after Structuralism. Ithaca:
Cornell University Press.
Du Bois, John. 1986. ‘Competing motivations.’ In John Haiman, ed., Iconicity in Language.
Amsterdam: John Benjamins BV. (Typological Studies in Language)
Fromkin, Victoria. 1985. ‘Linguistics’, in: ‘Major Trends in Research: 22 Leading Scholars
Report on Their Fields.’ The Chronicle of Higher Education Sept. 4, 1985, pp. 12-13.
Hopper , Paul J., and Janice Martin. 1987. ‘Structuralism and Diachrony: the development of
the indefinite article in English.’ in Ramat, Anna Giacalone, Onofrio Carruba, and Guiliano
Bernini, eds. Papers from the 7th International Conference on Historical Linguistics.
Amsterdam: Benjamins.
Lambrecht, Knud. 1984. ‘Formulaicity, frame semantics, and pragmatics in German
binomial expressions.’ Language 60,4: 753-796.
Lambrecht, Knud. 1987. ‘On the status of SVO sentences in French discourse.’ In R. Tomlin,
ed., Coherence and Grounding in Discourse. Amsterdam: John Benjamins BV.
Morgan, Jerry. 1981. ‘Some observations on discourse and sentence grammar.’ Studies in the
Linguistic Sciences 11: 137-144.
Newmeyer, Frederick. 1983. Grammatical theory: its limits and possibilities. Chicago:
University of Chicago Press.
Plummer, Charles, ed. 1892. Two of the Saxon Chronicles, Parallel. ... Edited on the basis of
an edition by John Earle. Oxford: Clarendon Press.
Radford, Andrew. 1981. Transformational Syntax: A Student's Guide to Chomsky's Extended
Standard Theory. Cambridge University Press. (Cambridge Textbooks in Linguistics)
Sadock, Jerry. 1984. ‘Whither radical pragmatics?’ In D. Tannen, ed., Georgetown
University Round Table Conference on Linguistics, 139-149.
Smith, Barbara Hernstein. 1978. On the Margins of Discourse. University of Chicago Press.
Staten, Henry. 1984. Wittgenstein and Derrida. University of Nebraska Press.
Watkins, Calvert. 1962. Indo-European Origins of the Celtic Verb. Dublin: Institute of
Advanced Studies.
Wittgenstein, Ludwig. 1958. Philosophical Investigations. Trans. G. E. Anscombe. 3rd
edition. New York: Macmillan.
Footnotes

Page 17
1
I would like to thank Sandy Thompson for written comments on the conference version of
this paper. It goes without saying that numerous others have directly or indirectly
influenced the paper, but its general debt to Sandy and to Pete Becker, Ranjit Chatterjee,
Jack Du Bois, Barbara Fox, Talmy Givon, Knud Lambrecht, Larry Roberts, Deborah
Tannen, and especially Dwight Dolinger, is surely both pervasive and obvious; none of them
should be blamed for any errors or excesses in the use I have made of their ideas, and I
apologise to them if, because of space restrictions, I have not always cited their published
work when I might have.
2
This was pointed out to me independently by Catherine Lutz and Deborah Tannen.
3
Again, Derrida's proposal for a typology of ‘grafts’ seems closely relevant here.