Some Notes on 'Deep' Grammar
Author(s): M. A. K. Halliday
Source: Journal of Linguistics , Apr., 1966, Vol. 2, No. 1 (Apr., 1966), pp. 57-67
Published by: Cambridge University Press
Stable URL: https://www.jstor.org/stable/4174917
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access
to Journal of Linguistics
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
Some notes on 'deep' grammar
M. A. K. HALLIDAY
Department of General Linguistics, University College, London W.C. I
(Received 29 November I965)
In the representation of syntagmatic relations in language, we may distinguish
between a linear sequence of classes, such as 'adjective followed by noun', and
a non-linear configuration of functions, such as 'modifier-head relation' or
simply 'modification'.1 Both of these have been referred to as 'structure',
although this term has also been extended to cover paradigmatic as well as
syntagmatic relations. For Hjelmslev, for whom 'structure' was not a technical
term (see e.g. i96I: 74 (=I943: 67)), 'the structural approach to language ...
[is] conceived as a purely relational approach to the language pattern' (I948:
quoted in Firth, 195 I: 220); among others who have emphasized the relational
aspect of such studies are Firth2 (I957: I7 ff., I951: 227-8; cf. Robins, I953;
Palmer, i964a), Tesniere (cf. Robins, I96I: 8 i ff.) and Pike (cf. Longacre, I964
i6). Chomsky's (I964: 32) distinction, using Hockett's terms, between 'sur-
face structure' and 'deep structure', 'structure' here going beyond syntagmatic
relations, is extremely valuable and widely accepted: the surface structure of a
sentence is defined as 'a proper bracketing of the linear, temporally given se-
quence of elements, with the paired brackets labelled by category names',
while the deep structure, which is 'in general not identical with its surface
structure', is 'a much more abstract representation of grammatical relations
and syntactic organization'.
A representation involving the concepts of CLASS and SEQUENCE may thus be
said to be a representation of surface structure. Here the ordering, if each pair
of brackets is said to enclose an 'ordered set' of classes, is interpreted in the usual
sense of the word, as linear successivity, or sequence. Such an interpretation
does not preclude discontinuity or fusion of constituents, nor is it affected by
the depth of the bracketing imposed: both the more copious bracketing of IC-
type representations and the much sparser bracketing of, for example, a tag-
memic analysis can adequately specify the relation of sequence in a surface
structure. The labelling attached to the entities specified as entering into this
relation of sequence may then be 'class'-type labelling and interpreted as such,
'the class "adjective" ' being the set whose members are good, bad, . . .;
although functional labels have also been introduced: for example Nida (I960)
[i] This paper was first presented at a meeting of the Linguistics Association, Newcastle,
March I965. I am grateful to R. M. W. Dixon and R. D. Huddleston for their sub-
sequent valuable comments and suggestions.
[2] Haas (I966: 13I) points out that Firth's position here has been misinterpreted; this
may be partly due to his use of the term 'element of structure' as a functional term.
57
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
JOURNAL OF LINGUISTICS
states generalized syntagmatic relations, such as hypotaxis, within the frame-
work of an IC analysis.
If the representation of syntagmatic relations is merely in terms of this type
of surface structure, sequence is then the only determinable relation. A con-
siderable amount of bracketing may be introduced in order to give as much
information as is possible, within this limitation, about the syntactic relations
involved. Class labels do not by themselves reduce the bracketing required,
since classes do not fully specify syntactic function. Such labels may be con-
ventionally interpreted as functional, but if so their correct interpretation
depends on their association with a designated pair of brackets; for example
'adjective' is to be interpreted as 'modifier' when attached to a particular node
in the tree. This adds considerably to the syntactic information; but if the tree
itself represents sequence at the surface its application is limited (cf. Palmer,
I964a).
It has always been recognized that the concepts of class and sequence alone
are inadequate for the representation of syntagmatic relations in language.
Indeed the development of modern structuralism may be seen as having taken
place in the context of a tradition in which it was the more surface elements
that had remained least explicit. Relational terms like 'subject' and 'predicate'
have always co-existed with class names such as 'noun' and 'verb'; and the
definition of the classes has rested at least in part on syntactic criteria, so tha
the designation of an item by its class name indicates something of its poten-
tiality of syntactic function. Classes were not thought of as specifying actual
syntactic function within a given sentence, since the theory also recognized the
deeper syntactic relations into which the classes entered; the attempt to com-
bine morphological with syntactic criteria in the definition of the classes (since
morphological 'types' have to be accounted for somewhere), while it may lead
to difficulties, is entirely explicable within such a framework.
While the underlying syntagmatic relations have been recognized as non-
linear, or at least as not manifested in the linear sequence of the linguistic
items, their representation, as Palmer (i964a: I25) points out, has usually in-
volved some form of linear notation. Since there is also a level of abstraction at
which the relevant syntagmatic relation is one of sequence, it may be important
to recognize that two different kinds of representation are involved. In this
sense class and sequence are inherently surface concepts, specifying the items
of the language and their arrangement; this is no less true of syntactically
defined than of morphologically defined classes, the former being merely sets of
items identified as relevant to the deeper representation. For terminological
simplicity we might perhaps here follow one tradition in referring to an
arrangement of classes in sequence as a SYNTAGM, reserving the term STRUC-
TURE for a configuration of functions. If then function-type labels such as
'modifier' are introduced, whether as such or as conventional interpretations of
58
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
SOME NOTES ON 'DEEPs GRAMMAR
class-type labels, they will not be located in the syntagm, since their defining
environment is not stated in terms of (its) sequence. This holds true even if in
a given language (say) a modifier-head structure is always realized as a syntagm
of adjective followed by noun; a structure is not defined by its realizations.
The ordering that is ascribed to structure may be thought of in dependency
terms, or in constituency terms as an underlying sequence which does not
(necessarily) correspond to syntagmic sequence, or as mere co-occurrence or
absence of ordering.3 In all cases it is of a different nature from syntagmic
sequence, in that the components are functions, not sets of items. If (with
Lamb) we use * to represent configuration, this being interpreted as 'un-
ordered with respect to syntagmic sequence, whether or not any other form of
ordering is considered to be present', then a structural representation may
take the form m-h, or interchangeably h-m (modifier-head); this contrasts
with a syntagmic representation of the form adj^n (adjective followed by noun).
Representation such as m^h and adj-n would then appear as mixed types,
where deep (structural) labels are attached to surface (syntagmic) relations, or
vice versa. These might be given conventional interpretations, perhaps for
example m-h as 'modifier-head structure with realization by sequence alone'
(i.e. where modifier and head are realized by the same class), adj-n as 'modi-
fier-head structure with realization by class alone' (i.e. where the classes may
occur in either sequence); but these would be merely a shorthand for com-
bining two types of representation.
While many other formulations are possible, the recognition in some form
or other of two distinct types of representation, linked by some form of
'realization' relation,4 is relevant to the understanding of syntagmatic pattern
and the distinction can be made and discussed solely in terms of relations on
the syntagmatic axis. Clearly, however, it is relevant also to relations on the
paradigmatic axis. It may be helpful to relate this point to the distinctions
made by Hjelmslev and by Firth. In Hjelmslev's terms (I96I: 38-9=(I943:
3S-6)), linguistic function embraces both relation and correlation; relation
syntagmatic, within the semiotic process, or the text, while correlation is
paradigmatic, within the semiotic system, or the language. While his view of
the relation between the two axes was somewhat different from that of Hjelms-
lev, Firth likewise makes a terminological distinction, referring (I957: 17) to
syntagmatic relations as relations of structure and to paradigmatic relations as
relations of system.
[3] Firth, perhaps somewhat confusingly, reserved the term 'order' precisely for this non-
linear relation among the components ('elements') of a structure, contrasting it with
'sequence'.
[4] I use Lamb's term 'realization' instead of the earlier 'exponence'. Lamb's term is
more widely known; it also corresponds closely to my own use, whereas as Palmer
(i964b) pointed out my use of 'exponence' differed materially from that of Firth.
59
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Provided there is at least some syntactic element in the definition of the
classes concerned, even a syntagmic, class-sequence representation already
gives some information about the paradigmatic relations into which an item
enters, but to a very limited extent. This limitation is again inherent in the
nature of the class-sequence concept: paradigmatic 'relatedness' depends on a
functionally defined environment. Two entities can only be said to contrast if
they have a functional environment in common,5 and this environment is
generally specified in terms of syntagmatic function; it presupposes therefore
a representation of structure - that is, of 'deep' syntagmatic relations. The
structural representation thus specifies the environment both for sets of para-
digmatic relations and for further networks of syntagmatic relations, those
within the lower-order constituents; for example the function of 'subject',
itself specified syntagmatically in clause structure, defines an environment
both for the syntagmatic relation of modifier-head and for the paradigmatic
relation of singular/plural. For paradigmatic relations in the highest unit there
is no functionally defined environment in this sense, so that if the sentence is
said to be 'EITHER declarative OR interrogative' this either/or relation rests on
other grounds: the sentence as a primitive term, or some postulated higher
unit not yet structurally described; the appeal to contextually-defined sentence
functions such as 'statement' and 'question' is not one of these, this being
rather a way of saying that declarative and interrogative have no environment
in common.
The paradigmatic contrasts associated with a given, defined environment
may be thought of as being accounted for either in a single representation of
'deep' grammar, in which are incorporated both syntagmatic and paradigmatic
function, or in a separate form of statement, distinct from, but related via the
specification of the environment to, the statement of syntagmatic relations.
Firth's concept of the system embodies the second approach. The 'system'
may be glossed informally as a ' "deep" paradigm', a paradigm dependent on
functional environment; in a sense, and mutatis mutandis, the relation of system
to paradigm is analogous to that of structure to syntagm as these terms were
used above. One could think of 'paradigmeme' as a possible tagmemic term.
In Hjelmslev the 'system', likewise a paradigmatic concept, is defined as a
'correlational hierarchy', the underlying notion being that of commutation
(Hjelmslev I96I: 73-4=(I943: 66-7)). A system is thus a representation of
relations on the paradigmatic axis, a set of features contrastive in a given
environment. Function in the system is defined by the total configuration, for
[5] This may be interpreted either as 'if there is at least one set of conditions under which
both could occur' or as 'if both could occur under the given set of conditions'. It is the
latter interpretation which I take to be the basis of (one aspect of) Firth's 'polysystemic'
approach. Firth himself was inconsistent in referring to a 'system' of word classes noun,
verb, etc.
6o
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
SOME NOTES ON 'DEEPs GRAMMAR
example 'past' by reference to 'present' and 'future' in a three-term tense
system, as structural function is defined by reference to the total structural
configuration, for example 'modifier' by reference to 'head'.
If paradigmatic relations are represented separately in this way, this implies
that the full grammatical description of a linguistic item should contain both
a structural and a systemic component. It may be useful therefore to consider
the notion of a 'systemic description' as one form of representation of a linguis-
tic item, the assumption being that it complements but does not replace its
structural description. The systemic description would be a representation of
the item in terms of a set of features, each feature being in contrast with a
stated set of one or more other features: being, in Firth's terms, a 'term in a
system'. This is exactly the sort of characterization that has been familiar for
a long time in the form of 'this clause is interrogative, finite, present tense,
... ', given that we are told somewhere in the grammar not merely what other
tenses, moods, etc. are found in the language but also which of them could
have occurred in this particular clause all other features being kept con-
stant.
There is, however, one modification of a traditional 'systemic description'
of this kind which may need to be considered. This concerns the ordering o
the features listed. In the traditional version they are unordered; but if the
grammar specifies not only relevant systems but also their interrelations with
one another, in particular their hierarchization on what I have called elsewhere
(I96I: 272, I964: i8) the 'scale of delicacy', then partial ordering is introduced.
Any pair of systems, such that a feature in one may co-occur with a feature in
the other in a systemic description, may be hierarchical or simultaneous; if two
systems are hierarchically ordered, features assigned to these systems are
ordered likewise. So for example the system whose terms are declarative/
interrogative would be hierarchically ordered with respect to the system in-
dicative/imperative, in that selection of either of the features declarative and
interrogative implies selection of indicative. If this is taken together with
another system, unpredicated theme/predicated theme, likewise dependent on
indicative, then the item 7ohn has seen the play may be represented in respect
of these features as:-
(indicative : (declarative/unpredicated theme))
where : indicates hierarchy and / simultaneity. Then It's John who has seen
the play contrasts with it in respect of one feature:
(indicative : (declarative/predicated theme) )
and is it John who has seen the play in respect of two features:-
(indicative : (interrogative/predicated theme) ).
The systemic description would represent a selection from among the possi-
bilities recognized by the grammar. As far as these examples are concerned,
the grammar would show that in a given environment selection is made between
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
JOURNAL OF LINGUISTICS
indicative and imperative; and if indicative is selected, there is also simultane-
ous selection between declarative and interrogative and between unpredicated
theme and predicated theme, the two latter selections being independent of
one another. Any item thus contrasts with others in respect of such features
and combinations of features as the ordering of the systems permits.
For any set of systems associated with a given environment it is possible to
construct a system network in which each system, other than those simultane-
ous at the point of origin, is hierarchically ordered with respect to at least one
other system. The point of origin is specified syntagmatically, so that all
features are associated with a syntagmatic environment; at the same time the
system network provides a paradigmatic environment for each one of the
features, specifying both its contrastive status and its possibilities of combina-
tion.
It is not the aim here to present in detail the properties of a systemic des-
cription, but rather to discuss it in general terms. Systemic description may
be thought of as complementary to structural description, the one concerned
with paradigmatic and the other with syntagmatic relations. On the other hand
it might be useful to consider some possible consequences of regarding sys-
temic description as the underlying form of representation, if it turned out that
the structural description could be shown to be derivable from it. In that case
structure would be fully predictable, and the form of a structural representa-
tion could be considered in the light of this. It goes without saying that the
concept of an explicit grammar implied by this formulation derives primarily
from the work of Chomsky, and that steps taken in this direction on the basis
of any grammatical notions are made possible by his fundamental contribution.
My own more specific debt here is to Lamb, whose formalization of stratifica-
tion theory is based on general notions closely akin to those which I had
adopted (cf. Hockett, 1965: 198). The present paper, however, attempts no
more than an informal discussion of the question of a grammatical description
in terms of features, here based on the notion of a feature as one of a set of
contrastive 'terms in system'.
Presenting the systemic description of a linguistic item as the underlying
grammatical representation of that item would seem to imply that its paradig-
matic relation to other items of the language was in some way its more funda-
mental property, from which its internal (syntagmatic) structure is considered
to be derived. This would seem to be Hjelmslev's view, in his discussion of
system and process (I96I: 39-40= 1943: 36-7)). (But the priority which is im-
plied is not one between paradigmatic and syntagmatic relations as such, but
rather between the external relations, both paradigmatic and syntagmatic, into
which an item enters (the point of origin of a system network being defined in
syntagmatic function) and its internal relations of structure. If one talks of
simplicity, this means the simplicity of the whole description; underlying
62
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
SOME NOTES ON 'DEEPs GRAMMAR
grammar is 'semantically significant' grammar, whether the semantics is
regarded, with Lamb, as 'input' or, with Chomsky, as interpretation. What is
being considered therefore is that that part of the grammar which is as it were
'closest to' the semantics may be represented in terms of systemic features.
This would provide a paradigmatic environment for the 'relatedness' of lin-
guistic items, a contrast being seen as operating in the environment of other
contrasts. Structure would then appear as the realization of complexes of
systemic features, involving in places both neutralization and diversification as
defined in Lamb's terms (Lamb, I964a: 64).
If the structural representation is not required to account for paradigmatic
relations, the question of how 'deep' it needs to be is determinable by reference
to other considerations: it should give an adequate account of syntagmatic
relations, and permit the explicit realization of the systemic description in
terms, ultimately, of a sequence of classes. This may be illustrated from the
example It's John who has seen the play. Leaving aside variation that is im-
material to the discussion, there would seem to be three possible representa-
tions of its structure:
(i) it subject, is predicator, J'ohn who has seen the play complement;
(2) it ... who has seen the play subject, is predicator, John complement;
(3) it's John who subject, has seen predicator, the play complement.
(i) would presumably be an attempt merely to state the simplest sequence of
classes in the syntagm, although it could be shown to be unsatisfactory even
on class-distributional grounds. (2) is distributionally acceptable and would
account adequately for the syntagmatic relations; but it fails to account for the
paradigmatic relations in that it does not show the 'relatedness' of this clause to
John has seen the play, etc. If the structural description is required to show the
paradigmatic as well as the syntagmatic relations of the grammar we need
some representation such as (3) in which John is the subject. This leads to
complexity in the realization, since a nominalization of the form it's John who
seems to add no new insight elsewhere in the grammar. A more serious diffi-
culty arises in relation to the element 'subject' in English, which is a complex
element within which it is possible to distinguish three components, or
features; each of these may contrast independently of the other two, although
there is a general, and generalizable, tendency to co-variation among them.
The three contrasts can be seen independently in (i) John has seen the play,
with tonic on play, versus, respectively, (ii) The play has been seen by John
(subject as actor versus subject as goal); (iii) The play 7ohn has seen (= 'the
play, John has seen, but . . . ', subject as theme versus subject nonthematic);
(iv) 7ohn has seen the play (with John tonic; subject as 'given' versus subject as
'new'). Each of these three is related paradigmatically to the original item, and
each of them contrasts with it in respect of one feature only. By a further con-
trast, that of 'unpredicated theme' versus 'predicated theme', (iv) is related to
63
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
JOURNAL OF LINGUISTICS
(v) It's John who has seen the play with John tonic.6 T
in constituent structure, (v) differs from (iv) in respect of only one feature.
Such patterns, where different complexes of (paradigmatic) features may be
combined in what is syntagmatically one and the same element of structure,
here the 'subject', involve some complexity for a structural description; if they
were handled in systemic terms, the structure need represent only their real-
ization in syntagmatic relations. We could then adopt a form of structural
representation such as (2) above.
The examples cited might be regarded as irrelevant on the grounds that they
do not involve cognitive distinctions and therefore belong to the realm of
stylistic variation. But this is to assume that it is the task of a grammatical
theory to differentiate between these different types of distinction. Such prob-
lems seem to me to fall more properly within the domain of a semantic theory,
where the selection of a particular variable, such as paraphrase, as a basis for
the classification of distinctions is not arbitrary as it seems to be in the gram-
mar. This is not to deny that the speakers of a language recognize some dis-
tinctions as 'more important' than others, and that this may depend at least in
part on a concept of paraphrase. The hierarchization of systems in delicacy, in
a system network, does seem to reflect some notion of the relative importance
of the systems involved; this is an instance of the convergence of semantic and
distributional criteria referred to by Lyons in his important discussion of sem-
antics and grammar (I963: chapter 2). Even if a clear answer can always be
given to the question 'is a a paraphrase of b or not?', or to other questions
where this is irrelevant (e.g. in the distinction between John has seen the play
with Yohn tonic and with play tonic, which answer different questions), the
place of a given distinction in the grammar would, as I see it, depend on its
environment in terms of other distinctions, this presupposing also its syntag-
matic environment, rather than on a classification of its semantic function.
To return to the discussion, another relevant factor here would be the desire
to incorporate into the grammar phonological realizations of grammatical
features, particularly (in English) those of intonation and rhythm. Such
features may be assigned a place in a syntagmatic representation, either as
superfixes in a syntagm or as elements in a structure; but the assignment of,
say, a pitch contour as a constituent to a specified place in a structural rep-
resentation, while it may be necessitated by the realization requirements,
seems in other respects rather arbitrary. Intonation, in English, provides
instances of both neutralization and diversification; one and the same feature
[6] Since the subject normally has the feature 'given', that of 'new' being realized in other
elements, the realization of the feature 'new' in the subject is often accompanied by its
predication as in It's Yohn who has seen the play. This explains why It's John who has
seen the play usually, though not obligatorily, has John and not play tonic, while the
opposite is true of John has seen the play.
64
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
SOME NOTES ON 'DEEP GRAMMAR
may be realized in some environments by a structural pattern and in others by
intonation, and a given intonation pattern may realize different features in
different environments. In other words, intonation is not predictable from its
structural environment. It can, however, be shown to be predictable in the
grammar if it is regarded as a form of the realization of systemic features, at
the same degree of abstraction (same stage of representation) as the struc-
tural elements but without constituent status.
Intonation, however, is merely a special, if clear, case of a more general
point, namely that if a representation in terms other than of constituent struc-
ture is adopted for the statement of paradigmatic relations, and is then made
to determine the constituent structure, then provided the structural descrip-
tion adequately handles the syntagmatic relations there is no need for every-
thing to be accounted for at a constituent stage of representation. This is most
obviously relevant to phonological features of the prosodic type, but could be
extended also to items identified as being markers of, rather than elements in,
syntactic relations.
The crucial factor in the designation of any feature as present in the gram-
mar would thus be its assignment to a place in the systemic network. A puta-
tive feature which could not be shown to contrast independently with one or
more others at some point would not be a distinct feature; each feature that is
recognized is thus a term in a system, which system is located in hierarchial
and simultaneous relation to other systems. The location is 'polysystemic': the
recognition of a system, and the assignment of a feature to it, depends on the
potentiality of contrast in the stated environment. For example, there might
seem to be a proportionality in English such that he can go is to can he go as
He is wondering if he can go is to He is wondering can he go: but the related
features are different in the two cases: that is, the two syntagmatic environ-
ments determine different sets of paradigmatic relations. The ordering of the
systems in delicacy would thus be important in the identification of the sys-
temic features.
It would be necessary also to specify the syntagmatic environment, in order
to define the point of origin of a system network. This can be done in terms of
the notion of rank, where the initial identification and labelling of certain
stages in a constituent hierarchy in such general terms would provide a starting
point for the delimitation of different more specific environments. The desig-
nation of rank, in other words, is a possible first siep in the specification of
what Haas (I966: 125) calls 'functional relations', relevant here in that it makes
possible the assignment of a system to a place determined solely by constituent
status (e.g. all clauses) and allows further specification of the environment to
be in terms of features: a feature x may be associated with constituents having
the features y and z rather than with constituents having a given syntagmatic
function. The possibility of contrast between active and passive in the
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
JOURNAL OF LINGUISTICS
clause depends on other features of the clause, not on its function in the
sentence.
In stratificational terms, rank defines an inner series of strata, or sub-strata,
within the outer grammatical stratum, with each rank characterized by a
different network of systems. While many, though by no means all, features
would be present at more than one rank, with consequent pre-selection at
certain points, an important distinction is to be made here between pre-selection,
where the choice of a feature at one rank determines the choice of a feature at a
lower rank but the two operate in different paradigmatic environments, and the
realization of a feature at a lower rank than that in which it has its environment.
The latter includes such familiar instances as the realization in the structure of the
word of a choice, such as that of number, associated with the group. There seems
no reason to assume a necessary relationship between the rank at which a fea-
ture has its environment and that in whose constituent structure it is realized.
The relevance of the concept of rank in this connection would thus be that
it is as it were neutral between system and structure. While clearly a con-
stituency notion, reflecting here the speaker's awareness of the hierarchic
organization of linguistic items, it imposes a minimum of bracketing and in
this way facilitates the interrelating of paradigmatic and syntagmatic modes
of representation. The discussion of 'systemic description' here has been in
terms of a rank-type constituent structure, since this would be one way of
defining a point of origin for a system network: each system, like each structure,
would be assigned to a given rank as its most generalized functional environ-
ment. It is not implied that a description in terms of features would necessitate
a rank-type constituent structure, but rather that the status of constituents in
the grammar would need to be brought into the discussion.
Palmer (I964: I30) wrote: 'Perhaps we need a pre-grammatical statement
in which order is utterly divorced from sequence'. In this paper I am following
up Palmer's conclusion by asking whether such a statement could be thought
of as a representation of the 'deep' grammar. If deep grammar is equated with
deep structure, in the sense of being thought of as relations of the constituency
type, it may be difficult to avoid connotations of sequence and to solve some
of the problems Palmer raised. If the underlying 'order' is thought of as
systemic, the more abstract representation of grammatical relations carries no
implication of sequence. Sequence can be stated by reference to these, a
'linguistic element whose exponent is sequence' having a status no different
from that of others. Such a description is in a sense of the WP type, with word
replaced by unit, or constituent, and paradigm by system. It is not suggested
that paradigmatic relations are somehow 'more important' than syntagmatic
ones; but merely that a description in terms of features, if it can be made
explicit, may help in bringing the 'unidimensional time sequence' of language
into relation with its deeper patterns of organization.
66
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
SOME NOTES ON DEEP GRAMMAR
REFERENCES
Chomsky, N. (I964). Topics in the Theory of Generative Grammar. (To appear in T. A.
Sebeok (ed.), Current Trends in Linguistics, 3. The Hague: Mouton.) (Preliminary ver-
sion, mimeographed.)
Firth, J. R. (I95i). General linguistics and descriptive grammar. TPhS 69-87. (Reprinted
in Papers in Linguistics I934-I951. London: Oxford U.P., I957. 2i6-228.)
Firth, J. R. (I 957). A synopsis of linguistic theory. Studies in Linguistic Analysis. (Special
volume of the Philological Society.) Oxford: Blackwell. I-32.
Haas, W. (I966). Linguistic relevance. In C. E. Bazell, J. C. Catford, M. A. K. Halliday &
R. H. Robins (eds.) In Memory of J3. R. Firth. London: Longmans. Il6-I47.
Halliday, M. A. K. (I96I). Categories of the theory of grammar. Word 17. 241-292.
Halliday, M. A. K. (i 964). Syntax and the consumer. In C. I. J. M. Stuart (ed.) Report of the
Fifteenth Annual (First International) Round Table Meeting on Linguistics and Language
Studies. (Monograph Series on Languages and Linguistics, 17.) Washington, D.C.:
Georgetown U.P. I I-24.
Hjelmslev, L. (I947). Structural analysis of language. SL x. 69-78.
Hjelmslev, L. (I96I). Prolegomena to a Theory of Language. (Translated by F. J. Whitfield.
Revised English edition.) Madison: Wisconsin U.P. (Originally published (1943) under
the title Omkring Sprogteoriens Grundlaeggelse. Copenhagen: Munksgaard.)
Hockett, C. F. (I965). Sound change. Lg 41. I85-204.
Huddleston, R. D. (i965). Rank and depth. Lg 41.
Lamb, S. M. (I964a). The sememic approach to structural semantics. AmA 66. 57-78.
Lamb, S. M. (I964b). On alternation, transformation, realization and stratification.
In C. I. J. M. Stuart (ed.) Report of the Fifteenth Annual (First International) Round
Table Meeting on Linguistics Language Studies. (Monograph Series on Languages and
Linguistics, 17.) Washington, D.C.: Georgetown U.P. I05-I22.
Longacre, R. E. (I964). Grammar Discovery Procedures. (Janua Linguarum, Series Minor
33). The Hague: Mouton.
Lyons, J. (I963). Structural Semantics. (Philological Society Publications, 20.) Oxford:
Blackwell.
Nida, E. (I960). A Synopsis of English Syntax. (Linguistic Series, 4.) Norman: Summer
Institute of Linguistics.
Palmer, F. R. (1964a). 'Sequence' and 'order'. In C. I. J. M. Stuart (ed.) Report of the
Fifteenth Annual (First International) Round Table Meeting on Linguistics and Language
Studies. (Monograph Series on Languages and Linguistics, 17.) Washington, D.C.:
Georgetown U.P. I23-I30.
Palmer, F. R. (1964b). Grammatical categories and their phonetic exponents. In H. G.
Lunt (ed.) Proceediings of the Ninth International Congress of Linguists. The Hague:
Mouton. 338-344.
Robins, R. H. (953). Formal divisions in Sundanese. TPhS I09-I42.
Robins, R. H. (I96I). Syntactic analysis (review article). ArchL 13. 78-89.
67
This content downloaded from
3.128.143.42 on Wed, 24 Apr 2024 06:25:24 +00:00
All use subject to https://about.jstor.org/terms
