LECTURETTE #2
ASSUMPTIONS OF THE STANDARD THEORY
What are the basic claims and assumptions of REST? Let's start with
those that it has in common with the Aspects model, Standard Theory sensu
stricto. These are summarized in (1).
(1) a. All syntactic relations are describable in terms of
Constituent Structure.
b. Constituents can be moved form one part of a structure
to another. As a result, one constituent structure
can be transformed into another. The generation of a
particular syntactic string involves a sequence of
levels distinguished from each other by such movement-
transformations. Such a sequence is called a 'deriva-
tion'.
c. The syntactic component of a grammar consists ulti-
mately of a set of well-formedness conditions on
constituent structures, on transformations, and on
derivations.
Now for some definitions and exposition of what these statements really
mean. The term 'Constituent Structure' is a brief label for a complex
concept that we will need to refer to over and over again, so we need
such a label for it. Constituent Structure is a way of organizing the
words and other constituents in a string that involves a conjunction of
two logically distinct relations, Dominance and Precedence.
Precedence is the easier of the two to explain in ordinary prose. It
involves linear ordering; two constituents are in a Precedence relation
to each other if one of them precedes the other in linear order.
Dominance is more complicated; it involves the notion of one constituent
being, in a pre-theoretical sense, contained within another (the verb
'contain' has a more precise meaning in PPA literature, to which i shall
return later but will ignore for now). Morphemes are contained within
words, words are contained within phrases, etc. Dominance relations are
typically represented by what are called 'tree diagrams', although these
diagrams are usually represented with their 'roots' at the top. A typi-
cal, but generic, tree diagram is given in (2).
(2) A
/ | \
/ | \
/ | \
B C D
/ \ |
/ \ |
E F G
In this diagram the letters A-G represent as *nodes*, and these nodes all
represent constituents of the string described by the diagram. A is the
*root*; B, C, and D are the constituents that make up A. We say that A
*immediately dominates* B, C, and D because there are no intervening
nodes, no nodes other than A that dominate any or all of the nodes B, C,
or D and are themselves dominated by A. B in turn immediately dominates
E and F, and C immediately dominates G. D doesn't dominate anything,
and is therefore a *terminal node*, as are E, F, and G in this diagram.
Note that it is possible for a node to dominate only a single other node,
as is the case with C and G; in this case C is said to be a *non-bran-
ching* node. Nodes like A and B which dominate more than one other node
are described as *branching*. It is also, a priori, possible for a node
to dominate more than 2 other nodes, as is the case with A. Richard
Kayne has argued in several of his works that, for the purposes of syn-
tactic theory, this option shouldn't be allowed. I am inclined to dis-
agree with him, mainly because my studies of 'free word-order' languages
such as Latin and Sanskrit have convinced me that it is difficult in many
cases to, e.g., distinguish between two different NP objects of a verb
such that one is consistently 'closer' to the verb in terms of constitu-
ent structure than the other. But Kayne's 'binary-branching' hypothesis
is often assumed as a default, and there is no question that it makes
for simpler trees. Whether it makes for a sufficiently accurate descrip-
tion of human linguistic behaviour is another question.
For historical reasons that are unclear to me, feminine terminology is
used in talking about nodes: two nodes one of which immediately dominates
the other are referred to as *mother* and *daughter* respectively, while
nodes which share a mother are *sisters*.
It is conventionally assumed that any given node may have at most one
mother. There are alternative frameworks of syntactic theory in which
this assumption does not hold, but it does in REST.
Note that both Dominance and Precedence are transitive relations: if A
dominates/precedes B and B dominates/precedes C, then A dominates/
precedes C. The relation of immediate dominance, of course, is not
transitive. There is an approximate parallel along the Precedence
dimension, namely Adjacency; two nodes are said to be adjacent if there
is no third node that precedes one but follows the other. The Adjacency
relation is important for certain details of the theory.
The other important thing to be noted about the Dominance and Precedence
relations is that they are mutually exclusive. Two nodes may be in
either a dominance or a precedence relation, but cannot be in both.
However, it is considered desirable, in a fully-specified constituent
structure, that Dominance and Precedence relations be maximized in their
domains: any two nodes must be in either a dominance or a precedence
relation. This is effected by claiming that, e.g. in (2), the nodes B
and G are in a precedence relation, specified by the fact that all the
daughters of B (i.e., E and F) precede G, and therefore B precedes G.
Note that this definition implies that the association-lines leading from
mother nodes to daughter nodes may not cross. This, too, is an assump-
tion typical of REST that is occasionally contested in some other frame-
works.
So much for the thumbnail sketch of Dominance and Precedence. POP back
to Constituent Structure. Constituent Structure is that which is exhaus-
tively describably in terms of Dominance and Precedence. If you can draw
a tree diagram of it, that tree diagram represents its constituent struc-
ture.
So far, everything i've been saying about Constituent Structure is rele-
vant to a wide range of syntactic-theoretical frameworks. The claim in
(1a) is that *all* syntactic relations are describable in terms of Con-
stituent Structure, i.e., in terms of Dominance and Precedence rela-
tions. This is a fundamental claim of Standard Theory which is not
shared by certain other frameworks. To give an example of what this
means, in the Standard Theory a 'direct object' is *defined* as an NP
immediately dominated by a VP node, while a 'subject' is defined as an NP
immediately dominated by an S node. When terms like 'subject' and 'ob-
ject' turn up in the Standard-Theoretical literature, they are merely
shorthand for the longer but (from the theoretical point of view) more
precise descriptions.
It would be an understatement to say that there is a lot of talk in the
Standard-Theoretical literature about 'movement'. I should point out as
early as possible (i.e., right now) that this is a metaphor. What is
critical is the assumption that the proper analysis of a syntactic string
may, and often does, involve not just one but several constituent struc-
tures; the set of all constituent structures involved is called the deri-
vation of the string. These constituent structures are all obviously
related to each other in that they share a common skeleton (in a little
while we shall get to discussing exactly to what extent they do) and,
more importantly, the same lexical material. But certain items occupy
different positions in different constituent structures of the derivation.
In the Aspects model, the derivation was defined as an sequence of con-
stituent-structures; it was assumed that a linear order could be imposed
on the 'levels', such that any level might be immediately 'preceded' by
at most one level and immediately 'followed' by at most one level (as we
shall see, in REST while the levels are still viewed as essentially or-
dered with respect to each other the ordering is not the simple, one-
dimensional ordering assumed in the Aspects model; and in some respects
the situation in the more recent 'Minimalist' work is even more complica-
ted). Such a linear ordering, especially when coupled with the fact
that different levels differed precisely in that identical items occu-
pied different positions, invited a view of the derivation as a process
progressing through time, with constituents moving from one position to
another as the derivation progressed from one level to the next. But
this view, though occasionally useful, is only a metaphor. It would be
at least equally correct to regard all levels as being generated simul-
taneously, in terms of real time. As long as each of them is well-formed
and they can all be arranged in an order that is also well-formed, each
level being related to the one preceding it in a legitimate fashion, the
derivation as a whole is grammatical. (Again, well-formedness conditions
in recent 'Minimalist' work are somewhat different, relating more to the
derivation as a whole and less to individual levels and transformations
within it, except insofar as these affect the Gestalt of the derivation,
which of course they do. To a great extent the difference is merely one
of perspective upon the theoretical objects rather than in the objects
themselves.)
Best,
Steven
---------------------
Dr. Steven Schaufele
712 West Washington
Urbana, IL 61801
217-344-8240
fcosws@prairienet.org
**** O syntagmata linguarum liberemini humanarum! ***
*** Nihil vestris privari nisi obicibus potestis! ***