Title: Discourse functions of the active-passive dichotomy in English
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00099357/00001
 Material Information
Title: Discourse functions of the active-passive dichotomy in English
Physical Description: viii, 177 leaves : ill. ; 28 cm.
Language: English
Creator: Coleman, Douglas Wells, 1953-
Copyright Date: 1982
Subject: English language -- Voice   ( lcsh )
Genre: bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by Douglas Wells Coleman.
Thesis: Thesis (Ph. D.)--University of Florida, 1982.
Bibliography: Includes bibliographical references (leaves 170-176).
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00099357
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000317384
notis - ABU4205
oclc - 08800893


This item has the following downloads:

discoursefunctio00cole ( PDF )

Full Text







Copyright 1982


Douglas Wells Coleman

To Lucy


I would like to thank the members of my supervisory committee

for their guidance and criticisms. They are Dr. William J. Sullivan,

Dr. Jean Casagrande, Dr. Chauncey C. Chu, and Dr. Ralph Selfridge.

Others who deserve thanks for their comments, questions, and

suggestions include Dr. Ira Fischler (UF), and Dr. John Albertini

and Dr. Gerald Berent (National Technical Institute for the Deaf).

For volunteering their classes for participation in the psycho-

linguistic experiment in Chapter 9, I wish to thank Abdu Benhallam,

Dr. Haig Der-Houssikian, Pat Sivinski, Margaret MacDonald, Charles

Hall, Dr. Roger Thompson, Dr. Robert Hamnond, Zayra Camacho, Dr. Donald

Dew, Dr. Howard Rothman, Lillian Huang (University of Florida), Larry

Arthur, Sam Holcomb, Barbara Ray Holcomb (National Technical Institute

for the Deaf).

- iv -








Notes 16


Notes 28


Notes 38


Case Grammar 41

Relational Grammar 50

Lexicalist Theory 54

Notes 61


Notes 93



Notes 108



Notes 114


Notes 124


Instrument 129

Subjects 130

Procedure 130

Results and Discussion 132

Notes 137


Notes 143


Notes 144







- vi -

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



Douglas Wells Coleman

May 1982

Chairman: William J. Sullivan
Major: Linguistics Department

It is hypothesized that the primary function of the active-passive

dichotomy in English is to control thematization.

In PART I, below, various linguistic-theoretical formalizations of

the active-passive relationship are explored. Included are the

'mainstream' transformational generative frameworks from Syntactic

Structures to the present, three 'alternative' transformational frame-

works (case grammar, relational grammar, and lexicalist theory), and

stratificational grammar.

In PART II, support for the above hypothesis is gained from

studies of various passive types: be vs. get passives, 'full' vs.

agentless passives, and 'true' vs. 'pseudo-' passives. Further support

is gained from an examination of actives and passives in different

environments, especially in embedded clauses. The psycholinguistic

literature provides valuable insights and evidence relevant to the

discourse function of the active-passive dichotomy. As part of the

current study, a psycholinguistic experiment is performed which further

supports the hypothesis.
vii -

A linguistic-theoretical framework must fulfill certain require-

ments in order to adequately describe the above discourse function

of the active-passive dichotomy. It must, for example, permit a unified

treatment of different passive types in English. Other requirements

are discussed as well. Of those considered, the stratificational

framework seems to be the only one capable of meeting all the require-

ments described.

- viii -


Previously, in virtually all studies of passive voice (and in

virtually all 'mainstream' linguistic studies, for that matter) the

assumption was for some time that--in one way or another--the

sentence is the 'basic' unit of language. Consideration was

primarily of the grammaticality of sentences (e.g. Jacobs and

Rosenbaum 1968:275), the structure of sentences, and so on.1 Many

inferred from this that the sentence is the largest unit of language

which should be of concern to the linguist.

This has provided a relatively narrow context within which to

develop an analysis of the interpretation of language structures.

It is, in fact, too narrow a context, as some have pointed out (e.g.

Rommetveit 1974; Longacre 1976; Clippinger 1977).

As an example of this rather narrow view, consider the way the

function of the English passive voice construction has 'tradi-

tionally' been interpreted. Most grammarians, and most linguists,

have interpreted this construction as fulfilling the function of

'emphasizing' some noun phrase by making it the 'subject of the

discourse'.2 See, for example, the discussion in Lyons (1968),

section 8.1.2. This view is based on sound intuitions, but looks

no farther than the bounds of a single sentence.

And, it is a highly oversimplified view of the function of

passive voice. It will be seen that passive functions in more

- 1 -


than one way to control thematization; further, these functions

depend upon discourse contexts.

Chapters 1 5 present, without much comment (most of which is

withheld until later chapters), the development of several formaliza-

tions of the active-passive relationship. These chapters comprise


PART II is an extended discussion of the function of the active-

passive dichotomy. 'Truncated' (i.e. agentless) passives are dis-

cussed first, in Chapter 6, but more space is devoted to the con-

sideration of 'full' passives (those which include an agent or an

agent-like element in a by phrase), i.e. Chapters 6 10.

The final chapter unites PART I and II in a discussion of the

characteristics required in an adequate formalization of the active-

passive relationship.


Such phrases seem to include the assumption that 'the gram-
maticality of sentences' is equivalent to 'the grammaticality of
utterances', the 'structure of sentences' equivalent to 'the struc-
ture of utterances', and so on.

2While the noun phrase is often a patient, it is not always;
it is always an NP for which realization as the subject of the
sentence is highly marked, however.


The average Ph.D. thesis is nothing but a
transference of bones from one graveyard
to another.

J. Frank Dobie (1980)
A Texan in England


In this chapter, I have attempted to trace the major developments

in 'mainstream' transformationalist formalizations of the active-pas-

sive relationship up to and including that of Chomsky's Aspects of the

Theory of Syntax (1965). This and the immediately following chapters,

which present further developments in formalizations of the relation-

ship between active and passive sentences, are to serve primarily as

background. Evaluation of the ability of particular formalizations to

meet our descriptive needs in terms of discourse functions of the

active-passive dichotomy is left to still later chapters.

Until the work of Chomsky (1957), no attempt was made to

formalize the relationship between an active sentence like (1.1) and

its passive counterpart (1.2).

(1.1) John hit Fred.

(1.2) Fred was hit by John.

Bloomfield (1933), for example, characterized sentences as being

of particular 'sentence types'. The two types relevant to this dis-

cussion are the 'actor-action' and the 'goal-action' sentence types.

Bloomfield (1933:173) gave the Tagalog sentence (1.3) as an example of

the former, and (1.4) as an example of the latter.

-4 -


(1.3) /sja j pu'mu:tul nan 'ka:huj/

'he cut some wood' (actor-action)

(1.4) /pi'nu:tul nja ao 'ka:huj/

was-cut by-him the wood' (goal-action)

Sentence (1.1), above, is thus of the 'actor-action' sentence

type, while (1.2) is of the 'goal-action' type. Use of Bloomfield's

'immediate constituent' analysis to produce tree diagrams of (1.1) and

(1.2) would yield (1.5) and (1.6) respectively.

(1.5) John hit Fred.


John / y

hit Fred

(1.6) Fred was hit by John.


Fred/ b

was hit by John

Bloomfield (1933) did not explicitly and formally relate

sentences like (1.1)--which could be analyzed into immediate

constituents as in (1.5)--with sentences like (1.2)--which could be

analyzed into immediate constituents as in (1.6).

In all fairness, Bloomfield did say (1933:267) that forms such as

Fred in (1.1) and (1.2) 'occur in...a position...with a positional

meaning'. In a sentence like (1.1), the 'positional' or 'class'

meaning of John would be 'performer of an action' or 'actor'. Pre-

sumably, the class meaning of John would be the same in (1.2). He


also described the class meaning of a form like Fred in (1.2) as

'undergoer of an action'.

The notions 'performer of an action' and 'undergoer of an action'

are analogous to those currently referred to by the terms 'agent' and

'patient'. Bloomfield, however, denied the relevance of class-

meanings when he said (1933:267-268) that

class meanings are not clearly-definable units which could
serve as a basis for our work, but only situational
features, undefinable in terms of our science.... Form
classes, like other linguistic phenomena, can be defined,
not in terms of meaning, but only in terms of linguistic
(that is, lexical or grammatical) forms.

This self-imposed restriction, which arose from his behaviorist

slant, prevented Bloomfield from attempting to formalize any relation-

ship between active sentences and their passive counterparts.

More generally, this same restriction implicitly defined the kind

of grammar Bloomfield used to describe natural language. This

grammar, though not yet formalized as such, was a limited phrase

structure grammar which Chomsky later (in Syntactic Structures) showed

to be inadequate.

Lyons (1968:210 ff.) put this aspect of Chomsky's work into its

proper perspective. In his Introduction to Theoretical Linguistics,

he helped elucidate the direct link between the kind of immediate

constituent analysis done by Bloomfield and its extension and further

formalization by Chomsky. The continuity of thought is greater than

many might suspect, since this point has often been missed. In his

discussion of phrase structure, Chomsky did not explicitly state what

links his work had with the older literature. It appears that, as a


result, many readers have missed the real point of his discussion of

phrase structure grammars in particular.

Chomsky was able to relate active and passive sentences, but not

because he rejected Bloomfield's arguments against the relevance of

class meanings. In general, Chomsky did not reject these arguments.

His own assumption was that 'the semantic component of a generative

grammar, like the phonological component, is purely interpretive' and

that 'all information utilized in semantic interpretation must be

presented in the syntactic component'. In this, he was suggesting

that semantics should not be relegated to 'unanalyzed semantic intui-

tion' (Chomsky 1965:75).

In Syntactic Structures, Chomsky (1957:42 ff.) introduced a

transformational rule for passivization as an example to demonstrate

the inadequacy of a phrase structure grammar alone for the description

of a natural language. This claim of inadequacy was based not on the

desire to include class meanings into consideration, but on a desire

for greater simplicity in the grammar. The simplicity gained was the

elimination of some major redundancies in the statement of

cooccurrence restrictions. It was gained at the expense of the intro-

duction of a transformational component.

Chomsky noted that many cooccurrence restrictions would have to

be placed on subject, verb, and object to include (1.7) (1.10), yet

exclude (1.11) (1.14).

(1.7) John admires sincerity.

(1.8) Sincerity frightens John.

(1.9) John plays golf.


(1.10) John drinks wine.

(1.11) *Sincerity admires John.

(1.12) *John frightens sincerity.

(1.13) *Golf plays John.

(1.14) *Wine drinks John.

Such cooccurrence restrictions present no problem, until passive

sentences are taken into consideration. In passives, cooccurrence

restrictions on subjects are like those on objects in their active

counterparts. Similarly, cooccurrence restrictions on some noun

phrases in by phrases in passives are like those on subjects in the

active counterparts. See (1.15) (1.22). This leads to what Chomsky

(1957:43) referred to as an 'inelegant duplication'.

(1.15) Sincerity is admired by John.

(1.16) John is frightened by sincerity.

(1.17) Golf is played by John.

(1.18) Wine is drunk by John.

(1.19) *John is admired by sincerity.

(1.20) *Sincerity is frightened by John.

(1.21) *John is played by golf.

(1.22) *John is drunk by wine.

What is preferred, is a simpler description of the same

linguistic data in a way that allows greater generalizations to be

made. Chomsky gave us this by the exclusion of passives from deep

structure. Instead, passives were described as introduced via an

optional rule of the form (Chomsky 1957:112):





Some argued, however, that active and passive should not be

derived from the same deep structure, since the two are not synony-

mous. A particularly eloquent argument for the nonsynonymy of active

and passive is presented in Ziff (1966).

In addition, it had by the time of Aspects (1965) become clear

that all of Chomsky's 'optional' singulary transformations (e.g.

passivization) had to be reformulated as obligatory, to be 'triggered'

by the presence of a marker in the deep structure which was generated

by the phrase structure rules for this purpose. Lees (1960), for

example, had shown that this was necessarily so for the negation

transformation. Katz and Postal (1964) outlined the more general

principle that, in Chomsky's words (1965:132), 'the only contribution

of transformations to semantic interpretation is that they interrelate

Phrase-markers'. For an extended discussion, see Chomsky (1965:132


This is the point at which it was decided that transformations

should not alter meaning.

But, since passive sentences were to obligatorily result from the

presence of a passive 'trigger' generated by the phrase structure

component, they were differentiated from active sentences at a deep

structure level. The new passive transformation was thus to be 'mean-

ing-preserving' and would not conflict with Katz and Postal's above-

mentioned principle.

- 10 -

Chomsky (1965) discussed Katz and Postal's principle specifically

in regard to passive (see Chapter 2 of Aspects). Actually, it was

Hockett (1961) who had first suggested that passivization depend on a

'trigger' in the deep structure. Hockett's suggestion preceded the

proposal of the Katz-Postal Hypothesis. Chomsky, therefore, repeated

Hockett's suggestion merely in a footnote, saying that Hockett gave

'no supporting argument' and that the suggestion was 'no more than a

notational innovation' (1965:223).

It having been established by transformationalists that there

should be a passive 'trigger' generated by the phrase structure com-

ponent, the next question was, 'where in the deep structure should the

passive "trigger" be specified?'

Lees (1960:8), as noted by Chomsky (1965:103), observed that

verbs which do not take manner adverbials freely also 'do not undergo

the passive transformation'. Chomsky compared verbs like resemble,

cost, marry, and weigh. Sentences with the former two do not have

passive counterparts. See (1.23) and (1.24). The latter two verbs

(marry and weigh) have no passive counterparts when used in structures

which do not allow manner adverbials freely, e.g. (1.25) and (1.26),

but can 'undergo passivization' when used in structures that freely

take manner adverbials, e.g. hastily and carefully in (1.27) and


(1.23) (a) John resembles Alan.

(b) *Alan is resembled by John.

(1.24) (a) This book costs ten dollars.

(b) *Ten dollars is cost by this book.

- 11 -

(1.25) (a) John married Mary.

(b) *Mary was married by John.

(1.26) (a) This car weighs two tons.

(b) *Two tons is weighed by this car.

(1.27) (a) The preacher married John and Mary hastily.

(b) John and Mary were married hastily (by the
(1.28) (a) John weighed the letter carefully.

(b) The letter was carefully weighed (by John).

Chomsky deduced from this data that since the ability of a verb

to undergo passivization depended on the verb's ability to freely take

manner adverbials, the phrase structure component should enforce this

cooccurrence restriction. His suggestion (1965:103-104) was to write

the phrase structure rule for manner adverbials so as to include the

dummy element passive (1.29). The passive transformation was to

obligatorily apply when the structual description (1.30) was met.2

(1.29) Manner by passive

(1.30) NP-Aux-V-...-NP-...-by passive-...

The placement of the by passive dummy element in the manner adver-

bial was also intended to account for so-called 'pseudopassives' such

as (1.31).

(1.31) The boat was decided on by John.

The location of the NP to become the surface subject after pas-

sivization was to the left of the by passive 'trigger'. This was a

critical feature of the structural description. By a slight modification

- 12 -

of the passivization rule, this NP was permitted to be contained

inside a prepositional phrase, in turn permitting the generation of

such 'pseudopassives'. However, the structural description and the

rule had to exclude the possibility of the application of passiviza-

tion to NP's in certain other types of prepositional phrases.

Sentence (1.31) is an acceptable passive counterpart to (1.32) if the

latter is a paraphrase of (1.33)(a), but not if it is a paraphrase of


(1.32) John decided on the boat.

(1.33) (a) John chose the boat.

(b) John decided while on the boat.

Chomsky's phrase structure rules placed a 'prepositional phrase'

like on the boat in (1.32)--paraphrasable as (1.33)(a)--before the

manner adverbial. This kind of 'prepositional phrase' was involved

with verb subcategorization rules, he argued, and should be placed

separately from prepositional phrases which were not. See (1.36).

(1.34) S NP ^Predicate-Phrase

(1.35) Predicate-Phrase Aux VP (Place) (Time)


be Predicate

(NP) (Prep-Phrase) (Prep-Phrase)^
VP-r V Adj
(like) Predicate-Nominal

- 13 -

(1.37) S/
NP Pred-Phrase
N Aux VP inner
/ I / A
John past V Prep-Phrase by passive
decide Prep NP
on the boat

(1.38) S
NP Pred-Phrase
N Aux VP Place
/ I I I
John past V Prep-Phrase
I o k
decide Prep NP
on the boat

- 14 -

Thus, those 'prepositional phrases' involved in verb subcategoriza-
tion fell to the left of the passive 'trigger' and were thus contained

in NP's able to undergo the passivization rule. Those prepositional
phrases not involved in verb subcategorization would fall to the right

of the 'trigger' and were not able to undergo passivization.

Tree diagram (1.37)--previous page--shows the deep structure of
the passive sentence (1.31), according to the Aspects formalization.

The deep structure of the active (1.32) with paraphrase (1.33)(b) is

shown in (1.38). As (1.39) shows, the passive counterpart to (1.38)
cannot occur because its phrase structure tree is both ill- formed3
and does not provide the environment for the passive transformation to

(1.39) *ill-formed

NP Pred-Phrase

N Aux VP Manner Place
/ / I I
John past V by passive Prep-Phrase
decide Prep NP
on the boat

The Aspects formalization of other passives (as opposed to 'pseudo-
passives') was very similar, e.g. (1.40) and (1.42). The deep struc-

ture for the active counterpart (1.41) would simply be given as lacking
the by passive 'trigger' in the manner adverbial.

(1.40) Fred was hit by John.
(1.41) John hit Fred.

- 15 -

(1.42) S

NP Pred-Phrase

N Aux VP NP Manner
/ / / I A
John past V N by passive
hit Fred

'Truncated' (or 'agentless') passives, in the same 1965 formaliza-

tion, included a transformation to delete the 'logical subject'


The recoverability principle (see above at footnote #1 to this

chapter) required that no deletion occur in such a way that the deleted

element not be 'recoverable'. In order to guarantee recoverability,

Chomsky decided that

a deletion operation can eliminate only a dummy
element, or a formative explicitly mentioned in
the structure index... or the designated repre-
sentative of a category. (1965:144)

In the case of passive sentences such as those under consideration, a

recoverable deletion could apply only to a dummy element.4 As Chomsky


the use of the dummy symbol A has been extended
here to the case of various unspecified elements
that will be deleted by obligatory transformations.

An agentless passive like (1.43) would--in the Aspects formalization--

be derived from the deep structure (1.44). Both the passive and dele-

tion transformations would obligatorily apply.

- 16 -

(1.43) Fred was hit.



NP Pred-Phrase

A Aux VP
past V NP
hit N by



This formalization more or less accounts for the base data. But,

whether it is more adequate and simpler than Bloomfield's account

remains moot.


1This, of course, implies the recoverability condition (Chomsky
1965, Chapter 3, especially 132, 137-138, 144-147, and footnote #3 on
page 222; Chapter 4, section 2.2), which will be of special interest
later on, relative to the discussion of agentless passives.

2As Chomsky states, the leftmost elipsis cannot contain an NP.

3Sentence (1.38) is ill-formed because the verb subcategorization
rules Chomsky mentions would block the verb decide from taking Manner
unless it also had a Prep-Phrase under the VP.

4The dummy element A was to be introduced by rule, in the cate-
gorical component, a simple phrase structure grammar 'with all the
lexical items mapped into the single symbol A' (1955:222). See
Chomsky (1965:esp. 222 ff.) for an extended discussion of the dummy
element and its relevance to deletion transformations.


In this chapter, the various suggestions relevant to the

formalization of the active-passive relationship which were made

after the appearance of Chomsky's Aspects of the Theory of Syntax

and prior to the emergence of the extended standard theory are sum-

marized. Only those writings within the 'mainstream' of

transformational generative grammar will be considered; other

relevant work will be discussed in subsequent chapters.

Hasegawa (1968) questioned the identification of the by

passive 'trigger' as a manner adverbial.

He argued initially that this aspect of the formalizaton

required further study, 'since there may be passive forms of verbs

which usually do not co-occur with manner adverbials, and the

notion of "manner adverbial" is itself not very clear' (p. 230).

However, he did not pursue this line of argument.

Hasegawa left the passive trigger in the manner adverbial,

though for reasons which are not clear, introduced notational

changes.I Rules (2.1) and (2.2), from the categorical component of

a transformational grammar, were given to describe the manner

adverbial (1968:237).

(2.1) Man + (Manner) (Ag)

(2.2) Ag-* of D
by J

- 17 -

- 18 -

The bulk of Hasegawa (1968) was devoted to the consideration

of two points of apparent inadequacy in the passive formalization

of Katz and Postal (1964) and Chomsky (1965). First, Hasegawa

(p. 321) argued that 'the status of the adjoined passive formative

been' in the underlying structure was 'not sufficiently clear'.

Also, he pointed out (p. 232) that the earlier transformationalist

formalizations did not include any attempt to generalize on the

relationship between been and get+en (the latter being the get


In reference to the lack of clarity in the status of been

(and get+en) in the underlying structure, Hasegawa discussed three

possibilities. The been (or get+en) might occur in the derived

phrase structure under the VP, under Aux, or under MV ('main verb').

The three possibilities (p. 231) are illustrated in (2.3), (2.4),

and (2.5).

(2.3) VP VP

Aux MV Aux been MV

(2.4) Aux Aux

Aux e Aux been

(2.5) MV MV

e MV been MV

For reasons sketched out near the beginning of this article (pp.

230-232), Hasegawa rejected the first two of these.2

Taken together with (2.1) and (2.2), this fragment of the cate-

gorical component proposed by Hasegawa formalized his view of

- 19 -

(2.13) 5


John Aux MV

Past V Cmp

get C S

en NP VP

Bill Aux MV

Past V NP Man
/ I I
see John Ag

by D

(2.14) S


he Aux MV

Past V Cmp

start C S

ing NP VP

he Aux MV
/ I
Past V

- 20 -

the status of been and get+en as part of MV:3
(2.6) S NP + VP

(2.7) VP + Aux + MV
(2.8) MV J be + C # S # )
f MV, (Loc) (Time)
(2.9) MV be + Pred
V (Particle) (NP) (Cmp) (Man)

(2.10) Cmp C # S #
(Prep Phr) (Prep Phr)

He first suggested been and get+en as V's which could take sentential

complements, e.g. in a structure V#S#.

Tree diagram (2.13)--facing page--shows the deep structure of

(2.11), as given by Hasegawa (p. 235). This structure for the get
passive is analogous to that for those verbs taking sentential comple-
ments in general; see (2.14)--also facing page--for an example of a

verb (start) which takes an ing complement.

(2.11) John got seen by Bill.

(2.12) He started singing.
Hasegawa was forced to formalize the be passive in a somewhat

different way, 'since be does not behave as a member of V in a

number of well-known transformations' (p.236). The passive be

appears in rule (2.8), above. In accordance with the above phrase
structure rules, the tree for a be passive counterpart to (2.11),
i.e. (2.15), would be as in (2.16)--on the following page.

(2.15) John was seen by Bill.

- 21 -

(2.16) S


John Aux MV

Past be C S

en NP VP

Bill Aux MV

Past V NP Man

see John Ag

by D
This brings us back to Hasegawa's second criticism of the Katz-

Postal-Chomsky formalization. Katz and Postal (1964) and Chomsky

(1965), he observed, had not generalized the be and get passives as

one 'process'. Although Hasegawa had one set of transformations

for both,4 the deep structures were given as quite different.

Thus, Hasegawa's formalization was at least partially subject to

the same criticism he applied to that of Katz and Postal, and,


There are other criticisms which could be levied against the

Hasegawa formalization.

There was necessarily a redundant specification of tense (via

Aux) in the matrix sentence and in the embedded sentence.

There was also a redundant occurrence of the patient noun

phrase in each.6

Unlike the Katz-Postal-Chomsky formalization, in which been

was introduced via rule, Hasewaga's formalization required redundant

specification in the deep structure of been (or get+en) as well as

by D.

- 22 -

Hasegawa also sought greater generalization in identifying be

and get with verbs taking sentential complements. But, be and get

would be the only English verbs taking en complements. Further,

the form of complementation for each of these two verbs would be

different. This hardly seems the path toward greater generaliza-


When all of these factors are taken into account, Hasegawa's

identification of be and get with verbs taking sentential comple-

ments seems poorly motivated. This is especially so in light of

the fact that the earlier formalization of Katz and Postal (1964)

and Chomsky (1965) was not subject to any of these latter


Kac (1969:146), also in opposition to Chomsky (1965) and Katz

and Postal (1964), suggested

that a somewhat reactionary view of transformations be
taken coupled with a drastic re-evaluation of the role of
consituent structure in generative grammars.

Specifically, he proposed that transformations 'become uniformly

optional and allow for the full range of operations (i.e. adjunc-

tion, permutation, etc.)'.8

His arguments were based on a misunderstanding of Chomsky's

revised treatment of the passive transformation. Kac apparently

thought that Chomsky intended an active-passive pair like (2.17)

and (2.18) to be derived from the same source, i.e. (2.19)--see

following page.

- 23 -

(2.17) John ate the watermelon.

(2.18) The watermelon was eaten by John.

(2.19) S
NP Pred P

N VP '0P Manner

John Aux V t Det N
was eat en the watermelon by passive

However, Chomsky (1965:132) stated that

many of the optional singulary transformation of Chomsky
(1955, 1957, 1962) must be reformulated as obligatory
transformations, whose applicability to a string is
determined by the presence or absence of a certain marker
in the string.

(The emphasis in the quotation is mine.) Chomsky's intention was

that the deep structure of a passive would contain the passive

trigger; in the active counterpart, the passive trigger would be

absent. See (1.40) (1.42), above.

The article by Kac (1969) is thus a non sequitur.

R. Lakoff (1971), in 'Passive Resistance', reviewed Hasegawa

(1968) and dealt with many of the same concerns.

She argued against Hasegawa's analysis, primarily on the basis

that he treated be and get passives in approximately the same way.

That is, the deep structures he suggested for (2.20) and (2.21)

would be essentially as in (2.22).

(2.20) Bill was hit by John.

(2.21) Bill got hit by John.

- 24 -

(2.22) S
Bill V' NP
be S
get[ / v

John V/ NP
hit Bill
She pointed out the nonequivalence of be and get passives. In
general, the get passive, she said (1971:160), 'often suggests the

active involvement, emotional or otherwise, of the superficial sub-
ject; the be passive, on the other hand, does not'. Based on the
differences she observed, Lakoff suggested that although Hasegawa's
treatment was a 'good candidate' for the deep structure of the get

passive, the G. Lakoff-Ross-Postal formulation was a better one for
be passives.9 Thus, a get passive sentence like (2.21) would have
the deep structure (2.22). But, according to the Lakoff-Ross-

Postal formulation, the be passive (2.20) would have the deep
structure (2.23).

(2.23) S


S be
John V NP
hit Bill

- 25 -

Chu (1973) examined the passive in Chinese and English. He

made this interesting observation (p. 444) about the be and get

passives: the imperative be passive is ungrammatical, while the

imperative get passive is at least sometimes grammatical. Compare

(2.24) (2.27).

(2.24) *Be arrested by the police.

(2.25) *Don't be arrested by the police.

(2.26) ?Get arrested by the police.10

(2.27) Don't get arrested by the police.

Largely based on this evidence, and on data presented in Lakoff

(1971), Chu tentatively proposed that the get passive 'is actually

a higher verb over the passive sentence' (1973-467). For the get

passive he suggested a deep structure like (2.28). The inchoative

get, he proposed, differed only in structure (i.e., not lexically),

and might be as in (2.29).11

(2.28) S
get x S

be x S

y x

- 26 -

(2.29) v ip
get x S

be y S
z y

In Syntactic Theory, Bach (1974) summarized the 'state-of-the-
art' in the transformationalist formalization of the active-passive

relationship. He gave several reasons why passive structures must
be derived, and not generated by the phrase structure component of
the grammar.
The first was the well-known argument involving the interaction
of passive and equi (p. 158).

The second reason (also p. 158) involved dative movement. The

absence of a passivization rule would greatly complicate the lexicon
to account for the ability of ditransitive verbs such as give to
appear in the passive with an object, e.g. (2.30). Further complica-

tions would result if the grammar were to include sentences like
(2.30) but exclude sentences like (2.31)--impermissible in some

(2.30) Sally was given the book by John.
(2.31) The book was given Sally by John.

- 27 -

Third, Bach (p.160) cited Postal's (1971) Crossover Principle

to explain the grammaticality of (2.32) and the ungrammaticality of

(2.33).12 'If both actives and passives were present in the base,

then we would need an entirely different explanation for the

ungrammaticality of these examples'--such as (2.32) (Bach

1974:160). Also, Bach stated that the grammaticality of (2.32) and

ungrammaticality of (2.33) established the directionality of the

rule relating actives and passives, i.e. from active to passive.

(2.32) John shaved himself.

(2.33) *John was shaved by himself.

Finally, Bach demonstrated why Hasewaga's formulation of the

be passive with a higher verb 'runs into insuperable difficulties'

(p. 161). He showed that it could not account for passives involv-

ing idioms composed of a verb and object noun of NP in cases where

the object appeared only in highly restricted environments.13 If

the restrictions were to be handled by lexical insertion of the

verb-object combination as a single item, then we could not account

for the existence of passives like (2.34). If such restrictions

were not to be handled by lexical insertion of this kind of idiom

as a single unit, the separate entries would have to be marked in

the lexicon to restrict them to the proper environments. Further,

he considered sentences such as (2.35). As he pointed out, Hasegawa's

'suggested formulation of Passive would have to find some ad hoc

- 28 -

means of getting the subject there of this sentence into its posi-

tion as an independent subject of a higher sentence. Notice that

there is not even present in deep structure' (Bach, 1974:162).

(2.34) Tabs were kept on his movements by the FBI.

(2.35) There was believed to have been a fire in the

It would seem that the best pre-extended standard theory formula-

tion of be passives was along the lines of that in Chomsky's Aspects

(1965). For reasons discussed above, the get passive should not

involve a higher verb get; the best formulation of the get passive

would appear to have been along the lines of a suggestion rejected

by Hasegawa, i.e. that get be introduced via rule to replace the

passive be. The best available formulation of the inchoative get

followed Chu (1973)--if separated from the passive get. However,

none of these descriptions is equally convincing with regard to all

parts of the data. For more discussion, see Chapter 10, below.


'Ag' represents an agentive by phrase; 'D' is the dummy element
(Chomsky's 'A', in Aspects); 'Man' is the new label for the manner
adverbial containing '(Manner)' and '(Ag)'.

2Akmajian (1977) offered independent evidence that been should
not be included under Aux; see esp. pp. 435-436.

- 29 -

3Hasegawa (1968) also included this fragment from the lexicon:

get, be [+_En#S#,... ]

begin, start [+_[aC1] #S#,...]

tend, try [+_to#S#,...]

quit, stop [+ Ing#S#,...]

En [+C,...]

0 [+C,...]
to [+C,...]

Ing [+C,...]

4Hasegawa's (1968) formulation actually divided the passive
transformation into three parts,

TAg -- 'replacement of the agentive dummy by the subject

NP' (p. 239)

TVC -- 'substitution of C for Aux of the embedded sentence'

(p. 140), where 'C' is be or get

Terase -- the subject of the embedded sentence is deleted

(p. 241); Terase was to be generalized to an

all-purpose equi-like rule

See also Chomsky (1972a:42), footnote #28) for further
Consider all other verbs in English which occur in structures
where equi obligatorilyy' applies, for example (i). In such cases,
equi applies if the environment for the rule's application is present.
If the environment is not present, it is still possible to have a
well-formed structure, e.g. (ii). Compare this to a

(i) Ike wants to leave right away

(ii) Ike wants Ted to leave right away.

- 30 -

parallel situation that would present a problem in Hasegawa's
analysis of passives, e.g. (iii) and (iv). Note that if the envi-
ronment for equi does not exist, the resulting structure is ill-
formed. In other words, unlike all other English verbs, the pas-
sive be would--in Hasegawa's formulation--require a deep structure
subject to equi.

(iii) Ira was hit by Tom.

(iv) *Ira was Bob hit by Tom.

7Further, inclusion of get passives under the Katz-Postal-
Chomsky formulation would not necessarily be so 'expensive' as
Hasegawa (1968) suggested (p.232).
8This is obviously in direct conflict with the Katz-Postal
Hypothesis. See below.

9R. Lakoff's (1971) observations generally seem to be valid.
However, the deep structures she proposed did not represent the
only, nor necessarily the best, possible formulations consistent
with the data. See below.

10Although Chu (1973:444) marked (2.26) ungrammatical in
isolation, I find it marginal, at worst. It seems perfectly
grammatical in the appropriate context. Consider:

SECRET AGENT: OK, boss, what should I do now?
CHIEF: We've got to find out about police corrup-
tion in Homersville.
CHIEF: Here's what you need to do. Go to Homers-
ville tomorrow morning. Get arrested by
the police. I don't care how you do it,
but get a look at the inside of that jail.


FATHER: Listen, son, I've been trying to
Do you want a criminal record?
SON: But, look Dad, everybody smokes dope.
FATHER: All right, I don't care! Get arrested by
the police.

A sentence like (2.25) is usually 'a strange thing to say' (I quote
a native speaker informant). This, I think is very likely to be
influencing judgements of grammaticality of such sentences.

11It should be noted that an English deep structure of VSO was
assumed by Chu (1973). He also used a modified version of Hasegawa's
treatment of the be passive, with be as a higher verb.

- 31 -

Chu's suggestion that the passive get and the inchoactive get
do not differ lexically does not seem to work. Consider, for one
thing, the ungrammaticality of (i) as opposed to the grammaticality
of (ii). Chu's original suggestion requires that the derivation of
(ii) be from a structure like (2.88) and that it include the applica-
tion of equi-NP deletion. If equi does not apply, then we get (i).
How can we block the generation of (i)? We could insist

(i) *The window got itself broken.

(ii) The window got broken.

(iii) John got himself punched in the nose.

(iv) John got punched in the nose.
that equi be obligatory. But, then we cannot explain the gram-
maticality of both (iii) and (iv). Sentences (iii) and (iv) seem
to require that equi be optional. But, consider (v) and (vi), which,
along with (i) and (ii), show that it is not. By this reduction ad
absurdum proof, we see that the passive get and the inchoative get
must exhibit a greater difference than that described by Chu in

(v) *John wanted himself to go fishing.

(vi) John wanted to go fishing.
Consider (vii) and (viii). These two sentences illustrate a lexical
difference between the two verbs get, above.

(vii) *John got himself punched in the nose, but it wasn't
his fault.

(viii) John got punched in the nose, but it wasn't his

12Sentence (2.33) was given by Bach (1974) as having unmarked
stress, since only sentences with unmarked stress are subject to
the Cross-over restriction.

13Bach (1974:161) gave as examples keep tabs on, make headway,
and take umbrage.


Three problems prominently involving the description of the

active-passive relationship have contributed to major revisions in

the 'standard theory' (the phrase with which Chomsky has referred to

his Aspects model). First, Chomsky (1972b, c) concluded that the

standard theory required revision with regard to the level at which

rules of semantic interpretation might apply in a derivation. This

problem area centrally involves consideration of the Katz-Postal

Hypothesis. Second, closely following Hasegawa (1968), but primarily

for different reasons, Chomsky (1972a) divided the passive transforma-

tion into two parts. This question--of a two-part passive rule--has

since been tied to the attempt (in the 'extended standard theory')

to simplify the types of rules in a transformational grammar. This,

in turn, closely concerns the third problem area, i.e., the division

of transformational rules into two types (as discussed by Emonds,

1970): 'root' and 'structure-preserving'.

The question of how the active-passive relationship is involved

with rules of semantic interpretation, and further, with the Katz-

Postal Hypothesis was considered in Chomsky (1972b) and (1972c).

Sentences such as those in (3.1) were presented there.1 The

importance of this data to Chomsky lay in the fact that (1972b:104)

'the scope of negation will be determined by the position of not in

- 32 -

- 33 -

surface structure'. This is important because the standard theory

stated that 'the grammatical relations that enter into semantic

interpretation are those represented in deep structure' (1972b:102).

Chomsky (1972b, c) presented active-passive data of this sort as

'cases in which semantic interpretation seems to relate more directly

to surface structure than to deep structure' (1972b:88).2 In earlier

work cited by Chomsky, Jackendoff (1969) had discussed the role of

surface structure in semantic interpretation. Jackendoff's arguments

were not widely distributed until Chomsky's papers appeared (those

which were reprinted by Mouton as Chomsky 1972b and 1972c); they

were given in considerable detail, later, in Jackendoff (1972).

(3.1) Not many arrows hit the target.

?The target was hit by not many arrows.

Many arrows didn't hit the target.

The target was not hit by many arrows.

?Not many arrows didn't hit the target.

*The target wasn't hit by not many arrows.

Chomsky (1972a:41) assumed an underlying structure for (senten-

tial) passives as in Aspects, i.e. (3.2). However, in a major theo-

retical departure, he divided the passive rule into two steps:

'agent postposing' and 'NP proposing'.3

(3.2) NP-Aux-V-NP-by

The justification for this division was based on observations

of 'passive-like' structures in complex noun phrases.

(3.3) the enemy's destruction of the city

(3.4) the enemy destroying the city

(3.5) the city's destruction by the enemy

- 34 -

First, Chomsky (1972a) established that structures like (3.3)

could not be derived from 'parallel' structures like (3.4). A struc-

ture like (3.4), he argued, had as its deep structure an S, but (3.3)

was underlyingly an NP. A sentence like (3.5) was taken to be 'only

apparently the nominalization of a passive' (Chomsky 1972a:43).

The 'passive-like' paraphrase of (3.3)--(3.5)--could not involve

the rule of passivization as presented in Aspects.4 The structural

description for the Aspects rule of passivization took as its domain

the structure under an S-node. Since NP's like (3.3)--and thus

(3.5)-- were underlying NP's, the earlier passivization rule could

not apply to (3.3) to yield (3.5). This situation seemed to require

that the relationship between (3.3) and (3.5) be totally different

from that between (3.6) and (3.7).

(3.6) The enemy destroyed the city.

(3.7) The city was destroyed by the enemy.

The division of passivization into two components permitted a

resolution of this problem.5 The environments for agent postposing

and NP proposing could be reformulated so that they could apply in

the domain of 5, or, within a complex NP.

Chomsky cited other motivation for separation of the rule into

two parts. Within complex NP's, either agent postposing or NP prepos-

ing could apply independently.

Consider the underlying structure (3.8). With or without NP

proposing, destroy will be realized phonologically as destruction.

The preposition of will be inserted via 'a general rule applying to

N-NP constructions' (Chomsky 1972a:41-42). Without NP proposing,

- 35 -

(3.8) would yield (3.9). If NP proposing applied, we would get


(3.8) the (destroy, +N) the city

(3.9) the destruction of the city

(3.10) the city's destruction

Consider (3.11). Here, agent postposing can apply independently

to yield (3.12).

(3.11) the enemy's (destroy, +N) the city by

(3.12) the destruction of the city by the enemy

Note that to complete the generalization, NP proposing must be

extended 'so that it can apply not only in cases given before, but

also before agent phrases'. Only with this extension can both NP

proposing and agent postposing occur together. See (3.13) (3.15).

Given this extension of NP proposing, Chomsky was able to use the

same two rules in the formalization of the active-passive relation-

ship, whether the domain was an S or a complex NP.6

(3.13) the enemy (destroy, +N) the city by

(3.14) the destruction of the city by the enemy

(3.15) the city's destruction by the enemy

(3.16) the John's (offer, +N) by

(3.17) the offer by John

The division of passivization into NP proposing and agent post-

posing had one other important theoretical consequence. Consider

(3.16) and (3.17). According to Chomsky (1972a:42), 'of the two

components of the passive transformation, only NP-preposing and not

agent-postposing requires the presence of an object (more generally,

- 36 -

a noun phrase, as in the "pseudo-passives" John was laughed at,

...approved of, etc.) in the position following the verb'.7

The question of a two-part passive rule later became critical

in the development of the S and X' convention. Jackendoff (1974)

suggested an equivalence of S and V" in order to generalize notions

such as 'subject of' and 'object of' to NP's. This generalization

was intended to include rules of semantic interpretation, as well as

selectional restrictions (p. 14).

Hornstein (1977), however, argued that this identification of S

and V" was incorrect. As he pointed out, such a generalization of

the projection rules 'would state, in effect... that the relation

between an NP and its predicate in a sentence is the same as that

between an NP head and its nominal phrase in an NP. Semantically,

however, it is not clear that they are parallel' (Hornstein, 1977:141).

Hornstein argued for the elimination of NP postposing. Citing

examples such as (3.18), he showed that some instances of by NP could

not be the result of NP postposing 'as there is no place that...'

such NP's 'could have been moved from' (p. 144). In cases

such as (3.18), then, the by phrase had to be generated in the deep

structure. Hornstein also pointed out (see esp. p. 145) that NP

postposing did not always apply in parallel fashion to structures of

the form NP's N and within an S. See (3.19) (3.21). Thus, Horn-

stein suggested, all by phrases in complex NP's could be generated in

the deep structure. He proposed a deep structure for NP's as in


(3.18) John's photograph of Mary by Warhol

(3.19) (a) *the sense of danger by John

- 37 -

(b) Danger was sensed by John.
(3.20) (a) *the fear of Harry by John
(b) Harry was feared by John.
(3.21) (a) *the respect for Mary by John

(b) Mary is respected by John.
(3.22) NP N (of) NP (by) NP
+possession +theme +agent
In 'On wh-movement', Chomsky (1977) still assumed a formulation
following the Aspects model. He gave the deep structure for (3.23)
as (3.24), using a notation that incorporated no significant departure
from past practice. The passive transformation was the result of
two applications of the rule Move NP.10 Thus, (3.24) becomes (3.25),
then (3.26).
(3.23) Bill was killed by John.

(3.24) (S(NP John) (ypbe en kill (NP.Bill)bY(NPke)))
i j k
(3.25) (S(NP.e) (ypbe en kill(Np Bill)by(NP John)))

(3.26) (S(NP.Bill)(vPbe en kill(NP e)by(NP John)))

The agentless passive, e.g. (3.27), was similarly derived by
one application of Move NP. The deep structure of (3.27), i.e. (3.28),

results in (3.29).11
(3.27) Bill was hit.

(3.28) (S,(S(NPe)(VPwas(APen(vp(Vhit)(NpBill))))))

(3.29) (S'(S(NpBill)(vpwas(Apen(Vphit(Npe)))))

Note that the rule (or rule sequence) which resulted in
'passivization' was at this point no longer viewed as obligatory.
The by phrase no longer served as a passive 'trigger'.12

- 38 -

Both Jackendoff (1972) and Chomsky (1977) cited Emonds work on

the notion of structure-preserving rules. As Jackendoff put it


it is no accident that the deep object of a passive sentence
comes to occupy subject position rather than perhaps a
position between the auxiliary and the main verb, where no
noun phrase can be generated in the base. Likewise, it is
no accident that the deep subject of a passive ends up in
a prepositional phrase which is like all other preposi-
tional phrases, rather than in some altogether new kind of

Chomsky gave an example (namely the derivation of (3.23), above) to

demonstrate the notion of structure-preservedness as it applies to

passive. In (3.24), we have the deep structure for (3.23). After

NP postposing, (3.25) results. After NP proposing, the result is

(3.26). Notice that an empty NP is not obliterated during the deriva-

tion, but is preserved as an indexed NP with null terminal.13


In earlier work cited by Chomsky, Jackendoff (1969) had dis-
cussed the role of surface structure in semantic interpretation.

2Chomsky's own (English) counterexamples to the standard theory
all involve 'the scope of logical elements' (1972b:106), i.e. nega-
tion and quantification.
3This is essentially an adaptation of Hasegawa (1968) to fit
within the framework of the (extended) standard theory.

4Thus, while NP's like (3.4) were formulated as nominalizations
of passive S's, NP's like (3.5) were formulated as passivess of base-
generated derived nominals' (Chomsky, 1972a:43).
5But introduced another. See Sullivan (1976:119).

61n Chomsky (1972a) the X' convention was introduced. See espe-
cially pp. 52 ff. In this article, Chomsky suggested that rules
whose domain was S should also apply within the domain of N'.

- 39 -

7As Chomsky noted, Hasegawa's (1968) agent postposing transforma-
tion also lacked any requirement for the presence of an object. In
the case of Hasegawa's formalization, the 'erase' transformation
performed the necessary filtering function.

8There are two main problems with a preposing-only account.
First, as Hornstein himself noted (p. 149), proposingg from the
of-NP slot is not always so good'. The same can be said for on-NP.

(i) John was done a favor.
the favor for John
*John's favor

(ii) Poland was attacked.
the attack on Poland
*Poland's attack

Similarly, locatives, which can supposedly be proposed in the domain
of S to yield so-called 'pseudopassives', cannot be proposed in the
domain of NP. Compare:

(iii) The bridge was flown under.
*the bridge's flight under

Consider also the case of certain temporal expressions, in which the
situation is reversed:

(iv) *Tuesday was observed the Moon.
Tuesday's observation of the Moon

The second problem is one of an unnecessary complication of
selectional restrictions. If all by phrases in NP's are
base-generated, two choices result. We could accept non-parallel
selectional restrictions on agent NP's in the domain of S vs. those
in NP's with derived nominals (e.g. destruction). Or, we could
attempt to derive the sentential active from an S with the agent N
in a by phrase. Both alternatives carry serious consequences; see
Chomsky (1957:42 ff.).

9Trace theory also entered the picture at this point. Consider
Hornstein's remarks:

If... NP-Postposing were dropped and we adopted proposing
rules exclusively, traces would never be left stranded and
subsequently obliterated by spelling rules... also...
spelling rules applied to traces could probably be dispensed
with. (p. 146)

Rightward NP movement rules and trace theory do not mix.
(p. 147)

40 -

It is interesting to note that a trace theory without
spelling rules would predict that no rightward NP movement
rules exist. (p. 147)

1Agent postposing and NP proposing are two specific examples
of Move NP.

11In (3.29), (Npe) is the trace of Bill. See below.

12Chomsky and Lasnik (1977:431): 'The transformational rules of
the core grammar are unordered and optional.... Adjacency of cate-
gories cannot be stipulated and no more than one element of the con-
text in which the operation applies may be specified'.
Chomsky (1980:6): '...All aspects of obligatoriness of
syntactic rules, contextual depencies, and ordering fall in a natural
way under local surface filters....'
13This is to satisfy the constraint on structure-preserving rules
specified by Emonds that--in the words of Jackendoff (1972:13)--'with
a certain class of exceptions... the output of a transformation must
be a structure that can be independently produced by a base rule'.
Or, as Chomsky and Lasnik (1977:432) stated, 'we assume that a move-
ment rule always leaves a trace....'


This chapter deals with the formalization of the active-passive

relationship in three 'alternative' transformational theoretical


Case Grammar

First, this chapter will explore the formalization within the

framework set forth by Fillmore in various articles from 1966 to


Fillmore presented his alternative transformationalist framework

because he did 'not believe that "subject" or "object" are to be

found among the syntactic functions to which semantic rules must be

sensitive' (1966:21). To illustrate this point, he cited sentences

such as the following:

(4.1) The door opened.

(4.2) The janitor opened the door.

(4.3) The janitor opened the door with this key.

(4.4) The key opened the door.

(4.5) The door was opened with this key.

(4.6) The door was opened by the janitor.

As he stated, 'the semantically relevant relation common to...'

sentences (4.1) and (4.2) '...is that between the subject of the

- 41 -

- 42 -

intransitive verb (in (3.1)) and the object of the transitive verb

(in (4.2)), not between the subjects of the two sentences' (p. 21).
To account for such data, Fillmore proposed a different base

structure from that seen in Chomsky (1965). Rules (4.7) (4.10)
illustrate this.1

(4.7) S + Mod Aux Prop

(4.8) Prop -V (Erg) (Dat) (Loc) (Inst) (Ag)
(4.9) {Erg, Dat, Loc, ... ) NP

(4.10) NP P (Det) (S) N

Two major features distinguished Fillmore's earliest version of

the case grammar from an Aspects-like framework. Both concerned the
treatment of the noun phrase. First, the noun phrase was to be

dominated by a functional-category maker, i.e. 'Erg', 'Dat', 'Loc',

'Inst', or 'Ag'. Second, the underlying representation of every

noun phrase contained a preposition ('P' in rule (4.10), above).

The preposition provided the realization of the underlying function

of the NP (e.g. by for Ag, with for Inst or Erg). Preposition-

deletion rules applied in certain cases, based on syntactic

environments--to most direct objects, and to all subjects; see
(4.11) and (4.12).

(4.11) bees swarm in the garden
Erg Loc

(4.12) Lthe garden swarms with bees
Loc Erg

- 43 -

As of Fillmore (1966), the rule proposed to relate active and

passive was 'ergative fronting in transitive sentences' (p.30). This

rule is given in (4.13), where the '2' in the SC is the 'Erg' in the


(4.13) SD: Aux V-Erg-Y (Inst) (Ag)
(X Be en)

SC: 1-2-3 2-1-3

As Fillmore explained, this rule was not totally adequate. Two

problems were primary. First, it did not handle 'dative-subject

passives' (e.g. Bob was sent the letter). Also, it did not account

for certain 'non-deletable prepositions' (as in It was looked at).

The primary advantage of Fillmore's framework over that of

Chomsky's Aspects was that the former was seen to relate not only

(4.14) and (4.15), but further, to relate (4.16) and (4.17) with

them and with each other.

(4.14) John broke the window (with a hammer).

(4.15) The window was broken (by John) (with a hammer).

(4.16) The window broke.

(4.17) The hammer broke the window.

In 'The case for case' (1968), Fillmore further expanded and

formalized his framework.

He distinguished 'case', 'the underlying syntactic-semantic

relationship', and 'case form', 'the expression of a case relation-

ship in a particular language' (Fillmore 1968:21).

- 44 -

He modified the base rules as well.2 Rule (4.7), above, was

replaced by (4.18), (4.8) by (4.19), (4.9) by (4.20), and (4.10) by

(4.21). The 'case categories' C1 to Cn were to be selected from the

set in (4.22).

(4.18) S + Mod + Prop

(4.19) Prop V + C + ... + C

(4.20) {A, I, D, F, L, 0 } K + NP ('K' for

(4.21) NP + (Det) (S) N (assumed)

(4.22) Fillmore's (1968:24) cases:

Agentive (A)
Instrumental (I)
Dative (D)
Factitive (F)
Locative (L)
Objective (0)

The set of verbs possible in a given deep structure depended on

the 'case frame' of the sentence, i.e. on the combination of cases

present in the deep structure. For example, the verb break could

occur in any of the frames [ + 0 ], [_ + 0 + A], [___ +

0 + I ], or [___ 0 + I + A ]: see (4.14) to (4.17), above.

'Optional' elements were indicated by parentheses; thus, the 'case

frame' for break could be given as [___ 0 (I) (A) ].

- 45 -

Trees (4.23) (4.27) show the derivation of the active sentence
(4.14) in the 'Case for case' framework.
Mod Prop
I V 'A
Det | N
SI I o I I
past break 0 the window by John
(4.24) by 'subject-fronting' from (4.23)
A Mod Pr

K NP Det N
by John past break 0 the window
(4.25) by 'subject-preposition deletion' from (4.24)
NP Mod ~Prop
V 0
Det N
John pt break the
John past break f the window

- 46 -

(4.26) by 'object-preposition deletion' from (4.25)

NP ModS Prop
Det N
John past break the window
(4.27) the surface structure derived from (4.23)
NP Prop
Det N
John broke the window
The passive construction (4.15) was accounted for 'via the
association of the feature [+passive] with the V' (p.37). The
feature [+passive], in Fillmore's framework, permitted 0 or D to be
realized in subject position despite the presence of A,3 provided
the passive verb form, and blocked application of the object-
preposition deletion rule. Trees (4.28) (4.32) show the deriva-
tion of the passive (4.15) (without I).
(4.28) c.f. the active (4.23)
Mod pProp
K NP by NP

a I I I I I
past break 0 the window by John

- 47 -

(4.29) via 'subject-fronting'

0 Mod 0Prop

Det N
0 the window past break by John
(4.30) via 'subject-preposition deletion'
Det N K NP
1 I I I
the window past break by John
(4.31) conditioned insertion of be

N P M o d V 0 P r o p

Det N K NP
I I I I I y
the window past be break by John

- 48 -

(4.32) conditioned morphological change

NP Mod Prop

+passive K \
/ I /\
Det N K NP
t I I I I
the window was broken by John
The above examples illustrate the basic aspects of the formaliza-
tion of the active-passive relationship in Fillmore (1968). Passive
sentences with D (dative) subjects were to be derived in a similar
fashion (see Fillmore pp. 39-40, esp.)
Sentences (4.33) and (4.34) represented a related problem that
had to be dealt with.
(4.33) The rats were killed by fire (*by the pest-
control officer).
(4.34) The rats were killed with fire (by the pest-
control officer).
Fillmore's solution, presented in 'Toward a modern theory of
case' (actually a rewritten and expanded version of Fillmore (1966)),
was to introduce the following lexical insertion rule: 'The Instru-
mental preposition is by if there is no Agentive present, otherwise
it is with' (1969:374). The deep structure for (4.33) would be
(4.35), that for (4.34), (4.36).

- 49 -


Mod "" Prop


I 0 t I I
past kill 0 the rats by fire
Mod Prop

Det N
I eI I I I
past kill 0 the rats with fire 0 0
Thus, (4.33) and (4.34) were distinguished on the following basis.
Sentence (4.33) marked the I with by because there was no A in the
deep structure (4.35). However, (4.34) contained a dummy NP under A
in the deep structure; this dummy element was deleted after it condi-
tioned the choice of insertion of by or with.4
By Fillmore (1977:65), the initiall association of the preposi-
tion by with the agent case had to yield in favor of a more com-
plicated principle associating by with the highest-ranking case in
the sentence'. This was necessary in order to account for

- 50 -

such varied types of by phrases as in eaten by George, destroyed by

fire, assumed by everybody--containing, respectively agent, instru-

ment, and experience5 noun phrases. Fillmore himself pointed out

(1977:65) that

in the end... such provisions did not look any better than
an account according to which the preposition by gets
introduced by means of a Passive transformation.

Relational Grammar

In 'relational grammar' (or 'RG') a new transformationalist

school arose which took 'subject', 'object', 'verb', etc. as the

grammatical categories relevant to the formalization of grammatical

rules. Advocates of relational grammar have asserted that such gram-

matical categories are linguistic universals, and that the most

universal formulations of syntactic rules will take these categories

as their arguments.

Johnson (1975) was probably the earliest work in relational

grammar which was widely read.6

In Johnson's framework, both the active (4.37) and its passive

counterpart (4.38) had the same underlying representation,7 i.e.


(4.37) John hit Fred.

(4.38) Fred was hit by John.


John hit red
John hit Fred

- 51 -

In (4.39) 'S' represents the relation 'subject', 'V the relation

'verb', and 'DO' the relation 'direct object'. The passive (4.38)

was to have the 'derived relational network' (4.40). The underlying

DO was 'promoted' to S. 'X' represents the grammatical relation of

an item 'demoted' from S.8
(4.40) V

?2 1
Fred hit John

In (4.40), the feature [+passive] is not to be confused with

Fillmore's lexical feature [+passive] Johnson's feature [+passive]
was introduced by rule to a derived relational network. As such, it
constituted 'an ad hoc device to distinguish active and passive verbs'


The passive rule was given by Johnson (1975:34) as (4.41). 'RD'

is the 'relational description' which is the input to the rule; 'RC',

the relational change', describes the output from the rule. The

lower-case 'x' and 'z' are variables over 'relational subnetworks'.

(4.41) RD RC

S(x,Vi) X(x,Vi)
DO(z,Vi) S(z,Vi)

- 52 -

The basic form of rule (4.41) was taken to be universal. However,

Johnson recognized that the passive rule would have to possess

language-specific features. For example, the English passive rule

'must specify that the passive verb form is BE + V + past participle

and that the ex-subject is marked with the preposition by' (p. 47).

The above rule (4.41) had two parts. The first was the 'demo-

tion' of the subject to the 'X' relation. This corresponded

--in English--to agent postposing at the sentential level, half of

the process of passivization in the EST framework. The second part

of the rule corresponded--again, with regard to English, at least--

to sentence-level NP proposing in EST.

Keenan (1975) demonstrated clearly that any universal form-

alization of the active-passive relationship could not be defined in

purely structural terms, i.e. 'in terms of the changes induced in

the dominance and linear ordering relations of P-markers' (p. 340).9

He used such arguments to support a 'relational' definition of pas-

sivization, one defined 'in terms of the changes it induces in the

grammatical relations... which NP's bear to their verbs' (also

p 340).

Perlmutter and Postal (1977) presented many of the same argu-

ments against a structural approach to the description of the active-

passive relationship. Like Keenan, they proposed an approach based

on the use of notions such as 'subject', 'direct object', and so on.

They introduced the notion of 'strata' in a relational grammar.10

For example, consider the passive sentence (4.42). According to

Perlmutter and Postal's notion of 'strata', the relations of the

- 53 -

NP's that book and Louise change from one 'stratum' to another as

shown in Fig. (4.43). At one 'stratum', Louise has 'relation 1'

(subject) and that book has 'relation 2' (direct object). At the

second 'stratum' that book has 'relation 1' and Louise has the 'X

relation' (chomeur).11

(4.42) That book was reviewed by Louise.

P 1 2

P X 1X

reviewed Louise that book

Those working in relational grammar have generally accepted

this type of formalization of the active-passive relationship.

Consideration of the relevance of the 'Noun Phrase Accessi-

bility Hierarchy' to the RG formalization of the active-passive rela-

tionship is beyond the scope of the present study. For some discus-

sion of this matter, see Keenan and Comrie (1977, 1979), and Comrie

and Keenan (1979).

- 54 -

Lexicalist Theory

A major divergence from Aspects--though still within a genera-

tive framework--was proposed by Freidin (1973).12 He stated (1973:1)

that the general acceptance of the derivation of passives via a tran-

formational rule was 'based on the assumption that the active-passive

relation is structural in nature and is therefore best handled by a

transformation'. Freidin claimed that this assumption was erroneous

and that it entailed basic problems for the theory.

One criticism he levelled against the 'standard theory' involved

the assumed synonymy of active and passive counterparts. If synonymy

is assumed, he argued, then (4.44) (a) and (b) must be synonymous.

This is a problem, since 'such a claim might lead us to wonder how a

manner adverb--in this instance by A --could be semantically empty'

(p.2). However, in light of ample data against the exact synonymy

of actives and passives (even excluding cases involving quantifica-

tion, negation, etc.), this argument seems to be a non sequitur.

(4.44) (a) S

(b) S


by A

- 55 -

Another argument by Freidin against the Chomskyan (in this case
Aspects) formalization of the active-passive relationship was equally
irrelevant. This argument concerned the recoverability condition
imposed by the acceptance of the Katz-Postal Hypothesis. Freidin
claimed that a sentence like (4.45) should be assigned two possible
readings by the standard theory model, if it involved deletion either
of by someone or by something. But (4.45), he pointed out, is not
so clearly ambiguous. Further, such an analysis violated the recover-
ability condition. Aspects, however, did not contain any proposal
for such an analysis; rather, in it Chomsky suggested something like
(4.44) (b).
(4.45) Bill was hit on the nose yesterday.
Freidin (pp. 6-8) did also present arguments against the empty-

node analysis of (4.44) (b).
In the first place, he asserted, such an analysis postulated
'two distinct underlying structures for passives' (p. 6), i.e.
(4.46) (a) and (b). (The '#' represents some lexical item.)
(4.46) (a) S

S/ /\
S # i by

(b) S
NP 00*VP

I 0
# #

- 56 -

While this may form the basis for a theoretical objection, of prime
consideration should be the question of whether both underlying struc-

tures are empirically motivated. It is argued in Chapter 6, below,
that two (such) sources are required for English passives.

Freidin's second objection to the formulation (4.46) (b) also

involved a fallacy. He argued that two applications of NP proposing
on (4.47) would yield either (4.48) (a) or (b). But, if NP proposing
into an empty NP node resulted in passivization, then (4.48) (a)

could not result from (4.47). Thus, Freidin's apparent 'counter-

example' was not one, unless it were required that all deletion rules
operate on the empty node.13
(4.47) N SV

expect S

nominate Harry

(4.48) (a) Harry expects to be nominated.

(b) Harry is expected to be nominated.

Freidin did point out some very real problems with the descrip-
tion of the active-passive relation in (purely) structural terms.
All of these problems resulted from the assumption that the ability
to passivize a sentence depended on some lexical

- 57 -

property of its main verb. This lexical property was either defined

in terms of strict subcategorization features (as in Aspects) or in

terms of a rule feature [-passive].

The strict subcategorization approach involved the assignment

of by passive to the manner adverb slot by the base rules. As

Freidin observed, some very serious problems had surfaced with regard

to the description of cooccurrence restrictions on verbs, manner

adverbs, and by passive in the standard theory. Certainly, the prob-

lems were serious enough to make the standard theory formalization


In the rule feature approach, unpassivizable verbs which per-

mitted a following NP (i.e. those which satisfied the structural

description for the application of the passive rule) were marked

[-passive] in the lexicon. The feature [-passive] blocked appli-

cation of the passive transformation. This 'solution', as Freidin

noted, was an ad hoc one, since it 'is no more revealing than a list

of verbs which are exceptions to the passive rule' (p. 10).

Freidin thus came to the same conclusion as did those in rela-

tional grammar: 'structurally there is no motivation' for excluding

sentence (4.49) (b) from undergoing passive (p. 10). (The

emphasis in the quote is mine.) Therefore, 'it is not strictly true

that transformations apply blindly to phrase-markers of the proper

form' (also p. 10).14 And (p. 11), 'there seems to be no non-ad hoc

way for PASSIVE to filter out the ungrammatical strings' such as

(4.49) (b).

- 58 -

(4.49) (a) Max resembles Harry.

(b) *Harry is resembled by Max.

Freidin's lexicalist proposal described the active-passive rela-

tionship in the lexicon. The proposal consisted of two main parts.

According to Freidin's preliminary proposal, the lexical entry

for give would resemble (4.50).

(4.50) /gIv/

+MOTIONAL: (involves the movement of a physical
object (the obligatory NP in the
strict subcategorization features
below) to a particular location (the
optional NP in the strict subcategoriza-
tion feature for +V, or the NP in an
optional PP where the preposition is

Item::: +V: [+ (NP) NP (PP)]; ...

+N: /gIft/; [+Det __ (PP) (PP)]; ...

Snom: N = object Ved

+A: Ma; [+ NP]; ...

This part of the proposal required the mirror-image cooccur-

rence restrictions for actives and passives argued against in Chomsky

(1957). To avoid this problem, Freidin suggested a redundancy rule

which would predict actives and their selectional restrictions from

entries for passives given in the lexicon. The redundancy rule (p.

16) was stated as (4.51).

- 59 -

(4.51) Mpass entails the existence of an active

verb where (V-active NPx -- NP ) if

(V-passive NP NPx) and V-active and
V-passive are semantically equivalent.15
Given this redundancy rule, it becomes unclear how examples such as

/gIv/ are to be handled. Notice that there is a third NP with any
such ditransitive verb, NPz, the receiver of the object. The possible

orderings are shown in (4.52). Further, note (4.52) (c). The problem

of the existence of this second possible passive was left unresolved

by Freidin.

(4.52) (a) V-active: NP ___NPy (prep + NPZ)

(b) V-passive1: NPy (prep + NPz) (prep +


(c) V-passive2: NP____NP (prep + NPx)
The second part of the proposal suggested that passives should

be analyzed as the main verb be plus a predicate adjective (the past

participle). Given this part of Friedin's proposal, the question of

where been was to be assigned (as part of Aux, MV, etc.--see the

discussion of Hasegawa (1968), above) disappeared.

Where one problem disappeared, another reappeared (but was not

observed by Freidin): the problem of mirror-image selectional restric-

tions. How could the listing of an adjective (if passive past par-

ticiples were to be treated as such) in the lexicon imply--via any
sort of redundancy rule--the existence of an active verb and its

cooccurrence restrictions?16

- 60 -

Despite the problems with the statement of the redundancy rule

for cooccurrence restrictions, this lexicalist proposal attempted to

capture the following generalization in formal terms:

NP's perform the same semantic functions even though the
form of the predicate has changed from active to passive
and the syntactic positions of the NP's have been reordered.
(Freidin, 1973:12)

Freidin's proposal attempted to capture this generalization in

the lexical entries of verbs which possessed both active and passive

forms. See again the example for /gIv/. Two sets of selectional

restrictions were stated. One set was stated with regard to semantic

features (above) in prose. The other set of selectional restrictions

(below, at 'Item:::', in standard TG notation) dealt with base struc-

ture. In this view, 'every predicate... governs a particular set of

semantic relations' (p. 12), and every verb governs a particular set

of structural relations.

In this sense the lexicalist proposal amounted to an implicitly

stratified model with regard to syntactic and semantic relations.

The model formally and explicitly related actives and corresponding

passives in terms of semantic relations.17 It formally

distinguished actives and passives syntactically.

There are, however, potentially grave problems arising from the

type of structure assumed for 'the lexicon' in this and other genera-

tive models. For a detailed discussion, see 'A stratificational

view of the lexicon' (Sullivan, 1977b).

- 61 -

An alternative to the various generative views of 'the lexicon'

has been suggested by stratificationalists.18 This alternative

(stratificational) formalization explicitly and formally separates

the two sets of selectional restrictions, without incurring the other

problems with 'the lexicon' which were alluded to above. This separa-

tion of the two sets of restrictions is inherent in the stratifica-

tional approach to 'the lexicon' and requires no additional descrip-

tive machinery. In this respect, it differs significantly from the

lexicalist position. The resulting formalization of the active-

passive relationship is discussed in the following chapter (Chapter 5).


1The following are abbreviations used in this chapter:

Prop 'proposition'
Erg 'ergative'
Dat/D 'dative'
Loc locativee'
Inst/I 'instrument'
Ag/A 'agent'
P 'preposition'
Mod 'modality'
0 'objective'
F 'factitive'
K 'kasus' (case marker)

2Fillmore's case grammar contained two types of rules. Rules
of type I mapped case systems into phrase-markers. Rules of category
II were transformations which generated derivations in the usual
way. Thus, passive was a rule of category I.
Chomsky (1972c:173) asserted that 'it would be quite consistent
with his theory of case grammar to regard passive as a transformation,
rather than a rule mapping case structure onto phrase-markers'. Exactly
what Chomsky meant by this is not altogether clear.

- 62 -

Consider. A phrase-marker (upon which rules of:category II
operate) indicates a 'default' linear order. This 'default' order
is to obtain if no (further) reordering transformations apply. The
deep structures in Fillmore's case grammar possessed no linear order-
ing, default or otherwise. For this reason, and because Fillmore
considered passive to be a rule of category I--rather than a trans-
formation of category II--'there are clear empirical differences...'
between Fillmore's proposal and that of Chomsky (1965) '...concern-
ing the status of passive' (Chomsky, 1972c). But, what happens to
these empirical differences if Chomsky's suggestion is followed and
the rules of category I are reformulated as transformations is that
the empirical differences dissolve. The implication of Chomsky's
(1972c) statements is that he would be able to find Fillmore's case
grammar acceptable. But this could be the case only if it were re-
duced to a 'mere notational variant' of the extended standard theory.
Then, of course, it could be rejected anyway because it would contain
no empirically significant divergence from the extended standard
Chomsky (1972c) devoted several pages to a discussion of the
differences between Fillmore's case grammar and the extended stan-
dard theory (pp. 173-180). His arguments against semantically-based
grammar do not hold up. See for example, one straw man he easily
knocked down in Chomsky (1972b:85):

If the concept of 'semantic representation' ('reading') is
to play any role at all in linguistic theory, then these
three expressions must have the same semantic representa-

(33) John's uncle

(34) the person who is the brother of John's mother or
father or the husband of the sister of John's mother
or father

(35) the person who is the son of one of John's grandparents
or the husband of the daughter of one of John's grand-
parents, but is not his father

In the same article, he discussed the question of notationall variants'
of EST. He said, 'it is senseless to propose as an alternative a
"semantically-based" conception of grammar in which S (the semantic
representation) is "selected first" and then mapped onto the surface
structure P and ultimately P (the phonetic representation)'
(ibid.). See Chapter 5, below.
3Fillmore intended 0 and D to be--along with A--'optional
choices provided by a Subject Selection Rule' (1977:69).

4As Fillmore said later (1977:71), deciding on the case status
of some NP's, e.g. fire in (3.33), 'is like deciding on the "-emic"
status of the stopped consonant in spy'.

63 -

5Fillmore (1971:42) introduced the case 'Experiencer'.

6Johnson (1975) cited earlier work by Keenan and Comrie, as
well as talks given by Postal and Perlmutter (see Johnson, 1975:3).

7This underlying representation was termed by Johnson (1975:31)
the 'underlying relational network' (or 'URN'). It should be noted
that such trees in RG do not indicate any linear order, but only
functional relations. Linearization is imposed by a sequencing rule
which enforces the order S-V-DO-(X); see Johnson (1975:36).

8For arguments concerning the justification of the 'X' relation
(chomeur), see Sheintuch (1976).

It is primarily in such structural terms that EST currently
defines the active-passive relationship.

10This notion of strata is not equivalent to that used by
stratificational linguists. In the RG framework, the set of NP rela-
tions is universal and is the same at all strata, except that the
'X' relation (chomeur) cannot occur in the first stratum. The 'strata'
in RG are analogous to stages in a TG derivation.

11According to the Stratal Uniqueness Law, 'only one dependent
of a clause can bear a given term relation in a given stratum'
(Perlmutter and Postal, 1977).
12Other similar lexicalist proposals have been set forth on
similar lines. For the sake of brevity, the discussion will be
limited primarily to Freidin (1973).

13In effect, Freidin argued against the empty node analysis on
the basis that such an analysis would not work for the Complement
Subject Delection, involved in (4.48) (a). This, of course, has
nothing to do with whether or not such an analysis could work for

14See Chomsky (1972c:197), quoted by Freidin: 'Each transforma-
tion applies to a phrase-marker on the basis of the formal configura-
tions expressed in it, and quite independent of the meanings or gram-
matical relations expressed by these formal configurations'.

1Freidin asserted that this resolved the problem of rule
feature analysis, which marked unpassivizable transitive verbs with
the feature [-passive].
But, his formulation required the listing in the lexicon of
which verbs can be passive. Which is to be preferred--marking the
marked condition, or marking the unmarked condition? The rule
feature analysis (almost) seems preferable on this basis.

64 -

16More recently, those taking the lexicalist position have tended
to differentiate between 'adjectival passives' and 'verbal passives'.
See, for example, Bresnan (1980:21 ff.).

17This formally-specified relationship is to be preferred to the
EST formalization, which assigns the similarity in their interpreta-
tion to a vague, unspecified set of 'projection rules'.

18Sullivan (1977b) explained this alternative in terms not
restricted to the stratificational point of view. Several articles
in Makkai and Lockwood also deal with this issue. See also Lockwood
(1972), Chapter 2: 'A preliminary sketch of linguistic structure',
pp. 14-29.


To avoid what has been a continuing misunder-
standing, it is perhaps worth while to
reiterate that a generative grammar is not a
model for a speaker or hearer. (Chomsky,

When we speak of a grammar as generating a
sentence with a certain structural descrip-
tion, we mean simply that the grammar assigns
this structural description to the sentence.
When we say that a sentence has a certain
derivation.., we say nothing about how the
speaker or hearer might proceed, in some
practical or efficient way, to construct such
a derivation. (Chomsky, 1965:9)

Lamb is trying to develop an analogical model
for the production and comprehension of
speech, a theory that will not only define
and describe the texts of a language, but
will do so in a way that explains how human
beings themselves produce and understand such
texts. (Algeo, 1972:10)

Transformational linguists have pretended
that the uncertain applicability of their
theory in the last area (language production
and understanding--DWC) is not a shortcoming.
(Chafe, 1967:89)

Our goal is to model natural language
behavior. (Reich, 1970:21)

As Reich (1972a:85) has said, SG is 'a theory which accounts for

BOTH the structural descriptions AND the psycholinguistic data'.

As the goals of stratificational linguistics differ from those of

transformational generative grammar, so must the type of formalism

- 65 -

- C6 -




be going to

remainder of
the VP



be going to


- 67 -

- 68 -

it employs. Motivations (in addition to the general goal suggested

in Reich's last statement) for the stratificational formalism are

discussed in Lockwood (1973), Lamb (1973), and Reich (1972a), for

example. The most up-to-date overview of the notational system--and

the framework in general--is presented in Lockwood (1972, currently

in revision). For a brief explanation of the notational system, see

Sullivan (1978).1 Current views on the stratificational approach in

general are expressed in Sullivan (1980).2

The remainder of this chapter deals with the treatment of the

active-passive relationship in stratificational linguistics.

Reich (1970) dealt with the place of passive marking within the

system of English auxiliary verbs.

For example, consider the auxiliaries which occur in (5.1),

which 'consist of a complicated interlocking chain of discontinuous

elements' (Reich, 1970:25).3

(5.1) This problem has been being researched for too long.

I passive



concord (singular)

Newell (1966:82) proposed the portion of the (lexotactic) diagram

(5.2) for nonfinite verb phrases in English. Reich (1970:29) amended

the diagram to include finite verb phrases.

Here are some comments by Reich concerning the be of be ing


The be is the same be of the passive be en, and the en of
the passive is the same as the en of have en. What is
meant by this is that no matter whether the be came from be

- 69 -





- 7n -


- 71 -

en or from be ing, it will be realized in a particular way,
depending upon person, type I tense, and number (e.g., am,
is, are, etc.), but not depending on what construction it
came from. Similarly, en has many alternative realizations,
depending on what verb it is to be associated with, but not
depending upon whether it came from be en or have en.
(Reich, 1970:30)

For clarification on this point, see Sullivan's (1977b) 'A

stratificational view of the lexicon'. This article also contains

points relevant to the consideration of elements marking realizations

of deep case relations, e.g., the postverbal agentive marker by or

instrumental marker with. See especially p. 14 for discussion

relating directly to the agentive by.

Lockwood (1972) discussed the active-passive relationship in

terms of general clause structures.

Diagram (5.4) is adapted from Lockwood (p. 150). It shows the

predication structure (semotactics) for some participants accompany-

ing the sememe 5/chase/.5

Diagram (5.5), also adapted from Lockwood (pp. 151 and 153),

shows the relevant portion of the lexotactics providing the realiza-

tion of the sememes in (5.4).6 Note that available details and

generalizations given by Newell (1966) and Reich (1970) are left out

for the sake of clarity.

(5.6) *The window broken.

(5.7) *The window broke by John.

(5.8) *The window was by John.
(where John is Ag)

(5.9) *John was broken the window.

The lexotactics as described by Lockwood does not prohibit

ungrammatical sequences such as (5.6) (5.9), where John is Ag

- 72 -


4 be ed
to have en
get ing
go s

- 73 -

c c



- 74 -

(agent) and window Pa (Patient). Rather, such sequences will be blocked

by the interaction of the sememic and lexotactic strata. For example,

/en/ occurs in the VP only as part of the realization for /perfect/

(L//have/en//) or S/focus/ (L//be/en//).7 This will block (5.6), since
it has L/en/ without L/have/ or L/be/. Similarly, the postverbal

agentive marker L/by/ occurs only in the composite realization of

Pa + focus as L//f/be/en//. (See again diagram (5.5).) 'It is a

general principle in stratificational theory that restrictions dealt

with on a higher stratum do not need to be repeated on a lower one'

(Lockwood, 1972:151). In a still more general way, the principle can

be stated as (5.10):

(5.10) A restriction dealt with on one particular stratum
or between two particular strata need not be
repeated elsewhere.

In other words, a restriction should be stated (ideally) only in one

way, i.e., either as a syntagmatic constraint or as a realizational

rule, and only one time. A restriction expressed at some point in

the relational network thus can have a 'filtering' effect (in much

the sense of Chomsky and Lasnik's filters) on the inputs and outputs

of other strata, especially adjacent strata.

Application of the above principle (5.10) allows a great deal of

simplification relative to Reich's description of the VP.

The relevant portion of the lexotactics can be reduced to the

configuration shown in (5.11).

The semonic knot pattern which 'feeds' this portion of the

lexotactics will restrict the output of the lexotactics to only the

- 75 -








of the
verb phrase

- 76 -

the car

- 77 -

grammatical combinations of elements. It will also enforce the correct

order of elements within the complex realization of a single sememe.

For example, an ordered AND node in the semonic knot pattern enforces

the correct order L//be/en// within a passive verb phrase. Diagram

(5.12) shows a portion of the semonic know pattern. (The sememe

realized as L//be/go/ing/to// is arbitrarily designated S/BGIT/ for


Hierarchical ordering in the semotactics enforces the overall

ordering of tense, aspect, and focus within the verb phrase. Diagram

(5.13)--based on (5.2), but with the 'ordering' taken care of by the

semotactics--shows in general how this could be done, following Reich's

(1970) formulation. The present formulation seems incomplete, however.

Its inadequacies should be solvable in the semotactics, assuming appro-

priate semonic knot patterns to provide the correct realizations.
Turning now from the problems of verb phrase morphology, the

des- cription of overall clause stuctures will be considered. Sulli-

van (1976) modified Lockwood's description of the lexotactics, giving
the network shown in diagram (5.14) (p. 130). Diagram (5.14) shows

the lexotactics for (5.15) and (5.16).
(5.15) John drove the car.

(5.16) The car was driven by John.

Sullivan's description of the lexotactics of the active-passive

relationship--diagram (5.14)--diverges from Lockwood's (see again

(5.5)) in one major respect: 'the only syntactic alternation pro-

vided' in (5.14) is 'the realization of patient in subject position.

Everything else that occurs is a derivative of this. Since

- 78 -

c, .

- 79 -

stratificational theory presupposes a network of relationships, the

realization of patient in subject position is sufficient to trigger a

series of subordinate realizations' (Sullivan, 1976:127).10 If

/passive focus/ occurs, it causes the Pa to be realized in subject

position. The Ag, therefore, is realized in the next available slot,

i.e., in the first complement position (Cl). The postverbal agentive

marker L/by/ is provided automatically as a by-product. The object

marker /obj/--equivalent to Lockwood's L/m/--was treated as a

syntactic redundancy (which it is), not as part of the realization of

Sullivan (1976) further suggested that there are two possible

semotactic configurations for Ag-Pa predication. They are shown in

diagrams (5.17) and (5.18). The former shows Ag-dominant predica-

tion, as suggested by (5.20) and (5.21). On the basis of (5.22), I

would like to withhold consideration of (5.19) for the present, as

S/drive/ seems to be a class of verbs which require the presence of

an adverbial for a structure of this sort. The sememe S/drive/ can

now be considered together with such predicates as S/attack/ and

S/explain/. These suggest the Pa-dominant structure of (5.18).

(5.19) The car drives well.

(5.20) John is driving.

(5.21) John drives.

(5.22) *The car drives.

The same information is included in diagram (5.23). The semo-

tactic diagram (5.23) accounts for (5.19) and (5.20) (agent-predicate),

- 80 -





- 31 -



- 82 -




- 83 -

(5.24) (agent-predicate-patient), and (5.25) (agent-predicate-patient-

focus). It could also be modified to account for (5.26) (predicate-

patient-focus; where SN/focus/ is a determined element). See diagram

(5.24) John drove the car.

(5.25) The car was driven by John.

(5.26) The car was driven.

In (5.28) the possible combinations of Ag, Pa, and focus with

this class of predicates are summarized.
(5.28) (a) Ag-Pred

(b) Ag-Pred-Pa

(c) Ag-Pred-Pa-focus
(d) Pred-Pa-focus
Another class of predicates, S/fix, destroy, .../ does not permit

(5.28) (a). For example, (5.29) (a) and (b) are ungrammatical.
Diagram (5.30) shows the cooccurrence restrictions on Ag, Pa, and

focus with these predicates.
(5.29) (a) *John fixes.

(b) *John is destroying.
Finally, there is a third class of predicates that permit the

combinations shown in (5.31), S/break, tear, burn, .../. Diagram

(5.32) includes the necessary cooccurrence restrictions.

(5.31) (a) Ag-Pred-Pa

(b) Ag-Pred-Pa-focus
(c) Pred-Pa

(d) Pred-Pa-focus






- 84 -

- 85 -






- 86 -

Diagram (5.33) summarizes (5.27), (5.30), and (5.32). The possi-

bility of a null Ag without conditioned passive focus (AO) is

restricted by the diamond node above the predicate class S/break,

tear, burn, .../. The possibility of null Pa (P0) is restricted by

the diamond node dominating the class of predicates S/drive, attack,

explain, .../. The ordered OR nodes above the diamond nodes at A0

and P0 and their right-hand lines are the equivalent expansions of

the corresponding circled lines in (5.27), (5.30), and (5.32).

Diagram (5.33) leaves a great deal unexpressed with regard to

other types of passives, especially the so-called 'pseudopassives'.

This area is beyond the scope of this chapter. However, the descrip-

tion of other passive types in English is to be added to the semo-

tactics presented here, such a description will make use of the

realizational (upward) OR node at the right of diagram (5.33).

In this way, 'passivization' is presented as a unified phenomenon.
'Splitting the passive transformation into parts and incorporating

them into three complementizing transformations (Hasegawa, 1968) seems

to imply that the active-passive pairs are related only fortuitously,

that passivization is not a unified phenomenon' (Sullivan, 1976:119).

The same criticism is applicable to the work of more recent versions

of Chomskyan transformational generative grammar (EST and REST).

The relationship of (5.34) to (5.35) and (5.36) to (5.37) is a

related matter. Transformationalists have based some dubious claims

about the active-passive relationship on data such as this. 'It turns

out that in fact there is no consistent way to characterize the way

- 87 -

the passive... changes meaning, so there cannot be a projection rule

for the passive' (Jackendoff, 1972:9).

(5.34) Some man loves every woman.

(5.35) Every woman is loved by some man.

(5.36) Every man loves some woman.

(5.37) Some woman is loved by every man.

The matter should be more carefully considered. Sullivan

(1976:135) observed that the relation of (5.34) to (5.35) is one of

inclusion, as is the relation of (5.37) to (5.36). In other words,

there is a sense in which (5.34) and (5.37) are ambiguous. Sentences

(5.35) and (5.36), respectively, are very close paraphrases of one

meaning of each. An expanded paraphrase of (5.35) might be For every

woman, there is some man who loves her. More than one man may be

involved, but as the lack of grammatical plural indicates, no more

than one man for each woman. This does not require a one-to-one

mapping of men to women, but could involve a mapping like: woman1-

man1, woman2-manl, woman3-man2, woman4-man3, ... womann-manm. A

similar situation exists for (5.36). For every man, there is some

woman that he loves is an expanded paraphrase of (5.36). Again, more

than one woman may be involved, and the mapping need not be one-to-

While the interpretations of (5.35) and (5.36) require this 'some/

every-mapping', those of (5.34) and (5.37) do not. In (5.34) and

(5.37), the interpretation of some may involve such mapping, but need


- 88 -




all some

(5.43) Ms ome
M Vsome

- 89 -

Now consider (5.38) (5.41). In each, some indicates one indefi-

nite man/woman is involved.

(5.38) Some man loves Ann.

(5.39) Ann is loved by some man.

(5.40) John loves some woman.

(5.41) Some woman is loved by John.
In general, then this some indicates one indefinite member of a

class. However, every can condition some/every-mapping. This mapping

is required in the interpretation of postverbal some if there is an

every in the subject. The quantifier all seems to exhibit a similar

effect on the interpretation of postverbal some when substituted (for

every) into sentences (5.34) (5.37). The quantifier each requires
this mapping in all environments. Thus, I will call this phenomenon

some/V-mapping ('V indicating universal quantification).

Diagram (5.42) shows a small portion of the semotactics relevant

to the description. The S/M/ is the sememe that controls some/V-


The interpretation of L/some/ may include S/M/. This is shown

in diagram (5.43).

Diagram (5.44) shows a portion of the lexotactics. Here, the

occurrence of a quantifier including V in subject position can condi-

tion some /V-mapping in Cl. The realizational lines marked 'a' and

'b' connect with those similarly labelled in (5.43).

(5.45) *S//(Ag/every/man)/love/(Pa/some/woman)//

(5.46) S//(Ag/every/M/man)/love/(Pa/some/woman)//

(5.47) S//(Ag/every/man)/love/(Pa/some/woman)/focus//

- 90 -




- 91 -

The phenomenon of some/V-mapping interacts only indirectly with

the active-passive dichotomy. For example, the semotactics and lexo-

tactics will work together to block (5.45), which lacks the appropriate

some-V mapping.11 In this case, without passive focus, the Pa would

be realized in Cl. But the lexotactics requires L/M/ with L/some/

under C1 because the realization of S/every/

is in subject position. Both (5.46) and (5.47) can be realized gram-

matically. The former contains some/V-mapping; L/some/ is realizable

in Cl. The latter contains S/focus/; this realizes the Pa in subject

position, where the L/M/ is optional.

Recall Jackendoff's (1972:9) assertion--quoted earlier--that

'there is no consistent way to characterize the way passive... changes

meaning, so there cannot be a projection rule for the passive.' Now,

there is no way this assertion can be faulted, given what Jackendoff


However, it contains one assumption that need not be taken as a

'given' of the problem set. Jackendoff assumed a priori that in order

for passive to have a rule of semantic interpretation, it must change

meaning in a consistent way. It is this assumption which I challenge.

The assumption is a necessary one only if rules of semantic interpreta-

tion are required to take as operands (a) ordered strings of elements

(i.e., the nodes of deep structures) or (b) rules which rearrange

the elements of those strings (into other ordered strings).

Stratificational theory relates the 'meaning' and 'surface struc-

ture' of the English sentences (5.34) (5.37) through sememic-to-

lexemic realizational rules, which combine with the 'filtering'


University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs