Linguistic Society of America

Spatial Networks as a Site for the Study of Language and Thought

Author(s): Charlotte Linde and William Labov
Source: Language, Vol. 51, No. 4 (Dec., 1975), pp. 924-939
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/412701 .
Accessed: 29/03/2011 02:59

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=lsa. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org
 SPATIAL NETWORKS AS A SITE FOR THE STUDY OF LANGUAGE
AND THOUGHT
CHARLOTTE
LINDE and WILLIAMLABOV
Hunter College, CUNY University of Pennsylvania
A techniquehas been developedfor observingthe translationof cognitive input into
language in a spontaneous, practiced speech event: descriptions of the lay-outs of
apartments.
The greatmajorityof such lay-outs are imaginarytours whichtransformspatial lay-
outs into temporally organized narratives. The set of discourse rules which govern
these speech events is given, and the rules which determinea number of features of
sentence grammar are outlined. This study represents an initial description of the
links between cognitive input, discourse rules, and the rules of sentence grammar.*
Some of the most complex problems that linguists face today arise from the
relations of such simple sentences as
(1) There's a kitchen and a dining room to the left.
(2) There's a kitchen with a dining room to the left.
Do these sentences mean the same thing ? Many linguists feel that the two sentences
can be derived from a single underlying form by a transformation which does not
affect meaning:1 this would then be another case of the massive free variation which
is characteristic of a grammar containing many unconstrained optional processes.
In this paper we will show that such apparent free variation can be constrained
by principles of considerable linguistic significance-once we move beyond sentence
grammar to the study of discourse, and move beyond intuitional semantics to a
descriptive semantics based on observation and natural experimentation.
Sentences 3 and 4 show another case of related sentences, normally not considered
a problem at all:
(3) And then you come to a kitchen with a dining room to the left.
(4) And then you come to a kitchen with a closet to the left.
Given a sentence of a certain structural type, we can test the stability of our judg-
ments of well-formedness by substituting other equivalent members of the word
classes involved. Thus 3 and 4 seem to us equally well-formed. Nevertheless,
though diningroom and closet share grammatical and semantic features, one cannot
be substituted freely for the other in this construction. In fact, we will show that 3
is not well-formed in one important sense.
* A preliminaryversion of this paperwas presentedat the LSA SummerMeeting,Ann Arbor,
1973. The data and much of the analysis draw upon Linde 1974.
1 This point of view was initiated by Gleitman 1965, and was developed further in Lakoff
& Peters 1966 and in Lee 1967. On the other hand, Dougherty 1971 objects by arguing that
A and B mix with C cannot be derived from A and C mix and B and C mix. His argument is
based upon the general claim that conjunct reduction results from an operation on binary
coordinates; even if this is granted,it would remainto be shown that the differencebetweenthese
complex forms is due to a differencein the meaning of withand and. We might also point to the
difference between The play was written by John Smith and George Jones and The play was written
by JohnSmithwithGeorgeJones. The differencein the attributionof authorshipdoes not auto-
matically generalizeto a differencein meaning for all sentences containing with and and.
924
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 925

It is now generally agreed that research on the relationships of sentences cannot

proceed without deeper insight into semantics. We share the general view that
language is a method of relating meaning to sound for the purposes of social
communication; and we would further argue that a substantive theory must be
informed by empiricalstudies of phonetics, social variation, and semantic structures.
Some progress has been made in replacing impressionistic data on phonetics and
the social use of language with empirical evidence; but the third area of substance,
semantics, is still limited to speculative and philosophical study. The descriptive
semantics called for by Weinreich, which we need as a basis for sound theoretical
work, has not yet been developed.

1. It has been difficult to evaluate or build on the many controlled laboratory

studies which have attempted to focus on the relationships of language and thought.
They typically test the limits of production and perception, rather than study
central linguistic skills. They may introduce new and unpracticed tasks whose
purpose is socially uninterpretable.2They may focus on testing skills that are well-
developed only for a small part of the population.3 For some time, we have been
searching for a site for experimental study which is free from these defects, which
would allow one to observe the encoding of thought into language as directly
as possible in a setting where the ordinary, unconstrained, and spontaneous flow
of language can emerge.
We have located such a site in the course of a sociolinguistic study of attitudes
toward urban life.4 Sentences 1 and 4 were produced by New York City residents
in answering a' question which was found to trigger an unexpectedly well-formed
speech act. This act emerged as a remarkably useful site for the study of the trans-
lation of thought into language; it is an answer to a very ordinary question:
Could you tell me the lay-out of your apartment?5
No one answering this question felt that he was the subject of a psycholinguistic
experiment. But as we examined the first set of answers, we found that this speech
act has the fdllowing properties of a paradigmatic 'natural experiment':
(a) It is GENERAL: everyone who has lived in an apartment for more than a few
days has an equally good knowledge of his apartment's lay-out.
(b) It is WELL-MOTIVATED:people are extremely interested in talking about their
apartments.

2
In some of the most influentialparadigms,subjectshave been asked to listen to dichotically
presentedsentences and clicks and to locate the clicks in the sentences (Fodor & Bever 1965),
to memorize a sentence plus as many unrelatedwords as possible (Savin & Perchonok 1965),
and to monitor a sentence for the first occurrenceof a given phoneme (Foss &,Lynch 1969).
3 For example, Gleitman & Gleitman 1970 have shown that the ability to paraphrasecom-
pound noun phrases is related to the amount of higher education of the subject.
4 This researchstrategywas first developedin a course on the study of the speech community,
as a means for eliciting complex syntactic forms in spontaneous speech. When it appeared
that the speech event of providing an apartmentlay-out was surprisinglywell-structuredand
uniform, Linde was able to develop this technique as a means of analysing the way in which
discourse constraints determinesentence formation.
5 It should be noted that couldand can are social variantsin colloquial New York City speech.
926 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

(c) It is CONTROLLED: we can observe different people coping with the problem
of converting the same spatial network into discourse.6
(d) It is OBJECTIVE:
we have almost direct access to the semantic input,7 and can
study it through another modality-the subject's drawing of the apartment.
(e) It occurs in a CONVENTIONALSOCIALSETTING, rather than in an experimental
framework, and corresponds closely to an activity of an ordinary conversation.
(f) It elicits a WELL-PRACTICED linguistic skill, a process of linguistic encoding
in which speakers follow an intricate set of rules governing their syntactic choice.
This is the crucial property which makes this event useful to linguists. A consider-
able body of evidence shows that the systematic character of language emerges
most clearly when speakers are dealing with familiar topics, following intricate
but well-practiced routines.8 But when speakers begin to wrestle with new notions
and unfamiliar tasks, the high degree of grammaticality of everyday speech can be
greatly reduced.9
2. Let us now examine the material and see how these properties emerge. Ex.
5 is typical of the form in which most lay-outs are cast:
(5) You walk into a long, narrow foyer,10
leading into a smaller, squarer foyer, eating place, dinette-area.
And-uh-to the right is my kitchen,
and-uh-to the left, my living room ...
Subjects begin at the outside entrance, and proceed to name each room, along with
instructions on how to reach that room. In 5 we observe one of the most character-
6
This can be done by obtaining lay-outs from several people who live in the same apartment
or from those who live in identical apartmentson differentfloors of the same building. Both
strategies have been utilized in this research.
7 The access is not entirely direct, since both the speakers' verbal representations and the
maps which they draw of their apartments are not completely accurate, and contain some omis-
sions and distortions. This is a phenomenon which has been found in other investigations of
mental images. Lynch 1960, speaking of the mental maps which city-dwellers form, states:
'The image itself was not a precise miniaturizedmodel of reality, reduced in scale and con-
sistently abstracted. As a purposive simplification, it was made by reducing, eliminating, or
even adding elements to reality, by fusion and distortion, by relating and restructuringthe
parts ... However distorted, there was a strong element of topological invariancewith respect
to reality.It was as if the map were drawn on an infinitelyflexiblerubbersheet; directionswere
twisted, distances stretched or compressed, large forms so changed from their accurate scale
projection as to be at first unrecognizable.But the sequence was usually correct, the map was
rarely torn and sewn back together in another order.'
Similarly, Downs & Stea 1973 report that people's mental images of spatial configurations
on a variety of scales of size are schematizationsrather than exact replicationsof the physical
reality.
8 As reportedin Labov 1966, about 75% of utterancesin most conversationsare well-formed

by any criterion (when rules of ellipsis and general editing rules are applied, almost 98%
would fall in this category). When non-academic speakers are talking about their everyday
experience,this proportion approaches90%7.
9 This seems to be particularlycharacteristicof transcriptionsof learned conferences (such
as those studied in Maclay & Osgood 1959), and may be responsiblefor the common opinion
that spoken language provides only degeneratedata.
10Foyer is a term common in New York City, referringto any entrance hall which is wider
than a corridor.
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 927

istic structures of lay-out sentences: locative, copula, noun phrase: To the right
is my kitchen.
Another form of this basic structure is 6, in which each room is identified with an
independent clause.1l The spatial representation is transformed into a temporal
sequence-a pseudo-narrative:12
(6) You walked in the front door.
There was a narrow hallway.
To the left, the first door you came to was a tiny bedroom.
Then there was a kitchen,
and then bathroom,
and then the main room was in the back, living room, I guess.
As we examine these apartment descriptions, we immediately ask: 'How good
are they?' How well are these apartment-dwellers representing the spatial con-
figurations in which they live ? Do they have the linguistic competence to transmit
this knowledge to others ? As we attempt to draw the apartments, we begin to have
doubts. Some apartments are very complex, so that we wouldn't be surprised
that a speaker might fail. Such a case is the following:
(7) The main entrance opens into a medium-sized foyer.
Leading off the foyer is an archway to the living room which is the fur-
thermost west room in the apartment.
It's connected to a large dining room through double sliding doors.
The dining room also connects with the foyer and main hall through two
small arches.
The rest of the rooms in the apartment all lead off this main hall which
runs in an east-west direction.
The first room to the right is a kitchen.
Then the first room to the left is a large child's bedroom.
Next, to the left, is a short hall at the end of which is a full bath,
and leading off of which to the right is a large study or office.
Continuing down the main hall to the right is a door which leads into a
suite of three small rooms, two small bedrooms connected by a short
hall, off which is a laundry room.
At the end of the main hall are two large bedrooms, one to the left and
the other at the end proper.
To the right at the end of the hall is a second full bath. That's it.
Any attempt to draw this apartment will illustrate the great difficulties in recon-
structing a conventional floor plan from the verbal description. An effort to draw
the similar example 6 will illustrate just what kinds of information are present in
the description. If the reader attempts this before turning to Figure 1, it will be
11But see below for modificationsof this statement in regard to marginal spaces like bath-
rooms and closets.
12 In Labov & Waletzky 1967 and Labov 1972 (Ch. 9), a narrativeis defined as a sequence
of preterit clauses in which the order of the clauses follows the order of the actual events.
The lay-outs which appear in this form may be referred to as PSEUDO-NARRATIVES,since no
such event has actually occurred; they are hypotheticalconversions of a spatial configuration.
The term 'pseudo-narrative'is from Wald 1973.
 That one was also a one-bedroom
apartment. It was a brand new
building. We were the first
occupants in that apartment.

You walked in the front door.

There was a narrow hallway.

- ^_^ W_~~I |To the left, the first door

_^^~~~~~ ~you came to was a tiny bedroom.

f r Then there was a kitchen,

/"3 fand
K then bathroom,

and then the main room was in

the back, living room, I guess.

Lst \
FIGURE1. Floor plan of the apartmentdescribedin Example 6
as drawn by the tenant.
928
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 9929

possible to compare the re-translation into spatial terms with the actual floor plan
as drawn by the speaker of 6. A number of uncontrolled features then appear:
the air shaft between the kitchen and the bathroom cannot be reconstructed from
the account, nor most of the details of the room shapes or the way they fit together.
Let us examine another example which seems more precise:
(8) I'd say it's laid out in a huge square pattern, broken down into four
units.
If you were looking down at this apartment from a height, it would be
like-like I said before, a huge square with two lines drawn through
the center to make like four smaller squares.
Now, on the ends-uh-in the two boxes facing out in the street you have
the living room and a bedroom.
In between these two boxes you have a bathroom.
Now, between the next two boxes, facing on the courtyard, you have a
small foyer and then the two boxes, one of which is a bedroom and
the other of which is a kitchen and then a small foyer-ah-a little
beyond that.
We will call this type of layout a MAP,which can be contrasted with two other types.
One is an existential construction which predominates in 5-7, and is fully exempli-
fied as follows:
(9) The entrance is into the kitchen.
And through an archway is the living room,
and then off through a separate doorway, a closing doorway, is the
bedroom.
I'd say the bedroom is about just a little bit smaller than the living room.
I think it's a pretty good size for two people.
The third basic type uses verbs of motion, as the speaker conducts the hearer on
an imaginary tour:
(10) As you open the door, you are in a small five-by-five room which is a
small closet.
When you get past there, you're in what we call the foyer which is about
a twelve-by-twelve room which has a telephone and desk.
If you keep walking in that same direction, you're confronted by two
rooms in front of you ... large living room which is about twelve by
twenty on the left side.
And on the right side, straight ahead of you again, is a dining room which
is not too big.
And even further ahead of the dining room is a kitchen which has a
window in it.
And the back, the farthest point of the kitchen, is at the same depth as
the farthest point of the living room.
In other words, the dinette and the kitchen are the same length as the
living room.
Now, if you turn right before you went into the dinette or the living room,
930 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

you, you would see a bedroom which is the small bedroom going into-
going in on the right ...
And if you kept walking straight ahead, directly ahead of you, you would
find a bathroom.
And on your left you would find the master bedroom, which is a very
large bedroom,
and there are closets all around.
Of these three types, we might ask: which is most common? which would we use
ourselves ? Most of the linguists and psychologists to whom this question has been
presented respond as we did ourselves: we might use any of these. Like many
intuitive reactions, these judgments turn out to be incorrect, and the actual facts
are quite surprising.

3. There are only two types of apartment descriptions, not three; analysis of
5, 6, 7, 9, and 10 shows that these all have the same structural organization, which
may be called the TOUR.The MAPtype of 8 is a rare exception, representing only
3% of all the lay-outs gathered. Not only is the tour the predominant type, but
it also appears to be strikingly well-formed, once we grasp the nature of the in-
formation the speaker gives us.
If the aim of the lay-outs were to instruct us how to make drawings of the outer
shape of the apartment and the configuration of its rooms, then we would have to
say that almost every speaker has failed. But as we examine the actual structure
of the discourse which speakers use in the tours, we find that each speaker has
succeeded in showing us something else: how to enter each of his rooms. More
precisely, a tour is a speech act WHICH PROVIDESA MINIMALSET OF PATHS BY WHICH
EACH ROOM COULD BE ENTERED. The path is presentedas a series of units of the
form DIRECTION (or VECTOR) and ROOM. There are two basic kinds of vectors:
the static type (to the right, straight ahead of you, off of the X) and the mobile type
(you keep walking straight ahead, now if you turn right).
The imaginary tour does not always take the listener into each room. In 6,
the listener walks in the front door and is in the hallway. He comes to the first
door, but he doesn't enter it. He passes the kitchen and the bathroom, sees how to
enter them but does not do so, and finishes in the living room. He may be conducted
into rooms, as in 10: Whenyou get past there, you're in what we call thefoyer. But
static vectors can also take him into rooms-again, from 10: and straight ahead .
is a dining room ...; and evenfurther ahead ... is a kitchen ...; now if you turn right
before you went into the dinette ...
The task of the speaker, then, is to traverse for this imaginary visitor the spatial
network formed by these vectors and rooms. The questions arise: are there any gen-
eral principles which govern where the traversal is to start? what forks are taken in
what order? when are rooms entered, and when are they not? Detailed analysis of
the 72 lay-outs has revealed a striking series of regular principles deciding these
questions (Linde 1974:55-114):
(a) The imaginary tour begins at the front door of the apartment.
(b) If the visitor comes to a one-room branch, he does not enter it.
(c) If he comes to a branch with rooms beyond the first room, he always enters.
(d) When he reaches the end of a branch, and there are other branches to be
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 931

traversed, he does not turn around and go back; instead he is moved back in-
stantaneously to the fork point where the other branches originate.
These principles reflect a formal organization of the speech act which is no less
complex and regular than other sets of formal linguistic rules. First, we may want
to characterize those apartment features which are the input to the description.
The path to be traversed is a branching network, consisting of an entrance point,
markers of directionality-the vectors-and rooms to be identified.
4. Before we can formally characterizethe traversal of a network, we need some
mechanism for displaying that network. There are obviously an unlimited number
of possible apartment forms. Phrase-structure rules are convenient tools for de-
scribing branching structures; we may use them here to enumerate all possible
networks, and to generate samples for display. We do not, of course, suggest that
such phrase structures for apartments are needed in a grammar; the recognition
of these hierarchic arrangements must be a part of the speaker's non-linguistic
knowledge as well.
The networks can be included in the subsets of the total branching process of
Rules A and B:13
RULE A: L- (V) E B
RULEB: B-V R BO
There is, of course, no suggestion here that we have a 'grammar of apartments'.
These rules are simply devices which permit one to construct any number of apart-
ment lay-outs of unlimited degrees of complexity; Rules 1-12 below for describing
apartment lay-outs can be tested against any such construction. The lay-outs are
not sets of square boxes attached to each other, but routes showing how access
is obtained to rooms. If we symbolize a vector as an arrow which generally points
left, right, or straight ahead, we can sketch some members of this unlimited set.
Corresponding to the apartment of ex. 6, we have Figure 2.
R2 RA R6
t t t
E-R R3R5-7 -* -*

FIGURE2. (R1=R3=R5).

Before the information represented by such a structure can be translated into

language, it must be re-organized in a linear form. In his imaginary traversal of
the apartment, the speaker must mention one room at a time, in an order which
systematically includes each room at least once. The subscripts indicate this order.
Again, the network corresponding to the more intricate lay-out of ex. 7 is
Figure 3 (p. 932).
The symbol R is a superordinate term which includes halls, hallways, foyers,
corridors, and passageways, as well as rooms proper, such as living rooms, kitchens,
bedrooms etc. Halls are rooms which lack any of the built-in fixtures or specifica-
tions of function which identify rooms proper. A necessary condition on the rules
permits the generation of only those halls which terminate in further branches;
this corresponds to the architectural fact that halls do not lead nowhere. Where
13 L = lay-out; V = vector; B = branch; E = entrance; R = room. Subscripts indicate
the minimum number of elements which can be present, and superscriptsthe maximum. Thus
B? indicates 'from 1 to n branches'.
932 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

R2 -> R3 R7 R R11 RI18

t /4 t t t t
E -> R- R>R - - -> R12 R1? RI1
4, 4, 4,
R5 R13 R20
4,
R16 - R14

Rl5
FIGURE3. (R4= Re =R8R2=R17).
R =

several halls are connected in a straight line and the branches from them are oblique,
the connected segments will be identified as a single hall. In Fig. 2, the units R1,
R3, and R5 are identified in this way. In Fig. 3, there is an unusually long hallway,
R4 =R=Rg=R12 =R17 The identity of these segments is seen in anaphoric
reference:14 R4 = main hall ... this main hall; R12 = the main hall; R17 = the
main hall ... the hall.
The two networks of Figs. 2-3 show the regular order of traversal which governs
the description produced by all speakers. The first five rules which account for or
generate the lay-outs themselves are as follows:
RULE 1. From the outside, go into the first room.
RULE 2. In any room, select a branch with no branch.
RULE 3. If there is none, select a branch and enter it.
RULE 4. If there is none, return to the last room with an unselected branch.
RULE 5. If there is none, terminate.
Rule 1 is not as obvious as it might appear. We know the apartments we live
in from all directions, and can find any one room from any other room. The
selection of the outside entrance as the starting point is a general strategy which
sharply reduces the number of options. This decision also permits the use of the
relative directions left and right (which are probably not characteristicof the stored
input),15 rather than compass directions.
Rule 2 gives first preference to one-room branches. This is a very powerful
principle, exemplified in 86 cases, with no counter-examples. The rationale for
this important generalization will be given below.
Rule 3 introduces the term ENTER: the first room of the complex branch must be
entered in the imaginary traversal so that the visitor can see how to enter the
succeeding branches. This is true even when the branches are arranged in a straight
line, so that one could see several rooms at once. That is, the tour is designed to
show HOWto enter rooms: one enters a room to see how to enter other rooms.
The contrast between entering a room and not entering it can be seen below:
14A numberof interestingdevices allow the speakerto indicate that he is still in a segment of
the hall without naming it again. In fact, named cases like R12and R17prove to be the marked
case, and are provided for by a more general rule which applies to rooms as well as halls-
returningfrom complex branches(see below).
15It is possible that structures which are always approached from the same direction are
stored directlyin terms of left and right: e.g., the dashboardof an automobile may be remem-
bered using relativedirections,since it is always orientedin the same way. But as we haveshown,
apartment lay-outs are not like this, since the speakercan traversethem from any orientation.
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 933

(11) ENTRY: You come smack into this nice long living room that
ends in a window.
(12) NON-ENTRY:a. ... and THEKITCHENWASOFFTO THELEFTOFTHAT
and you kept on going, just a short little hallway,
into the living room.
IN FRONTOF YOUWASA LIVINGROOM,
b. ... and DIRECTLY
and you took a right down this long narrow hallway,
and on the right was the kitchen.
Rule 4 is brought into play only when we reach the end of a complex branch,
but areas of the apartment remain un-entered. This case deals with the intriguing
and unpredictable problem of returning to some prior choice point. In a real-life
tour of an apartment, one would reach the end of a complex branch, turn around,
and proceed back; the vectors of left and right would then be reversed. But this
reversalnever takes place in an apartmentlay-out description.16Instead, the speaker
returns instantly to a choice point which is lexically marked:
(13) Oh, going back to the hall, as you come into the hall, the kitchen is to
the left.
(14) So in other words you went to your left through a hall,
and you made a right and into a large bedroom.
Now as you're coming into the front of the apartment, if you go straight
rather than go right or left you come into a large living room area.
Rule 5 exemplifies the same principle in a more general and obvious way. In
an actual tour, one would have to be taken back to the entrance, but the tour is
terminated at the last room.
The rule-governed character of this process is clear, and we will see how it
constrains the distribution of information-and, consequently, syntactic choice. To
observe its connection with sentence grammar,we first add four rules which further
specify vectors and apartments. The first two are entirely determined by the physical
configuration of the apartment. None of the higher-level strategies for traversing a
network (Rules 1-5) would require us to distinguish in the topography the features
[ ? oblique]-straight ahead vs. left or right-or even to distinguish left from right;
but the linguistic forms available are organized in ways that make these distinctions:
RULE 6. V -*[ ? oblique]
RULE 7. [+ oblique] -({iht})

The features left and right are realized in this discourse as morphemes with that
shape;17 but if Rule 7 is not selected, the speaker can realize [+oblique] with a
morpheme such as off which is neutral to the left-right distinction.
Rule 8 expresses a distinction which is primarilyin the realm of social categoriza-
tion, and only partly determined by the physical facts:
RULE 8. R- [? major]
16
In some cases, the rooms on one side of a hallway have not all been described when the
other side is completed. The speaker may then refer to the remainingside with the term back,
indicating that he has not reversedhis own orientation.
17 In other contexts, however, they may be realized differently, e.g. port vs. starboardor
windwardvs. leeward.
934 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

The distinction between major and minor rooms will be crucial in the rules which
link the discourse strategies with sentence formation. We are necessarily involved
in a taxonomy of differentiated spaces. Major rooms are kitchen, dining roam,
living room, and bedroom. Minor rooms are all others: study, den, laundry room
etc. Halls are rooms of a special type; some of their peculiar properties have been
indicated above. Bathrooms have a marginal status; unlike other minor rooms,
they need not be included in the traversal of the apartment. Otherwise, all rooms
must be specified in the lay-out description. Closets are rarely mentioned and never
entered; though these are demarcated spaces, they are not categorized as rooms in
apartment lay-outs.
Rule 9 registers a distinction which is entirely a matter of linguistic expression:
the way in which the direction is encoded in linguistic form. At first glance, a speaker
seems to have free choice of static vs. mobile vectors, as exemplified above. But
Linde's analysis of this choice (1974:132-44) shows that it is in fact highly con-
strained. It is always possible to choose a static vector such as On the left is ....
Straight ahead is ... But the choice of a mobile vector is limited to three specific
contexts: at the entrance which begins the tour, on returning from a long branch
by Rule 4, and immediately after another mobile vector on the same branch:
f-_E 1
RULE 9. V -* ([+ mobile]) / g(VR)2 _
[+ mobile] R __J
-> [-mobile]
The most general situation described by this rule is that a speaker begins with
a mobile vector (You walk in the door and ...); then he continues with mobile
vectors until he optionally switches to a static vector. He never reverts to the
mobile vector unless he returns from a long branch (Now if you turn right before
you went into the dinette ...)
Rule 9 has considerable significance for the relationships between discourse
rules and sentence grammar. It has the general form typical of many discourse
rules: one set of options is possible under a specific set of circumstances, but is
not required under any of them.18 The other option is possible under all circum-
stances, including those covered under the first sub-rule.19
In most discourse rules, we find a detailed set of social conditions for interpreting
specific linguistic forms; but the number of such forms which could realize any
given speech act seems to be large and unlimited.20Therefore the articulation be-
tween discourse rules and the grammar of English remains at present in a condition

18
The decision to switch to a static vector on a single branch appearsto be unmotivatedby
any factor which we have yet been able to discover. Some speakers continue with mobile
vectors to the end; others switch to static vectors, and then continue.
19Another such example is the rule governingthe choice of the map or the tour strategy.The
relatively rare map strategyis possible only when the apartmentto be describedhas a nameable
outer shape; but even in this situation, the tour strategyis possible, and is in fact preferred.
20 For example, rules for promising, such as those set forth by Searle 1969. Searle gives many

conditions which allow us to recognize an utterance as a valid promise; however, except for
the actual use of performativeverbs in a performativecontext, there is no route from the speech
act to the choice of a specific linguistic form.
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 935

best described as vague and anecdotal. Rule 9 seems to us a promising advance

toward forming the necessary connections. Many unconstrained options still
remain for expressing directions in the apartment descriptions. But a major cate-
gorical feature, [ ? mobile], is here seen to be well governed and motivated, so that
we begin to see the decision procedure which the speaker seems to follow in choosing
a linguistic form.
As we approach the details of linguistic specification, we see that Rules 8-9
set up conditions for a number of important syntactic choices; e.g., the distinction
between major and minor rooms will eventually have the following effects:
(a) Major rooms may be introduced with definite articles, minor rooms are not.
(b) Major rooms may be introduced as subject noun phrases, minor rooms only
in complements.
But perhaps the most far-ranging linguistic decision for sentence grammar is
the implantation of the sentence boundaries themselves. This process is not con-
sidered or required in most grammars, which begin and end with the boundaries
of the sentence. In those recent studies of anaphoric referenceand other grammatical
phenomena which go beyond the individual sentence (Dressler 1970, Grimes 1971,
Gunter 1966), sentence boundaries are also treated as given. The propositional
content of these discourses is usually provided in advance, and thus we obtain only
limited insight into the process which transforms cognition into propositions and
propositions into sentences.
The apartment lay-outs are transformations of an input which is obviously not
propositional: it is a state of the world which people encounter, and integrate
into their cognitive structure. They have available a well-rehearsed linguistic
strategy for describing this structure, and the propositions seem to be formed as a
part of this strategy. There is no reason to think that they store permanently the
divisions of the network which correspond to the sentence boundaries of the lay-
out descriptions. We thus have a fortunate opportunity to observe empirically
the principles which determine the division of information into sentences.
The following two rules schematize constraints on the process of implanting
sentence boundaries, as developed by Linde in her investigation of apartment
lay-outs:

RULE10. 0 / (V[m r }

[ +majorJ

RULE11. 0 { rV J
(7)I aJor]

[ [+mobile] --)
Rule 10 combines three obligatory environments. First, it uses the neighborhood
convention to show boundaries before and after L, i.e. at the beginning and the end
of the description. Second, sentence boundaries are placed at both ends of a long
branch, i.e. around two or more VR units. Third, sentence boundaries are always
placed before a branch with a major room.
936 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

The simplest form of the third condition is exemplified by the majority of the
sentences: from 6, Then there was a kitchen; from 10, And on the right side, straight
ahead of you again, is a dining room which is not too big. On the other hand, minor
rooms are often introduced in relative clauses or prepositional phrases: with a
little dressingroom off that; with a foyer at the front; with a bathroomon the right;
off which is a laundryroom.
Rule 11 shows two cases where sentence boundaries are implanted optionally.
The first is the converse of the last condition in Rule 10. Though minor rooms can
be incorporated within a sentence without being the main constituents, they can
also be identified in separate sentences: e.g., from 6, You walked in thefront door.
There was a narrowhallway. The second condition of Rule 11 allows the possibility
of separating a mobile vector from its associated room. Typically we have: You
walk in the door and there's a kitchen.
These rules do not account for all the sentence boundaries in an apartment
lay-out, since considerable descriptive information is often added to the basic
framework; e.g.,
(15) The kitchen was off to the left,
and it was a decent-sized kitchen.
It wasn't fantastic,
but it was THERE, you know.
Closets are not rooms in the sense defined here, but occasionally a full sentence
is devoted to them; e.g., from 10, And there are closets all around.Thus Rules 10-11
have not exhausted the options that are open for the implantation of sentence
boundaries: they state the major constraints for the basic structural information
of the apartment lay-outs.
The following lay-out illustrates the problem of conjoining and enumeration:
(16) It's got a long hall about thirty feet long ...
And then the hall turns left, and goes for maybe twenty feet,
and then runs into a kitchen ... pink.
There are three rooms off to the right of the second hall,
umm ... the first one of which-well, the door faces, like the long hall,
the first hall ... [interruption]
Okay. The first one is little, and the door faces the long hall as well as
the little hall ... that is to say, the intersection of the two halls.
Under certain conditions, we can conjoin major rooms: in Linde's analysis,
the vectors must be identical, and the conjoined rooms must each be one-room
branches. Such sentences as 17a can then be transformed into the quantified 17b;
and the order of the constituents can be reversed, with there-insertion, to put the
simpler constituent first:21
(17) a. Off to the right of the second hall is my bedroom and Steve's bedroom
and Annie's bedroom.
21
We are not suggestingthat the sentencesare formed in this way; this is one of many general
strategies which ensure right-branchingsurface structuresrather than left-branching,and this
canonical form may dictate the process of sentence composition as it actually occurs.
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 937

b. Off to the right of the second hall are three bedrooms.

c. There are three rooms off to the right of the second hall.
In these networks, both options are governed by the following rule, which pro-
duces conjunction or quantification:
RULE 12. # - (&) / (VIR) (V,R)
A sentence boundary is optionally replaced by & between two short sister branches.
This & will conjoin rooms and appears as and; in some cases, enumeration takes
place, and the phrase is converted by a later rule to a quantified noun phrase.
5. We are now ready to approach the problem of the well-formedness of ex-
amples 1-2. Ex. 1 is the result of the operation of Rule 10, which plants sentence
boundaries between two major rooms, and of Rule 12, which replaces the boundary
by &. On the other hand, ex. 2 is a type which never occurs in our data. It is not
generated by the rules we have written: the implantation of sentence boundaries
before dining room is obligatory by Rule 10, since dining room is a major room,
and the with-phraseis a non-sentential adjunct to a noun phrase, an option avail-
able only when sentence boundaries do not intervene.
Ex. 3 is not well-formed, and no such types occur in our data. Ex. 4 is well-
formed, since closets are not rooms, and are not recognized by these rules for
implanting sentence boundaries or by any other rule we have stated. Though later
optional rules can create separate sentences for closets (as at the end of lay-out
10), this option of introducing closets by a with-phrase is normal and typical:
one-third of the mentions of closets are linguistically realized in this form.
The same considerations provide for the well-formedness of the following:
(18) On the left side of the hallway is a bedroom with a closet off of it.
(19) On the left side of the hallway is a bedroom with a study off of it.
Here 18 shows the normal way of handling closets, and 19 is well-formed as a
minor possibility. Note that Rule 11 states that implantation of sentence bound-
aries before non-major rooms is optional, not obligatory. The great majority of
minor rooms are given separate sentences. But a few of them, particularly small
ones, are retained in the same sentence with a preceding room, and in this case the
relationship is linguistically realized by a with-phrase.The following five cases show
all the instances of this minor rule in the 72 lay-outs studied by Linde:
(20) And there's another little bathroom WITH A LITTLEDRESSINGROOM
OFF THAT.
(21) And that's the three bedrooms WITHTHEBATHROOM.
(22) And then off the living room was a stairway up WITH A SMALLHALL,
nothing, no use, it wasn't functioning.
(23) I entered a large living room WITHTHEKITCHENoff the back.
(24) You walked to a bedroom and like A LITTLEHALLWAYwith a ... and a
little bathroom coming off of the hallway.
Sentences 20-22 are straightforward cases of non-major rooms adjoined by with-
phrases, but 23 is an apparent counter-example. This seemed to us at first to be a
serious challenge to the force of our distinction between the obligatory and optional
placing of boundaries. But in the subject's drawing of his apartment, we then saw
938 LANGUAGE, VOLUME 51, NUMBER 4 (1975)

that the 'kitchen' was not a room at all, but merely a row of kitchen appliances at
the end of the living room.
Finally, in 24, we see a revealing self-correction of the type which we study
carefully in phonology and sentence grammar. It would have been possible to
complete 24 as follows:
(25) and like a little hallway WITH a little bathroom coming off of it.
But the speaker corrected in the direction of the favored pattern, replanting the
boundary and thus finishing with
(26) AND a little bathroom coming off of the hallway.
6. In this presentation we have opened a theoretical perspective which is striking-
ly different from that of recent grammatical inquiry. It would be possible to take
each sentence in our data and analyse it into the maximal number of possible
predications. We might then erect an abstract and elaborate structure in which each
noun phrase, adjective, verb, adverb, or even conjunction was seen as a separate
logical predication. But we doubt if such structures, without strong empirical
support, can lead us to understand the actual semantics and grammatical opera-
tions that govern the language. We hope to clarify the semantics of natural lan-
guage by distinguishing the cognitive input-the perceived state of affairs-from
the selection and evaluation procedures imposed by discourse rules. It appears
that a very large part of sentence grammar is dependent on and arises from rules
of social interaction, social conventions, and abstract discourse processes which we
have exemplified here.
There are many other aspects of the grammar of our apartment lay-out data
in which the influence of discourse rules may be seen; a number of these are exam-
ined in Linde's dissertation. A very large part of the grammatical apparatus is
still unspecified by our rule system; but we have for the first time established the
position of discourse rules as a necessary link between cognitive representationsand
the syntax of sentences.

REFERENCES
DOUGHERTY, RAY C. 1971. A grammar of coordinate conjoined structures,II. Lg.
47.298-339.
DOWNS,ROGER,and DAVID STEA. 1973. Cognitive maps and spatial behavior. Image
and environment,ed. by R. Downs & David Stea, 11-13. Chicago: Aldine.
1970. Towards a semantic structure of deep discourse grammar.
WOLFGANG.
DRESSLER,
Papers from the 6th Regional Meeting, Chicago LinguisticSociety, 202-9.
FODOR,J. A., and T. G. BEVER.1965. The psychological reality of linguistic segments.
Journalof Verbal Learningand Verbal Behavior4.4-20.
Foss, D. J., and R. H. LYNCH,JR. 1969. Decision processes during sentence compre-
hension: effects of surface structureon decision times. Perception and Psycho-
physics 5.145-8.
GLEITMAN, LILA. 1965. Coordinating conjunctions in English. Lg. 41.260-93.
, and HENRYGLEITMAN. 1970. Phrase and paraphrase. New York: Norton.
GRIMES,JOSEPH E. 1971. Kinds of information in discourse. Kivung 4.64-74.
GUNTER,RICHARD.1966. On the placement of accent in dialogue. Journal of Linguistics
2.159-79.
LABOV,WILLIAM.1966. On the grammaticality of everyday speech. Paper presented
at the LSA Annual Meeting, New York.
 A SITE FOR THE STUDY OF LANGUAGE AND THOUGHT 939

--. 1972. Language in the Inner City. Philadelphia: University of Pennsylvania

Press.
--, and JOSHUA WALETZKY. 1967. Narrative analysis. Essays on the verbal and visual
arts, Proceedings of the 1966 Spring Meeting of the American Ethnological Society,
ed. by June P. Helm, 12-44. Seattle: University of Washington Press.
LAKOFF,GEORGE,and STANLEYPETERS.1966. Phrasal conjunction and symmetric
predicates. (Harvard Computation Laboratory Report NSF-17.) Cambridge,
Mass.
LEE, GREGORY.1967. The English preposition with. Working Papers in Linguistics,
Ohio State University, 1.30-79.
LINDE,CHARLOTTE. 1974. The linguistic encoding of spatial information. Columbia
University dissertation.
LYNCH,KEVIN.1960. The image of the city. Cambridge, Mass.: MIT Press.
MACLAY,HOWARD,and CHARLES E. OSGOOD.1959. Hesitation phenomena in spon-
taneous English speech. Word 15.19-44.
SAVIN, H. B., and E. PERCHONOK. 1965. Grammatical structure and the immediate
recall of English sentences. Journal of Verbal Learning and Verbal Behavior
4.348-53.
SEARLE, JOHN.1969. Speech acts. Cambridge: University Press.
WALD, BENJI V. 1973. Variation in the system of tense markers of Mombasa Swahili.
Columbia University dissertation.
[Received21 October 1974.]