...... "'.

~-----

Frame Semantics
Chin-W. Kim
110

8.249-335. . I I a ' vol 1 Linguistic Research

· I t T 197;_. Problems ill the Theory oj P 10/10 0",), .'
LIgl ner,. Charles J. Fillmore
Inc., Edmonton, Canada. d lIabification Linguistic Inquiry 10.443-466.
McCarthy, J. 1979a. On stress a~l SSY 'f PI on~logy and Morphology. Unpub- University of California, Berkeley
1979b. Formal Problems III elm IC 1
--iished MIT Ph.D. dissertation. . I I of Tiberian HebrClf. Ullpub-
P r. m c A 1975 The Phonology and MOIp 10 ogy
e,.· . 1. Introduction
lished MIT Ph.D. dissertatIOn. . . n Linguistic Inquiry 11.511-562.
. 1980. A metrical theory for Estolllall qu~n l:r~trical Structure, and Harmony With the term 'frame semantics' I have in mind a research program in
Sfi K d 1979. Papers on Syllable Structwe, . , empirical semantics and a descriptive framework for presenting the results
a 1r, . e . W k" P pers in LingUlstlcs, vol. 1).
Processes (= M IT or 111g ~ . ' English word stress. Linguistic of such research. Frame semantics offers a particular way of looking at
Selkirik, E. 1980. The role of prosodic categofles 111 word meanings, as well as a way of characterizing principles for creating new
Inquiry 11.563-605. . . ' " The Relation Between Sound and Stmc- words and phrases, for adding new meanings to words, and for assembling
. forthcoming. Phonology and S)'nta.\ . the meanings of elements in a text into the total meaning of the text. By
.-- ture. MIT Press. . of svllabic phonology. Linguistische Berichte the term 'frame' I have in mind any system of concepts related in such a way
Vennemann, T. 1972. On the theOlY ..
that to understand anyone of them you have to understand the whole
18.1-18.
1aunony. University of Massachu-
mal theory of vowe II" structure in which it fits; when one of the things in such a structure is in-
R 1979 . A for
Vergnau d , J -.
setts Occasional Papers 5. 135-.143. . 1 I bo Linguistic Inquiry 7.463-
troduced into a text, or into a conversation, all of the others are automatically
Williams. E. 1976. Underlying tone 111 Marg! anc g . made available. I intend the word 'frame' as used here to be a general cover
484. term for the set of concepts variously known, in the literature on natural
language understanding, as 'schema', 'script', 'scenario', 'ideational scaf-
folding', 'cognitive model', or 'folk theory'.!
Frame semantics comes out of traditions of empirical semantics rather
than formal semantics. It is most akin to ethnographic semantics, the work
of the anthropologist who moves into an alien culture and asks such ques-
tions as, 'What categories of experience are encoded by the members of this
speech community through the linguistic choices that they make when they
talk?' A frame semantics outlook is not (or is not necessarily) incompatible
with work and results in formal semantics; but it differs importantly from
formal semantics in emphasizing the continuities, rather than the discontinui-
ties, between language and experience. The ideas I will be presenting in this
paper represent not so much a genuine theory of empirical semantics as a
set of warnings about the kinds of problems such a theory will have to deal
with. If we wish, we can think of the remarks I make as 'pre-formal' rather
than 'non-formalist'; I claim to be listing, and as well as I can to be describ-
ing, phenomena which must be well understood and carefully described
before serious formal theorizing about them can become possible.

[Ill]

The Linguistic Society of Korea (ed.), 1982,

Linguistics in the Morning Calm, Seoul: Hanshin Publishing Co.
 112 Charles J. Fillmore Frame Semantics 113

In the view] am presenting, words represent categorizations of experience, he doesn't live with her." The substitution in this frame of BUT and YET
and each of these categories is underlain by a motivating situation occurring suggests that these two words have (by this diagnostic at least) very similar
against a background of knowledge and experience. With respect to word functions; insertion of MOREOVER or HOWEVER suggest the existence of
meanings, frame semantic research can be thought of as the effort to under- conjunctions functioning semantically similarly to BUT and YET but re-
stand what reason a speech community might have found for creating the quiring sentence boundaries. The conjunctions AND and OR can meaningfully
category represented by the word, and to explain the word's meaning by be inserted into the frame, but in each case (and in each case with different
presenting and clarifying that reason. effect) the logical or rhetorical 'point' of the whole utterance differs impor-
An analogy that] tlnd helpful in distinguishing the operation and the goals tantly from that brought about by BUT or YET. In each of these cases, what
of frame semantics from those of standard views of compositional semantics one came to know about these words was the kind of structures with which
is between a grammar and a set of tools-t06ls like hammers and knives, they could occur and what function they had within those structures.
but also like clocks and shoes and pencils. To know about tools is to know In the early sixties, together with William S-Y. Wang and eventually D.
what they look like and what they are made of-the phonology and morphol- Terence Langendoen and a number of other colleagues, I was associated with
ogy, so to speak-but it is also to know what people use them for, why the Project on Linguistic Analysis at the Ohio State University. My work on
people are interested in doing the things that they use them for, and maybe that project was largely devoted to the classification of English verbs, but
even what kinds of people usc them. ]n this analogy, it is possible to think now not only according to the surface-syntactic frames which were hospitable
of a linguistic text, not as a record of 'small meanings' which give the to them, but also according to their grammatical 'behavior', thought of in
interpreter the job of assembling these into a 'big meaning' (the meaning of terms of the sensitivity of structures containing them to particular grammat-
the containing text), but rather as a record of the tools that somebody used ical 'transformations.' This project was whole-heartedly transformati onalist,
in carrying out a particular activity. The job of interpreting a text, then, is basing its operations at first on the earliest work on English transformational
analogous to the job of figuring out what activity the people had to be grammar by Chomsky (1957) and Lees (1961), and in its later stages on
engaged in who used these tools in this order. advances within the theory suggested by the work of Peter Rosenbaum
2. A Private History of the Concept 'Frame' (Rosenbaum 1967) and the book which established the standard working
paradigm for transformationalist studies of English, Chomsky (1965). What
I trace my own interest in semantic frames through my career-long interest animated this work was the belief that discoveries in the 'behavior' of partic-
in lexical structure and lexical semantics. As a graduate student (at the ular classes of words led to discoveries in the structure of the grammar of
University of Michigan in the late fifties) I spent a lot of time exploring the English. This was so because it was believed that the distributional
co-occurrence privileges of words, and] tried to develop distribution classes properties of individual words discovered by this research could only be
of English words using strings of words or strings of word classes as the accommodated if the grammar of the language operated under particular
'frames' within which] could discover appropriate classes of mutually sub- working principles. My own work from this period included a small mono-
stitutable elements. This way of working, standard for a long time in phono- graph on indirect object verbs (Fillmore 1961) and a paper which pointed to
logical and morphological investigations, had been developed with particular the eventual recognition of the transformational cycle as an operating
rigor for purposes of syntactic description by Charles Fries (Fries 1952) principle in a formal grammar of English (Fillmore 1963).
and played an important role in the development of 'tagmemic formulas' The project's work on verbs was at first completely syntactic, in the sense
in the work of Kenneth Pike (Pike 1967), the scholars who most directly that what was sought was, for each verb, a full account (expressed in terms of
intluenced my thinking during this period. Substitutability within the same subcategorization (features) of the deep structure syntactic frames which
'slot' in such a 'frame' was subject to certain (poorly articulated) conditions t were hospitable to it, and a full account (expressed in terms of rule features)
of meaning-preservation or structure-preservation, or sometimes merely of the various paths or 'transformational histories' by which sentences con-
meaningfulness-preservation. In this conception, the 'frame' (with its single taining them could be transformed into surface sentences. The kind of work
open 'slot') was considered capable of leading to the discovery of important I have in mind was carried on with much greater thoroughness by Fred
functioning word classes or grammatical categories. As an example of the Householder and his colleagues at Indiana University (Householder et al
workings of such a procedure, we can take the frame consisting of two 1964), and with extreme care and sophistication by Maurice Gross and his
complete clauses and a gap between them, as in "John is Mary's husband- team in Paris on the verbs and adjectives of French (Gross 1975).
 114 Charles J. Fillmore Frame Semantics 115

Tn the late sixties I began to believe that certain kinds of groupings of rate GIVE from SEND, just as there seemed to be semantic commonalities
verbs and classifications of clause types could be stated more meaningfully between ROB and STEAL, BUY and SELL, ENJOY and AMUSE, etc.,
if the structures with which verbs were initially associated were described in which were lost in the syntactic class separation of these verbs.
terms of the semantic roles of their associated arguments. I had become My ultimate goal in this work in 'case grammar' (as the framework came
aware of certain American and European work on dependency grammar to be called) was the development of a 'valence dictionary' which \vas to
and valence theory, and it seemed clear to me that what was really important differ importantly from the kinds of valence dictionaries appearing in Europe
about a verb was its 'semantic valence' (as one might caB it), a description of (e.g., Helbig and Schenkel 1973) by having its semantic valence taken as
the semantic role of its arguments. Valence theory and dependency grammar basic and by having as much as possible of its syntactic valence "ccounted
did not assign the same classificatory role to the 'predicate' (or 'VP') that for by general rules. (Thus, it was not thought to be necessary to explain,
one found in transformationaJist work (see, e.g., Tesniere 1959); the kind of in individual lexical entries, which of the arguments in a [V A P IJ predica-
semantic classifications that I needed could be made more complete and tion of the type described above was to be the subject and which was to be
sensible, I believed, if, instead of relying on theoretically separate kinds of the object, since such matters were automatically predicted by the grammar
distributional statements such as 'strict subcategorization features' and with reference to a set of general principles concerning the mapping from
'seJcctional features: one could take into account the semantic roles of all configurations of semantic cases into configurations of grammatical rela-
arguments of a predication, that of the 'subject' being simply one of them. tions.)
Questioning, ultimately, the relevance of the assumed basic immediate- Although the concept of 'frame' in various fields within cognitive psychol-
constituency cut between subject and predicate, I proposed that verbs could ogy appears to have origins quite independent of linguistics, its use in case
be seen as basically having two kinds offeatures relevant to their distribution grammar was continuous, in my own thinking, with the use to which I have
in sentences: the first a deep-structure valence description expressed in terms put it in 'frame semantics'; In particular, I thought of each case frame as
of what 1 called 'case frames', the second a description in terms of rule characterizing a small abstract 'scene' or 'situation', so that to understand
features. What I called 'case frames' amounted to descriptions of predicating the semantic structure of the verb it was necessary to understand the proper-
words that communicated such information as the following: 'Such-and-sllch ties of such schematized scenes.
a verb occurs in expressions containing three nominals, one designating an The scene schemata definable by the system of semantic cascs (a system
actor who performs the act designated by the verb, one designating an object of semantic role notions which 1 held to be maximally general and defIning a
on which the actor's act has a state-changing influence, and one designating minimal and possibly universal repertory) was sufficient, I believed, for
an objcct through the manipulation of which the actor brings abollt the understanding those aspects of the semantic structure of a verb which were
mentioned state change.' In symbols this statement could be represented as linked to the verb's basic syntactic properties and to an understanding of
[- A P I], the letters sta nding for' Agent', 'Patient' and 'Instrument'. Actually, the ways in which different languages differently shaped their minimal
the kind of description I sought distinguished 'case frames' as the structures clauses, but they were clearly not adequate for describing with any com-
in actual individual sentences in which the verbs could appear from 'case pleteness the semantic structure of the clauses containing individual verbs.
frame features' as representations of the class of 'case frames' into which This theory of semantic roles fell short of providing the detail needed for
prrticular verbs could be inserted. In the description of 'case frame features' semantic description; it came more and more to seem that another inucpcnd-
it was possible to notice which of the 'cases' were obligatory, which \vere ent level of role structure was needed for the semantic description of verbs
.optional, what selcetionaJ dependencies obtained among them, ~l11d so on in particular limited domains. One possible way of dcvising a fuller account
(sec Fillmorc 19(8). of lexical semantics is to associatc some mec1wnism for deriving cds of
"We were developing a kind of mixed syntaclic-semantic valence descrip- truth conditions for a clause from semantic information individually "tt"chcd
'lion of verbs, and we noticed that the separate valence patterns seemed to to given predicates; but it seemed to me more proiitable to believe that there
characterize semantic types of verbs, such as verbs of perception, causation, are larger cognitive structures capable of providing a new layer of ~;elJlantic
movement, etc. Within these synt8.ctic valence types, however, it seemed role notions in terms of which whole domains ofvo','abulary could he SCIil:il1-
that some semantic generalizations were lost. There seemed to be important tically characterized.
differcnces between GIVE IT TO JOHN and SEND JT TO CHICAGO My flrst attempt to describe one sllch cognitivc structure was in a paper
that could not be illuminated merely by showing what syntactic rules sepa- on 'Verbs of judging' (Fillmore 1971)--vcrbs like BLAME, ACCUSE,
-

116 Charles J. Fillmore

CRITICIZE-for which ! needed to be able to imagine a kind of 'scene

schematization' that was essentially different from the sort associated with
'case frames'. In devising a framework for describing the elements in this
I
l( and motivation for the categories whieh these
word 'frame' for the structured '
Frame Semantics

. words represent. Using the

/17

y
class of verbs, I found it useful to distinguish a person who formed or ex- remembered we can say that t1 fwa In whIch the scene is presented or
' le rame structures tl . d '
pressed some sort of judgment on the worth or behavior of some situation or t h at the word 'evokes' the frame le wor -meanIngs, and
individual (and I called such a person the Judge); a person concerning whose The structures T have mentio~ed so far ca b
behavior or character it was relevant for the Judge to make a judgment (I the categories speakers wish t b" n e thought of as lllotivatino-
. 0 nng II1to play wh d 'b" b
called this person the Defendant); and some situation concerning which it t Ilat mIght be independent f tJ ' en escn II1g SItuations
o le actual speech SIt t' h
seemed relevant for the Judge to be making a Judgment (and this I called context. A second and equall . , u a Ion, t e conversational
y Important kJnd of f ' .
simply the Situation). In terms of this framework, then, I chose to describe t 1le actual communication situat' Wh ramlllg IS the framing of
Ion. en we underst d '
ACCUSE as a verb usable for asserting that the Judge, presupposing the guage, we bring to the task both ou b'l't . an a piece of lan-
r a I I Y to aSSign I ' .
badness of the Situation, claimed that the Defendant was responsible for the p h ases or components of the 'w Id' tl sc lernatlzatlOns of the
' , or lat the text s e1 '
Situation; I described CRITICIZE as usable for asserting that the Judge, ~n our abIlity to schernatize the situatio'
d . on: l~W charactenzes,
presupposing the Defendant's responsibility for the Situation, presented IS being produced, We have b t l ' n, l,n whIch thiS pIece of language
f ' 0 1 cognItIve frames' d" ,
arguments for believing that the Situation was in some way blameworthy, rames , the latter having to do 'tl h an IllteractlOnal
The details of my description have been 'criticized' (see esp, McCawley 1975), b etween the speaker and the I WI 1 ow we concept
b
r I
ua Ize w lat IS going on
'
but the point remains that we have here not just a group of individual By the early seventiec I had b learer,' or fl
etween the tl
au lOr and the reader.
f, , , 3 ecome In uenced by 'k
words, but a 'domain' of vocabulary whose elements somehow presuppose ormatlvlty, and prao-matl'cs I' I WOI on speech acts per-
, b n genera and had b .,'
a schematization of human judgment and behavior involving notions of field 111 the form of a numb f ,,' egun contnbutlllg to this
, < er 0 wntlllgs 0 . , ~
worth, responsibility, judgment, etc" such that one would want to say that e,g" Fillmore 1975). Knowled g e of deic~ presuPP~sltlOns and deixis (see,
nobody can really understand the meanings of the words in that domain standing of the ways in whl'ch t tIC categones requires an under-
, . enses person ma k'
who docs not understand the social institutions or the structures of experi- stratlve categories etc schem t' tl' , rIng morphemes, demon-
, ' "". a Ize le comm ' . .
ence which they presuppose. of Jllocutionarv poin ts princi.pl f ~l1Icatlllg SItuatIOn; knowledge
, J, '. es 0 conversatIOnal '
A second domain in which 1 attempted to characterize a cognitive 'scene' I~ed speech events, contribute to the full unde co~peratlon, and routin-
with the same function was that of the 'commercial event' (see Fillmore tlOnal exchanges, Further know' tl t ' rstandlllg of most conversa-
• IIlg la a text IS sa b'
]977b), In particular, I tried to show that a large and important set of of marriage, a business contract or a f, Ikt I' y, a,n 0 Ituary, a proposal
English verbs could be seen as semantically related to each other by virtue how to interpret particular passa~es in it \0: e, prOVIdes knowledge about
of the different ways in which they 'indexed' or 'evoked' the same general and how to know when it I'S fi ' h d 1" to expect the text to develop
nls e, t IS freq tl I '
'scene', The elements of this schematic scene included a person interested in expectations combine with tIle t I ' uen y t le case that SUcll
, . ac ua matenal of tl t 1
exchanging money for goods (the Buyer), a person interested in exchanging correct IIlterpretation And 0 'I" le ext to ead to the text's
, ' ... nee agall1 t lIS IS ac l' I .
goods for money (the Seller), the goods which the Buyer did or could acquire mtnd an abstract structure of exp t t' . COl,llp IS led by haVing in
ec a Ions which bmw 'tl'
(the Goods), and the money acquired (or sought) by the seller (the Money). poses, natural or conventional I' d ~s WI 1 It roles, pur-
ze sequences of eve t t
Using the terms of this framework, it was then possible to say that the verb o f the apparatus that ViP wish t ' . n ypes, and all the rest
BUY focuses on the actions of the Buyer with respect to the Goods, back- i
In the mid-seventie", carne 0 tassoclate Wlt~l the notion of 'frame',
k f
grounding the Seller and the Money; that the verb SELL focuses on the (R osch 1973) and that of BrentIIIB0 contact ('
WIth the
Wor o· Eleanor Rosch
er 111 and Paul K'1\1 (B I'
actions of the Seller with respect to the Goods, backgrounding the Buyer and began to see the importance ofth ,< <J er.1I1 and Kay J 969)
e notIon of'pr t t ..
and the Money; that the verb PAY focuses on the actions of the Buyer with t Ile nature of human c'ltegorl' . t' TI . 0 0 ype In understanding
• . < za Ion Hough tl . k ~
respect to both the Money and the Seller, backgrounding the Goods, and (ZImmer 1971) and Pamela D .' . le Wor of Karl Zimmer
. . ' ownIng (DownIng 1977)
so on. with such verbs as SPEN D, COST, CHARGE, and a number of others categorIZIng ~
contexts. to prI'nc'Ip Ies 0 f word-f, ~. - , ' on the. relevance of
somewhat more peripheral to these. Again, the point of the description was reflects fruitful collabontion WI't! P I K OImdtlon and, In work that
, 1 au ay and G L k If
to argue that nobody could be said to know the meanings of these verbs who propose descriptions of w o r d ' eorge a ,0 . I began to
. meanings that made f I
did not know the details of the kind of scene which provided the background notIon, One generalization that s .d /. use 0 t le prototype
eeme va Id was tInt f
or background against which t l ' < very 0 ten the frame
~ le meanlllg of a word is defined and under--
 118 Charles J. Fillmore

stood is a fairly large slice of the surrounding culture, and this background
understanding is best understood as a 'prototype' rather than as a genuine
I Frame Semantics

sleep through the morning, wake up at three 0 'clock in the afternoon and
119

sit dO\vn to a meal of eggs, toast, coffee and orange juice. and can that'meal
body of assumptions about what the world is like. It is frequently useful, 'break:as~', shows that the 'early morning' character of t hc category is also
when trying to state tmtn conditions for the appropriateness of predicat~ng not cnter~al; and last~y, the fact that a person can sleep through the night,
the word of something, to construct a simple definition of the word, allowmg wake up \l1 the mor11lng, have cabbage soup and chocolate pic 'for break-
the complexity of fit between uses of the word and real world situations to f~lSt', shows that the 'breakfast menu' character of the concept is also not
be attributed to the dctails of the prototype background frame rather than criteria!' (This in spite of the fact that an American restaurant that advertises
to the details of the word's meaning. Thus we could define an ORPHAN as its willingness to serve breakElst at any time is referring precisely to the
a child \-vhose parcnts are no longer living, and then understand the category stereotyped breakfast ingredients.) What we want to say, when we observe
as motivated against a background of a particular kind: in this assumed usage phenomena like that, is not that we have so far failed to capture the
background wo~'ld, children depend on their parents for care and guidance tru~ core of the wor,d's meaning, but rather that the word gives us a category
and parents accept the responsibility of providing this care and guidance which can be llsed 1Il many different contexts, this range of contexts deter-
without question; a rerson without parents has a special status, for society, mined by the multiple aspects of its prototypic use-the use it has when the
only up to a particular age, because during this period a society needs to conditions of the background situation more or less exactly match the de-
provide some special way of providing care and instruction. The cateogy fining prototype.
ORPHAN does not have 'built into it' any specification of the age after which The descriptive framework which is in the process of evolving out of all
it is no longer relevant to speak of somebody as an orphan, because that of the above considerations is one in which words and other linguistic forms
understandi~lg is a part of the background prototype; a boy in his twenties and categories are seen as indexing semantic or cognitive categories which
is generally regarded as being able to take care of himself and to have passed are themselves recognized as participating in larger conceptual structures of
th~ age where the main guidance is expected to come from his family. It is some sort, all of this made intelligible by knowing something about the kinds
that background information which determines the fact that the word OR- of settings or contexts in which a community found a need to make such
PHAN vvould not be appropriately used of such a boy, rather than informa- categories available to its participants, the background of experiences and
tion that is to be separately built into a description of the word's meaning. practices within which such contexts could arise, the categories, the contexts,
In the prototype situation, an orph"\I1 is seen as somebody deserving of pity and the backgrounds themselves all understood in terms of prototypes. '
and Ci.mcern; hence the point of the joke about the young man on trial for
3. Further Illustrations and Some Terminological Proposals
the murder of his parents who asked the court for mercy on the grounds that
he was an orphan: the prototype scene against which society has a reason to A 'frame', as the notion plays a role in the description of linguistic mean-
categorize some children as orphans does not takc into ,lccount the case in ings, is a system of categories structured in accordance with son;e motivating
which a child orphans himself. context. Some words exist in order to provide access to knowledr!e of such
As a second cxaml,!e of a category that has to be fitled onto a background frames to the participants in the communication process, and simultaneouslv
of inslitutions and practices we can consider the word BREAKFAST. To serve to perform a categorization which takes such framing for granted ..
understand this word is to understand the practice in our culture 01 having The motivating conte:~t is some body of understandings. some pattern of
three Illea!s a day, at more or less conventionally ec;tablished times of the practices, or some history of social institutions, against which we find intelli-
cia\', and lo!' one of these: meaL; to be the one which is eaten early in the gible the crec\tion of a particular category in the history or the lan!?:uage
d'i~. ~lfler a period 01 sleep, and for it to comisl of a somewhat unique menu community. The \-vord WEEK·END convcrys what it convey~; both bc::au~c
(tl:c detail:, of which can vary from community to cOlllmunity). Wh;l!. is inter- of the calendric seven-day cycle and because or a partic~t1ar practice of
c;;tin!! 'lhOlilthc \vord BREAKFAST is that each of the three conditions most devoting a relatively brger continuous block of davs within such a cvcle
typic~;\1v ,I:,sl'ciatcd with it can be independently ,lbsent still allowing na,live to public work and two continuous days to one's pri\;ate lire, rrwe had ~nlv
speakers tl' use the word, The f,ICt that someone can work through th,e mght one 'day of rest' there would be no need for the word 'week-end': one could
withoul sleep, and the ') at sun-up have a meal 01 eggs, toast. conee and simply use the name of that day. If we had three days or work and four days
orangeiuice, and cedi lint meal 'breakfast', shows clearly that the 'post- of rest, then too it seems unlikely thal the name 1'01' the period devotee! (0
sleep' ch,:r~lcter of the category is not criterial: the fact that someone can one's private life would kIVe been given Ihat name (If the work week is
 120 Charles J. Fillmore

gradually shortened, the word 'week-end' might stay; but it is unlikely that
, Frame Semantics 121

cept. That is, nobody schematizes the physical world in a way that would
the category could have developed naturally if from the start the number of give a reason to speak of part of it as 'phlogiston'.
days devoted to work were shorter than the number of the remaining days. To illustrate the point with items from everyday language, we can con-
An acquaintance of mine who works only on Wednesdays, pleased at being sider the words LAND and GROUND (which I have described elsewhere
able to enjoy 'a long week-end', recognizes that the word is here being used but cannot forego mentioning here). The difference between these two words
facetiously. ) appears to be best expressed by saying that LAND designates the dry surface
The word VEGETARIAN means what it means, when used of people in of the earth as it is distinct from the SEA, where as GROUND designates
our culture, because the category of 'someone who eats only vegetables' is the dry surface of the earth as it is distinct from the AIR above it. The words
a relevant and interesting category only against the background of a com- 'land' and 'ground', then, differ not so much in what it is that they can be
munity many or most of whose members regularly eat meat. Notice that used to identify, but in how they situate that thing in a larger frame. It is
the word designates, not just someone who eats plant food, but someone by our recognition of this frame contrast that we are able to understand that
who eats only plant food. Furthermore, it is used most appropriately for a bird that 'spends its life on the land' is being described negatively as a bird
situations in which the individual so designated avoids meat deliberately and that does not spend any time in water; a bird that 'spends its life on the
for a purpose. The purpose might be one of beliefs about nutrition, or it may ground' is being described negatively as a bird that does not fly.
be one of concerns for animal life; but the word is not used(in a sentence like Though the details are a bit tricky, the two English words SHORE and
"John is a vegetarian.") to describe people whose diet does not include meat COAST (not differently translatable in many languages) seem to differ from
because they are unable to find any, or because they cannot afford to buy it. each other in that while the SHORE is the boundary between land and water
Occasionally one comes upon a term whose motivating context is very from the water's point of view, the COAST is the boundary between land and
specific. One such is the compound FLIP STRENGTH, used, I am told, in water from the land's point of view. A trip that took four hours 'from shore
the pornographic literature business. Some publishers of pornographic novels to shore' is a trip across a body of water; a trip that took four hours 'from
instruct their authors to include a certain quota of high interest words on coast to coast' is a trip across a land mass. "We will soon reach the coast"
every page, so that a potential customer, in a bookstore, while 'flipping' the is a natural way to say something about a journey on land; "we will soon
pages of the book, will, no matter where he opens the book, find evidence reach the shore" is a natural way to say something about a sea journey.
that the book is filled with wonderful and exciting goings-on. A book which Our perception of these nuances derives from our recognition of the different
has a high ratio of nasty words per page has high flip strength; a book which ways in which the two words schematize the world.
has these words more widely distributed has low flip strength. As I under- The Japanese adjective NURUI is another example of a framing word.
stand the word, an editor of such a publication venture might reject a manu- Although not all Japanese-speaking informants support this judgment,
script, requesting that it be returned only after its flip strength has been raised. enough do to make the example worth giving. In the usage that supports
With this last example, it is extremely clear that the background context my point, NURUI, used to describe the temperature of a liquid means
is absolutely essential to understanding the category. It is not that the condi- 'at room temperature', but it is said mainly of liquids that are ideally hot.
tions for using the word cannot be stated without this background under- "Kono ocha ga nurui" (this tea is lukewarm) is an acceptable sentence in the
standing (relative flip strength of novels could easily be determined by a idiolects that support my point, but "kono biiru ga nurui" (this beer is luke-
computer), but that the word's meaning cannot be truly understood by warm) is not. It will be noticed that the English word LUKEWARM does
someone who is unaware of those human concerns and problems which not 'frame' its object in the same way. A cold liquid and a hot liquid can
provide the reason for the category's existence. both become lukewarm when left standing long enough; but only the liquid
We can say that, in the process of using a language, a speaker 'applies' a that was supposed to be hot can be described as 'nurui'.
frame to a situation, and shows that he intends this frame to be applied by A large number of framing words appear only in highly specialized con-
using words recognized as grounded in such a frame. What is going on here texts, such as the term FLIP STRENGTH discussed earlier. The I~gal term
seems to correspond, within the ordinary vocabulary of a language, .to DECEDENT gives us another example of such context specialization. Ac-
lexical material in scientific discourse that is describable as 'theory laden': cording to my legal informants (and my available law dictionaries) the word
the word 'phlogiston' is 'theory-laden'; the reason it is no longer used in DECEDENT is used to identify a dead person in the context of a discussion
serious discourse is that nobody accepts the theory within which it is a con- of the inheritance of that person's property. (The word DECEASED, as in
 122 Charles J. Fillmore

the phrase 'the deceased', is also limited to legal or journalistic contexts,

but it is not limited to any particular subdomain within the law.) Another
I! Frame Semantics

the current scene-something that might be visible in a pictorial representa-

tion of tbe scene-but is that of a much larger framework. Thus, the descrip-
123

example is MUFTI. Mufti, in the sense it once had in the military service, tion of someone as a HERETIC presupposes an established religion, or a
refers to ordinary clothing when worn by somebody who regularly wears a religious community which has a well-defined notion of doctrinal correct-
ness. In a community lacking such beliefs or practices, the word has no pur-
military uniform. If we sec two men wearing identical suits, we can, referring
to their clothing, say that one of them is 'in mufti' if that one is a military \ pose. Sometimes a word situates an event in a history wider than the history
of the ongoing narrative. In speaking of locations within North America,
officer. The property of being 'in mufti' is obviously a property that has
relevance only in the context of a military community. the expressions OUT WEST and BACK EAST arc frequenlly used. The
Given all these examples of clear cases of terms linked to highly specific terms h,ne the form they do because for a large portion of American families
cognitive frames, we can see that the process of understanding a text involves the settlement history of the country traced its way from the east coast to the
retrievi ng or perceiving the frames evoked by the text's lexical content and west coast. European immigrants first landed on the east coast: some of them,
assembling this kind of schematic knowledge (in some way which cannot be or some of their descendants, gradually migrated westward. The easten~
easily formalized) into some sort of 'envisionment' of the 'world' of the part of the country, where these immigrants or their ancestors once were,
text. If I tell you (to be somewhat ridiculous) that the decedent while on was BACK EAST; the western part of the country. not yet reached, wa~
land and in mufti last weekend ate a typical breakfast and read a novel high OUT WEST. The expressions are used today by people whose families did
in Hip strength, you know that I am talking about a now-dead naval officer not share in this general westward movement themselves, but the terms recall
who during the period including last Saturday and Sunday read a pornogra- the historical basis of their creation.
phic novel: and you know a few other things about the man, about how he Earlier I spoke of the notion of deep cases as oflering an account of the
spent his time, and about the setting in which this report of his activities is semantic aspects of single-clause predications which flgured in the basic
given. The sentence did not give you this information directly: you had to grammatical structure of clauses. A broader view of the semantics of gram-
'compute' some of it by constructing. in your imagination, a complex context mar, one which owes a great deal to the work of Leonard Talmy (see Talmy
within which each of the lexically signaled framings was motivated. We see 1980) and Ronald Langacker (Langacker forthcoming), sees lexical framing
in this way that there is a very tight connection between lexical semantics and providing the 'content' upon which grammatical structure performs a 'con-
text semantics, or, to speak more carefully, between lexical semantics and figuring' function. Thinking in this way, we can sec that any grammatical
the process of text comprehension. The framing words in a text reveal the category or pattern imposes its own 'frame' on the material it structures.
multiple ways in which the speaker or author schematizes the situation and For example, the English pluperfect can be described as having as its role,
induce the hearer to construct that envisionment of the text world which in structuring the 'history' of the text world, that of characterizing tbe situa-
would motivate or explain the categorization acts expressed by the lexical tion at a particular time (the narrative time) as being partly explained by the
occurrence of an event or situation that occurred or existed earlier on. The
choices observed in the text.
The interpreter's envisionmcnt of the text world assigns that vvorld both a progressive aspect, in its turn, schematizes a situation as one which is con-
perspective and a history. A report of somcbody buying something evokes tinuing or iterating across a span of time. Thus, a sentence in a narrative of
the frame of the commercial event, but sees that event, for the moment at the form "She had been running," a form which combines the progressive
least. from the point of view of one of its participants. Describing somebody and the pluperfect forms, can bave the function of explaining why, at the
as being ON LAN D locates the scene in the history of a SC,) voy,'ge, by narrative time point, "she" was panting, or sweating, or tired. Thus we see
noticing that it is relevant to dcscribe the location in this way only if this that the cognitive frames which inform and shape our understanding of
pcriod ie; seen as ;1\1 intcrruption of a period of sea tnwel. Saying that some- language can differ greatly in respect to their generality or specificity: a lexical
body is AT BAT locates an event as one part of a particular baseball game. verb like RUN can give us a specific kind of physical activity image, while
Describing: cofTee, in j,lpanese, ;IS NURUI recognize; that it \\as once hot the pluperfect and the progressive combine, each in a gener,;] and~ abstr,lct
,111d has been allowed to 'coor. One knows that the coffee is currently at way, to shape the image of running in a way that fits the current SitU,ltiol1
room tempcrature. but also thal it did not get that way by starting out as and to situate the event of running both temporally and in 'relevance' into
the ongoing history of the text world.
iced cofTee.
Sometimes the perspective which a word assigns is not a perspective on It is necessary to distinguish two importantly different \\,;1\S in \\iJich the
 Frame Semantics 125
Charles J" Fillmore
124
h\lternate Fram"lOgs 0 f a Smgle
. Situation
cognitive frames we call on to help us interpret linguistic texts get introduced
into the interpretation process. On the one hand, we have cases in which the From a frame semantics point of view " "
the same 'facts' can be present d "tl"' It I.S frequently possible to show that
lexical and grammatical material observable in the text 'evokes' the relevant k I e WI 1111 different fra" f .
ma. e t lem out as different 'fact'S
s. omebody who sl mll1gs. . ramll1gs
. " which
frames in the mind of the interpreter by virtue of the fact that these lexical
give out money in a partieula " t ' " 10WS an unwilhngness to
forms or these grammatical structures or categories exist as indices of these STINGY . r Sl uatlon might be descr'b d b
. (111 which case the behavior i . l e y one person as
frames; on the other hand, we have cases in which the interpreter assigns and by another as THRIFTY (" .s contrasted With being GENEROUS)
coherence to a text hy 'invoking' a particular interpretive frame. An extremely WA 111 which case a c t . a'
r
ass~~:sPfh:tS.~h.e ST~NGY: con~
STEFUL). The speaker wi on rast IS made with bein
important difference between frames that are evoked by material in the text
trast to a way of behaving GENEROUS
and frames that are invoked by the interpreter is that in the latter case an
the behaver's treatment of" II h I IS to e evaluated with respect to
'outsider' has no reason to suspect, beyond a general sense of irrelevance or 1e ow umans' wh h
ates the behavior by applying to it a THRI;;as! e speaker who evalu-
pointlessness in the text, that anything is missing. To repeat an example that
assumes that what is most im t' Y. WASTEFUL contrast
I have used elsewhere, a Japanese personal letter in the traditional style is or
played in the use of money tr talnt IS a measure of the skill or wisdom dis-
supposed to begin with a comment on the current season. Somebody who o ler resources.
knows this tradition is able to sense the relevance of an opening sentence in
a letter which speaks of the garden floor covered with leaves. The kind of
*" 'Contrast Within Frames' versus 'Contrast Across Frames'

understanding which allowS such an interpretation comes from outside of T~e fact that a single situation can be 'frame ,. "
possIble two ways of presentin . d JI1 contrastll1g ways makes
. g a negatIon or an ..
contrasts mtroduced in the 1 t .
the text itself.
Invoked frames can come from general knowledge, knowledge that exists .
.
as paragraph If I say of b ~
opposItIOn. Using the
s Ingy-he's really generous" II ' some ody, "He's not
independently of the text at hand, or from the ongoing text itself. t
to measure him and I in"o' lave ah cce?ted the scale by which you choose
I. ' l' rm you t at m my ..
4. Frame-Semantic Formulations of Empirical Semantic Observations t lIS scale was in error. If on tl tl h 0p1l110n your application of
thrifty", what I am doing is le o. ler and I say "He's not stingy-he's
.v I~ questIon is not
In this section I examine a number of observations about lexical meaning t b propOSIng that the beha"lor "
o e evaluated along the STINGY' GENE
or text interpretation which permit formulations in terms of notions from THRIFTY:WASTEFUL d' .'. ROUS dImenSIOn but along the
frame semantics. In the following section I examine a number of traditional . ImenSlon In the first I h
partIcular standard in th J" . case ave argued for a
topics in standard semantic theorizing and raise questions about the impor- e app Icatlon of an accept d I .
case my utterance argues for .tlle ]rre. Ievance of one seIsca de;I In the second
tance they would be given in an account of linguistic meaning of the sort we ness of another. ca e an t,le appropriate-
have been exploring.
~ polysemy Arising from Alternative Framings of the Same Lexical Item Word Sense Creation by Frame Borrowing
WI?~n a speaker wishes to talk about somethin
f. g.for wlll~h an appropriate
For many instances of polysemy it is possible to say that a given lexical .
cogl1ltJve frame has not been establisl d
item properly fits either of two different cognitive frames. One possibility is a novel schematization he ca ' le. ' or OJ whIch he Wishes to introduce
that a word has a general use in the everyday language but has been given a
separate use in technical language. For example, we might wish to say that
. .. ' n sometImes acc
t le lIngUIStiC material associated "'tl . f
hI ' . (
o~p J"IS I1 t 1"liS byI transferring
WIld rame whIch mak d'
e s mterested in onto the new 't t" . . (es t le Istinctions
the English word ANGLE is understood in connection with a perceptual . Sl ua Ion relyIng on th . t
appropnateness of the tnnsfer C t "' e In erpreter to see the
frame as a figure made by two lines joined at a point in a way suggested by d ' . er all1 new senses of d
un erstood as having originated in this wa . .' wor s can be best
y, we might expect that such W~lS
a bent stick. Presented in terms of a competing procedural frame, an angle
the case in the importation of tJ t
is thought of in terms of the rotation of a line about a point, the angle itself . le erm BACHELOR' t I · '
appropnate
. to fur seal society
" t I
0 use t le exampJe made III 0 t le terminology
.
visually represented as the line before and after its rotation. In the procedural
semantIcs discussion from the remil d . K ,common 111 lexical
1 e~, matz and Fodor (l963), of the lIse
frame the notion of a 180 degree angle is intelligible, as is the notion of a of the word BACHELOR t d "

360 degree angle. Within the perceptual frame such notions do not fit. (The th' 0 ~slgnate a male fur se I 'tl
p

e matlllg season' Lakoff d J I a WI 10ut a mate during

example is from Arnheim 1969, p. 182f.) . an 0 lnson ( 1980) have made lIS a ware of the
 126
Charles J. Fillmore

value of metaphor in concqJtualization and communication, making the

persuasive case that in a great many domains of experience metaphors pro-
I Frame Semantics

perienc~ of al~ acquaintan~e of mine-when talking about very young

females, n:y fnend found hllllself, several times, using the word WOMAN
127

when tal~lllg about an eight-year-old girl. The fact that this friend would
vide us with the only way of communicating about those experiences. (Some never aCCIdentally use the word MAN when talking about an eight-vear-old
details are to be found elsewhere in this volume.) b~y show~ that the c.hange in question is not of the reschematization type
{hscussed 111 the prevIOus paragraph. An equally clear example of the same
Reframing a Lexical Set
phenomenon (as I have discussed elsewhere-Fillmore 1972) is in the use of
Various kinds of semantic change can be illuminated by considering the the word SUSPECT where the speaker or writer might have been inclined to
phenomena in frame semantic terms. One important type of change consists use such a word as BURGLAR
.' , MURDERER
. _ , ARSONIS1'
. , OJ, 1110re gener-
in reconstituting the motivating circumstances while preserving the lexical
a!ly, CU.LPRIT. ConscIOus of the legal doctrine that a person is to be con-
item and its basic fit with the associated scene. People observing certain
~Idere~ mnocent u.ntil pr?ven guilty, and conscious too of the danger of
usages of English with an eye to feminist concerns have noticed tendencies on commlttmg lIbel, JournalIsts and police officers have learned to identify
the part of many speakers to have certain asymmetries in the sets of condi- persons accused of crimes but not (yet) legally held to be guilty of them a-s
tions for using the words in the proportion BOY :MAN: :GIRL:WOMAN. SUSPECTS. ~ change. in usafc which would clearly reflect the- adoption of
In particular, in the usage pattern that I have in mind, males appeared to be the legal.d.octnne mentIOned, bove about guilt and innocence as the under/y-
classified as MEN at an earlier age than that at which females are classified mg cO~l1ltJve frame would not result in some of the frequent mistakes peoole
as WOMEN. A number of people, sensing that this usage pattern revealed make 1I1 the use of the wo~d SUSPECT. The word SUSPECT is supposed' to
attitudes toward females (or a history of attitudes toward females reflected be ~sed of a person who IS suspected of committing the crime in question;
in current conventional usage possibly in independence of the user's own for It to be ~sed appropriately, there has to be some specific person of whom
attitudes) which ought to be corrected. A number of speakers have succeeded
It.~an be said that th~t pers~n .is suspected by someone of committing the
in modifying their usage in a way which established the age boundary between 'Clime. The current J?urnalIs(]c use of SUSPECT even "vhen nobody has
the BOY to MAN transition at the same place as that between the GIRL to been 3.ccused of the cnme shows that the change is of the superficial kind,
WOMAN transition. The semantic change in this case is a real one, which foll~wJl1g the apphcatlOll of a rule of thumb that says, "Wherever I am
needs to be explained. But it would not be satisfying to see the explanation mclmed to say CULPRIT (etc.), 1 should instead say SUSPECT." I h,lve
solely in changes of the meaning of the words GIRL and WOMAN; the full 111 mmd such usages as can be found in reports like "Police investigating the
explanation must assign the change to the underlying schematization on the murder have found no clues as to the identity of the suspect."
part of the language user. The realities (of poeple of both sexes getting older)
havc not changed, nor have the availa bJc choices of linguistic material; Miscommunication by Frame Conflict
what has changed (in some speakers) is the underlying schematization, the
The law provides many contexts in which specific new framin2s need to
for familiar words. The notion INNOCENT. .mentit;Iled co~t)l,\C
circumstances motivating the category contrasts.
be constructed
• . •.
Relexicalizing Unchanged Frames lS an example. In both everyday language andlcgallanguage there is a con-
tradIctory opposition between INNOCENT and GUILTY. In everydav
A second kind of semantic change, which oddly call be illustrated with
language, the dJfference depends on whether the individual in question di~l
the same words, is one in which the links betvveen words and their frames
or dl~ not commIt the crirne in question. In legal language, by cOlltr(1st.
arc changed, but the underlying schematization remains unchanged. The
the dlfference depends on whether the individual in questi()]; has'or has !wt
errort to respond to society's new sensitivity to the connections between
been declared quilty by the court as ,\ result oflegal action within the criminal
language and attitudes is perhaps easiest to manage in the short run if it does
Justice system. This disparity of schunatization is responsible for frequent
not require something as deeply cognitive as a reschel11,lllzatJon of the
llllsunderstandings in the lise of these words. An example of such misu;1dcr-
domain. A superficial rule-orO-thumb for bringing about the appearance of a
standmgs (which I have discussed in Fillmore 1978) was in a conversati<'11
raised consciousness in the realm of language and sexism is a mechanical
betwe~n a prospective juror and lawyers in a voir dire hearing in a l11unicip;d
principle like "Where I am inclined to say GIRL I should instead say
:,ourt 1I1 Berkeley. The attorney for the defense asked the prospective juror
WOMAN". A person who adopts this rule may find that 111 most cases It
Do you accept the AmerIcan legal doctrine that a m,l1l is inllocent ulltil
performs very wcll; hut one sometimes finds oneself trapped-as in the ex-
 128 Charles J. Fillmore
'I
>

Frame Semantics 129

proven guilty?" The citizen answered that a person should be treated as

innocent until proven guilty, but that it would be strange to sa~ that he was
r flght, I would lose), then too you can be charged with FORCIBLE ENTRY.
A third example is ORAL AGREEMENT. Basically an ORAL AGREE-
actually innocent. The attorney asked again, say!ng, "I'm talkmg about the MENT is a contract or agreement which two parties entered into orally,
doctrine that a man IS innocent until proven gUilty. Do you or do you not that is, without putting the agreement in a written form and without signing
accept that doctrine?" The citizen answered that if the man IS innocent, our names to it. The importance of the notion ORAL AGREEMENT in the
then there is no need for a trial. (This rude answer excused the man from law is that the conditions of its authenticity and its bindingness distinguish
jury duty.) This little bit of miscommunicating could easily have been it from agreements that are fully written out and signed. The critical differ-
avoided. The citizen was not really being asked whether or not he accepted ence, for the given lcgal purposes, is the presence or absence of the signatures
a particular legal doctrine, but whether or not he was willing to adopt ~or the of the principals. The important part of the contrast, then, is that between
purpose of discussion in the trial which was about to start t~e .fram.m g .of being signed and not being signed. Accordingly, provisions made in the law
the words INNOCENT and GUILTY provided by the cnmmal Justlce for ORAL AGREEMENTS also apply to written agreements which happen
institutions in place of the everyday use of these same words. not to be signed. The prototype background in which the notion ORAL
AGREEMENT is motivated, is one in which agreements are either made by
Reformulations in Technical Language
word of mouth or by means of documents which are written and signed.
Legal contexts give us further ways of seeing chang~s .between general and In situations which depart from the prototype the law has needed to deter-
special-purpose framings of words. In many case.s t~IS IS because the ev~ry­ mine which aspect of the prototype contrast is legal1y the most salient (the
day sense of a word does not cover all cases 111 which It should be appropnate presence or absence of the signatures supporting a written document) and
to use the word. In the prototype case of events fitting the ,,:ord MURDER, let that be the criterion which specifies the contrast.
one person (A), intending to kill a second person (B), acts 111 su~h a ,:ay as
Frames for Evaluation
to cause that person to die. This prototype does not co~er a cas.e 111 whIch A,
intending to kill B. aims his gun at B, and kills C (who IS standmg next to B) One important area in which semantic interpretation depends crucially
instead. Some of the properties of MURDER relate A and B; o~hers relate on lexical framing is that of attributions of value. Evaluative adjectives can
A to C. The question somebody needs to answer, of course, IS whether, contain in their meanings reference to the dimensions, scales, or standards
for the purposes of the law, it is proper to say that A murdered.C. !he law according to which something is evaluated, as with adjectives like FRA-
does this, not by modifying the definition of MURDER so that It wIll c.over GRANT, TASTY,EFFICIENT,lNTELLIGENT,etc.In many cases,however,
this 'wrong-target' case, but by adding to the system of leg~1 sema~tlcs a an adjective is abstractly evaluative (as with the English words GOOD and
statutory interpretation principle caUed 'Transfer of Intent accordll1~. to BAD) and interpretations of their attributive use depend on knowledge of
which A's intcnt to kill B is fictitiously transferred to C so that ~he defil1ltlOn the ideational frames to which they are indexed. The fact that speakers of
of MURDER can fully fit what A did to ~. With res~ec~ t~ Judgments of English are able to interpret such phrases as A GOOD PENCIL GOOD
reprehensibility and legal provisions for pU11lshment, A s klllmg of C should COFFEE, A GOOD MOTHER, A GOOD PILOT, etc., shows that they
be treated in the same way as A's successful killing of B would have been. are able to call into their consciollsness for this purpose the fact that a
The Transfcr of Intent principle makes it possible for the non-prototypic pencil is used for writing and can be evaluated for how easy or efficient it is
case to faU under the same definition. . . to write with it, or how clearly its traces appear on the paper, the fact that
. . '111 t I1e 1aw" ,ll-e equally founded on mtentJons
Other such re111terpretatlons coffee is a drink and can be evaluated for its taste, its contribution to the
. . . I ' t f FORCIBLE ENTERY drinker's alertness, etc., that mothers and pilots do what they professionally
associated With the prototypical case. T 1e conccp 0 ,
involves one person gaining entry to another person s prope~·ty, by over- and convcntionally do and can be cvaluated for how easily, how effectively,
coming the resistance of persons trying to prevent th~t person s entIy. The and how effLciently thcy do it. The point was made earlier that cognitive
' .. f' I~ORCI BLE ENTRY however ll1c1udes not only the frames cal1ed on to assist in text interpretation may derive from genend back-
usual d etimtlon 0 . , ' - I
situation in which the intruder physically overpowers the othel, bU~. a ~~ ground knowledge or may be brought into play by the textual context. This
~he situation in which, as it is usually put, "resistan.ce woul? be unavaJ, ll1g . is particularly true in the casc of the interpretation of evaluative adjectives,
If you being twice my size and strength, insist on bemg admItted t~ my apart- since some nouns have frames associated with them whose evaluative dimen-
ment,'and I meekly let you cnter (on the reasonable grounds that If we had a sions are provided in advance, while others designate things that could be
 130 Charles J. Fill mOl e

evaluated only if the CO:ltext provided somc basis for the evaluation. When
we come across the phrase A GOOD STICK we expect to find in the context
I
{ 5.
Frame Semantics

Frame-Semantic Formulations of Issues in Technical Semantics

131

In this section I examine a small number of topics that one traditionally

some explanation of a situation within which one stick could function better
finds in standard treatises on technical semantics: proportionality, para-
than another (for propping a window open, for repelling a raccoon, for
digms, taxonomies, syncategorematicity, the supposed contrast between
skewering marshmallows, etc.). A general concept of 'framing' involves con-
'dictionary' and 'encyclopedia', the goal of descriptive simplicity and redun-
textualiz~lg or situating events in the broadest sense possible; within linguis-
dancy elimination, and, lastly, the troubled notion of ' lexical presupposition'.
tic semantics proper the concern is with patterns of framing that are already
established and which are specifically associated with given lexical items Proportionality
or grammatical categories. One of the most frequently used heuristic devices for discovering and
demonstrating the existence of semantic features in the vocabulary of a
Script Evocation
language is that of setting up a proportionality involving four words and
I said earlier about cognitive frames that to speak of one of its elements asking for intuitive agreement about the identity of pairwise differences
is to speak of the others at the same time. More carefully put, to speak of among them. Believing that man is to woman as boy is to girl, we set up the
one part of a frame is to bring to consciousness, or to raise into question, ratio MAN:WOMAN: : BOY: GIRL. Others frequently used are COME:
its other components. This effect is particularly striking in connection with GO: : BRING: TAKE, LOOK: SEE: : GLANCE: GLIMPSE. INHALE:
the kinds of frames known as 'scripts', frames whose elements are sequenced EXHALE: :SNIFF: SNORT, and MAN :WOMAN: :BACHELOR: SPIN-
types of events. Text understanding that makes use of sc~ipt~l knowledge STER. The approach which sees the basic semantic relations as holding
(on which see Schank and Abelson, 1977) involves the activatIon of who1e- among words taken in isolation fails to help us become aware of the possihly
scale scripting of cvents on the presentation of an event that can be seen to quite separate ways in which individual members of these proportions are
part of such a script. Thus, in a textlet like fitted onto, or frame, their reality. I have already pointed out that in many
"He pushed against the door. The room was empty." people's speech the differentiating criterion for BOY 'Is. MAN might be
we make the two sentences cohere by assuming that the goal somebody importantly different from that for GIR L 'Is. WOM AN; BRING is separate
might have in pushing against a door is to get that door open, and that if enough in its semantics from COME for it to have acquired quite sCpcHate
on~ succeeded in getting the door open by such an act, one could then be patterns of dialect variation; and the motivation for the categorics BACHE-
in a position to l~otice whether the room was empty. Reading betvveen the LOR and SPINSTER appear to be considerably different, in spite of one's
lines, we expand the text to mean: inclination, as a systematizer, to put the two words together. One might wish
"He pushed against the door. THE DOOR OPENED. HE LOOKED to propose that the abstract structural patterns underlying these word groups
INSIDE. H E SAW THAT The room was empy." are simple and straightforward, in the ways suggested by the proportions,
even though certain facts about the world make the domain look less orderly.
Frames for Texts I think such a proposal is not helpful, because it is not one which asks the
Discussion of text structure on the part of Robert Longacre and others analyst to look for the background and motivating: situations which sepa-
shows that languages or cultures can differ with respect to the ways in which rately give reasons for the existence of the individual categories. (One by one.
texts v/ith particular communicative goals can have particuhlr convention-
Paradigms
alized forms. Recipes in English make consistent use of imperatives. In Hun-
aarian recipes, first person plural descriptions are the norm. And Longacre A prime example of semantic structure among lexical items is the 'para-
has described (ill conversation) a language lacking in procedural discourse digm'; and the best example of a lexical-semantic paradigm i,; tlw kind cf
uses narrative form for :iuch purposes. Here it would be difficult to believe display of livestock terms represented by Table I.
cattle sheep horse S\V! nc
that languages c1iffer from each other in the presence of material usable for
particul~r kinds of discourse, it seems rather to be the case that traditions cow ewe mare
of language usc within the culture devclop In different ways 111 texts With bull ralll stallion bo;_,r
steer wether gelding baITO\\
dilTerent communicative goals.
Table
 Frame Semantics 133
Charles .J. Fillmore
132

experiences. VERTEBRATE and MAMMAL arc terms whose enmlovment

Hcre the propos,,] that we luvc a closed system of terms tied together by
fits a particular kind of interactional or contextual schema (that of' sci~lltijlc
such "eatmes as General, Female, Male, and Neuter. cross-cut by feat~res
discourse),while RETRIEVER as a category occurs most naturally as an answer
identifying species (Bovine, Ovine, Equine, Porcine), seems very attractive.
to a question about what kind of a dog one has. Suppose that you, hearing
Unfortunately the display disguises many facts about both these words and
a splash. in my back yard" were to ask me what that noise was, and suppos~
the domain which they appear to cover. CATTLE and SWINE are plurals;
the fact IS that my pet retnever fell in the family swimming pool. As a way of
SHEEP and HORSE are not. The words WETHER and BARROW are
explaining the source of the noise, it would be natural for me to say '~An
known onh to specialists. In the case of CATTLE, COW and BULL appear
animal fell 111 the pool" or "A dog fell in the pool", but it would be very
to have the'statu,; of ' basic level objects' (in the seme of Rosch 1973), whereas
unnatural for me to say "A vertebrate fell in the pool" or "A mammal fell
the !:!cneral terms have that function in the case of SHEEP and HORSE.
In tl~e case of SWINE, a word not in the table, namely PIG, is the best candi-
in the pool", and unnatural in a different way for me to say "A retriever fell
in the pool". The latter three terms seem to appear more natural in utterances
date for 'basic level object' status. used in acts of classifying, but seem unnatural when used in acts of referring.
In short, the regularities apparent in the paradigm (and this set. of terms-
This functional difference is not revealed within the logic of a stand'lI:-d
together with terms for young, newborn, etc.-make up wha~ IS g~nerally
taxonomic tree.
accepted as the best example of a semantic paradigm) are rr:lsieadillg. To
which we ought to add the Neuter category of the wo~ds 111 the ?ottom Syncategorematic Terms
row is not just a 'neutral' category operating in the same Ime o~ busl11~SS as
. It has frequently been discussed (e.g., Austin 1964, Lecture VII) that a wore!
the categories Female and Male. The category is differently n:otlvated III the
lIke .IMI~ATION does not semantically modify a word it grammatically
dilTerent species, which is another way of saying that one has dlff~rent reasons
modIfies In the standard 'set intersection' way. Rather, it combines with the
for castrating a bull and a horse, one might do it at different (relatIve) ages, etc.
mean,ing of its partner to form a fairly complex concept. Something correctly
descnbed as IMITATION COFFEE looks like coffee and tastes like coffee
Taxonomies
The next most common kind of lexical semantic forma! structure is the and it looks and tastes like coffee not by accident, but because somebod;
'semantic taxonomy', a semantic network founded on the relation 'is a kind manufactured it so that it would have these properties; but, whatever it is
it is not made of coffee beans. Understanding the category, in fact, require~
of'. Scientific taxonomies have obvious uses in scientific discourse, and
understanding the role of coffee in our lives and (perhaps) the reasons
research that has led to the uncovering of folk taxonomies has been among
someone might have for making a coffee substitute.
the most important empirical semantic research yet done',But there are two
Ry contrast a word like REAL appears to contribute nothing at all to
aspects of taxonomic structures that argue against regardll1g them as repre-
the noun to which it is attached as a modifier. To describe something as
senting merely a formal system of relationships founded on a sll1gle clear
REAL . COFFEE is to do nothing more than to assert that somethinGc is coffee'
.,
semantic relation, The llrst is that at difTerent levels 111 a taxonomy the com-
agamst the background of (the possibility of) somebody's suspicion that it
Illunity might have had different reasons for introducing the categones; the
is imitation coffee. As with IMITATION, a p(lrt of a full understanding of
second is tll".\ the usual tree-form display of the elements of a tax~nomy
an expression with REAL is knowing the reasons one might have for ~ro­
docs not show how it is that particular elements in the taxonomy are cog11l-
tively privileged categories' in important ways ..Both of these pOints b~
C,'ll: viding substitutes for the thing in question. The notion REAL COFFEE
illustrated with a 'path' in ~1 taxonomy of zoological terms In EnglIsh, ndn.e1:
makes sense to us becausc we know that in somc settings cofTee is scarce,
and we know that some people find coffee damaging to their health or held
ANIMAL
offensive by their religion. We can understand a category like REAL GOLD
VERTEBRATE
or REAL DIAMOND because we can imagine a reason why somebodv
MAMMAL
might choose to produce fake gold or fake diamonds, and we GIn ima!.'ill~
DOG
why someone might have doubts about the authent;c:ty of particular SaInl~les.
RETRIEVER
By contrast, a notion like REAL PANTS is unintelligible, because it is imnos-
Of this set of words, DOG and ANIMAL seem to be the cogniti'vely privi-
sible to imagine something looking like pants and functioning like P~lilts
leeed cateeories, privileged in the sense that tbey are the words t.h~t woul,d
which do not, by virtue of those properties alone, count as being genuine
m"asl C'rdil~~1rily be used when in everyday natural talk one is descnbll1g one s
 134 Charles .J. Fillmore Frame Semantics 135

pants. approaches value simplicity and frame-semantic approaches do not, there is

another sense in which simplicity of description is enhanced bv the frame
Redundanq Elimination
semantics approach. A recent lively discussion between Paul Ka~ and Linda
A common goal in structural semantics is the elimination or minimization Coleman on the one hand (Coleman and Kay 1981) and Eve Sw~etser on the
of redundant information in the semantic description of lexical items. Fre-
quentlya semantic theorist will declare that the goal of a 'semantic dictionary'
1i other h,md (Sweetser 1981) concerns the possibility of a prototype back-
ground of assumptions (or, as Sweetser calls it, a 'folk theory') as providing
is that of saying just enough about each word in the language to guarantee the grounding for a simplified definition of the noun UE. On the Kay/
that it is semantically in contrast with each other word in thc languagev Coleman account, a LIE is something which is (I) false in fact, (2) believed
(Bendix 1966). It is a goal which presupposes the analyst's ability to have by th.e speaker to be false, and (3) said in order to deceive. Sweetser's sugges-
an overview of the entire lexical repertory of the language. Such a goal is tIon IS that if we can characterize a folk theory of human communication
completely antithetical to the goals of frame semantics, since frame semantics involving cooperation, expressing what one believes, etc., then jt is possible
aims at discovering what categorizing functions the word serves in the con- to describe a UE as simply a 'false statement". those other understandings we
texts in which its use is motivated. This kind of knowledge is in principle have about the concept falling out through an understand in!! of whv one
attainable independently of knowledge about other words in the language, would bother to produce a false statement. ~.
except for those relatively few cases in which the 'mosaic' image is appro- Presupposition
priate, the image by whieh the meaning given to anyone word is dependent
on the meanings of its neighboring words (as in Trier 1931). Claims about 'presuppositional' information being associated with indi-
vidual lexical items have not received a good press. I find that within frame
Dictionary vs. Encyclopedia sema~1tics, the concept of lexical presupposition does not seem unjustified.
The various structuralist approaches that lind a goal of redundancy eli- ConSIder the case of a verb like English CHASE, a verb for which a lexical
mination relevant, also find it intelligible to draw a clear distinction between presuppositionist might be incl ined to say that when it is used of two beings
'dictionaries' and 'encyclopedias'. In particular, certain scholars insist on a ~oving in the same course, the movement ofthe.on~ in front is presupposed,
distinction between purely semantic information about words and encyclo- Illd~pendently of w~1ether the movement of the mdlvldual designated by the
pedic information about the designata of words. Somebody holding this vie\v subject of the verb IS asserted, denied, questioned, or supposed. ]n a settin!! in
might expect to be able to justify certain characteristics ofcaJventers (or the :Vhich.one person is running, especially where it is understood that that per~on
concept CARPENTER) as belonging to the semantic category of the noun, IS ~eell1g, 1t IS relevant to consider whether some other person is or is not
other distinct characteristics of carpenters as simply being true of the indivi- ~omg to try to prevent that first person from getting away. (My illustration
duals who satisfy the criteria associated with the category. A frame-semantic IS With people, but that's not an important condition.) The verb CHASE
approach would rather say that communities of men contain individuals who eXIsts a~ a ~ateg~ry by recognition of such relevance. If 1 ask. "Did anybody
bv tradc make things out of wood, using particular kinds of tools, etc., etc., chase hlm?'- or If! say "We didn't chase him", our reason for understanding
a;1d would note that these people are called CARPENTERS. The possibility that 'he' was running (fleeing) is that we know the kind of situation <wainst
of separating some features of a full description of what carpenters do as re- which the category CHASE has a reason for being. It is in that sense. it ;eems
lated to the concept and others as related to the people does not seem impor- to me, that one can talk about lexical presuppostions. .
tant. Therc is a distinction to be made between knowledge about words and 6. Concluding Remarks
kno\\led!!e abollt things. but it is not to be made in a way that serves the
interests ~of thc semanticists [ have just been describing. True 'encyclopedic' In this paper I have argued for a view of the description ofme;1J11ng-bearing
i 11 formal ion about Cl rpenters as people might say something about wages, elemcnts 111 a language accordll1g to which words (etc.) come int'J being only
union ~1l111iations. job related diseases, etc.: such information is not a matter for a reason. that reason bemg anchored in human experiences and human
of dispute. institutions. In this view, the only way in which people can tJuly be said to
understand the lise to which these meaning-bearing elements aJ:e being put
Simplicity of Description In actual utterances IS to understand those experiences and institutions and
While in respect to redundancy elimination it has appeared tbat standard to know why such experiences and institutions gave people rea~'om to create
 Charles J. Fillmore Frame Semantics 137
136

the categories expressed by the words. The semanticist's job is to tease out Current Issues in Linguistic Theory. Bloomington: Indiana University Press.
thc precisc nature of thc relationship betv.'een the word and the category, and Fillmore, Charles J. (1978), On the organization of semantic information in the
the precise nature of the relationships between the category and the back- lexicon. In Papers from the Parasession on the Lexicon. Chicago: The Chicago
ground. I believe that somc of the examples 1 have offered have shown the Linguistic Society.
advantages of looking at language in this way. Fries, Charles C. (1952), The Structure of English. New York: Harcourt, Brace &
World.
Gross, Maurice (1975), Methodes en Syntaxe. Paris: Hermann.
Note
1. For a rCCCill attempt to dilfcrentiJte these terms, sec Beaugrande i981, p. 303. I Helbig, Gerhard and Wolfgang Schenkel (1973), Worterbfich zur Valenz und
Distribution deutscher Verben. Leipzig: VEB Verlag Enzyklopa

I
Householder, Fred W., et al (1964), Linguistic Analysis of English, Final Report
on NSF Grant No. GS-J08.
References
Katz, Jerrold J. and Jerry A. Fodor (1963), The structure of a semantic theory.
Arnheim. Rudolf (1969), Visual Thinking. Berkeley: University of California Press. Language 39.170-210.
Austin, John L. (1964), Sense and Sensibilia. (Reconstructed from manuscript Lakoff, George and Mark Johnson (1980), Metaphors We Live By. Chicago:
notes by G. J. \Varnock.) Oxford University Press. University of Chicago Press.
Beaugrande, Robcrt de (1981), Design criteria for process models of reading. Langacker, Ronald W. (forthcoming), Foundations of Cognitive Grammar.
Reading Research Quarterly 16.2.261-315.
Bendix, Edward H. (1966), Componential Analysis of General Vocabulary: The
Semantic Structure of a Sel of Vcrbs in English, Hindi and Japanese. Bloom-
ington: Indiana University Press.
Berlin, Brent and Paul Kay (1969), Basic Color Terms. Berkeley: University of
II Lees, Robert B. (1960), The Grammar of English Nominalizations. The Hague:
Mouton
McCawley, James D. (1975), Verbs of bitching. ]n David Hockney et al. (eds.),
Contemporary Research in Philosophical Logic and Linguistic Semantics. Dor-
drecht: Reidel. pp. 313-332.
California Press. Pike, Kenneth L. (1967), Language in Relation to a Unified Theory of the Structure
Chomsky, Noam A. (1957), Syntactic Structures. The Hague: Mouton of Human Behavior. The Hague: Mouton.
Chomsky, Noam A. (1965), Aspects of the Theory of Syntax. Cambridge (Mass.): Rosch, Eleanor H. (1973), On the internal structure of perceptual and semantic
M.1.T. Press categories. In Timothy E. Moore (ed.), Cognitive Development and the Acquisi-
Coleman, Linda and Paul Kay (1981), Prototype semantics, Language 57,26-44. tion of Language. New York: Academic Press.
Downing, Pamela (1977), On the Creation and Use of English Compound Nouns. Rosenbaum, Peter S. (1967), The Grammar of English Predicate Complement
Language 53. 810-842. Constructions. Cambridge (Mass.): M.LT. Press
Fillmore, Charles J. (I 96 \), Indirect Object Constructions in English and the Order- Schank, Roger C. and Robert P. Abelson (1977), Scripts, Plans, Goals and Under-
ing of Transformations. The Hague: Mouton. standing: An Inquiry into Human Knowledge Structures. Hillsdale: Lawrence
Fillmore. Charles J. (1963), The position of cmbedding transformations in a gram- Erlbaum.
mar. Word 19.208-301. Sweetser, Eve E. (1981), The definition of Lie: An examination of the folk theories
Fillmore, Charles J. (1968), The case for case. In E. Bach and R. Harms, eds., underlying a semantic prototype. Unpublishel tns.
Universals in Linguistic Thcory. New York: Holt. Rinehart & Winston. Talmy, Leonard (1980), Grammar and cognition. MS. University of California at
Fillmore, Charles J. '( Ill71), Verbs of judging: an exercise in semantic description. San Diego, Cognitive Science Program.
In C. J. Fillmore and D. T. Langencloen, cds., Studies in Linguistic Semantics. Tesniere, Lucien (1959), Elements de Syntaxe Structurale. Paris: Klingksieck.
New York: Holt, Rinehart and Winston. Trier, .lost (1931), Der deutsche Wortschatz im Sinnbezirk des Verstancles. Heidei-

I
Fillmore. Charles J. (1972). On generativity. In Stanley Peters, ed.. The Goals of berg.
Lin51uistic Theoi'Y. Englewood ClifTs: Prentice Hall. Wilson, Deirdre (1975), Presupposit ions and Non-Truth-Conditional Semantics.
Fillmt;rc. Charles J. ( 1975), Santa Cruz Lectures on Deixis. Bloomington: Indiana London: Academic Press.
University Linguistics Club. Zimmer, Karl E. (1971), Some general observations about nominal compounds.
Fillmore, Charles .f. (1977a). The case for case reopened. In P. Cole and J. Sadock, In Working Papers 011 Language Universals, 5. Language Universals Project,
(cds.). Syntax and Semantics X: Grammatical Relations. New York: Academic Conlmittee on Linguistics, Stanford University, pp. 1-24. Reprinted in Wort-
Press. bildung (1981), Wissenschaftliche Buchgesellschaft Darmstadt, pp. 233 257.
Fillmore, Charles J. (1977b), Topics in Lexical semantics. In Roger W. Cole (cd.),