Approaches To The Study of Word Meaning The Legacy of Historical Semantics
Approaches To The Study of Word Meaning The Legacy of Historical Semantics
Approaches To The Study of Word Meaning The Legacy of Historical Semantics
Curs 1
Morriss description of the functioning of the sign: the presence of the form
causes the concept, while the concept sends to an object. Forms are indirectly related
to objects, by means of concepts.
There are several points of entry in the study of meaning, which define several
major approaches to semantics.
a. Referential Semantics is devoted to the study of the relation between
linguistic forms and the extra-linguistic entities they denote.
Referential semantics is primarily concerned with sentence meaning, word meaning is of
interest only to the extent the word contribute to the denotation and truth of the sentence.
Here are a few familiar statements that are part of referential semantics:
(2)
He kissed Mary.
He wished to kiss Mary.
>
He kissed Mary.
Since my arms are full, will you open the door for me?
My arms are
full.
b. Structural Semantics is devoted to the study of the relation between
linguistic forms and the corresponding concepts. Indeed, in the picture above, if the
object is ignored, the familiar Saussaurean definition of the sign emerges:
Syntax
Semantics
Pragmatics
form
concept
Defining the sign in the semiotic tradition (The scholastic tradition, Morris (1933)
Foundation of the Theory of Signs)
(2)
concept
object
signifier
signified
form (word)
a) they are part of lexical paradigms (semantic fields)
a.
b.
c.
d.
trajectory (TR) for the moving entity and landmark for (LM) for the background against
which movement occurs. Over has 3 related senses:
a) the above-across sense
b) the above sense
c) the covering sense
2. Lexical Semantics vs. Historical Semantics
2.1 Prior to the advent of structuralism, which brought about a clear separation of
synchrony and diachrony, but also after that, meaning had been and still is investigated
from a historical perspective: two disciplines dealt with this problem:
a) historical semantics (lexicology) the evolution of meaning, the typology of meaning
changes, the causes of semantic changes;
b) etymology (it reflects the hermeneutic bias for the earliest meaning as the truest,
unadulterated meaning of the word)
lord/lady are etymologically related to loaf of bread
2.2 Historical semantics
Semantics was first used to refer to the development or change of meaning, the attempt
to classify semantic changes and to establish the causes (internal and external) of
semantic changes. A representative work in this line is Sterns Meaning and Change of
Meaning (1931).
The great storehouse of semantic change in English is the OED, originally, New
English Dictionary on Historical Principles (1883-1928), containing 411, 825 words + a
Supplement of 63 000 words in (1972-1986)
2.3 Ullmans typology of semantic change
A possible typology of semantic changes is that proposed by Stephen Ullman (Semantics,
An introduction to the Science of Meaning, 1963).
Meaning is defined by Ullman as the relation between name and sense (under the
influence of de Saussure). Ullmans postulate is Leibnizs Natura non facit saltus, in
other words, no matter what causes bring about the change, there must always be some
connection, some association between the old meaning and the new one.
Semantic changes naturally fall into two categories:
a. those based on an association between the senses;
b. those based on an association between the names (forms);
Within each category, the association may be one due to similarity or due to
contiguity. A four fold division of changes emerges:
1. CHANGES BASED ON SIMILARITY OF SENSES (metaphors); a metaphorical change
is mediated by some common ground or similarity
(6)
Understanding spatial preposition requires the possession of certain image
schemata. In our particular case what is required is a path image schema: using the terms
common ground
/
tenor
\
vehicle
a fox ( a sly person); a bear ( a morose person); a book worm; a cat (a spiteful
woman), a calf ( a beginner), a chicken (a young woman), a monkey (a naughty
child), an oyster ( a silent person).
c. synaestehtic metaphors:
(13)
Sally is a block of ice
a. part used for the whole or whole used for the part (synaechdochy)
Ullman (1931) also describes a few categories of metaphors common in the vocabulary of
most languages.
a. antropomorphic metaphors are based the features of body parts:
a.1 they extend the human body to the realm of animals and plants:
(8)
brow of a hill, ribs of a vault, mouth of a river, head of a bed / head of a table, the heart of
matter, etc.
(14)
to be a good sword
the horse (the cavalry)
the crown (the King)
c.
the holder for the thing held, the container for the contained
(16)
a.2 the opposite tendency is shown in metaphors where the body parts are named
after plants, objects, animals:
(9)
muscle [ Lat. mus, mouse]; apple of the eye, ear-drum
b. animal metaphors are based features properties of familiar animals.
b1. similarities between animals and plants, the latter being named after the
former:
(10)
dogs tail
cocks foot
goats beard
(11)
pipetnri
golomoz, nuduroasa
barba caprei, creua
cock of a gun
crane
cat-o nine tails
cat-head
(17)
cocosul armei
boon (AS)
1.
prayer
2.
a blessing, something to be grateful for; the second
meaning developed under the influence of the homonymous French adjectives boon (bon,
good)
industrial action
strike
recession
depression
explosive device
bomb
liquidation
murder
strategic weapon
nuclear missile
The above classification of semantic changes calls for the following comments:
a. Only the first two conceptual types are productive and somewhat systematic.
b. Many changes fit more than one class:
(19)
(23)
He bought a Picasso.
(Is it ellipsis, i.e. contiguity of names, or metonymy, i.e. contiguity of senses?)
(24)
Changes of denotation
a. Words may undergo generalization they take on a broader range of
meanings:
2.4.1
educationally subnormal
certically challemged
(25)
BOX
AS
a container made of box-wood, normally safe-keeping of something
precious, such as ointment, jewelry (currently, casket, chest)
A box at the opera is a facetious extension of the basic sense.
gear box technological extensions:
BED
Flower-bed
bird
weak, feeble
simple, ignorant
feeble-minded
(26)
ENGINE
medieval sense:
mechanical contrivance often for war or torture cotton
gin
mechanical source of power
steam engine
FUGOL
OE
1. bird
Mod E 2. fowl
STEORFAN OE
1. die
Mod E 2. starve, die/suffer from hunger
DEOR
800
1000
1225
OE
1. beast
Mod E 2. a certain beast, the deer
Changes of connotation
From the multifarious changes in connotation that may affect a word, at this point
we focus on the following to: a) changes in emotional overtones (Ullman) b) changes of
register.
2.4.2
blame
(28)
2.
1.
2.
3.
1
2.
ME
ModE
KNAVE
(27)
CHURL
OE
ceorl
1.
a servant
a very high executive
1.
room-mate (Spanish: camarada)
friend
plague (the bubonic plague
garden pests (insects, animals R daunatori)
annoying person
blasphemy
blame
Deterioration
a word takes on pejorative association
SILLY (OE saellig)
900
OE
saelig
happy
1200 ME
seely
happy, blessed
1500 ModE
deserving of compassion
10
2.5
11
Of the original Anglo-Saxon word-hoard, one third has survived, some of them
have shifted their denotation as shown above) or connotation. Notice thus that the
updating of forms and meanings of major terms from AS heroic poetry produces some
strange results. By such means it might be said that Beowulf took lust (AS lust pleasure,
joy) in dreary (A-S drearig, bloodstained) battle, was moody (AS, modig, brave,
spirited), rode a mare (AS mare, steed), yelped ( gielpan, to boast or challenge), fazed
his enemies (AS fesian, to put to flight), and then cringed (AS cringan, to die, fall in
battle). What are the current meanings of these terms?
(32)
Examples of AS terms
man, wife, child, house, bench, meat, grass, leaf, fowl, bird; good, high,
strong; eat, drink, sleep, live, fight,
functional words:
be, have; back; that; ic, thou, he, we, etc.
sky, skin, skill, scrape, scrub (Compare shirt (AS) / skirt (Scandinavian))
bloom, gift, bank, birth, crook, dirt, egg, fellow kid, leg, sister, skill, root,
reindeer, slaughter, snare; old, sly; take, thrive, thrust
12
(36)
Norman
question
Latin
feline
porcine
equine
Doublets
(38)
Synonymic patterns
AS
ask
AS
catty
piggish
horsy
(37)
ascend
physician
These borrowings should not mislead us. For essential purposes we are still
dependent on the Anglos Saxon core of words and the ancient assimilated borrowings,
which represent a nucleus or central mass of many thousand words whose Anglicity is
unquestioned.
AS
calf
cow
sheep
deer,
boar
pig
mount
doctor
In modern times, the lexicon has reflected the ages of exploration and colonialism
by becoming increasingly polyglot and cosmopolitan. The following everyday words,
erstwhile linguistic tourists, now jostle together with the natives as assimilated
immigrants:
(34)
rise
leech
Latin, Greek
interrogate
13
Warm, rich and full of golden-goodness will give your furry friend health,
strength and get-up and go;
The patient is experiencing a potentially fatal hemorrhage situation
(the patient is bleeding to death)
The familiar characterization is that high register terms are polysyllabic, while low
register terms are commonly short is essentially true. This is seen most simply in the
evolution of terms by clipping:
(39)
ME physiognomy becoming phiz (1688), subsequently fizz;
ME lunatic, becoming vulgar loony (1872);
fanatic becoming fan (US)(1889);
Renaissance obstreperous becoming stroppy (1951) and so on.
14
(39)
literary
archaism
higher
academic technical
central
or
neutral word
neologism
lower
slang colloquialism
1. business
2. budget
(40)
3. duty
4. embezzle
insane
of unsound mind
not in full possession of ones faculty
possessed
neurotic
psychotic
demented
unhinged
bereft of reason
non compos mentis
mental
barmy
cuckoo
gaga
batty
3. Groups of terms support the familiar generalization that capitalism liberated private
enterprise which had hitherto been restricted under feudalism. The more restrictive terms,
denoting traditional obligations to the Crown and the State are en force up to 1500: yield,
tithe, debt, tax, levy, revenue, customs, excise, duty, monopoly, trademark, account.
maladjusted
MAD
account
bananas
bonkers
crackers
round thee bend
off ones chump
debt
embezzle
custom
potty
loony
(41)
(42)
tithe
yield
nuts
daft, crazy
4. From 1500 on one finds a growth of terms or of dominant senses suggestive of money
as a source of freedom: capital, cash, purchase, credit, finance, fortune, invest,
concession, speculate, entrepreneur.
15
16
5. Institutional terms: guild (1000), exchequer (1300) ( from the quadrangular square on
which the transaction takes place, exchange (1335 le table deschange soit a Dovorri
Act 9 of Edward III), bank (1550), trade, traffic (1550). It is around 1700, the time when
the Bank of England is founded, that much of the institutional vocabulary emerges, in the
form of specialized terms: economy, fund, security, concern note, stock, cheque,
consumption and demand; soon to be followed by budget, currency and draft.
1760
1800
1850
1900
chattel
capitalist
scab
exploitation
trade union
entrepreneur
boom
devaluation
reschedule
ME
1.
cattle
2.
chattels (movable property)
feoh (cattle, property, money) ModE (fee) money
feolaga (business associate, one who lays down feoh, or money)
farm a sum of money paid as a rent tax or the like
Mod E lease; tract of land for cultivation
17
18
Curs 2
The Development of Structuralist Semantics
Basic Concepts
1. Aim of the course: an introduction to Lexical Semantics.
The domain of Lexical Semantics is the study of word meaning as embodied in the words
of any natural language (English, Romanian). As the name lexical semantics suggests,
this discipline is a merger of two fields of study: lexicology and semantics
1.1 Lexicology
is the synchronic or diachronic study of the vocabulary of the
language. Problems such as the constitution and growth of the vocabulary, means of
enriching the vocabulary, including word formation rules and borrowings, lexical
relations , all belong yo the domain of lexcology.
1.2 Semantics and Semiotics
1.2.1 Semantics is generally defined as the science of meaning (from GK. semantikos
significant). Semantics as an independent discipline came into being in the second half
of the 19-th century. Michel Bral coined the term in the title of his pioneering study, La
Smantique, in 1895. Semantics was primarily viewed as the science of word meaning,
and according to Bral semantics was not primarily concerned with the historical change
of meaning.
1.2.2 Before the advent of linguistic semantics, the problem of meaning has been a
traditional concern of philosophers, logicians and psychologists.
In the XX-th century, semantics is approached from the more general perspective
of semiotics, the study of signs, including linguistic signs.
Here are a few familiar statements that are part of referential semantics:
(2)
He kissed Mary.
He wished to kiss Mary.
>
He kissed Mary.
Since my arms are full, will you open the door for me?
My arms are
full.
b. Structural Semantics is devoted to the study of the relation between
linguistic forms and the corresponding concepts. Indeed, in the picture above, if the
object is ignored, the familiar Saussaurean definition of the sign emerges:
Syntax
Semantics
Pragmatics
form
concept
Defining the sign in the semiotic tradition (The scholastic tradition, Morris (1933)
Foundation of the Theory of Signs)
(2)
concept
object
signifier
signified
form (word)
Morriss description of the functioning of the sign: the presence of the form
causes the concept, while the concept sends to an object. Forms are indirectly related
to objects, by means of concepts.
There are several points of entry in the study of meaning, which define several
major approaches to semantics.
19
20
a.
e.
f.
g.
21
22
(1)
single structure the tri-functional systems of sovereignty, war and fertility. The
same tri-functional structure is further modeled by social relations:
b. Lexical gaps One language may lexicalize a meaning that fails to be lexicalized
in another language. Think of the English heartburn or the Romanian dor.
c. Incongruity of lexical bounderies
In the general case, the boundaries between the meaning(s) of what seem to be
equivalent lexemes from different languages are incongruent. The whole problem of what
constitutes semantic equivalence between lexemes of different languages is complex and
controversial; as will eventually appear, it ultimately depends on the cultural equivalence
of objects, institutions and situations. Denotational equivalence is the only side of
meaning which is relatively independent of the cultural context. Consider the possible
French/Romanian translation of the harmless Russellian sentence:
(2)
Is it a door mat that is being referred to (Fr paillasson) or a bedside mat (Fr
descente de lit) or a small rug (Fr. tapis). There is a series of lexemes in English: mat,
rug, carpet and a series of lexemes in French, tapis, pailleson, carpette, descente de lit
and none of the French words has the same denotation as any one of the English lexemes.
The denotation of a lexeme is limited by the relations of sense which hold
between it and other lexemes in the language. The denotation of mat is limited by its
contrast in sense with rug and carpet.
2.2 (optional) On the notion of structure The relevant notion of structure for
structuralism is that of very abstract mathematical structure.
Bourbaki We can now clarify what is to be understood by mathematical
structure. The feature common to the various notions ranged under this generic heading
is that they all apply to sets of elements, the nature of which is not specified; in order to
define a structure, one or more relations involving these elements may be taken into
consideration (it may then be postulated that this or these relations fulfill certain
conditions, to be enumerated), relations which are the axioms of the structure envisaged.
To develop the axiomatic theory of a given structure is to deduce all the logical
consequences of its axioms, forbidding oneself any other hypotheses concerning the
elements under consideration, and especially any hypotheses with regard to their
particular nature.
In other words, structural analysis begins with the structure, with relations defined
in a purely formal way by certain by certain properties; these relations characterize a set
of elements the nature of which is not specified. From the basis of that structure, thus
established, the analysis demonstrates that a certain cultural content (a myth, a kinship
term system) is a model of that structure or a representation of it. What has been
demonstrated is only that this content is isomorphic to other contents, all of which model
the same relations. Structure is precisely that which holds good in an isomorphism
between two sets of contents. Structuralism is a comparatist method. The cultural
anthropologist Dumzil compared one pantheon (set of Gods of a culture) with another,
rather than one God with another. All the contents analyzed appeared as models of a
23
(3)
----------------------------------------------------------------------------------------------------------structure
sovereignty
war
fertility
mythology
Jupiter
Mars
Quirinus
-------------------------------------------------------------------------------------------------------------social order
oratores
bellatores
laboratores
--------------------------------------------------------------------------------------------------------------------To quote Descombes (1980: 85) For a given cultural content, whether it be god, table, or
basin, an analysis is structural when and only when it presents this content as a model.
Against this notion of structure, one can better understand the Saussurean dichotomies
2.3 The Saussurean Dichotomies and the Study of Meaning
a. The synchronic/diachronic dimension
The synchronic linguistic system is a
theoretical construct of the linguist and it rests upon the more or less deliberate and to
some degree arbitrary discounting of variation in the language behavior of those who are
held, pretheoretically, to speak the same language.
The need to separate these two dimensions is apparent in the etymological fallacy: the
common belief that the meaning of words can be determined by their origins. In fact, the
etymology of a lexeme is, in principle, synchronically irrelevant. What can we make of
the information that curious form Latin careful, fastidious.
b. The substance form dichotomy
Language results from the imposition of structure on two kinds of substance:
sound and thought. The phonological composition of a word-form is a complex of
phonemes, each of which derives its existence and essence from the structure imposed by
the language system upon the continuum of sound substance.
The conception of a language as form is clearly expressed in Hjelmslevs view of
the sign. Hjelmslev retains the opposition between the signifier and the signified. He
conceives of it as a relation between two levels of structure: Expression and Content,
arbitrarily defined by their mutual opposition. Any expression is the expression of a
content and, conversely, any content is the content of an expression. The expression
denotes a content. A language is thus viewed as a denotative semiotics:
(4)
Language
Expression
Content
24
for instance, the individual pronunciation of the phoneme [ ]. The form is defined as
constant and abstract.
At the level of content, the substance is whatever several inter-translatable
sentences have in common: I do not know.// Nu tiu.// Non so.// Ich weiss nicht.// Je ne
sais pas. The quoted expressions are the forms of this content, arbitrary relative to this
content; these forms inform the substance, producing the several substances of content.
Likewise, at the level of the expression, the substance could correspond to the
continuum of sound differently segmented by the phonology of each language. The signfunction is the association of the expression-form with the content form:
(5)
Substance (conceptual substance)
Content
Form (lexical concepts, syntax)
Language
Sign function
Form (phonology)
Expression
Substance (phonetics)
The relation between form and substance is one selection and manifestation, where the
substance manifest the form, and the form selects the substance
c. The paradigmatic/ syntagmatic opposition.
As far as lexical structure is concerned the following statements are relevant
1. The structure of language depends at each point (Merge) on the
complementarity between combination into a syntagm (V+ DP) and selection out of a
lexical paradigm: eat + DP/ [V eat]+ [DP an apple] // [V eat]+ [DP many cakes] //
2. The selection of one unit rather than another from a paradigmatic set is relevant
informationally: eat apples/plums/ cherries
3. It is often the case that one language will pack into a single lexical item (i.e.,
paradigmatically) information which is conveyed by means of a syntagm in a different
language:
Par arbitraire, lauteur entend quil est immotiv, cest a dire arbitraire par
rapport au signifi avec lequel il na aucune connexion naturelle dans la ralit.
The reasoning behind this famous assertion is, however, fallacious, through the
tacit resort to the 3d term which is not contained in the initial definion: the thing itself.
This is the relation of significance (denotation)
If one rightly admits that language is that language is form, not substance one has
to admit that linguistics is the science of forms exclusively: Or cest seulement si lon
pense lanimal boeuf dans sa particularit concrte et substantielle qu lon est fond
juger arbiraire la relation entre /bf/, /oks/ une meme ralit. There is, thus, a
contradiction between the way in which De Saussure defines the sign and the
fundamental nature that he attributes to it.
Between the signifier and the signified, the relation, far from being arbitrary, must
be described as necessary. Ensemble ils svoquent en toute circonstance. Lesprit ne
contient pas de formes vides, de concepts inomms. Cette consubstantialit du significant
et du signifi assure lunit structural du significant et du signifi.
If one accepts both the necessary relation between signifier and signified and the
arbitrary relation between the sign and the referent, one understand why the sign is both
immutable and mutable. The sign is immutable because, as it is arbitrary with respect to
the referent, the relation sign/referent, and implicitly signifier/signified, holds by
convention. A convention is a regularity of thought and behavior known by the members
of a community and which is observed only because everybody else observes it, since in
itself, it is not motivated by any other reason.(Lewis). The signifier/signified relation
cannot be questioned or challenged, hence the immutability of the sign at any given
synchronic moment.
At the same time, the sign is historically mutable. There are frequent changes of
meaning and form; that this is possible is also the consequence of the arbitrary nature of
the sign.
To quote Beneveniste once more:Ce nest pas entre le signifiant et le signifi que
la relation en meme temps se modifie et reste immobile; cest entre le signe et lobjet,
cest en dautres termes la motivation objective de la designation [the convention] qui est
soumise comme telle laction de divers facteures historiques. Ce que Saussure dmontre
reste vrai mais de la signification, nom du signe.
3.2.1 Absolute motivation the sign iconically reflects properties of the denoted object.
a) Primary onomatopoeia is defined as the imitation of (natural) sound by sound; the
referent itself is or implies an acoustic experience which is more or less closely imitated
by the structure of the word (signifier)
25
26
The distinction between opaque and transparent words is not absolute. It is customary to
distinguish between absolute and relative motivation
(6)
buzz, crack, fizz, hum, flop, flap, plop, roar, squeak, squeal, gong, hoot, clank
coo, mioow
(9)
under+stand
under+sign
doctor
(metaphor)
c. Mixed motivation Morphological and semantic motivation more often than not occur
together, this is mixed motivation (the compositional analysis of words), but they are in
principle independent.
(11) mixed motivation
blackbird, writer,
morphological motivation alone
conceive,
confer,
deceive,
conception, undergo
semantic motivation alone
leg of a table
Motivated words evince different degrees of semantic transparency, from fully
transparent to fully opaque:
(12)
fully transparent
degree of opacity
partly transparent
partly opaque
Remark Sounds are not expressive in themselves. Onomatopoeia will arise only
when the expressive potentialities latent in a given sound are borught to life by contact
with a congenial meaning. This can best be seen by contrasting certain pairs of
homonyms, one of which is onomatopoeic, while the other is not.
(8)
understand
undersign
doc
fully opaque
non-smoker, illegal
blackbird
(two semantic indicators)
blakboard
ladybug (buburuza) one semantic indicator
cranberry (afina)
gooseberry
red herring
brown study
undergo
Morphemes like cran- (cranberry, bil- (bilberry, afina), rasp- (raspberry), goose
(gooseberry, agrisa); bull- (bullfinch (botgros), cf. finch pasare cantatoare), dor(dormouse), pad- (padlock), gang (gangway), under-(understand) are called semantic
tallies (cf. Cruise 1986) Semantic tallies combine with categorizers to form a semantic
constituent, by entailment:
27
Its a dormouse
Its a bilberry
Its a mouse
Its a berry
hrciog
afin
28
(15)
Thus, the general problem is that of compositionality, i.e., to what extent meaning
is compositional (transparent). Lack of motivation or compositionality is also the main
difference between idioms and metaphors
An idiom is an expression whose meaning cannot be inferred from the meaning of
its parts: to pulls someone leg, to cook someones goose, to be off ones rocker, round
the bend of the creak. An idiom is lexically complex but may count as one semantic unit:
to kick off the bucket = to die. Idioms may evince syntactic constraints (they may resist
word order changes, modification, etc), as a consequence of being apprehended as a unit
with semantic and formal internal cohesion. Compare:
(16)
(19)
Languages differ regarding the proportion of transparent and opaque words they
possess. For instance, French is more opaque than Latin (loss of motivation due to sound
changes) and also more opaque than English. English is more opaque than German. A
typical case of lack of motivation is the borrowing in English of derivatives from Latin /
French, while not also borrowing the root as well (confer/ conceive/ conception, defer,
deceive, deception, conduct, deduct, conductive, deductive etc.)
German
English
Gezetz gezetzlich
law
legal
Kirche kirchlich
church ecclesiastical
Bischof
bischflich
bishop episcopal
Mund mndlich
mouth oral
French
loi
lgal
glise ecclsiastique
vque piscopale
bouche oral
(17)
(18)
Compounds
German
Finger-hut
Wasserleitung
Nilpferd
Formenlehre
Bedeutungslehre
English
thimble
aqueduct
hippopotamus
morphology
semantics
Derivatives
Ursache (original matter)
cause
French
d
aquduc
hyppopotame
morphologie
smantique
cause
Suffixes are more productive in German, in that they attach to a larger number of
La solidarit des termes dans le systme peut tre conue comme une limitation
de larbitraire, soit la solidarit syntagmatique, soit la solidarit associative
(paradigmatique).
Leech (1974) identifies several types of meaning, of which two directly relate to
de Sassures notion of value.
a. Reflected meaning is that type of meaning which arises in the case of
polysemantic words, whereby one meaning of the word has constituted itself on the basis
of different meaning of the same word. (Leech 1974, 19). Reflected meaning is the result
of the paradigmatic association of two or more meanings of the same word. The same
phenomenon had been noticed by Prof. Levitchi (1975): When a word is used in a
secondary meaning, that meaning is always connotatively accompanied by its main or
dominant meaning.
stems.
(20)
29
30
a.
b.
pretty girl/woman/flower/*boy
handsome man/boy/car/vessel/woman
The term collocation will be used to refer to sequences of lexical items which
habitually co-occur, but which are nonetheless fully transparent. Collocations have one
selectionally dominant term (the head of the phrase, which selects complements and
modifiers) and they exhibit semantic cohesion:
(22)
fine weather, torrential rain, light drizzle, high winds, rough sea, addled eggs,
rancid butter
From the description and the examples above there doesnt seem to be much
difference between s-selection and collocation. It is indeed true that collocations are also
instances of s-selection. In the case of collocations there is however more than semantic
selection. There is lexical selection, often out of a range of near synonyms. This is why
collocations must be observed in usage and learned as such. The semantic integrity of the
collocation increases when the selectionally dominant term has a narrow meaning and
imposes very specific choices ( see examples (23). At the limit only one item is allowed
in the collocation and the collocation verges onto an idiom or a bound collocation (24):
(23)
(24)
bound collocations
31
3. The code is independent from the emitter. The subject submits to the conventions of
the external code. There is a sharp contrast with phenomenology. Phenomenology placed
itself on the side of the speaking subject and regarded speech as one means among others
of corporal expression, as a manner of being in the world. In his verbal gesticulation, the
speaking subject was at the origin of the meaning of the utterance.
In contrast, semiotics places itself on the side of the receiver. The message
conveys information because it could have been different. The speaker has selected the
message he emitted from among all those which he could equally well have constructed
given a certain code. The code and not the emitter decides what shall and what shall not
be permitted. The meaning is not that of an experience; rather, it is the meaning that a
particular experience may receive in a discourse which articulates this meaning according
to the conventions of a particular code?
Problem: How can one account for the plasticity of meaning in context? How can
one account for the creative, figurative use of meaning in context.
One consequence is the very limiting view that figures of speech are deviations
from the norm.
32
5. Connotative semiotics
5.1 Denotation and connotation
Recall that Hjelmslev primarily defined language as a denotative semiotic, where
an Expression denotes a content. The relation between form and substance is one
selection and manifestation, where the substance manifest the form, and the form selects
the substance
(25)
Languages
Expression
Content
Hjelmslev was aware that the relation of denotation does not exhaust a text. Texts
often differ through their connotations.
Connotation, as opposed to denotation is that aspect of meaning which is
conveyed by the choice of a particular form rather than another to denote a particular
referent. Compare the members of the pairs below:
(26)
fall
autumn
horse
gee-gee
(27)
Expression
Content
Content
denotative content
connotative content
/ dji:dji:/
HORSE
BABY-TALK
33
lexical choice, text structure) build up a rhetoric which connotes an ideology. This is
Barthes position in Mytologie. The whole denotative system becomes a signifier which
is in relation to a signified represented by the associated connotations. Connotative
semiotic thus explicates what the text conveys over and above the primary meaning.
Here the system of denotation is a rhetoric and the connotative system is an ideology.
5.2
Rhetoric and Ideology. Myth Today: The ambiguity of connotation
(optional).
Consider the following picture in a French magazine.
The picture shows a
black man saluting the French tricolor.
At the level of denotation , there is the sign of salute formed between the
soldiers raised arm and the flag upom which is gaze is fixed, and secondly the sign of
color (black being selected from the paradigmatic color set.
However, when the whole denotative system is chosen as a signifier, it serves as a
rhetoric sustaining an ideology. So the sign connotes the free participation of subjected
peoples in the French empire. The picture makes an ideological statement. This is hardly
a novelty, and as Barthes (op. cit) shows, the picture supports several ideological
underpinnings and it is its ambivalence and ambiguity which makes it worth while.
a. One may treat the salute as simply an example of French imperialism. The
image is descriptive and it tells a story, stating the obvious.
b. One may treat the salute not as an example, but as an alibi of French
imperialism. In this case, the interpretation is critical, exposing the lies of French
imperialism. The critical reading presents itself as another true story.
Both of these readings are static, analytical, both of them are stories, not myths.
They destroy the myth either by making its intention obvious (first reading) or by
unmasking it. The former reading is cynical, the second is demystifying.
There is, however, a third type of focusing, a dynamic one; the interpreter
consumes the myth, living it as a story which is both true and unreal.
Conceived as an ambivalent story, the myth is true, because it expresses the
practice through which a form of domination functions ultimately. At the same time, the
myth is unreal because the practice naturalizes history and will be revealed as myth in
historical change. Notice that this complex dynamic interpretation eliminates the
dichotomy between synchrony and diachrony.
5.3 The structure of connotation in the lexicon. Some connotative features.
5.3.1 Diastratic and diatypic variation
Registers may be viewed as levels of formalities. All the features below describe
connotative content and may be applied to any frame:
34
(39)
higher
literary
archaism
academic technical
central
or
neutral word
American English
fall
truck
tap
neologism
b.
social stratification of the speech community, correlated, but not identical
with social rank and class structure (sociolects, idiolects)
Fadiman (1970)
speaks of a variety which he levels upper middle class (the
U variety) with specific pronunciation (RP with local coloring) and specific lexicosyntactic features, such as a tendency to avoid redundancy, to use understatement
lower
slang colloquialism
(40)
U
black-tie dress
curtains
children
parents
my wife
insane
of unsound mind
not in full possession of ones faculty
possessed
neurotic
psychotic
demented
unhinged
bereft of reason
non compos mentis
mental
barmy
cuckoo
gaga
batty
British English:
autumn
lorry
faucet/spigot
non-U
tux (tuxedo)
drapers
kids
folks
Mrs. Smith
maladjusted
The languages of various sub-cultures also belong here. The underground
language of today was initiated by hard-core drug addicts and criminals in the 1920s and
1930s. The underground language also created a feeling of cohesiveness, togetherness
and belonging with others who share common beliefs, traits and behavior patterns. A
recent underground dictionary identifies the following subcultures:
MAD
bananas
bonkers
crackers
round thee bend
off ones chump
()
Blacks, drug-users, motorcycle groups, prisoners and police; medical, scientific or
chemical personnel, college students, gays and prostitutes, street-gangs
potty
loony
(41)
(42)
nuts
daft, crazy
groups.
Who are you swinging with, man?
Why, I am swinging with the Bishops, man
Good, man! Were tight with the Bishops How about dropping a dime and
well get a bottle of sneaky pete.
35
(standard)
What street club do you belong to?
I belong to the Bishops.
Good. Were friendly with the Bishops How about contributing a dime and
well buy a bottle of wine. (Fadiman, 1986, 176)
36
Legalese
malicious intent
with malice aforethought
extenuating circumstances
contempt of court
37
a) the moralization of status words. Some of the social roles named by certain
terms disappear (ceorl), but whether the term does or does not continue to be used as a
social rank term, social term develop a secondary evaluative meaning, in a manner which
observes the position of the ranks on the social scale. High rank terms become positive
evaluative (nouns like prince, lady, gentleman, adjectives like noble, princely, gentle,
frank (not in serfdom), free (not in bondage), liberal (of the arts and sciences, worthy of a
free man, as opposed to servile or mechanical). Low rank terms become negative
evaluative (nouns like churl (serf), knave (a male chilepage), villain (low-born, baseminded), blackguard (the lowest menials of a household), vassal , wretch (exile, outcast),
slave (Lat. sclavus), etc.
Thus social status and moral worth are inextricably linked.
b) the secularization of religious terminology This trend is now so advanced
that it seems almost pedantic to point out the religious origin of: sanction, sanctuary,
doctrine, propaganda, novice, incumbent, conscientious, office, lobby, asylum, cell,
anathema, pittance, lesson, passion, mercy, etc.
The politicization of religious terms should be related to the enduring power
struggle between Church and State.
Notice not only an extension of the connotative feature from [+religious] to
[+religious] or [secular], but also process of deterioration in the perception of terms like
dogma, propaganda, dogmatic, pittance (originally a pious donation). An example of
decline is the conjuring term hocus pocus, originally hoc est corpus, used in the
Eucharist.
38
Curslexicalsemantics3
Componential Analysis (I)
Classical structuralism
(2)
0. Preliminaries
From previous study, it is known that linguistic analysis proceeds level by level,
specifying in each case: the primitives of the level, the combinatorial operations and
rules, and finally a representation of the utterance on that level (an L-marker). In the
particular case of the semantic level, one must specify: a) the sense components , the
constructional rules for building complex meaning out of the more elementary meanings
and a semantic representation of words and sentences. Recall that LF is a syntactic
representation relevant for interpretation in as much as: a) it specifies the constituents of
an utterance, i.e. the parts which go into the construction of meaning, b) the ccommanding relations between logical operators which indicate their relative scope; c) it
directly or indirectly specifies the semantic type of the constituents, and thus their
denotation.
LF has got nothing to say about the specific interpretation of lexical phrases. Thus
from the point of view of LF the following pairs of sentences are alike even if they
express vastly different things:
(1)
a.
b.
The sense of every lexeme can be analyzed in terms of more general sense
components, some, or all, of which will be common to several lexemes in the
vocabulary.
Aims of the method:
a.
to discover the elementary units of meaning, the invariants of the semantic
level, or perhaps the minimal units, the primes of the level.
b.
to show the systematicity of the vocabulary, by revealing the various
relations (similarity, incompatibility) holding between the lexical items.
Methodologically, CA exploits the hypothesis of the isomorphism of the linguistic
levels, a hypothesis which invites a transfer of methods from one level to another.
The methodological transfer, advocated by Hjelmslev, Coseriu, etc. is from
phoneme to the word:
The analogy between the phoneme and the word is apparent in the following properties
they share:
(3)
a. The lexeme, just like the phoneme may differ only with respect to some
distinctive feature from some other lexeme:
(i)
Lexical semantics precisely contributes an interpretation of phrases like: arrow,
student, novel, etc.
Structuralist semantics worked under the at the time novel hypothesis that meanings are
decomposable, and proposed two complementary methods of semantic analysis:
componential analysis (a paradigmatic method) and distributional analysis (a syntagmatic
method). In the coming lectures, stress is laid on the development of componential
analysis, starting with the classical structuralist period
1. Componential Analysis
1.1 History (optional) It is probably true to say that, currently, the majority of
semanticians implicitly or explicitly adopt a version of CA.
The earliest and most influential proponents in the post Saussurean tradition were
Hjelmslev and Jakobson. Both believed that the principles that Troubetzkoy (1939) had
introduced into phonology could be, and should, be extended into semantics. Foremost
among European theorists of CA are linguists like Greimas, Pottier, Coseriu.
In the US, CA appears to have developed independently. It was proposed by
anthropologists, as a technique as a technique for describing the vocabulary of kinship
terms in different languages. (See Goodenough (1963), Lounsbury (1964)). Only later
was it taken up and generalized by Nida (1964, 1975), Weinreich (1963), Katz and Fodor
(1963), Katz (1972), a.o.
SEX
stallion : mare
MALE : FEMALE
(4)
b. The same opposition is found in many pairs of lexical items; these pairs
establish a correlation in the lexicon, just as there are correlations in the phonological
system:
(i)
SEX
: FEMALE
stallion : mare
boy : girl
he
: she
he-goat: she-goat
jack-ass: jenny-ass
actor : actress
usher : usherette
MALE
(5)
c. Like phonological oppositions, semantic oppositions may be neutralized, with
the unmarked, extended term covering the semantic space named by the opposition. Here
are a few examples where the opposition is neutralized, respectively functional
man
MAN (human)
woman
39
old
OLD (age)
young
40
(i)
(ii)
(6)
d. Lexical semantics adopts the commutation test to investigate lexical paradigms
(subsystems of the vocabulary) uncovering the relevant semantic component features in
terms of which the units of the system are opposed.
The application of the commutation test makes possible the identification of the
distinctive features of a paradigm. For example, the terms on the left in (7) are opposed to
those on the right in terms of the feature [ CO-LINEAL). Replacement of one form by
another is semantically equivalent with substituting one semantic feature by another.
Similarly the feature [ APPROXIMATION ] is relevant for the correlation in (8):
(7)
[- CO-LINEAL] [+ CO-LINEAL).
my uncle
my father
my aunt
my mather
(8)
[- APPROXIMATION ] [+ APPROXIMATION ]
blue
bluish
red
reddish
young
youngish
Thus the terms blue/bluish are invariants of the lexical system standing in a
relation of commutation. On the other hand, pairs like the ones below are variants.
Substitution of one form by another does not lead to a change of content, in terms of any
opposition. They are variants contracting a relation of substitution.
(9)
he-goat
: billy-goat
she-goat : nanny-goat
1.3.
Units of the analysis
a.
The seme (distinctive feature, semantic marker, semantic feature) is a sense
component or constituent of a lexical meaning; it is a theoretical construct used to
characterize the vocabulary of a language. e.g. [approximation], [age].
b.
The sememe is a reunion (product, conjunction) of semes covering one lexical
meaning of a word.
c.
The lexeme is the association of a sememe and a phonological matrix; one word
may represent several lexemes. Most linguists, however, use the term lexeme in the
same acceptation as the term word.
The terms seme and sememe are part of the meta-language; in contrast
lexemes are part of the object language.
(boarding school)
Importantly Wordnet was used as a starting point for the construction of similar
on-line dictionaries. It was also used as a starting point for a multilingual dictionary
which employs English as a link.
(11)
1.4
copac
TREE
arbre
41
42
(12)
BOARD
(14)
a.
b.
c.
hotel
restaurant
p
[-voice]
blue
[- approximation]
kitchenette
[+small, size]
kitchen
[- small size]
The model of privative oppositions is also successful for triplets like those in (21),
where one term is positively marked [+F], while the unmarked term is underspecified,
functioning as [ F], therefore, the [F] contrast is neutralized. In the examples below,
woman, mare, etc is [+ female], while man, dog are [ female]. A redundancy rule of the
vocabulary stipulates that [+MALE ] [- FEMALE]
(21)
43
[consonant]
[+plosive]
b
:
[+voice]
(19)
(20)
On the other hand, the concept of causality is the subject of a theory of physics.
Concepts like object, entity substance are explained in physics, philosophy, etc. Similarly
the notion of approximation may be a primitive with respect to English adjectives, if we
accept that suffixes bear primitive meanings, through their generality. But the concept
of APPROXIMATION is explained in a theory of measurement. This is the consequence of
the fact that NL language provides its own meta-language.
Moreover the task of characterizing semes as minimal, or as conceptual primitives
collapses when confronted with semes like the following:
(16)
The strong thesis of binarism was that all oppositions are not only binary, but also
privative.
(18) Definition. An opposition is privative when one term of the opposition is marked
for a feature that the other term lacks.
{board, PLANK}
[ FEMALE].
man
horse
dog
[+ FEMALE ].
woman
mare
bitch
[- FEMALE].
man
horse
dog
44
Redundancy rule:
(22)
opposition between a and b. Similarly, liquid and solid are not directly opposed, but
through the mediation of the complex seme [ neither liquid nor gas] (i.e. solid).
An important ternary system, where the reduction to two binary oppositions
seems to be much better motivated than above, is that of the English pronouns:
[+MALE] [- FEMALE]
[MARRIAGE]
[+MARRIAGE]
married
husband
wife
[-MARRIAGE]
unmarried
bachelor, single
spinster, single
(26)
[-PERSON]
|
it/ what/somebody/no one
It should be obvious that not all oppositions may be privative or even binary.
Other types of binary oppositions are the so-called equipollent oppositions.
Definition The two terms of an equipollent opposition share the base and the
relevant dimension, but differ in terms of specific differential features. The form of the
opposition is thus aRb, rather than a R a (the privative opposition):
SEX
[- FEMALE]
girl
mare
b)
(25)
[+PERSON
|
who/somebody/nothing ..
|
[MALE]
[FEMALE
he
she
Two things appear to have been achieved: Apparently ternary oppositions are
reduced to two binary oppositions. Secondly, a hierarchy of structure of content, resulting
from a geometry of features is also achieved. Such a hierarchy, i.e, a cluster of lexical
items hierarchically ordered by the opposing features represents a semantic/lexical
system.
(23)
[+MALE]
boy
stallion
[ENTITY]
they
1.4.5 The semantic system, as defined above, is yet another unit of content of the
analysis, which offers a model of hierarchical organization of content.
Coming back to the problem of binary oppositions, the tendency towards binarism
is a fundamental trait of the human mind, apparent in basic dichotomies like, truth/falsity,
beauty/ugliness, good/bad; binarism is also essentially tied to the description of language
as discrete infinity. Syntactic features also come in binary pairs:
AGE
|
[NEITHER YOUNG NOR OLD] -----------------------
(27)
middle aged
However, it is beyond doubt that one must acknowledge dimensions having more
than two terms: colors, geometrical figures:
old
(28)
[FIGURE]
[TYPE OF BORDER LINE]
STATE
|
[NEITHER LIQUID NOR GAS]
|
solid
STRAIGHT -SIDED
CURVED
circle
ellipsis
(29)
45
THREE-SIDED FOUR-SIDED
triangle
square
quadrangle
FIVE-SIDED
pentagon
COLOR
red
violet black
46
1.5 Structuralists (e.g. Nida, 1975) propose that in the hierarchical organization of
content, there are two fundamental patterns of organization: the paradigm and the
taxonomy. They are seldom exemplified in a pure form, but still there are good
examples of both taxonomies and paradigms.
Both are examples of conceptual domains, and eo ipso, lexical sets characterized
by the fact that;
(32)
a) All members of the paradigm/taxonomy have at least one common semantic feature,
the root of the paradigm or taxonomy.
b) The meaning of every form differs from that of every other form in terms of one or
several additional features.
c) In the perfect paradigm features are unordered, all dimensions are relevant for all
terms, that is, the features of any dimension combine with all those of any other
dimension. The dimensions behave like the grammatical categories of inflectional
paradigms (see the paradigm of kinship terms below). In the perfect taxonomy, features
are hierarchically ordered. A given feature combines with only one feature from any
other dimension.
1.6 A good example of lexical paradigm: English kinship terms
Componential analysis is applied to sets of terms that represent culturally relevant
domains and proceeds by recognizing semantic distinctions which organize the set into
subsets.
CA tends to concentrate on denotative meaning, but connotative features may also
be introduced.
The aim of CA is twofold: a) to break down the meaning of a word into components, thus
offering componential definitions of the terms; b) to bring out the relationship between
terms, which can best be described in terms of their features.
The starting point of the analysis is an empirical (disjunctive) listing of the
denotations of the term, therefore an extensional definition of the term.
(30)
my cousin
Paradigm considered
grandfather, father, mother, grandmother, uncle, aunt, son, grandson, granddaughter, nephew, niece, uncle, aunt, son-in-law
Root of the paradigm: K = kinsman
G+3
G+2
great-grandfather
grandson
G+1
grandfather
G0
G-1
father ego
G-2
son
G-3
grandson
great-
d. LINE opposes the term designating kinship relations on a direct line (+L) to
those designating kinship relations on a collateral line (-L).
(34)
+L :
-L
[+D]
[-D]
(36)
Term Root
K
father +
mother +
uncle +
aunt +
brother +
sister +
cousin +
son
+
niece +
nephew
son-in-law
Consang
Line
+
+
+
+
+
+
+
+
+
+
-
+
+
+
+
+
-
+
+
+
+
-
Rank Sex
Male| Female G-2
m
f
m
f
m
f
X
m
f
m
m
Generation
G-1 G0
+
+
+
+
+
+
+
G-1
G-2
+
+
+
+
Some results:
47
48
1.7 Taxonomies
As already illustrated, in the perfect paradigm features are
unordered, all dimensions are relevant for all the terms. In the perfect taxonomy, features
are ordered in a hierarchy and combine with only one feature of any other dimension.
PLANT
FLOWER
|
Word meaning is represented as a conjunction of conceptual semes
TREE
GARDEN FLOWER
WILD FLOWER
FRUIT TREE
WILD
TREE
(38)
grand daughter
The method may be extended to other terms. Most semanticians regarding the
word as merely implying its definition, rather than being equivalent with it.
(39)
woman
spinster
bachelor
peony rose
fir
|
|
apple pear
|
|
cherry oak
tree
tree
tree
daisy violet
tree
tree
2.
Componential definitions serve as a basis for delimiting the class of analytical
statements and contradictory statements:
(40)
(41)
He is male
He is the same generation with me
x [[ x is my brother] [x is male]]
He is my brother
The same class of entailments may be used to state redundancy rules, i.e., specific
features guarantee the presence of hierarchically related dimensions/features
(43)
x [[ x is human] [x is animate]]
49
50
Curslexicalsemantics
b.
semantic markers are the features which characterize more than one item; these features express
similarities between items and therefore they show those aspects of the meaning which are in fact
systematic and uncover the structure of content; e.g., [+ Animate], [+ Colour], [+State], [+Gradable];
c.
distinguishers
a marker which reflects the idiosyncratic element in each lexical element; the
presence of a distinguisher shows the irreducible nature of words. Lexical items are not equivalent with
componential definition
d.
selectional features, the fourth type of information provided by the dictionary refer tot the
combinatorial abilities of lexical items: arguments are s-selected; modifiers s-select the modifee; e.g., drink
+ -- < LIQUID>NP.
Notational conventions: semantic markers = ()
distinguishers
= []
selectional restrictions = <>
4A
On semantic markers and distinguishers In principle the distinction is hard to maintain in particular cases,
because it would presuppose, in principle, an exhaustive analysis of the language to establish the nonsystematic features, which characterize only one item.
The distinction is not without theoretical interest: Semantic distinctions is expressible in terms of markers
or classemes, such as, person, become, human, animal, furniture, etc. have certain characteristic
properties: they are readily identifiable across languages, being less language dependent than contrasts
which require distinguishers to be described; the meanings of bachelor above corroborate that point. (ii)
such contrasts may be relevant for the lexical and grammatical system of the language, often being
lexicalized or grammaticalized: For instance, English embodies the [ person ] distinction in its pronominal
system through contrasts like: someone/ something, nobody/ nothing, etc. The feature [ cause] is
lexicalized in pairs like kill/ die; bring/ come (They brought him here/ He came here), take to/ go to, (They
took me there. I went there ), [ become] is embodied in pairs like: red/redden; black/ blacken, etc.
a)
grammatical markers, namely categorial and subcategorial features: +N, + V, [+__NP], etc.
b)
semantic markers are those semantic features which express a generalization, since they are
supposed to characterize at least two lexical items. Such markers measures the similarity of meaning
0. The contribution of GG to lexical semantics has been manifold:
a)
proposing the concept of mental lexicon and thus laying stress on the manner in which lexical
knowledge is stored and acquired; the generative tradition is an intrinsic part of the cognitivist paradigm.
b)
elaborating and refining a notion of lexical entry;
c)
describing the manner in which the lexicon stores syntactic features as well as the manner in
which syntactic information is used in syntax (the projection of lexical items)
d)
very generally an important component of various formal approaches to grammar is negotiating
the boundary between syntax and the lexicon.
()
bachelor
|
+Noun
grammatical
markers
[Det ---]
1. Katz and Fodor: Towards a Semantic Description of English
This is the first attempt of explicitly coupling grammatical and semantic description. Recall that the
Grammar had three components in the following organization, whereby the Semantic Component operated
on the Deep Structure produced by phrase structure rules:
(human)
(animal)
(male)
(adult)
(young)
(young)
Semantic Component
(never married) (knigh)t
Deep Structure
()
semantic
markers
(male)
..
Syntactic Component
(seal)
Surface Structure
[serving under
the banner of another]
Phonologic Component
The semantic component consists of consists of two elements
a.
a lexicon that provides a meaning for each lexical item
b.
a finite set of projection rules, which assign readings to syntactic phrases by combining the
readings assigned to their constituents.
It is the form of lexical entries that we are concerned with here, since it directly embodies a form
of componential analysis. Where K&F innovate is in that they propose a classification which is of interest
in the articulation of grammar and semantics.
Each lexical entry contains the following types of markers:
a.
grammatical markers, specifically the categorial and sub-categorial features (the c-selection
feature);e.g. +N, +V, + -- NP, etc.
51
[when without
a mate during
the breeding
season
distinguishers
selectional
restrictions
It has also been proposed that semantic markers, therefore those features which are repeatable in a
language and translatable cross-linguistically are taken from a set of semantic universals.
Leibniz, the first philosopher who proposed a form of Universal Grammar, put forth the
construction of a universal symbolic language. The symbols in this universal language should express
simple atomic ideas. The project was revived in a different by Carnap: Logische Aufbau der Welt (The
Logical Structure of the World). We have already rejected the project of atomic concepts.
There is a clear connection between CA and the project of semantic universals. One might propose
a very strong thesis of universalism claiming that: (i) there is a fixed set of semantic components, which are
universal in that they are lexicalized in all languages; (ii) the formal principles by means of which these
combine when they lexicalize are the same (=universal). The distinction between (i) and (ii) is a distinction
52
between substantive universals (universal concepts) and formal universals (universal principles of
combination. (iii) Meanings are decomposable without residue.
A few remarks are in order here:
a. None of the European structuralists) was a universalist. Hjelmslev maintained a relatively weak
thesis of formal universalism, but explicitly rejected the thesis of substantive universality. Indeed the
problem for structuralist semantics it that it has no place for semantic universals, since concepts, in
principle even those employed in a meta-language are language dependent.
More recent writers in the same tradition (notably Coseriu, Greimas, Pottier) also reject the thesis
of universalists. They apparently do not adhere to the division without residue thesis. Their adherence to
this principle is weakened by the methodological decision that meanings should not be split without limit,
but only to the point of any two lexemes are different from each other in terms of at least one seme.
Secondly, those analysts who recognize a difference between systematic (classemes) and idiosyncratic
semantic markers (= semes) implicitly reject the thesis of division without residue.
b. As already acknowledged classical structuralism cannot in principle guarantee the existence of
any substantive universals, and hence there is no theoretical foundation for mutual translatability and
mutual comprehension.
c. The Aspects Grammar also insisted on the existence of formal universals ( the types and the
format of the rules) According to K&F, the projection rules of the semantic component, the structure of the
lexical items are all found in all languages. It is one the merits of GB to insist on the existence of
substantive universals: all items which are [+V] in any language denote eventualities and have a-structure,
2. Lexical decomposition and the analysis of verbs
Generally the type of lexical entry proposed by K&F is still valid, but has been enriched, due to a
result more recently obtained in semantic analysis, namely that the internal semantic structure of words
depends on the part of speech they belong to.
Dowtys aspectual classes:
MILLER
In 1985, Princeton University, 95,600 word-forms, only five categories have been
considered: nouns, verbs, adjectives and adverbs
Function words are omitted on the assumption, supported by observations of the
aphasic patients, in Garett, 1982, that they are probably stored separately. The realization
that syntactic categories differ in subjective organization emerged from studies of word
association: (Fillenbaum &Jones 1965): a N probe obtained a N (80%), A/A (65%), V/V
(43%).
Nouns are organized as topical hierarchies, verbs are organized by a variety of
entailement relations, and adjectives and adverbs are organized as N-dimensional
hyperspaces. Each of these lexical structures reflects a different way of categorizing
experience. Languages have far fewer verbs than nouns. ( Collins: 43ooo nouns, 14,000
verbs. verbs are more polysemous. Nouns have, 1.74 sense on the averge, while verbs
have 2,11.
VERBS
Verbs meanings are more flexible, depending on the nouns they combine with ( I had a
headache / a shower/ a surprise, the high mutability of verbs) In anomalous V+N
combinations, the verb is assigned a new reading, rather than the noun.( grasp a hand/
grasp a meaning; drink water/ drink knowledge,
Light verbs: highly polysemantic: be, have, run, make, do, set, go, come, take. For these
polysemy basically arises by combinations with nouns.
Aspectual Hierarchy
Cause > Other
Recall that only an argument which is prominent on both hierarchies is an external argument in
syntax, whence the existence of verbs that have no external argument (unergatives):
()
53
cut
(Agent, Theme)
(Cause> Other)
(x (y))
Since what matters are degrees of prominence, the labels themselves are less important. The
relative prominence is dictated by the manner of projection, the head-complement and the head-specifier
relation.
Recently, Hale and Keyser propose to represent prominence relations using syntactic relations
defined on trees. They propose a conception according to which thematic relations are identified with NP
positions in syntactic projections, termed Lexical Relational Structures, which are defined by the lexical
entries of predicates. Roles derive from a certain manner of projection in LRS. Since the number of
syntactic configurations defined by x-bar theory is limited, there will be a limited number of theta-roles.
LRS representations are subject to general syntactic principles (government, head-to head
movement). LRSs include only the lexical categories N, V, A, P, each associated with a specific entity.
There is thus an ontology of parts of speech.
()
Thematic Hierarchu
Agent > Experiencer > Location/ Goal/ Source/ > Theme
V
event
N
entity
A
property
P
relation
54
Possible grammar rules define possible lexical items, since what is lexicalized is always a syntactic
component, a sub-tree.
Thematic relations and Theta roles:
1.
Why are there so few theta roles?
2.
The principles of Theta theory require a uniform projections of theta roles, so that the same role
always appears in the same configuration:
Our basic answer to the questions expressed above (why there are so few roles) is that in an
important sense there are no thematic roles. Instead, there are just the relations determined by the categories
and their projections and these are limited by the small inventory of lexical categories and by the principle
of unambiguous projection (binary branching, and the Single Complement Hypothesis).
We will start by looking at the concept of Agent. An Agent has always been described as a Causer,
the subject of the conceptual predicate CAUSE. Causation is a relation between two events, such that the
presence of the first one (the causing event) necessarily triggers the presence of the second event (the
effect). The cause unilaterally produces/ selects an effect. Since events are named by verbs, the syntactic
configuration relating two events in a cause relation will be the one in () below. The matrix V (the head, the
cause) selects and entails the presence of the effect denoting VP; causation appears as a head complement
relation, an example of unilateral selection. Semantically, the relation between e1 and e2 is entailement.
()
V
V
In this structure, the head-complement relation involves the categories V and P, with the latter
subordinate to the former. We will continue to assume that the semantic type of P is inter-relation, using r
to symbolize this. These basic semantic notions combine to assign an elementary semantic value to this
syntactic structure: a dynamic events implicates an interrelation , as expressed in (10). The most salient
meaning attached is change ( achievements). A P expresses the inter-relation between a trajectory and a
landmark [get mud on the wall, get a fly into the soup]. Using the same notation for subject > we get the
semantic formula of a change of location predication; the label attached to the specifier of this
configuration is Theme or an affected patient. Again these semantic roles are derivative of the syntactic
configuration:
(10)
(10)
(11)
NP
e ---------------- r
i > (e -------------- r)
An affected patient is the only role that undergoes middle formation:
()
e1-------------- e2
(7)
PP
P
VP
The syntactic structure above and the associated semantic relation comprise the LRS expression
of a causal relation. If this is true the subject of this configuration is an Agent. This NP bears a
syntactically unambiguous relation to the V of (4). Suppose we symbolize this relation as > and derive a
composite elementary semantic representation for the entirety of (4):
V
V
b)
In an accepted view of thematic relations, the Theme role and the associated elementary
semantic relation change extend to predicates of the type represented in (12):
(12)
i > (e1-------------- e2 )
a.
b.
c.
d.
a.
b.
c.
d.
VP
NP
gravy
()
V
V
thin
AP
t thin
i > ( e--- s)
V
V
VP
NP
V
tput
P
become
get
on
(dynamic event)
put
()
The lexical category AP is associated with the notion state, the configuration in (13) expresses a
change of state, as defined below. The subject is again an affected patient or theme. Embedding this under a
verb yields the parallel causative construction. Middle formation is again possible:
V
PP
NP
(16)
the shelf
VP
NP
(17)
V
V
the
books
put
PP
P
on
NP
the shelf
We have examined three of the complement types available in LRS representations, those
projected by the categories V, P, A. The fourth type is that projected by the category N, exemplified by the
unergative verbs of (16), and the simple transitives in (17):
55
56
(16)
a.
b.
c.
d.
a.
b.
c.
d.
(17)
NP
the
saddle
(18)
(4)
V
V
(19)
(20)
PP
P
the horse
Conflation
Lexical representation
The Generative Semantics program was motivated at least in part by the idea that the notion of possible
lexical item is defined or constrained by certain principles of grammar which also determine the wellformedness of syntactic structures.
VP
NP
a.
b.
c.
d.
V
V
NP
e----- i
I > (e----- i)
The lexical representation of any verb involves a system of relations (called Lexical Relational Structure)
having an essentially syntactic form of the type in (17).
Unergatives cannot be embedded under causatives: suggestion they do not have a subject in l-syntax, but
just a cognate object; they get their subject in syntax. This accounts for the following data:
(19)
VP
NP
(21)
a,
b.
V
V
VP
LRS representations represent a syntax and they are subject to general principle of syntax. LRS involves
the same categories defined in syntax: projection, government, complement, specifier.
By assumption, lexical categories contain only lexical elements: N, V, A, P.
LRS may contain unexpended XPs, which correspond to variables, as well as actual morphemes,
such as saddle, shelf, for verbs like to saddle, or to shelf.
a.
b.
VP
NP
(24)
VP
NP
V
V
V
V
P
AP
dark
porch
(25)
Consider locatum verbs, like saddle the horse, carpet the floor, etc. Two configurations seem plausible: It is
the one where the locatum is the lower PP which is actualized:
The horse saddles easily.
*The saddle horses easily
()
V
V
VP
NP
the
horse
()
V
V
tporch
NP
N
tporch
VP
V
V
porch
NP
PP
P
saddle
VP
NP
(newspapers)
The verbs shelve/ porch projected their structure in the general way, according to the canonical
structure realization of verbs in syntax. Thus the transitive verbs porch / shelve project in the usual manner.
The NP variable which it c-commands is realized as an internal argument, and the specifier will be realized
as an external argument subject: The mailman porched the newspapers. The rest of the lexical structure is
not represented in syntax. It is part of the lexicon, representing lexical knowledge. To understand a verb
like porch one has stored a LRS like (24).
VP
NP
NP
NP
()
PP
V
VP
NP
the
room
we
c.
D-structure representation
Consider a possible lexical representation of He shelved the books/ He porched the newspaper
V
V
VP
57
58
(15)
a.
locatum verbs (locatum = object located; carpet the floor, put the carpet on the floor, salt
the food, sugar the cake, paint the wall, paper the wall, etc.
b.
location verbs
(ground the planes, jail the prisoner, shelve the book, porch the
newspaper)
c.
duration verbs ( summer in France, weekend at the cabin)
d.
agent verbs
(buthcher the cow, jockey the horse)
e.
experiencer verbs (witness the accident)
f.
goal verbs
(fool the man, group the actors, pile the money)
g.
source verbs
(piece the quilt together, word the sentence)
h.
instrument verbs (nail a sign to the door, wedge the window open)
i.
miscellaneous (lunch, fish rain)
1. Simple verbs
The examination of the list above shows that Clark and Clark omit to name a very productive class of
verbs, the class of verbs which incorporate their internal (Theme) argument. This is a rich class, which we
discuss starting from the class of verbs of animal birth:
Verbs of animal birth-giving:
()
a.
b.
verbs of harvesting:
hay, berry, mushroom, fish, whale; get fish, get mushrooms
c.
weather verbs
()
Location verbs
These are modeled on the verb put , i.e., cause to be in a particular location. He shelved the book is
analyzed as He put the book on the shelf, or He caused the book to get/be on the shelf. Thus, the lexical
decomposition of verbs amounts to decomposing them in smaller verbs which the complex verb entailed.
If He shelved the book is true, then it is also true that The book /is on the shelf, The book got to be on
the shelf, He caused the book to be on the shelf.
VP
NP
V
V
foal
The parallel structure with a variable (rather than a constant) in the position of the DO is below, and
corresponds to simple transitive sentences.
()
()
a.
b.
c.
In this case, analyzing these verbs as unergative does not come from any common transitive
paraphrase, but follows from a grammatical principle that prevents specifiers to incorporate; conflation
cannot proceed from upwards, as a result, rain, thunder, are unaccusative and select an expletive subject to
satisfy the EPP property of Tense. Notice also occasional cognate object constructions
Flicka foaled
The cow calved
(a)
a.
b.
c.
Several other groups of verbs are part of the same unergative paradigm:
a.
()
lunch, dine, breakfast, picnic
foal, fawn, whelp (dog, wolf), calve, pup, lamb, wether (ram)
They are projections of a single lexical head and have one constant (nominal), which is
incorporated by head-to-head movement.
()
()
ground the planes, jail the prisoner, shelve the book, porch the newspaper
V
V
VP
NP
(b)
NP
V
V
V
have
jail
The English have corresponds to the most abstract construal of such verbs as produce, do,
make, effect, bring forth, issue and in general (b) corresponds to this meaning.
We assume that object incorporating verbs are unergative, illustrating the same class of verbs as
typical unergatives: laugh, sleep, snore; chirp, twitter, roar, etc. which also involve conflation of an internal
object. Most of these verbs may appear with cognate objects and have light verb paraphrases. They may
assign Accusative to the cognate object (Burzios Generalization) or to the subject of a small clause in a
causative construction:
(prisoners)
tjail
()
a.
b.
tjail
VP
V
V
VP
NP
V
V
tjail
NP
VP
NP
NP
VP
laugh, sleep, boo, cough, hiccup, wince, hiss, whistle, snore; chirp, twitter, roar,
()
PP
P
Evidence for the complex structure is also provided by adverbial modifiers and negation, which
may modify any one of the predicates in this lexical structure.
PP
()
VP
for
for
three
NP
laugh
59
V
V
PP
P
NP
60
years
jail
(prisoners)
tjail
tjail
tjail
Given (30b), the structures in (27) should be as in (31). Such structures are interpreted as activity-denoting
events.
(31)
VP
NP
V
V
PP
for three years
V
V
RH
tjail
Returning to the unaccusative structure of (26), two subtypes can be distinguished depending on the nature
of the XP:
(26)
[ VP D[ VP V XP
PP
P
tjail
NP
tjail
If the XP denotes a path with an endpoint the configuration indicates movement to a telos (Goal). Note that
the path may be a scale; movement along the scale is directed motion. Change of possession is directed
motion. State denoting predicates in change of state construction can also be viewed as directed motion:
(34)
[ VP D[ VP V XP
a.
b.
c.
[V
[V
[V
V
V
V
a.
b.
a.
b.
c.
(35)
a.
b.
[P P [ P P D]]
[PP
D]
(36)
a.
b.
(37)
[D[V[PP
D]]
[ John [ VP is [ P in [ the park]]]]
N]
sleep]
laugh]
While one can distinguish between static and dynamic events for PPs, how can one extend this to
APs, to differentiate between:
()
We may assume that run too has a hidden object: John ran a long distance, John ran a distance of
two miles. This would mean that manner of motion verbs are not denominal, in contrast to laugh, work,
sleep
We will endorse the view that the presence of an external argument is not regulated by the verb
per se. We will assume the existence of a vP, immediately above l-syntax. The head v license the external
argument and identifies transitivity as case assignment in the strucrure v V.
(30)
a.
b.
c.
It has been proposed that the P in directed motion cases is complec : a directional P+ locationP. In statives
the preposition is simple:
Probably not all unergatives are denominal verbs. Note that manner-of-motion verbs are
systematically ambiguous between an (acitivity) denoting unergative verb and an accomplishment denoting
transitive use, comparable to verbs of consumption and creation:
(28)
(29)
(33)
On the other hand if the XP complement in the unaccusative structure (26) denotes a location, then the
structure is stative:
The subject of an unergative is not an argument of the verb. It originates in the sentential part of the
grammar, i.e., in s-syntax. Furthermore unergatives are denominal verbs associated with the structure in
(27), involving conflation; in this process the head projects the categorial feature and the complement
provides the phonological content for the derived V.
(27)
[vP D[ v [ VP V D]]]
[vP D[ v [ VP V sleep]]]
[vP D[ v [ VP V laugh]]]
VP
VP
NP
jail
a.
b.
c.
a.
b.
a.
b.
[P D [P D
[aP D [a A]]
61
a.
62
b.
c.
(46)
The structure in (46) lacks a category that introduces the relation of central coincidence. In such
a case, A is associated with the following default interpretation:
(47)
If A is not dominated by little a, then it is interpreted as denoted the endpoint of a path.
(48)
If V immediately denotes a category that denotes the endpoint of a path (A or N), V denotes a
change of state or location.
Assuming that the line of analysis is correct, we derive:
(53)
(54)
Preliminaries
Word formation rules are rules that express generalizations about the actual and
potential words of a language. They have been characterized as a subset of redundancy
rules (rules that account for systematic relations between words).
WFR express formal as well as conceptual relations between atomic and complex words.
WFRs are meant to capture at the level of linguistic description a native speakers
intuition of lexical relatedness. Halle (1973) remarks that it is part of a native speakers
linguistic competence to possess knowledge not only about the words of the language
but also about the composition and structure of the words.
Generally, complex words are analyzable into constituent parts which contribute a
component of meaning to the whole: the prefix /in / (realized phonologically in various
ways) occurs in at least 2000 words and means:
(a) not (as in indirect, impossible, irregular, illiterate)
63
64
Precept
Detain
Intermittent
Insist
Aspect
predeinterinad-
before
away,down
between, among
in,on
to,towards
capere
tenere
mittere
stare
specere
take, seize
hold,have
send
stand
see,look
65
(i) Blending
+snort
(ii) Acronyms ATV (all terrain vehicle) D.I.Y (do-it-yourself ) LEM (lunar
excursion module)
(iii)Clipping lab, exam, con, bike, gym, mike, pub, cabby,movie
(iv)Reduplication
bible-babble (idle talk) tick-tack, zig-zag, wishy-washy (not
dependable) boogie-woogie.
4.0. AFFIXATION
(i)
Affixation rules do not generally involve phrasal categories : *washslowlyable.
They may involve bases that are free morphemes (happy-happi#ness) or bound roots, i.e.
morphemes that cannot be independent words, e.g. emphat+ic; electr+ic .
-(ii) suffixation generally involves a shift in the lexical category of the base (though
this is not always true: compare: / read/ V /read/ V#er/ N vs /glove/ N /glove/#er/ N
(iii) prefixation does not involve any shift in the category of the base, though this is
not always true; compare: happy/A/un /happy/ A vs /noble/AV/en/noble/A/ V
(iv)
diachronically speaking, affixes are of native stock (Germanic origin) or came
into the language through borrowing (Romance, Greek, etc. known as Latinate).
4.2.Non-category changing Prefixes
Preliminaries:
Prefixes in general do not change the category of the base they subcategorize for
(with a few exceptions: en- (enlarge, enslave) be- (behead)) but may bring about changes
in the subcategorization properties of the bases :
RE # attaches to verbs and contributes semantic information (ANEW;AGAIN) ; Consider
some examples:
He thought about the problem vs HE rethought *about the problem
He wrote me a letter vs He rewrote *me a letter
Descend from vs redescend *from; wash off the deck vs rewash *off the deck
The negative prefix MIS# (BADLY, WRONGLY) generally attaches to verbs that select
sentential complements; the attachment of MIS# requires that the verb take a DP
complement:
I understood that he was right vs I misunderstood him/the problem
I advise you that you should go vs I misadvised you
4.1.Origin of prefixes
Germanic:
Be- (about; over): to besprinkle; (thoroughly, completely) to besmear
For- (away, off): to forswear, to forbid
In- (into): insight, inlet
Mis- (badly,wrongly): to mislead, to misunderstand;
Un- (negative) : to undo, unfriendly, unnecessary
Out (out of): to outlive, to outshine
Over- (above, beyond): to overeat, to overestimate
With - (opposite, against): to withstand, withhold
Romance:
a-,ab-,abs- (from): to abuse, to abstain
ad- (to) (by assimilation: af-,an-,etc): to adhere, to affirm, to accredit
bi-, bis- (two,twice): bi-lingual, bi-monthly
66
67
In the case of un# and in# pairs: happy-unhappy; sane-insane, the pairs do not
share this property characteristic of complementary terms, i.e. the denial of one term is
not the assertion of the other term: not happy unhappy; not saneinsane (but the
relation holds the other way round: he is not unhappy is to assert that he is happy). Both
unhappy and insane are stronger in negative force than not happy/not sane, Un# and in#
derivatives share gradability in terms of more/less/very: very unhappy, less unhappy than
etc. The pairs built on in# and un# are said to be implicitly gradable.
The three negative prefixes behave similarly with respect to the possibility of
reversibility which can be tested with more and less: more obedient =less disobedient;
more happy=less unhappy; more sane=less insane. The test shows that the three pairs
built with un-/in-/dis-/ have no absolute value at one end of the scale along which they
are graded.
Non# is frequently contrasted with un# and in# in expressing binary (non-gradable)
contrast vs opposite end of the scale: compare:
non-scientific (contradictory of the base) vs unscientific (contrary or opposite to the
base);
non-human vs inhuman; non-Christian vs unchristian;
The four negative prefixes discussed here are all productive with adjective bases but
some of them have the structural property to attach to bases that belong to other syntactic
categories, in particular N and V.
Non# words represented originally loans from Old French mainly in the field of
jurisprudence and mostly nouns. Recorded since 14 cent: non-payment, non-tenure, nonresidence. Non# extended to adjectives in the 17 cent, and is highly productive.
UN#words are of Germanic origin and the prefix goes back to the same Indo-European
root as OGreek a-,an-, Latin in-. Very productive, chiefly with adjectives. There are few
un# derived adjectives formed from simple base adjectives.In point of semantic
interpretation the un#derivatives, together with their base adjectives, form antonymic
pairs.
Remark: there are some un#adjectives which do not have corresponding unnegated forms:
unrelenting, unabashed, unprecedented, uncouth (couth is the past participle of the OE
cunnan=to know), unkempt (kempt=combed has become dialectal).
Un# may also take verbs as bases to form negative verbs. The base verb and the
derivative form a pair of converse (antonymic) pairs: unbar, unbrace, unbuckle, undo,
unnerve, unhinge, unarm, unfold, untie, unveil,etc. In point of their semantics , the
negative derivatives imply the existence of a former state denoted by the base: to unpack
smth entails a prior state at which the object was packed.
IN# has several allomorphs which are phonologically conditioned (ir-,il-,im-,). It is of
Latin origin and was introduced in English via French (14th century):incomprehensible,
indiscreet,inflexible, insensible.
IN# attaches mostly to foreign bases claiming a restricted sphere chiefly learned and
scientific, hence its rival, the native un# has come to oust in# more and more. That this is
the case can be seen in the following instances where in#derivatives have given way to
un# words:
e.g.
Incertain-uncertain; inceremonious-unceremonious; incomfortableuncomfortable; ingrateful-ungrateful; intenable-untenable, etc.
68
There are also instances in which both types of derivatives exist (with no change in
meaning):
e.g.
Inalterable- unalterable; inorganic-unorganic; inapproachable-unapproachable;
infallible-unfailing; infructuous-unfruitful;
From a semantic point of view, in# builds negative derivatives that , given the semantic
properties of the base, can form either:
(i)
(ii)
French in origin and is not productive in present-day English; the existing word represent
wholesale loans: debark, degrease, dewax, etc.
From a semantic point of view the derived negative verbs constitute converse terms to
their bases: the negative entails the existence of the converse term.
5.0 Suffixation
5.1 Preliminaries
(i) As already mentioned an important property of suffixes is that they change the
syntactic category of the base, hence one can distinguish among noun-forming (#er, #ee, #ness,
#ship etc), verb-forming (#ize, #ify,#en) and adjective forming suffixes (#able, #ish;#less, etc).
This means that suffixes are assigned a syntactic category and function as syntactic heads.
(ii) a second important property is that suffixes (like prefixes) are selective as far as the
morphological make-up of the base is concerned: the noun forming suffix +ity selects adjectives
ending in #able but not adjectives ending in #ish; compare: perishable-perishability vs freakish*freakishity.
Adjectives ending in #ish will allow the noun-forming suffix #ness:freakishness.
This property of affixes can be accounted for in terms of the origin of the affix: native or foreign
(mostly Latinate). The constraint is that Latinate affixes of the +ity type cannot apply to
adjectives formed with the Germanic adjectival suffix #ish. Latinate affixes can follow other
Latinate affixes but cant follow Germanic affixes This means that, given their origin (Germanic
or Latinate) affixes are ordered. It is Latinate affixes that will mostly attach to bound roots (e.g.
author+ity, dign-ity, ent-ity). This generalization about the ordering of affixes can be
schematized as in (1) below:
(1) [[[root/stem] Latinate affix] Germanic affix]
From a semantic point of view the pair of terms display either antonymy relations
(converse relations ) or complementary relations (contradictory).
In the converse relation the derived negative describes the reversal of event or
state denoted by the base: array-disarray; band-disband; colour discolour; engagedisengage; embark-disembark; enchant-disenchant; entangle-disentangle; integratedisintegrate; mount-dismount; organize-disorganize; place-displace; use-disuse, etc.
The second semantic interpretation, i.e. the pairs form complementary terms, is
represented by examples like: agree-disagree (not to agree= to disagree, not to
disagree=to agree); disaffirm, disavow, disapprove, disclaim, dislike, displease,
dissatisfy, etc.
Due to the fact that the verbs in the dis#class represent French loans, some of the
derivatives have lost their structural transparency: disappoint, for instance, is formed
from the base appoint which has one of its senses come up to the expectation or hope,
so disappoint is the complementary term of appoint in this sense.
DE# is historically related to dis- i.e. it is a French loan. Marchand (1969) remarks that
English adopted the derivational pattern and not the words. The independence of de# is
proved by the fact that French words like doxygner, deoxyder appear as deoxygenate
and deoxidize around 1799. The prefix selects bases ending in ise/ize; -ify; ate; atomic
bases: demoralize; depopulate; declassify, devalue. The pattern with atomic bases is
What the generalization above says is that if a word has both Latinate and Germanic
derivational affixes the Latinate ones will occur inside the Germanic ones. Compare:
childishness vs *childishity
Other clues to the differences between the two sets of affixes are the phonological
changes that Latinate suffixes force on their base (including stress shift): +ant/ent may
attach to V that end in the suffix +ate (Latinate). In such cases a truncation rule applies in
order to produce the derived nouns: participateparticipant; administrate-administrant,
etc. Affixes of Germanic origin do not mess with their bases like that (Harley 2007).
69
70
(iii)
The distinction in terms of origin may account, partly, for the fact that different
affixes may have the same denotation: consider +ant/ent and #er. Both suffixes attach to verbs
to form agentive nouns: i.e they both designate persons connected with whatever the baseverb denotes: servant, aspirant, inflatant, resident, referent vs writer, baker instructor, teacher,
etc. +Ant/+ent is restricted to foreign bases (even bound roots) while #er will combine with both
foreign as well as native bases on condition they qualify as independent phonological words
(lexemes). This is why suffixes like #er will generally have the word boundary #.
5.2. Noun-forming suffixes
(i)
Latinate (productive): -Nouns denoting beings and objects: +ist (Gr); +ee (Fr); +ess
(Fr);+ant/ent
- abstract nouns : +ism (Gr);+age (Fr); +ance/ence (Fr);+ment;
+tion; #al; +ity
(ii)
(5) ist
Denominal person noun
Deadjectival person noun
guitarist, Marxist
purist, fatalist
Ever since the 80s Williams, Di Sciullo, Lieber, (among others) have argued that the
changes in the syntactic valency of base verbs can be accounted for by assuming that
derivational processes are morphological processes that operate on the argument structure of
bases, hence the behavior of particular affixes was accounted for in terms of a purely
argument-structure theoretic framework.
Rapaport Hovav&Levin (1992) (for English) and Booij (1986) (for Duch) give a very neat
account of er nominals. The argument goes that er binds the external argument of the verb
to which the affix attaches; appeal to a syntactic argument position makes it unnecessary to list
a variety of thematic roles (agent, experience, instrument, stimulus, etc) in the analysis of the
affix.
71
The framework, though, cannot account for the denominal forms (Londoner, villager).
Although it has been argued that nouns have arguments (Higginbotham 1985) the R argument
= referential argument), the R of London can not be interpreted as an agent or instrument, or
a theme , since nominal arguments do not have thematic interpretation in the way verbs do.
Yet, nouns like Londoner or freighter receive a personal interpretation and an instrumental
interpretation, respectively.
Moreover, this kind of framework, as clearly stated by Barker (1998, Lieber 2007), is not
adequate for all affixes: -ee, for instance has been assumed to bind the patient/theme argument
of the base verb, but it does not account for the cases when the referent of the ee noun is the
indirect object (addressee) the object of a preposition (experimentee) or the cases where the
ee noun receives an agentive, or subject-oriented interpretation (escapee, attendee). Barker
argues instead for a semantic analysis of ee in which the affix binds an argument of the verb
under 3 conditions:
(i) argument is active participant (i.e. it is episodically linked to the verb)
(ii) (ii) it must denote something sentient
(iii) (iii) it must lack volitionality.
In a canonical case like employee the suffix binds the internal argument because it has
the above mentioned features (+sentient, -volitional). For indirect object interpretations like
addressee the affix cannot bind the direct internal as it is not [+sentient] and the external cannot
be bound because it is {+volitional}, hence the only argument that can be bound is the indirect
argument [goal]. Barker acknowledges , nervertheless, that the requirement [-volitional] is
problematic in cases like escapee, and other subject ee nouns such as retiree and attendee.
In work in progress, Lieber (2007) argues that the analysis of the meaning of affixes lies
at the level of lexical semantics, in the sense that beyond the binding of a base argument by the
affix one of the tasks is to identify the actual semantic contribution of the affixes.
5.2.2. Featural decomposition and the analysis of affixes
5.2.2.1.Relying on work by Jackendoff (1990), Wierzbicka(1996) and Szymanek(1986), Lieber
(2007) assumes that the description of nominal semantics requires the identification of a set of
primitives or atoms of meaning (in the form of semantic features) that might be used to
characterize the affixal semantic contribution.
The two important features she assumes are: [dynamic] and [ material]. These
features are used: (i) in a cross-categorial way; (ii) in both a n equipolent and a privative way, i.e.
the features will be binary in value (i.e positive or negative), but they may be either present or
absent in the semantic skeleton of a given lexical item; absence will indicate the irrelevance of
the semantic feature for the item in question.
The basic conceptual categories characterized by the two features are:
(i)
SUBSTANCES/THINGS/ESSENCES the notion correspondent of the syntactic
category N
(ii)
SITUATIONS - V & A
[material] defines the conceptual category (in (i)) correspondent of N . The positive
value denotes the presence of materiality, characterizing concrete nouns. The negative
value denotes the absence of materiality; it defines abstract nouns
72
[IEPS =Inferable eventual position or state] that signals the addition of a PATH
component of meaning. Only [+dynamic] verb classes bear the feature [IEPS]. [+IEPS]
and [-IEPS] are distinguished in terms of the kind of PATH they imply. i.e. [+IEPS]
roughly [inchoatives &unaccusatives: descend, fall, grow) and [-IEPS] [manner of
change: walk, run, amble, waver, fluctuate). This feature permits us to distinguish a
number of verb classes:
(9)
STATES
[-dynamic]
EVENTS
[+dynamic]
SIMPLE ACTIVITY
[+dynamic]
eat
hear
vary
Remain
Own
Barbaras constant war against her neighbours was distressing (complex event)
That/one war caused much destruction (result)
Lieber argues that, since the 2 readings are induced in a larger syntactic context, nouns
do not come with positive or negative specifications for the feature [dynamic]. The skeletons of
processual SUBSTANCES/THINGS/ESSENCES come with the feature [dynamic] unspecified for
value, further specification of this feature taking place at a higher level of semantic
interpretation.
73
UNACC/INCH
MANNER
[+dyn.+IEPS]
[+dyn,-
descend
grow
walk
fall
go
run
amble
IEPS]
Be
Out of context derived nouns like expression, assignment, development are compatible with
either of the two readings.
According to Lieber (1999, 2007:28) the same might be said for processual simplex
nouns like war ; with full expression of arguments a complex event reading is prominent and
with determiners like that/one and absence of an internal argument the result reading is
highlighted:
CHANGE
[+dynamic IEPS]
OF CHANGE
(8) (i)
(iii)
SITUATIONS
hold
kiss
listen
Typical members of the classes above will have the following skeletons:
(10)
Know [-dynamic] ([ ] [ ])
Eat [+dynamic] ([ ] [ ])
Grow [+dynamic, +IEPS] ([ ] [path ])
Walk [+dynamic, -IEPS] ([ ] [path ])
The typology above does not include the class of verbs known as CAUSATIVES e.g. I
grow flowers. These verbs are not simple situations; they consist of two sub-events: an ACTIVITY
[x does something to y] and a RESULT [such that x causes y to become/go to z] (Dowty 1979 ,
Hovav&Levin 1995: LCS: [x DO y] CAUSE [y BECOME z].
It is interesting to note that the only productive verb-forming suffixes are ize
(hospitalize, -ify (purify), and maybe en- (entomb, imprison) which we could characterize as
74
causative. One of the reasons could be that English has a productive non-affixal means of
word-formation , namely conversion (to open, to clear, to jail, to shovel etc)
5.3.2.2. According to Lieber, affixes like simple lexemes have skeletons and the semantic part of
derivation involves adding the affixal skeletons to the skeletons of the base, thereby
subordinating that skeleton. Affixal skeletons consist of functions and arguments. Affixation
will require the coindexation or binding of an affixal argument with a base argument. The
general assumption is that the vast majority of category-changing affixes in English and
indeed in all languages - add a function that corresponds in featural content to one of the major
semantic categories of simplex lexemes (Lieber 2007:36).
The expectation is that some lexeme-derivation (affixation, compounding conversion)
should create lexemes that fall into those semantic classes that exist for underived lexical items
(see 11):
(11)
affixal skeletons
[+material, dynamic ([ ], <base>])
[-material,dynamic ([ ], <base>])
[+material([ ], <base>])
[-material([ ], <base>])
[[+dynamic([ ], <base>])
[-dynamic([ ], <base>])
[+dynamic, +IEPS([ ], <base>])
[-dynamic,-IEPS([ ], <base>])
Bipartite skeleton
(which may also be assumed to be a compounding stem) which fulfills this function: glassware,
tinware, hardware, software, Delftware, etc. Several answers have been given :
(i) The vast bulk of simplex nouns are concrete, i.e. the vast majority is monomorphemic,
abstract nouns being formed through a process of nominalization (Lyons 1977)
(ii) English has a highly productive alternate means of word-formation for creating concrete
nouns, namely root-compounding. A new name can be created for a thing by simply
putting together two already existing stem nouns, thereby extending the simplex
noun lexicon (Lieber 2007)
6. Conclusions
Given the featural system suggested by Lieber we can identify the common semantic
content that affixes like : -er, -ee, -ant/ent, -ist share, namely they are affixes which create
concrete and processual nouns, characteristics that agents,instruments, patients, experiencers
all have in common.
The affixes ment, -ation etc can be characterized as abstract processual [-material,
dynamic] : examination, refusal, closure, amusement. They preserve the processual nature of
their bases while also making abstractions of them.
As already mentioned, these affixes form a constellation of rival affixes, i.e. they have
the same effect on the meaning of their bases; the particular differences are accounted for by
phonological and morphological properties induced, generally, by their origin (as already
mentioned).
Affixes like ness (happiness) and ity (purity) create abstract [-material] nouns from
adjectives; -hood (knighthood) and ship (friendship) form abstract nouns from concrete nouns.
Although they take different bases (A vs N) both sets of affixes mean the same thing, namely :
abstraction having to do with X (where X=the denotation of the base).
Stative , i.e. [-dynamic], items appear in the form of adjectives formed by affixes like ic
(dramatic), -ary (visionary), -ive (attractive), -al (architectural), -en (golden), -ous (poisonous) y
(fishy). Such adjectives are often called Relational adjectives and are glossed in a variety of ways,
such as: in the nature of X, pertaining to x, characterized by X, having X, belonging to X etc
(Marchand 1969 apud Lieber).The semantic content of all these affixes is that they place the
nominal or verbal bases into the semantic category states .
An odd thing about English is that there seem not to be any bona fide affixes that create
simple concrete [+material] nouns or simple verbs of any sort. The only semi-affix is -ware
75
76
Curslexicalsemantics5-6
The Criticism of Componential Analysis. The causal Theory of Reference
1. Major difficulties of CA The arbitrariness of the solutions
1.0. More kinship semantics
One difficulty of CA that was soon discovered by linguists is that for the same limited set of
terms, it was possible to write several analyses, which were all formally correct, i.e., discrete (each term
differed from any other term in the set in terms of at least one feature), and non-redundant (each tern
differed form any other one in terms of at most one feature).
To prove this point we will mention several equally plausible analyses of English kinship terms.
The existence of this situation raises two points:
a.
How does one choose among rival anlyses?
b.
Why is it possible to have several analyses which are equally justified?
1.1 Paul Kays analysis
This is a rigorously linguistic analysis because it pays the greatest attention to the relation between lexical
meaning and word-form.
The analyst acknowledges the following aims:
a. The analysis should be generative, because the kinship terms form an open set. Since there are
indefinitely many kinship terms, the analysis should be recursive. As evidence consider the semes:
grandfather, great-grandfather, great-great-grandfather, etc.; similarly, consider, first cousin, second
cousin, third cousin, etc.
The analyst should produce rules that generate terms, rather than a finite output.
b. The analysis should be internally motivated; it must explicate all instances of co-variation between sound
and meaning: e.g., the semantic analyst must show that in English, the term brother-in-law is composed of
BROTHER-[ IN- LAW],(which is not the case in the Romanian equivalent
c. The analysis must provide an account of the relations of paraphrase, entailment, contradiction, a.s.o. that
obtain between sentences containing the kinship terms:
Lecture 5
COMPOUNDS
and
COMPOUNDING
(1)
Marion is male.
Marion is the brother of Sydneys mother.
Conceptual basis of Kays kinship semantics. The system has two primitives:offspring (x, y), spouse(x, y)
OFFSPRING
SPOUSE
x is an offspring of y
x is a spouse of y
The primitives are further used to define the concepts; blood-parent and law-parent, since P (parenthood)
can be either blood parenthood or law-parenthood.
(1)
(2)
The relation of parenthood (=P) obtains between two individuals when they are in the relation x is
the blood-parent of y or x is the law-parent of y or both. In fact in the Western world there is a cultural
presupposition that:
BLOOD-P (x, y) LAW-P (x, y)
One may define the English terns parent and child which are converse terms. From this point
on, Kay uses an anthropological notation: any blood relationship is viewed as a sequence of child-of [Q],
77
78
parent-of [P], relations, down from and up to the common ancestor. The common ancestor is as usual the
zero point of the system:
(3)
A. Basic terms
(4)
Q0P0
self
0 1
parent-child
1 1
SIBLING
1 2
QP
NUNCLE-NIBLING
Q2P2
cousin
QP
QP
father
a. Ted is my father/ brother/ son, more precisely he is my step/ foster father/ brother/ son.
b. Jane is my mother/sister/daughter, more precisely she is my step/foster mother /sister/daughter.
c. From now on, I will no longer consider you may father/ brother/ mother/ siter.
d. How could you pretend to be my father/ mother/ sister when you did that?!
e. Mary is a good/ terrible/ bad mother/ sister/ teacher/ wife/ friend.
f. I certainly know my father; he did his best to raise me. I have, howeve, never met my genitor.
X
P1
X
Q1
X
sister
Basic terms:
X
P2
X
uncle/aunt
2
P
X
society properly.
Q1
X
Q
X
Kinship terminology is integrated in a larger theory of social roles which is large enough to accommodate
some of the supplementart meanings of the kinship terms.
cousin
PARENT (x, y)
=
FATHER/MOTHER (x, y)
=
CHILD (y, x)
=
SON/ DAUGHTER (y, x)
=
FOSTER FATHER/MOTHER (x, y)
Parent-child (x, y)
just if P (x, y)
just if P(x, y) and FEMALE (x)
just if P (x, y) and MALE (x)
just if CHILD (y, x) and MALE (y)
just if CHILD (y, x) and FEMALE (y)
Q1P1
BROTHER (x, y)
SISTER (x, y)
=
=
Q1P2
NUNCLE (x, y)
NIBLING (y, x)
UNCLE (x, y)
AUNT (x, y)
NEPHEW (y, x)
NIECE (y, x)
=
=
=
=
=
=
C. English kinship terms via semantic functors. Functors are items which operate on the meaning
of an argument-term to yield another term; in the kinship system, great, in-law, foster are functors. For
instance, great introduces one more generation on the parent line
GREAT (Q1Pj)
Q1P j+1 =
e.g. great-grandfather = great (Q0 P2) = Q0P3
Evaluation of the analysis
a. The analysis is morphologically motivated.
(6)
79
RAISE (x, y)
MALE/FEMALE (x) & RAISE (x, y)
BE RAISED y BY x
MALE/FEMALE (y) & BE RAISED y BY x
= MALE /FEMALE (x) & RAISE (x, y) & ADOPT
(x, y)
1.3.
The biological model: Wierzbieckas (1982) attempt to find true semantic primitives
Goals of the analysis: to uncover the basic concepts that are cognitively relevant in various languages. She
finds fault with the features traditionally used by componential analysts.
I would maintain that male and female, child of, parent-of, ancestor, descendant,
young and adult should not be regarded as primitive semantic units for the following reasons.
1) All these notions are highly specific, confined to one particular domain of the vocabulary; it is
to be expected that the genuine semantic primitives will turn out to be extremely versatile, pervading all
areas of the vocabulary; part, become are notions of this kind, but not male, female, parent,
adult.
2) Words such as male, female, adult, parent are highly abstract and learned they do not belong
to the basic vocabulary of language users; in fact, they are incomparably less basic and intuitive than the
words man, woman, mother, father, boy, girl, which they purport to explain.
3) The attractive simplicity, symmetry and elegance of the binary oppositions (male/female,
young/ adult, child-of/ parent-of correspond more to the longing of researchers than to the realities of
language and cognition. The idea that a mother is a kind of female father allows for considerable symmetry
and simplicity of description, but it also seems to be crassly false.
It is unnatural to say that a child is a non-adult human being, a child is a boy or a girl. But even
that does not do. Boy and girl may be used about young people who have ceased to be children long
ago. In contrast to child, kitten does not mean he-kitten or she kitten, but small young cat.
80
The author goes on to present the following types of definitions, which she founds cognitively
natural
(8)
JOHN S MOTHER the human being inside whose body (once) there was something that was becoming
Johns body'
JOHN S FATHER the human being who once caused a woman to have inside her body something that was
becoming Johns body
JOHN S PARENTS Johns father or mother
JOHN S PARENT one of Johns parents
WOMAN
human being that could be someones mother
MAN
human being that could cause a woman to be someones mother
BOY
young human being that one thinks of as becoming a man
GIRL
young human being that one thinks of as becoming a woman
CHILD
small young human body
1.4. Discussion
All the authors are aware of the danger of arbitrariness, hence all seek to find motivation for the
semes they employ:
a.
by increasing the range of the data they consider (for instance, the inclusion of
supplementary meanings of the terms), and by strictly observing the formal requirements of the analysis,
and increasing the sophistication of the linguistic tools employed in the analysis.
b.
by using an empirical methodology, so as to eliminate subjectivism in the choice of the
primitives;
c.
by appealing to an idea of naturalness and cognitive psychological relevance.
2. Componential definitions as eliminative definitions. Classical categorization: the model of
necessary and sufficient conditions
2.1 CA and classical categorization
As known, one of the acknowledged aims of CA is to propose componential definitions which specify the
necessary and sufficient conditions that an object must satisfy if it is to be designated by the analyzed term/
things and indicate their individuality and whose destruction causes destruction of the whole (Metphysics,
8.8.3). Accidents are incidental properties which play no part in determining what a thing is. Accidents
means that which applies to something and is truly stated, but is neither necessarily, nor usually
(Metaphysics: 5.301). As an example, the essence of man: a two-footed animal, a rational creature. That a
man is white or cultivated is accidental; these attributes might be true of an individual, but they are
irrelevant in determining whether that individual is indeed a man.
Moreover, there is a similarity between lexemes and concepts: both the concept MAN and the
meaning of the lexeme man are defined by the same formula, logos, of the essence.
Later influential presentations of the same doctrine are available in Locks Essay on Human
Understanding(). See also Lewis (1943), Copi (1992).
The same view is coherently articulated in experimental psychology, in 1920, in Hulls
Quantitative Aspects of the Evaluation of Concepts.
Our presentation of the central features of this traditional view takes into account the
philosophically relevant semantic dichotomy, proposed by Carnap (1947), following Frege: intension
(sense) /extension (reference).
A first question is what is the relation between concepts and categorization? Briefly, concepts are
the result of categorization, of analyzing and classifying our experience and the objects around us,
The basic statements of the classical theory of the meaning of terms and, thus, of categories are the
following:
1. Each meaningful term has some concept or intension or cluster of features associated with it. It is this
meaning that is present to the mind when the term is understood. The mental representation of a concept is
a summary description, a pattern of recognition of an entire class.
2. The features in the intension/ concept are singly necessary and jointly sufficient to define the referent
(extension) of the term. For a feature to be singly necessary, every object that qualifies as a referent must
have it, and for a set of features to be jointly sufficient, every entity having that set of features must be in
the extension of the concept. Hence the name of the necessary and sufficient conditions model.
The model of NSC is easily verified for concepts like square:
(10)
(9)
father
+ Kinsman
+male
+ Generation +1
+Co-lineal
+Direct rank
squares:
closed figure
four-sided
equal angles
equal sides
We will refer to such necessary and sufficient features as defining features. Let us also recall,
Frege-Carnaps thesis that intension determines extenstion. Satisfaction of the features in the intension
cluster determines membership in the extension class.
3. Since the features are necessary, representing defining or criterial attributes, they offer the basis of
analytic statements.
If P is a property in some concept T, then the statement All Ts are P is true by definition
(11)
All Ts are P
Remark. Aristotle distinguished between the essence of thing and its accidents. The essence is
that which makes a thing what it is; the essence is defined as all parts immanent in things, which define
Criterial attributes are essential in the Aristotelian sense. Notice the difference between accidental
and essential properties: Dogs are animals./ Dogs are good pets.
This is also the distinction between the lexicon and the encycloapedia, which is richer, containing
essential as well as accidental or contingent attributes. Remember the structuralist view (Coseriu) of
meaning:
Le sens dun mot est un ensemble de traits rferentilelles, de traits que doit possder un segment
de la ralite pour tre dnomm par ce mot.
4. This conception of meaning leads to certain characteristics of categories:
81
82
(i)
The conception that categories have clearly delimited boundaries.
(ii)
Membership in a category is a yes/no matter; thus, a certain n geometric figure either is or is not a
square, according as it does or does not satisfy all the necessary and sufficient conditions for membership in
the category square.
(iii)
All the members of a category have equal status; because each member possesses all the required
properties.
The model of the NSC has come under the fire of severe criticism from linguistic, philosophical
and psychological quarters, all of which have recently adopted a cognitivist perspective.
3. Against the NSC model
The model of NSC both as a model of conceptual analysis and as a model of lexical analysis has been
found to be deficient in various ways.
3.1 One type of criticism addresses CA as a model of lexical analysis, and addresses the
controversial issue of the synonymy of the CA definition with the lexical term:
(12)
bachelor
UNMARRIED MAN
Since the definition includes only NSCs (avoiding redundancy), the definition should be minimal;
componential definitions do not include supplementary, accidental meaning components.
Componential definitions include only obligatory components, which, regardless of any context,
represent what every speaker must know From a different perspective, we hit upon the same distinction
between the minimal componential definition which represents linguistic knowledge and encyclopedic
entries which contain non-obligatory semantic components, encapsulating knowledge of the world.
This position is clearly stated and illustrated by Bendix (1970) in an essential study on verbs of
possession and transfer of possession.
The meanings of the terms in the language are presented as standing in opposition nto one
another within the system of the language and as being distinguished by discrete semantic components
acting as distinctive features. Our aim is to restrict ourselves to minimal definitions, employing only
criterial attributes.
A minimal definition of the meaning of an item will be a statement of the semantic components
necessary and sufficient to distinguish meaning paradigmatically.
In accordance with this view, one may analyze the verb give as cause to have, rather than
cause to get. Get entails the notion of change which have does not, hence cause to get is
semantically more complex than cause to have. If definitions are minimal, then give means cause to
have, rather than cause to get, because there is no other English word which means cause to get, rather
than cause to have. In agreement with his theoretical assumptions, Bendix offers the following analysis of
the verb lend:
C has B before time T
(13)
C lends B to A
C causes A to have B after T
B is not As
study which explicitly adopts this position is Millers analysis of English verbs of movement, conducted
with experimental, psycholinguistic methods.
An incomplete definition is a suitable paraphrase that has a more general meaning than the word it
replaces: the word implies the incomplete definition, but the incomplete definition does not imply the word
defined.
travel CHANGE LOCATION
change location travel
Thus the gist of the criticism is that, in the first place, minimal definitions do not do justice to our
intuitions about words mean, about how we understand and use them. Secondly, as a general theoretical
result, we retain that as a rule definitions are incomplete (exceptions are few and motivated, see below).
3.2. Does intension (properly) determine extension?
One thesis, implicitly present in the CA model, and which has been intensely scrutinzed from philosophical
quarters is that intensions (concepts) viewed as bundles of NSC correctly determine extension. In the
seventies and eighties this thesis has been challenged by philosophers within the framework of the so-called
causal or historical theories of reference.(Putnam, 1975, Donellan, 1982). The following objections were
brought against this thesis:
a. The first obvious objection was that not all objects which make up the extension of the term
have all or even most of the characteristics stipulated in the definition. Even if Lemons are yellow and
Tigers are striped and Cats are four-legged, there are green lemons, unstriped tigers and three-legged
cates.
b. It is not always the case that intension determines extension. An example which has become
famous is Putnams Twin Earth Example.
Putnams Twin Earth Story
Twin Earth is a planet in all respects like Earth. The only peculiarity of Twin Earth is that the
liquid called water is not H2O, but a different liquid whose chemical formula is very long, say XYZ. We
suppose that XYZ is indistinguishable from water at normal temperature and pressure. In particular, it
tastes like water and it quenches thirst like water.
A spaceship from Earth first decides that water has the same meaning on TE and E. This
supposition will be corrected when it is discovered that water on TE is XYZ and the Earthian spaceship will
report something like the following: on the basis of the following observations:
(15)
Eearth
Twin Earth
water
water
LIQUID
COLORLESS
TASTELESS
ODORLESS
.
.
.
H2O
Fillmore (1976) rightly complains that this analysis ignores the borrowers obligation to return the
object, a fact which is essential to the lending-borrowing transaction.
The answers to the problem of minimal definitions were varied.
It was admitted that even in minimal pairs words may differ in terms of more than one feature. Heres an
examples from Bidu-Vranceanus analysis of manner of speech verbs:
(14)
a vocifera :
a sa rasti :
a vorbi
a glasui
83
LIQUID
COLORLESS
TASTELESS
ODORLESS
.
.
.
XYZ
On TE, the word water means XYZ, determining the extension XYZ.
Conclusions : a. The same intension (cluster of properties), the same definition determines
different extensions, H2O vs XYZ. Therefore, one cannot say that intension correctly determines extension.
b. The situation is not improved if we equate intension with cognitive state of a speaker, as
determined when entertaining a concept. In that case, to have an intension/concept means to be in a certain
psychological state. Two speakers on E and TE may be in the same psychological state, they may have the
same concept of water and yet they determine different extensions.
3.3 This example raises two important questions: a) how does one actually determine the extension
of a term (if intensions are unreliable)? B) Secondly, what does the intension of a term describe, if it does
not always accurately describe the extension of the term. The answer is that the definition correctly
describes a prototypical instance of the class the term is purported to be true of.
3.3.1 The stereotype
84
The properties included in the definition are not always true of all the members that make up the
extension of the term. The properties included in the definition should not be regarded as offering the basis
for analytic statements.
(16)
Discussing the problem of analyticity in NLs, Putnam (1975) concludes that analytic definitions
can be given only for single-criterion concepts. For these, if we give up that single criterion, the identity of
the concept is destroyed.
(17)
bachelor
vixen
oculist
=
=
=
unmarried man
female fox
eye doctor
Natural kind terms and most general terms in NLs are cluster concepts, since their meaning is
given by a cluster of concepts (of criterial attributes).
Man, for instance, is defined as a set of propertiesP1,P2,.Pn. For each such property, we can ask
whether there could be a man without the given property and in many cases (but not all), the answer would
be affirmative, so that This man is P1. is not an analytic statement.
Moreover, the features that make up the intension are revisable, with the growth of knowledge. A
whale is a fish has become A whale is a mammal, atoms are divisible particle in spite of the fact that
their name was given at a time when they were thought to be the smallest, indivisible particles. Terms of
science, such as energy, electricity, light etc. have been defined in terms of a cluster of laws, some of which
have been abandoned, etc. Thus the different properties or laws that constitute, at any given time, the
meaning of a cluster concept represents that knowledge of the word that has passed into the meaning of the
words. The difference between such hypotheses (contained in the meaning of terms) and the hypotheses
expressible only in sentences of the language is pragmatic in nature and has been stated by Quine as a
difference between language and theory. But any of the hypotheses contained in the meaning of the word,
like any theoretical hypothesis may be subject to revision.
Conclusion
The features in the intension of a term represent a stereotype, that knowledge or
that part of the theory about the extension which has passed into the language.
3.3.2 The determination of extension. Putnams division of linguistic labor.
Recall that one problem raised by Putnams examples was: Since the intension does not determine
extension, how do we determine the extension of a general term?
Partial answers to this question were supplied in the framework of the causal or historical account
of reference. According to this orientation reference fixing mechanisms are typically non-definitional. For
most natural kind terms, reference is fixed through some kind of causal interaction between the users of the
term and samples of the natural kind to which the term refers. Reference is passed along the reference chain
of the community. The causal explanation of reference starts from an attempt of defending the position that
general terms must be understood referentially and that they name real essences, even if we do not possess
conventional unrevisable definitions for them. (recall the case of whale and atoms). Putnam suggests
that the correct position to adopt is that of metaphysical realism.
There are several non-definitional ceremonies of introducing new terms:
(a) ostensive definitions For instance, This is water with an accompanying gesture, The sentence is uttered
with the presupposition the body of liquiud that the speaker is pointing to bears a certain sameness relation
(x is the same liquid as y) to what must members of the community will call water.
(b) giving a stereotype Reference may be introduced via associated properties, which function as an
operational definition or a stereotype of the normal member of the extension. The members of the extension
are sufficiently the same as the typical member.
In both instances we are dealing with judgments of similarity evolving the notion same x as y.
But the relation same x as y is a theoretical relation; whether something is or is not the same liquid as this
may take an indeterminate amount of scientific investigation to determine, and depends on how advanced
85
our tools are etc.. (A spaceship from Earth before Lavoisier, would not have been in a position to describe
the difference between Earth water and Twin Earth water,
Coming back to Putnams water on TE examples, one might adopt two different attitudes to the
facts of the case:
(i)
We may adopt a conventionalist position In that case, we consider the properties in the intension
of water are taken as constant and unrevisable. Water is any stuff that satisfies the NSC. In that case, TE
water, i.e., XYZ counts as water, i.e., water has a constant meaning/intension and a world/relative
extension.
(ii)
We may adopt a realist stance In that case, we consider that the term water names a real essence of
this (our) world. a stuff with a particular hidden structure (described chemically as H20). But then, the
stuff called water on TE is not water. But then, in some sense, water no longer has the same meaning on E
and TW since different extensions are determined, and it appears that extension is part of the terms
semantics.
The two positions embody different epistemological attitudes. On the first, conventionalist
(traditionalist) position, water, gold, tiger etc. designate whatever satisfies the current operational
definitions, since the extension is just what the term is true of, and truth has become an intra-theoretic
notion depending on the particular definitions endorsed by our current theory. We call water a liquid which
bears the relation same liquid to water in the actual world. that is what we intend when we give an
ostensive definition.
Water is H2) is a metaphysically true statement (i.e., it is true in all possible worlds), although it
is epistemologically contingent.
3.3.3 It also appears that terms like water, gold, etc. have an unnoticed indexical component, water
is stuff that bears a certain similarity relation to the water around here. Water at another time or place must
bear the relation same Liquid to our water in order to be water. Thus Putnam concludes the theory that
(1) words have intensions, which are something like concepts associated with words by speakers and that
(2) intension determines extension cannot be true of natural kind terms like water, for the same reason that
it cannot be true of indexical words like I. We cannot ignore the contribution of the context, the real world
in determining the hidden structure, the essence of the term. When Archimedes asserted that something
was gold, he was not saying that it had the superficial characteristics of gold, he was saying that it had the
same hidden structure as any normal piece of gold. When we introduce the term, we need not know the
real essence of the substance we are naming. We merely hope that knowledge of the real essence will be the
product of empirical research. Once the term has been introduced, it will pass from person to person,
through a referential chain, which provides the historical link between the different uses of the term,
maintaining and intending to maintain reference.
3.3.4. The example also suggests that extension is not always determined at the level of the
individual speakers, but at the level of the community. To take one more example, two English speakers
would agree that the words elm-tree and beech-tree have different meanings and extensions, although, quite
possibly, they may not be able to tell an elm-tree from a beech-tree. The possibility of using such words in
true sentences would not exist, if there did not exist in the community experts who can identify elm-trees
and beech trees, thus, only experts can determine the extension of words like molybdenum, gold, hedgesparrow, etc.
There exists, Putnam believes, a sort of division of linguistic labor. Any community possesses at
least some terms whose associated criteria are known only to a subset of the speakers who acquire the term
and whose use by the others depends upon a structured cooperation between most experts and the experts.
Thus the correct correspondence between linguistic expressions and the world may be determined by the
speech community as a whole, but may not necessarily be reflected in the knowledge of each member of
the community.
3.4 More on componential definitions as descriptions of prototypes As mentioned above, one other
problem raised by Putnams story regards the role of the componential definition, Since it does not really
include the necessary and sufficient conditions for membership in the extension, it is no longer clear what
its functions is.
A natural suggestion is that a componential definition describes a typical member of the extension
a prototype (this is a cognitive notion, explained below). The linguistic description of a prototype is
dubbed by Putnam a stereotype; A stereotype is an operational definition; membership in category is based
on a judgment of family resemblance with the prototype. In terms of language use, the stereotype represents
amy speakers minimal linguistic obligation any speaker who knows the word is supposed to know the
86
stereotype. Thus, for most words, there is a certain minimum number of features (entailments) that all
speakers of the language are supposed to know, even though these features may happen not to single out
the extension reliably. For examples, English speakers are supposed to know that tigers are large, cat-like
animals that are orange and have black stripes, even if these features (entailments) are neither necessary,
nor sufficient to distinguish tigers from non-tigers accurately.
Putnam proposes to divide the description of the semantics of concrete terms into two parts:
(1)
the determination of extension this presupposes the sociolinguistic division of labor, it is a
matter of sociolinguistics and it is based on a structured co-operation between speakers.
(2)
the description pf semantic competence, the particular ideas and skills required for an individual to
use a word correctly; at least for some terms this knowledge is in fact knowledge of a stereotype, that is
knowledge of a description of a prototypical member. The stereotype is a linguistic entity, the prototype is a
cognitive, psychological entity. The concept of prototype originates in the psychological studies of
categorization (see Rosch and associates).
87
Lecture 6
Cognitive Semantics:
From Prototype Theory to
Frames and Idealized
Cognitive Models
88