Chapter 18 Lexical Functions: Description of Lexical Relations in a ...

Chapter 18 Lexical Functions: Description of Lexical Relations in a Lexicon.......................................................... 2
1 Lexical Correlates of a Lexical Unit ................................................................................................................... 2
1.1 Introductory Remarks ................................................................................................................................ 2
1.2 Semantic Derivations ................................................................................................................................. 4
1.3 Collocations................................................................................................................................................ 9
1.4. Overlap of Semantic Derivations and Collocations..............................................................................10
2. The Notion of Lexical Function .......................................................................................................................12
2.1. Introductory Remarks .............................................................................................................................12
2.2. Definition of Lexical Function...............................................................................................................15
2.3 Definition of Standard Lexical Function ................................................................................................19
2.4 Examples of Non-Standard Lexical Functions.......................................................................................20
2.5 Classification of Lexical Functions.........................................................................................................22
3 The List of Simple Standard Lexical Functions...............................................................................................26
3.1 Introductory Remarks ..............................................................................................................................26
3.2. Paradigmatic Lexical Functions (1-26)..................................................................................................27
3.3. Syntagmatic Lexical Functions (27-64).................................................................................................43
3.3.1. Nominal Lexical Functions(27)......................................................................................................43
3.3.2. Adjectival/Adverbial Lexical Functions (28-34) ..........................................................................44
3.3.3. Prepositional Lexical Functions (35-40)........................................................................................48
3.3.4. Verbal Lexical Functions (41-64) ..................................................................................................50
4. Special Phenomena Related to Lexical Functions ..........................................................................................66
4.1. Complex Lexical Functions....................................................................................................................66
4.2. Configurations of Lexical Functions .....................................................................................................67
4.3 Fused Elements of Values of Lexical Functions....................................................................................68
5. Presentation of the Values of Lexical Functions in the Lexical Entries of the Keywords ...........................68
5.1. Elements of the Value f(L) as Subentries within L’s Entry .................................................................69
5.2. Generalizing over the Values of Lexical Functions..............................................................................73
5.3. Ellipsis of the keyword of an LF............................................................................................................77
6 Linguistic Nature of Lexical Functions ............................................................................................................77
6.1 Semantic Aspect of Lexical Functions ...................................................................................................77
6.1.1 Lexical Functions Are Not Semantic Units ....................................................................................78
6.1.2 Semantic Subtypes of Lexical Functions........................................................................................80
6.1.3 Different Semantic ‘Facets’ of a Keyword.....................................................................................82
6.1.4 Lexical Functions and Their Relation to the Definitions of their Keywords ...............................83
6.1.5 LFs as Deep Lexical Units...............................................................................................................85
6.2. Phraseological Aspect of Lexical Functions .........................................................................................86
6.2.1. Graduality of Restrictedness of LF Expressions ...........................................................................86
6.2.2. LF Expressions and Collocations...................................................................................................87
6.3 Lexical Functions in Linguistic Representations ...................................................................................88
6.4 Universality of LFs ..................................................................................................................................89
7 New Simple Standard LFs .................................................................................................................................93
8 Lexical Functions in Computer Applications...................................................................................................95
8.1. LFs and Lexical Choices ........................................................................................................................96
8.2. LFs and Communicative Structure ........................................................................................................98
8.3. LFs and Text Cohesion ...........................................................................................................................99
Bibliographic Remarks........................................................................................................................................100
Notes.....................................................................................................................................................................100

Chapter 18
Lexical Functions:
—Chapter 18. Lexical Functions— 2
Description of Lexical Relations in a Lexicon
Žolkovskij & Mel´čuk 1966, 1967, Mel´čuk 1974: 78ff, Mel’čuk 1982, 1988a: 61ff, 1996,
[78], [79: 55-66], [35], Alonso Ramos 1993, Polguère 2000, Steinlin et al. 2004.
To speak of Lexical Functions, one needs first the notion of lexical correlates and
those of their two major types: 1) semantic derivations and 2) collocates; I begin
with sketching these three notions. Based on them, Lexical Functions are introduced and studied
in detail. Accordingly, Chapter 18 is organized in seven sections:
1. Lexical Correlates
2. The Notion of Lexical Function
3. Review of Simple Standard Lexical Functions
4. Other types of Lexical Functions
5. Presentation of the Values of Lexical Functions in an ECD
6. Linguistic Nature of Lexical Functions
7. Exhaustiveness of the Simple Standard Lexical Functions Inventory
1 Lexical Correlates of a Lexical Unit
1.1 Introductory Remarks
In the process of text production, the Speaker is faced at the outset with the problem of lexical
choices: in order to go from a given Sem(antic) R(epresentation) [= SemR] to a corresponding
D(eep-)Synt(actic) R(epresentation) [= DSyntR] of the sentence he intends to utter, the Speaker
has to select lexical units [= LUs]—lexemes and phrasemes (that is, words and idioms)—that he
will use to build his sentence. In this perspective, two types of LUs have to be distinguished.
• The LUs of the first type constitute the huge majority of the LUs of any language L. They
could be somewhat loosely called free LUs. In text production a free LU L is selected by the
Speaker strictly according to its meaning, and it is looked for—in the Speaker’s lexical stock—
independently of other lexical choices, previously made or to be made. Thus, if the Speaker needs
to refer to a (four-wheel motor vehicle designed for transport of a few people ...) he says CAR.
These are semantically-driven free lexical choices.

—Chapter 18. Lexical Functions— 3
• The LUs of the second type are different in this respect. The selection of such an LU is
contingent upon other lexical choices—such that its own meaning can be more or less disregarded
in the search and selection process. These are restricted LUs. The Speaker looks for a
restricted LU (in his lexical stock) based on some other LUs he has already chosen. These
lexically-driven restricted lexical choices are carried out along the two major linguistic axes:
paradigmatic and syntagmatic.
On the one hand, the Speaker, while speaking of cars, may need to find the name for a
(small business which sells gas and oil for automobiles ...) [= GAS STATION] or for a (little plaque
which is designed to be fixed on a vehicle and on which the vehicle’s registration number is
printed) [= LICENSE PLATE]; this is a paradigmatic choice, and the corresponding search is
done starting with CAR, AUTOMOBILE or VEHICLE, which must have pointers to GAS STATION
and LICENSE PLATE. On the other hand, the Speaker may need to find the name for an event or a
property involving a car, e.g., when (the wheels of a car lose their grip on the road, so that the car
slides sideways) [= The car SKIDS] or to say that his (car uses up too much gasoline) [= This car is
a GAS-GUZZLER]; this is a syntagmatic choice, the search for the necessary LU also being
conducted under the control of the LU CAR. Underlying both types of lexical choices are relations
that link LUs in a lexicon: in both cases what the Speaker needs are lexical correlates of
the starting LU he has in mind.
Lexical Functions [= LFs], the object of this chapter, are a set of formal tools designed to
describe, in a systematic and compact way, all types of genuine lexical relations 1 that
hold between LUs of any language. This amounts to specifying, for any given LU, all the lexical
correlates that might be needed in the process of speaking. LFs are also extremely convenient for
paraphrasing on the DSynt-level—a topic that is covered in Part IV, Chapter 14, p. 00ff.
Lexical Functions were first introduced in 1965 (Žolkovskij & Mel´čuk 1965, 1966 and
1967). 2 They are aimed at describing ‘in parallel’ both of the two above-mentioned types of
lexical correlates, that is, the paradigmatic and the syntagmatic choices of LUs which depend on
previously made lexical choices. These types have mostly been considered separately in
linguistics, but turn out to be of the same logical nature: both are readily amenable to a
description via the concept of function in the mathematical sense. Thus, for an LU L, the LFs
cover both types of lexical correlates: 1) paradigmatic lexical correlates, or semantic
derivations of L; and 2) syntagmatic lexical correlates of L, or collocates of L, which
form collocations with L.

—Chapter 18. Lexical Functions— 4
Let us first consider these two types of lexical correlates in turn and then show that there is
a considerable overlap between them.
NB : In this chapter, as in the whole book, we deal exclusively with lexical units—that is, welldistinguished
lexicographic senses rather than polysemous words. Therefore, to be really rigorous, I should
indicate for each word I use as an example the corresponding lexicographic number. This, however, is
practically impossible: on the one hand, there is no fully reliable dictionary from which such numbers could
be taken; on the other hand, lexicographic numbers would of course encumber the reading. As a result, the
reader must accept that some examples will be ambiguous. However, I try to organize my examples in such
a way that the polysemy is resolved as much as possible.
1.2 Semantic Derivations
Suppose the Speaker starts with the LU L, choosing it as the keyword—the headword
of a lexical entry «L» in the dictionary of language L, from which appropriate pointers refer the
user to L’s lexical correlates he is looking for. Paradigmatic lexical correlates {L´ i | 1 ≤ i ≤ n}PARAD
of L can be loosely qualified as being ‘(quasi-)synonymous’ with L or ‘(quasi-)derived’ from L.
If L refers to a situation/an object P, L´ can designate a situation/an object close to or identical
with P, a part of P, a participant in the situation P, a generic notion for P, etc. For instance:
• L = SCHOOLN
{L´ i }PARAD = {PRINCIPAL, TEACHER, STUDENT, SUBJECT, EXAM, LESSON, HOMEWORK, RECESS,
• L = ESCAPEV
MARK, CLASS, CLASSROOM, TEXTBOOK, COLLEGE, UNIVERSITY, kINSTITUTION OF
LEARNINGl, ...}
{L´ i }PARAD = {FLEE, kBREAK AWAYl [from N], ESCAPEN, JAILBREAK, ESCAPEE, FUGITIVE, kCORREC-
TIONAL INSTITUTIONl, kPLACE OF CONFINEMENTl, kTUNNEL OUTl [of N], PURSUE,
PURSUIT, kBE ON THE LAMl, ...}
Such lexical correlates are typically used in text instead of their keyword L; L´ and L are in
an “either—or” relation. If, in text synthesis, the Speaker decides to express the meaning (L) or a
related meaning via L´—one of L’s paradigmatic lexical correlates (rather than via L itself), then,
as a rule, the LU L does not appear in his sentence. Consider the meaning (John teaches math to
Mary); if we select the noun TEACHER or PUPIL to express it, then the verb TEACH cannot be
used, and we say John is Mary’s math teacher or Mary is John’s pupil in math.
Paradigmatic groupings of lexemes are based on the fact that the signifieds of any two LUs
L 1 and L 2 are linked by one of the four set-theoretical relations:
1) they can be equal: (L1) = (L2);
2) one can strictly include the other: (L1) ⊃ (L2) [and (L1) ⊄ (L2)];

—Chapter 18. Lexical Functions— 5
3) they can have a strict non-empty intersection: (L1) ∩ (L2) ≠ Λ [and (L1) ⊄ (L2), (L2) ⊄ (L1)];
or 4) they can have an empty intersection, i.e., be disjoint: (L1) ∩ (L2) = Λ.
These relations between signifieds give rise to several lexical relations well-known in linguistics;
for instance:
• Lexical synonymy (exact and approximate semantic equality): exact synonyms: PUMA ~
COUGAR ~ MOUNTAIN LION, TORNADO ~ TWISTER; approximate synonyms: FIREV [upon N] ~
SHELLV [N], MACHINE-GUNV [N]; OBSTACLE ~ SNAG.
• Lexical antonymy (a special case of semantic inclusion: the extra element in one of the two
antonyms is negation): ALLOW ~ FORBID (as in Mom allowed/forbade me to go to this party: (X
allows Y to Z) ≈ (X says that Y may Z), and (X forbids Y to Z) ≈ (X says that Y may not Z));
NOISE ~ SILENCE ((noise) ≈ (presence of more or less intense sound ...), and (silence) ≈ (no
presence of sound ...)).
• Lexical conversion (semantic equality with the permutation of Deep-Syntactic actants):
FOLLOW ~ PRECEDE ((X follows Y) = (Y precedes X)); LONGER ~ SHORTER ((X is.longer.than Y)
= (Y is.shorter.than X)).
• Several derivational relations (semantic identity or inclusion): action noun (RUNV ~ RUNN),
agent noun (RUN ~ RUNNER), patient noun (PAY ~ PAYEE), place noun (SKATE ~ SKATING RINK),
etc.
Some of these relations between LUs in some languages are regular and morphologically
marked: thus, runn+er and swimm+er, as thousands of similar English nouns, have a suffix that
identifies them as agent nouns derived from run and swim. These are instances of morpholo-
gical derivation. However, in quite a number of cases when an important lexical relation
is present between L1 and L2 the corresponding link is either morphologically not marked at all or
is marked in an irregular way. Thus, someone who steals is not a *steal+er, but a thief, and
someone who flees is not a *flee+er, but a refugee or a fugitive. Here are three examples of this
phenomenon, which will interest us now.
1) In Arabic, the noun mi+ft-a-ħ (key) lit. (opener) is formally derived from the verb
fataħ(+a) ([to] open), while the nouns Eng. key or Fr. clé are not formally derived from any verb;
nonetheless, the semantic relation between the noun (key) and the verb ([to] open) is the same in
all these languages (roughly, (key) = (implement for opening a lock in a particular way)):
(miftaħ) (key) (clé)
_ ——— = ——— = ———
(fataħa) (openV) (ouvrir)

—Chapter 18. Lexical Functions— 6
2) In German, Schül+er is formally related to Schule (school), but Eng. student is not for-
mally related to school; for Ger. Häftling (prisoner) the situation is inverse: it is formally unrelat-
ed to Gefängnis (prison), but Eng. prison+er is formally related to prison. However, taken seman-
tically, student is to school what prison+er is to prison, that is, (person who is acted upon by N):
(student) (prisoner) (Schüler) (Häftling)
———— = ———— = ———— = —————
(school) (prison) (Schule) (Gefängnis)
3) Bedroom translates in Russian as spa+l´nja, from spa(-t´) (sleep), while dining room is
stol+ovaja, from the noun stol (table), rather than from the verb es(-t´) (eat); but again semantical-
ly these pairs are proportional:
(bedroom) (dining room) (spal´nja) (stolovaja)
———— = —————— = ———— = ————
(sleep) (eat) (spat´) (est´)
In the last example the meaning added in the numerator—i.e., in the meaning of the noun
L´ with respect to the meaning of the nominator verb L—is the same in all pairs: (room designed
to L in it). This ‘added’ meaning appears in many other LUs (cf. sitting/living room, study,
kitchen, bathroom, nursery, play room, ball room, storage room, etc.) of many languages. At the
same time, in English the meaning (room designed to L in it) does not have a standard expression
which could be specified by a lexical entry or a general (derivational) rule. Therefore, the
corresponding semantic relation between L and its lexical correlate L´—as well as all similar
relations—must be explicitly shown in the dictionary under L (in this case, under ROOM/Rus.
KOMNATA (room)). This can be done in the form of cross-references—from L to all its paradig-
matic lexical correlates {L´ i }. Paradigmatic lexical correlates {L´ i }PARAD of L are semantic
derivations of L.
Definition 18.1: Semantic derivation of a lexical unit L
Let there be, in the lexicon of language L, two LUs—L and L´; L is taken as the starting
point, i.e., the keyword.

—Chapter 18. Lexical Functions— 7
The LU L´ is a semantic derivation of the LU L if and only if [= iff] it satisfies
simultaneously the following three conditions:
1. Semantic inclusion of the keyword L
(L´) ⊃ (L) [the meaning of L´ includes the meaning of L such that L´ is lexicographically
defined in terms of L].
2. The regularity of the semantic relation between L´ and L
The semantic difference (δ) between (L´) and (L) [(δ) = (L´) – (L)] either (i) appears
between members of many lexical pairs in L [that is, (L´i) – (Li) = (δ), where i is
sufficiently large], or (ii) is negligible (≈ neutralizable).
3. The formal expression of the semantic difference (δ)
(δ) is formally expressed in L at least in some cases.
If L´ is a semantic derivation of L, L is called the base of L´.
Comments on Definition 18.1
1. Condition 1 covers also an important particular case: (L´) = (L). This is the case of full syno-
nyms (PUMA ~ CUGUAR) and of syntactic derivatives—deverbal nouns (INCLUSION from IN-
CLUDE, ATTACKN from ATTACKV), denominal adjectives (NATIONAL from NATION, MATERNAL
from MOTHER) , deadjectival adverbs (FULLY from FULL), etc. In these cases, (δ) = Λ.
2. Condition 2 is crucial: for L´ to be a semantic derivation of L, the semantic difference (δ)
between L´ and L must be systematic, or regular, in L. If it is not, it must be neutralizable:
(i) If the difference (δ) between (L´) and (L) is neutralizable, it is, in a sense, as if it did not exist
—that is, as if it were empty. And empty (δ)’s are regular in any language—there are, after all,
some full synonyms in any language, and their semantic difference is empty.
Now, the neutralization of a semantic difference is understood as the
possibility of the substitution of a sign A by a non-synonymous sign B in a context C1 ... C2 such
that (C1AC2) = (C1BC2). Thus, VIEW [= A] and OPINION [= B] are not synonymous; however, there
exist contexts in which one can be substituted for the other without changing the meaning of the
whole, for instance, the adjective POLITICAL [= C]: (his political views) ≈ (his political opinions);
therefore, VIEW can be considered to be a semantic derivation of OPINION. (Cf. 3.2, Item 1, p.
00.)
(ii) If the difference (δ) between (L´) and (L) is not neutralizable it must be regular enough in L.
Thus, the semantic difference between KNIFE and CUTV is the same as between NEEDLE and SEW,

—Chapter 18. Lexical Functions— 8
PEN and WRITE, SPADE and DIGV, etc.—namely, the former is a (tool designed for) the latter;
therefore, KNIFE is considered to be a semantic derivation of CUTV, etc.
3. The relation (X is a semantic derivation of Y) is symmetrical in some cases (as in quasi-sy-
nonymy of the type CAR ~ AUTOMOBILE or in antonymy of the type ALLOW ~ FORBID) and anti-
symmetrical in others (as in pairs of the type CAR ~ DRIVER or KNIFE ~ CUTV); we cannot, how-
ever, develop this topic here in more detail. (But in the list of Lexical Functions in Section 3 the
logical properties of lexical relations underlying a given LF—including its symmetry—will be
explicitly indicated.)
4. Semantic derivation is a more general notion than what is commonly known as morpholog-
ical derivation: morphological derivation presupposes semantic derivation, but not the other way
around. Morphological derivation requires both regular semantic and formal links between the
base and the derived LU, and these links must be parallel; for semantic derivation, just semantic
links are sufficient. Thus, run ~ runner represents the morphological derivation RUNNER which is
also a semantic derivation from the base RUNV; in steal ~ thief we see the semantic derivation
THIEF, which does not of course show morphological derivation from STEAL.
Examples of semantic derivations
The particular case: (L´) = (L) (i.e., (δ) = Λ)
Action noun for a verb
Fr. prendre (take) ~ prise; acheter (buy) ~ achat; laver (wash) ~ lavage; s’empresser (hurry) ~
empressement; tomber (fall) ~ chute; arrêter (arrest) ~ arrestation; ... [most of these are morpho-
logical derivations, but so irregular that they have to be listed in the dictionary]
The common case: (L´) ⊃ (L) (and (L´) ≠ (L))
Diminutive ((δ) = ((nice) little [L]))
Rus. grob (coffin) ~ grób+ik; dub (oak) ~ dub+ók; tramváj (tram) ~ tramváj+čik; ózero (lake) ~
ozer+k(-ó)/ozer+c(-ó); mál´čik (boy) ~ mal´č+ónok; ... [all of these derivations, highly produc-
tive, are also morphological derivations; some of them are quite regular and need not be listed,
while some others must be because of irregularities]
Female ((δ) = (female [of L]))
Tiger ~ tigr+ess; elephant ~ elephant cow; horse ~ mare; emperor ~ empr+ess; king ~ queen;
duke ~ duch+ess; cousin ~ cousin; ... [some of these are, and some are not, morphological deri-
vations]

Means noun ((δ) = (means [for L]))
—Chapter 18. Lexical Functions— 9
Sew ~ thread; wash ~ water, soap, shampoo, detergent; fry ~ oil, butter; shoot ~ ammunition; ...
[none of these is a morphological derivation]
1.3 Collocations
Syntagmatic lexical correlates {L´i | 1 ≤ i ≤ n}SYNTAGM of L, combined with L, form collo-
cations, that is, phraseological expressions of the type of those underscored in (1) [L is in
small capitals]:
(1) a. The President imposed a CURFEW on three areas ... in order to stamp out 〈= put
down〉 the VIOLENCE.
b. The panel issued a REPORT to the Secretary of State.
c. The President rejected PLEAS to open TALKS with striking US air traffic controllers.
d. The heaviest prison TERMS in Kansas history have been handed down against two men.
e. Government troops have spread a DRAGNET across the country in a SEARCH for three
heavily ARMED guerillas. The FARC has claimed RESPONSIBILITY for the ATTACK
launched Tuesday in which four ROCKETS were fired at an ARMY camp.
Syntagmatic lexical correlates can be considered as a sort of ‘addition’ to L or ‘charac-
terization’ of L. If L designates a situation or an object, an L´ designates a property or a state of L
or else an action that the referent of L can accomplish or that can be carried out on the referent of
L. Thus:
• L = SCHOOLN
{L´ j }SYNTAGM = {HIGH (school), ELEMENTARY/GRAMMAR (school), TEACH (elementary/high
• L = ESCAPEN
school), GO (to school), GRADUATE (from a school), ...}
{L´ j } SYNTAGM = {DARING (escape), GREAT (escape), MAKE (an escape), THWART (an escape), ...}
These lexical correlates of L are, as a rule, used in text together with L.
Texts—from colloquial to artistic to journalistic to technological—swarm with expressions
of the type illustrated in (1): phrases containing L and lexical correlates of L that show colloca-
tional relationships with L. Collocations are described in Chapter 19; here I will limit myself to
simply reproducing, for the reader’s convenience, the definition of the notion.

Definition 18.2 [= 19.11]: Collocation of the lexical unit L
—Chapter 18. Lexical Functions— 10
Let there be, in the lexicon of language L, two LUs—L and L´; L is taken as the starting
point, that is, as the keyword.
A phrase L—L´ = 〈(S) ; /L/ ⊕ /L´/ ; Σ L⊕L´ 〉 which is a semantic phraseme of L is called a
collocation (of L) iff it satisfies simultaneously the following three conditions:
1. The meaning of the keyword L is the semantic pivot of the expression
The signified (S) of the expression L—L´ includes the signified of L as its semantic
pivot [see Def 19.11]: (S) = (C)((L)).
2. The keyword L is selected independently
In the process of text synthesis the LU L is selected by the Speaker independently of
L´—for its own signified (L).
3. The keyword L controls the selection of the other constituent L´
L´ is not selected unrestrictedly: it is selected for the signified (C) as a function of L.
The LU L is the base of the collocation L—L´, and L´ is its collocate.
NB: Note the two senses of the term base: the base of a semantic derivation vs. the base of a collocation.
Although the two notions are not identical, they are intimately related, and this fact will be reflected in
using the same term for both when representing the semantic derivations and the collocations via Lexical
Functions: both will be called keywords.
For all comments, explanations and examples with respect to collocations, the reader is
referred to Chapter 19, pp. 00ff.
1.4 Overlap of Semantic Derivations and Collocations
An important property of semantic derivations and collocations is that the two sets have a very
significant overlap: on the one hand, there are linguistic expressions that look like semantic
derivations but in fact manifest ‘fused’ collocations, while, on the other hand, some expressions
that are collocations actually hide underneath ‘split’ derivations. Thus:
• The verb BLASTV ([to] attack vigorously), as in The council blasted the university administra-
tion for this botched attempt, could be considered a semantic derivative of the verb ATTACKV
(verbally)—the relation between BLASTV and ATTACKV corresponds to Def. 18.1. However, the
semantic difference (d) in this case is (intensely), and this meaning—intensification—is currently
expressed in English by separate LUs: [to] ATTACK vigorously, [to] APOLOGIZE profusely, ravenous
APPETITE, close SHAVE, THIN as a rake, ANXIOUS as a cat on a hot tin roof, etc. (all boldfaced LUs
are in fact expressions of the Lexical Function Magn, see…). Therefore, it is preferable to consider
BLASTV as a ‘fused’ collocation of the type ATTACKV VIGOROUSLY and treat it in the dictionary as a

—Chapter 18. Lexical Functions— 11
manifestation of a collocation (see below, on fused elements of values of LFs, p. 00ff). BLASTV,
considered with respect to ATTACKV, adds a characterization of the meaning of (attack): to attack
how?—Vigorously; from the viewpoint of its semantic role it is more similar to collocations.
• On the contrary, in the collocations ACCOUNT holder, PRISON guard or party to the AGRREMENT
the boldfaced collocates do not in any sense characterize the base. Here, the collocation as a
whole serves as the name of an element of the situation referred to by the base: namely, of the 1st
actant of the situation ((bank) account), (prison) and (agreement). In this, the collocations at hand
are similar to thousands of English LUs that name the 1 st actant of a situation: ADMIRER for
(admire), PORTER/REDCAP for (carry luggage), DOCTOR for (hospital), DEBTOR for (debt), CLERK
for (sell [in a store]), PAINTER for (paintV), etc., ad nauseam. As a result, the collocations of this
type are better presented and processed as ‘split’ semantic derivations.
In this way, the border between monolexemic semantic derivations, which are paradigmatic
lexical correlates of the headword L, and polylexemic collocations, which are L’s syntagmatic
correlates, seem to become fuzzy. In a sense, this is true, but only from the formal point of view.
The semantic contrast between the two types of lexical correlates remains clear-cut:
Paradigmatic lexical correlates of L are involved in selecting LUs in the process of
naming something; they are needed to answer questions of the type ‘What do you call an
entity or a situation L´, related to L?’—while speaking of L´ rather than of L.
Syntagmatic lexical correlates of L are involved in combining selected LUs in the
process of predicating something of L; they are needed to answer questions of the type
‘What do you call the action 〈characteristics, attribute, etc.〉 L´ of L?’—while speaking of L
rather than L´.
Each type of lexical correlate fulfills its role independently of the formal way it is expressed—by
a monolexemic LU or by a plurilexemic LU. Therefore, such verbs as BOMBV, OILV, WATERV,
SALTV, etc., which mean (apply bombs/oil/water/salt/… to Y), are considered syntagmatic corre-
lates of BOMBN, OILN, WATERN and SALTV—in spite of their being monolexemic; the Speaker
needs them to predicate something of L—to name correctly a typical action done with L, i.e.,
(applying L in a proper way). At the same time, such collocations as WORM OF DOUBT, PANGS OF
REMORSE, FLAME OF PASSION, CLOUD OF SUSPICION, etc., which are used to refer to (doubt),
(remorse), (passion) and (suspicion) ((worm of doubt ≈ (doubt), (pangs of remorse) ≈ (remorse, etc.)
are taken to be paradigmatic correlates of DOUBT, REMORSE, PASSION, and SUSPICION—in spite
of their being polylexemic; the Speaker uses them simply to name the corresponding psycholo-

—Chapter 18. Lexical Functions— 12
gical states in a fancy way (by journalistic clichés). In Subsection 2.4 I will briefly return to the
problem of the overlap between the expressions of paradigmatic vs. syntagmatic lexical correlates
in connection with formal encoding two different types of the elements of values of LFs (see p.
00).
2 The Notion of Lexical Function
2.1 Introductory Remarks
I will begin with the general notion of LF and then proceed to a particular case—the simple
standard LFs, which are of special interest in the present framework. The term function is
used, as pointed out above, in the mathematical sense: f(x) = y.
Formally, an LF f is a function that associates with a given LU L a set {Li} of lexical
expressions that express, contingent on L, the meaning (f) associated with f and bearing on the
meaning (L):
f(L) = {Li}, such that an Li expresses (f)((L)).
L is called the argument, 3 or keyword, of f; f(L) = {Li} is the value of f applied to L; and
an Li is an element of this value.
Informally, an LF f is, roughly speaking, a meaning (f), coupled with a deep-syntactic role,
such that its expression is not free (in contrast to all ‘normal’ meanings), but depends on the LU
to which this meaning applies.
The core idea of Lexical Functions is lexically bound expression of some meanings.
To put it differently, a prototypical LF, namely a simple standard LF, corresponds to a very
general, ‘poor’ meaning that can be expressed in a large variety of ways, the choice of the
expression being determined in each particular case by the LU L with which this meaning is
being expressed.
Using a convenient term taken from L. Wanner (1996: 1ff), it can be said that LFs are
designed to represent, in a formal and systematic way, institutionalized lexical
relations—regular relations that hold between LUs L1 and L2 of language L and are based on
L1’s and L2’s signifieds, being at the same time constrained by L1’s and L2’s formal and
combinatorial properties, Thus, as we see, these relations are exactly lexical rather than purely
semantic; they are institutionalized in that they hold between the members of numerous pairs of
LUs in L.

Here are some examples of LFs.
—Chapter 18. Lexical Functions— 13
(f) = (the one who/which undergoes L) [patient noun; a semantic derivation]
S 2(shootV) = target S 2(serveV) = client
S 2(hotel) = guest S 2(prison) = prisoner
S 2(doctor) = patient S 2(hair-dresser) = customer
(f) = (intense(ly) L), (very L) [intensifier; a collocate]
Magn(shaveN) = clean, close, smooth Magn(naked) = stark
Magn(easy) = as pie, as 1-2-3 Magn(skinny) = as a rake
Magn(condemnV) = strongly Magn(relyV) = heavily
(f) ≈ (do L), (perform L) [support, or light, verb; a collocate]
Oper 1(cryN) = let out [ART ~] Oper 1 (figureN) = cut [ART ~]
[He cut a fine/miserable figure]
Oper 1 (strikeN) = be [on ~] Oper 1 (visitN) = pay [ART ~]
Oper 1 (tripN) = take [ART ~] Oper 1 (orderN) = give [ART ~]
(f) = (realize L), (fulfill [a requirement of] L) [realization verb; a collocate]
Real 2(testN) = withstand [ART ~]
Real 2(jokeN) = get [ART ~]
Real 2 (examN) = pass [ART ~]
Real 2(mineN) = hit, strike [ART ~]
[Their car struck a land mine]
Real 2(ambushN) = fall, step, run, walk
[into (ART) ~]
Real 2(challengeN) = meet [ART ~]
[The expression in square brackets represents the Government Pattern [= GP] of the keyword of the LF; the
symbol ART indicates that an article or a grammatically equivalent determiner should be used according to
grammatical rules; the tilde ~ stands for the keyword. For more about GPs of LFs, see below, 4.4, p. 00]
The following examples show how the meaning (very) can be expressed in different langua-
ges in order to modify the meaning (thin/skinny); they demonstrate the linguistic variability and
wealth of LF expressions:
Arabic [Standard]
nāħif nāħāfa(t) lit. (skinny [with-]skinniness)
Z#adīd annāħāfa(t) lit. (strong [with-]the-skinniness)
ka/anna+hu Z#ild ?alā ?az`m lit. (as-if-he-were skin on bones)

—Chapter 18. Lexical Functions— 14
[Lebanese]
dà?īf miθla al?ūda(t) (skinny as the-twig)
[Egyptian]
/āmil zaZ3#
al/asāZ3#a(t)
lit. ([he] makes as the-stick)
Catalan prim com un fideu; semblar (skinny as a noodle; seem a noodle, be as a noodle)
un fideu, estar com un fideu
Chinese gu‹shòurúchái ([be] skinny as [a] piece-of-firewood),
lit. (skeleton-skinny-like-piece.of.firewood) [shòu = (skinny);
gu‹ and chái are bound radicals that appear just in such expres-
sions; the whole is a compound qualifying verb, which corres
ponds to adjectives in European languages]
Dutch mager als een lat (skinny as a wood-strip)
English skinny/thin as a bone 〈as rake,
as a stick〉, stick-thin, rail-thin
French maigre comme un clou/ (skinny as a nail/as a stake)
comme un échalas
Georgian gamxdari rogorc čxiri (skinny as [a] small stick ≈ as [a] switch)
German dünn wie ein Spargel, (skinny as an asparagus, as a line)
wie ein Strich
Greek adínatos san odóntoglyfida (skinny as [a] toothpick)
Hebrew raze kmo xut (skinny as [a] thread)
Italian magro come un’acciuga (skinny as an anchovy)
Japanese kirigirisu-no yooni yasete (skinny as [a] grasshopper)
lit. (grasshopper like skinny)
Korean jotkarak chorom malassta ([be] skinny as [a] chopstick)
lit. ([he/she] chopstick like skinny.became)
Malagasy mahia hoatra ny tahon-katsaka ([be] skinny as [a] corn seedling 〈a broomstick, a ghost,
〈ny tahon-kifafa, ny angatra, a phantom〉
ny kinaoly〉
Polish chudy jak patyk (skinny as [a] twig)
Russian xudoj kak ščepka, kak skelet (skinny as [a] sliver/as [a] skeleton)
Serbian mršav kao štap, kao motka, (skinny as [a] walking-stick/as [a] stick/
kao glista as [a] worm)
Spanish flaco como un fideo (skinny as a noodle);
hecho un fideo 〈un palillo〉 (made a noodle 〈a little.stick〉);
en los huesos (in the bones) = (very skinny)
Swedish mager som en skrika (skinny as a jay)
Turkish iskelet gibi olmak (be as [a] skeleton) = lit. (skeleton like be)

—Chapter 18. Lexical Functions— 15
A linguistic expression that constitutes an element of the value {L´i} of an LF f when
applied to L (f(L) = {L´ i }) can be, generally speaking, of one of the five types, that is, L´ can be:
— a lexical unit
1) a lexeme : Magn(smoker) = heavy [smoker]
2) a phraseme (including
collocations) : Magn(smokeV) = [smoke] klike a chimneyl
Magn(rainV) = [rain] kcats and dogsl 4
— a morphological element
3) a compounding stem : Magn(smoker) = chain-[smoker]
4) a derivational affix : S 1(smokeV) = [smok]+er
— a free phrase
5) : Magn(smokeV) = [smoke] more than a battalion of firefighters,
without stopping, lighting up a cigarette from
the other, ...
(Cf. Her marriage is as DEAD as the remains of some Wooly Mammoth unearthed by researchers
in Siberia, or some such place; the underscored free phrase intensifies the adjective DEAD.)
Now the two relevant definitions can be formulated: Lexical Functions and Standard
Lexical Functions.
2.2. Definition of Lexical Function
As was already stated, a Lexical Functions f is a correspondence between lexical expres-
sions (of language L) that is associated with a meaning (f) and a (deep-)syntactic role. According
to this characterization, f is applicable or not applicable to LUs of particular semantic/
syntactic classes. The notion of applicability of an LF f to an LU L is based on semantic compati-
bility of f and L—that is, on the compatibility of the meanings (f) and (L). Formally, the applicabi-
lity must be explicitly specified for each f by indication of the semantic class of the correspond-
ing LUs—i.e., LUs that can be arguments of f. If f is applicable to L, we will say that f is
defined for L.

Definition 18.3: Lexical Function
—Chapter 18. Lexical Functions— 16
A correspondence f that associates a set f(L) of lexical expressions with an LU L is called a
Lexical Function [= LF] iff it satisfies either conditions A1-A3 or condition B.
A. f is applicable to several LUs and:
1. Semantic homogeneity of f(L)
For any two different LUs L1 and L2, if f(L1) and f(L2) both exist, then any L´1 ∈ f(L1) and
L´2 ∈ f(L2) bear an (almost) identical relationship to L1 and L2, respectively, as far as
their meaning and the DSynt-role are concerned:
2. Maximality of f(L)
L´1 L´2
—— ≈ ——
L1 L2
For any two different LUs L´1 and L´2, if L´1 ∈ f(L1) and L´2 ∉ f(L2), then L´2 does not
stand to L2 in the same relationship as L´1 to L1:
3. Phraseological character of f(L)
a) At least in some cases f(L1) ≠ f(L2); and
L´1 L´2
—— ≠ ——
L1 L2
b) at least for some f(Li) some elements of f(Li) cannot be specified without mentioning
an individual LU Li.
B. f is applicable to only one LU L (or perhaps to a few semantically close LUs).
In f(L), the LU L, which is the argument of f, is called the keyword 5 of f, and f(L) = {L´i}
is f’s value.
An LF that is applicable to several LUs—satisfying Conditions A1-A3—is called normal;
an LF applicable to only one LU (or two or three semantically close LUs)—satisfying Condition
B—is degenerate. (Degenerate LFs are an extreme case of non-standard LFs, see 4.5.2, p.
00.)
Thus, a normal LF f specifies for its keyword L a set of lexical expressions
f(L) = {L´i | 1 ≤ i ≤ n} SYNTAGM
such that:
• all its elements stand to L in (almost) the same semantic/syntactic relation R;
• it includes all the expressions that bear the relation R to L;

—Chapter 18. Lexical Functions— 17
• at least some of its elements cannot be specified without a direct reference to L as an indivi-
dual LU.
A degenerate LF specifies for L a lexical expression L´ that appears, generally speaking,
with no other LU, because L´’s meaning includes the meaning of L ((L´) ⊃ (L)).
Comments on Definition 18.3
1. The value of an LF f for the LU L—f(L)—is, generally speaking, a SET of linguistic expres-
sions {L´i}. An element of f(L) can be, as we have seen, an LU (a lexeme or a phraseme), a
morphological sign (a compounding stem or an affix), or else a free phrase. The elements of the
value f(L) need not be strictly synonymous: they must be only more or less synonymous—shar-
ing an important semantic component (f), which is associated with f (for more details, see 5.1.1, p.
00).
2. Condition A1 guarantees the homogeneity of f(L): for any two keywords, the semantic
and/or syntactic relationships between any element of f’s value and the keyword remains approxi-
mately the same. The researcher is free to determine himself how approximate this may be—
according to the task he sets himself. The identity of the semantic relationship between L and an
element of the value f(L) is relative: it depends on the degree of granularity chosen by the resear-
cher as well as on the level of analysis. Consider two collocations: The PREDICTION comes true [=
‘begins to be reality’] and The FILM opens [= ‘begins to be screened’]; both underscored express-
ions are elements of the value of the LF IncepFact 0 for the respective keywords (in small caps).
Of course the semantic relationships between, on the one hand, come true and prediction and, on
the other hand, open and film are not literally the same. They can, however, be considered the
same on a more abstract level: in both cases, the entity denoted by the keyword ‘does what it is
expected to do’: thus, a prediction is expected to become reality, and a film, to be screened. The
lexicographic definitions of the LUs PREDICTION and FILM must include the corresponding
components.
3. Condition A2 establishes the saturated character of f(L): f is maximal in that it returns ALL
possible elements of any given value for any LU. Thus, any intensifier of L must belong to the
value f(L) of f that covers the intensifiers—this f is the LF Magn (see below). An L´ that does not
belong to Magn(L) cannot be an intensifier of L.
4. Condition A3 ensures the phraseological character of f(L). Conditions A1 and A2 characte-
rize f as a potential LF: they require only the existence of the proportion L´1 : L1 = L´2 : L2
and the inclusion of all appropriate elements in the value. Condition A3, on the other hand,

—Chapter 18. Lexical Functions— 18
characterizes f as an actual LF: it requires that in L, for (at least) some pairs of LUs 〈L1,
L2〉 the value of f be phraseologically bound by its keyword. To put it differently, if f
has several keywords, then the selection of the elements of f(L1) and f(L2) is contingent, at least
in some cases, on particular L1 and L2, respectively. If the distribution of the elements of f(L1)
and f(L2) can be stated in terms of a general rule, Condition A3 is not satisfied and f is not an LF.
For instance, the idea of (good) is expressed as GOOD with nouns and as WELL with verbs; but this
fact alone does not allow us to associate the meaning (good) with an LF. (There are other factors
that argue for such an LF—see Bon, No. 33 in 4.4.2, p. 00; namely, there are many phraseologic-
ally bound elements of its possible values.) If f has just one keyword L, the selection of f(L) is
trivially contingent on L.
NB: Generally speaking, an LF f returns for an LU L not only phraseologically bound elements of the value f(L),
but all the free expressions with the corresponding meaning. Therefore, the phrase L—f(L) can be a genuine
collocation (if the element of the value is phraseologically bound by L) or a free phrase that is considered to
be collocation by analogy (otherwise).
5. For each LF f, as indicated above, its domain of applicability is specified—a set of LUs for
which f is defined. An LF may be applicable to, or defined for, any LU (as is the case with Syn),
to very many LUs (such as the LF Magn, applicable to any LU whose meaning includes a grad-
able component), to few LUs (such as the LF Pos, which applies only to LUs referring to evalua-
tions), or even to one LU (such as the non-standard LF F 1 = which has 366 days, applicable only to
YEAR: F 1(year) = leap [year]).
If the LF f is applied to an LU which does not belong to its domain, the result is absurdity:
*S 2(sleep) or *Oper i(tree). But even if applied inside its domain, f may return an empty value:
Real II 3(invitation) = Λ (to express the meaning (Y does what Y is invited to do) English does not
say something like *follow/execute an invitation, as, e.g., Russian does: posledovat´ priglašeniju;
English has, however, Labreal II
3(invitation) = take [NX] up [on Aposs-X ~]).
A particular type of LFs is of special interest—Standard LFs, which form a proper
subset of normal LFs (several examples of standard LFs have been already shown above).

2.3 Definition of Standard Lexical Function
Definition 18.4: Standard Lexical Function
—Chapter 18. Lexical Functions— 19
A normal LF f is called standard iff it satisfies the following two conditions:
1. Broadness of f’s domain
The meaning (f) associated with f is sufficiently general (non-specific) to be applicable to
many other meanings ((f) can even be empty), so that f is defined for a relatively large
number of keywords.
2. Broadness of f’s range
The expressions of the meaning (f) are sufficiently variegated so that f has a relatively
large number of elements in its possible values.
Condition 1 characterizes an LF f as a potential standard LF, and Condition 2—as an
actual standard LF; it means that the set of all f(Li), for a vast variety of Lis, is relatively rich.
Both 1) normal LFs that do not satisfy at least one of Conditions 1 and 2 of Definition 20.4
and 2) degenerate LFs are called non-standard.
The expression relatively large used twice in Definition 18.4 is intentionally vague: the
distinction between standard and non-standard LFs is in basically quantitative (it concerns the
number of possible keywords and possible value elements). In this sense, the borderline between
standard and non-standard LFs is to be established by the researcher according to his practical
goals, his ideas about the compactness and elegance of the resulting description, etc.; cf. the dis-
cussion of LF standardness in Polguère 200?. (In the MTT approach, the number of possible key-
words of a typical standard LF must be in the hundreds, and the number of possible different
elements of the values, at least a few dozen.) The difference between standard and non-standard
LFs will be discussed in 2.5, Item 2, p. 00; here, I will limit myself to a few examples.
S 2, Magn, Oper 1 and Real 2, cited above, are standard LFs:
1) The meanings associated with them are very poor (= general), so that each of these LFs is
defined for a very large domain. Thus, S 2 is defined for all LUs which have (at least) two
Semantic Actants [= SemAs]; Oper 1, for all nouns whose meaning is a predicate in the logical
sense; and Real 2, for all LUs whose meaning includes (with the goal of ...) or (designed for ...).
2) Each of these LFs has—in English, but in many other languages as well—a very large
range, that is, the number of expressions for them is high (in the thousands).

LFs.
—Chapter 18. Lexical Functions— 20
To better delineate the concept of standard LF, I will present in more detail non-standard
2.4 Examples of Non-Standard Lexical Functions
Here are four explained examples of non-standard LFs.
Example 1
The meaning (without addition of a dairy product) has in many languages two expressions:
a special, phraseologically bound expression used with the noun meaning (coffee), as in black
COFFEE, Fr. CAFÉ noir, Rus. čërnyj KOFE lit. (black coffee), Sp. CAFÉ sólo lit. (coffee alone), and a
free phrase without milk (and its equivalents in other languages) used in other contexts; thus, tea
without milk is not called *black tea, *thé noir, *čërnyj čaj: it is simply tea without milk.
Remarks
1. (Coffee with milk) has a free expression in English and Russian—with milk and s molo-
kom, but not in Serbian and French: Serb. bela kafa lit. (white coffee), Fr. café au lait ou café
crème, see below.
2. Note a phraseological expression of both meanings in (some dialects of) English, but in a
different syntactic position: I prefer my tea black (without milk) and I prefer my coffee light (with
milk).
The meaning (without addition of a product modifying the taste) also has two expressions in
French: THÉ nature lit. (nature tea) = (regular tea), and WHISKY sec lit. (whisky dry) (cf. also have
one’s WHISKY straight (up)).
The expressions of these two meanings are lexically distributed: BLACK with COFFEE, but
WITHOUT MILK with anything else (similarly in other languages); Fr. NATURE with THÉ, but SEC
with WHISKY. Therefore, the meanings (without addition of a dairy product) and (without addition
of a product modifying the taste) satisfy Def. 18.3: they are associated with an LF. However, they
fail Condition 1 of Def. 18.4: these meanings are too specific and applicable only to a few names
of beverages and dishes. (They fail Condition 2 of Def. 18.4 as well of course.) The meanings in
question correspond to non-standard LFs in English, French, and Russian.
Example 2
In French, the meaning (with the addition of a little milk) has at least two different expres-
sions, which are also lexically distributed: CAFÉ crème 〈*café à la crème, *café avec de la crème,
*café lait〉 (coffee with milk) vs. THÉ au lait 〈*thé crème〉. 6 Therefore, the meaning in question

—Chapter 18. Lexical Functions— 21
determines a lexical dependency which satisfies Conditions A1-A3 of Def. 18.3: it is a Lexical
Function. However, it fails to satisfy Conditions 1 and 2 of Def. 18.4: it also corresponds to a
non-standard LF.
Example 3
In Russian, the meaning (of brown color) has five different expressions depending on what
it characterizes: for anything but human eyes, human hair and horses the adjective KORIČNEVYJ
is used; but (brown eyes) are KARIE glaza 〈*koričnevye glaza〉, (brown hair) is TËMNORUSYE or
KAŠTANOVYE volosy 〈*koričnevye volosy〉, and (a brown horse) is GNEDOJ kon´ 〈*koričnevyj
kon´〉. Therefore, the meaning (of brown color) determines in Russian a lexical dependency which
satisfies Conditions A1-A3 of Def. 18.3: it is an LF. However, although unlike the meanings
(without addition of a dairy product), (without addition of a product modifying the taste) and (with
the addition of cream), the meaning (of brown color) satisfies Condition 1 of Def. 18.4 (a huge
variety of things can be of brown color!), it fails Condition 2: it has only five different expres-
sions and four of these (KARIJ, TËMNORUSYJ or KAŠTANOVYJ, and GNEDOJ) are used with very
few specific arguments each (eyes, hair, horses). This meaning equally is a non-standard LF of
Russian.
Example 4
The meaning (which has 366 days) cannot be applied to anything but YEAR: it constitutes a
degenerate—and therefore non-standard—LF. Its expression appears only with YEAR and is fair-
ly idiosyncratic: leap YEAR; cf. similar expressions Ger. SchaltJAHR lit. (switch year), Rus. visokos-
nyj GOD [the adjective VISOKOSNYJ cannot be used outside of this phrase], Serb. prestupna
GODINA [the adjective PRTESTUPNI cannot be used outside of the given phrase, either]. In French,
this meaning is applicable to two nouns: AN (year) and ANNÉE ((duration of a) year); but it pro-
duces a correct collocation only for ANNÉE: année bissextile, *an bissextil.
Degenerate non-standard LFs are characterized by unique cooccurrence of the collocate: it
combines with just one base (or with very few near-synonymous bases). Thus, bleary/beady EYES,
aquiline NOSE, husky VOICE, artesian WELL, bubonic PLAGUE, LOOK for a needle in a haystack are all
examples of degenerate non-standard LFs of English (the keywords—collocation bases—are
capitalized).
@@Easternn Europe vs. East Africa, Afrique orientale
Although theoretically the distinction between normal non-standard and degenerate non-
standard LFs is quite clear, in practice, i.e., with respect to specific collocations, it is not always

—Chapter 18. Lexical Functions— 22
obvious where the line has to be drawn. The solution depends on the semantic description of the
value expression: if its meaning includes the meaning of the keyword, we have a degenerate LF;
otherwise, it is a normal LF. Thus, I think that LOOK daggers [at N], blind DATE, ruddy CHEEKS,
purple PROSE, puppy LOVE and white LIE are normal non-standard LFs: the meaning of the collo-
cate [daggers = (angrily or hatefully), blind = (with an unknown person), ruddy = (of reddish
color), purple = (overornate and too elaborate), puppy = (occurring for the first time in very young
people—quite innocent), white = (told so as not to hurt someone)] does not include an explicit
mention of the keyword and therefore semantically it is applicable to other LUs.
Because of their rather specific meanings, non-standard LFs are more closely related to par-
ticular languages (than standard LFs); that is, they are not universal. Because of that:
• non-standard LFs cannot be logically predicted and organized into an overall system;
• non-standard LFs are very numerous, vary from language to language and extremely caprici-
ous—they have to be discovered and collected empirically.
The good news about non-standard LFs is, though, that they are very limited as to their
domain; they concern, as a rule, highly specialized situations. This greatly facilitates their ana-
lysis and description.
This chapter deals with the standard LFs only.
2.5 Classification of Lexical Functions
For a better surveyability of LFs, they can be classified according to different parameters;
these classifications facilitate the task of the user and thus possess a pedagogical value. Five
classification parameters of LFs will be considered here:
• Paradigmatic vs. syntagmatic LFs.
• Standard vs. non-standard LFs.
• Simple vs. complex vs. mixed LFs.
• Nominal vs. verbal vs. adjectival vs. adverbial LFs.
• Semantic groupings of LFs.
1. Paradigmatic vs. syntagmatic LFs. The opposition ‘paradigmatic vs. syntagmatic LFs’
corresponds to to the opposition between paradigmatic and syntagmatic lexical correlates that has
already been introduced above, so that here it is sufficient to concentrate on a formal complica-
tion:

—Chapter 18. Lexical Functions— 23
An element of the value of any LF f can in principle come in two syntactic ‘varieties:’
1) With some LU L1 the element L´1 ∈ f(L1) appears instead of L1; such an element is
called a fused element of the value of f.
2) With some other LU L2 the element L´2 ∈ f(L2) appears together with L2; this is a
non-fused element of the value of f.
This particularity of the elements of the value of an LF is indicated in the lexicographic
description of each L´i (that is, in the lexical entry of the keyword L); see 5.3, p. 00.
As a result, we have to consider two logically independent oppositions:
• paradigmatic vs. syntagmatic LFs;
• non-fused vs. fused elements of the values of LFs.
The fused elements are specified by the symbol “//” placed before them. Thus, in the
subentry of the LF f in the dictionary, the non-fused elements of f are found to the left of //, and
the fused elements to its right. The meaning of a fused element L´i of the value f(L) includes the
meaning of f—(f); the meaning of a non-fused element L´i of f(L) may include or not include the
meaning of L.
The paradigmatic LFs have, as a rule, fused elements in their values, while the syntagmatic
LFs tend to have non-fused elements of the values. Nevertheless, this correspondence is logically
not necessary and can be violated: namely, both syntactic ‘varieties’ of value elements, fused and
non-fused, are encountered for both types of LFs. For instance, the syntagmatic LFs Magn and
Labreal ij both have non-fused and fused elements in their values:
Magn(applause) = frenzied, thunderous [applause] //(standing) ovation
or
[(thunderous applause) ≈ (ovation)]
Labreal 12(asphaltN) = cover [NY with asphalt] //asphaltV [NY]
[(They covered the road with asphalt) = (They asphalted the road)].
Such ‘deviations’ do not undermine the inherent paradigmatic vs. syntagmatic nature of a given
LF. If an LF f is supposed to help the user to select an LU in order to refer to a situation (L) or to
a component thereof, that is, to name something, f is paradigmatic, no matter how its value
elements behave (that is, independently of whether they are rather fused or non-fused). If f serves
to select an LU in order to characterize L, to add something to it, it is syntagmatic, again inde-
pendently of the fused or non-fused character of the elements in its value.

—Chapter 18. Lexical Functions— 24
2. Standard vs. non-standard LFs. These are different in two major respects:
i) Quantitatively: a standard LF has many possible keywords and many possible value
elements, while a non-standard LF has very few keywords and very few value elements.
ii) Qualitatively: in principle, a standard LF (with a few exceptions) participates in the general
rules for synonymic DSynt-paraphrasing, while a non-standard LF never does
For more on the difference between standard and non-standard LFs, see Polguère 200?.
3. Simple vs. complex LFs. Standard LFs are divided into simple and complex. A simple
standard LF f can be of one of two types:
• Either f is elementary, i.e., within the existing system of LFs it cannot be represented in terms
of other LFs. Such are, for instance, Anti, Conv, Magn, Oper 1, Incep, Caus, Son.
• Or f is not elementary, that is, it can be represented in terms of other standard LFs, but it is
introduced anyway as a convenient abbreviation—because f is often used in natural language as
one simple unit. Such are, for instance:
—Oper 2, which can be represented as Conv 321Oper 1;
—Degrad, representable as IncepPredPejor;
—Liqu, which corresponds to CausFin.
Several syntactically linked simple standard LFs may form a combination—product of
several LFs—called a complex LF, such as AntiMagn, IncepOper 1, CausFunc 0, CausPred-
Plus, etc. (see 4.5.1, p. 00).
The borderline between simple and complex LFs is drawn based on the convenience of the
description and manipulation: simple standard LFs tend to have more frequently independent lin-
guistic expressions and are associated with relatively simple meanings.
As we will see later, there are also mixed LFs (combinations of syntactically linked
standard and non-standard LFs) and configurations of LFs (= combinations of syntactically
non-linked LFs that share the keyword).
4. LFs according to their part of speech. At the DSynt-level of representation, where LFs
appear, four deep-syntactic parts of speech are distinguished: Nouns (notation: S, from
‘substantive’), Adjectives [notation: A], Adverbs [notation: Adv], the latter covering genuine
adverbs of all types, prepositions, conjunctions, and interjections, and Verbs [notation: V]. As a
result, LFs are readily classified into nominal, adjectival, adverbial (including prepositional), and
verbal ones.

—Chapter 18. Lexical Functions— 25
5. 10 semantic groupings of (simple standard) LFs can be distinguished:
Paradigmatic LFs
1) Basic LFs: Syn(onym), Anti [= antonym], and Conv(ersive) ij . They correspond to the main
three semantic relations that play a special role in MT-theory—synonymy, negation, and con-
verseness (X precedes Y ≡ Y follows X).
2) (Semantic) derivations are of two subtypes:
—Structural derivations represent nominalization S 0 (REJECTION from REJECT), adjectivali-
zation A 0 (URBAN from CITY), adverbialization Adv 0 (WELL from GOOD, FIVE TIMES from FIVE),
and verbalization V 0 (ATTACKV from ATTACKN).
—Meaningful derivations are, roughly speaking, agent noun S 1, patient noun S 2, active adjec-
tival A 1 (in search [of N] from LOOK [for N]), passive adjectival A 2 (under construction from
BUILD), place noun S loc (bedroom from SLEEP), instrument noun S instr,
(syringe from INJECT),
active potential adjective Able 1 (inquisitive from ASK: ≈ (who tends to ask)), passive potential
adjective Able 2
(reliable from RELY), etc.
3) Generics: hyperonym Gener (LIQUID substance) and metaphoric denotation Figur (flame of
PASSION).
Syntagmatic LFs
4) Quantifiers: singulative Sing (speck of DUST) and collective Mult (pride of LIONS).
5) Modifiers: cliché Epit (proud PARENTS), intensifier Magn, Plus/Minus ((more Magn), as in
growing CONCERN, and (less Magn), as in falling DISCIPLINE), objective qualifier Ver (restful SLEEP,
(healthy, nutricious MEAL), and subjective qualifier Bon (valuable CONTRIBUTION, exquisite MEAL).
6) Semi-auxiliaries (= support, or light, verbs): semantically empty, or emptied, verbs linking
a DSynt-Actant [= DSyntA] of L to L, namely Oper 1/2 (give/take an EXAM), Func 0/1/2 (A CHANGE
occurs/comes from the boss/affects our plans) and Labor 12/21 (submit the thief to an INTERROGA-
TION/leave Mary a WIDOW).
7) Phasals: verbs denoting the three phases of an event—the beginning (Incep), the end (Fin),
and the continuation (Cont). These LFs are often used combined with other verbal LFs to produce
complex LFs, e.g., IncepOper 1, ContFunc 0, etc.

—Chapter 18. Lexical Functions— 26
8) Causatives: verbs denoting the three possible types of causation, namely—the causation of
existence (Caus), the causation of non-existence (Liqu), and the non-causation of non-existence
(Perm).
The phasals stand in the antonymous relation to each other; the same is true of causatives:
Incep = AntiFin, Liqu = AntiCaus, etc. Furthermore, causatives and phasals are also related,
because you can cause the beginning, the end or the continuation of an event (see 4.4.1.2, Remark
3, p. 00).
9) Realizations: Real 1/2, Fact 0/1/2, Labreal 12/21.
10) Varia: Prox, Involv, Son, Imper, Degrad, Manif, Obstr, Sympt.
3 The List of Simple Standard Lexical Functions
3.1 Introductory Remarks
A full list of simple standard LFs known to date follows (see as well in Mel’čuk 1974: 82-
99, 1994, 1996, Mel’čuk & Zholkovsky 1988, Mel’čuk et al. 1984, 1988, 1992, 1999, Mel’čuk et
al. 1995).
The set of simple standard LFs has been established empirically as a subset of standard LFs
such that it has proven to be particularly convenient for describing paradigmatic lexical choices
and syntagmatic restricted lexical co-occurrence, as well as paraphrasing. In other words, what is
a simple LF and what is not is decided by the researcher according to the considerations of con-
venience and elegance of description. Each simple LF is identified by an individual name, which
is, as a general rule, a four-letter abbreviation of a Latin word (actual or concocted for the occa-
sion); it is treated as a unit. The number of simple standard LFs in all languages of the world is
about sixty; they constitute the core of the proposed system of LFs.
As indicated above, several simple LFs are not elementary—they are related semantically
to other LFs, and it is possible to represent them in terms of the latter. Thus, e.g., Oper 2 =
Conv 321Oper 1 (for a tri-actantial Oper 1, as in John[I] gave[Oper 1] Mary[III] a kiss[II]. ~ Mary[I]
received[Oper 2] from John[III] a kiss[II]. Semantic relations among simple LFs are used in para-
phrasing rules.

—Chapter 18. Lexical Functions— 27
A standard LF f is, by definition, a Deep Lexical Unit, 7 appearing in the representation of
sentences at the DSynt-level. To characterize its value f(L), the following points need to be
specified:
about f
1) f’s meaning and its semantic relation to the keyword L;
2) its Deep-Syntactic part of speech (S, V, A, or Adv);
3) the DSynt-relation linking an element of f(L) to L—for the syntagmatic LFs only.
about L
4) L’s semantic properties necessary for f to be applicable to L;
5) L’s DSynt-part of speech, which can also be needed to define f’s applicability to L;
6) L’s Government Pattern [= GP(L)] (for GP, see ??, p. 00).
meters.
In the list of LFs that is presented below, the LFs are described according to these para-
The order of presentation of LFs is as follows:
• Paradigmatic LFs: first basic, then non-basic LFs.
• Syntagmatic LFs: nominal, adjectival, adverbial and then verbal LFs.
Notation
If an element L´ of the value f(L) of the LF f is used syntagmatically, that is, together with
L and in such a way that L is a DSynt-Actant of L´, then L´ is accompanied by an expression in
square brackets that indicates the syntactic link between L and L´; among other things, this
expression presents the reduced Government Pattern of L´, the appropriate determi-
ners, etc. Example: the record
IncepOper 1(attackN) = launch [ART ~]
shows that the light verb LAUNCH takes its keyword ATTACKN as its DSyntA II [≈ Direct Object]
and the resulting phrase presupposes the use of an article (or an equivalent determiner) with
ATTACKN—according to general rules of English.
3.2. Paradigmatic Lexical Functions (1-26)
LFs 1-3 are basic LFs: Syn, Anti and Conv ij correspond to the three central semantic
relations of the Meaning-Text framework, namely synonymy, antonymy, and conversiveness.
These three LFs are relatively well-known in linguistics. 8 Let me simply point out their four most
important properties:

—Chapter 18. Lexical Functions— 28
• They are, in a sense, meta-LFs: they can apply not only to ‘normal’ LUs, but also to LFs (in
sharp contrast to other LFs that apply only to ‘normal’ LUs). For instance, we have such equiva-
lences as the following ones:
Degrad ≡ Syn(IncepPredPejor)
Liqu ≡ Anti(Caus)
Oper 1
≡ Conv 21(Func 1)
• They can be semantically exact or approximate: the elements of their values can have a richer
( ⊃ ), poorer ( ⊂ ), or intersecting ( ∩ ) meaning with respect to the meaning (L). In this case, they are
quasi-synonyms, quasi-antonyms, and quasi-conversives of L.
NB: The same set-theoretical subscripts are also used for other LFs in order to show their semantic approximateness,
when appropriate.
• They do not have a part of speech of their own: they are of the same part of speech as their
keyword L. (Again, all other LFs have their own part of speech.)
• Lexical relations underlying them are symmetrical; roughly:
if X = Syn(Y), then Y = Syn(X);
if X = Anti(Y), then Y = Anti(X); and
if X = Conv(Y), then Y = Conv(X);
1. Syn [Lat. synonymum]: synonym
This LF corresponds to the basic relation of synonymy, or identity of meaning, which is
so important in MTT. Syn(L) has the same or approximately the same meaning as L: (Syn(L)) =
(L) or (Syn(L)) ≈ (L), and the same Deep Part of speech; the LF Syn is applicable to LUs of any
meaning and any part of speech—to lexemes as well as to phrasemes.
NB: (Being synonyms) is a particular case of (being synonymous). Two linguistic expressions of any type can be
synonymous, but only two LUs, which have a part of speech, can be synonyms.
Two major type of synonyms are distinguished: full, or absolute, synonyms ((Syn(L))
= (L)) and approximate, or quasi-, synonyms ((Syn(L)) ≈ (L)). Quasi-synonyms, in their turn,
are
• richer, or more specific: Syn ⊃
• poorer, or less specific: Syn ⊂
• intersecting: Syn ∩

—Chapter 18. Lexical Functions— 29
The lexicographic definition of a richer synonym of L includes L’s meaning, so that if L1 =
Syn ⊃ (L2), then (L1) ⊃ (L2). Syn ⊃ (L) logically entails L, but not vice versa, while Syn ⊂ (L) is
logically entailed by L, but not vice versa.
NB: Syn ⊃ and Syn ⊂ are reciprocally inverse. Thus, if Syn ⊃ (L1) = L2, then L1 = Syn ⊂ (L2).
Examples
Syn(helicopter) = chopper, copter
Syn(policeman) = cop
Syn(telephoneV) = phoneV
Syn(heart) = ticker
Syn(standard) = garden-variety
Syn(stopV [N]) = pull the plug [on N]
Syn ⊃ (fireV [upon N]) = shellV [N] [shellV [N] entails fireV [upon N], but not vice versa]
Syn ⊂ (break away) = escapeV [break away entails escapeV, but not vice versa]
Syn ∩ (escapeV) = elude, avoid
Syn ∩ (kick the bucket) = bite the dust
Remarks
• Lexical synonymy is considered in MTT with respect to specific lexicographic senses only,
i.e., with respect to separate LUs. Therefore, as stated above, all LUs in our examples should
carry distinguishing lexicographic numbers; this is not done for simplicity’s sake.
• Lexical synonymy is considered with respect to meaning only, the differences in distribution
and style/register being immaterial from this stand; they are taken care of in the lexical entries of
synonyms. TICKER is an absolute synonym of HEART, in spite of the fact that the former is a
slang expression, while the latter, a neutral term.
• It is generally assumed that absolute synonymy—like Fr. VÉLO (bike) ~ BICYCLETTE (bike),
or BICYCLE ~ BIKE—is rare. This is true, but its rarity should not be exaggerated: the number of
absolute synonyms in a ‘standard average’ European language is about a few hundred. 9 However,
it is true that approximate synonymy is by far more important in human languages and, as a
consequence, quasi-synonyms are much more widespread.
• Quasi-synonyms must be neutralizable in at least some contexts. For instance, LUs CAR and
VEHICLE are quasi-synonyms because in some contexts one of them can be used instead of the
other: The police chased the blue car throughout Manhattan, but the vehicle disappeared.
WEAPON is a Syn ⊂ of GUN, while FIREARM is not, because the former, but not the latter, can
anaphorically replace GUN in texts: Lucky Joe pulled his gun, but the sheriff wrestled the weapon
〈*the firearm〉 from him; The criminal had two weapons 〈*two firearms〉 hidden in his socks.

—Chapter 18. Lexical Functions— 30
Thus, the LF Syn does not represent a purely semantic relation: it stands for a LEXICAL relation, in
the same vein as Anti and Conv.
2. Anti [Lat. antonymum] = antonym
This LF corresponds to the second important relation dealt with in MTT: roughly speaking,
that between X and not X, i.e., between an expression and its negation. Anti applies to any LU L
whose meaning (L) contains inside a functor which accepts the negation, i.e., (L) admits a nega-
tion; Anti returns for L an LU L´ such that the meanings (L) and (L´) differ only by a negation in
one of them. Anti does not have its own part of speech, but L can be of any part of speech, and
Anti takes on the part of speech of L.
The negation that distinguishes an LU L1 and its antonym L2 can be positioned at different
depth within the meaning (= semantic decomposition) of (L2), and this leads to four major types
of antonyms—especially since they behave differently under paraphrasing (Apresjan 1974: 288-
302, Apresjan & Cinman 2002: 112-116 and Milićević 2003a: 195ff): contradictory Anti - ,
contrary Anti neg, inverse Anti > and opposed Anti opp.
1) Anti - is a contradictory antonym (of L1)—with an ‘internal’ negation, that is, a
negation that bears on a semantic component WITHIN (L1):
if L2 = Anti - (L1), then (L2) = (L´1 ... not L´´1).
This type of antonym corresponds to the mathematical minus in negative numbers; hence the
notation.
Examples
openV Y = (cause that Y becomes open),
and Anti - (openV) = closeV ≡ (cause that Y becomes not open)
allow Y to Z = (say that Y may do Z),
and Anti - (allow) = forbid ≡ (say that Y may not do Z)
Contradictory antonyms Anti - often form triplets in which an Anti - (L) contrasts with a
negation of L:
L Anti - (L) negation of L
X likes Y ~ X dislikes/hates Y ~ X does not like Y
X builds Y ~ X destroys Y ~ X does not build Y
X is joyful ~ X is glum/gloomy ~ X is not joyful
Remark
Note an interesting complication: in some cases, the negative particle signals an Anti - , so
that He did not allow me to leave means that he forbade me to leave, not that he simply did not

—Chapter 18. Lexical Functions— 31
give me his permission. This phenomenon is known as Neg-Transportation, and it is
found with several verbs in different languages. Thus, in Russian, the verb PRIKAZAT´ ([to] order)
does not admit the Neg-Transportation and the Russian sentence On ne prikazal mne uxodit´ (He
did not order me to leave) means exactly what its English gloss says. But VELET´, a quasi-syno-
mym of PRIKAZAT´, behaves differently: On ne velel mne uxodit´ means (He ordered me not to
leave), rather than *(He did not order me to leave). Such cases must be stated in the dictionary: for
VELET´, we have Anti - (velet´) = ne velet´; similarly, for ALLOW: Anti - (allow) = do not allow.
2) Anti neg is a contrary antonym—with an ‘external’ negation, that is, negation that bears
on the semantic head of (L2): if L2 = Anti neg(L1), then (L2) = (not L1).
Examples
wet ≈ (having liquid on its surface),
and Anti neg(wet) = dry ≈ (not having liquid on its surface)
married ≈ (linked to a person by a special rite ...),
and Anti neg(married) = single ≈ (not linked to a person by a special rite ...)
3) Anti > is an inverse antonym based on the (more) ~ (less) contrast:
if L2 = Anti > (L1), then (L1) ⊃ (... α > β ...) and (L2) ⊃ (... α < β ...).
The (more) ~ (less) contrast is, in its turn, reducible to conversion, see below, No. 3:
Example
(X is.more.than Y) = (Y is.less.than X).
heavy X = (the weight of X is.more.than the norm of weight for Xs),
Example
and Anti > (heavy) = light X = (the weight of X is.less.than the norm of weight for Xs).
4) Antiopp is an opposed antonym based on the semantic component (opposite [direction]):
if L2 = Anti opp(L1), then (L2) ⊃ (… opposite to L1).
North = (direction to the left of the speaker when he is facing the sunrise),
and Anti opp(North) = South = (direction opposite to North). 10
In the text, that is, for paraphrasing, an element of the value Anti(L) accompanied by
negation can be used instead of the keyword L: e.g., Alain is still young. ≅ Alain is still not old.
To give the reader a better overview of antonyms, I will give below parallel examples of all
four types of antonymy:

Anti - (hireV) = fireV [an employee]
Anti - (begin) = cease
Anti > (high) = low
Anti > (precede) = follow
—Chapter 18. Lexical Functions— 32
Anti neg(stayV) = leave
Anti neg(dead) = alive
Anti opp(left) = right
Anti opp(backN [human]) = belly
In some languages, one of possible meanings of Anti - , namely, (undo the result of the
preceding action), i.e., (nullify the effect of what has been done before), is grammaticized as a
morphological meaning of reversive, a morphological form found, e.g., in Bantu languages, for
instance, in Swahili [the letter j stands for /z/]:
fung-(a) (fasten) ~ fung+u-(a) (unfasten)
kunj-(a) (fold) ~ kunj+u-(a) (unfold)
tat-(a) (tangle) ~ tat+u-(a) (untangle)
fumb-(a) (close) ~ fumb+u-(a) (open)
teg-(a) (set trap) ~
teg+u-(a) (release trap)
va-(a) (put on clothes) ~
v+u-(a) (take off clothes)
English has a similar formation, too: BUTTONV ~ UNBUTTON, LOCKV ~ UNLOCK, CLIPV ~ UN-
CLIP, etc.; this formation is, however, not as productive and systematic as the reversive in Bantu.
The LF Anti is also used within the system of simple standard LFs, since many such LFs
stand to each other in the relation of antonymy: e.g., Fin ≡ Anti(Incep), Magn vs. AntiMagn, etc.
This will become clearer after the corresponding LFs are introduced.
3. Conv kij [Lat. conversivum] = conversive
Conversives correspond to the third important relation exploited in MTT: that between two
equisignificant LUs whose DSynt-Actants do not correspond. More precisely, this LF returns for
L an LU L´ with the same meaning as L but with its DSynt-Actants [= DSyntAs] i, j and k
permuted with respect to its Sem-actants, so that they become, e.g., the DSyntAs k, i and j [i→k,
j→i, and k→j]. The relation of CONVERSION underlies the deep syntax in MTT. 11 For instance, if
L = PRECEDE and Conv 21(L) = FOLLOW ((A precedes B) = (B follows A)), we have:
A
I
L
II
B
!
A
Conv (L)
21
II I
B

—Chapter 18. Lexical Functions— 33
As Syn and Anti, the LF Conv has no part of speech of its own: it ‘borrows’ the part of
speech of L. It is applicable to an LU L of any meaning and any part of speech, provided that L
has Sem- and DSynt-actants.
Remarks
1. Interestingly, for conversion to be possible, it is enough if L has just one Sem- and one
DSynt-actant. The DSyntA I can be permuted to become II, such that L remains without DSyntA
I. Thus, the expression She [I] vomits corresponds in Russian to Eë [II] rvët—without the DSyntA
I and with a dummy zero Surface Subject. The Russian verb RVAT´ is Conv Δ1(VOMITV): the
(translation of the) expression that is the DSyntA I of VOMIT becomes the DSyntA II of RVAT´,
which has no DSyntA I. To show that an actantial position of L is filled up by a surface dummy,
the symbol Δ is used.
2. Under conversion, DSyntAs can be lost. Thus, COSTV = Conv 421(PAYV), where the
DSyntA III of PAY (to whom?) cannot be expressed with COST: John [I] paid Mary [III] $100 [II]
for the album [IV] ~ The album [I] cost John [III] $100 [II] (*to Mary).
In quite a few languages the permutation of DSyntAs of L with respect to its SemAs can be
expressed morphologically: it is the passive voice, or, to be more precise, passive voices (since
there are different types of passives). For instance, Lat. Marc+us [I, NOM] Petr+um [II, ACC]
occidi+t+Ø [ACTIVE] (Mark is-killing Peter) = Petr+us [I, NOM] a Marc+o [II, ABL] occidi+t+ur
[PASSIVE] (Peter is-being-killed by Mark).
The LF Conv is crucial in the system of simple standard LFs, since many standard LFs
stand in the relation of conversion to each other. Thus:
Func 1 ≡ Conv 21(Oper 1), Labreal 213 ≡ Conv 321(Fact 2), etc.
This fact is widely exploited in paraphrasing.
Examples
Conv 21(include) = belong
[This set includes the element E = The element E belongs to this set.]
Conv 21(wife) = husband
Conv 21(precede) = follow
Conv 231⊃(opinion) = reputation
Conv 21(behind [NY]) = in front [of NX]
Conv 2(fireV [an employee]) = get the ax
Conv 2(prepare) = be in the making
1) X’s [= I] opinion of Y [= II] as Z [ III] ~ Y’s [= I] reputation of Z [= II] among Xs [= III].
[(Reputation), in contrast to (opinion), is necessarily held by several people and concerns the social image of
a person only; this is why it is a more specific ( ⊃ ) conversive.]

—Chapter 18. Lexical Functions— 34
2) In the case of FIREV and PREPARE, the conversive loses one DSynt-Actant of L.
Conv and Anti stand to each other in one of two relations: depending on L, either Conv 21(L)
= Anti > (L) or Conv 21(L) ≠ Anti - (L). Let me illustrate both cases:
Conv 21(L) = Anti > (L)
Anti > (precede) = follow = Conv 21(precede)
Rus. Anti > (prevosxodit´) = ustupat´ = Conv 21(prevosxodit´)
(surpass) (be inferior)
Anti > (bigger) = smaller = Conv 21(bigger)
Anti > (bossN) = subordinateN = Conv 21(bossN)
Conv 21(L) ≠ Anti - (L)
Anti - (give) = take back vs. Conv 321(give) = receive
Anti - (build) = destroy vs. Conv 21(build) = be built
Anti - (wife) = girlfriend, mistress 12 vs. Conv 21(wife) = husband
Anti - (friend) = enemy vs. Conv 21(friend) = friend
LF 4 is not very current; it is used mostly for artistic effects.
4. Contr [Lat. contrarium] = contrastive
This LF returns for L a standard contrastive expression L´ used in the text along with L. A
contrastive LU is superficially similar to an antonym, yet essentially different from it in that it
does not necessarily include a negation.
Generally speaking, the LF Contr does not have its own part of speech and takes the part of
speech of its L. But in most cases known so far, it applies to nouns and thus is itself a noun.
Examples
Contr(earth) = sky Contr(day) = night Contr(head) = heart Contr(land) = sea
LFs 5 and 6 are kinds of hyperonyms of L: No. 5, a neutral hyperonym, and No. 6, a
metaphorical one (but see below).
5. Gener [Lat. genus] = generic concept for L
Let it be emphasized that the LU L´ which is a Gener(L) is just not a simple hyperonym of
L—the name of the corresponding higher class. Hyperonymy is a genuine semantic or conceptual

—Chapter 18. Lexical Functions— 35
relation, so that if L´ is a hyperonym of L in L, the translation of L´ will be a hyperonym of the
translation of L in any other language. But the LF Gener models a LEXICAL relation, and it is
required that the elements of its value satisfy a particular textual-syntactic condition: Gener(L)
must be admissible into certain constructions in L; namely, the elements of its value must be able
to appear in at least one of the following two constructions:
1) Gener(L)−ATTR→DER(L) [where DER is any structural derivation, see below, Nos. 7-10].
The meaning of the whole expression must be the same as that of L:
(Gener(L)−ATTR→DER(L)) = (L).
2) L, X1, X2, ..., Xn and other (kinds of) Gener(L).
Semantically, L is not restricted. Syntactically, L can be of any part of speech, but most
commonly, it is a noun. Gener has the part of speech of L. The elements of the Gener(L) value
normally combine with L (or L’s structural derivation) in the text (i.e., they are non-fused), yet it
is a typical paradigmatic LFs: it is used to construct a complex name for an entity or a situation.
Examples
Gener(republic) = state [republican state = republic]
Gener(liquidN) = substance [liquidA substance = liquidN]
Gener(carrots) = vegetables [carrots, spinach, zucchini and other vegetables]
Gener(arrestN) = reprisals [arrests and other (kinds of) reprisals]
Gener(whisperV) = say [say in a whisper = whisperV ]
Gener(yellowA) = color [yellow, green and other colors]
6. Figur [Lat. figuraliter (figuratively)] = standard metaphor
L´ which is Figur(L) is a clichéed, automatic metaphor for L; the elements of the Figur
value appear in the construction Figur(L)−I→L, so that Figur(L) is used in the text together
with L, forming a collocation. The meaning of this collocation is roughly equal to that of L:
(Figur(L)−I→L) ≈ (L).
Semantically, L can be anything; syntactically, L and Figur(L) are both nouns.
Examples
Figur(remorse) = pangs [pangs of remorse ≈ remorse]
Figur(passion) = flame [flame of passion ≈ passion]
Rus. Figur(blokada) = kol´co [kol´co blokady lit. (ring of blockade) ≈ blokada (blockade)]
Rus. Figur(tuman) = pelena [pelena tumana lit. (shroud of fog) ≈ tuman (fog)]

AntiMagn:
—Chapter 18. Lexical Functions— 36
Quite often the LF Figur is used in configurations (see 4.2, p. 00) with LFs Magn and
[Magn + Figur](fog) = wall [wall of fog ≈ dense fog]
[Magn + Figur](rainN) = curtain [curtain of rain ≈ heavy rain]
[AntiMagn + Figur](doubt) = shadow [a shadow of a doubt ≈ slight doubt]
[AntiMagn + Figur](hopeN) = shred [a shred of hope ≈ a little hope]
The elements of the Figur(L) are non-fused: they always combine with L in the text, but
the expression Figur(L)−I→L is just another name for (L), so that, again, this is a paradigmatic
LF.
LFs 7-22 correspond to derivational meanings of natural languages, i.e., they are
semantic derivations of L (see above, 1.2, Def. 18.1, p. 00ff). The values of the deri-
vational LFs do not necessarily have regular morphological links to the keyword; we find here
many cases of derivational suppletion (such as city ~ urban, sun ~ solar, heavens ~ celestial).
LFs 7-10 are structural derivations of L, i.e., LUs which have the same meaning
as L but are of a different part of speech and therefore are used in different types of DSynt-confi-
gurations. Thus, (S 0(LV)) = (LV), but of course LV and S 0(LV) are syntactically in complementary
distribution. (In point of fact, a structural derivation of L more often than not has a meaning
different from that of L as well, so that it is but an approximate derivation; for a discussion of
possible semantic shifts under structural derivation, see, e.g., Padučeva 1991 and Uryson 1996.)
7-10. S 0, A 0, V 0, Adv 0
= structural derivations
The LFs S0, A0, V0, Adv0 are, respectively, a noun (= S(ubstantive)), an adjective, a verb and
an adverb that has the same meaning as L:
S 0(analyze) = analysis V 0(analysis) = analyze
A 0(sadness) = sad S 0(sad) = sadness
A 0(city) = urban S 0(urban) = urban character
Adv 0(follow [NY]) = after [NY] V 0(after [NY]) = follow [NY]
Adv 0 (neighborV [on NY]) = close [to NY] V 0 (close [to NY]) = neighborV [on NY]
LFs S0, A0, V0 and Adv0 are applicable to an LU L of any part of speech different from that
of the LF in question (i.e., S 0 is applicable to any LU except a noun, A 0—to any LU except an

—Chapter 18. Lexical Functions— 37
adjective, etc.), provided the meaning of L is a functor, i.e., the name of an event, process,
relation, property, etc. All structural derivations have their inverse functions: if S 0(LV) = L´, then
V 0(L´) = LV, etc.
However, A 0 is an exception in this respect: it can apply to the names of concrete objects,
and when an A 0 is applied to a concrete noun, it has no inverse function, as can be seen from the
example with A 0(city) above.
An overview of theoretically possible structural derivations is presented in Fig. 2 below.
DER(L)
L
S0 A0 V0 Adv0 S A (honesty) = honest
0 V0(attackN) = attackV Adv (rapidity) = fast
0
A S 0(honest) = honesty V 0(sick) = vomitV Adv 0 (broad) = broadly
V S 0(attackV) = attackN A 0 (vomitV) = sick
Adv S0(fast) = rapidity [room] A0(broadly) = broad V0(fast) = rushV
Figure 18-1: Theoretically Possible Structural [= DSynt-] Derivations
Adv 0 (rushV) = fast
LFs 11-20 are meaningful derivations of L—i.e., LUs which add something to
the meaning of L. As a rule, the meaning of such a derivation of L includes that of L:
(DER meaningful (L)) ⊃ (L).
LFs 11-20 are applicable to any L which expresses a functor—that is, denotes a si-
tuation. Among them, LFs 11-16 are actantial and circumstantial nouns. S1 is applicable to LUs
that have at least one Sem-Actant, S 2 to LUs that have at least two SemAs, etc.
11. S i = actantial nouns
The LF S i is the standard name of the i-th DSynt-actant of L; S i are actantial nouns:
principally, agent noun [(the one who L-s)] and patient noun [(the one whom someone/ something
L-s)].
The elements of the value of an S i(L) are of two syntactic types discussed above: fused and
non-fused. A fused element is used in the text instead of L (especially if this L is a verb); a non-
fused element is used in the text together with L, taking it as its own DSyntA II: S i−II→L.

Examples
for the verb TEACH:
(person X [I] teaches subject Y [II] to people Z [III])
S 1(teach) = //teacher
S 2(teach) = //(subject) matter [in high school]
S 3(teach) = //student; pupil
—Chapter 18. Lexical Functions— 38
for the noun LETTER:
(letter by person X [I] to person Y [II] about Z [III])
S 1(letter) = writer [of the letter]; sender [of
the letter]
S 2(letter) = addressee [of the letter]
S 3(letter) = contents [of the letter]
In the left-hand column, one sees fused elements of the value of the corresponding LFs, and
in the right-hand column, non-fused elements (see above, 2.4, Item 1, p. 00, and below, 5.3, p.
00). The fused character of an LF value element is shown—here and below—by the symbol ‘//,’
preceding all the fused elements. Note also that in cases where the value of an LF contains more
than one element we put commas between elements that are (almost) synonymous; those that
show a more serious semantic difference are separated by a semicolon (thus, strictly speaking, the
(writer of a letter) ≠ (the sender of this letter)).
12-16. S instr, S med, S mod, S loc, S res = circumstantial nouns
The LFs S instr, S med, S mod, S loc, S res are the standard names of main (Deep-Syntactic)
circumstants of L; they are circumstantial nouns: respectively, the standard name of
instrument, of means, of mode, of location, and of the result presupposed by the situation (L) (L
is, as a rule, a noun or a verb). Like actantial nouns, circumstantial nouns normally are used
instead of their keyword L; if they don’t, they also take L as their DSyntA II: S instr−II→L.
Examples
Sinstr⊃ (shoot) = //firearm
S instr(murderV,N) = murder weapon
Smed⊃ (shoot) = //ammunition
S mod(consider) = //approach
[I consider this problem as follows: ...
≈ My approach to this problem is as follows: ...]
S loc(fightV [two armies]) = //battlefield
S loc(war) = theater of war
Sres⊃ (learn) = //knowledge, skills
Sres⊃ (explosion) = shockwave
S res⊃ (copyV) = //copyN
LFs 17 and 18 specify properties of potential and probable DSynt-actants of L; syntactical-
ly, both are adjectives and both replace L in the text. Both apply only to LUs whose meaning is a
real functor (action, state, process, property, …) and thus presupposes actants.
17. Able i [Lat. habilis (able, manageable)] = potential actantial adjectives

—Chapter 18. Lexical Functions— 39
L´, an element of Able i(L), names the determining property of the (referent of the) i-th
potential DSynt-actant of L. More precisely, Able 1 means (which can/tends to L), and Able 2,
(which can/tends to be L-ed); thus, the meaning of Able i(L) includes that of L.
Examples
Able 1(cryV) = tearful Able 1(varyV) = variable
Able 2(proveV) = provable Able 2(trustV) = trustworthy
Rus. Able 2(somnevat´sja (doubtV)) = somnitel´nyj lit. (doubtable)
Rus. AntiAble 2(somnevat´sja (doubtV)) = nesomnennyj lit. (undoubtful) =
(such that one cannot doubt it)
Cf. also:
S 1Able 1(cryV) = crybaby
S 1Able 1Caus(cryV) = tearjerker [a movie, a book]
S 1Able 2(laugh2V) = laughing stock
18. Qual i [Lat. qualitas (quality)] = qualifying actantial adjectives
L´, an element of Qual i(L), names the determining property of the i-th probable DSynt-
actant of L, that is, the property such that it entails the tendency of the referent of this actant to
L/to be L-ed; the meaning of Qual i(L) does not include that of L.
Examples
Qual 1(cryV) = sad
Qual 1(laugh1V) = cheerful
Qual 1(win [over NY]) = strong
Qual 1(prove) = logical
Qual 2(doubtV) = implausible
Qual 2(laugh2V) = awkward; absurd
Qual 2(win [over NY]) = weak
Qual 2(prove) = obvious
Qual i entails the probability of Able i but not vice versa: if something is implausible it
tends to be doubtful, but a doubtful statement is not necessarily implausible; an obvious state-
ment is likely to be provable, but something provable is by no means always obvious; etc.
LFs 19 and 20 are, roughly speaking, ‘quantifiers:’ a singulative and a collective. They are
inverse with respect to each other:
(Sing(Mult(L1))) ≈ (L1) and (Mult(Sing(L2))) ≈ (L2).

—Chapter 18. Lexical Functions— 40
Syntactically, both are nouns; Sing takes its L (when it does not replace it) as its DSyntA II, and
Mult takes its L as DSyntA I: Sing-II→L, Mult-I→L. 13 L also is a noun, denoting a collectivity/a
set/a substance/a state (in case of Sing) or an object/a person/an event (in case of Mult).
In some languages, the meanings of Sing and Mult are expressible via morphological
elements—grammemes/derivatemes—of singulative and collective.
19. Sing [Lat. singulus (singular, unique)] = singulative
L´ which is a Sing(L) means (one instance/one unit of L); it designates, so to speak, a
‘quantum’ of (L).
Examples
Sing(fleet) ≈ //ship
Sing(sand) = grain [of ~]
Sing(thunderN) = [~]clap [= thunderclap]
20. Mult [Lat. multum (multitude)] = collective
together.
Examples
Sing(air force) ≈ //military aircraft
Sing(anger) = fit, gust [of ~]
Sing(automatic fire) = burst [of ~]
L´ which is a Mult(L) means (an aggregate/a set of Ls); it designates several Ls taken
Mult(ship) ≈ //fleet
Mult(military aircraft) ≈ //air force
Mult(criticism) = barrage, chorus [of ~s]
Mult(wolf) = pack [of ~s]
Mult(lie) = pack [of ~s]
Mult(lion) = pride [of ~s]
LFs 21 and 22 designate two elements important in the structure of an organization or a
team denoted by L—respectively, the ‘boss’ and the ‘personnel.’ These LFs have a rather narrow
sphere of application: their Ls can be only nouns that denote things having a ‘boss’ and a ‘person-
nel.’ Both LFs are nouns and take L as their DSyntA II: Cap−II→L and Equip−II→L.
21. Cap [Lat. caput (head)] = (the head of …)
Examples
Cap(university) = president [of ART ~];
rector [of ART ~]
Cap(tribe) = chief [of ART ~]
Cap(aircraft) = //captain
Cap(department [at a university]) = head, chairman,
chair [of ART ~]
Cap(empire) = //emperor
Cap(catholic church) = //pope

—Chapter 18. Lexical Functions— 41
22. Equip [Fr. équipage (crew)] = (the staff/the crew of …)
Examples
Equip(university) = staff [of ART ~];
faculty [of ART ~]
Equip(theater) = company [of ART ~]
Equip(aircraft) = crew [of ART ~]
Equip(hospital) = personnel, staff
[of ART ~]
Equip(empire) = population, subjects
[of ART ~]
Equip(catholic church) = clergy [of ART ~]
Cap and Equip both have mostly non-fused elements of the value—L can be added to L´
almost everywhere, as we see in the above examples; as a rule, L´ forms with L a collocation.
Nonetheless, Cap and Equip determine particular names (‘What do you call the head of L?’,
‘What do you call the personnel of L?’), and thus both are paradigmatic LFs.
LFs 23 and 24 correspond to syntactic inflectional meanings: those of the participle and
of the deverbal adverb, respectively. Both are applicable to LUs whose meanings are functors and
have actants.
23. A i = actantial adjectives
The LF A i denotes the determining property of the i-th Deep-SyntA of L [= DSyntAi(L)]
from the viewpoint of its role in the situation (L). Semantically, A 1 is roughly equivalent to an
active participle (≈ (which is L-ing)), and A 2, to a passive participle (≈ (which is being L-ed)).
Syntactically, A i is an adjective; it is prototypically used, as is expected of a Deep Adjective, to
modify a noun.
Examples
A 1(angerN) = in [~] //angry
A 1(importance) = of [~] //important
A 1(knowV) = //aware
A 1 (searchV[for NY]) = //in searchN [of NY]
24. Adv i= actantial adverbs
A 2(fireV [upon NY]) = //under fire
A 2(conduct [an orchestra]) = //under the baton [of NX]
A 2(crossfire) = caught [in ~]
A 2(analyze) = //under analysis
The LF Adv i denotes the determining property of the action by the i-th DSyntA of L
according to the role of the DSyntAi(L) in the situation described by L. Semantically, Adv 1 is
roughly equivalent to an active deverbal adverb (≈ (while L-ing)), and Adv 2, to a passive deverbal
adverb (≈ (while being L-ed)). Syntactically, Adv i is a Deep adverb; it is prototypically used to
modify a verb, an adjective or another adverb.

Examples
—Chapter 18. Lexical Functions— 42
Adv 1(anger) = with [~] //angrily Adv 2(bombard) = //under bombardment
[They fled under heavy bombardment.]
Adv 1 (speed) = at [a ~ of ...] Adv 2(permit V ) = //by permission
Adv 1(decreaseN,V) = //down
[Sales amounted to 6 million, a decrease of 2.7% =
... to 6 million, down 2.7%]
Adv 2(applause) = to [the ~ of NX]
Adv 2(charges [jur.]) = on [ART ~]
Three Deep-Syntactic subtypes of Adv i are distinguished from the viewpoint of the ‘alter-
nation’ between an Adv i and the Main Verb; namely, I Adv 1 , II Adv 1 and III Adv 1 :
(2) a. It is strange [L]−I→that John left. ~ Strangely [ I Adv 1(L)], John left.
b. John hurried [L]-II→to leave. ~ John left in a hurry [ II Adv 1(L)].
| ———III——— ↓
c. John starts [L] the paper by giving an example. ~
At the beginning[ III Adv 1(L)] of his paper, John gives an example.
The left superscript identifies the DSyntA(L) that becomes the Main Verb when L is replaced
with Adv i(L). This encoding is used by Lexical Paraphrasing Rules Nos. 46-48, see Part IV,
Chapter 16, p. 00.
LFs 25 and 26 reflect what in many languages are SEMANTIC INFLECTIONAL MEANINGS of the
verb. They are applicable only to a verbal LU.
25. Imper [Lat. imperāre ([to] command, order)] = imperative
L´ as an element of value Imper(L) is a received imperative expression meaning (do L!). It
corresponds to the (IMPERATIVE) grammeme of the inflectional category of mood; its Deep-Syntactic
part of speech is clausative—an expression that constitutes a full clause by itself.
Examples
Imper(shoot) = Fire!
Imper(speak quietly) = Sh-h-h!
Imper(seize arms) = Take up arms!
Imper(stopV [to a horse]) = Whoa!
26. Result [Lat. *resultāre ([to] result)] = resultative
L´ of the LF Result is necessarily a verb meaning (be the expected result of L). This LF
corresponds to the (RESULTATIVE) grammeme of the inflectional category of aspect—in languages
that have this category with this grammeme.

Examples
—Chapter 18. Lexical Functions— 43
Result(buy) = ownV Result(lie down) = be lying
Result(learn) = know [how], have (necessary) skills [for N/Vinf]
Cf. the LF S res (No. 16, p. 00): S res(buy) = [a] buy, S res(learnt) = knowledge, skills, as well as
Real 1 of a verb, No. 52, p. 00.
3.3. Syntagmatic Lexical Functions (27-64)
Syntagmatic LFs can be naturally subdivided according to their Deep-Syntactic part of
speech: they can be nominal, adjectival/adverbial, prepositional (a subclass of adverbial LFs), and
verbal.
3.3.1. Nominal Lexical Functions(27)
27. Centr [Lat. centrum (center)] = (the highest point of …)
L´ as an element of the value Centr(L) means (the center/the culmination of L); syntac-
tically, it is a noun that takes the name of L as its DSyntA II: Centr-II→L. This LF applies to an
LU that has a rather specific meaning—such that it makes sense to define for it a culminating
part.
Examples
Centr(forest) = the thick [of the ~]
Centr (crisis) = the peak [of the ~]
Centr (struggleN) = climax [of ~]
Centr (glory) = summit [of ~]
Centr (life) = prime [of ~]
Centr (desert) = the heart [of the ~]
Centr is used, as a rule, in combination with the LF Loc in (No. 35, p. 00):
Loc inCentr(forest) = in the thick [of ART ~]
Loc inCentr(glory) = at the summit [of ~]
Loc inCentr(life) = in prime [of ~]
Loc inCentr(winter) = in the dead [of ~]
NB : Although this LF seems to be formally similar to Cap and Equip, it is in fact very different from them.
While Cap and Equip return the NAMES to be used for the (head) and the (personnel) of (L) (and are thus
paradigmatic), Centr returns an expression L´ that serves not to name anything, but to CHARACTERIZE (L)
(and is thus syntagmatic). The difference manifests itself in the following fact: the elements of the values of
Cap and Equip are readily used without L (the rector, the emperor, ..., the crew, the personnel, ...), but the
elements of the value of Centr cannot be so used (*the thick, *the peak, *the prime).

3.3.2. Adjectival/Adverbial Lexical Functions (28-34)
—Chapter 18. Lexical Functions— 44
LFs 28-34 express different QUALIFICATIONS of L: Epit is an empty cliché modifier, Magn,
Plus and Minus are quantitative modifiers (“(L)—to what extent?”), Ver is an objective qualifier
(“Does (L) correspond to what the speaker believes to be the norm for L’s referents?”), while
Bon/Pos are subjective qualifiers (“Does the speaker/S 1((L)) like (L)?”). All of them are DSynt-
adjectives and/or DSynt-adverbs and they depend on L via the DSynt-Relation ATTR:
Magn←ATTR–L, etc.
28. Epit [Lat. epitheton] = redundant cliché
The LF Epit returns, for a given L, the expression L´ that is a common (almost) empty
epithet for L. Its value is a cliché adjective (that in many cases simply repeats a component of L’s
meaning and thus does not contribute to the meaning of the whole expression) which is employed
along with L according to the linguistic usage. L is a noun, so that Epit(L)←ATTR–L.
Examples
Epit(winner) = lucky [speaking of a lottery] Epit(parent) = proud
Epit(ocean) = immense Epit(oaf) = clumsy
Epit(interest) = vested Epit(intrigue) = underhand
Epit(opinion) = humble Epit(understanding) = humble
29. Magn [Lat. magnus (big, great)] = intensifier
The LF Magn means (very), (to a (very) high degree), (intense(ly)), (completely). Its L can be
any LU that has a gradable component inside its meaning;L’s part of speech does not matter.
Examples
Magn(naked) = stark
Magn(laughV) = heartily < one’s head off
Magn(abyss) = yawning
Magn(injury) = serious
Magn(patience) = infinite
Magn(skinny) = as a rake
Magn(price) = high < exorbitant
Magn(powers) = vast
The symbol ‘

—Chapter 18. Lexical Functions— 45
A curious case of Magn is known in Wolof, where many so called descriptive verbs—verbs
denoting properties— have each a special intensifier (to facilitate the reading, Wolof words are
given in phonological transcription rather than in national spelling; letter doubling indicates long
phonemes):
Magn(xonq (be red)) = čur, čojj
Magn(weex (be white)) = tal
Magn(≠uul (be black)) = kukk
Magn(leer (be light [in color])) = na≠
Magn(set (be clean)) = weč
Magn(tang (be warm)) = zur
Magn(tak (be burning)) = zeret
Magn(xasaw (be smelly)) = xu≠
The expression xonq čur thus means (very-very red), set weč—(absolutely clean), and tak zeret—
(burn very intensely) = (blaze).
30-31. Plus/Minus = comparison markers
Plus and Minus mean, respectively, (more) and (less) ((to a greater/lesser extent) [than
something else]; these are, so to speak, comparative degrees of Magn). Plus and Minus are mostly
used in complex LFs, combined with Pred and a phasal/causative verb. They apply to LUs to
which Magn can apply.
Examples
IncepPredPlus(prices) = skyrocket, soar, spike
IncepPredPlus(temperature) = increases, rises
Caus 1PredPlus(aggression) = escalate [ART ~]
IncepPredMinus(pressure) = decreases
IncepPredMinus(health) = is failing
CausPredMinus(blowN) = soften [ART ~]
[for the LFs Pred, Incep and Caus, see below, Nos. 42, 46 and 49].
NB: In the examples, an LF verb whose keyword is its SSynt-Subject is quoted in the form of the 3rd person (in
the present of the indicative); otherwise, LF verbs are quoted in the infinitive.
32. Ver [Lat. verus (real, genuine)] = objective qualifier
L´ of this LF means (as it should be), (meeting intended requirements); its L, which can be
of any part of speech, must contain in its meaning a component able to receive such characteriza-
tion.

Examples
Ver(surpriseN) = sincere, genuine, unfeigned
Ver(punishment) = well-deserved
Ver(demandN) = legitimate
Ver(container) = leakproof; airtight
33. Bon [Lat. bonus (good)] = subjective qualifier
—Chapter 18. Lexical Functions— 46
Ver(instrument) = precise
Ver(walkV) = steadily
Ver(citizen) = loyal
Ver(argue) = convincingly
Here, L´ means (good), i.e., this adjective is a received praise for L coming from the Speak-
er. Semantically, the LF Bon applies to any LU whose referent can in principle be evaluated; L
may belong to any part of speech.
Examples
Bon(cutV) = neatly, cleanly
Bon(struggleN) = heroic
Bon(proposal) = tempting
Bon(analysis) = fruitful
Bon(service) = first-class
Bon(behavior) = good < excellent
Bon(aidN) = valuable < invaluable
Bon(place) = prominent
The symbol ‘

—Chapter 18. Lexical Functions— 47
[Pos + Oper 2](voteN) = sweep [~] [He is expected to sweep vote in this region.]
[Pos Real 2](pressN) = have [(a) good ~] [He played all games and had good press.]
Remarks
1. The LFs Magn, Ver, Bon and Pos are often combined with the LF Anti to produce complex
LFs. Thus:
Magn(temperature) = high, while AntiMagn(temperature) = low
Pos(opinion) = good, high, while AntiPos(opinion) = bad, low
Pos(reviewN) = good, positive < rave, while AntiPos(review) = bad, negative < damning,
scathing
[Bon(review) = good < excellent, brilliant(ly written); a brilliant(ly written) review may well be
positive or negative, i.e., Pos or AntiPos.]
Note an interesting LF configuration for REVIEW:
[between Pos and AntiPos](review) = mixed
Rus. [between Pos and AntiPos](recenzija (review)) = kislosladkaja lit. (bitter-sweet).
Pos(criticism) = favorable AntiPos(criticism) = adverse
Cf. as well:
AntiBon(car) = //jalopy, lemon, kold cratel
AntiBon(start) = rocky
AntiBon(dwelling) = rundown, shabby
AntiBon(hotel) = seedy, sleazy //kflea bagl
2. The LFs Magn, Ver, Bon and Pos are semantically close to each other, and with some
keywords they can be indistinguishable. It stands to reason that something which is (as it should
be) [= Ver] is (good) [= Bon] because of this; (positive) opinions [= Pos] are generally considered
as (good); something which functions intensely [= Magn] tends to be perceived as something that
is (as it should be) [= Ver] and (good) [= Bon]; etc. Thus, GOOD in the collocation good health is at
the same time Magn, Ver and Bon of HEALTH. With some other keywords, these LFs can, how-
ever, be contrasted, as, for instance, with LUs OBSERVER and OUTPUT:
2003.
OBSERVER OUTPUT
Magn close high < huge
Ver objective, unbiased stable, steady
Bon keen excellent, high-quality
Such overlaps happen with other LFs as well; see the discussion of this topic in Kahane

3.3.3. Prepositional Lexical Functions (35-40)
—Chapter 18. Lexical Functions— 48
Semantically, LFs 35-40 express different localizations of something with respect tO (L) or
the role (L) plays in the situation in question. Syntactically, Loc in, Loc ad and Loc ab are locative
prepositions (with the meanings (in/at L), (into L), (out of L)), Loc
causative preposition ((by using L) and (being the cause of L)). All of them constitute a special
!
subclass of DSynt-adverbs; they take L as their DSyntA II:
temp
in
is a temporal preposition
((during L) or ( at the moment L)), while Instr and Propt are respectively an instrumental and a
Loc in−II→L, etc.
The keyword L of these LFs is a noun with an appropriate meaning, that is, whose referent
admits localization with respect to it.
35-37. Loc in, Loc ad, Loc ab
[Lat. locus (place)] = localizations in space
The LF Loc expresses a spatial or abstract localization of something with respect to L—
with the corresponding directionality, i.e., (being in) ( in), (moving into) ( ad), (moving out of) ( ab).
Examples
Loc in(height) = at [a ~ of ...] Loc ad(height) = to [a height of ...]
Loc in(levelN) = on [ART ~] Loc ad(levelN) = to [ART ~]
Loc ab(height) = from [a height of ...]
Loc ab(levelN) = from [ART ~]
L designates a spatial or abstract entity where something can be located.
To give a few interesting examples of Locs from different languages:
English Italian
You put gloves on your hands Mettete i guanti sulle mani
and the ring on your finger. e il anello al dito.
German
on holiday, in the country, at a party auf Urlaub, auf dem Lande, auf einer Party
Russian
v otpuske, v derevne, na večerinke
French
en conge, à la campagne, na večerinke
Russian
on the list, in the diagram, at the airport v spiske, na risunke, v aèroportu

!
!
38. Loc
temp
in
= localization in time
situation
!
with respect to (L).
Examples
—Chapter 18. Lexical Functions— 49
L´ = Loc (L) is a preposition governing L and designating a temporal localization of a
in
!
temp
temp
Loc (arrestN) = at the time [of ART ~]
in
Loc
temp
in (capture) = at the moment [of ~]
Loc (analysis) = in the course [of ART ~]
in
Loc
temp
It is interesting to compare the values ! of Loc from different names of points in time:
in
[John will arrive] on Monday at 3P.M. next week in the evening in 2006
temp !
As one can see, for Loc the keyword L refers to a fact or to a time ‘segment’ during
in
the occurrence of which something happens. !
temp
temp
temp
in (capitalism) = under [~]
Loc , with the meaning (reaching a point in time), and Loc ((starting from a point in
ad ab
!
time)) are theoretically possible, but I do not have examples illustrating sufficient variability and
phraseological character of their expressions.
!
39. Instr [Lat. instrūmentum (instrument)] = by means of
L´ of Instr(L) is a preposition meaning (by means of [L]); L refers to a physical entity or a
fact which is used in order to bring about a desired state of affairs.
Examples
Instr(typewriter) = on [ART ~] Instr(satellite) = via [~]
Instr(mail) = by [~] Instr(argument) = with [ART ~]
40. Propt [Lat. propter (because of)] = because of
Semantically, the LF Propt is related to the LF Caus [≈ (cause)), see below, No. 49, p. 00.
More precisely, Propt is an abbreviation for Adv 2Caus [≈ (being caused by...)]. Syntactically, L´
= Propt(L) is a preposition meaning (because of/as a result of [L]); L itself is a noun.
Examples
Propt(fearN) = from, out of [~] Propt(loveN) = out of [one’s ~ of ...]
Propt(lackN) = for [~ of NY] Propt(fireN) = in [ART ~] [John was killed in a fire.]
Rus. Propt(revnost´ (jealousy)) = iz (out of) [revnost+i]
Rus. Propt(strax (fear)) = so (from) [strax+u]
Rus. Propt(bol´ (pain)) = ot (of) [bol+i]
!
temp

3.3.4. Verbal Lexical Functions (41-64)
—Chapter 18. Lexical Functions— 50
The LFs 41 and 42 are copular expressions; their keyword L is a nominal.
41. Copul [Lat. copula] = copula
Copul is a verb meaning (be [something]). It takes L as its DSyntA II: Copul(L)−II→L.
Examples
Copul(teacher) = be [a ~], work [as a ~]
Copul(example) = be, represent [an ~], serve [as an ~]
IncepCopul(ill) = fall [~]
IncepCopul(sourAdj) = go, turn [~] //sourV, turn
42. Pred [Lat. *praedicātum ([a] predicate)] = verbalizer
L´ = Pred(L) is a verb meaning (be (an) L). Thus, semantically, Pred covers the syntactic
configuration Copul(L)−II→L. Pred converts a non-verbal L into a syntactic verbal predicate. In
other words, Pred(L) is nothing but a fused expression for Copul(L)−II→L, needed for the con-
venience of some synonymic transformations:
Its value is thus always fused.
Examples
Pred(L) ≡ //Copul(L)−II→L.
Pred(drunkard) = //drinkV [as in He drank all his life.]
Pred(commander) = //commandV
Pred(close) = //borderV [on N] [His attitude borders on insubordination.]
Pred(far) = //be a far cry [from N][N designates a property, a state, etc.]
Rus. Pred(bol´noj (ill)) = //bolet´ (be ill)
Rus. IncepPred(bol´noj (ill)) = //zabolet´ (fall ill)
ing ones:
Pred is often used in complex LFs; see a few examples in Nos. 30-31 and also the follow-
PredA 2(threaten) = //be under the gun
PredAble 2(doubtV) = //invite doubts
PredAntiAble 2(doubtV) = //be beyond doubt
PredMagn(easy) = //be a piece of cake
PredMagn(hungry) = //could eat a horse
[In I am starving! we see the adjective starving = //Magn(hungry)]

Remarks
—Chapter 18. Lexical Functions— 51
1. Cf. similar, but different collocations of the type Animosity ran rampant or The traffic is
flowing smoothly, which are described by configurations of LFs (4.2, p. 00):
[Magn + Func 0](animosity) = runs rampant 〈high〉
[Ver + Func 0](traffic) = flows smoothly
2. Copul and Pred stand in a restricted synonymy relation with the LF Oper 1, see immediately
below. Namely, for L which is an adjective or a relational noun, Copul(L) = Oper 1(L).
The verbal LFs 43-54 are, respectively, support verbs, phasal verbs, causation
verbs, and fulfillment verbs. The LFs of these four groups are most naturally introduced by
triplets (why this is so will be explained below).
43-45. Support verbs
The LFs Oper i, Func i and Labor ij are support (or ‘light’) verbs (cf. Gross 1981);
they are semantically empty (or at least emptied) in the context of L, which is their keyword. The
keyword of a support verb is necessarily a noun whose meaning is or includes a functor, thus
presupposing actants. In other terms, the keyword of these LFs is, as a general rule, the name of
an action, an activity, a state, a process, a property, a relation, etc. It can also be the name of a
concrete object, which is defined by its role in a situation. Such is the case, e.g., of a body part or
an organ: they represent what is called ‘inalienable possession’ and have as the value of the LF
Oper 1 the verb HAVE or its equivalent: for instance, Oper 1(HEART) = HAVE [ART ~]. Another
example is the noun OBSTACLE: an obstacle can well be a physical object (a rock, a fallen tree, a
ditch), yet it has support verbs: e.g., Oper 2(obstacle) = face [ART ~]. The noun BOOK in the
context of a library (as an object of lending/borrowing) also has support verbs: Caus 2Func 2(book)
= borrow [ART ~], while in Russian the corresponding support verb is brat´ lit. (take). And so
forth.
The support verbs serve to link, on the DSynt-level, (the name of) a DSynt-actant of L to L
itself; they thus play an important semantic-syntactic role and can be loosely called semi-auxilia-
ries.
Remarks on semantic emptiness of support verbs
Saying that the support verbs Oper i, Func i and Labor ij are ‘semantically empty,’ I mean
the following:

—Chapter 18. Lexical Functions— 54
43. Oper i [Lat. operāri (do, carry out)] = support verb with L as 1st Object
The DSyntA I of an Oper i verb = L ´(and its SSynt-Subject) is the phrase that is described
in the Government Pattern of L as the i-th DSyntA(L), and Oper i’s DSyntA II (= its main S(urf-
ace)Synt-Object is L itself. 15 Further DSyntAs of an Oper i , if any, are the phrases described in
the GP of L as further DSyntAs of L.
Examples
Oper 1(blowN) = deal, strike [NY ART ~] Oper 2(blowN) = receive, take [ART ~ from NX]
Oper 1(supportN) = give, lend, offer [~ to NY] Oper 2(supportN) = receive [~ from NX]
Oper 1 (orderN) = give [ART ~ to NZ] Oper 3 (orderN) = receive [ART ~ from NX]
Oper 1(resistance) = put up [ART ~], Oper 2 (resistance) = meet [ART ~], run [into ART ~]
offer [ART/Ø ~]
Oper 1(controlN) = have [~ over NY] Oper 2 (controlN) = be [under NX’s ~]
Oper 1 (examN) = give [ART ~] Oper 2(examN) = take [ART ~] Oper 3(examN) = be [on ART ~]
The expression in brackets following each element of the value of the LF illustrated is in
point of fact its reduced Government Pattern—its lexical subentry; for a discussion of
such subentries, see 5.1, p. 00.
In case an Oper i verb admits only a dummy subject (a purely grammatical lexeme,
which has no meaning of its own), the subscript ‘ 0’ is used: Rus. Oper 0(zapax (smell)) = tjanut´
[~om] [Iz podvala tjanulo zapaxom moči lit. (From the-basement [it] pulled with-the-smell of-
urine), with a zero dummy subject Ø3sg.neu], Fr. Oper 0(baisse (decrease)) = on constate une baisse
lit. (One sees a decrease) [with the subject ON], Ger. Oper 0(Problem (problem)) = Es gibt ein
grosses Problem lit. (It gives [a big problem]) [with the subject ES].
Let it be recalled (p. 00, NB under Conv, No. 3) that all Oper is are counted as one LF, and
so are all Func is and all Labor ijs, see below.
44. Func i [Lat. *functionāre ([to] function) = support verb with L as Surface Subject
The DSyntA I of the Func i verb (and its SSynt-Subject) is L itself, and its DSynt-actant II
(= its main SSynt-Object) is the i-th DSyntA(L).
Examples
Func 1(blowN) = comes [from NX] Func 2(blowN) = descends [on NY], falls [upon NY]
Func 1(proposal) = comes, stems [from NX] Func 2(proposal) = concerns [NY]

—Chapter 18. Lexical Functions— 55
NB: Let it be remembered that LF verbs taking their keyword as SSynt-Subject are quoted in the 3rd person
singular: such is, e.g., the case of all Func; otherwise, they are in the infinitive.
is used:
In case there is no object at all, i.e., Func i is an absolutely intransitive verb, the subscript 0
Func 0(snowN) = falls [At night, the snow started to fall] Func 0(war) = is on
Func 0(silence) = reigns Func 0(smellN) = lingers
45. Labor ijk [Lat. laborāre ([to] work, toil)] = support verb with L as 2nd Object
The DSyntA I of a Labor verb (and its SSynt-Subject) is the i-th DSyntA(L), its DSyntA II
(= its main SSynt-Object) is the j-th DSyntA of L, its DSyntA III (= its secondary SSynt-Object)
is the k-th DSyntA of L, and its further DSyntA (= its tertiary SSynt-Object) is L itself.
Examples
Labor 12(interrogation) = subject [NY to an interrogation, where the keyword INTERROGATION is
DSyntA III of the verb to subject]
Labor 321(leaseN) = grant [NY to NZ on lease, where the keyword LEASEN is DSyntA IV of
the verb grant; the underlying expression is X leases Z from Y]
Rus. NASLEDSTVO (inheritance) (the underlying expression is X nasleduet Y ot Z-a (X
inherits Y from Z)) has the following three Labors:
Labor 123(nasledstvo) = polučit´ [N acc ot N gen v ~] lit. (receive NY from NZ in inheritance)
Labor 213(nasledstvo) = dostat´sja [N dat ot N gen v ~] lit. (come to NX from NZ in inheritance)
Labor 321(nasledstvo) = ostavit´ [N acc N dat v ~] lit. (leave NY to NX in inheritance)
In all these examples, the phrase v nasledstvo (in inheritance) is the DSyntA IV of Labor.
Oper 0/i, Func 0/i and Labor ijk
can be paired in converse relations:
Oper 1 = Conv 21(Func 1); Labor 12 = Conv 132(Oper 1); etc.
These relations may be diagrammed—for a two-actant LU, in this case ANALYSIS—as follows:
Oper 0
Func 2
L = ANALYSIS
Func 0
Oper Oper
Func
2
1 2
Labor
12
Labor
21
I = JOHN II = PHENOMENON
Figure 18-2
Support Verbs and Their DSynt-Relationships with their Keyword

—Chapter 18. Lexical Functions— 56
Fig. 18-2 presents the nominal lexeme ANALYSIS with two the DSynt-actants: I = JOHN and
II = PHENOMENON; the whole means (John analyzes the phenomenon). The arrows represent the
LF values, i.e., the support verbs in question; the arrow’s tail indicates DSyntA I (= SSynt-
Subject) of the corresponding support verb, the head pointing to its DSynt-actant II (= Main
Object). Thus:
Oper 1(analysis) = carry out [John carries out the analysis of the phenomenon]
Oper 2(analysis) = undergo [The phenomenon underwent (careful) analysis (by John)]
Func 1(analysis) = — (it could be something like *The analysis of this phenomenon is due
to/stems from John]
Func 2(analysis) = covers, concerns [John’s analysis concerns this phenomenon]
Labor 12(analysis) = submit [John submits this phenomenon to a (careful) analysis]
Labor 21(analysis) = — (it could be something like *The phenomenon prompts John to do its ana-
lysis)
Func 0(analysis) = takes place, occurs [John’s analysis of the phenomenon is taking place]
Oper 0(analysis) = — (it could be something like One sees an analysis of the phenomenon by
John, with an ‘impersonal’ expression one sees)
A different way to express the same idea is by using the following table:
DSynt-role with respect
to the support
verb V LF
Support verb V LF
DSynt-actant I of
V LF is:
DSynt-actant II of
V LF is:
Oper0/1/2 dummy/Ist/II
DSyntA of L
L
Func0/1/2 L
none/Ist/IInd
DSyntA of L
Labor12/21 Ist/IInd DSyntA of L IInd/Ist DSyntA of L
Figure 18-3: Definitions of the Support Verbs
DSynt-actant III/IV of
V LF is:
________
________
From the diagram of Fig. 18-2 and the table of Fig. 18-3 it is easy to see why the support
verbs are presented as three LFs: these LFs are distinguished according to their syntactic behavior
with respect to the major sentence elements, and there are just three such elements: SSynt-
Subject, Main (roughly, Direct) Object and Second (roughly, Indirect or Prepositional) Object.
The phrases formed with a support verb have always attracted the attention of linguists and
have been relatively well studied: see, e.g., Polenz 1963 (German), Deribas 1975/1983 (Rus-
sian), 16 Giry-Schneider 1978 (French), Cattell 1984 (English), and Günther & Förster 1987 (Ger-
L

—Chapter 18. Lexical Functions— 57
man-Russian). Now the reader has a detailed account of support verbs based on Spanish (Alonso
Ramos 2004).
Let us now return to the survey of verbal LFs.
46-48. Phasal Verbs
The LFs 46-48 represent the meaning of what are often called phasal verbs:
Incep [Lat. incipere] carries the meaning (begin),
Cont [Lat. continuāre] means (continue),
and Fin [Lat. finīre] means (cease).
They are semantically full and linked by obvious semantic relations:
Fin(P) = Incep(NonP) [(He ceased to sleep) = (He began not to sleep)];
Cont(P) = NonFin(P) = NonIncep(NonP) [(He continues to sleep) = (He does not cease to sleep)].
Once again, one can easily see the reason for a tripartite organization of these LFs (the initial
Incep plus two possible negative ‘derivations’).
The three phasal LFs are applicable to verbs or verbal expressions only. Therefore, to apply
one of them to a noun, even a predicate noun, a support verb is needed: (begin negotiations) must
be written as IncepOper 1 (negotiations), etc.; as a result, they are most often used in combina-
tion with other LFs, i.e. in complex LFs, see below, 3.1, p. 00. A phasal verb takes L as its
DSyntA II: Incep−II→L, etc.
Examples
IncepOper 1(fireN [shoot]) = open [~ on NY]
IncepOper 2(power) = fall [under the ~→of NX]
FinOper 1(power) = lose [one’s ~→over NY]
ContOper 1(power) = retain [one’s ~→over NY]
[The expression ‘~→NP’ in the reduced Government Pattern means that the NP in question
syntactically depends on the key lexeme L and not on the support verb.]
ContFunc 0(offer) = stands
ContFunc 0(smellN) = lingers [Locin N], lingers on [here, we see an obligatory circumstan-
tial (locative or temporal), which is by no means a DSyntA of SMELL]
However, in principle, these LFs can be used as well outside of complex LFs, i.e., directly
with verbal keywords, if the language allows for this: Rus. Incep(goret´ (burn [intrans.])) =
//za+goret´+sja, Incep(plakat´ (cry)) = //za+plakat´, Incep(spat´ (sleep)) = //za+snut´,

—Chapter 18. Lexical Functions— 58
Incep(ljubit´ (love)) = //po+ljubit´, ... [‘//’ indicates, as was said several times, fused elements of
an LF value]. Compare as well: Incep(burn [intrans.]) ≈ //catch fire; flare up (with addition of
(intensely) = Magn, i.e., we have here [Magn + Incep](burn)); or else Fin(sleepV) = //awakeV.
49-51. Causation verbs
The LFs 49-51 express the meaning of causation: 17
Caus [Lat. causāre] means (cause) [≈ (do something so that a situation begins occurring)]
Perm [Lat. permittere] means (permit/allow) [≈ (do nothing which would cause that a situation
stops occurring)]
Liqu [Lat. *liquidāre] means (liquidate) [≈ (do something so that a situation stops occurring)]
Thus, these LFs are semantically full.
negation:
Very much like phasal LFs, these LFs are also linked by semantic relations based on
Liqu(P) = AntiCaus(P) = Caus(NonP)
Perm(P) = NonLiqu(P) = NonCaus(NonP)
And also like phasal verbal LFs, the causation LFs are often used in combination with other
verbal LFs.
Examples
CausOper 1(opinion) = lead [NX to ART ~]
PermFunc 0(aggression) = condone [ART ~]
CausFunc 0(crisis) = bring about [ART ~]
LiquFunc 0(aggression) = stop [ART ~], put an end [to ART ~]
LiquFunc 0(traces) = wipe out [ART ~]
CausFunc 1(hopeN) = raise [~ in NX]
Yet they can also be found (depending on the language) outside of complex LFs: thus, in
Russian we have:
Caus(pit´ (drink)) = //poit´ Caus(serdit´sja (be angry)) = //serdit´
Caus(spat´ (sleep)) = //usypit´ Caus(katit´sja (roll [intrans.])) = //katit´
Remarks
1. Causation LFs and the actantial structure of complex LFs with their participation. Unlike all the
other LFs, which never change the actantial structure of the LU L referring to a given situation, a
causation LF introduces, as a general rule, a new SemA and, as a consequence, a new DSyntA:
the Causer/the Cause. The Causer/the Cause is expressed as DSyntA I of the causation LF, and

—Chapter 18. Lexical Functions— 59
the inherent DSyntAs of L are all shifted, or permuted, with respect to the initial SemAs. One can
easily see this in the most trivial example of the regular causative construction in French, where
the inherent DSyntA I becomes the DSyntA III of the causative verbal expression:
(4) Jean [= I] mange la soupe [= II]
John eats the soup
vs.
La mère [= I] fait manger la soupe [= II] à Jean [= III].
The mother makes eat the soup to John
The permutation of DSyntAs entailed by a causation LF is represented by means of the
empty verbal LFs, i.e., the support verbs Oper i, Func i and Labor ij. (In other words, we use com-
plex LFs of the type CausOper i, LiquFunc i, etc.) In order to show more clearly the systematic
character of this permutation, I will analyze an example in detail: the French noun ENVIE (desire
caused by a necessity). ENVIE has two DSyntAs: I is the Experiencer and II, the Object [of the
desire], as in (5):
(5) Pierre [I] a [Oper 1] ENVIE d’y aller [II].
Peter has desire to go there
Sentence (5) can be embedded in a causative expression, as, e.g., in (6):
(6) a. Cette aventure a privé Pierre de l’ENVIE d’y aller.
This adventure has deprived Peter of the desire to go there
b. Cette aventure a ôté à Pierre l’ENVIE d’y aller.
This adventure has taken from Peter the desire to go there
These two sentences are equivalent in their propositional content, and the verbs PRIVER (deprive)
et ÔTER (take from) are of course values of LFs of ENVIE. Now, what should be their symbolic
description in terms of LFs?
Semantically, both verbs mean the same thing: (cause the non-existence [of Peter’s desire to
go there]); this meaning has to be described by the LF Liqu.
Syntactically, however, the two verbs differ by their government; to express this difference,
we need the support verb LFs—Oper i, Func i or Labor ij. More precisely, priver Pierre de l’envie
d’y aller is interpreted as (cause that Peter has no more desire); since a [= (has), AVOIR] is Oper 1
of ENVIE, priver is written as LiquOper 1(envie). In its turn, ôter à Pierre l’envie d’y aller is
interpreted as (cause that the desire is no more with Peter); est à [= (is with), ÊTRE à] is Func 1 of
ENVIE, 18 so that ôter is written as LiquFunc 1(envie). Similarly, in the collocation plunge Susan
into a rage, the verb PLUNGE is CausOper 1(rage), while the verb OPEN in open a new perspective
for him is CausFunc 1(perspective). Here are a few more examples:

CausOper 1(despair) = reduce [NX to ~], throw [NX into ~]
CausFunc 2(difficulty) = create, present [ART ~ for NY]
LiquFunc 0(assembly) = dissolve [ART ~]
Caus 1Func 2(supremacy) = maintain [NX’s ~ over NY]
LiquFunc 2(attention) = divert [NX’s ~ from NY]
Caus 2Func 2(attention) = attract, draw [NX’s ~]
Perm 1Fact 0(angerN) = let go, let out [NX’s ~]
Compare the following three-way contrast:
CausOper 2(controlN) = put [NY under NX’s ~]
vs.
Caus 1Oper 2(controlN) = bring [NY under NX’s (own) ~]
vs.
IncepOper 2(controlN) = get [under NX’s ~]
—Chapter 18. Lexical Functions— 60
2. Actantial structure of causation LFs. Causation LFs may have no actantial subscripts: this is the
general case, where an LF of the type Caus introduces an additional actant—new with respect to
the inherent actants of the keyword L. This additional actant is the external Causer/Cause of the
situation (L). Nevertheless, it is also possible that the Causer/Cause of (L) is internal: it is one of
the actants of L. This fact, as we have just seen, must be indicated by the corresponding actantial
subscript. Cf. Caus 2Func 2(attention) and Perm 1Fact 0(anger) above, where the Causer/Cause is at
the same time an actant of the keyword: the person, thing or event that atract/draws someone’s
attention is necessarily the object of this attention; the person who allows his anger to ‘realize’
itself is the Experiencer of the anger.
3. Causation LFs and phasal LFs. For the three causation LFs, the following convention is
adopted. What can be caused is either the beginning (Incep), or the continuation (Cont), or else
the cessation (Fin) of a state or an event P; therefore, Caus, Perm and Liqu should always be
followed by one of the phasal LFs. But as a convenient abbreviation, we omit Incep after Caus
and Cont after Perm and Liqu, so that one can write Caus(P) instead of CausIncep(P), Perm(P)
instead of PermCont(P), and Liqu(P) instead of LiquCont(P): causing the beginning, on the one
hand, and permitting or putting an end to the continuation, on the other, are considered as non-
marked, or default, cases. All the other combinations of a causative LF with a phasal one must be
stated explicitly:
CausCont(P), PermIncep(P), PermFin(P), LiquIncep(P) and LiquFin(P).

NB: CausFin(P) means—by definition—Liqu(P).
52-54. Realization verbs
—Chapter 18. Lexical Functions— 61
The LFs 52-54—Real 0/i, Fact 0/i and Labreal ij, are realization, or fulfillment,
verbs; they mean, very roughly, (fulfill the requirement of L) = (do with L what one is supposed
to do with L) or (L fulfills its own requirement). The ‘requirements’ differ with respect to diffe-
rent Ls: thus the ‘requirement’ of a hypothesis is its confirmation, the ‘requirement’ of a disease
is the malfunctioning/death of the organism affected, and the ‘requirement’ of an artefact is that it
be used according to its intended function. Real 0/i [Lat. realis (real)], Fact 0/i [Lat. factum (fait)]
and Labreal ij
[a hybrid of Labor and Real] are (more or less) synonymous full verbs, different
with respect to their syntax only; their actants are actantial nouns whose meaning includes the
component corresponding to a ‘requirement’: (supposed to ...), (designed to ...), etc.
In sharp contrast to the support verbs, which accept as their keywords basically abstract
nouns, the fulfillment verbs can have both abstract and concrete keywords, provided the latter are
actantial and imply a ‘requirement’. Such concrete nouns are necessarily the names of artefacts or
organs, which are by definition (designed to ...).
Syntactically, Real 0/i, Fact 0/i and Labreal ij are fully analogous to the LFs Oper 0/i,
Func 0/i and Labor ij, respectively. This means that the keyword L and its DSyntAs fulfill with
respect to Real i the same syntactic roles as they do with respect to Oper i, etc. Therefore, they are
linked to their keywords in the following way:
Real 0/i−II→L, Fact 0/i−I→L, and Labreal ij−III→L.
Examples
Real 1(accusation) = prove [ART ~] Real 2(law) = abide [by ART ~]
Real 1(illness) = succumb [to ART ~] Real 2(demand) = meet [ART ~]
Real 1(car) = drive [ART ~] Real 2(hint) = take [ART ~]
Real 1(bus) = drive [ART ~] Real 2(bus) = ride [(on) ART ~] 19
Compare:
Oper 1(obstacle) = face [ART ~], but Real 1(obstacle) = //turn back
Oper 2(attack) = be [under ART ~], but Real 2(attack) = fall [to ART ~ →of NX]
Oper 2(exam) = take [ART ~], sit [in ART ~], but Real 2(exam) = pass [ART ~]
Fact 0(hope N ) = comes true Fact 0(film N ) = is playing, is on
Rus. Fact 0(nadežda (hopeN)) = opravdyvaetsja Fact 0(podozrenie (suspicion)) = podtverž-

—Chapter 18. Lexical Functions— 62
lit. (justifies itself) daetsja lit. (confirms itself)
Rus. Fact 0(mečta (dream)) = sbyvaetsja lit. (realizes itself)
ContFact 0(luck) = holds Rus. ContFact 1(udača (luck)) = ne pokidaet [NXacc]
lit. (does not abandon)
Labreal 12(saw) = cut [NY with ART ~] //saw [NY]
Labreal 12(gallows) = string up [NY on ART ~], hang [NY from ART ~]
Labreal 12(asphalt) = cover [NY with ~] //asphalt [NY]
The LFs 55-64 express new situations related to the situation referred to by the keyword L.
55. Involv [Lat. involvere (drag along)] = involvement verb
L´ = Involv(L) is a verb meaning (involve Y), (concern Y); it links L and the name of a
non-participant Y which is affected or acted upon by the situation (L); Y is DSyntA II of Involv,
and L (= the keyword), its DSyntA I.
Examples
Involv(light) = floods [NY, e.g. the room]
Involv(snowstorm) = catches [Nhum-Y Locin N; e.g. … caught John in the woods],
hits [N area -Y; e.g., Then the snowstorm hit!, where hit = IncepFunc 0 ]
Involv(explosion) = rocks [NY, e.g. the building]
Involv(river) = floods [NY, e.g. the area]
AntiBon 1Involv(car) = smashes [into NY, e.g. a telephone pole]
AntiBon 2Involv(car) = hits, runs over, strikes [NY, e.g. a pedestrian]
56. Manif [Lat. manifestāre (manifest)] = manifestation verb
Manif is a verb meaning (L manifests itself [≈ becomes apparent] in Y). The keyword L, a
noun, is DSyntA I of Manif, and Y (= in which L manifests itself), its DSyntA II.
Examples
Manif(distrust) = lurks [in Andrew’s eyes]
Manif(joy) = explodes [in them]
Manif(scorn) = is dripping [from every word]
Caus 1Manif(gratitude) = express, show [NX’s ~]
Caus 1Manif(relief) = heave a sigh [of ~]

—Chapter 18. Lexical Functions— 63
The two following LFs—Prox and Prepar—have no actantial structure of their own and
therefore with a nominal L they appear only with LFs of the type Oper i or Real i, i.e., they are
always used in complex LFs; with a verbal L they can appear alone. Their keyword L is a noun or
a verb with an predicate meaning.
57. Prox [Lat. proximāre (approach)] = proximative
Examples
Prox is a verb meaning (be about to …) ≈ (be on the verge of …).
Prox(die) = //be clinging to life, be fighting for life
ProxOper 1(despair) = be on the edge [of ~]
ProxOper 1(disaster) = head [for ~] < be on the brink [of ~]
ProxOper 1(tears) = be on the verge [of ~]
ProxFunc 0(storm) = brews, gathers
58. Prepar [Lat. praeparāre] = preparative
Prepar is a verb meaning (prepare N for ...) ≈ (get N ready for normal use or functioning);
Prepar1 is used to express the idea of (prepare oneself for ...).
Examples
PreparFact 0(car) = fill up [ART ~]
PreparFact 0(ears) = prick up [ART ~s]
PreparFact 2(program) = load [ART ~ into a computer]
PreparFact 2(ambush) = set [ART ~ for NY]
Prepar 1Real 1(ambush) = lay [in ~ for NY]
Prepar 1Real 2(plane) = board [ART ~]
Remark
Prepar and Prox stand in an obvious semantic relation:
Prepar(Y) ≈ consciously CausProx(Y).
That is, Prepar implies teleological, purposeful activity; that is why it was decided to keep it as a
separate LF.

—Chapter 18. Lexical Functions— 64
59. Degrad [Lat. degradāre (lower, degrade)] = degradative
Degrad is a verb meaning (degrade) ≈ (become permanently worse or bad). It takes its
keyword L, which can be any noun that designates something that can become worse, as its
DSyntA I: Degrad−I→L.
Examples
Degrad(milk) = goes 〈turns〉 sour, turns Degrad(discipline) = weakens
Degrad(teeth) = decay Degrad(house) = becomes dilapidated
Degrad(clothes) = wear out Degrad(patience) = wears thin
Degrad(temper) = frays Degrad(health) = deteriorates
Remark
Degrad can be represented as IncepPredMinusBon (= IncepPredPejor) or IncepPred-
MinusVer; however, this formulation is considered to be too complicated.
60. Son [Lat. sonāre ([to] sound)] = sound verb
Son is a verb meaning (emit characteristic sound). It also takes its keyword L, which most
often, but not necessarily, is a concrete noun, as its DSyntA I: Son−I→L.
Examples
Son(dog) = barks Son(banknotes) = rustle Son(waterfall) = roars
Son(teeth) = chatter Son(whipN) = cracks Son(elephant) = trumpets
Son(thunder) = rumbles Son(wind) = howls Son(hurricane) = roars
Caus 1Son(fingers) = snap [NX’s ~] Caus 1Son(tongue) = click [NX’s ~]
The three following LFs—Obstr, Stop and Excess—all denote temporal change in
functioning of something; as their keyword, they take a concrete noun denoting a functioning
being or object (= an organ or artefact) or else an event that is taking place. As for their syntax,
they take, by default, their keyword L as their DSyntA I. If, however, their DSyntA I must denote
the person involved (= DSyntA I of L), this must be indicated by the subscript ‘ 2’. (Thus, Obstr 2
is a conversive of Obstr.)
61. Obstr [Lat. obstruere (obstruct)] = obstructive
Obstr is a verb meaning (function with difficulty) = (function being obstructed).
Examples
Obstr 2(breath) = [NY] is short [of ~] Obstr(eyes) = do not see

—Chapter 18. Lexical Functions— 65
Obstr 2(speech) = //[NY] stutters, stammers, Obstr(saliva) = //[NY’s] mouth is dry, [NY’s]
mumbles tongue sticks to the roof of [NY’s] mouth
Obstr(negotiations) = are stalled Obstr(economy) = stagnates
62. Stop [Lat. *stuppāre (stop up, plug)] = cessative
Stop is a verb meaning (stop functioning).
Examples
Stop 2(breath) = [NY] loses [NY’s ~] Stop(voice) = breaks
Stop(heart1) = stops Stop(heart2) = breaks
63. Excess [Lat. *excessus] = excessive
Excess is a verb meaning (function in an excessive way).
Examples
Excess 2 (heart1) = [NX] has palpitations Excess(engine) = races
Excess(sweat) = rolls down across [NX’s] Excess motor
2
forehead, trickles down the back of
(teeth) = [NX] grinds [NX’s ~]
[NX’s] head, pours off [NX’s] brow Excess motor
2 (sweat) = [NX] is bathed [in ~]
[For the superscript motor , see below, 5.1.3.]
64. Sympt [Lat. *symptoma] = symptom
L´ = Sympt(L) is a verbal expression meaning (symptom of); more precisely, Sympt denotes
a bodily reaction that is a symptom of an emotional or physical state L. Sympt takes three
DSyntAs, the actantial subscripts being distributed as follows:
itself;
—the subscript ‘ 1’ corresponds to the body part in which the symptom in question manifests
—the subscript ‘ 2’ corresponds to the person who ‘owns’ the body part involved and who is
the Experiencer of the emotion or physical state L;
—and the subscript ‘ 3’, to the emotional or physical state L.
In accordance with our technique of expressing syntactic conversion (= the permutation of
DSynt-Actants), the order of actantial subscripts specifies the SSynt-role of the actants: the sub-
script placed as first corresponds to the SSynt-Subject of the verbal expression which is the value
of Sympt; the second subscript corresponds to its Main Object; and the last subscript, to its
Secondary Object. The LF Sympt is used, as a general rule, together with Obstr, Stop and
Excess (as well as with some non-standard LFs), forming with one of them the configurations of
the form F—Sympt, where Sympt specifies that state (L) whose symptoms interest us, and the LF

—Chapter 18. Lexical Functions— 66
F indicates the body part or body function affected by (L) as well as the way it is affected (func-
tions with difficulty, stops functioning or functions excessively).
Examples
Obstr(speech) —Sympt 23(anger) = [NY] sputters, splutters [with anger]
Obstr(breath) —Sympt 23(anger) = [NY] chokes [with anger]
Stop(speech) —Sympt 2(amazement) = //[NY] is dumbstruck
Excess motor (eyes) —Sympt1(amazement) = //[NY’s] eyes start from their sockets
Excess motor (mouth) —Sympt 213(amazement) = [NY] opens [NY’s] mouth wide [with amazement]
Excess motor (mouth) —Sympt 13(astonishment) = [NY’s] jaw drops [in astonishment]
Excess motor (hair) —Sympt 1(fear) = [NY’s] hair stands on end [with fear]
4. Special Phenomena Related to Lexical Functions
Three lexical phenomena related to LFs and their presentation in an ECD are relevant to
this presentation: complex LFs, configurations of LFs, and fused elements of
the values of LFs.
4.1. Complex Lexical Functions
A complex LF is a combination of syntactically related LFs that has a single expression
covering the meaning of the combination as a whole. A complex LF is written as a sequence of
simple standard LFs, e.g., fg or fgh, etc. In point of fact, such sequences encode particular
DSynt-relations between constituent LFs; the corresponding relation is specified for each simple
LF (see the list in 2.2). Thus, generally speaking, fg(L) = L´ means
• either f(L)–r 1 →g(L) = L´, such that (f(L)−r 1 →g(L)) = (L´),
• or f(L)←r 2 –g(L) = L´, such that (f(L)←r 2 −g(L)) = (L´);
which is the case is indicated in the description of f and/or g.
More precisely, the internal syntax of a complex LF—except for Anti, Conv and structural
derivations (S i, A i and Adv i) as left members of complex LFs—is specified by straightforward
rules valid for each simple LF (F stands for any LF):
MagnF = Magn←ATTR−F; IncepF = Incep−II→F; CausF = Caus−II→F; etc.
However, Anti, Conv, Si, Ai and Advi are used within complex LFs as derivational affixes; thus
S i, A i and Adv i, added on the left of an LF-formula, convert it into an N, an A or an Adv,

—Chapter 18. Lexical Functions— 67
respectively; e.g., Adv 1IncepReal 1(L) is a DSynt-adverb meaning (while beginning to do with L
what L is designed for) [ (while ...-ing) is the ‘meaning’ of Adv 1].
I have already given several examples of complex LFs; let me quote here some more
Complex LFs of the same type (with the keyword in small capitals):
AntiMagn : scattered APPLAUSE, weak ARGUMENTS, low TEMPERATURE, negligible LOSSES, ...
AntiVer : false SHAME, a wrong CONCEPTION, unfounded SUSPICION, ...
IncepOper 1
CausOper 2
Caus 1Oper 2
AntiReal 2
: acquire POPULARITY, sink into DESPAIR, embark upon the path of TREASON, ...
: put [NY] under [NX’s] CONTROL, plunge [NX] into SLAVERY, ...
: bring [NY] under CONTROL, put [NY] up for SALE, force [NY] into EXILE, ...
: fail an EXAMINATION, reject a PIECE OF ADVICE, turn down an APPLICATION, ...
Note the crucial difference between a complex LF fg(X) and a composition (in the
mathematical sense) of LFs f(g(X)): generally speaking, fg(X) ≠ f(g(X)). Thus:
Magn(applause) = thunderous, AntiMagn(applause) = scattered,
but Anti(thunderous) ≠ scattered;
Oper 1(despair) = be [in ~], IncepOper 1(despair) = sink [into ~],
but Incep(be [in]) ≠ sink; etc.
A complex LF can also combine standard and non-standard LFs; such an LF is called
mixed. For instance:
after FinOper 1, again IncepOper 1(consciousness) = regain [~]
to attempt to CausFunc 0(elections) = call [ART ~]
being too many, IncepOper 1(market) = flood [ART ~]
For more examples, see 5.1.1, p. 00.
4.2. Configurations of Lexical Functions
A configuration of LFs is a combination of syntactically unrelated simple or complex
LFs which have the same keyword such that there exists a single lexical expression covering the
meaning of the combination as a whole. For example:
[Magn + Oper 1](laughter) = roar [with ~],
where roar means ≈ (do [= Oper 1] big [= Magn] [laughter])
[AntiMagn + A 2Manif](irony) = tinged [with ~],
where tinged means ≈ (in which manifests itself [= A 2Manif] small [= AntiMagn] [irony])
[tooMagn quant + IncepOper 1](market) = flood [ART ~],

—Chapter 18. Lexical Functions— 68
where flood means ≈ (too many [= tooMagn quant ] begin to be on [= IncepOper 1] [market])
In a configuration of LFs, the syntactically central LF, which determines the part of speech
of the whole configuration and of its value L´, is written rightmost; the configuration is enclosed
in square brackets, while its components are separated by ‘+.’ The LFs that make up an LF confi-
guration can be both standard and non-standard.
4.3 Fused Elements of Values of Lexical Functions
A fused element of the value f(L) of an LF f, or a fused expression of f, is a lexical
unit L´ that formally does not include the keyword L itself but semantically covers both the
meaning of f and that of L, so that (L´) = (L + f). As stated earlier, a fused value element is
indicated by a double slash ‘//,’ which separates the non-fused elements (on its left) from the
fused ones (on its right). For example:
Remarks
Magn(rainN) = heavy //downpour [downpour ≈ (heavy rain)]
Magn(laughV) = uproariously //split one’s sides
[split one’s sides ≈ (laugh uproariously)]
AntiBon(car) = //lemon [lemon ≈ (bad car)]
1. The empty LFs cannot have fused expressions. Since an empty LF, that is, a support (=
‘light’) verb, does not have a meaning of its own, there is nothing to be fused with the meaning of
the keyword. Therefore, the expression of the form ‘//Oper i(L)’ is considered absurd; V 0(L)
should be used instead.
2. The scope of the ‘//’ symbol reaches up to the symbols ‘

—Chapter 18. Lexical Functions— 69
• proliferation of common elements in the values of f for different keywords, which leads to
the necessity of generalizing them.
Let us take these two problems in turn.
5.1. Elements of the Value f(L) as Subentries within L’s Entry
Logically speaking, an element L´ in the value of the LF f(L) can itself be an LU of lan-
guage L and have a lexical entry of its own, where its properties are fully specified. Yet it is also
quite possible, and it actually happens more often than not, that as (an element of) the value of a
particular LF f(L1), this L´ possesses some properties concerning, e.g., its Government Pattern [=
GP] or even its inflection, which it does not possess as an element of the value of f(L2) or g(L1)
or else as a ‘free’ LU. The observable peculiarities of L´ are phraseological; they do not, strictly
speaking, accrue to L´ as such, but are due to its being part of the collocation L´ + L. These pecu-
liarities must be indicated in the lexical entry of L—where L´ itself is specified. This means that
an element L´ of the value f(L) of the LF f given in the entry of f’s keyword L can form an
embedded subentry which carries whatever is individual and idiosyncratic information for the
particular collocation; these data override the indications found in L´’s own entry (where we have
to look for the information which is not peculiar to the collocation in question—the phonological
form, morphological data, etc.).
The information specified in the subentry of an element of an LF f(L) value concerns those
of its syntactic properties that are specific to the given collocation; let me briefly review them. An
LF-value element subentry, i.e., the data accompanying the element L´ ∈ f(L), may contain the
following types of linguistic information:
1) A reduced GP of L´. This reduced GP specifies the form of SSynt-Actants of L´: governed
prepositions and conjunctions, grammatical cases, and the like, their obligatoriness, their cooccur-
rence with each other, etc. Moreover, the reduced GP may be affected by the following complica-
tion: As a general rule, the reduced GP for a given element L´ of the value f(L) specifies the
government of L´ only; L retains its own GP, which, therefore, does not need to be repeated in the
subentry of L´. But in some cases, the GP of L changes in the phrase L´ + L as well, and this fact
has to be shown somehow in the reduced GP of L´. The change in question can be twofold:
• Either some of the expressions of the DSyntAs of L are lost; this is indicated by a special
condition of the type “CII(L) = Λ” [‘DSyntA II of L cannot be expressed’], which accompanies
the subentry of L´.

—Chapter 18. Lexical Functions— 73
7) Indications on morphological peculiarities of L´ (special forms, missing forms—e.g., “| no
plural”, or “| no passive”, ...).
8) Indications on stylistic properties of L´ (informal, slang, literary, ...).
To sum up: In the general case, if the LU L´ appearing as an element of the value of an LF
differs from the free LU L´ in its GP or in some other details, it does not necessarily have a full-
fledged entry of its own but may be described by a subentry within the entry of its keyword L.
(L´ receives its own entry only if it appears with exactly the same properties in several lexical
entries as an element of the value of the same LF.)
5.2. Generalizing over the Values of Lexical Functions
In spite of the basically idiosyncratic character of LFs, in many cases a given LF f may
have the same values for quite a few different keywords: you take a BATH, a SHOWER, and a
REST; you are in DESPAIR, LOVE, and a RAGE; you follow the RULES, PRESCRIPTIONS, RECOM-
MENDATIONS, and INSTRUCTIONS. As is clear, the reason is mainly semantic proximity: seman-
tically related LUs tend to feature the same values for a given LF. This fact can be accounted for
by the Principle of lexical inheritance (Mel’čuk & Wanner 1996):
Principle of Lexical Inheritance
All lexicographic data systematically shared by a family of semantically related LUs should
be stored just once—under one LU of the corresponding vocable or under the generic LU
of the corresponding semantic field; these data are inherited by the LUs involved.
In the domain of LFs, this principle applies in two major ways. Namely, the elements of the
values of the LF f shared by several keyword LUs can be generalized by extracting them from in-
dividual keyword entries and storing them together just once—either on the keyword side (for a
group of LUs) or on the LF side (attaching them to the LF). Of course, in practice both ways can
be combined in different proportions. Each of these approaches to LF value generalization has, in
turn, two further versions, so that we have, all in all, four techniques for Lexical Inheritance. To
wit, the shared LF values can be grouped:
Generalizing on the keyword side
1. under a convenient LU (this can be the basic lexeme of the same vocable or a lexeme of a
different vocable); or
2. under the generic LU of the corresponding semantic field.
Generalizing on the LF side
3. under the LF itself; or

—Chapter 18. Lexical Functions— 74
4. under the vocable to which the given element of an LF value belongs.
Let us consider these cases in turn.
1. Shared Values of Lexical Functions and Cross-references to a Convenient Lexical Unit of a Vocable
If the LU L1 of the vocable {L} has the same values of some LFs as L2, another LU of {L}
or of a different vocable {L´}, these values are given only under L2, with a cross-reference under
L1: ↑L2, i.e., ‘see L2’. For instance, under ESCAPEN, we find in the LF zone:
LFs S 1Perf, S 1Able 1, S usual
2, S 3, A 1: ↑ESCAPEV,
which means (the values for the LFs S 1Perf, S 1Able 1, S usual
2, S 3 and A 1 for the LU ESCAPEN are
identical to the values for the same LFs for the LU ESCAPEV).
This is a rather straightforward technical generalization.
2. Shared Values of Lexical Functions and the Generic Lexical Unit of the Semantic Field
LUs belonging to the same semantic field may have the same values for the same LFs. Of
course this is only a tendency, since there are numerous exceptions and deviations, 21 yet in some
cases this tendency is quite clear. If shared values of LFs are semantically determined—that is, if
the keywords belong to the same semantic field, the generalization can be achieved without expli-
cit cross-references. It suffices to store all common values of LFs supplied with necessary indica-
tions only once—under the generic lexical unit of the corresponding semantic field.
The methodology (proposed in Mel’čuk & Wanner 1996) runs as follows:
1) Each semantic field is identified by its generic LU—LGENER. For instance, for the field CAR,
TRUCK, BUS, CARAVAN, MOTORBIKE, ... the LGENER is MOTOR VEHICLE; this LGENER appears in the
definitions of all the LUs of the field.
2) The LGENER of a semantic field can have, in addition to its own ‘private’ entry, another one,
called ‘public’, where the shared elements of the values of the LFs associated with the indivi-
dual members of the field are stored. They are not repeated in individual entries, which thus
contain only those LF value elements which are idiosyncratic to particular LUs.
3) When the user needs the value of a particular LF f as applied to a particular LU L, he is sup-
posed to look for it in the entry of L. If, however, the value f(L) is not given in L’s entry or if it is
given only partially, it has to be looked up in the public entry of the LGENER of the corresponding
field (LGENER is specified in L’s definition). Thus, there is no need to repeat Real 1 = drive and
Real 2
= ride for every particular name of a motor vehicle: both verbs can be indicated only once,
in the ‘public’ entry of MOTOR VEHICLE. From there, they are inherited by all LUs of this field.

—Chapter 18. Lexical Functions— 75
The noun CANCER has Oper 1 = suffer [from ~], Real 1 = succumb [to ~], AntiReal 1 = defeat
[NX’s ~], ... But so do the names of (theoretically) all serious diseases: John suffered from tuber-
culosis, succumbed to AIDS, defeated his pneumonia. Therefore, these LFs should not be repeat-
ed in the entries of individual diseases’ names, but concentrated in the ‘public’ entry of the noun
DISEASE, which is of course the LGENER of the semantic field of diseases (perhaps these common
LFs should be supplied with the constraint “for grave diseases”).
In a Russian ECD, with emotion nouns, Func 1 = oxvatyvat´ (seize) can be given only once
—under ÈMOCIJA (emotion), which is the LGENER for all emotion nouns, with the following
semantic condition (L is the name of the eventual keyword):
oxvatyvat´ | (L) ⊃ (intense) and (L) (caused only by intellectual evaluation).
In addition, we have to mark individually all emotion nouns that satisfy the above semantic
condition and yet do not accept oxvatyvat´. If such lexical exceptions are marked, Func 1 =
oxvatyvat´ is good for any emotion noun which is inherently or actually (intense), but does not
have in its definition the component (caused only by intellectual evaluation): STRAX (fear), PANI-
KA (panic), GNEV (anger), GRUST´ (sadness), DOSADA (annoyance) can all oxvatyvat´, but not, e.g.,
VOSXIŠČENIE (rapture) or (krajnee) IZUMLENIE ((extreme) amazement), see Apresjan 1992: 26-
27. Note that OXVATYVAT´ does not combine with the lexeme ÈMOCIJA itself, although it does
with its near-synonym ČUVSTVO; it does not combine with LJUBOV´ (love) and, for many speak-
ers (but not for me), with NENAVIST´ (hatred) or OTVRAŠČENIE (revulsion), either. All these are
lexical exceptions which have to be marked in the corresponding lexical entries.
The technique proposed makes up part of what is known as Lexical Inheritance methodo-
logy. The idea itself seems to be fruitful (cf. Borillo 1992: 29-31); but for many particular seman-
tic fields it may give only limited results due to a very high degree of idiosyncrasy.
A rather straightforward particular case of this type of generalization is the use of LFs with
proper names. If the keyword of an LF f happens to be a proper noun Nprop, the value of f must be
computed substituting for Nprop the LU L that refers to the entity called by Nprop. In other words,
the values of LFs for Rhine, Ganges, Volga, Thames, Mississipi, etc. are those of the LFs for
RIVER (or STREAM):
The Rhine flows from Switzerland to North Sea.
The Ganges empties into the Bay of Bengal.
The Oka is a tributary of Volga.
The Thames was swollen and flooding the riverbanks.

—Chapter 18. Lexical Functions— 76
In much the same way, any particular ship name accepts the same semantic derivations and
collocates as the noun SHIP:
The “Saratoga” sails tonight.
The “Shawinigan” dropped anchor at Malta.
The BISMARCK was sent to the bottom by Royal Navy shellfire the next day.
The ‘Satu Maru’ was torpedoed and went down with all her crew.
All of them navigate/sail, can be scuttled, call/put to port, heave, pitch and roll, have a
hull, decks, a bow and a prow, holds, etc.
3. Shared Values of Lexical Functions as Universal Elements of LF Values
The level of generality of elements of LF values might be such that it becomes feasible to
associate with a given LF some universal, or default, elements of its value—specifying of
course the semantic conditions under which the restricted lexical cooccurrence will be observed.
Thus, we may indicate that universal elements of the Oper 1 value are MAKE, GIVE, HAVE, CARRY
OUT and maybe some more; for each of those the semantic restrictions have to be specified
(something along the lines of Wierzbicka 1982). For Func 0, universal value elements are, among
others, kTAKE PLACEl, HAPPEN, BE HELD [speaking of a conference], etc.; see a study of semantic
distribution of the corresponding verbs in French in Gaatone 1992. For Magn, such universal
elements are (the keyword being a noun) BIG, HIGH (for non-linear physical parameters such as
PRESSURE, SPEED, TEMPERATURE), UTMOST/UTTER (for strong feelings), etc., or (the keyword
being an adjective) VERY. Such an approach presupposes the existence of special lexical entries
for LFs.
4. Shared Values of Lexical Functions as Separate Lexical Entries
Another possible avenue for a sufficiently general description of shared elements of LF
values is compiling for them individual lexical entries, where all semantic and syntactic condi-
tions on their use could be specified. Thus, the verb GIVE as an element of the value of the LF
Oper 1 of many action nouns (give a talk/a look/a kiss) receives or several lexical entries of its
own in the vocable GIVE; accordingly, it should not be necessarily specified in all keyword
entries but could be selected and used following its own entry.
Note that determining whether a given L´1 as an element of f(L1) and a homophonous L´2 as
an element of f(L2) represent one LU or two different ones is far form simple (cf. Alonso Ramos
1995, especially p. 104ff)

—Chapter 18. Lexical Functions— 77
NB: What has just been said should by no means be construed as a claim to the effect that it is always or even
fairly often possible to state the semantic conditions for the distribution of LF shared values. Quite on the
contrary, I think that, as a general rule, such description is unfeasible—because of a very high degree of
idiosyncrasy in lexical cooccurrence. However, some regularities or tendencies exist beyond any doubt, and
it is worth the researcher’s effort to capture them.
5.3. Ellipsis of the keyword of an LF
In texts, sentences are often seen in which the keyword of an LF has been elided, while the
LF’s expression, i.e., an element of its value, is present. For instance, sentence (12) is quite clear:
(12) 1.10$ a liter and climbing [ a newspaper title].
It is about the donate blood, gasoline price; but how is the verb CLIMB selected if the noun PRICE
is not in the sentence? To produce (12), its DSyntS must contain the LU PRICE, which will
determine the correct choice of the element for the LF IncepPredPlus:
IncepPredPlus
I
PRICE
!
CLIMB
subjectival
PRICE
Only when the proper element of the value of the LF is in the SSyntS can the ellipsis take place.
As one sees, such ellipses do not create any additional difficultes for the use of LFs.
6 Linguistic Nature of Lexical Functions
To characterize the linguistic nature of Lexical Functions in a more profound way, I will
discuss now the semantic aspect of LFs, their phraseological aspect, their place in utterance
representations, and their universality.
6.1 Semantic Aspect of Lexical Functions
Under this heading, five points have to be made:
• an LF is by no means a semantic unit;
• an LF can have several semantic subtypes;
• an LF can return semantically different sets of elements of its value contingent on different
semantic ‘facets’ of the keyword;
• some syntagmatic LFs have to be related to some components in the definitions of their
keyword;
• LFs are Deep Lexical Units.

6.1.1 Lexical Functions Are Not Semantic Units
fact:
—Chapter 18. Lexical Functions— 78
To avoid misunderstanding, I have to insist—for the umpteenth time—on the following
Lexical Functions are by no means semantic entities; they are not precise enough semantic-
ally, so that when using them, a linguist should not pursue semantic precision.
LFs are Deep Lexical Units, and are treated as such. Their ‘duty’ is simply to point to the
family of LUs that in principle may fulfill the same semantic-syntactic role with respect to a
given LU L; the actual choice of a specific element within the value f(L) = {L´i} of the LF f is to
be made under the control of the semantic description of each of the elements in this value.
To put it differently, the elements of the value of a given LF f for a given keyword L need
not be fully synonymous: they can even contrast in meaning. The only requirement is that they
share the ‘core’ meaning associated with f.
Suppose that in the ECD entry of the adjective OPPOSED [to N] we have the following:
(13) Magn : adamantly, bitterly, consistently, deeply, resolutely, steadfastly, strongly,
vehemently, vigorously
This should by no means be construed as implying absolute synonymy of all the elements in this
value of Magn. The value Magn(opposed) simply lists the adverbs that could be used to intensify
the meaning of (opposed)—perhaps in quite different respects. All these adverbs have their own
entries in the ECD; in order to choose what is actually needed in a particular case in the process
of lexicalization their definitions should be consulted. To facilitate the work of the user, some
semantic distinguishers could be put directly into the value of Magn: adamantly [(unyieldingly)],
bitterly [(exhibiting or proceeding from animosity towards Y)], consistently [(continuingly, at all
times)], etc.
The elements of the value of an LF from a given LU L which are not sufficiently synony-
mous are separated by semicolon; this indicates that more semantic information is needed in
order to make the right choice, and this information is to be obtained from the lexical entries of
these elements.
There exists a different, but equivalent way of representing the non-synonymy of the ele-
ments in the value f(L). Namely, the semantic ‘surplus’ found in the elements of the value f(L)
can be represented in terms of non-standard LFs—to form together with f a mixed complex LF f´.
Thus, instead of one Magn, we will have in the entry of OPPOSED one Magn and four mixed
complex LFs with Magn as their central element:

OPPOSED [to N]
unyieldingly + Magn : adamantly
exhibiting or proceeding
from animosity to Y + Magn : bitterly
exhibiting or proceeding
from resolution + Magn : resolutely
—Chapter 18. Lexical Functions— 79
Magn : deeply; strongly, vigorously < vehemently
continuingly + Magn : consistently; steadfastly
In the value of Magn, as indicated above, the symbol ‘

—Chapter 18. Lexical Functions— 80
ATTACKII ≈ (verbal aggression), as in Their troops launched a large-scale attack on the liberated
area vs. The committee launched a scathing attack on the Medical Research Council. (The
adjectives devastating and violent can apply to both ATTACKI and ATTACKII; for a more
precise description of the lexical cooccurrence, further senses should be distinguished—thus,
under ATTACKI one has to isolate (military operation) vs. (physical aggression) etc.
— Give a KISS vs. blow a KISS: here we also have two different noun lexemes: KISS1 (X’s ges-
ture of affection toward Y done with X’s lips touching Y’s body ...) vs. KISS2 (X’s gesture of af-
fection toward Y done at a distance: X kisses X’s own fingers and shows them to Y as if X were
throwing X’s kiss1 to Y). Then give = Oper 1(kiss1 N ) and blow = Oper 1(kiss2N). Cf. in German:
einen KUSS (kiss1) geben (give) vs. eine KUSSHAND (kiss2) zuwerfen (throw at).
— Give a LOOK vs. have a LOOK (from Cattell 1984: 84-86):
(15) a. John gave 〈*had at〉 Sue a look, as if he wanted to kill her 〈... *to see what color her
eyes were〉.
b. John had 〈*gave〉 a look at Sue to see what color her eyes were 〈... *as if he wanted to
kill her〉.
c. give [N] 〈*have [at N]〉 a blank look.
Here, too, we have two noun lexemes: LOOK1, which is given to express or communicate some-
thing, and LOOK2, which is had to see or learn something.
6.1.2 Semantic Subtypes of Lexical Functions
Another source of non-synonymy of the elements in the value f(L) is the existence of
semantic subtypes for several LFs. These subtypes are marked by way of superscripts.
• For fulfillment verbs (Real, Fact, and Labreal), numerical Roman superscripts ‘ I ’, ‘ II ’, and
‘ III ’ indicate the degree of realization of the ‘requirement’ in question. Thus, Real I 1(invitation)=
accept, while Labreal II 12(invitation) = take [NY] up [on NY’s invitation]; in point of fact, you have
first to accept an invitation, and then you do whatever you are invited to do (if nothing interferes
of course). Cf. as well Real I
3(suggestion) = accept, while Real II
3(suggestion) = act [on the sug-
gestion]; Real I
3(advice) = accept, take and Real II
3(advice) = follow; etc.
The superscript ‘ I ’ denotes the minimal degree of realization (say, at the mental level only),
the superscript ‘ II ’, the middle degree of realization (the final degree at the mental level or the
initial degree at the physical level), and the superscript ‘ III ’, the maximal degree of realization (at
the physical level).

—Chapter 18. Lexical Functions— 81
The same degree superscripts ( I , II , and III) can also be used for the three qualifying LFs:
Magn, Bon and Ver: Magn II (woundN) = serious vs. Magn III (woundN) = mortal; 22 or else
Magn I/II (overhaulN) = substantial vs. Magn III (overhaulN) = complete; etc.
• For the intensifier Magn, alphabetical superscripts specify, when relevant, the aspect of inten-
sification:
quant = (with respect to quantity) : Magn quant (traffic) = heavy
[heavy traffic means (MANY vehicles)] 23
temp = (with respect to duration) : Magn temp (illness) = long, Magn temp (joy) = enduring
Magn temp (rainN) = steady
The same superscripts can be used of course for all other LFs, when needed.
NB : Actually, such a superscript indicates the meaning which is not part of the definition of the keyword, but on
which the meaning of Magn bears; see for more the end of Subsection 6.1.4, p. 00.
• For the LFs Obstr and Excess, alphabetical superscripts specify the aspect of obstruction
or excessive Functionsing:
color color
= (with respect to color) : Excess (cheeks) = glow
Excess color
1(cheeks) = be red-cheeked
dim = (with respect to dimension/
size) : Excess dim (eyes) = are like saucers (in NX’s head)
fulg = (with respect to brightness) : Excess fulg (eyes) = flash, glitter
motor = (with respect to movements) : Excess motor (eyes) = pop out on stalks
Excess motor (heart1) = pounds, races
stat = (with respect
to the vertical position) : Obstr stat (body) = crumples
Obstr stat (knees) = give way
trem = (with respect to trembling) : Excess trem (hands) = shake
t 0
= (with respect to temperature) : Excess t0
(cheeks) = burn, flame
• For the adjectives Magn, Bon and Ver actantial subscripts may indicate the DSyntA on
which the LF bears: Magn quant
1= general [general panic means (MANY panickers)]
• For various LFs (S i, Oper i, Fact i , ...), alphabetical superscripts actual and usual specify,
respectively, whether the situation denoted is actual (i.e. is referential, can be directly perceived,
has a definite duration, etc.) or usual (i.e., non-referential).

6.1.3 Different Semantic ‘Facets’ of a Keyword
—Chapter 18. Lexical Functions— 82
The elements of the value f(L) can further be non-synonymous because the keyword L can
be considered in different situations or, better, under different facets—and in these different
facets it has different values with a given LF f. Two major classes of cases must be distinguished:
a facet is specified a) by the instantiation of a SemA of L or b) by a particular angle from which
the referent of L is considered.
Facets defined by the value of a SemA
Consider, for example, HEART2 (X’s imaginary organ of emotion Y toward Z ...).
• If the emotion is love, i.e., Y = LOVE, then Fact 3(heart2) = belongs [to NZ].
• If the emotion is joy, then Fact 0(heart2) = sings.
• If the emotion is sorrow, then Fact 3(heart2) = sinks [at NZ] | NZ = idea,thought, ...
• If the emotion is grief, then Fact 0(heart2) = bleeds.
For Y = SORROW, Caus 3Fact 1(heart2) = break [NX’s ~];
for Y = PLEASURE, Caus 3Fact 1(heart2) = warm [(the cockles of) NX’s ~];
and for Y = PAINFUL FEELING, Caus 3Fact 1(heart2) = gnaw [at NX’s ~], pierce [NX’s ~].
As can be easily seen, the elements of the values of different LFs differ, in their turn,
according to the SemA Y of HEART2.
Facets defined by the perspective on L
The referent of L may be used in different situations in various ways. Because of that, L
may have different clusters of semantic derivations and collocates, each cluster corresponding to
a particular situation. To put it differently, L has different restricted cooccurrence as a function of
different situations of its use—of its different facets. To account for this, the lexical entry of L
has to be divided according to these facets. Let us illustrate this phenomenon with two examples:
nouns BLOOD and NOSE.
In English, the noun BLOOD is used under six different facets: as a ‘carrier’ of life (= blood
circulation), as a signal of injury or disease (bleeding), as the target of medical treatment, as a
symbol of temperament and feelings, as a symbol of life, and as a ‘carrier’ of biological origin.
Its lexical entry will show the divisions along these lines.
BLOOD (fluid circulating in the body...)
1) Circulation
blood vessels, arteries; B. flows/runs in X’s veins, rushes to X’s cheeks, ...

—Chapter 18. Lexical Functions— 83
(red/white) blood cells; B. coagulates; hot-blooded [animal]; blood-shot [eyes]
2) Bleeding
bleed; B. oozes, pours, spurts, squirts; draw, spill B.; blood-stained; B. cakes;
puddle of B.
3) Medical Treatment
blood pressure; transfusion; blood test; donate B., donor, blood-drive; blood clot;
hematology
4) Temperament/feelings
cold/hot B.; B. boils/freezes; in cold B.; blood-curdling
5) Symbol of life
shed/spill B.; to the last drop of B.; crush/drown in B.; rivers/sea of B.; bloodshed,
bloodbath; bloodthirsty
6) Biological origin
of the same B.; German and Armenian B.; blue, good, pure, royal B.; [prince] of the
B.; related by B.; run in the B.
NB : I by no means insist on this description. It seems preferable, as many English dictionaries do, to separate the
noun BLOOD into several LUs: BLOOD1 (body liquid), BLOOD2 (B.1 as symbol of emotions), BLOOD3 (B.1 as
symbol of life), and BLOOD4 (B.1 as symbol of biological origin). But even then BLOOD1 will have three facets. This
example is purely illustrative.
For the LU NOSE, at least two facets must be distinguished: the nose as an organ and as an
esthetic element.
NOSE (part of the face...)
1) Anatomy/physiology
nostrils, nasal cavity; bridge of the N.; breathe through the N.; head cold, running,
stuffed N., snot; blow/wipe one’s N.; pick one’s N.; handkerchief; kleenex
2) Esthetics
Greek, Roman, aquiline N.; pug, snub, bulbous, turned-up N.
6.1.4 Lexical Functions and Their Relation to the Definitions of their Keywords
The whole specificity of LFs follows from their intimate semantic relatedness to their key-
words. For instance, if a lexical unit L has a Magn, its meaning (L) tends to include a gradable
component, which is to accept the intensification and to which Magn actually applies (but see
below, at the end of this subsection). In X’s serious INJURY, what is intensified by SERIOUS is the
component (interference with X’s normal functioning) (in the meaning of INJURY); in X highly
RECOMMENDS Y to do Z, HIGHLY intensifies the component (X incites Y to do Z); etc. If for f(L)

—Chapter 18. Lexical Functions— 84
the meaning (L) contains only one component that can be the scope of (f), this component is
picked out by default. But if (L) contains more than one component feasible for (f), then it
becomes necessary to specify explicitly which of the available components is targeted by f. Thus,
the meaning of HOPEV includes two gradable components: (likely) and (want); that is, X hopes that
Y ≈ (X believes that Y, which X wants, is likely). Therefore, in the entry for HOPEV one has to
indicate—with a subscript—which component the given Magn applies to:
Magn [(likely)] (hopeV) = confidently, firmly
Magn [(want)] (hopeV) = dearly, deeply, fervently, passionately
At the same time, there may be a Magn that applies to both gradable components together,
i.e., which does not differentiate between (likely) and (want); then no subscript is used:
Magn(hopeV) = strongly, to God, to goodness
Here is another example. In French, ATTENDRE (wait, await) has two different Magns,
semantically bearing on two different components in the verb definition: (desirability) of the thing
awaited and (readiness) of the subject to deal with it:
Magn [(want)] (attendre) = avec impatience (with impatience)
[but not *impatiemment (impatiently)]
Magn [(ready)] (attendre) = de pied ferme lit. (on steady foot) (= be ready and waiting)
Such use of semantic subscripts makes the system of LFs even more discriminatory and
powerful. As we see, the elements of the value of an LF f for the keyword L may be semantically
different because f targets different components in (L).
However, this is not always the case that a LF bears on an ‘internal’ component in the
definition of its keyword L—this is, as was said above, no more than a strong tendency. In some
cases an LF can be semantically related to a meaning (σ) that is not part of (L) but caracterizes (L)
from the ‘outside’. For instance, in enduring joy, the adjective, which is a Magn(JOY), indicates
the long duration of the feeling, i.e., intensifies (period of time), and this meaning is not present
inside the meaning (joy)—it is an external caracterizer. Similarly, AntiMagn(RAINN) light in light
rain bears on the size of individual drops: (rain whose drops are small); again, (size (of drops)) is
not part of the meaning (rain), but caracterizes it for the outside. To accound for such links, the
name of the LF must be supplied with an alphabetical superscript specifying the external carac-
terizer: Magn temp (JOY), AntiMagn dimens (RAINN), etc. (For more on this topic, see Iordanskaja &
Polguère 2005.)

6.1.5 LFs as Deep Lexical Units
—Chapter 18. Lexical Functions— 85
Summing up what has just been said, simple standard LFs are, in an important respect,
similar to grammatical signifieds—inflectional (= grammemes), such as nominal number or
verbal tense, or derivational (= derivatemes), such as Agent or Location nouns. Namely, very
much like grammatical signifieds, LFs are not genuine, bona fide meanings. For instance, the
grammeme (PLURAL) is a label for a set of correspondences between a few meanings—(more than
one) (trees), or (different sorts of) (French wines), or else (great quantity/extent)(the sands of the
desert)—and a few morphological markers—-s, -en (oxen), -i (foci), -a (phenomena), etc. In just
the same way, the LF f is a label for a cluster of correspondences between several related mean-
ings (= having a common core) and several lexical expressions. Therefore, a complete description
of an LF must include all its alternative semantic representations (which is not done in the list in
Subsection 2.2). In order to give an idea of what this could look like, let me give three examples.
NB : The underscoring in a SemS indicates the semantically dominant node: a node in the assertional
part of it which resumes, in a sense, the whole SemS; the rest of the SemS hinges, so to speak, on the
dominant node, so that a SemS is reducible to its dominant node. 24 ]
1. The LF S 1:
LF-Sem-Rule 1
2. The LF Magn:
LF-Sem-Rule 2
2
(P) # (magnitude )
(P)
1
(person)
(very)
1
(!)
!
"
S (L(P))
1
L(P)
ATTR
Magn
The meaning (person who P-es) can be
expressed by a derived noun S 1, some-
thing of the type P-er (smoker, skier,
student, blacksmith, etc.).
The meaning (value a of magnitude P
is elevated) can be expressed by a
Magn, modifying the noun L(P) that
lexicalizes the meaning (P).

LF-Sem-Rule 3
(intensely)
1
!
—Chapter 18. Lexical Functions— 86
L(P)
ATTR
(P) Magn
The meaning (intensely P) is expressed
by a Magn modifying the LU L(P) that
lexicalizes the meaning (P).
There are also other sources for Magn, but these examples seem to be sufficient here.
3. The LF Real 3:
LF-Sem-Rule 4
(P) " (prescription)
2 3
1
(Y) (Z)
example.
!
L(Z)
Real 3
I II
II
L(Y)
L(P)
The meaning (Z does Y that some-
one’s prescription P tells him to do)
can be expressed by a Real 3,
applied to the noun L(P) that
lexicalizes the meaning (P).
The LF Real i has other semantic sources, too; but as above, I will limit myself to one
Where subtler meaning distinctions concerning LFs are needed, semantic distinguishers,
superscripts and other devices, as presented above, are used.
According to the given definitions, the LFs are selected, in the process of synthesis, and
inserted into the DSyntS of the sentence being constructed—very much like all other Deep
Lexical Units. 25
6.2. Phraseological Aspect of Lexical Functions
6.2.1. Graduality of Restrictedness of LF Expressions
In some cases, an element of the value of a given LF f is highly idiosyncratic: e.g., pay
ATTENTION, where PAY = Oper 1(attention) is a rare choice. But in numerous other cases elements
of an LF are not idiosyncratic at all: for example, attract X’s ATTENTION, where ATTRACT =
Caus 2Func 2(attention) is semantically completely motivated. However, this expression still must
be described by a lexical function: with some other nouns, you cannot use attract in order to

—Chapter 18. Lexical Functions— 87
express the meaning (Y causes X’s L to be on Y) [= Caus 2Func 2]. Thus, Y compels 〈*attracts〉
X’s respect, Y draws 〈*attracts〉 criticisms/cheers from X, Y excites 〈*attracts〉 X’s suspicion,
etc. The meaning corresponding to Caus 2Func 2
is such that at least in some unpredictable cases it
is expressed by phraseologically bound LUs; as a result, its expression has to be explicitly speci-
fied everywhere, since otherwise one cannot know whether this expression is free or not in a
specific case.
To sum up: The value of a given LF from a given keyword is not always necessarily very
idiosyncratic. The lexical-functional expressions form a cline—from the expressions strongly
phraseologically bound by their keyword (i.e., highly restricted) to the expressions whose phrase-
ological character is only determined par ricochet, i.e., they are phrasemes (= collocations) only
because the corresponding meaning participates systematically in collocations. (This is what was
called above a ‘collocation by analogy’.) Otherwise, they have almost zero restrictedness. Thus,
the phrase very tired, which is fairly free, is a collocation and has to be described by the LF Magn
because there are other Magns and also because one cannot say *very wounded or *very armed
(these impossibilities cannot be reduced to semantic or formal rules and therefore have to be
listed).
Note, however, that one does not need to stock all the expressions with very in the dictiona-
ry: it is enough to give the general rule (VERY modifies adjectives and adverbs) and then list only
those cases where very is impossible. Cf.:
Magn(wounded) = badly, seriously < critically < fatally, mortally
Magn(armed) = heavily, to the teeth
The absence of VERY in these series means that it cannot be used. On the other hand, for TIRED
and RESTED, the dictionary must show VERY:
Magn(tired) = very < bone-, dog-, to the bone
Magn(rested) = very, well < completely
6.2.2. LF Expressions and Collocations
What is the relation between LF expressions and collocations? The answer is that the two sets
intersect. On the one hand, some LF expressions do not describe collocations: they describe
semantic derivations; on the other hand, some collocations are not described by LFs. It is these
mavericks that interest us here. What are they?
They are collocations in which the ‘disturbing’ element (= the collocate) is an actant of the
keyword (= the base). Let me cite a few examples from French. In French, such expression as

—Chapter 18. Lexical Functions— 88
assurance maladie lit. (illness insurance) = (medical insurance) and assurance vie (life insurance)
are collocations: in the first case you insure against illness, in the second case you insure your
life. (Other cases of the same type are assurance incendie lit. (fire insurance) vs. assurance
logement lit. (apartment insurance).) Or else Fr. auto-école vs. driving school; or Fr. [un]
condamné à vie lit. (condemned to life) = (person condemned to life in prison) vs. [un] condamné à
mort lit. (condemned to death); cf. life sentence vs. death sentence. In all these collocations the
problem is created by a semantic actant expressed idiosyncratically (it is shown in boldface). The
technique used in MTT to describe such phenomena is to use the Government Pattern, in which
all such idiosyncrasies are explicitly stated. Thus, the collocations of this type are covered by
GPs rather than by LFs. (See Chapter 17, 5, p. 00ff.)
6.3 Lexical Functions in Linguistic Representations
As far as the participation of LFs in linguistic representations is concerned, the standard and the
non-standard LFs are treated differently.
The standard LFs are considered to be Deep Lexical Units; therefore, they appear in the
DSynt-Structure of the sentence under processing. Thus, during synthesis, the standard LF f is
selected based on its semantic source in the initial SemS. For instance, for the expression high
temperature of the liquid we have the correspondence (16):
(16) (temperature )
(liquid)
1 2
1
(!)
(very)
TEMPERATURE
"
I
LIQUID
SemS DSyntS
ATTR
Magn
The LF Magn is computed according to LF-Sem-Rule 2, see above, 5.1.5. (The adjective HIGH is
computed later—at the SSynt-level as an element of the value Magn(temperature); it is taken from
the lexical entry of TEMPERATURE and inserted into the SSyntS of the sentence.)
The non-standard LFs are not Deep Lexical Units and they do not appear in any
representation. They are simply used in the lexicon to ensure the correct lexical choice in the
process of synthesis. For instance, for the expression leap year we have the correspondence (17):

(17)
(year)
(include)
1 2
—Chapter 18. Lexical Functions— 89
1
(366)
(day)
YEAR
! ATTR
LEAP
The meaning (that includes 366 days) is specified in the lexical entry for YEAR as a non-standard
LF with the value LEAP:
YEAR
…
that has 366 days : leap
This non-standard LF allows for the transition SemS ⇒ DSynt in (17); as a result, the LU LEAP
appears in the corresponding DSyntS.
6.4 Universality of LFs
One of the most important features of LFs is their linguistic universality: they can be used to
describe both semantic derivation and restricted lexical cooccurrence (= collocations) in any
human language. Interestingly, they correspond to meanings that receive special treatment in
natural language: to what is called grammatical meanings, i.e., inflectional and/or
derivational meanings (grammemes and derivatemes). Thus, the LFs provide lexical
expressions for meanings out of a ‘privileged’ set, which under different circumstances are
expressed morphologically. I cannot go into greater detail here, but I think that it could be useful
to illustrate LFs in languages other than English. Therefore, I present below some examples taken
from thirteen different languages.
1. English
Magn(rain) = heavy
Magn(argument) = convincing, strong, weighty,
Oper1(trip) Oper1(deal) = take [ART ~]
= strike [ART ~]
< knock-down
Magn(applause) = thunderous, deafening, boisterous,
whirl-wind, …
Oper1(apologies) = offer [~s]
Magn(resistance) = dogged, ferocious, fierce, stiff, ... Oper1(resistance) = offer, put up [~]
2. French
Magn(pluie (rain)) = forte (strong), violente (violent)
< torrentielle (torrential)) Oper 1(voyage (trip)) = effectuer (carry out),
faire (make) [ART ~]
Magn(argument (argument)) = fort (strong),
convaincant (convincing), massue lit. (club) Oper 1(accord (deal)) = arriver (arrive), parvenir
(achieve) [à ART ~]

—Chapter 18. Lexical Functions— 90
Magn(applaudissements (applause)) = nourris lit. (fed),
Frénétiques (frenzied), à tout casser (to break all) Oper1(excuses (apologies)) = présenter
(present) [ART/A
poss
~ à N]
Magn(résistance (resistance)) = acharné (dogged),
farouche (ferocious) Oper1(résistance (resistance)) = opposer
(oppose) [ ART ~ à N]
3. Russian
Magn(dožd´ (rain)) = sil´nyj (strong) < prolivnoj, lit. Oper1(putešestvie (trip)) = soveršit´ (accomplish) ~e]
(showery)
Magn(dovod (argument)) = veskij (weighty),
ubeditel´nyj (convincing) Oper1(soglašenie (deal)) = pridti (come to) [ k ~ju],
zaključit´ (conclude) [~e]
Magn(applodismenty (applause)) = burnye lit. (tempestuous),
gromovye (thunderous) Oper1(izvinenija (apologies)) = prinesti (bring) [~ja]
Magn(soprotivlenie (resistance)) = jarostnoe (furious)
Oper 1(soprotivlenie (resistance)) = okazat´ ≈ (give) [~e]
4. German
Magn(Regen (rainN)) = starker (strong)
< Platz-( burst-) Oper1(Reise (trip)) = [ART ~acc] machen (make)
Magn(Argument (argument)) = gewichtiges (weighty),
schlagende (beating), unschlagbares (beating) Oper1(Übereinkunft (deal)) = [über ART ~acc]
erzielen (obtain)
Magn(Applaus (applause)) = tosender (roaring) Oper1(Entschuldigung (apology)) = [Ndat ART ~acc]
~acc] vorbringen (forebear)
Magn(Widerstand (resistance) = erbitterter (embittered) Oper1(Widerstand (resistance)) = [Ndat ART ~acc]
leisten (supply), entgegensetzen (set against)
5. Polish
Magn(deszcz (rainN)) = silny (strong) <
ulewny (showery) IncepOper 1(podróż (trip)) = wybrać się [w ~]
(get oneself ready for))
[NB: Polish does not have Oper 1 for this lexeme]
Magn(argument (argument)) = silny (strong),
mocny (powerful) Oper 1(porozumienie (deal)) = dojść [do ~a] (come to)
Magn(oklaski (applause)) = burzliwe (tempestuous),
olbrzymie (enormous) Oper 1(przeprosiny (excuses)) = Λ
[NB: no Oper 1 for this lexeme, either; instead, the verb
przepraszać (offer one’s apologies) is used—it is V 0(przeprosiny)]
Magn(opór (resistance)) = zacięty (persistent) Oper 1(opór (resistance)) = stawiać (put up) [Ndat ~]
6. Hungarian
Magn(eső (rainN)) = zuhogó (torrential) Oper 1(utazás (trip)) = [~t] tenni (make)
Magn(érv (argument)) = komoly (serious) Oper 1(lépés (step)) = [~t] tenni (make)
Magn(taps (applause)) = viharos (whirly), vas- (iron) Oper 1(hatalom (power)) = [~t] birtokolni (possess)
Magn(ellenállás (resistance) = heves (fiery) Oper 1(ellenállás (resistance)) = [~t] kifejteni (develop)

—Chapter 18. Lexical Functions— 91
7. Arabic
Magn(matàr (rain N )) = ġazīr (abundant) Oper 1(safar (trip)) = qāma [bi ~] (go on)
Magn(ħuzzat (argument)) = dāmiġat (hitting),
qawijja (strong) Oper1(?ittifāq (deal)) = tawas`s`àla
[?ila ~]
(come to)
Magn(tas>fīq (applause)) = ħārr (warm) Oper1(/i?tiDārāt (apologies)) = qaddama [ART ~]
(advance) [trans.]
Magn(muqāwamat (resistance)) = /arisat (savage),
/anifat (violent), /arī/at (audacious) Oper1(muqāwamat (resistance)) = /abdā [ART ~]
(show, demonstrate)
See Ibrahim 2002 for some particular techniques of building collocations of the type
Oper 1(S 0) + S 0 in Arabic.
8. Somali
Oper 1(birmad (attack N )) = ki?i (lift), dhufan (strike) Oper 1(jawabta (answerN)) =?elin (turn)
Oper 1 (rajoda (hope N )) = qabi (keep) Oper 1(fiiro (attention)) = lahaan (have), yeelan (make)
Oper 1(dagal (struggle N )) = jiri (find oneself in) Oper 1(moqif (positionN)) = taagan (stand in)
9. Albanian (Hendriks 1980)
Oper 1(besim (confidence)) = ka (have) Oper 2(qotek (beating)) = ha (eat)
Oper1(be (oath)) = bën (make) Real1(borxhin (debt)) = bën (make)
IncepOper1(bela (problems)) = bie [në ~] (fall into) [this espression means (pay a debt), rather than *(make a
debt)]
10. Persian
Oper 1(kotak (beating)) = zadan (strike) Oper 1(qose (troubleN)) = xordan (eat)
Oper 1(ġalabe (victory)) = kardan (make) Oper 1(fähm (understanding)) = dāštan (have)
Oper 1(soāl (questionN)) = kardan (make) Oper 1(kālāme (word(s))) = hārf zadan (talk)
Real 1(zaru (broom)) = kardan (make) [= (sweep)] Real 1(čašm (eye)) = duxtan (sew) [= (observe)]
In Persian, the majority (about 90 %) of verbal meanings are expressed not by individual
lexemes but rather by noun-verb collocations of the type illustrated, which include the LFs such
as Oper and Real. In this language, LFs occupy a very special place indeed. Persian has slightly
over 100 simple verbs that are actively used as Opers and Reals; some of them are pretty rare
(Karimi 1997).
11. Turkish
Magn(yag* *mur (rainN)) = siddetli (violent) Oper1(yolculuk (tripN)) = [~] yapmak (do)
Oper 1(seyahat (tripN)) = [~] etmek (make)
Magn(kanIt (argumentN)) = sag*lam (solid),
kesin (decisive) Oper 1(anlaşma (understanding)) = [~ya] varmak (reach)
Magn(alkIş (applause)) = kuvvetlI (strong) Oper 1(özür (apology)) = [~] dilemek (ask for))
Magn(direniş (resistance)) = büyük (big) Oper 1(direniş (resistance)) = [~] göstermek (show)

—Chapter 18. Lexical Functions— 92
12. Chinese
Magn(yu‹ (rainN)) = dà (big) Oper 1(lü]tú (tripN)) = tàshàng [~] (walk on)
Magn(lùnjù (argumentN)) = yo‹ulì-de (having strength) Oper 1(xiéyí (dealN)) = dáchéng [~] (reach)
Magn(zha‹ngshēng (applause)) = léidòng (thunderous) Oper 1(qiàn (apology)) = dào [yīge ~] (say) [an apology])
Magn(di#kàng (resistance)) = wánqiáng-de (firm, solid) Oper 1(di#kàng (resistance)) = jìnxíng [~] (proceed to)
For LFs in Korean, see Hong 1995 and Lee, Park & Kim 2001.
13. Malagasy
Magn(orana (rainN)) = be (abundant),
mivatravatra (violent) Oper 1(dia (tripN)) = manao [~] (do)
Magn(hevitra (argumentN)) = mitombina (convincing) Oper 1(fifanaraha (dealN)) = manao [~] (do)
Magn(tehaka (applause)) = mirefodrefotra (frenzied) Oper 1(azafady (apology)) = manao [~] (do)
Magn(fanoherana (resistance)) = lehibe (grande) Oper 1(fanaherana (resistance)) = manao [~] (do)
Collocations described by LFs are not only wide-spread in modern languages—they were
present already at the earliest stages of language development. Thus, Gamkrelidze & Ivanov
1984: 833-834 show the existence of LF-collocations in Proto-Indo-European, for instance:
(18) a.*nom(e/o)n *d h eH- lit. (put [a] name) = (establish/give a name)
[*d h eH = CausFunc 1(*nom(e/o)n)]
b. *k h leu̯o- *d h eH- lit. (put glory) = (acquire glory)
[*d h eH = Caus 1Func 1(*nom(e/o)n)]
After the demonstration of the importance and the power of LFs, a few words of caution are
in order. What I said in this subsection should by no means be construed as implying that the
same set of simple standard LFs is valid for any language: this is simply not the case, and this is
so for two reasons.
• First, a simple standard LF can be inapplicable in some languages. Thus, Bambara does not
have readily available nominalizations (S 0), so that collocations of the type put up resistance,
offer apologies or get a refusal are mostly impossible in this language; finite verbal forms are
used instead. Therefore, support verb LFs (Oper i, Func i, Labor ij) would be practically useless in
Bambara: they cannot have enough keywords.
• Second, a simple standard LFs that is applicable in the language can have mostly standard,
non phraseologized expressions and, as a result, it is of only limited relevance. Suppose that in
language L the LF Oper 1 is expressed by three support verbs: (make), (have) and (say), distributed

—Chapter 18. Lexical Functions— 93
as follows: (make) combines with the names of actions, (have)—with the names of states, and
(say) —with the names of sounds. In L then this LF, although formally valid, is practically useless:
the elements of its value are not bound phraseologically. Similarly, again in Bambara, the LF
Magn is more often than not expressed by the augmentative suffix -bá: (heavy rain) is sánjí+bá,
(thunderous applause)—tégéré+bá, etc. Under these circumstances, the importance of Magn is
much lower in Bambara than it is, say, in English or French.
The universality of LFs boils down to their following three aspects:
1) Generality and versatility of the concept itself of LF: LFs can be used to describe both
semantic derivations and restricted lexical cooccurrence in any human language, and they allow
the linguist to do this in a homogeneous and systematic way. Crucially, as I have already said,
LFs correspond to meanings that receive special treatment in natural language: namely, to what
are called grammatical meanings, i.e., inflectional and/or derivational meanings
(grammemes + derivatemes). To put it differently, what is an LF in language L can well be
a grammeme or a derivateme in L itself or in a different language L´ and then will be expressed
by morphological means. Thus, the LFs provide lexical expressions for meanings out of a
‘privileged’ set, which under different circumstances are expressed morphologically. (For more
on the role of LFs for the description of collocations, see Chapter 20, especially 3, p. 00ff.)
2) The necessity of the LF list proposed: In most languages and in most cases of restricted
lexical cooccurrence within a given language, the simple standard LFs listed in this chapter will
prove applicable and useful.
3) The sufficiency of the LF list proposed: It is unlikely that in any L there can be found some
regular series of collocations for the description of which a simple standard LF absent from our
inventory will be needed. In other words, I believe—until proven wrong—that the list of simple
standard LFs is exhaustive. This is of course an empirical question: the study of new languages
under the angle of restricted lexical cooccurrence may bring in new LFs, while some of those that
have now the status of non-standard can be promoted to standard. Such an event is by no means
excluded, but I find it highly improbable that many new standard LFs will turn up. However, it is
worth a bit more discussion.
7 New Simple Standard LFs
The question is asked over and over again: How can one know that the list of simple standard
LFs is complete? Is there a way to prove or disprove it? Unfortunately, the answer is a clear No.

—Chapter 18. Lexical Functions— 94
The inventory of LFs is established empirically, and no logical reasoning can guarantee the non-
existence of other LFs.
First, a language that has not yet been explored for restricted lexical coccurrence could
have an LF which is not known today, something, say, like (of a big size) or (deal with Y in a
friendly/unfriendly manner). True, the probability of finding such a function cannot be very high,
because the most abstract and general meanings appearing as LFs are already, so to speak,
involved and an LF by definition tends to be associated with a very abstract and general meaning.
Second, the borderline between standard and non-standard LFs is strictly quantitative,
which means that it is rather fuzzy: the researcher has to decide whether a given LF should be
treated as a standard or non-standard one, and in many cases the decision is far from obvious.
Thus, consider three candidates for Simple Standard LFs: (female of) = Fem, (male of) = Masc, and
(young of) = Juven. All the three are defined for all animal names:
Fem(horse) = mare Masc(horse) = stallion Juven(horse) = foal
Fem(chicken) = hen Masc(chicken) = rooster Juven(chicken) = chick
Fem(cat) = (pussy)cat Masc(cat) = (tom)cat Juven(cat) = kitten
Fem(goose) = goose Masc(goose) = gander Juven(goose) = gosling
Fem(elephant) = elephant cow, Masc(elephant) = elephant bull, Juven(elephant) = baby
cow elephant bull elephant elephant
Fem(tiger) = tigress Masc(tiger) = tiger Juven(tiger) = tiger
cub
Do they qualify? I think that in English rather not: their keywords are limited in number
and very specific: just animal names. In addition, they do not participate in paraphrasing. (But I
by no means insist on this opinion.) However, nothing precludes us from discovering a language
where these LFs will embrace also human beings, including names of inhabitants; in such a
language it would be more justified to consider Fem, Masc and Juven as standard LFs.
For the same reason, I do not think that it is worth introducing as a Standard LF the LF
Pecun = (currency of): USA ~ dollar, Germany ~ mark, Ukraine ~ karbovanec, Norway ~ crown,
Greece ~ drachma, Japan ~ yen, Israel ~ shekel, Paraguay ~ guaraní, Samoa ~ tala, ...: the set of
possible keywords is closed and they are highly specific (names of countries). In addition, the LF
Pecun is too sensitive to ‘external’ change: thus, the advent of the euro wiped out many of its
possible values. And, most importantly, Pecun again cannot participate in paraphrasing.

—Chapter 18. Lexical Functions— 95
There are two more candidates for the status of simple standard LFs: Germ and Nocer,
which have been played with for a long time; but today I think that it is not worth trying to
introduce them into the inventory of Standard LFs.
Germ (Lat. germen (embryo)) meaning (germ, beginning) is poorly represented:
Fr. Germ(colère (anger)) = ferment, levain (yeast) [de la ~], Germ(idea) = germ [of an ~]—and
that’s about it: I cannot think of other good examples.
Nocer (Lat. nocēre (do harm)) denotes a harmful action that the referent of L is supposed to
do to a ‘victim’: in English, a bee stings and a snake bites, as well as a dog, but in French, a dog
(bites) (mord), while a bee or a snake (picks) (pique 〈*mord〉); in Russian, a dog (bites) (kusaet),
but a bee or a snake both (bites) (kusaet) or (stings) (žalit); in English, a car runs over/down a
pedestrian, in Russian, it (crashes) (davit) him, and in French, it (mows) (fauche) him; and so
forth. But, on the one hand, it seems that the possible arguments and values are not numerous
enough; on the other hand, the necessary meaning can be probably better described by a
combination of two LFs: AntiBon and Involv. Indeed, AntiBon 1Involv(automobile) could be
strikes Y [a tree, a cement wall], smashes into Y ((bad for itself) = (does harm to itself)), while
AntiBon 2Involv(automobile) would then stand for runs Y over, strikes Y [a pedestrian], etc. I
prefer the latter description as more analytical.
For the reasons indicated, I did not put these two LFs on my list.
But, as has been said above, it is immaterial—in both practice and theory—whether we
decide to treat an LF as standard or non-standard (cf. Polguère 200?).
8 Lexical Functions in Computer Applications
LFs were invented in 1961, during my participation—as a hired worker—in a geological
expedition in Southern Kazakhstan semi-desert. Walking up and down stony hills, I was return-
ing in my head one of well-known problems of machine translation: how to avoid or at least
reduce tedious search for lexically-determined contextual Russian equivalents of such English
lexemes as IMPORTANT, EXTENSIVE, HEAVY, HIGH, on the one hand, and DO, MAKE, GIVE, GET,
etc. on the other. That is how I figured out what later became Magn and Oper 1. This goal was
clearly stated in Žolkovskij & Mel´čuk 1967; cf. also a specific proposal to this effect in Kula-
gina & Mel´čuk 1968: 301-302.
It turned out later that LFs have, at the same time, an important theoretical status. For one
thing, they constitute the ‘missing link’ in the theory of phraseology, since they are necessary to

—Chapter 18. Lexical Functions— 96
describe rigorously and systematically collocations, or what can be also called semi-phra-
semes. That theoretical contribution is dealt with in some detail in Chapter 21. Furthermore,
they are essential in the theory of syntax, since several syntactic regularities have to be described
in terms of LFs. Cf., e.g., (19) [from Abeillé 1988]:
(19) a. King John launched [= IncepOper 1] an attack against the city.
vs.
Which city did King John launch an attack against?
It is against this city that King John launched an attack.
b. King John watched an attack against the city.
vs.
*Which city did King John watch an attack against?
*It is against this city that King John watched an attack.
Extraction in such constructions is possible/impossible according to whether the verb in question
is/is not an LF of its Direct Object (in launch an attack it is, in watch an attack it isn’t).
Here, however, I will limit myself to shedding some light on the use of LFs in computa-
tional linguistics.
Basically, I know of three major ways in which LFs prove useful (or even indispensable)
for computer-oriented linguistic descriptions:
• As a tool for determining correct lexical choices within phrases (all LFs): COLLOCATIONAL
aspect.
• As a tool for determining lexical choices necessary to adapt the DSynt-Structure of the
sentence to be synthesized to its Communicative Structure (some verbal syntagmatic LFs,
namely—support, causative and fulfillment verbs): COMMUNICATIVE aspect.
• As a tool for determining lexical choices necessary to ensure the coherence of the text to be
synthesized (some paradigmatic LFs): LEXICAL-COHESIONAL aspect.
8.1 LFs and Lexical Choices
The use of LFs for finding the correct collocates is rather straightforward. Thus, under Machine
Translation at the level of DSynt-Structure (i.e., without going through a SemR), it suffices to
reduce the source language collocation (= which is to be translated) to its LF-representation, then
translate the keyword only and then, finally, to select the value of the LF for the translation of the
keyword in the target language. For instance, Fr. Jean m’a détourné de cette habitude is
reduced—via a monolingual French dictionary which lists the values of all LFs for all head LUs
to its DSyntS (20). Then, on the stage of transfer, the French DSynt-tree (20) is replac-ed with the

—Chapter 18. Lexical Functions— 97
English DSynt-tree (20)´, which is ensured by a relatively simple and straight-for-ward bilingual
index:
(20) LiquOper 1
I II III
JEAN MOI HABITUDE
(20)´ LiquOper 1
I II III
JOHN I HABIT
In the process of synthesis, (20)´ will be realized as John broke me of this habit, this time using
an English monolongual dictionary where the values of LFs are specified.
In phrases of this type only the noun requires actual transfer, i.e., looking up of its English
equivalent(s). The search for the ‘bizarre’ correspondence DÉTOURNER = BREAK in the context of
HABIT is avoided altogether: BREAK will be computed as an element of the value of the LF
LiquOper 1 (HABIT) in a monolingual English dictionary. Normally, multilingual translation in all
possible directions between n languages requires 2n! dictionaries (n! is the number of language
pairs, and there are two directions; for five languages, 2 × (5 × 4× 3 × 2 × 1) = 240 transfer
dictionairies are needed. However, if we are using LFs, for the transfer of collocations it is
enough to have detailed and accurate MONOLINGUAL dictionaries with LFs; LFs appear then as a
convenient transfer interlingua.
To make this procedure clearer, let me cite a series of correspondences that can be easily
expressed in terms of LFs (the example is adapted from Fontenelle 1993):
Eng. HABIT ⇔ Fr. HABITUDE
IncepOper 1 acquire, develop, form [ART ~], contracter, prendre [ART ~]
get [into ART ~], take [to ART ~]
FinOper 1 drop [ART ~], get out, get rid abandonner, perdre [ART ~]
[of ART ~], ...
LiquOper 1 break [NX of ART ~], wean [NX from détacher, détourner [NX de ART ~]
ART ~]
Liqu 1Oper 1 break (off), kick, shake (off), throw se débarrasser, se défaire [de ART ~],
off [ART ~] renoncer [à ART ~], rompre [avec ART ~]
CausFunc 1 instill [ART ~ in(to) NX] inculquer [ART ~ à NX]
Figure 20.4: Lexical Correspondences Expressed in Terms of LFs

—Chapter 18. Lexical Functions— 98
Interestingly, equations relating the LFs allow for necessary syntactic transformations.
Thus, Rus. On vzjal zverja na mušku lit. (He took the-beast on bead) = (He took aim at the beast)
is reduced to
| ————I II ————————— ↓
ON←I−Labreal 12(MUŠKA)−II→ZVER´ MUŠKA,
which is replaced, using the dictionary, by the English lexemes as follows:
| ————I II ———————— ↓
HE←I−—Real 1(BEAD)—−II→BEAD BEAST
This DSynt-Structure is readily realized as He drew a bead on the beast. In this way, LFs take
upon themselves even the syntactic adjustments needed to carry out the transfer between
languages—in cases where the LFs are implicated.
If, however, the Machine Translation system under consideration proceeds via a SemR,
then the task (in regard to restricted lexical cooccurrence) is to establish the relevant LF starting
from the initial SemR and then to compute its value for the given L, based on a monolingual
dictionary of the ECD type. Of course the same procedure is needed for text generation, whatever
its underlying representation.
8.2 LFs and Communicative Structure
The use of LFs for an appropriate expression of the Communicative Structure of the sentence
through its lexico-syntactic structure is dealt with in Wanner & Bateman 1990. A detailed
presentation of how this could be done requires a description of both the Paraphrasing System
(Mel’čuk 1992) and Communicative Structure, which is impossible here. Therefore, the reader
has to be satisfied with an example (adapted from Wanner & Bateman 1990). Suppose the text-
generation system has to verbalize the meaning of the sentence (21):
(21) a. The adjective ‘electronic’ indicates to the reader that the dictionaries are dedicated
to computers.
If in the SemS of (21) the meaning of the phrase the adjective ‘electronic’ is specified as the
Theme of the sentence to be synthesized, then sentence (21)a can be produced. But if the Theme
is the meaning of the phrase the reader, a different syntactic structure is needed, which will even-
tually lead to (21)b:
b. The reader gets an indication that the dictionaries are dedicated to computers from
the adjective ‘electronic’.

—Chapter 18. Lexical Functions— 99
To replace indicate with get an indication, one needs the paraphrasing equivalence of the follow-
ing type:
V ≡ S 0(V) + Oper 2(S 0(V))
The application of this equivalence can be illustrated by the following paraphrases:
X analyzes Y ≡ Y undergoes an analysis by X,
X resists Y ≡ Y runs into a resistance by X,
X orders Y to Z ≡ Y receives from X an order to Z, etc.
Most importantly, to successfully use such equivalences, one of course needs a dictionary which
specifies, for each L, the values of all LFs applied to it—that is an Explanatory Combinatorial
Dictionary.
For specific rules controlling the use of LFs to express the Communicative Structure of
the sentence, see Iordanskaja et al@@@. The linguistic paraphrasing rules described in terms of
Lexical Functions are presented in Chapter 14; the structure of an Explanatory Combinatorial
Dictionary is the subject of Chapter 15.
8.3 LFs and Text Cohesion
LFs prove equally useful in selecting the referring expressions in anaphorical links in such a way
as to avoid tedious repetitions and guarantee, at the same time, the maximum cohesion of the
resulting text (see Lee & Evens 19??, Tutin 1992, and Alonso et al. 1992: 160-165). Thus,
speaking of an ambush, you can refer back to it by calling its participants attackers:
(22) An Indonesian patrol was caught in an ambush. The attackers fired three rockets at
the soldiers and sprayed them with automatic fire.
Here, attacker = S 1(ambush), and soldier = S 1(patrol). This lexical knowledge is used to con-
struct the sentence sequence (22) in an obvious way. Another example:
(23) Sales increased slightly in Quebec and Ontario. Modest gains were also reported in
British Columbia.
Instead of simply repeating the same phrase and saying Sales also increased slightly in British
Columbia, the Speaker chooses to use S 2 (increase) = gainN [(the amount by which X increased)]
inm the second sentence, and this allows him to produce a more varied and elegant text.

Bibliographic Remarks
—Chapter 18. Lexical Functions— 100
For more on LFs, see Žolkovskij & Mel’čuk 1965, 1967, Mel’čuk 1974: 78-109, 1982, 1996,
2003a, b, Mel’čuk et al. 1984, 1988, 1992, 1999, Mel’čuk et al. 1995: 125-151, Mel’čuk &
Zholkovsky 1984, 1988, Wanner (ed.) 1996. New: Kahane & Polguère 2001
Notes
1 (1.1, p. 00) LFs are not designed for the description of pseudo-lexical relations, such as, e.g.,
the part-whole relation: this relation holds between things, and not between words (i.e., not
between LUs). This is not to say that the part-whole relation is irrelevant for text production nor
that it should not be covered in a comprehensive linguistic description: as with many other
encyclopedic data, it is very important for the use of language. I am only saying that LFs are not
intended for this type of relation.
2 (1.2, p. 00) The concept of Lexical Functions itself, as well as the LF inventory, was developed
jointly by Alexander Zholkovsky and the present writer in 1963-64. In this chapter, I draw heavi-
ly not only on our above-mentioned publications but also on the rich experience of our common
work as authors and editors of the Russian Explanatory Combinatorial Dictionary (Mel’čuk &
Zholkovsky 1984). I seize this opportunity to thank Prof. Zholkovsky for many things that I have
learned from him in the process.
3 (2.1, p. 00) The term argument is ambiguous: argument1 of a functor vs. argument2 of an LF. To
avoid this ambiguity, we propose to use—instead of argument2—the term keyword (of an LF).
Other current terms for what is called keyword and value (of a lexical function) are base and
collocates (of a collocation).
4 (2.1, p. 00) Because of their unique cooccurrence (only with SMOKEV and RAINV, respectively),
the expressions kLIKE A CHIMNEYl and kCATS AND DOGSl are actually pseudo-phrasemes of
English—that is, they need not have separate entries in an English dictionary. For more on
pseudo-phrasemes, see Chapter 21, 3, Comment 2 on Definition 21.11, p. 00.
5 (2.2, Def. 18.1, p. 00) The term keyword (of a LF) is used in order to avoid the inflated poly-
semy of the term argument in linguistics (an argument of a semantic predicate).

—Chapter 18. Lexical Functions— 101
6 (2.3, p. 00) More precisely, café crème (or [un] crème) signifies (strong expresso coffee with a
little milk). There is also the expression café au lait, meaning (not so strong coffee in a bowl with
much hot milk). As one can easily see, the meaning under consideration implies considerable
specification.
7 (3.1, p. 00) The discussion of LFs requires several concepts related to the DSynt-representation
of sentences; all these concepts have been introduced in Chapter 11, p. 00ff.
8 (3.2, p. 00) Synonymy and antonymy have been known since Antiquity; conversives as a sepa-
rate class of lexical correlates were explicitly introduced in Lyons 1963: 72, under the name of
converse terms (see also Apresjan 1974: 256-257).
9 (3.2, No. 1, p. 00) As a curiosity, see the note Suárez 1971 on the numerosity of absolute
synonyms in Tehuelche.
10 (3.2, No. 2, p. 00) The opposition ‘> vs. Y) ≡ (X and Y being on a direct line on the same side of zero, X - Y is positive)
(X < Y) ≡ (X and Y being on a direct line on the same side of zero, X - Y is not positive)
(ii) X in opposite direction to Y ≡
(X and Y being on a direct line, X is on one side of zero, and Y is not on the same
side of zero)
11 (3.2, No. 3, p. 00) When counting LFs, Conv21/Conv 231/Conv 321, etc. are taken to be ONE LF:
Conv. The same thing is done with Oper 1/Oper 2, ..., Real 1/Real 2, etc., see Nos. 41-43, 50-52. In
other words, actantial variations in an LF do not change its status as one particular LF. What
counts here is the syntactic role of the keyword L: all Oper i take L as the DSyntA II, and
therefore they can be counted as one LF; with a Func i the keyword is the DSyntA I, so that all
Func i constitute one LF different from Oper i.
12 (3.2, No. 3, p. 00) The definition of (wife) includes the component (linked by the ritual of
marriage), which is negated in the definitions of (girlfriend) and (mistress).
13 (3.2, No. 18, p. 00) For an explanation of DSyntA numbering with the names of sets (this is the
case of Mult), see Chapter 16, 00.0.

—Chapter 18. Lexical Functions— 102
14 (3.3.4, p. 00) In a particular case, a different solution of the problem of non-empty support
verbs is possible. A noun of COLÈRE type can be defined as ((a) X’s emotion … or (b) X’s
general state caused by this emotion). With such a disjunctive definition, we could postulate that
être [en ~] works for the ‘global’ meaning of COLÈRE, while éprouver et ressentir are con-
strained: “only COLÈRE(a)”.
15 (3.3.4, No. 43, p. 00) A main Surface-Syntactic Object of an LU L is either its Dir(ect) O(b-
ject) (if L can have a DirO), or its Indir(ect)O (if L cannot have a DirO), or the strongest Prep(o-
sitional)O (in the absence of both DirO and IndirO).
16 (3.3.4, p. 00) Interestingly, the book by Deribas presents 4962 phrases using 723 different sup-
port verbs. This clearly illustrates the wealth of support verbs and corresponding constructions—
at least, in Russian.
17 (3.3.4, Nos. 49-51, p. 00) In conformity with Kahane & Mel’čuk 2006, we call a causation
verb a verb that expresses the meaning of causation, without indicating the result of the latter;
thus, [to] CAUSE, FORCE [Y to do Z], LEAD [Y to do Z], PROVOKE, PRODUCE, ALLOW, STOP [Y
from doing Z], etc. are causation verbs.
18 (3.3.4, Nos. 49-51, p. 00) ÊTRE à does not appear as such in French as the value of Func1 of
ENVIE: *l’envie d’y aller est à Pierre.
19 (3.3.4, p. 00) An excellent illustration of values of the LF Real (and AntiReal) in English is
offered in Montreal’s The Gazette, Sept. 28, 1985, in a letter to the columnist Ann Landers,
which I reproduce here:
Dear Ann:
I discovered this item stuck in an old book. It was written in longhand and there is no sign
of an author. I hope you will print it. I found it highly inspirational. —T.D.
Dear T. D.: So did I. Here it is:
What is Life?
Life is challenge ... meet it.
Life is a gift ... accept it.
Life is an adventure ... dare it.
Life is a sorrow ... overcome it.
Life is a tragedy ... face it.
Life is a duty ... perform it.

Life is a game ... play it.
Life is a mystery ... unfold it.
Life is a song ... sing it.
Life is an opportunity ... take it.
Life is a journey ... complete it.
Life is a promise ... fulfil it.
Life is a beauty ... praise it.
Life is a struggle... fight it.
Life is a goal ... achieve it.
Life is a puzzle ... solve it.
—Chapter 18. Lexical Functions— 103
20 (5.1, p.00) An interesting analysis of some semantic factors underlying the deviant use of the
passive in French ‘support verb + action noun’ collocations is presented in Anscombre 1986; cf.
such cases as *Solution en a été donnée par Fermat (Its solution was given by Fermat) vs.
Démonstration en a été faite par Fermat (Its demonstration was produced by Fermat), etc.
21 (5.2, Item 2, p. 00) Here is a good example of the LF CausOper1 expressed differently with the
LUs of the same semantic field: send into ECSTASIES vs. fill with FEAR vs. plunge into DESPAIR
vs. drive to DISTRACTION vs. throw into CONFUSION.
22 (6.1.2, p. 00) The nouns like WOUND (in literal sense) cannot have Magn I , since there cannot be
an intensification on a moral level.
23 (6.1.2, p. 00) The expression heavy traffic also admits of a different description: the definition
of (traffic) may include the component (number [of vehicles]), since a lonely car traveling on a
street cannot be called traffic; then Magn should bear exactly on this gradable component ((heavy
traffic) = (LARGE number of vehicles)), and the superscript quant is not needed. The same is true
about the phrase thickly populated [area]: (populated) might include the component (number [of
people]), because a vast area where only one or two persons live cannot be called populated.
24 (p. 00) The concept of dominant nodes in SemSs of utterances as well as in lexicographic
definitions has been formally introduced by L. Iordanskaja and A. Polguère in their joint work
(Iordanskaja 1990: 35-36, Mel’čuk & Polguère 1991: 206-207, Polguère 1992: 117); they have
demonstrated its crucial importance for the transition SemR ⇔ DSyntR, as well as for the writing
of lexicographic definitions. The concept goes back to pioneering ideas of A. Zholkovsky about
‘semantic underscoring’ (1964: 10-12).

—Chapter 18. Lexical Functions— 104
25 (p. 00) Lexical functions are lexical units to such an extent that it is even possible to write
poetry in the ‘language’ of LFs! Thus, in May 1971, during a Linguistic Summer School in Dili-
zhan, Armenia, Alexander Zholkovsky (with some technical assistance by Nikolaj Pertsov) wrote
an amusing pastiche of the famous Pushkin’s poem ‘Na xolmax Gruzii (On Georgia’s Hills...)’. I
find it opportune to quote below Zholkovsky’s poem, along with a literal translation, and
Pushkin’s original, also with a close translation:
V gorax Armenii — i Func, i Fact, i Son. ‘In Armenia mountains there are Func, and Fact, and Son.
V GARAS-e MagnFact 0(struktura). In GARAS* MagnFact 0(structure).
Pečal´ moja ne Ver; mne Bon i AntiBon. My sorrow is not Ver; I’m feeling Bon and AntiBon.
Pečal´ moja polna Figur-a. My sorrow is full of Figur.
Degrad, odin Degrad... GARAS IncepPred(xaos), Degrad, only Degrad... GARAS IncepPred(chaos),
IncepPred(vystavka risunka). IncepPred(drawings exposition**).
I serdce3 vnov´ FigurFunc 0 + Adv 2Caus, And my heartII again FigurFunc 0 + Adv 2Caus,
Čto ne PredAble ot AntiFunc-a! Since it is not PredAble of AntiFunc!’
* GARAS is an abbreviation for ‘Group of English-Russian Automatic Dictionary.
** Exposition of children’s drawings organized by proud parents.
A. S. Pushkin
Na xolmax Gruzii ležit nočnaja mgla,
Šumit Aragva predo mnoju.
Mne grustno i legko; pečal´ moja svetla,
Pečal´ moja polna toboju,
Toboj, odnoj toboj... Unyn´ja moego
Ničto ne mučit, ne trevožit.
I serdce vnov´ bolit i ljubit — ottogo,
Čto ne ljubit´ ono ne možet.
Close translation
Dark falls upon the hills of Georgia,
I hear Aragva’s roar.
I’m sad and light, my grief—transparent,
My sorrow is suffused with you,
With you, with you alone... My melancholy
Remains untouched and undisturbed,
And once again my heart ignites and loves
Because it can't do otherwise.