See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/46071785

Classification of compounds

Article  in  Lingue e Linguaggio · January 2005

Source: OAI

CITATIONS READS

32 5,966

1 author:

Sergio Scalise
University of Bologna
36 PUBLICATIONS   1,061 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Sergio Scalise on 10 January 2017.

The user has requested enhancement of the downloaded file.

 The classification of compounds
Antonietta Bisetto* and Sergio Scalise∗

0. Introduction

Compounds are generally acknowledged to be formed by at least two words1

(e.g. Eng. taxi-driver, postman, It. divano-letto ‘sofa-bed’, camposanto ‘cemetery’, Fr.
tire bouchon ‘cork-screw’, timbre-poste ‘stamp’, etc.). Though this definition does not
always hold true2, it is nonetheless descriptively valid for the core of compounding
processes.
Since we have reasons to believe that the classifications of compounds provided
so far are not adequate, in this paper we would like to propose a new classification
which is at the same time based on consistent criteria and not just valid for one language
or for a family of languages, but hopefully universally valid3.
The paper is organized as follows: in section 1 we will discuss the main
problems with current and previous classifications of compounds; in section 2 we will
present our proposed classification and in section 3 we discuss some residual problems
and draw some provisional conclusions.

∗
University of Bologna

We would like to thank Geert Booij, Emiliano Guevara, Rochelle Lieber and Nigel Vincent for reading a
previous version of this paper and giving us valuable suggestions. Usual disclaimers apply. This paper is
part of a larger research on compounding made possible also thanks to funds of the University of
Bologna.

1
Cf. among others Allen (1978), Selkirk (1982), Scalise (1992), ten Hacken (1994).
2
Leaving aside recursive composition, there are at least two cases where this definition is not true: a) in
the so called neoclassical compounds and b) in phrasal compounds. The former are constructions such as
anthropology and insecticide where one or both constituents are elements of Greek or Latin origin
(anhtropo- and -logy in the first example and -cide in the second one). The latter are those compounds
found in Germanic languages whose first constituent is a phrasal construction as in [God is dead]
theology (cf. Lieber 1992).
3
For this article, we will use data from the Morbo/Comp project, a data base of compounds under
development at the Department of Foreign Languages in Bologna. We have tested our classification of
compounds on the languages analyzed so far within the Morbo/Comp project (Italian, English, Dutch,
Spanish, Chinese, French, German, Bulgarian, Norwegian, Russian).
1. Problems in compound classifications

Classifications of compounds found in the current literature present a range of

different problems, namely: terminological problems (1.1.), problems of neglected
lexical categories (1.2.), and problems of consistency of the criteria used (1.3., 1.3.1.).

1.1. Terminological problems

Terminological problems can be said to fall into two distinct types, basically
caused by a) meaning shifts of the definitory terms in the course of time and b) the
language specific nature of many terms.
To illustrate the first type of problems, one example will suffice. The term
bahuvrihi, originally used to designate a possessive exocentric compound ((one who
has) much rice) with time ended up having the only generic meaning of «exocentric».
Or, as Bauer (2001:700) puts it, this term ended up applying «to any compound which is
not a hyponym of its own head element»4.
As for the second type of problems, the Anglophone tradition in particular has
taken into account almost exclusively two types of compounds: root (or primary)
compounds and synthetic (or secondary) compounds. The notion of root compound,
however, cannot be applied to all languages; Romance languages, for instance, do not
have compounds based on roots because Romance lexemes, besides the root, can
contain another element with grammatical information. In an Italian compound like
capostazione ‘station master’, for example, the two words capo and stazione are formed
by the two roots cap- and stazion- plus a grammatical morpheme, –o and –e
respectively.
Also the label «synthetic compound» used for constructions such as taxi driver,
is not universally applicable because, still in Romance languages, this type of
compound does not seem to exist. Root compound and synthetic compound are thus
language specific terms. This is probably the reason why these terms, while largely
utilized during the '70/'80 (cf., among others, Allen 1978 and Selkirk 1982) have
thereafter been abandoned in the morphological literature

4
Cf. for example Booij (2005: 80) according to whom bahuvrihi compounds «are sometimes considered
to form a subset of the exocentric compounds».
 Also problematic are labels such as «phrasal» and «neoclassical». In fact,
phrasal refers to the (syntactic) nature of the non-head, while neoclassical refers to the
(Greek or Latinate) nature of the head or of both constituents.

1.2. Neglected categories

Morphological research (especially on English compounds) has favoured some

categories of compounds, in particular those formed by two Nouns or by an Adjective
and a Noun (both having categorial output Noun). Other lexical categories (either as
input or output) have often been neglected. This is the case a) of compound adjectives
(e.g. the A+A (bitter-sweet), A+N (It. giallo limone ‘lemon yellow’), V+A (Du.
druipnat 'drip-wet') and N+A (girl crazy) types), b) of compounds containing i) adverbs
(It. sottosopra 'upside down'), ii) prepositions (Fr. sans papier lit. ‘without documents’),
iii) pronouns (self-determination), iv) particles (make-up) or v) verbs (It. portacenere
'ash tray').
A quick look at the data analyzed in the framework of the Morbo/Comp project
reveals that besides the structures [N+N] and [A+N], the world’s languages can have
many other compounding structures. Besides the ones just mentioned, compound
structures such as [V+V]V, [V+V]N, [A+V]V, [N+V]N, [N+V]V, [V+N]N, [V+N]V,
[Pro+N]N, [Pro+V]V, [P+A]A, [Adv+V]V etc., can be found whose properties have been
only sporadically studied. Among them, there are highly productive structures, as is the
case for Italian V+N compounds.

1.3. Classificatory criteria

The most salient problem in the classification of compounds has to do with the
heterogeneous nature of the criteria adopted. It is to this problem that we will devote our
attention in the following section.
To illustrate the point, let us take into account the notions of
endocentricity/exocentricity and the notion of coordination. The first two notions define
compounds on the basis of the presence vs. absence of a head constituent; the notion of
coordination is based on the grammatical relation holding between the two constituents
of the compound. However, if we do not allow any kind of intersection of the different
notions involved, we would be forced to say, for example, that coordination has no
relation whatsoever with endocentricity/exocentricity. But, as we will see below,
intersection of defining criteria is fundamental in order to obtain a descriptively
adequate classification.

1.3.1. Previous and current classifications

In order to illustrate the problems just mentioned, we will take into account an
indicative if not exhaustive survey of some different proposals (that we illustrate in (1)
making use of trees)5:

1)
a) Bloomfield (1933)

constructions

exocentric endocentric

subord coord

loudmouth love story bittersweet

(bahuvrihi) (tatpurusa) (dvandva)

b) Bally (1950)

‘de coordination’ ‘d’accord’ ‘de rection’

sourd-muet chaleur solaire maison de campagne

5
The representations given in (1a-i) are not taken in this form from the quoted works, but are the result of
our interpretation, which in some cases is undoubtedly partial and schematic. In general, quoted works do
not intend to propose a real classification of compounds but they represent, however, a state of the art of
our knowledge on the topic.
c) Marchand (1969) (only endocentric compounds)

synthetic compounds non verbal-nexus compounds

copula c. rectional

subsumptive attributive additive steamboat

oak tree girlfriend (dvandva)
blackboard fighter b.
Austria-H.

d) Spencer (1991)

endocentric exocentric (bahuvrihi) dvandva

head-modifier predicate-argument mother-child
(attributive?) pickpocket Austria-Hungary
appositional
learner-driver
e) Fabb (1998)

no head one head two heads

exocentric endocentric (co-ordinate/appositional/dvandva)

f) Olsen (2001)

determinative copulative possessive

coffee cup poet-doctor greybeard

g) Haspelmath (2002)

endocentric exocentric affix comp. coord. (additive) appositional

lipstick lavapiatti green-eyed elun-ai poeta pintor
‘adult and child’
h) Bauer (2001)6

tatpurusa dvandva bahuvrihi synthetic c.

determinative/endoc. copulative possessive (more recent)
karmadharaya aggregative exocentric
coordinative Kickkopf
Schl.Holstein

A+N N+N
blackbird woman doctor

i) Booij (2005)

endocentric exocentric bahuvrihi copulative

lavapiatti Kahlkopf
auricomus
dvandva7 appositive
candra-ditya-u Fürstbischof
‘moon-sun-DUAL’

Let’s first examine a set of proposals that seem grosso modo equivalent to each
other, namely those of Spencer, Haspelmath, Booij and Bauer.
As can be easily observed, all four classifications consider the presence/absence
of a head constituent as a criterion of the same level as, for instance, copulative (Booij
and Bauer), dvandva (Spencer), coordinative and appositional (Haspelmath). In other
terms, these scholars seem to set apart endocentricity and exocentricity, not allowing
these two notions to extend across classes. Spencer, for instance, proposes that
compounds are classed into 3 groups: a) endocentric head–modifier constructions –
maybe also containing attributive compounds; b) exocentric (bahuvrihi) predicate-

6
Bauer’s paper is a typical example of what we have said in footnote 5. Bauer in fact acknowledges that
current classifications are problematic (for instance, in his view, a compound such as woman doctor
should be considered as coordinative and not as karmadharaya). On the other hand, Bauer’s paper has
typological rather than classificatory aims.
7
Booij points out that «the copulative/appositive compounds [such as Fürstbishof] are different from
dvandva compounds because their number is singular».
argument formations and c) dvandva compounds – group perhaps including
appositional constructions. But, separating endocentric and exocentric compounds from
dvandvas has the undesired consequence that the latter seem to be unanalyzable on the
basis of the presence or absence of the head. But dvandvas and appositionals are
different just in that the former are exocentric and the latter endocentric. Not only head-
modifier compounds, in fact, are endocentric and not only predicate-argument ones are
exocentric.
Also Haspelmath's groups (endocentric, exocentric, coordinate, appositive and
«affixed compounds») seem to obscure the fact that both affixed and coordinate
(additive) compounds of the type adult-child are exocentric while appositional ones are
endocentric.
In Booij's arrangement, though dvandva compounds of the Sanskrit type (with
dual or plural inflection) are correctly separated from copulatives (which have singular
number) the separation of endocentric and exocentric compounds from bahuvrihis and
copulatives causes redundancy in that dvandva copulative compounds are exocentric
while appositive copulative compounds are endocentric.
Similar observations can be made about the classification extrapolated from
Bauer's work. Also in Bauer's taxonomy the notion of head does not apply with the due
extension to all types of compounds.
Besides the problem we have just seen, which derives from the layering of
notions like endocentricity and exocentricity on a par with other notions like dvandva
and synthetic, a problem of inconsistency arises from these classifications.
Consider, for example, Haspelmath's proposal. His classification uses different
classifying criteria: a) presence/absence of a head (giving rise to the distinction between
endocentric and exocentric compounds), b) formal structure of compounds (introducing
a class of «affixed compounds») and c) syntactic-semantic relation between constituents
(determining the class of appositional compounds). As a consequence, it is not easy to
understand whether or not criterion (a), i.e. presence/absence of a head, can be applied
to compounds classed on the basis of criterion (b), i.e. formal structure, or whether there
is any possible relation between affixed compounds and appositional ones.
Different observations can be made about Olsen's and Fabb's proposals. Fabb's
classification, though consistent in the sense that it makes use of a single criterion –
number of heads – is too restrictive if confronted with the variety of attested
compounds.
Olsen’s classification has the advantage of using the notion of determinative as
opposed to the notion of coordinative compounds. However, besides these two classes,
Olsen introduces the class of (exocentric) possessive compounds, which is clearly a
class based on a different criterion. An undesired consequence of this mixture of criteria
is that it is not clear whether or not the notion endocentric/exocentric does apply to
determinative and copulative compounds. Furthermore, while in determinative and
copulative compounds the relevant relation is the one between the two constituents, in
possessive compounds the «possessive» relation is the one between the whole
compound and the absent head.
In passing, it can be noted that a label is not always used to indicate the same
type of compounds (though this is normal in any scientific taxonomy). For example,
what Olsen calls copulative compound (i.e. poet-doctor) is called appositional by
Haspelmath and Spencer (i.e. poeta-pintor, learner-driver).
Let us now have a look at some previous «traditional» classifications.
Bloomfield used consistently the notions of «subordinate» and «coordinate»; however,
the structure of his classification misses the fact that both subordinate and coordinate
compounds can be exocentric.
Marchand's proposal is articulated, rich and based on consistent criteria but it is
applicable only to endocentric compounds. Marchand, in fact, maintains that what is
known in the literature as exocentric compounds are to be analyzed as containing some
sort of (categorizing) zero suffix and, as such, they are formations not pertaining to the
compounding domain but to derivation.
Finally, the classification proposed by Bally seems to be based on a unique
criterion, that of the grammatical relations holding between head and non-head
constituents, viz. relations ‘de coordination’, ‘d’accord’ and ‘de rection’. This proposal
is consistent because it is based on a single criterion but, unfortunately, it is insufficient
because of the absence of the notion «head». And, in some sense, we could say that
Bally’s and Fabb’s proposals are complementary. A further problem with Bally’s
proposal, however, is the fact that included in the domain of compounds there are forms
that today would be clearly considered as phrases (maison de campagne).
2. A new proposal

We propose a novel classification of compounds which is based on a very simple

assumption. What is special about compounds is the fact that the two constituents are
linked by a grammatical relation which is not overtly expressed (cf. apron string vs.
string of the apron). Therefore, we would like to suggest that the classification of
compounds be uniquely and consistently based on this criterion8. The possible
grammatical relations holding between the two constituents of a compound are basically
the relations that hold in syntactic constructions: subordination, coordination and
attribution. The classification of compounds we propose is thus the following:

2) compounds

subordinate attributive coordinate

endo exo endo exo endo exo

Compoundss are classified as ‘subordinate’ whenever there is a ‘complement’

relation between the two constituents. In a compound such as taxi driver, taxi is clearly
the «complement» of the deverbal head. We argue that this is also the case in
compounds such as apron string, where apron is in an «of relation» with string.
Complement relations, however, are not exclusively ‘of relations’: apron string, in fact,
can have different interpretations: ‘string of an apron’, ‘string on a apron’, ‘string in a
apron’, etc. Nevertheless, the relation between the two constituents is always a
complement relation, namely a subordinative one.

8
This position is not completely new. For example Marchand (1969:18) observed that all compounds can
be explained on the basis of the syntactic relations underlying the corresponding sentences. This
observation is interesting, even though not workable within the lexicalist framework which rejects
derivation of compounds from sentences.
 Note that there is a clear subordination relation also when the head is not
present, as in cut throat or in lavapiatti and not only when the compound is endocentric
as taxi driver.
Attributive compounds are formed either by a noun and an adjective, as in blue
cheese (where the adjective expresses a property and is in a modifier relation to the
noun) or by two nouns, where the non-head very often is used somehow metaphorically,
expressing an attribute of the head (cf. snail mail, sword fish).
Coordinate compounds are those formations whose constituents are tied by the
conjunction «and». They are potentially recursive even in Romance languages (cf. It.
poeta pittore regista 'poet-painter-director') where recursion in compounds is not usual.
From a semantic point of view, such compounds can be considered as having two heads
(poet painter is both a «poet» and a «painter»).
Compounds of these three classes can be both endocentric and exocentric.
Therefore, all the compounds exemplified in (1) can be accommodated in six classes, as
illustrated below:

3)
SUBORDINATE ATTRIB COORDINATE
endo exo endo exo endo exo
love story loudmouth bitter sweet Bl.
steamboat blackboard oaktree Austria-H. March.
girlfriend
fighter-bomber
pickpocket learned-driver mother-child Sp.
coffee cup greybeard poet doctor Olsen
lipstick lavapiatti greeneyed poeta pintor elun-ai Hasp.
blackbird Kickkopf woman doctor Sch.H Bauer
lavapiatti Kahlkopf Fürstbischof candra-ditya-u Booij
auricomus

Since the compounds in (3) are not representative of all possible types of compounds
(note, for example that under the column ‘attributive/endocentric there are no
compounds with a Noun as non-head) we will enrich the list with some other examples,
as in (4):

(4)

SUBORDINATE ATTRIB COORDINATE

endo exo9 endo exo endo exo

apple cake kill joy blue cheese white collar actor author mind brain
brain death cut throat atomic bomb green house priest hermit mother child
finger print back yard pale face singer bassist north east
mail man French kiss long legs dancer singer
sun glasses ape man free lance artist designer
water pipe ghost writer fighter bomber
taxi driver key word king emperor
stone cutter public opinion merchant tailor
arm control sword fish
baby care
agora phobia

In a framework like the one we are proposing, neoclassical compounds -

involving that particular kind of elements called «semi-words» (sW) (Scalise 1984) -
are to be considered as subordinate. In this way, the use of a label which is only
«descriptive» of the origin of constituents can be avoided. Subordination is, in fact, the
relationship tying together the two constituents of these formations, as the examples in
(5) show:

5) neoclassical compounds

sW+sW sW+N N+sW

calligraphy apiculture aerodrome
hydrology hydrophobia insecticide
philosophy antropomorphism parkingmeter

In fact, hydrology is the ‘science of water’, hydrophobia is the ‘fear of water’, and so
on. Also the so-called phrasal compounds can be inserted into one of the proposed
classes. Consider the following examples:

6) phrasal compounds

[floor of a birdcage] taste

[punch in the stomach] effect
[pipe and slipper] husband

9
The low figures of exocentric compounds is intended to reflect the fact that in Germanic languages
exocentricity is not so widespread as in Romance languages. This is probably due to the fact that
Germanic languages are head final. In a Romance language such as Italian, on the contrary, the most
productive type of compounds is the exocentric V+N compound (like scacciapensieri).
 According to the criterion underlying our classification, these compounds have
to be seen as belonging to the attributive class: in fact, the non-head phrases involved in
these constructions function as properties qualifying the head nouns (a terrible taste, a
painful effect, a boring husband). In these compounds the non head has a metaphorical
interpretation: to say that something created a «punch in the stomach effect» on you
means that you can feel like having received a punch in the stomach but you do not
really need to actually have received it; and the meaning does not change if instead of
«pipe and slippers» a husband is of the kind «newspapers and TV».

2.1. Selection

The distinction we draw among subordinate, coordinate and attributive

compounds can be further supported by the modality by which the head selects the non
head in each of these three classes (cf. Scalise, Bisetto, Guevara 2005). If we represent
compounds making use (in a slightly modified way) of Lieber’s (2004) theory of lexical
semantics representation, according to which every lexeme is represented with a
skeleton (containing grammatical information) and a body (containing encyclopaedic
information), the following picture can be drawn (arrows indicate matching features;
heads are underlined):

(7) selection in compounds

a) coordinate compounds
actor director
[Thing[+com, –abst, +an]([x],[ ]),[Event ACT [x])]] [Thing[+com, –abst, +an]([x],[ ]), [Event DIRECT [x])]]
<human, professional> <human, professional>
<show business> <show business>
<works in theatres, films, etc.> <works in theatres, films, etc.>
<receives directions> <gives directions>
<…> <…>

b) N+N subordinate compounds

apple cake
[Thing [+com, –abst, –an] ([ ])] [Thing [+com, –abst, –an] ([ ])]
<physical> <physical>
 <shape> <shape>
<edible> <edible>
<can be an ingredient> <made with ingredients>
< .... > <baked>
<made for parties>
c) Attributive compounds
snail mail
[Thing [+com, –abst, +an] ([ ])] [Thing [+com, –abst, –an] ([ ])]
<gastropod> <institution>
<secretes slime> <means of communication>
<very slow> <takes time>
<….> <….>

Coordinate compounds (in 7a) are characterized by a virtual identity on both

levels of representation: matching of the skeletons of the two constituents and a high
level of matching features in the encyclopaedic body.
On the contrary, in subordinate compounds like in 7b), both endocentric and
exocentric, the skeleton seems not to play a significant role: what really matters in this
type of compound is the set of encyclopaedic features of the two bodies. At least one of
the features of the head constituent must be matched by the encyclopaedic features
characterizing the non-head constituent (such as <edible> and <made with ingredients>
in the example above).
In attributive compounds with a noun as non-head the skeleton plays no
significant role (as in subordinate compounds). What matters is that the non-head
matches at least one of the encyclopaedic features of the head. The only information
pertaining to the non-head that is present in the output is the matched feature: the rest of
the information is ignored (for instance, the feature <gastropod> is not present at all in
snail mail). The non-head has the sole function of specifying an attribute of the head’s
body (in this case «slowness»). In other words, the non-head is almost «adjectival»: in
LCS terms, it is interpreted as a Property and not any longer as a Thing (two lexical-
conceptual categories, related to, but not to be confused with, the syntactic categories
Noun and Adjective, cf. Jackendoff 1990: 43–58). The attributive relation is obviously
self-evident if the non-head is an Adjective (e.g. blackboard).
 Different types of compounds have thus a different mechanism according to
which the non head constituent is selected.

3. Some residual problems and some provisional conclusions

The proposed classification is generally clear and it seems to work properly. There are
of course borderline cases where the analysis is more complicated. For example,
compounds such as greybeard or greeneyed have been classified here as attributive
(exocentric). Now, the attributive relation is indeed present between the two free
constituents of this structure (grey-beard; green eye). But if one takes into consideration
their whole structure, it can be observed that in grey beard, for example, the relationship
between [grey beard] and the (non realized) external head is not the same. This is also
true for green eyed, where besides the attributive relation between green and eye, there
is another grammatical relation between the (realized) head -ed and green eye, probably
a subordinative relationship. There is, then, a case where two different types of
relationships are present in the same compound: which of the two do we consider
primary?
The proposed classification considers a first level of analysis, that is the grammatical
relationship between the two constituents. We do believe that this first step is basic and
that it should be kept separated from other possible criteria such as the internal
structure, the semantic relation between the constituents, the origin of compound
constituents or the categorial status of the constituents;, all these criteria have to be
ordered, so to speak, after the grammatical level of classification.
Attributive compounds, for example, do not constitute a homogeneous class because the
non-head can be either an adjective (blackbird), or a noun (sword fish) or a verb (play
ground). This means that the classification could be possibly enriched by a further
subdivision based on categories.
It is also likely that within each of the three identified macrocategories there will be
need for further semantic analysis in order to capture Marchand's intuition according to
which what he calls copulative compounds can be further distinguished into, for
example, subsumptive and additive compounds. In our classification, we do not
distinguish among coordinate compounds (they all are linked by an «and» operator) but
the semantic relation in subsumptive and additive compounds is clearly not identical
(hyponymy in one case, synonymy in the other).
In conclusion, in this paper we have suggested that an adequate classification of
compounds has to be done primarily and consistently on grammatical grounds. And
exactly the grammatical relations between the constituents of a compound can allow an
homogeneous grouping (at least on a first level) of compounds of different languages.
However, we do not exclude (to the contrary) that there could be a second level of
analysis, based on more subtle distinctions.

References

Allen M.R. (1978), Morphological Investigations, PhD Dissertation, University of

Connecticut.
Bally,C. (1944), Linguistique générale et linguistique francaise, 2. ed. Berne,
A.Francke.
Bauer, L. (2001), Compounding, in Haspelmath, M. et al. (eds.), Language Typology
and Language Universals, Vol. I, Berlin, Walter de Gruyter, 695-707.
Bloomfield, L. (1933), Language, New York, Holt.
Booij G. (2005), The Grammar of Words, Oxford, Oxford University Press.
Fabb, N. (1998), Compounding, in Spencer A. and A. M. Zwicky (eds.) Handbook of
Morphology, Oxford, Blackwell, 66-83.
Hacken ten, P. (1994), Defining Morphology. A Principled Approach to Determining
the Boundaries of Compounding, Derivation, and Inflection, Hildesheim, Olms.
Haspelmath, M. (2002), Understanding Morphology, London, Arnold.
Jackendoff, R. (1990), Semantic Structures, Cambridge (MA), MIT Press.
Lieber, R. (1992), Deconstructing Morphology, Chicago-London,The University of
Chicago Press.
Lieber, R. (2004), Morphology and Lexical Semantics, Cambridge, Cambridge
University Press.
Marchand, H. (1969), The Categories and Types of Present Day English Word
Formation, München, Beck’sche Verlagsbuchhandlung.
Olsen, S. (2001), Copulative Compounds. A Closer Look at the Interface Between
Morphology and Syntax, in «Yearbook of Morphology 2000», 279-320.
Scalise S. (1984), Generative Morphology, Dordrecht, Foris.
Scalise S. (ed.) (1992), The Morphology of Compounding, special issue of «Rivista di
Linguistica», 4 (1).
 Scalise S., A. Bisetto and E. Guevara (2005), Selection in Compounding and
Derivation, in Dressler W. U., D. et al. (eds.), Morphology and its Demarcation,
Amsterdam/Philadelphia, John Benjamins Publishing Company, 133-50.
Selkirk, E. (1982), The Syntax of Words, Cambridge (MA), MIT Press.
Spencer, A. (1991), Morphological Theory, Oxford, Blackwell.

SUMMARY

In questo articolo si discute della classificazione dei composti a partire dalle diverse
analisi che ne sono state proposte, in particolare in questi ultimi anni. I problemi che le
classificazioni correnti pongono sono sia di terminologia che di coerenza di criteri
classificatori. In questo lavoro si propone una classificazione «di primo livello» che
utilizzi un criterio unico, quello del rapporto grammaticale tra costituenti. In questo
modo vengono identificati tre macroclassi di composti: subordinati, coordinati ed
attributivi. Le nozioni di endocentricità ed esocentricità, abitualmente considerate come
criteri di primo livello, sono in realtà subordinate rispetto al criterio grammaticale.

View publication stats