Compounds and multi-word expressions in Greek

Complex lexical units include compounds as well as multi-word expressions displaying mixed morphosyntactic properties.1 These mixed properties are determined by language-specific characteristics. Moreover, a diversity of properties is observed among the different types of multi-word expressions; in some cases even within the same type of structure. Therefore, their status is rather unclear, as is also revealed by the strong name variation among scholars (Hüning/Schlücker 2015: 450 f.), even within the same language. The different naming suggestions cannot be considered as one-to-one equivalents or synonyms. The selection of one of them is also determined by the theoretical approach adopted. Specifically, the selection or the creation of a new label depends on the type of grammatical model as well as on the role of the lexicon to the formation of new lexical units. Multi-word expressions in Greek have caught the attention of linguists in the twentieth century. This type of lexical unit has been used more often in the form of loan translations from English and French (Anastassiadis-Symeonidis 1986, 1994). Since then it has been rather prominent in many terminological domains as well as in media language. Moreover, it constitutes a commonly selected formation type of lexical units for the naming of new concepts or the translation of borrowed terms gaining ground over the formation of typical compounds. The phenomenon of terminological variation regarding multi-word expressions is also apparent in the literature of Greek. Different names that have been suggested among scholars are for instance lexical phrases (Anastassiadis-Symeonidis 1986; Ralli 1991), multi-word compounds (Ralli 1992; Anastassiadis-Symeonidis 1996; Christofidou 1997; Ralli/Stavrou 1998) and loose multi-word com­ pounds (Ralli 2005, 2007; Koliopoulou 2006, 2008, 2009). Ralli (2013a, 2013b; cf. also Bağriaçik/Ralli 2015) adopts in her later studies the term phrasal compounds, inspired by Booij’s (2009, 2010: 169–192) term phrasal names, in order to differen-


Introduction
Complex lexical units include compounds as well as multi-word expressions displaying mixed morphosyntactic properties. 1 These mixed properties are determined by language-specific characteristics. Moreover, a diversity of properties is observed among the different types of multi-word expressions; in some cases even within the same type of structure. Therefore, their status is rather unclear, as is also revealed by the strong name variation among scholars (Hüning/Schlücker 2015: 450 f.), even within the same language. The different naming suggestions cannot be considered as one-to-one equivalents or synonyms. The selection of one of them is also determined by the theoretical approach adopted. Specifically, the selection or the creation of a new label depends on the type of grammatical model as well as on the role of the lexicon to the formation of new lexical units.
Multi-word expressions in Greek have caught the attention of linguists in the twentieth century. This type of lexical unit has been used more often in the form of loan translations from English and French (Anastassiadis-Symeonidis 1986, 1994. Since then it has been rather prominent in many terminological domains as well as in media language. Moreover, it constitutes a commonly selected formation type of lexical units for the naming of new concepts or the translation of borrowed terms gaining ground over the formation of typical compounds. The phenomenon of terminological variation regarding multi-word expressions is also apparent in the literature of Greek. Different names that have been suggested among scholars are for instance lexical phrases (Anastassiadis-Symeonidis 1986; Ralli 1991), multi-word compounds (Ralli 1992;Anastassiadis-Symeonidis 1996;Christofidou 1997;Ralli/Stavrou 1998) and loose multi-word com pounds (Ralli 2005(Ralli , 2007Koliopoulou 2006Koliopoulou , 2008Koliopoulou , 2009. Ralli (2013aRalli ( , 2013bcf. also Bağriaçik/Ralli 2015) adopts in her later studies the term phrasal compounds, inspired by Booij's (2009Booij's ( , 2010 term phrasal names, in order to differen-tiate specific types of complex lexical units from typical one-word compounds which are morphological objects. However, the use of the term phrasal compound to refer to this type of structure can be misleading, since it is also used to denote another kind of structure, namely compounds with a phrasal element at the nonhead position, like chicken and egg situation in English. Such structures are not possible in Greek (cf. Section 2.1).
In this study, I adopt the term multi-word expression as a term that is general and theory-neutral -also suggested by Hüning/Schlücker (2015: 451) -to refer to different types of complex lexical units in Greek sharing morphological and syntactic features in various proportions. The aim of this study is to analyze their complicated properties and compare them to typical compounds without letting theoretical considerations override the data. After having analyzed in detail the different types of multi-word expressions in Greek, I will come back to more theoretical considerations regarding their interrelation with other comparable lexical units as well as their locus of realization in grammar.
Specifically, this study is structured as follows: Section 2 gives an overview of various complex lexical units found in Greek. Typical compounds, multi-word expressions as well as phrase-like structures are analyzed in detail and compared to each other. Section 3 discusses the interrelation between the various types arguing that they coexist in the lexicon as complementary resources of nominal naming units. However, coexistence in the lexicon does not exclude competition among types. Section 4 deals with the question of how complex lexical units can be accounted for in the lexicon and in grammar. Finally, Section 5 summarizes the conclusions.

Typical compounds vs. other complex lexical units
Compounding can be considered as the output of a morphological operation situated closer to syntax than any other morphological formation (Scalise 1992: 4). As a result of this closeness, it is sometimes rather difficult to differentiate compounds from phrases 2 , and even more from intermediate structures displaying a 2 Many studies have been carried out on the distinction between compounds and phrases based on selected criteria mostly concerning the formal properties of a compound con trary to those of a syntactic phrase (e. g. Borer 1988;Scalise 1992;ten Hacken 1994;Bisetto/Scalise 1999;Bauer 2001;Olsen 2001;Donalies 2004;Gaeta/Ricca 2009;Schlücker/Hüning 2009

Typical compounds
Compounding is one of the most productive morphological processes in Greek. One-word compound formations mostly built up from stems are found in both spoken and written language in various types of texts. The spontaneous creation of compounds that in some cases succeed to be established and to enter the speakers' mental lexicon is not rare. Compounds in Greek involve all major lexical categories, namely nouns (1), adjectives (2) and verbs (3) Nominal compounds consisting of two nouns are the most productive ones (1), as in a number of other languages, for instance in German. Verbal compounding is very productive in Greek in comparison to other European languages, either in the form of determinative structures, as in (3), or in the form of coordinative structures (e. g. πηγαινοέρχομαι 'come and go'). In German, for instance, the limited number of verbal compounds is the result of a backformation process from nominal compounds (Becker 1992: 20 f.;Günther 1997: 6). With regard to their structural properties, compounds in Greek usually consisting of stems form one phonological word written as one graphemic unit (cf. (1)- (3)). This phonological word has one main stress assigned either on the antepenultimate syllable of the entire compound formation (1), or on the regular stress position of the right-hand constituent (2, 3). Stress assignment is determined by two specific phonological rules applicable to all compound formations (Nespor/Ralli 1994: 201, 1996. The form of these rules will not concern us here. Moreover, compounds in Greek constitute one morphological unit, to which syntactic operations do not have access. In the following, I contrast the properties displayed by a compound formation (4) with those of a syntactic phrase (5), both consisting of an adjective and a noun, so that the analysis is comparable. The first indication of the word atomicity displayed by compounds is related to the fact that word internal inflection is not allowed (4b), contrary to syntactic phrases, whose components are inflected. With regard to the last test, according to which the order of the constituents of syntactic phrases can be reversed (8b), it should be mentioned that this possibility increases the emphasis on the syntactic phrase. Specifically, the property designated by the adjective is highlighted by this stylistic variation (ten Hacken/Koliopoulou 2016: 131 f.). On the contrary, the word order of compound components is fixed (8a). Even in compounds consisting of components with the same lexical category, like noun-noun compounds, the change of the order of the two components is -at least in Standard Modern Greek -ungrammatical (cf. (1) κεφαλόσκα λο/*σκαλοκέφαλο 'upper/wider step').
A further distinctive characteristic is related to the type of constituents participating in syntactic phrases or compounds. Phrases consist of words while compounds in Greek usually consist of stems. However, the possibility of a word constituent in one of the two positions or even in both positions of a compound formation cannot be excluded. Since both types of free lexemes can occupy any constituent position, four structural patterns result from all possible combinations (cf. Ralli 2005Ralli : 237 f., 2013aKoliopoulou 2013: 24 f.). The most productive pattern is that of stem-stem formations (9a), since stem-constituents are preferred in Greek compounds. The preference for a specific type of constituent in the compound formations constitutes an important parameter determining various structural characteristics of compounds (cf. Koliopoulou 2013Koliopoulou , 2014a, among others the possibility of the appearance of a linking element. Specifically, in case the first constituent is a stem (9a), (9b), the two constituents are linked to each other with the element -o-. 6 Its appearance is obligatory and rather systematic, motivated by the fact that compound constituents are usually stems.
The impact of the "word-vs. stem-based parameter" in compounding becomes obvious if we compare Greek with German compounds. 7 German compounds are mostly built out of words, without excluding cases of a stem constituent in the first position of the compound, as in Stimmabgabe ('voting', Stimm(e) 'vote/voice', Abgabe 'delivery'). They are also characterized by the appearance of a linking element as for instance Arbeit-s-ablauf ('workflow').
However, the linking element in German compounds displays very different properties compared to the Greek linking element -o- (Koliopoulou 2014b). Therefore, the appearance of a linking element is not systematic, its form is variable, while compounds without linking element are very productive (e. g. Stimm-AE-ab gabe 'voting').
The preference of a particular language to build compounds out of words or stems affects further characteristics of the compounding process, for instance the possibility of recursion. Specifically, compounds in German tend to be expanded through recursion either in the non-head or the head position, as shown in (10a) and (10b) respectively (cf. Bauer 2009: 350;Neef 2009: 386;Koliopoulou 2017: 123). By contrast, recursion in Greek compounds, as illustrated in (11) The difference in the degree of recursion between German and Greek compounds is related to the type of constituents. Specifically, I argue that stem constituents exhibit more restrictions than word constituents, whose more independent character allows the connection with further compound members, either in the head or in the non-head position.

Multi-word expressions
Another type of complex lexical unit composed of free morphemes are multi-word expressions. These peculiar structures, found also in Greek, have been already studied by many scholars (cf. literature mentioned in Section 1 as well as Anastassiadis-Symeonidis 1986: 138-143, 203 ff.; Koliopoulou 2012: 862) in comparison to one-word compounds, to syntactic phrases and even to each other, since the various types display different characteristics. Specifically, contrary to typical compounds constituting one morphological word to which syntax has no access, multi-word expressions in Greek are structures with some morphological properties (cf. Section 2.2.1) without though preventing syntax from having access to their internal structure. They can be considered as intermediate structures, since they behave similarly to one-word compounds, but they also bear features typical for syntactic phrases. Their mixed properties vary not only within the different types of intermediate structures, but in some cases even among the various examples of the same type (cf. (26)- (27)). Specifically, multi-word expressions in Greek are nominal structures 8 composed either of an inflected adjective and a noun or of two nouns. They look like syntactic phrases since their components are independent phonological words, contrary to one-word compounds constituting a single phonological word, regardless of the type of the compound constituents. Moreover, multi-word expressions consist of two inflected words. Compounds, by contrast, are usually formed out of stems linked by the element -o-. Compound formations are inflected at the right edge of the structure.
To be more specific, there are four types of multi-word structures 9 : a) [A N] expressions composed of an inflected adjective and a noun (12)  Multi-word expressions constitute naming units, many of them displaying an idiomatic meaning. The degree of semantic opacity is in some cases comparable to that of typical compounds. Consider, for instance, the example μαύρη αγορά ('black market', (12c)), denoting a very specific type of market, or the example άρση βαρών ('weightlifting', (13d)) denoting an athletic discipline. However, as stated e. g. by Gaeta/Ricca (2009: 36), the semantic criterion is unreliable and can even be misleading for the demarcation between morphological and syntactic structures (cf. Section 2.2.1). Therefore, the present analysis is mainly based on formal criteria. With regard to headedness, all four types display the same order as comparable adjective-noun (16a) and noun-noun syntactic phrases (16b). A further property that some multi-word expressions share with typical compounds is that they can be input to a derivation process, specifically to suffixation (cf. Koliopoulou 2006Koliopoulou : 49, 2009Koliopoulou : 62, 2012Ralli 2007: 232 f., 2013a: 247 f., 266; ten Hacken/Koliopoulou 2016: 132 f.). Specifically, one-word compounds, regardless of the type and the lexical category of constituents they consist of, can become bases for derivational suffixation, cf. (17). The most common derivational suffix added to a complex base is the adjectival suffix -ικ(ός), as shown in the examples below.  (13) can also be input to a suffixation process, as shown in (19). The suffixes that take part in this derivational process are the adjectival suffix -ικ(ός) and the nominal suffixes -ίτ(ης) and -ίστ(ας). The possibility to become input to derivation is not applicable to all [A N] or [N N GEN ] expressions. Μεγάλη οθόνη ('cinema', (12d)) or αγορά εργασίας ('job market', (13a)), for instance, cannot be input to any derivation process. Although there are no certain criteria determining which structures can participate to further derivation processes, it can be argued that these structures share more morphological features with typical compounds.
[N N App. ] structures are different from the other types of multi-word expressions with regard to headedness. Particularly, the two components share the same formal and semantic properties and thus the head properties as well. Since the two components display the same lexical category, it is possible to reverse their order, as shown in (20) 2009: 30). In the case of Greek, there is a clear demarcation between the two types of formation since coordinative compounds constitute one phonological and morphological word (21). In contrast, [N N App. ] expressions consist of two phonologically and morphologically independent words. Moreover, coordinative compounds are not characterized by an appositional relation between the components. The most common type of semantic relation found in Greek coordinative compounds is the additive one (Ralli 2007(Ralli : 80 f., 98, 2013aKoliopoulou 2013: 297 ff.).
Since multi-word expressions always consist of two inflected words, they do not display the morphological properties of one-word compounds (cf. Anastassiadis-Symeonidis 1986: 149, 174, 196). Particularly, the inflected components of [A N] expressions agree in gender, case and number, as shown in (22)

Syntactic fixedness
Despite the fact that multi-word expressions share basic properties with regular syntactic phrases, they share many properties with typical compounds as well. Specifically, all four types of multi-word expression in Greek display a certain degree of syntactic fixedness. Some expressions are more restricted than others with regard to the degree of access to syntactic operations, as illustrated by the result of applying a number of tests concerning their internal properties i. e. their degree of lexical integrity (Anderson 1992: 84). Their mixed morpho-syntactic properties have been studied in detail (Anastassiadis-Symeonidis 1986, 1994, 1996Ralli 1991Ralli , 1992Ralli , 2005Ralli , 2007Ralli , 2013aRalli , 2013bChristofidou 1997;Ralli/Stavrou 1998;Koliopoulou 2006Koliopoulou : 43-56, 2008Koliopoulou , 2009Koliopoulou , 2012; Bağriaçik/Ralli 2015; ten Hacken/Koliopoulou 2016). In most of these studies, the degree of lexical integrity of the multi-word expressions has been analyzed on the basis of diagnostic tests exploring how many properties they share with regular syntactic formations. In the following, I use the tests applied to typical compounds (cf. (6)-(8)) in the previous section in order to determine the degree of syntactic fixedness displayed by the different types of multi-word expressions found in Greek (cf. (12)-(15)). Moreover, I use an additional test regarding the possibility of adjective-noun syntactic phrases to double the definite article for emphatic reasons, which is only applicable to [A N] expressions. I summarize the tests under (28) It is obvious from the negative response of the [A N] expression to all diagnostic tests that the structure displays a certain degree of lexical autonomy, contrary to the corresponding syntactic phrase, which allows access of all syntactic operations to its structure.
In (30), I test the structural properties of the [N N GEN ] expressions by applying the tests (28a-c). Specifically, I take as an example the expression αγορά εργασίας ('job market', (13a)) contrastively to the syntactic phrase αναζήτηση εργασίας ('job search') which bears the same non-head (cf. Koliopoulou 2009: 63;Ralli 2013a: 248 ] expressions to the same tests is not different from that of the structures tested above, as illustrated in the following on the basis of the example φόρος φωτιά ('very high tax', (14c)).
(31a) *φόρος μεγάλη φωτιά foros megali fotia tax big fire (31b) *φόρος φωτιά και καπνός foros fotia ke kapnos tax fire and smoke (31c) *φωτιά φόρος fotia foros fire tax With regard to the possibility of reversing the order of the constituents, most of the examples belonging to this type of expression have a negative response, proven by (31c) as well as by (32a΄-c΄). However, there are a few exceptions, e. g. (32d΄), in which the inversion of the two constituents is allowed (Koliopoulou 2006(Koliopoulou : 52, 2009 The frequency of use or the degree of semantic compositionality (cf. Fellbaum 2011) are possible parameters that influence the varying degree of syntactic fixedness determining which structure may be characterized by a free word order.
The last type of multi-word expressions displays an appositional relation between the constituents. As already mentioned in the previous section, these expressions are double-headed. Therefore, the tests listed under (31) are almost inapplicable. Specifically, the application of the tests regarding the coordination of compounds (31b) as well as reversing of the order (31c) would not make much sense, since appositional structures are recursive and can be coordinated with further constituents attached to any of two members. Moreover, the order of the constituents is reversible (cf. (20)), since the structures are double-headed, despite the semantic restrictions.
I consider test (28a) in that I check the possibility of independent modification of one of the two constituents, although none of them is a non-head (33a). Moreover, I apply a further diagnostic test in order to investigate the degree of lexical integrity in their internal structure of the [N N App. ] expressions. Specifically, in (33b) I test the possibility of insertion of an uninflected adjective (πρώην 'former'), while in (33c) I test the possibility of insertion of a parenthetical element between the constituents.
(33a) *μεταφραστής ικανός διερμηνέας metafrastis ikanos diermineas lit. translator capable interpreter (33b) ?μεταφραστής πρώην διερμηνέας metafrastis proin diermineas lit. translator former interpreter (33c) ?ο μεταφραστής, όπως βλέπετε, διερμηνέας είναι … o metafrastis, opos vlepete, diermineas ine … lit. the translator, as you see, interpreter is … As illustrated above, the independent modification of one of the two members by a qualifying adjective is not possible, cf. (33a). However, this type of expression displays a limited degree of syntactic fixedness in comparison to the other types of multi-word expressions, since an element can intervene in their internal structure, as shown in (33b) and (33c).

Summary
In ( Greek. Not only with regard to their inflectional properties but also with regard to their behavior in the diagnostic tests, they show the lowest degree of syntactic fixedness among all types of expressions considered in this study. However, they still show some signs of lexical autonomy, according to which their classification as multi-word expressions is justified.

Phrase-like structures
Although there is a clear distinction between typical compounds and multi-word expressions in Greek, the variety of structures sharing properties with compounds as well as with syntactic phrases creates a certain difficulty in differentiating them from each other and classifying them into distinctive types.  (29) and (30). However, due to the argument structure displayed by these structures it can be argued that they are of a different nature from common syntactic phrases. Particularly, their structure resembles that of compounds consisting of a relational adjective and a noun (ten Hacken 1994: 89-98;Bisetto 2010: 65-85). Moreover, they constitute naming units which also supports the view that they are of a different nature than common syntactic phrases. Therefore, they constitute a further type of complex lexical units, which on the one hand differs from regular syntactic phrases, and on the other cannot be classified as belonging to the set of the multi-word expressions analyzed above. Moreover, they display more syntactic properties than the multi-word expressions. Thus, their demarcation from syntactic phrases is a rather difficult task, since it is only based on a few minor distinctive characteristics and not on their response with regard to the diagnostic tests of syntactic fixedness.

Complementation vs. competition
I have argued above for a distinction between three types of complex lexical units in Greek: one-word compounds, multi-word expressions and phrase-like structures. All three constitute nominal structures sharing a function, i. e. to name concepts, particularly complex concepts. Regarding their function, they are clearly different from syntactic phrases, which describe a concept but do not name it. Since they provide further means for naming concepts associated with various terminological areas, the set of naming devices in the nominal domain of the lexicon is extended through their existence. In this sense, the three types of complex lexical units constitute complementary resources of nominal naming units.
Complementation in the lexicon with regard to different naming strategies does not exclude competition among structures. Specifically, typical one-word compounds and multi-word lexical units do not exist in Greek side by side, although this scenario cannot be excluded for all languages. Take for instance lexical units in German (cf. ten Hacken/Koliopoulou 2016), like grüner Tee and Grüntee ('green tee') or schwarzer Markt and Schwarzmarkt ('black market'), coexisting synchronically. Their coexistence is explained by Hüning/Schlücker (2015: 459) on the grounds of stylistic variation and/or diachronic change arguing that the structure schwarzer Markt, for instance, has been gradually replaced by the compound Schwarzmarkt, which is synchronically more frequent than the equivalent phrase.
In Greek, the three types of complex lexical units compete with each other. However, there is no evidence supporting the existence of a blocking mechanism (cf. Rainer 2016), although the formation of typical compounds is much more productive and regular than the formation of multi-word structures. Moreover, I claim that the selection of a possible naming strategy depends on the characteristics of the concept. Specifically, a borrowed nominal concept or a complex concept meant for terminological use is a possible candidate for a type of multi-word lexical unit.

Complex lexical units in lexicon and grammar
In Greek, there is a clear demarcation between compounds and other complex lexical units, i. e. multi-word expressions and phrase-like structures. Among these, compounds are the only type of complex lexical units built in morphology. Taking into consideration the various types of multi-word formations and in some cases their variable features, the question arises how they can be accounted for.
They are neither morphological structures nor regular syntactic phrases; they are rather situated in between. Therefore, multi-word expressions in Greek have been often assigned to a continuum situated between the two components. In this sense, different grammatical models supporting the interaction between the two domains (Kiparsky 1982;Bybee 1985;Borer 1988) have been adopted by many scholars as the most sufficient way to deal with multi-word expressions and their variable features in Greek (cf. Ralli 1991Ralli , 1992Ralli , 2007Ralli/Stavrou 1998;Koliopoulou 2009Koliopoulou : 69, 2012. In a similar context, Ralli (2013aRalli ( : 261 f., 266 ff., 2013b, based on Borer's (2009) analysis of comparable nominal constructs in Hebrew, argues that multi-word expressions in Greek are derived within the syntactic domain which interacts with morphology. Her argument is rather justified, since multi-word expressions and phrase-like structures in Greek look like syntactic phrases that consist of two phonologically and morphologically independent words. However, they are different from regular syntactic phrases, since their structure is not accessible to all syntactic operations. Moreover, they display a certain degree of lexical integrity coinciding in many cases with a non-compositional meaning, also displayed by typical compounds.
The fact that there is strong variation among the different types of multi-word structures with regard to their mixed morphosyntactic properties supports the view that there is no clear borderline between morphology and syntax and that the two domains are situated on a continuum. 11 Multi-word expressions in Greek which display a varying degree of structural visibility to syntactic operations occupy different positions on this continuum.

[A N] and [N N GEN ] expressions in
Greek are clearly nearer to the morphological domain, i. e. to typical compounds, than any other multi-word expression. The fact that some [A N] and [N N GEN ] formations can be input to a derivational process is a further argument in favor of the interaction between morphology and syntax, since structures generated in syntax are turned into one complex stem in order to undergo a morphological operation (cf. (18)-(19)). The other two types of nominal expressions are widespread on the continuum, specifically between the [A N] and [N N GEN ] formations and regular syntactic phrases. Phrase-like structures are situated near to the syntactic domain.
The various approaches that argue in favor of the existence of a continuum between the two grammatical components or the interaction among them are based on the assumption that the two grammatical domains are distinct. Although they may account for structures like multi-word expressions in Greek displaying mixed morphosyntactic properties, they do not throw any light on the grey zone between morphology and syntax. In this respect, the question arises whether the two grammatical domains are actually distinct and if not what kind of demarcation would allow us to differentiate between typical morphological structures, syntactic phrases and intermediate structures.
In order to distinguish compounds from phrases as well as from the in-between formations, Gaeta/Ricca (2009: 38 f.) propose another type of demarcation. They argue in favor of a four-scaled classification based on two criteria: a) 'morphological', i. e. the output of a morphological operation and b) 'lexicalized', i. e. attributed to the lexicon taking into consideration not only idiosyncrasy but also token frequency and/or naming force. In this respect, typical compounds are characterized as [+ morphological] and [+ lexical], whereas syntactic phrases have a negative sign in both properties. Multi-word expressions -or phrase-like units in Gaeta/Ricca's terminology -are non-morphological but lexical units. In this view being a lexical unit is independent from being an output of a morphological operation.
On a similar basis, ten Hacken/Koliopoulou (2016: 134 ff.), dealing with [A N] multi-word expressions in various languages, argue that the main criterion to demarcate [A N] intermediate structures from adjective-noun syntactic phrases is related to the function of these structures. Structures constituting a naming unit are lexical units, while descriptive phrases belong to the syntactic domain.
With regard to Greek, the different types of multi-word expressions and phrase-like structures, despite their varying morphosyntactic features, sometimes even within the same type, share the naming function (cf. Anastassiadis-Symeonidis 1986: 142 f.). They are lexical units with a rule-based formation extending the naming device of the lexicon. This extended view of the lexicon is also supported by approaches such as the Parallel Architecture (cf. Jackendoff 2010) and Construction Morphology (cf. Booij 2010, this volume) on the basis of comparable multi-word, intermediate structures.

Conclusions
The demarcation between the various types of complex lexical units is primarily a language specific matter, although most of the criteria used to differentiate morphological from syntactic structures apply at an abstract level to all languages. It actually depends on the particular characteristics of wordhood and compoundhood, as displayed in each language. These two basic characteristics determine the morphological structures and the lexicon. The degree of resemblance between typical morphological structures and other complex lexical units specifies the form of the lexicon in a particular language and the possibility of interaction between the grammatical domains.
Multi-word expressions and phrase-like structures in Greek are clearly distinct from typical compounds: their constituents are phonologically and morphologically independent words, a linking element is not required, they display head properties similar to syntactic phrases as well as internal inflection. In Greek, the degree of syntactic fixedness depends on the type of expression one deals with. Sometimes, there is variation of the syntactic characteristics even among the different examples of the same type of structure (cf. (26)-(27), (32)). Despite the fact that multi-word expressions and phrase-like structures in Greek cannot be assigned to morphology like typical compounds, all three types of complex lexical units share the same function, i. e. the naming function. They are generated by different lexical unit formation patterns which extend the naming strategies of the lexicon. The outcome of this formation process is lexical units stored in the speakers' mental lexicon.
Compounding in Greek is a very productive process and thus a main language naming device. However, new concepts have been introduced to the language in the last decades through the form of a multi-word expression or a phrase-like structure mostly found in specialized or newspaper texts. The appearance of such a lexical unit is an indication for native speakers of the terminological use of the concept. The emergence of various types of lexical units other than compounds shows a clear tendency to different types of naming units and indicates a silent process of language change regarding the naming of concepts, especially those borrowed from English.