2. What is a compound?

Scholars have provided different definitions for compounds (Marchand Reference Marchand1960: 11, Fabb Reference Fabb, Zwicky and Spencer1998: 66, Olsen Reference Olsen2000: 280, Carstairs-McCarthy Reference Carstairs-McCarthy2002: 59, Ralli Reference Ralli2013: 10). These definitions agree that compounds consist of two or more words, stems, or roots which are combined to form a new complex word, for example school bus, classroom or egghead. Two main types of compounds exist in the relevant literature, based on the presence (endocentric) or absence (exocentric) of a semantic head. An endocentric compound is semantically headed since it denotes a hyponym of the head element in the compound, that is, the semantic head is inside the compound (Lieber Reference Lieber, Štekauer and Lieber2005: 178, Bauer Reference Bauer, Lieber and Štekauer2009: 348), for example sailboat, whereas an exocentric compound is not semantically headed.Footnote ² It does not denote a hyponym of either of its elements; that is, the semantic head is outside the compound, for example scarecrow (Fabb Reference Fabb, Zwicky and Spencer1998: 66-67, Booij Reference Booij2002: 81, Haspelmath Reference Haspelmath2002: 87-89).

In English, endocentric compounds are normally right-headed based on the Right-Hand Head Rule (RHHR), first suggested by Williams (Reference Williams1981: 248), who states that “in morphology, we define the head of a morphologically complex word to be the right-hand member of that word.” For example, a bookshop is a shop. Conversely, Arabic tends to place the head element in endocentric compounds in the initial/left position, as illustrated in (1) (Altakhaineh Reference Altakhaineh2016: 14).

The compound minʃaar lxaʃab “the wood saw” is a hyponym of minʃaar “saw”. To make the compound plural, the word minʃaar ‘saw’ is pluralised rather than lxaʃab ‘the wood’. Therefore, the head of the compound is minʃaar “saw” (Altakhaineh Reference Altakhaineh2016). It has been argued that the position of the head is a parameter; that is, it can be either the right or the left element of a construction in a language (Selkirk Reference Selkirk1982).

Another difference between English and JA is the form of the compound. As mentioned previously, compounds in JA are similar to phrases in form, that is, both types of compounds are SGCs. Some criteria have been proposed to distinguish between the two constructions. Specifically, Altakhaineh (Reference Altakhaineh2019: 33) suggests that a JA compound consists of at least two adjacent words, where the non-head is normally non-referential. In other words, adjacency and referentiality can be used to draw the line between these two constructions: the two elements of compounds are inseparable (3), while those of phrases are separable (2).

Example (2) shows that the elements of a phrase in JA are separable in the sense that an element can be inserted between them (i.e., kull ‘all’), while those of a compounds in (3) are adjacent or inseparable. Regarding the property of referentiality, the non-head of phrases is referential (4), while that of compounds is not (5):

Example (4) shows that the non-head of phrases in JA is referential, as evidenced by the use of the demonstrative haaða ‘this’. In contrast, in (5), the demonstrative cannot be used with the non-head of the compound because it is non-referential.

The semantics of these two constructions is also different. The elements of phrases denote a relationship of possession, whereas those of compounds do not. This relationship can be shown using the word tabaʕ ‘of/for’ which can act as a possessive marker in JA. This maker appears in phrases but not in compounds, especially non-compositional ones, as in (6–7):

Other criteria were found to be useful in differentiating phrases from non-compositional compounds, but not from compositional ones, as follows:

Examples (8–9) show that the head of phrases, such as galam ‘pen’ in (8), and the head of compositional compounds, such as findʒaan ‘cup’ in (9) can be coordinated using the conjunction wu ‘and’ with other nouns, without any syntactic or semantic problems. Conversely, the head of non-compositional compounds cannot be coordinated with other nouns. In addition, coordination can be used as a criterion to differentiate between compositional compounds, on the one hand, and semi-compositional and non-compositional compounds, on the other. The head of semi-compositional and non-compositional compounds cannot be coordinated with other nouns. This can be seen in the following examples:

Examples (10) and (11) show that the heads of non-compositional compounds, such as ʕaruus ‘bride’ in (10) and ʃaʕar ‘hair’ in (11) cannot be coordinated with other nouns without resulting in an impossible reading. Similarly, examples (12–13) demonstrate that the heads of semi-compositional compounds, such as ħajar ‘stone’ in (12) and baab ‘door’ in (13) cannot be coordinated with other nouns. Thus, examples (10–13) demonstrate that the possibility/impossibility of head coordination in N + N compounds can be used to distinguish between compositional compounds, on the one hand, and semi-compositional and non-compositional compounds, on the other. In addition, coordination of the non-heads is impossible with both semi-compositional and non-compositional compounds:

Examples (14) and (15) demonstrate that if the non-head of non-compositional compounds, such as lbaħir ‘the sea’ in (14) and lbanaat ‘the girls’ in (15), is coordinated, the compound loses its non-compositional reading. The same applies to examples (16–17) where the coordination of the non-head results in an anomalous reading.

3. Compositionality and its types

Neef (Reference Neef, Lieber and Štekauer2009: 394) states that a linguistic expression is considered compositional if its meaning is derived from the meanings of its parts and the way it is structured. For example, the English compound school bus is compositional because its meaning is determined by its components; school and bus. Neef (Reference Neef, Lieber and Štekauer2009: 395) points out that the compositional meaning of a compound with the constituents AB is ‘B that has something to do with A’. That is, every compositional compound which consists of two elements can be interpreted in an associative way, especially N + N compounds. For instance, the German compound Fisch•frau, lit. fish•woman ‘is a woman that has something to do with fish’ (Neef Reference Neef, Lieber and Štekauer2009).

In English, three levels of semantic compositionality in compounds can be distinguished. The fact that there are levels within compositionality has been acknowledged by Fernando (Reference Fernando1996: 36), who examined the levels of compositionality in idioms. In particular, Fernando argues that in addition to a pure idiom, which is completely non-compositional, there is another type of idiom, which he refers to as semi-idiom. The latter refers to a sequence that has one or more literal elements and at least one element that has a non-literal sub-sense. The semantics of a semi-idiom is not as complex as that of a pure idiom, since its meaning is partially transparent. For instance, one can infer from the idiom “to promise someone the moon” that something is being promised.Footnote ³ Similarly, Dirven and Verspoor (Reference Dirven and Verspoor1998: 60) argue that compounds can be placed on a cline of transparency, which includes transparent compounds, partially transparent compounds, and non-transparent compounds. Examples of the three classes are apple tree, high street, and red tape, respectively. The three levels of compositionality are presented in Table 1.

Table 1:

Levels of compositionality in English by Dirven and Verspoor (Reference Dirven and Verspoor1998: 60)

In the first level, the compounds are completely compositional in the sense that the meaning of the whole compound is the total sum of its parts. For instance, the compound bookshop is a shop that has something to do with books. In the second level, the meaning of the compound is not completely the total sum of its parts, but the head makes a clear contribution to the meaning of the whole compound. For instance, the compound high street is a street, even though it is not necessarily high. The compounds in the third level are completely non-compositional in the sense that the meaning of the whole compound cannot be derived from the sum of its parts. For example, egghead refers to neither a head nor an egg. Its meaning, ‘intellectual’, is not related to both elements combined together.

Semantically speaking, by applying the ‘IS A’ condition, which was suggested by Allen (Reference Allen1978: 11), it seems that English compounds are usually semantically headed, but there are some that are headless. This principle is normally used to differentiate between endocentric and exocentric compounds. Allen's ‘IS A’ condition is given in (18).

This can be seen in the endocentric compounds in examples (19) and (20):

This principle can be used to show that faintheart and airhead are exocentric compounds, as in (21) and (22):

Bauer (Reference Bauer1998: 67) suggests that non-compositional compounds are listed in the dictionary, whereas syntactic constructs such as phrases are not, although he points out that this can be considered a lexicographical rather than a linguistic criterion. In particular, Bauer states that “many linguists seize one aspect of listedness – namely idiomaticity – and use that as a criterion for compound status”. Examples would be a compound like high street. Later, Kavka (Reference Kavka, Lieber and Štekauer2009: 33) argues that compositionality is the most important criterion that distinguishes compounds from free combinations, claiming that, like idiomatic expressions, compounds are non-compositional. Kavka (Reference Kavka, Lieber and Štekauer2009: 33) suggests that “their status will be understood more readily if they are viewed as parts of concrete, contextually defined utterances”.

On the other hand, Lieber (Reference Lieber, Štekauer and Lieber2005: 376) points out that compounding in many languages is highly productive and new compounds are very often compositional in meaning, especially when context is taken into account. In other words, it is easy to dismiss this non-compositional criterion for compoundhood, at least in languages like English; the more productive the process of compounding in a language, the less likely individual compounds are to be lexicalised or listed (Lieber and Štekauer Reference Lieber, Štekauer, Lieber and Štekauer2009: 7). The same applies to phrases: since old hand and green fingers are non-compositional, whereas beautiful house, long journey and tall man are compositional. Therefore, compositionality is not a reliable criterion for distinguishing compounds from phrases in English. In other languages, things may of course be different. For instance, Borer (Reference Borer, Lieber and Štekauer2009: 491-2) shows that compositionality in Hebrew is a reliable criterion for distinguishing between compounds on the one hand (e.g., beyt (ha)sefer, lit. house (the) book ‘(the) school’) and R-constructs (i.e., possessive constructs such as beyt (ha)mora, lit. house (the) teacher ((the) teacher's house)) and M-constructs (i.e., modification constructs such as kos (ha)mic, lit. glass (the) juice ‘(the) juice glass’), on the other. Borer suggests that the meaning of productive syntactic constructs (i.e., R-constructs and M-constructs) is completely predictable from their components, while the meaning of compounds in Hebrew is not predictable from the individual N components; hence, it is non-compositional. In JA, Altakhaineh (Reference Altakhaineh2019: 33) suggests that the non-compositionality criterion can only identify non-compositional compounds; it cannot differentiate between P-constructs (phrases), on the one hand, and compositional compounds, on the other. Therefore, it is partially applicable. Borer's M-construct, whose meaning is totally predicted from its parts, is a compositional compound in Arabic. Thus, in Arabic, R-constructs are phrases, while M-constructs are compositional compounds. What Borer refers to as compounds in Hebrew, which are entirely non-compositional, are also found in Arabic. This suggests that while compositionality can be used to pinpoint compounds in Hebrew, it cannot be used as a criterion in Arabic, since both compositional and non-compositional compounds can be found.

In relation to degrees of semantic compositionality, Bauer (Reference Bauer1983: 56) argues that compounds that exhibit different degrees of compositionality are attested in the literature.Footnote ⁴ For example, unlike understand, which has been completely lexicalised, compounds like playboy and bedstead still have some relation to the meaning of their parts; they are only partially compositional (Bauer Reference Bauer1983: 56–57). According to Marelli and Luzzatti (Reference Marelli and Luzzatti2012: 653) and Ji et al. (Reference Ji, Christine L. and Thomas L.2011), semantic transparency or degrees of semantic compositionality is all about how well the combination of the two elements of the compound, rather than the two individual words, determines the meaning of the whole compound. Along these lines, semantic compositionality can be said to lie on a synchronic and diachronic continuum, with full compositionality at one end and no compositionality at the other end, and with various degrees of compositionality in between (Fernando Reference Fernando1996, Dirven and Verspoor Reference Dirven and Verspoor1998).

Similarly, Kavka (Reference Kavka, Lieber and Štekauer2009) argues that compositionality can be regarded as a scalar phenomenon, in which multiword expressions are viewed as fully compositional, such as shoot a bird and red ink; semi-compositional, such as shoot a film and red carpet; or non-compositional, such as shoot the breeze and blue blood. With respect to the difference between fully compositional and semi-compositional, Kavka (Reference Kavka, Lieber and Štekauer2009) states that playground is fully compositional, since it literally means a ground to play on, while life boat does not mean ‘*a living boat’, rather ‘a boat used for saving lives’. Similar examples are bulldog, horse-fly, stone-fish, etc. He also indicates that the sequence of elements in the majority of non-compositional compounds is fixed, as in lazybones (not *boneslazy). With regard to their lexical flexibility, modifications are always external, that is, they modify the whole compound, rather than one or the other element, as in an intolerable lazybones, the word intolerable describes the whole compound.

According to Kavka (Reference Kavka, Lieber and Štekauer2009), the meaning of a multiword expression may become fixed through time, which reduces the degree of compositionality, resulting in it having a figurative meaning. This may suggest that the expression moves along the scale of compositionality. Historically, compositionality can be regarded as a cline on which multiword expressions are situated. Expressions which are referred to as idiomatic are on the non-compositional side of the cline.

Fernando (Reference Fernando1996), and Dirven and Verspoor (Reference Dirven and Verspoor1998)'s classifications, as well as Kavka's (Reference Kavka, Lieber and Štekauer2009) proposal, which are quite similar, can also be applied to compounds in JA. If we also take into account that compositionality can be regarded as a scalar phenomenon with fully compositional at one end, and fully non-compositional at the other end, with some intermediate degrees in between, and if such degrees of compositionality can be based on the semantic contribution of the head and the non-head to the meaning of the whole compound, then a new, slightly modified, classification can be proposed for both English and JA. That is, if only the head of the compound contributes to the meaning of the whole compound, then it is semi-compositional, whereas if only the non-head of the compound contributes to the meaning of the whole compound, then it is semi non-compositional. This is illustrated in Figure 1 below.

Figure 1:

Levels of compositionality vs. non-compositionality in compounds

Examples of each type of compound described in Figure 1 from English are provided below:

Examples (23) and (24) show two levels of non-compositionality. The meaning of the compound swansong ‘farewell performance’ in (23) is derived from neither swan nor song. That is, swansong IS NOT A song, which means that the compound swansong is semantically headless and does not denote a hyponym of its head song. The compound scarecrow ‘an object to scare birds’ in (24) is also semantically headless since scarecrow IS NOT A crow, which means the compound does not denote a hyponym of its head. However, the meaning of the non-head, towit scare, seems to contribute to the whole meaning of the whole compound.

In addition, two levels of compositionality can be depicted in English compounds in (25) and (26). In (25), the compound bookshop is endocentric, that is, semantically headed, since bookshop IS A shop, suggesting that the compound denotes a hyponym of its head. The meaning of the compound bookshop ‘a shop where books are sold’ is a sum of its two parts book and shop; thus it is completely compositional. In contrast, high street in (26) is also endocentric since high street IS A street, which means that the compound is a hyponym of its head. However, high street is not necessary high, which means that only the head contributes to the meaning of the compound, making it semi-compositional.

4. Methodology

The analysis in the current study is focused mainly on JA compounds in comparison with English compounds. For the purpose of this study, the examples that are analysed were collected from various resources, including: (1) some research articles on JA compounding such as Altakhaineh (Reference Altakhaineh2016, Reference Altakhaineh2017, Reference Altakhaineh2019) and Zibin and Altakhaineh (Reference Zibin and Altakhaineh2018); (2) some social media websites, such as Facebook and Twitter; and (3) my intuition as a native speaker of Jordanian Arabic (see Altakhaineh Reference Altakhaineh2016). I collected around 60 examples of JA compounds. These compounds were classified into four degrees of compositionality, namely, completely compositional, semi-compositional, semi non-compositional and completely non-compositional, based on the judgement of 40 native speakers of JA. The 40 informants recruited in the current study are residents of Amman (the capital of Jordan) and are part of the same social network. They are students majoring in linguistics at the University of Jordan. Their mean age is 21. They are all native speakers of JA with working knowledge of Modern Standard Arabic (henceforth, MSA). I informed the participants that taking part in the study was voluntary and they could withdraw at any time during the experiment. Each participant was given a compositionality judgement task in which he/she was asked to classify 60 JA compounds as: 1) completely compositional; 2) semi-compositional; 3) semi non-compositional; and 4) completely non-compositional. To ensure that the participants understood the task, the compounds were all contextualised, and I explained to the informants the notion of compositionality and the four above degrees with illustrative examples. The informants were also asked to explain their choice briefly. Only the compounds that were agreed upon by 80% of the informants (50 out of 60 examples) were included in the study.

The current study adopted a corpus-driven approach to data analysis. According to Tognini-Bonelli (Reference Tognini-Bonelli2001: 84), more insightful results can be yielded when the researcher adopts a corpus-driven approach in comparison with a corpus-based approach. This is because in the corpus-driven approach, the researcher analyses all the data in the corpus, which provides more accurate results. In corpus-driven studies, the researcher usually does not have preconceived hypotheses which he/she intends to test based on the corpus. Conversely, in this approach the researcher starts with the corpus itself (Tognini-Bonelli Reference Tognini-Bonelli2001: 84). Nevertheless, some researchers (e.g., Zibin Reference Zibin2018) have argued that in reality the majority of corpus-driven studies do start with some preconceptions about what they are going to find in the corpus, which means that these studies at least have an initial prediction with regards to the findings. Still, in these kinds of studies, initial predictions are examined by analysing the entire corpus. (Deignan Reference Deignan and Gibbs2008: 282). That is, since the corpus size is small, the researcher can manually analyse the entire corpus and even if he/she has preconceived notions, these can be examined thoroughly. In large-size corpora, the researcher is unable to manually examine the entire corpus. Rather, an automated software-based search is normally performed. Studies which adopted a corpus-driven approach indicated that more accurate results are elicited when the researcher analyses all the data in the corpus (e.g., Moon Reference Moon1998, Zibin Reference Zibin2018). Hence, in this study, I analysed all the instances of compounds collected from different sources and made judgements regarding their compositionality, as well as consulting with the native speaker informants who took part in the study.

5. Compositionality in JA

This section classifies the JA compounds into four levels, based on their compositionality. For JA, examples of each type of compound described in Figure 1 are provided below (see Altakhaineh Reference Altakhaineh2019 for more examples of N + N compounds in JA):

5.1 Completely non-compositional compounds (neither element)

Examples (27–29) were judged by 90% of the informants to be completely non-compositional, since the meaning of compound is not derived from either of its elements.

In example (27), ʃaʕar lbanaat ‘candyfloss’ IS NOT ʃaʕar ‘hair’. Semantically, it is neither related to ʃaʕar ‘hair’ nor to lbanaat ‘the girls’. The same applies to examples (28–29) where the meaning of the compound liʕb ʕyaal ‘something which is insignificant’ in (28) is neither liʕb ‘playing’ nor ʕyaal ‘children’ and the meaning of ʔabuu lhool ‘the sphinx’ in (29) is not related to either of its constituents. This suggests that when there is no contribution of the meaning of either the head or the non-head to the meaning of the entire compound, this compound can be regarded as completely non-compositional. This also suggests that this type of compound is semantically headless since the compounds ʃaʕar lbanaat ‘candyfloss’, liʕb ʕyaal ‘something which is insignificant’, and ʔabuu lhool ‘the sphinx’ do not denote hyponyms of either element, that is, the semantic head is outside the compound.

5.2 Semi-non-compositional compounds (only the non-head)

The following are examples of semi-non-compositional compounds in JA.

lit. pioneer of the space

Comparing examples (27–29) with (30–32) shows that two levels of non-compositionality can be distinguished in JA compounds. Specifically, JA compounds can be completely non-compositional when the meaning of the whole compound (e.g., candyfloss) is not related to the meaning of both elements combined. Conversely, ʕaruus lbaħir ‘mermaid’, in example (30), IS NOT A ʕaruus ‘bride’. However, ʕaruus lbaħir ‘mermaid’ is a mythical creature that lives in lbaħir ‘the sea’. Thus, at least one element of the compound, in this case the non-head, contributes to the meaning of the whole compound. Hence, it is semi-non-compositional. The same can be noted in example (31) in which the non-head lfadˤaaʔ ‘the space’ contributes to the meaning of the compound raaʔid lfadˤaaʔ ‘astronaut’ since he/she works in space despite not being a pioneer. Similarity, in (32) faras nnahir ‘the hippopotamus’, the non-head nnahir ‘the river’ contributes to the meaning of the compound, since hippos usually inhabit rivers. These examples were judged by 80% of the informants as semi-non compositional. It can be observed that both completely non-compositional and semi-non-compositional compounds are exocentric. In particular, both types are semantically headless, since they do not denote hyponyms of either of their elements. The interesting point here is that even when compounds are semantically headless (i.e., exocentric), in semi-non-compositional compounds, the non-head still contributes to the meaning of the compound even if it does not denote a hyponym of either of its elements. This may suggest that the lack of a semantic head does not necessarily imply that the meaning of the compound is not derivable from the meaning of its non-head.

It can also be noted that some compounds, such as ʕaruus lbaħir ‘the mermaid’, are based upon metaphors. The issue of whether metaphoric/metonymic compounds are endocentric or exocentric has been subject to a great deal of debate in the relevant literature. For some researchers (e.g., Guevara and Scalise Reference Guevara, Scalise, Scalise, Magni and Bisetto2009), compounds based on metaphor and/or metonymy are regarded as exocentric. Bauer et al. (Reference Bauer, Lieber and Plag2013: 478) suggest that bahuvrihisFootnote ⁵ are exocentric, but leave open the possibility of analysing them as endocentric because of their figurative reading. They indicate that bahuvrihis of this type are productive in English, for example air head, bone head, acid head, cheesehead, butterhead, bottle head and redhead, which are all metonymic in nature. Bauer et al. (Reference Bauer, Lieber and Plag2013: 478) conclude that the main difference between bahuvrihis and endocentric attributive compounds is that the referents of bahuvrihi compounds are characterized figuratively, normally being metaphorical or metonymic. Therefore, analyzing them as regular endocentric compounds with a metonymic or metaphorical interpretation of the head noun is possible (Bauer et al. Reference Bauer, Lieber and Plag2013: 478-9). Andreou and Ralli (Reference Andreou, Ralli, Bauer, Körtvélyessy and Štekauer2015) argued that nominal bahuvrihis should be analyzed as endocentric compounds based on metonymy. Other researchers (e.g., Benczes, Reference Benczes2006) proposed the umbrella term “creative compound” for any metaphorical and/or metonymical compound, and discarded the traditional “endocentric” and “exocentric” labels. It can be observed that whether compounds based upon metaphors are endocentric or exocentric is non-conclusive; yet, based on the analysis of JA compounds in this study, the compounds based upon metaphor usually lack a semantic head, and as such can be regarded as exocentric.

5.3 Completely compositional compounds (both elements)

Examples (33–35) are of completely compositional compounds.

Two levels of compositionality can be detected in compounds in JA. For example, ʕasˤiir ttuffaaħ ‘the apple juice’, in example (33), IS A ʕasˤiir ‘juice’. Semantically, ʕasˤiir ttuffaaħ is juice made from apples. Hence, this compound is completely compositional. Similarly, findʒaan ʃʃaay in example (34) IS A findʒaan ‘cup’ and both elements contribute to the meaning of the whole compound. The same applies to example (35). These examples are also endocentric, which means that they are semantically headed and the compound denotes a hyponym of its head. That is, ʕasˤiir ttuffaaħ ‘the apple juice’ is a hyponym of ‘juice’, findʒaan ʃʃaay ‘the tea cup’ is a hyponym of ‘cup’ and muʕallim lfiizyaaʔ ‘the physics teacher’ is a hyponym of ‘teacher’.

5.4 Semi-compositional compounds (only the head)

The following are examples of semi-compositional compounds.

The compound ħajar zzaawiyah ‘the cornerstone’, in example (36), IS A ħajar ‘stone’, which is used as a foundation in any building. In other words, it is not part of the second element zzaawiyah ‘the corner’; the relationship between the elements is determinative rather than associative (see Adams Reference Adams1973, Olsen Reference Olsen2000: 279). Thus, this compound is semi-compositional, because only the head contributes to the meaning of the whole compound (see Figure 1). The same applies to example (37) baab lʕaamood which IS A baab ‘door’ but is not part of the second element lʕaamood ‘the pillar’ and thus was judged by 90% of the informants as semi-compositional. Example (38) ʔaklit lmuluuk IS A ʔaklih ‘meal’ but is not related to kings, at least not contemporarily. It can be observed that both completely-compositional and semi-compositional compounds are endocentric, as they are semantically headed and they denote hyponyms of their heads. This can be seen as the exact opposite of completely non-compositional and semi-non-compositional compounds, since the latter are both exocentric; and while in non-compositional compounds neither elements contribute to the meaning of the compound, in semi-non-compositional ones, only the non-head does. When the compound is semantically headless, that is, exocentric, neither element, or only the non-head, can contribute to the meaning of the compound, whereas when the compound is semantically headed, that is, endocentric, either both elements, or only the head, contributes to the meaning of the compound. It can be observed that there is an interplay between the notions of endocentricity and exocentricity, on the one hand, and semantic compositionality, on the other.

Note that many examples of non-compositional compounds in JA may have been borrowed from other languages, such as English, and integrated into JA at some point in history. JA does not have one-to-one equivalents for words like ‘mermaid’, that is, ħuuriyyat lbaħr ‘beautiful woman in the sea’.Footnote ⁷ Therefore, JA uses two words to describe one lexical item. This can explain why N + N constructs are very productive. Note, also, that the outcome of this integration is, in many cases, non-compositional.

Out of the total 50 examples judged by the 40 informants based on their level of compositionality, four compounds were found to be completely non-compositional, seven semi-non compositional, 33 completely compositional, and six semi-compositional. Based on the data collected in the current study, JA compositional compounds were more frequent than non-compositional ones, and by extension endocentric compounds were more frequent than exocentric ones. This tentative result may have implications for the productivity of these types of compounds. That is, compositional compounds in JA, both completely and semi, could be more productive than non-compositional ones, both completely and semi (see Lieber Reference Lieber, Štekauer and Lieber2005: 376).