Proposal of a catalogue of criteria for the identification and classification of metaphorical collocations

2.1 Our concept of collocation

In the course of research on collocations, countless definitions have been proposed. Overall, however, these can be roughly grouped into two major approaches: namely, a quantitative-statistical and a qualitative-semantic one (see, e.g., Mellado Blanco, Mollica and Schafroth 2021: 8; Konecny 2021: 145, 2023: 5–7). Our understanding of collocations aligns primarily with the qualitative-semantic approach, drawing heavily on the foundational works of Franz Josef Hausmann (e.g., 1985, 1993, 2004, 2007), although analytical tools from the quantitative-statistical approach are also employed to ensure empirical grounding and cross-linguistic comparability in the analysis of the four languages under study (see Section 3).

Methodologically, Hausmann’s division of collocational components into “base” and “collocator” proves to be particularly useful for the MetaKol project. These two components stand in an asymmetrical relationship to each other (see, e.g., Hausmann 1985: 119, 2004: 312). In the German combination den Tisch decken ‘lay the table’ (lit. “cover the table”), the noun Tisch is the base that expresses the essential content and from which the encoding of an utterance normally starts. The verb decken, on the other hand, is the collocator (see Hausmann 2007: 218). If we compare the equivalents in the three other languages under investigation (Engl. lay the table, Ital. apparecchiare la tavola [lit. “equip the table (with tablecloth, crockery etc.)”], Croat. postaviti stol [lit. “set/lay the table”]), we see they are completely identical at the base level. For this reason, within the MetaKol project, we use Sketch Engine to generate collocation profiles of selected nouns functioning as bases and then analyse their collocators semantically. According to Hausmann, the meaning of the collocator is determined by its combination with the base to which it is contextually linked (see, e.g., 1993: 475). Here, “context” refers to the lexical environment of a word at the level of parole, i.e., not to the situational context. The verb decken, for instance, is polysemous in itself, but when combined with Tisch its ambiguity is resolved. Hence, following Heid (1992: 532), a collocation can be viewed as a conceptual unit. According to Hausmann (1993: 475), the base transfers its basic meaning to that of the entire collocation, while the meaning of the collocator may vary depending on the context. The collocator is therefore often subject to semantic modification compared to its primary or original meaning (see Konecny 2010a: 287–289). Accordingly, collocators can be classified by the degree of their semantic alteration (see, e.g., Stein 1998: 68–69): in some combinations, the basic meaning of the collocator hardly changes, such as in große Angst ‘great fear’, where there is only a (broadly metaphorical) transfer from space as a concrete domain to the abstract domain of emotion. In other examples, the collocator maintains only a part of its meaning, often a more general seme, as in großes Unwetter ‘great storm’, where the feature of intensity is preserved. In other cases, only a vague or even no connection between the original and the actual meaning of the collocator is recognisable, as in the German collocation eingefleischter Junggeselle ‘confirmed bachelor’ (lit. “a bachelor down to the flesh”). Thus, in these examples, the metaphor is usually faded or worn out. Such categorisation suggests different degrees of semantic transparency in collocations, with the shift in meaning affecting only the collocator in most cases.

It is important to point out that in the MetaKol project – in accordance with Hausmann (2004: 313–317) – the criteria of polylexicality and idiomaticity are interpreted broadly, which is why certain marginal phenomena are still regarded as collocations:

1. As far as the first criterion is concerned, compounds or one-word phrasemes are sometimes considered collocations (see Hausmann 2004: 317; Konecny and Autelli 2015: 583). The requirement is that a structure consisting of “base + collocator” must be identifiable, i.e., that at least one component (interpretable as the base) is used in a primary meaning and can be translated (more or less) directly or without major problems into another language. Two German compounds cited by Hausmann (2004: 317), which in his view can be regarded as collocations, are Schiebedach ‘sliding roof, sun roof’ and Briefkastenfirma ‘letterbox company’, because the nominal elements here can be interpreted as bases and their translation does not pose any difficulties. The determining constituents (Schiebe- and Briefkasten-), on the other hand, function similarly as collocators because they are unpredictable and the equivalents in other languages may be based on different conceptualisations. In this context, Hausmann mentions the divergent French equivalents toit ouvrant (lit. “openable roof”) and société écran (lit. “screen/camouflage company”). It should be noted that in these examples in French – as is generally the case in Romance languages – multi-word expressions are typically used. Hence, from the morphological point of view, these are not compounds, but compound equivalents that (as in German) denote a certain subtype of roof or company. Polylexicality, which is actually a defining criterion for collocations, can therefore be interpreted broadly in such cases, thus also encompassing compounds and their components. As the examples given show, such a view makes sense from a contrastive-cognitive as well as from a learning/teaching and lexicographical perspective, because in this way it is possible to uncover interlingual divergences that may be caused by metaphorisation, among other things.^[8]

2. As far as idiomaticity is concerned, it should be emphasised that semi-idiomatic phrasemes (see Burger 2015: 27) are considered a subcategory of collocations in the MetaKol project. In this case, too, the decisive criterion is that there must be a constituent that can be interpreted as the base and that represents the non-idiomatic part (see Hausmann 2004: 313–315; Konecny 2010: 102, 2021: 148). The idiomatic part, on the other hand, is constituted by the collocator, which in the respective idiomatic meaning usually occurs in only one or in just a few combinations. One of the examples given by Hausmann (2004: 314), which is also cited in several other works (e.g., Hartenstein 1996: 86; Konecny 2011b: 122; Häcki Buhofer 2023: 209), is Germ. blinder Passagier ‘stowaway’, i.e., ‘passenger travelling without ticket’ (lit. “blind passenger”). Additionally, Hausmann also cites a number of more complex cases, such as the pseudo-triplet structure Geld zum Fenster hinauswerfen ‘to throw money down the drain’ (lit. “to throw money out of the window”), where money is the base and zum Fenster hinauswerfen is the complex idiomatic collocator (Hausmann 2004: 316). Furthermore, he also considers comparative phrasemes as semi-idiomatic collocations, such as Germ. passen wie die Faust aufs Auge ‘to be a perfect match’ (lit. “to match like the fist on the eye”), where the verb passen represents the base and the comparative constituent the idiomatic collocator (Hausmann 2004: 314). In line with this extended concept of collocation, which also includes semi-idioms, pagare sull’unghia ‘pay (cash) on the nail’ is cited below in Section 4.1 as an example of a collocation, because pagare ‘pay’ is clearly used literally and thus represents the base, whereas sull’unghia ‘on the nail’ is an idiomatic collocator, which is based on a metonymy in this case. At the same time, such partial idiomatic collocations are characterised by the maximum possible degree of shift in meaning on the part of the collocator.

Apart from the essential criterion of the collocator’s shift in meaning, several other proposals of criteria for identifying collocations can be found in the relevant literature, e.g., in Heid (1992: 534), Manning and Schütze (1999: 165), Caro Cedillo (2004: 93), and Roth (2014: 120–122). With regard to the objective of the present article, however, a fundamental problem is that the works in question are often based on a different, usually broader conception of collocation, which is statistically oriented and sometimes also includes fully idiomatic expressions. Moreover, they do not focus on metaphorical collocations. Within the framework of the MetaKol project, it is clear that the criteria mentioned in the literature for defining metaphorical collocations must be regarded as inadequate. For this very reason, the aim of the project is to develop suitable criteria for identifying metaphorical collocations on the basis of extensive corpus analyses. Since, in this respect, language comparison provides important methodological insights into different collocational patterns and types, the following subsection will take a closer look at the role of idiosyncrasy and the relevance of interlingual considerations.

2.2 On the role of idiosyncrasy and the relevance of comparative linguistic analysis

A typical characteristic of collocations is generally considered to be idiosyncrasy, which means that the combination of two lexemes in a collocation is arbitrary. Accordingly, it is almost impossible for non-native speakers to predict which word combinations are “allowed” or common in a language and which, on the other hand, are unacceptable (see Cop 1991: 2776). Konecny (2010a: 176–184) distinguishes between collocations with different degrees of idiosyncrasy, emphasising that examining their formation process can reveal possible original motivations for their emergence. As an example, she cites the German verb lichten, which represents a component of the collocation den Anker lichten ‘to loose the anchor, to weigh anchor’ that is not used outside of this specific combination. From a diachronic perspective, however, it can be observed that lichten etymologically relates to leicht ‘light’ and originally meant ‘to lighten’, which can in turn be associated with raising the anchor. This suggests that the combination was previously motivated and became consolidated over time while the verb lost its original meaning. As a result, den Anker lichten is now considered a fixed phrase, specific to the German language and, therefore, idiosyncratic in this sense.

In addition to diachronic considerations, interlingual comparisons can also make a significant contribution to identifying collocations as such, as their collocation status often only emerges from language comparison (see, e.g., Mellado Blanco, Mollica and Schafroth 2021: 7). In the following, some German “triple collocations” (see Hausmann 2004: 216) with the components schwer ‘heavy’ and Zeit ‘time’ are contrasted with their possible equivalents in English, Italian, and Croatian, which have been identified on the basis of corpus analyses:

Comparative linguistic analysis shows that the equivalents can all be regarded as collocations in a broad sense, which can be justified by the fact that they cannot be translated one-to-one, as the languages differ in their structure and lexical usage (although some combinations are similar and there is more than one way of expressing a given concept). What all the examples have in common, however, is that they are diachronically based on the process of metaphorisation^[9] and thus exhibit a certain degree of metaphoricity (even if synchronically this is difficult to determine), which at the same time indicates that they are not unmotivated. Ultimately, though, the selection of the respective metaphor and its concrete linguistic realisation are the result of language-specific and often also cultural particularities, which leads to differences that only become apparent when contrasting languages (see Stojić and Košuta 2020: 152–153). This means that in collocations only the act of verbalising something in a certain way is actually idiosyncratic. Therefore, in this respect language is, following Dobrovol’skij (2001: 90), ‘arbitrary and non-arbitrary at the same time: arbitrary as regards impredictability, and non-arbitrary as regards the retrospective interpretability of relevant phenomena’.^[10]

2.3 From conceptualisation to metaphorical collocation

As can be seen from the examples in Table 1, the metaphorical collocations represent partial equivalents in the four languages analysed. This is due to the fact that the same extralinguistic reality is conceptualised differently and, therefore, also lexicalised in a different way. In cognitive linguistics, conceptualisation is described as a process in which abstract concepts are defined by concrete concepts using metaphors or metonymies. According to Lakoff and Johnson (1980), metaphors and metonymies are not only linguistic phenomena, but also cognitive mechanisms of thought. Despite numerous differing definitions for both metaphors and metonymies, the common denominator is that metaphors are cognitive projection processes between two domains, while metonymies represent cognitive mappings within a single domain (see Sulikowska 2020: 142).

Metaphors transfer the meaning of a concept from a source domain to a target domain on the basis of similarities. Metonymies, on the other hand, are based on the experiential proximity of concepts; they are usually more anchored in our physical experience, less abstract and more fundamental. In some cases, they lead in turn to the formation of metaphors. Grady (1997: 20) calls these basic correlational phenomena primary scenes and notes they result without conscious awareness from the neuronal co-activations of the source and target domains. Taylor (1995: 138) explains the primary metaphor as a metonymy-based experience, emphasising that the level of metaphor can only be reached through schematisation, leading to an interaction between metaphor and metonymy, which Goossens (1990: 369) calls metaphtonymy. Radden (2002: 411), who defines the interaction between metaphor and metonymy as a process, analyses adjective-noun collocations based on the concepts Up(Height) and More. His model distinguishes four phases of conceptualisation: (1) in its literal sense, Updenotes vertical direction (e.g., high tower); (2) in a partial metonymic meaning, it denotes both verticality and quantity (e.g., high water level); (3) the process of so-called conflation leads to complete metonymy where Upmeans More (e.g., high temperature); (4) finally, deconflation marks the emergence of the conceptual metaphor Up is more (e.g., high quality). According to Radden (2002: 411), expressions like high price fall along this continuum, challenging classification due to the blending of metonymy and metaphor. The assignment to a phase or type of conceptualisation depends on whether one looks at ‘height’ (of a price) and ‘quantity’ (of money) as belonging to the same experiential domain (up for more) or two distinct domains (up is more). Hence, the conceptual relationship between Upand More is not clear or obvious; nevertheless, prices still have a common experiential basis (e.g., the fact they are expressed in numbers or charts). Radden (2002: 412) introduces the concept of “metonymy-based metaphor”, which denotes the central phenomenon of this type. In a metonymy-based metaphor, the target and source domains can be reduced to one common domain.

In contrast to conceptualisation, a specific construction expresses how a speaker decides to present a particular conceptual idea linguistically. In this way, the construction influences the conceptualisation as it shapes the latter in the mind of the speakers (Evans and Green 2006: 536). One specific point at which linguistic conceptualisation and construction overlap is the process of lexicalisation. This process is often based on metaphors and metonymies and is characterised by repeated use of speech that causes the original concept to fade, leading to the emergence of the so-called “entrenched” metaphors (see Lakoff and Johnson 1980; Kövecses 2002). These metaphors, which are firmly embedded in the language system, become increasingly specialised; although their original meaning fades, it remains recognisable to the average speaker, as can be seen, for example, in the metaphorical collocations Germ. absehbare Zeit – Croat. dogledno vrijeme – Engl. foreseeable time – Ital. tempo prevedibile, which are all based on the conceptual metaphor Time is a moving object.

The process of demetaphorisation culminates in dead metaphors, which are semantically faded and conventionalised to such an extent they are no longer recognised as metaphors unless their etymology is known. Müller (2008: 15) emphasises they cannot be reduced to conceptual metaphors, even if both reflect each other. Lakoff and Turner (1989: 129) argue that conceptual metaphors are “alive”, while lexical, i.e., grammaticalised metaphors, are considered “dead”. However, they criticise the term dead metaphor itself, pointing out that such metaphors are the most active type, as they control our thinking unconsciously and require less cognitive effort than new, creative metaphors.^[11] These considerations led to a classification of metaphors into three categories based on different degrees of transparency of meaning (see Lakoff and Turner 1989: 129): (a) dead metaphors – conventionalised and no longer transparent, (b) rooted metaphors – conventionalised and transparent, (c) new or creative metaphors – non-conventionalised and transparent.

The MetaKol project starts from the assumption that these mechanisms and their effects – both in synchronic and diachronic terms – are essential factors in the creation of collocations. Furthermore, it is assumed that metaphorical collocations are always lexicalised (i.e., non-spontaneous) metaphors or metonymies that exhibit varying degrees of synchronic perceptibility. The analyses show that it is not always easy to draw a clear line between different degrees of metaphoricity and, consequently, between different types of metaphorical collocations. Nevertheless, such analyses point to the important role that the mechanisms of conceptual metaphor and metonymy play in the formation of collocations. The aim of the MetaKol project and of the present article is precisely to highlight and analyse this crucial role in detail. Although different levels regarding the lexicalisation of metaphors can be identified when comparing languages, all examples of metaphorical collocations are ultimately conventional, well-established combinations, which shows that the principles of the emergence of collocations are inherent to the languages studied (which suggests the universality of mechanisms). The process of lexicalisation, on the other hand, is idiosyncratic and depends on language use (thus giving the impression of arbitrariness), confirming, in turn, the initial hypotheses of the project.

4.1 Criteria for identifying metaphorical collocations

The inductive examination of collocational profiles in the analysed corpora and the consideration of information from the secondary literature revealed that different aspects should be taken into account to determine whether a combination of lexemes is to be regarded as a metaphorical collocation. These aspects are illustrated below:

1. Congruence with conceptual metaphors and cognitive plausibility: When analysing concrete examples, it is necessary to check whether a collocation can be integrated into a superordinate conceptual metaphor (see Lakoff and Johnson 1980), even if the collocation only helps to relate an abstract concept to a concrete, more descriptive concept. This can immediately be regarded as a manifestation of the general metaphor Abstract is concrete, which gives rise to a metaphorical collocation. To identify collocations, it was necessary to determine the types of conceptual metaphors according to standardised criteria. For this purpose, we resorted to the Master Metaphor List (Lakoff et al. 1991) in its modified form. From the methodological point of view, inspecting bases that belong to the same semantic field intralingually proved to be useful to our analysis. The detailed lexico-semantic analysis of metaphorical collocations from the semantic field ‘time’ in the Croatian corpus (Stojić 2024) revealed analogous processes in the formation of collocations, which manifest themselves in the same collocators being linked to different bases of the same semantic field. The second step included interlingual comparisons to determine the differences or similarities between the four languages with regard to conceptual schemes, and to find out whether, in the latter case, the verbalisations are convergent, similar or divergent on the lexical and morphosyntactic level. A collocation that can be integrated into an established cognitive scheme is also characterised by a higher degree of cognitive plausibility and is more likely to create a link between the collocator and the base in the minds of speakers.

2. Inclusion of metonymic aspects: As the corpus analysis has shown, metonymies and superordinate metonymic schemes or conceptual metonymies (see Kövecses 2002, 2010) can also play a role in the creation of collocations (similar to idioms). They often even occur in combination with the process of metaphorisation, which justifies our decision to include them in the analysis. The metonymic origin of collocators has already been addressed in several studies, e.g., Turk (2000: 480) and Konecny (2021: 147). In the latter article, the Italian semi-idiomatic, colloquial collocation pagare sull’unghia (lit. “to pay on the nail/claw”) is cited as an example, where unghia ‘(finger)nail (also: claw)’ is used partonymically for the whole hand. In English and German, this sequence corresponds to the (partially) equivalent collocations pay (cash) on the nail and bar auf die Kralle zahlen. Concerning the German equivalent, in this case the additional explicit mention of the adverb bar ‘cash’ is obligatory and Kralle (unlike Italian unghia and English nail) cannot refer to a person’s fingernail in regular usage, but only to an animal’s claw. In the German collocation, therefore, in addition to the metonymic component, a metaphorical transfer from the source domain Animals to the target domain Humans can be observed, which results in mostly negative (and sometimes ironic) connotations of the expression. This example also illustrates the blurred boundaries between metaphor and metonymy: in this specific case, the claw stands metonymically for hand, but at the same time the projection processes encompass two domains, Humans and Animals. The fact that metonymies can generally occur in collocations is also mentioned by Siller-Runggaldier (2008: 595), although in her examples battere un concorrente/un avversario, una squadra, la Ferrari, un record (‘beat a competitor/opponent, a team, Ferrari, break/beat a record’) the metonymic relation exists between the potential collocational bases and does not concern the collocator battere (lit. “to beat (sb.) up”). When analysing the corpora of the four languages studied, it is thus necessary to check whether metonymies affect the collocational base or the collocator and whether/how they are linked to metaphorical transfers.

3. Semantic modification: It should also be investigated whether and how the meaning of the collocator has changed within the word combination compared to its primary meaning. If its use within the combination has led to an expansion or shift in meaning, this may generate a metaphorical collocation. In addition to the meanings of collocators, this also affects the meanings of nominal bases in many cases, as these are already used in (one of) their lexicalised secondary meaning(s) in some contexts. Such an approach thus allows for the lexico-semantic analysis of the behaviour of collocators when they occur with bases that have already undergone the process of metaphorisation or metonymisation (Stojić and Matešić 2024: 55), an innovation that proved to be a useful strategy in the MetaKol project. The identification of secondary, metaphorically (or metonymically) conditioned meanings was based on the MIP method (Pragglejaz Group 2007), a metaphor identification procedure that focuses on the comparison between the primary and current contextual meaning of lexical units. According to MIP, a lexical unit is considered metaphorical if the contextual meaning differs significantly from the primary meaning given in dictionaries. To check this, certain contextual meanings of the lemmas determined via WordSketch were compared with the primary meanings listed in the relevant dictionaries for the respective project languages. For Croatian, we resorted to the online portal Hrvatski jezični portal; for German, to the DWDS (and in part also Duden); for English, to the Merriam-Webster Dictionary; and for Italian, to the Vocabolario Treccani. Investigating the process of the creation of collocations is particularly important for the purpose of detecting the possible loss of their former primary meaning, thus revealing an original metaphorical or metonymic motivation (see 4.2).

4. Deviation from linguistic conventions: It should also be ascertained whether elements of a word combination deviate from usual linguistic conventions. Metaphorical collocations can often occur when an otherwise uncommon use of certain words or phrases is present and when selectional restrictions are violated in this way. A case in point includes the uncommon use of the prepositional phrase auf die Kralle in the above-mentioned collocation bar auf die Kralle zahlen (‘pay (cash) on the nail’). Further illustrative examples are Germ. Zeit totschlagen (‘to kill time’, lit. “to beat time to death”) and its Italian equivalent ammazzare il tempo (“to kill the time”), whose verbal collocators outside the collocation are generally restricted to objects of the classes [humans] and [animals]. Due to the contextual combination with Zeit/tempo, the verbs are semantically modified and take on the special reading ‘not use sensibly’. The degree of change in the meaning of the collocator according to Stein (1998: 68–69) is, therefore, to be regarded as particularly high here, especially since the meaning in question only comes into play in connection with a few semantically related bases. The greater the deviation from linguistic conventions and the more restricted the collocational range of the collocator is in its meaning within the collocation, the more likely it is for the metaphor to be conspicuous (even for L1 speakers) and for the collocation to approach the category of idioms.

5. Consistency in use: It is also necessary to check whether a collocation is used consistently in different contexts. Metaphorical collocations or their collocators tend to recur in certain metaphorical patterns or contexts, as can be seen in the German example die Zeit verfliegt and its English equivalent time flies by, whose verbs are not limited to Zeit/time as prototypical bases, but also occur with other nouns (in the singular or plural) denoting the time dimension, e.g., die Jahre/Monate/Wochen/Tage/Stunden/Abende verfliegen (‘the years/months/weeks/days/hours/evenings fly by’). The example above, Zeit totschlagen, displays a similar behaviour, as witnessed by combinations such as den (halben/ganzen) Tag/Nachmittag/Abend totschlagen (“to kill, i.e., not use sensibly the (half/entire) day/afternoon/evening”). This criterion is, therefore, linked to the fact that the possible representatives of bases are often metonymically related to each other. As Siller-Runggaldier (2008: 593) demonstrates on the basis of extensive sample material from Italian, the bases of the same collocator can often stand in different lexico-semantic relations (synonymy, antonymy, hypo-/hyperonymy) to one another or belong to the same semantic field, e.g., feelings or emotions in the case of the Italian collocations trattenere le lacrime/il pianto/le risate/la rabbia and their German equivalents die Tränen/das Weinen/das Lachen/seinen Zorn zurückhalten/unterdrücken (“hold back/suppress the tears/crying/laughter/one’s rage”).

6. Frequency of use: According to the narrow, semantically oriented concept of collocation (see 2.1), the (absolute) frequency in itself is not actually a compulsory criterion for classifying a combination of lexemes as a collocation. However, the categorical exclusion of frequency measures can now be at least partially relativised by today’s possibilities of extensive corpus analysis, which can be used to determine not only the pure frequency but also the typicality values of the co-occurrence of words (indicated as “score” in Sketch Engine). Taking into account how often a word combination occurs in authentic texts and how high its degree of typicality is can actually impact its perception as a related unit that should not be underestimated. In some cases, this can therefore help in deciding whether a sequence of lexemes should be categorised as a free word combination or a collocation. As a result, frequency may suggest that a certain expression represents a metaphorical or metonymic collocation (see the example of wise words discussed below in 4.2.2).

7. Contextual adaptation and possible polysemy of collocations: Furthermore, how the collocation is used in context should be taken into account because metaphorical collocations can develop a special metaphorical power in specific textual environments. In line with other studies (see Konecny 2018: 154–158), the analyses in the MetaKol project have found that, in addition to individual collocational components, collocations can also be polysemous as a whole. This has been observed, for instance, with the Italian lemma tempo, which can be read not only as ‘time’ but also musically as ‘beat’. This gives rise to ambiguities in some collocations. In order to determine the specific meaning activated, it is necessary to consider the exact context, as can be seen from the example battere il tempo. In relation to the clock or the clock pendulum in the subject position, this collocation means ‘to beat/indicate the time’, whereas in a sporting context it has the meaning ‘to beat someone’s time’, and in a musical context ‘to beat time’. Other examples, such as the Croatian radno vrijeme, also show the important role of semantic readings and pragmatic usage features. The collocation in question can refer to the time during which work is done at the workplace, as well as to the opening hours of shops, offices or other facilities during which they are accessible to customers or employees. In the context of medical services, however, radno vrijeme also means ‘consultation time’. The respective meaning can therefore only be disambiguated depending on the linguistic and situational context.

It is assumed that only by combining the above criteria can we have a comprehensive assessment of whether a word combination qualifies as a metaphorically and/or metonymically motivated collocation. These criteria, though interconnected and sometimes overlapping, provide a robust framework for analysing the formation and usage of collocations across languages.

4.2 Identified collocation types

On the basis of the qualitative, manual analysis of the collocational profiles of the most frequent noun bases in the four investigated languages, nine subtypes were identified within metaphorical and metonymic collocations, depending on the underlying semantic-cognitive processes. Several concrete examples found in the corpora for different noun bases have been used to illustrate these types in the following sections. Specifically, collocations that are motivated by the same or similar semantic processes in more than one or all four languages have been selected; sometimes, however, only examples from individual languages are given.