The key terms are defined straightforwardly as in (4)–(6). It can be noted that these definitions are completely parallel to the definitions of synexpression and its subtypes (see §5 below).

(4) coexpression (of two meanings A and B):

= expression of either A or B by a minimal form or a construction

(5) colexification (of two meanings A and B):

= expression of either A or B in a root

(6) cogrammification (of two meanings A and B):

= expression of either A or B in a grammatical marker

The term colexification was coined by François (2008), and it was generalized to coexpression by Hartmann et al. (2014). The two main types of linguistic expression are expression by lexical forms (or lexification) and expression by grammatical forms, which we can call grammification (a neologism introduced here for terminological symmetry). Then, by analogy with colexification, the availability of different meanings in a grammatical marker can be called cogrammification (rather than “grammatical polysemy,” as in much earlier work such as Emanatian, 1991; Goddard, 2003).

For simplicity, this paper only mentions situations where two meanings are (or can be) expressed by a minimal form, but of course there can be more than two meanings, both in coexpression patterns and in synexpression patterns.

As I mentioned, lexical and grammatical coexpression patterns have often been described with the term polysemy, which goes back to Bréal (1897) (e.g., Nerlich et al., 2003; Gries, 2015). But as we saw, there is often no reason to think that the forms that are compared across languages must have several meanings from a language-internal perspective, so coexpression is a better term for contexts in which different languages are compared. It is well-known among semanticists that distinguishing between polysemy (or ambiguity) and indeterminacy (or vagueness) is rather difficult (e.g., Geeraerts, 2001; Riemer, 2010: §5.3), so it is often more practical to ignore this distinction, as is done with the term coexpression (and especially colexification, when applied to word meanings).⁷ Another term that linguists sometimes use is multifunctionality, where the vague term function seems to stand for a comparison meaning. Finally, Bybee et al. (1994, p. 44) talk about different uses of tense-aspect-mood forms in the world’s languages, by which they also mean comparison meanings.

Another widely used term in comparative semantics is categorization: Linguists often say that different languages categorize particular domains (e.g., the human body, kinship relations, perception) differently by their lexical items (e.g., Koptjevskaja-Tamm et al., 2016, p. 434), or that they have different semantic categories (Evans, 2011: §2). And often they say that meanings are lexicalized in different ways in different languages. As these terms (categorization, lexicalization) are used with other senses elsewhere, it would be clearer to talk about different lexifications in different languages. (And more generally, “lexical typology” is perhaps better called lexification typology).

For grammatical markers, the term syncretism has been used for quite some time in a synchronic sense (e.g., Plank, 1991; Baerman, 2007). For example, Latin homo ‘human, man’ has the inflectional paradigm in (7). The singular has five different forms, but in the plural, nominative and accusative are “syncretized” (i.e., they are cogrammified), as are dative and ablative.

(7)singularplural

nominative homo    homin-es

accusative  homin-em  homin-es

genitive homin-is homin-um

dative homin-i homin-ibus

ablative homin-e   homin-ibus

The term syncretism has also been used for coexpression patterns of grammatical markers beyond inflectional paradigms (e.g., complementizers, Baunaz and Lander, 2018; voice markers, Bahrt, 2021), but it is a rather odd and opaque term (originally referring to mixtures of beliefs), and it can easily be replaced by coexpression, or more specifically cogrammification. (Instead of saying that a marker is syncretic, we can say that it is coexpressant, but it should always be made clear whether this is meant with respect to general meanings or language-particular meaning distinctions.)

The opposite of coexpression can be called “disexpression,” and in particular, we can talk about dislexification (François, 2022):

(8) dislexification (of two meanings A and B):

= expression of A and B by two different roots

This term can only be used about languages; for example, we can say that German dislexifies ‘go by foot’ (gehen) and ‘go by vehicle’ (fahren). Another term is partial colexification (List, 2023):

(9) partial colexification (of A and B)

= expression of A and B in composite forms that contain the same root

An example is German Tuch ‘cloth,’ which partially colexifies ‘towel’ (Hand-tuch ‘hand-cloth’), ‘shroud’ (Leichen-tuch ‘corpse-cloth’) and ‘sheet’ (Bett-tuch ‘bed-cloth’).

While colexifcation patterns have become quite famous also among psychologists in recent years (e.g., Jackson et al., 2019; Xu et al., 2020; Brochhagen and Boleda, 2022), cogrammification was the first prominent domain of coexpression studies in linguistics. Georgakopoulos and Polis (2018) survey the tradition of “semantic map studies” going back to work such as Anderson (1982), van der Auwera and Plungian (1998), and Cysouw et al. (2010).⁸ Most of this work deals with meanings of tense-aspect-mood markers, case markers, connectives and other grammatical markers, but Haspelmath (2003) already mentions the colexification patterns as highlighted by Hjelmslev (1953). Since the term “semantic map” has rather different senses outside of linguistics, it seems best to call such representations “coexpression diagrams”:

(10) coexpression diagram

= a graphic representation of coexpression relationships

Two different types of representation have become well-known: On the one hand, connectivity diagrams, i.e., simple graph representations with connecting lines showing possible coexpression sets [as in Figure 1, from Malchukov, 2004]. These diagrams have been used for small datasets where semantic analysis is at the center of attention.

On the other hand, linguists have used proximity diagrams, based on clustering techniques such as multidimensional scaling. Figure 2 shows the range of uses of argument-coding markers in two different languages, based on the valencies of 80 comparable verbs in three dozen languages (Hartmann et al., 2014, p. 473). In such proximity diagrams, spatial closeness of two dots indicates that the meanings they represent are often coexpressed.

If a coexpression diagram is based on a (reasonably) representative set of languages, it embodies claims about universals of coexpression: Those meanings that are linked by a connecting line (or are in close proximity) are claimed to be very likely to be coexpressed in some language. In Haspelmath (1997) and some subsequent work (e.g., Hengeveld and van Lier, 2010), they are therefore called implicational maps. For example, the map in Figure 1 entails the implicational universal that if a connective marker has an adversative and a correction meaning, it also has a contrastive meaning.

When there is syncretism or coexpression within an inflectional paradigm, a common way of representing the pattern is by means of underspecification. For example, instead of positing two different suffixes -(e)n (1st and 3rd plural) in German verb inflection (as shown in 11), we can posit the underspecified suffix -(e)n [+pl] (see 12). Since the suffix -t₂ is more highly specified, it takes precedence over -(e)n in the second person plural.

(11) German verb inflection

(12) underspecification analysis (following Wunderlich, 1996, p. 95)

Underspecification allows formally elegant analyses of syncretism patterns in particular languages, but the cross-linguistic significance of such analyses is sometimes unclear (see Baerman, 2007; Hein and Weisser, 2021 for recent overviews of the literature in syncretism). These analyses are relavent to universals only if strong assumptions about an innate feature inventory are made.

In a different research tradition, Greenberg (1966) included inflectional syncretism in his discussion of “markedness” asymmetries,⁹ observing that it is especially in “marked” environments (such as plural or past tense) that inflectional syncretism occurs [both are illustrated by the German paradigm in (11)].

In yet another tradition (with its source in generative syntax), coexpression universals are expressed by the notion of a functional sequence, a set of hierarchically arranged functional categories that are assumed to be given in advance, presumably innately (see Cinque, 2013). For example, Baunaz & Lander (2018, p. 8) consider demonstratives, complementizers, relativizers, wh-pronouns and indeterminate pronouns, and posit a functional sequence “Dem > Comp > Rel > Wh > Indet” to explain the widely observed coexpression patterns in European languages, e.g., English that (Dem/Comp/Rel), Spanish que (Comp/Rel/Wh), Serbo-Croatian (Comp/Rel/Wh/Indet) što, German was (Wh/Indet). De facto, this approach is similar to implicational maps, but it is practiced by linguists from different communities (see also Hein and Weisser, 2021: §4).

The next key terms are defined straightforwardly as in (13)–(15), in a way that is fully parallel to the definitions of coexpression and its subtypes (see §3.1 above).

(13) synexpression (of two meanings A and B):

= expression of both A and B by a minimal form or a construction

(14) syllexification (of two meanings A and B):

= expression of both A and B in a root

(15) syngrammification (of two meanings A and B):

= expression of both A and B in a grammatical marker

The initial examples of syllexification that I gave in Table 1 and in (3) above were English bequest ‘give + as an inheritance,’ and kitten ‘cat + young.’ As the notion of syllexification is less familiar than colexification, I give more examples of syllexification in Table 2, with counterparts in other languages showing that many meanings need not be expressed as in English. Thus, the German Fahr-rad [ride+wheel] corresponds to syllexified bike in English, and the French fauteuil corresponds to the circumlexified arm + chair in English.

Syllexification can also be exemplified by suppletive comparatives such as English worse (‘more + bad’) and French mieux (‘more + good’; cf. bon ‘good’), or by suppletive ordinals such as Russian pervyj (‘one + − th’) and English second (‘two + − th’).

As in the case of coexpression, when we say that a form synexpresses several meanings, we do not claim that these meanings should exist in the language in question, but we are making statements on the basis of comparison meanings. There are probably many cases of synexpression of meanings that would not be easy to render in a non-synexpressed way in the same language, as when a language has a word for ‘float,’ which means ‘move on the surface of a liquid,’ and it lacks a general word ‘liquid.’ In such cases, a syllexification view is motivated only by a comparative perspective and cannot be justified language-internally.

Syngrammification (i.e., synexpression of grammatical meanings, or cumulation) is generally taken to imply that the meanings exist independently, e.g.,

(16) a. case + number:Latin omn-ibus ‘to all’ (cf. 3b)

b. tense + aspect:   French décriv-ai-ent ‘(they) described’ (cf. 3b)

c. tense + person:  Latin vid-i, vid-isti, vid-erunt

‘I saw, you saw, they saw’

However, case and number are never expressed separately in Latin, and imperfective aspect is never expressed separately in French, so from a language-particular perspective, one might treat these as single meanings. This is even more the case with person + number, which are very often syngrammified, not only in person indexes (bound person forms), but also in independent personal pronouns, as illustrated in (17). Only the first person plural independent person form is usefully analyzed as containing a plural suffix (-s), while the other forms show synexpression of person + number.

(17) Spanishindexesindependent pronouns

1 + sg-o/−eyo

2 + sg-s(te)tú

3 + sg-Ø/−oella

1 + pl-mosnosotras

The way linguists have talked about synexpression is even more varied than the terminology for coexpression. Hockett (1947, p. 333) used the term portmanteau morph for forms like French au [o] ‘to + the’ or Spanish -é ‘1sg + preterite’ (in canté ‘I sang’), and this term is still common.¹⁰ Another well-known term for syngrammification is Matthews’s (1972) term cumulative exponence (contrasting with separative exponence, Bickel and Nichols, 2007: §1.5.1).¹¹ In Distributed Morphology parlance, cumulative exponence has often been treated in terms of “feature bundling” (e.g., Matushansky, 2006).

For syllexification, there are a number of earlier terms which are familiar, but not often thought of as technical terms. In Talmy’s (1985, 2000) famous paper about “lexicalization patterns,” he talks about conflation of different meanings in a single form (e.g., French entrer ‘go into’ conflates motion and path meanings), and he also mentions incorporation (used by Gruber, 1965). Ullmann (1957, 1966) distinguished between motivated and unmotivated words (e.g., German Hand-schuh lit. ‘hand-shoe’ vs. French gant ‘glove’). In Seiler (1975), the distinction was called description vs. labeling (e.g., German Lehr-er ‘teach-er’ vs. Arzt ‘doctor’). Ježek (2016, p. 7–11) distinguishes between synthetic vs. analytic lexicalizations (e.g., Italian cenare vs. English have dinner). Urban (2012, 2016) talks about analyzability of lexical items.

The opposite of synexpression can be called circumexpression, and in particular, we can talk about circumlexification:

(18) circumlexification (of meanings A and B)

= expression of A + B in two roots corresponding to A and to B

For example, we can say that the meaning of French fauteuil ‘armchair’ (seen in Table 1) is circumlexified in English by means of the two roots arm and chair. (The term circumexpression is of course inspired by circumlocution.)

This subsection briefly introduces three main types of explanations for general coexpression patterns which appeal (i) to conceptual similarity, (ii) to the likelihood of semantic extension in diachronic change, or (iii) to the need for maintaining informativeness. These explanations need not be mutually incompatible.

This subsection briefly mentions two types of explanatory approaches that are even more tentative than the explanations suggested in §7.1. It should be noted that from the perspective of language acquisition (and the evolutionary emergence of complex language), synexpression is actually the default that we start out with, and what needs explaining is circumexpression (i.e., non-holistic, composite expressions). Thus, especially the second type of explanation (in terms of frequency) is really an explanation of circumexpression.