This collective volume focuses on connections between grammaticalization and (inter)subjectification and/or the directionality of change, both in early and late stages of grammaticalization. It divides into two halves each consisting of six articles, one concentrating on (inter)subjectification, the other on directionality, but some authors connect both issues. All contributions are written from a functional-cognitive perspective, and most are based on corpora. This review cannot do justice to the majority of specific problems raised in particular contributions, nor even address problematic premises and claims; see however the subsuming section at the end.

The first contribution by Heiko Narrog is meta-theoretical. It compares three concepts of (inter)subjectivity and (inter)subjectification and gives a synthesis exemplified through modal notions. The often vague usage of ‘subjectivityʼ comprises a cluster of concepts which should be delineated along three different dimensions: (1) speaker commitment/performativity, (2) accessibility of information, (3) construal. Dimension (3) is opposed to (1–2) and typical of Cognitive Grammar approaches (e.g., Langacker 1990). Narrog dubs them “conceptualist approaches“(p. 24), since only a top-down procedure is offered, and compares them with Traugott’s approach (e.g., Traugott 1995). Since dimensions (1–2) are easier to reconcile than either of them with (3), Narrog favours Traugott’s approach, as do the other contributors to the volume (except Chappell). Moreover, Narrog proposes to add “increase toward discourse-orientation” to subjectification and intersubjectification and subsumes all three under “speech-act orientation” (pp. 36-38).

There are three contributions dealing with the rise of complex modal predicates.

Karin Beijering presents a thorough diachronic case study of the Swedish cognate modals må ‘may, shouldʼ and måtte ‘may, mustʼ (< magha ‘have strength, powerʼ). In Modern Swedish they lost their dynamic and deontic meanings and postmodal (optative, concessive) meanings started to prevail, albeit with different distribution (pp. 52–55). Beijering claims that må and måtte have reached the final stages of grammaticalization, but since she does not specify properties of these final stages, one can only guess that she is referring to the stages before a unit becomes predominant in the post-modal domain (pp. 67–68).

Beijering also engages in an appealing theoretical discussion about the relation between (inter)subjectification, grammaticalization and other composite types of language change (pp. 68–75). Despite some weak points (not all notions are sufficiently defined) Beijering’s reasoning is sound, and she is very consistent in assigning features of primary and secondary grammaticalization to må and måtte (p. 69).

Adeline Patard and Johan van der Auwera present a careful diachronic corpus study on French semi-auxiliaries, dubbed ‘comparative modal constructions’ (CMCs): faire mieux de lit. ‘do better ofʼ, valoir mieux lit. ‘be worth betterʼ, falloir mieux lit. ‘must betterʼ. The CMCs have remained incipient and weakly integrated paradigmatically (pp. 108–9); in comparison to other modals (e.g., devoir ‘must’) their frequency is low, yet their distribution over TAM grams and extensions into postmodal meanings differs (pp. 86–88, 103–4). Importantly, the three CMCs have been different in frequency and grammatical distribution (including TAM grams and argument realization) from the late 13th century until today. This can be motivated largely by their original lexical meanings: „some of the synchronic morphosyntactic properties do not reflect a particular stage of grammaticalization, but rather mirror features of the source constructions“ (p. 101). Moreover, „entrenchment [measured in terms of frequency; BW] on the one hand and syntagmatic fixation on the other hand do not necessarily go hand in hand in cases of moderate grammaticalization“ (p. 101). All three CMCs have remained semantically transparent, so that „[o]ne must conclude that the grammaticalization of French CMCs has operated at the structural level but not at the semantic level“ (p. 105).

Thomas Egan has used corpus data to investigate the development of intersubjective meanings of Engl. fail to and not fail to since the late 14th century. His core hypotheses are: (1) intersubjective meanings are not extensions of subjective, but of objective functions; (2) while in objective and subjective meanings fail to and not fail to code semantic opposites (contraries), their intersubjective meanings are diachronically totally unrelated. Egan argues on the basis of Construction Grammar and uses Langacker’s notion of ‘entrenchmentʼ („intersubjective senses of a construction become entrenched in the grammar of a cross-section of the speakers of a language“). Nevertheless, he generally sticks to Traugott’s unterstanding of ‘(inter)subjectivity’ (pp. 169–172).

Egan’s hypotheses (1–2) are well-motivated. Although not entirely explicitly, his analysis demonstrates that the development of (not) fail to can be arranged in the order (a) ‘try and not succeedʼ > (b) ‘neglect a dutyʼ > (c) ‘fail to meet speaker’s expectationsʼ. Only (c) represents subjective usage. In general, the functional development of not fail to preceded that of fail to, at least until 1850. However, the first intersubjective uses of not fail to emerged in directive, not in commissive speech acts (pp. 174–5); they did not evolve from subjective usage. Egan’s findings correspond to Patard and Van der Auwera’s claim: the rise of intersubjective meanings of (not) fail to is not a result of grammaticalization, but simply a concomitant feature of their discourse functions at very early stages.

Besides (1–2) Egan shows that recently fail to has been grammaticalizing toward a mere negation marker, an extension which has been evolving from subjective use (‘disappoint speaker’s expectationsʼ), pp. 190–192. Relying on Boye and Harder’s (2009, 2012) understanding of grammaticalization as the conventionalization of discursively secondary usage, Egan shows that, although fail to has been subject to bleaching (the component of speaker expectation often proves absent from usage), it still can be targeted, i.e. behave as an item with primary discursive status. This proves that as negation marker fail to is only weakly grammaticalized.

Two contributions deal with clause linkage and discourse cohesion.

Gijsbert Rutten and Marijke van der Wal have investigated the change from d-forms into w-forms of relative adverbs and relative pronominal adverbs in Dutch relative clauses (e.g., Het huisdaar > waarik woon ‘The house there > where I liveʼ). The 17th-18th centuries are considered crucial for this process, so the analysis is conducted on the basis of a 17th c. corpus of private letters. The change was initiated by continuative relative clauses, as in She was found face down in the water and airlifted to hospital, where she died hours later. Such relative clauses demonstrate „a discrepancy between form and function“ (p. 114), inasmuch as „they signal to the reader that the discourse is coherent“, although the new clause typically supplies unexpected information (p. 133). Since this way of creating discourse continuity is sensitive to the addressee, one can treat it “as an intersubjective effort to create coherence“ (p. 113).

The analysis provides a textbook example of syntactic tightening ‘from discourse to grammar’. Following Givón (1979), the authors regard the development parataxis > hypotaxis > subordination (= embedding) as an instance of grammaticalization, since complexity increases (p. 116). They argue that continuative relative clauses do not represent a clear instance of hypotaxis and possibly do not even move toward subordination, but rather represent „dependency beyond the sentence“ (Mithun 2008: 69). They even consider these clauses to be an instance of insubordination (pp. 132–3). However, this explanation fails: the authors themselves argue (p. 133) that these clauses move into subordination, i.e. that they have not previously been subordinated (for relevant considerations with other cases cf. Wiemer 2017, 2019). Regardless, provided the change from d- to w-relativization qualifies as an instance of grammaticalization, this case study illustrates that intersubjectivity can arise as a by-product of grammaticalization.

Hilary Chappell’s article surveys discourse functions of say-verbs in Sinitic languages, in which mood and illocutionary distinctions are expressed only by constructional templates which differ in their arrangements of slots. Against this background Chappell reconstructs the evolution of the verb kong1 ‘sayʼFootnote 1 into a clause-final discourse marker in Southern Min (Taiwan, Sinitic). She identifies four clause-final usage types: (1) declarative assertions, employed to refute a presupposition inherent in the context; (2) suggestions in imperative form; (3) warnings in imperative form; (4) rebuttals in wh-interrogative form. The intersubjective component in all these utterance types amounts to the speaker’s assumptions about the interlocutor’s knowledge state; clause-final uses of kong1 „specifically involve the correction of a presupposition attributed to the other interlocutor“ (p. 157). Chappell regards this intersubjective component as more abstract (‘grammaticalizedʼ), but she also reconstructs structural changes which kong1-constructions must have undergone: (1) truncation in quotative remarks, (2) reanalysis of head-dependent relations and retainment of tone sandhi, (3) ensuing polyfunctionality of kong1 based on separate grammaticalization chains (complementizer, clause-initial topic marker and conditional conjunction, clause-final discourse marker); pp. 159–60. The analysis looks plausible. However, since the “chains” are rather short, one wonders whether the polyfunctionality of kong1 would not be better represented in a radial fashion.

Problems arise with Chappel’s notional apparatus. Apart from her maximally broad conception of ‘modalityʼ, which includes all sorts of rhetorical devices and illocutionary force (p. 142), she does not restrict ‘intersubjectivity’ to formally coded means of social deixis and face—as do Traugott and the majority of the contributors to the volume—but considers it „intrinsic to the communicative process, whereby pragmatic features of context that provide the conditions of use for a particular syntactic structure are necessarily coded as part of the constructional meaning“ (p. 140).

The second half of the volume concentrates on directionality of change.

Luisa Brucale and Egle Mocciaro present a study on the semantic network of per used as preverb (= prefix or verbal particle) and preposition in Early Latin (3rd–2nd century BC). Their study is based on Cognitive Grammar and pursues the central claim that preverb and preposition, though sharing the same semantic nucleus and network of spatial meanings, differed in their developmental paths, which arose from different collocations of erstwhile adverbs as modifiers either of verbs or of nouns. In Latin, we can observe only the outcomes of these processes (p. 201), in which the original adverb became subject to different syntagmatic constraints.

The authors argue for an interplay between grammaticalization and lexicalization, which is not always easy to follow, mainly because they do not elaborate much on the commonalities and differences between both processes. Their otherwise thorough and detailed analysis brings to light interesting findings, but they sometimes stop short before more far-reaching conclusions. For instance, the more abstract meanings of per- (e.g., durative/intensive) were restricted to a small number of verb stems, and per- did not develop beyond the stage of a telicizing morpheme (p. 205), i.e. it did not develop into a perfectivizer (= marker of mere limitation); cf. Arkadiev (2014, 2015), Wiemer/Seržant (2017) for a larger background. Now, the authors consider such restrictions of lexical input, together with semantic opacity of the preverb-stem complex, indicative of „a different process (…) which involves specific lexemes only and which can be analyzed in terms of lexicalization“ (p. 229). This conclusion is remarkable since it means that lexicalization can follow on (rudimentary) grammaticalization, while in other cases lexicalization has been identified as a result of change that may feed grammaticalization (Lehmann 2002). Thus, this case study supplies an argument for the claim that there is no unitary sequence between lexicalization and grammaticalization.

Andrzej M. Łęcki and Jerzy Nykiel deal with the rise of the English adverbial subordinators in order that and in order to, which appeared in the transition from Middle to Early Modern English (14th–17th c.). While in order to triggered the onset of relevant processes and preceded analogous processes with in order that by about 300 years, the latter, probably following the model of in order to, developed more quickly (after the mid-18th centutry, p. 246). Grammaticalization consisted in decategorialization, decrease of paradigmatic and syntagmatic variability, and in an increase of bondedness. Łęcki and Nykiel convincingly show that the source meaning for purpose was manner: ‘in order, in sequenceʼ > ‘in proper condition, correctlyʼ (p. 241). The transition to purpose consisted in syntactic reanalysis and was probably catalyzed by an element of intention ascribed to an agent in the clause to which in order attached as an adverbial modifier.

Łęcki and Nykiel assume that the grammaticalization of in order-expressions became manifest in a change from preposition to conjunction (pp. 247–253), but they do not ask how the increase of scope (over NPs > over clauses) can be made consistent with structural, Lehmann-like criteria of grammaticalization, by which the authors otherwise abide. Anyway, a check of the other criteria shows that in order first became a complex preposition before it was reinterpreted as a conjunction. Plausibly, in order that was modelled in analogy to in order to, accompanied by expansion of the model to finite clauses. We can thus, to some extent, speak of renewal, mainly in formal genres (pp. 247–50). However, Łęcki and Nykiel assume that renewal pertains to grammaticalization (p. 254); pace Lehmann (2015 [1995]: 20–1, 2004: 156–7, 160–2).

Björn Hansen develops a typology of six post-grammaticalization processes, which he applies to constructions with modals in Russian, Polish, Czech, and Serbian/Croatian.

Secondary grammaticalization is a process by which already grammaticalized markers/constructions expand into neighboring semantic domains. A case in point is the employment of the negated imperative form of the Serbian/Croatian auxiliary moći ‘can’ as a prohibitive marker (p. 263).

Marginalization consists in a decrease of frequency and leads to a peripheral position of elements in the language system (e.g., stylistic restrictions or limitation of lexical input); p. 264. This process may precede retraction (see below), and it correlates with lack, or decrease, of entrenchment. Hansen examines four examples, all of them related to Pol. mieć ‘have’.

Degrammaticalization, which has often been denied or regarded as very infrequent, is defined, following Norde (2009: 120), as „a composite change whereby a gram in a specific context gains in autonomy or substance on more than one linguistic level“ (p. 266). Hansen argues for Czech nemusím ‘I dislikeʼ (+ accusative complement) as a case in point. Concomitantly, the change ‘lack of necessity’ > ‘dislike’ includes also morphological decategorialization: nemusím ‘I dislike’ is very infrequent in the future tense and it does not occur in the infinitive.

Retraction is understood as the opposite of expansion (pp. 269–70). Hansen points out three cases: (1) general participant-external impossibility > circumstantial impossibility (Pol. niepodobna); (2) participant-external necessity + probability (i.e. epistemic) > loss of epistemic meaning (Russ. nado(bno)); (3) general participant-external necessity > deontic necessity (Russ. podobaet). The functions right to the arrow (>) are the ones that remain.

According to Brinton and Traugott (2005: 144), lexicalization implies loss of semantic, often also of morphosyntactic, transparency; in contrast to marginalization, lexicalization does not yield a new grammatical marker. Hansen points out two types from the domain of Slavic modals: (1) semantic specialization in the domain of epistemic modality; (2) transition into idiomatic phrases with fully lexical content. Hansen’s analysis (which, via objections, raises more intriguing insights into a typology of change) is most insightful.

Finally, in contrast to lexicalization, which needs some specific communicative context and is a gradual process, word derivation takes place on the level of the lexicon and occurs abruptly (p. 276). Among Slavic modals, this process is restricted to perfective-imperfective aspect, which is a grammatical opposition based on stem derivation (Wiemer/Seržant 2017). Perfective stems are predominantly derived via prefixes, and this device is applied to a small number of possibility modals (which count as imperfective verbs). Prefixes transfer them into the class of perfective verbs, with the consequence that they code factual events, e.g. Russ. s-moč’ ‘manage (to do)ʼ (< moč’ ‘can’). Thus, although these perfective verbs have the same syntactic properties as the imperfective modals from which they derive, they cease being modals from a semantic point of view.

Helle Metslang examines polysemy copying and its role in ‘forced grammaticalization’. The latter takes place if „a grammatical resource of a language is formed from the material of the language itself straight into its final-stage shape without a natural step-by-step grammaticalization process“ (p. 283). Contact-induced and language-internal forced grammaticalization may be distinguished, together with a further distinction (in either case) between (1) innovations based on foreign models from which the polysemy of some function word or a construction has been copied, and (2) abrupt innovations for which one cannot even assume intermediate stages. All these phenomena are attested in the history of Estonian, from which Metslang draws a series of case studies, among which are the emergence of articles and of the future tense in the early periods and five phenomena examined for the modern period (20th–21st c.). All innovations basically superseded analytic constructions (following Germanic models), and they were deliberately introduced in order to undo, or counterbalance, analytic structures.

Metslang’s presentation is rich in interesting details, concerning both the socio-political conditions of change and the structural changes themselves. Metslang is aware that polysemy copying differs from contact-induced grammaticalization (cf. Heine and Kuteva 2005). However, the formulation of ‘forced grammaticalization’ is no less problematic; in fact, Metslang uses this term in contexts when grammaticalization does not appear to be an issue. Simultaneously, one wonders how forced grammaticalization can be distinguished from the implied opposite: ‘natural evolution’, or ‘spontaneous grammaticalization’: which factors eventually determine „the general structural tendencies of a language“ (p. 304)?

Deborah Ziegeler’s challenging article on pattern replication integrates communicative and sociolinguistic factors of contact-induced grammaticalization with general cognitive tendencies assumed as operative in grammaticalization paths. Here I can summarize only the main threads and conclusions, a wealth of intriguing details and considerations cannot even be mentioned.

First, Ziegeler considers contact in grammaticalization to be the norm and, in accordance with Heine/Kuteva (2005), argues that grammaticalization always runs through the same stages regardless of the degree to which contact is involved. Contact can accelerate grammaticalization or slow it down, in either case communicative pressure is at play (pp. 314–5). Second, various kinds of ‘versals’ (Angloversals, Vernacular Universals, etc.) should be accounted for in terms of grammaticalization (p. 320) and given more prominence in explanatory and usage-based accounts of pattern replication. Third, Heine/Kuteva’s (2005) problematic conception of replica contact-induced grammaticalization (CIG), implying that speakers of the replica language repeat the entire grammaticalization process which they think to have taken place in the model language, makes sense under specific conditions. Such conditions apply in creoles, or creole-like New English) varieties. Correspondingly, Ziegeler produced corpus-based case studies on Singaporean English and Indian English, with East African English and British English as control varieties. She investigated (1) will as habitual aspect marker, (2) the extended use of the progressive, and (3) one as determiner in noun reference. She argues that these features are not replications of their model patterns triggered by substrates, rather they replicate developmental patterns in the lexifier (English), which existed many centuries ago. Naïve implications carried by the model of replica CIG can be overcome if we assume that the functional polysemy of lexical source expressions must have remained transparent, and that implicational hierarchies between more and less concrete meanings of a source expression always operate (p. 343).

Cognitive principles operative in implicational hierarchies can be forced under communicative pressure: “The replication of the grammaticalization path is thus not only cognitively motivated, but also functionally driven in the mixed language situation” (p. 345). This general conclusion gives explanatory models of convergence via contact a more pronounced sociolinguistic nuance, but it also raises the requirement that, instead of looking for innovative features in high-contact varieties, we should also look for possible archaic features whose origin might be found in the preservation of obsolete minor patterns or in the re-invention of “missing links” of grammaticalization processes.

Freek Van de Velde and Béatrice Lamiroy analyze external dative (or indirect) possessors in West Germanic and Romance. They claim that their decline should be viewed as a side-effect of a general long-term move toward NP-configurationality (p. 390) and that this movement is indicative of constructional grammaticalization (p. 377). Extant substrate theories must be dismissed as “naïvely simple”, too anglocentric, and because realistic periods of contact do not correlate with the demise of external possessors (p. 393). Instead, urbanization is a better predictor of why external possessors retreated more clearly in English, Dutch and French than elsewhere in Europe (p. 393). Moreover, accounts of external possessors (especially with dative pronouns) as a typical SAE-feature rather focus on an epiphenomenon for a variety of reasons, which the authors convincingly scrutinize (pp. 372–3, 382–3).

Van de Velde and Lamiroy also argue that the versatility of the external possessor is inversely correlated with the grammaticalization degree of the definite article (pp. 384–389). Their argument is convincing, except for the combinability of articles with possessive pronouns: many languages whose article systems are regarded as particularly advanced (e.g., Modern Greek, Armenian, Norwegian Landsmål) allow for a combination of the definite article and possessive adjective, or even require it, sometimes newer articles can be combined with older ones (as in Swedish). Possibly, the solution to this conundrum lies in assuming different roads along which a demonstrative can spread and develop into a definite article.

To sum up, this volume is a substantial, thought-provoking contribution to the fields indicated in its title. Unfortunately, notwithstanding its coherent macrostructure, the convergence of theoretical positions and conclusions is hardly supported by cross-references between contributions. With some authors core notions, including ‘grammaticalization’, remain vague (e.g. Narrog, Chappell, Metslang), while other authors are very clear and consistent (Beijering, Patard and Van der Auwera, Egan, Hansen) or admit that grammaticalization is debatable (Ziegeler, Van de Velde and Lamiroy). In conclusion, we may comment on the core notions as follows.

Grammaticalization among other processes Only Beijering proposes a comprehensive approach with critical meta-discussion; Hansen has elaborated on a typology of post-grammaticalization phenomena. Persistence, or layering (Hopper 1991), plays a prominent role in most contributions. In general, however, the placement of grammaticalization within a typology of language change (including “primitive” types, indicated by Beijering) still requires more consistent treatment.

Most authors agree that grammaticalization and (inter)subjectification are independent processes. Subjectification can accompany early stages of grammaticalization, while intersubjectification is at least more typical of late stages (Narrog), although it does not necessarily imply preceding subjectification; instead, both can develop in parallel if objective meanings “bifurcate” (Egan). Moreover, intersubjective meanings can develop without any significant grammaticalization process and result mainly from the strengthening of lexical source meanings (Patard and Van der Auwera). The relation of both to discourse-oriented uses seems to be least clear (compare, for instance, Narrog against Rutten and Van der Wal).

External influence on directionality has been treated in only two papers. Metslang shows that changes in Estonian that are not in agreement with expected directions of grammaticalization can be explained as results of codification or language planning aiming at the restitution of assumed core properties of the language family (Finnic). Rutten and Van der Wal argue that the change from d- to w-based relative clauses in Middle Dutch was stimulated by the written medium, supported by Latin models, which might have led to an increase of (already existing) continuative relative clauses. One wonders, of course, to which extent grammaticalization processes (as well as possibly correlated (inter)subjectification) examined in other contributions (e.g., Patard and Van der Auwera, Egan, Łęcki and Nykiel, Van de Velde and Lamiroy) might not have been supported by the written medium as well. Only Ziegeler’s article on New Englishes makes a dedicated (and substantial) contribution to tenets of contact-induced grammaticalization, while Metslang’s investigation of ‘forced grammaticalization’, despite this name, focuses on phenomena which run counter to grammaticalization or are hardly related to it.

Entrenchment can be evaluated on the basis of frequency increase, which however is not always accompanied by other processes associated with grammaticalization. Conversely, (increase of) frequency is not always a reliable indicator of grammaticalization, although emergent patterns may demonstrate intriguing distributions over grammatical contexts (e.g. modal concord or restrictions with TAM grams); see Beijering, Patard and Van der Auwera, Hansen. Alternatively, entrenchment can be judged from a sociolinguistic perspective (Metslang), as can marginalization (Hansen).

Finally, the volume provides hardly any meta-discussion on the issue of primary versus secondary (or derived) meanings; an exception is Egan’s paper (pp. 183–190). This is astonishing, given that many authors pay attention to persistence (see above), and some explicitly deal with shifts from pragmatic (i.e. context-driven) to semantic (i.e. coded) meaning components, which result from pragmatic strengthening.