Pseudo-grammaticalization: The anatomy of come in Modern Hebrew pseudo-coordination constructions

The paper examines the pseudo-coordination construction featuring the verb come preceding a lexical verb in Modern Hebrew, and shows that this is a mono-clausal mono-eventive construction, which did not emerge via a process of grammaticalization. That is, there is no tightening of internal dependencies between parts of the construction (Haspelmath 2004), nor evidence of a lexical unit starting to assume grammatical functions (Heine, Claudi & Hünnemeyer 1991). I go on to argue that, in this particular construction, the verb come is a “lexical restructuring verb” (Wurmbrand 2004, 2014), whose lexical properties do not differ from those of ‘simple’ change-of-location uses of come in that both feature a deictic meaning component. Particular attention will be paid to what looks like the absence of a motion component, suggesting that even if simple come selects for a prepositional complement, it does not necessarily encode a motion component, and therefore the absence of the PP, in a complex verb construction is not tied to loss of motion, but merely to a change in the type of complement. The current account provides substance to claims stressing a metaphorical relation between the two occurrences of come, since it points to the close similarities in the lexical-pragmatic properties of this lexeme in its two environments of use, and locates the difference between them in the choice of complement that produces the effect of transfer from the location realm to a more abstract one characteristic of metaphoric meanings.

2018, Rubinstein 2019, a.o.). Alongside these well-cited and recurring cases of grammaticalization, Devos & Van der Wal (2014) turn the attention to less studied ones, where the verbs come and go give rise to categories not directly related to TAM, such as passive forms and discourse markers, and where, at times, the construction resulting from grammaticalization seems less prevalent cross-linguistically.
However, not all complex verb sequences presenting mono-clausality and mono-eventivity are cases of grammaticalization. Butt and colleagues, in work on Indo-Aryan languages (Butt 2003, Butt & Geuder 2001, Butt & Lahiri 2002, Butt & Ramchand 2005 discuss constructions, in which the motion verb is a Light Verb, not on the way of becoming an auxiliary. Light Verbs are shown to be a category in their own right, presenting distinct properties compared both to lexical verbs and auxiliaries (cf. Snider 2018 on Light Verbs in Biblical Hebrew). Butt (2003) and Butt & Lahiri (2002) provide diachronic data to show the stability of the Light Verb construction over centuries of documented text, forcefully indicating the absence of a grammaticalization process, such as defined in Traugott & Hopper's (1993, p. 108) grammaticalization cline.
The aim of the current paper is to pursue this line of research, claiming that not every monoeventive complex verb construction involving a motion or position verb combined with a lexical verb is an immediate candidate for grammaticalization. Concretely, I examine pseudo-coordination constructions featuring the verb come in Modern Hebrew, and show that, while this is a mono-clausal mono-eventive construction, the distributional and semantic properties of come in this particular setting do not point to there being a tightening of internal dependencies between parts of the construction (Haspelmath 2004), nor a lexical unit starting to assume grammatical functions (Heine, Claudi & Hünnemeyer 1991). I go on to argue that the verb come is not a Light Verb, but a fully lexical verb (pace Bjorkman 2009, Bruening 2015 on V-V andV-N light verb construction in English). This is shown to be so on the basis of distributional properties, and the availability of meaning components characteristic of 'simple' change of location -centrally a deictic meaning component. As for the motion meaning component, I suggest that it is not absent, since it is not entirely clear that it is there to begin with. The picture that emerges from this discussion is that the subcategorization frame of the lexical verb come is such that it does not only select for locational complements, but also clausal ones, and that these selectional patterns are a stable property of come throughout the centuries. This way of viewing the relation between 'simple' come and pseudo-coordinative come clarify the structural and lexical semantic basis for a metaphorical interpretation: namely a transfer from the location realm to a more abstract one.
The paper is structured as follows: section 2 provides background on pseudo-coordination constructions, situating the Modern Hebrew (MH) ones among those discussed in the literature. The next three sections are each dedicated to showing that main features of grammaticalization cannot be identified in the MH pseudo-coordination construction. Section 3 lays out the distributional properties of the verb come in the discussed construction showing that it is a lexical restructuring verb in the sense of Wurmbrand (2004Wurmbrand ( , 2014, to be detailed below, and that therefore there is no change from a lexical into a functional category in this case. Section 4 shows that, contrary to a more general view in the literature on pseudo-coordination constructions, there is no TAM-related grammatical category encoded by the structure, nor is come a means to encode mirativity. Section 5 closes with a discussion of the lexical properties of MH ba 'come' pointing to a deictic and a doxastic modal component, crucially available in both the change-of-location construction and the pseudo-coordination one. This section also takes issue with the motion meaning component, which seems to be lacking in the pseudo-coordination construction, arguing that in fact come as an achievement does not encode a complex path, but a simple change of location. 2. Pseudo-coordination constructions. Pseudo-coordination constructions are attested in numerous languages, unrelated genealogically or geographically (cf. Cruschina 2017, Ross 2016, Biberauer & Vikner 2017, Wiklund 2008, de Vos 2004, Stefanowitsch 1999 et seq. among many others). 1 One of their most salient properties is the inferences of surprise, suddenness, unexpectedness that they give rise to. Here are some examples from English, Swedish, Malayalam and Palestinian Arabic: (1) a.
English ( Distributionally speaking, pseudo-coordination is not like regular coordination. De Vos (2004) shows that the first verb of the pseudo-coordination construction does not pattern like a lexical verb nor a modal; most tellingly, extraction from a conjoined VP is possible in the case of pseudo-coordination, contrary to regular VP coordination (see Ross 1967).
What has John gone and done __ now? b.
*What has John painted a house and eaten __? (De Vos 2004: ex. 4) Many of the above cited authors analyze these constructions as aspectual (de Vos 2004, Wiklund 2008, Gamliel & Mra'i 2015, mentioning ingressiveness or inception as the main meaning contribution of come/go and accounting for the 'special' inferences of suddenness and surprise as being a side effect produced because of the specific aspectual setting available in the construction. Others place the special inferences at the center and attribute a modal source to them, relying on metaphorical parallelisms between change of location come and its instantiation in the pseudo-coordination construction (Kuzar 2006, Bourdin 2014, Ross 2016. Colloquial Modern Hebrew has pseudo-coordination constructions featuring the verbs come and go (see Kuzar 2006, Gamliel & Mra'i 2015. These are exemplified in (3) and (4), respectively.
Modern Hebrew (Stefanowitsch 1999: p. 127 As the translations to these examples suggest, all convey an inference of unexpectedness or unacceptability in their context of use. Example (3a) says that Dan's opening of the box is unexpected given the reticence to do so on the part of the others in the context; (3b) indicates that the unaccepted/unexpected behaviour of the referent of the subject DP is counter to what he was asked to do. Next, (4a-b) differ from the cases in (3), since the underlying proposition is counter to some received norm held by the speaker, namely, that it is not acceptable to laugh at people's expense, but the referent of the subject DP does so nevertheless, or that it is not acceptable/expected to cancel leaves, and therefore the particular occurrence with the First Sergeant is counter to the speaker's expectation.
In other words, the examples with come and go seem to differ with respect to the type of unexpectedness concerned. Whereas the ones with the verb ba ve-'come and' convey that the underlying proposition is unexpected or unacceptable according to some contextual social norm relevant to the speaker, and possibly also to the addressee , in the ones featuring go, the unexpectedness is circumstantial, namely dependent on a particular contingent contextual setting. Thus, even though the two constructions seem very close in their meaning, ba ve-'come and' cannot replace halax ve-'go and' in the particular contexts provided in (3a-b), since this context does not readily allow the accommodation of an appropriate norm relevant from the speaker's point of view. 2 That only unacceptability/unexpectedness vis-à-vis some social norm is at stake can be observed when considering examples such as (5), where unexpectedness related to physical processes or natural occurrences cannot be conveyed using ba ve-'come and': (5) # ha-ʃamayim ba'u ve-hitkadru. the-skies came and-darkened Intended: 'The skies unacceptably/unexpectedly darkened.' Note that the same sentence with halax ve-'go and' is infelicitous as well, 3 and this might lead one to think that there is a ban on inanimate subject DPs.
(6) # ha-ʃamayim halxu ve-hitkadru. the-skies went and-darkened Intended: 'The skies unacceptably/unexpectedly darkened However, whereas inanimate subjects are strictly excluded in the case of halax ve-'go and' constructions, they are indeed attested with ba ve-'come and' constructions if the proposition includes an unaccusative/passive verb (7), or the DP subject refers to abstract notions such as the law, codex, and related ones (8). In these cases, the underlying proposition is understood as uttered against a normative background relevant from the speaker's point of view.
(7) kol ma ʃe-'ata bone be-hamon haʃka'a all what that-you build in-a.lot application {ba/*holex} ve-neheras lexa yom exad. comes/goes and-destroys to.you day one 'Everything that you build with great application gets destroyed on you one day.
[This is unacceptable from the perspective of the speaker]' 2 A similar contrast seems to hold in Palestinian Arabic (Abu-Gosh dialect, near Jerusalem; Nadine Abdel-Rahman, p.c.), where the presence of the verb come (i) conveys that the speaker conveys that the company's conduct is unacceptable, since, presumably, companies should not act this way, and in (ii) with the verb go the speaker focuses on the circumstances related to the subject DP, sympathizing with the singer. (8) CONTEXT: The criminal law seeks to forbid the marriage but… ha-miʃpat ha-ezraxi {ba/*holex} ve-makir bi-zxuyot mamoniyot the-law the-civil comes/goes and-acknowledges in-rights financial ʃe-tsomxot me-ha-keʃer ha-ʃeni. that-stem.PL from-the-relationship the-other 'The civil law (unexpectedly) acknowledges the pecuniary rights that stem from the other relationship.' Thus, example (7) is uttered against a normative background whereby things which are built with a lot of application should not be destroyed in one day, and therefore that it is unacceptable for such a thing to happen; similarly, example (8) indicates that the civil law contrasts with the criminal law in acknowledging pecuniary rights that stem from a certain type of relationship, possibly to the approval of the speaker. Crucially, in both (7) and (8) halax ve-'go and' is excluded.
For the sake of the current discussion, I set aside constructions featuring halax ve-'go and', and focus mainly on the ones with come. The basis for the analysis of these constructions set in , Boneh & Abarbanel, and Abarbanel (2019), to be further developed in section 5, is sketched here stating that in the case of examples in (4) the unexpectedness/unacceptability inference arises from adding the proferred content (namely he laughs at everyone's expense (4a); the First Sergeant cancelled all leaves (4b)) to the discursive background norm held by the speaker, and possibly also the addressee (respectively for (4a) and (4b) one should not laugh at others' expense; leaves should not be cancelled). This way of capturing the meaning of pseudo-coordination constructions featuring ba ve-'come and' takes inspiration from the discourse update literature, where the asserted proposition is added to a discursively available set (for different ways of going about defining the relevant discursive sets, see also Portner 2004Portner , 2007Farkas & Bruce 2009;Murray 2014;Bar-Asher Siegal & Boneh 2016). The claim is that ba 'come' can bring this about because of its particular lexical semantic properties, namely its status as a deictic verb with a doxastic meaning component (see Barlew 2015 and in section 5).
The next section shows that distributionally ba 'come' in the pseudo-coordination construction is a lexical restructuring verb. The subsequent section takes issue with analyses of particular interpretative properties of ba 'come', which set it as the source of the particular inference described above, excluding an analysis of this verb as an aspectual or a modal/mirative semiauxiliary.
3. come as a lexical restructuring verb. Issues of clause boundedness and the architecture of functional projections in the clausal spine are a main theme in the investigation of complex or periphrastic verb constructions. Within generative literature, many insights in this respect have been gleaned from Romance and Germanic languages (Cinque 1999, Wurmbrand 2001et seq., Laca 2004, providing tools to classify the distributional and interpretative properties of the first verbs in these verb sequences (henceforth V1s). While Cinque (1999Cinque ( : 106, 2006 with this extended clausal spine, takes all of them to be functional heads instantiating one of the multiple possible projections, Wurmbrand (2004Wurmbrand ( , 2014, as well as Laca (2004), identify two groups of V1s: lexical and functional. Wurmbrand distinguishes this way between lexical and functional restructuring verbs, whereas Laca, who focuses on various types of aspectual periphrases, delineates two tiers of auxiliaries -inflectional-like vs. derivational-like V1s.  (5)-(6)). However, when aspectual verbs are combined with another V1 that gives rise to recent past or the prospective, order restrictions become apparent, indicating that the two types of V1s belong to different categories.
(10) French (Laca 2004: 7d) a. Il vient de continuer de sortir. He come of continue of go.out 'He has just/recently continued to go out.' b.
*Il continue de venir de sortir. He continues of come of go.out Intended: 'He continues having just/recently gone out.' In turn, the retrospective and the prospective V1s cannot stack, and whichever one appears first, is the V1 of the periphrastic cluster, whereas the second verb can only be interpreted lexically. The examples are taken from Laca (2004: exx. 10a-b), and show that neither of the retrospective and the prospective can outscope the other, namely, they are in complementary distribution. These patterns are predicted if the functional/inflectional-like V1s are derivationally higher in the clausal spine, and the lexical/derivational-like V1s are closer to the VP. Furthermore, the functional/inflectional-like V1 occupies a specific slot in the clausal spine (cf. Cinque 1991Cinque , 2006 and therefore cannot be permuted, contrasting in this way from lexical V1s, which can freely combine, if their lexical selection restrictions are satisfied. There are no architectural constraints imposed in this case, only lexical ones. This picture is strengthened by two additional independent properties that set these two groups of V1s apart. Wurmbrand (2004) shows that the lexical V1 theta-marks its subject and presents lexical selection properties (see also Laca 2004 for the latter point), unlike functional ones. Laca (2004) shows that the inflectional-like V1s, contrary to the lexical/derivation-like ones (e.g. the retrospective and the prospective), cannot freely inflect for tense, allowing in the Romance languages only the present and the imperfective inflection, not the perfective/preterit (Laca 2004: ex. 16), further attesting their inflectional nature.
In what follows, I will apply the tools of stacking and order restrictions to probe the status of ba ve-'come and' as a V1. Therefore, the following sets of data present combination of ba ve-'come and' with three types of V1s: the phasal aspectual continue (12), the prospective/avertive be about to (13), and the modal can (14).
First, then, combining ba ve-'come and' with an aspectual verb is possible in either order. Importantly, as the translations to (12a-b) indicate, the inference of unacceptability/unexpectedness is available, and different orderings indicate which proposition the inference applies tonamely whether it includes himʃix 'continue' (12a) or not (12b).
hu ba ve-himʃix litsxok 'al xeʃbon kulam. he came and-continued to.laugh on expense everyone 'He unacceptably continued to laugh at everyone's expense.' b.
hu himʃix lavo ve-litsxok 'al xeʃbon kulam. he continued to.come and-to.laugh on expense everyone 'He continued to unacceptably laugh at everyone's expense.' Next, the prospective 'amad 'stand/be about to' differs from the aspectual verb exemplified above, it can only select ba ve-'come and' (13a), but not be selected by it. Normally, what is unacceptable is the actual cancelation of leave by the First Sargent, not the prospect of being about to do so. Reversing the order of the verbs, as in (13b)  Finally, under the assumption that modal auxiliaries are projected higher in the clausal spine (Cinque 1999(Cinque , 2006, the pair of examples in (14), similarly to the ones in (13), indicates that ba ve-'come and' can only scope below the modal (14a), mirroring the word order. Again, in (14b) only a (somewhat implausible) bi-eventive interpretation is available.
hu yaxol lavo ve-litsxok 'al xeʃbon kulam he can to.come and-to.laugh on expense everyone i.
'It is possible that he unacceptably laughs at everyone's expense.' ii.
#'It is unacceptable that he can laugh at everyone's expense.' b. # hu ba ve-yaxol litsxok 'al xeʃbon kulam. He comes and can to.laugh on expense everyone Intended: 'He unacceptably can laugh at everyone's expense.' The following representation (15) summarizes the data presented above, where both the prospective amad 'stand/be about to' and the modal yaxol 'can' scope above ba ve-'come and'; but not aspectual phasal verbs, which can stack with it just above the little vP, in XP* (the Kleene star represents recursion and free ordering of similar categories.) (15) Before concluding this section, it is important to note on the basis of the examples above that ba ve-'come and' freely inflects for tense and can appear also as an infinitive, indicating that it patterns like Laca's derivational-like V1s. This constitutes further evidence for ba ve-'s lexical status. I will not discuss ba's ability to assign theta-roles (recall Wurmbrand's criteria for lexical vs. inflectional headedness), since the verb come, in its 'simple' spatial use has been shown to be ambiguous between unaccusative and unergative meanings (see Levin & Rappaport Hovav 1992, Van Geldereen 2018, and as an unaccusative can be claimed to select for a small clause. I will entertain this possibility in the analysis I outline in section 5, but do not fully explore this in the current paper.

Interpretative properties: A new grammatical category?
The previous section suggests that ba ve-'come and' should be grouped with aspectual V1s such as continue, stop, begin, echoing suggestions in the literature as to the aspectual nature of V1 in pseudo-coordination constructions (de Vos 2007, Wiklund 2008, Gamliel & Mra'i 2015. For these authors, the particular aspectual import is responsible for the surprise or suddenness inference. Concretely, Gamliel & Mra'i (2015) proposed that the construction expresses the ingressive (for the Hebrew and Palestinian Arabic construction), and Wiklund (2008) claimed for Swedish that the meaning is that of the inceptive. Both these aspectual categories highlight the initial stage of the event, but only inceptives also entail their ending (cf. Wiklund 2008). Be this as it may, it indeed seems that quite a number of the examples provided so far can be understood in this way, since the start of an underlying eventuality is perceptible. This is evident in examples cooccurring with stative verbs such as (16-17) below, which are understood dynamically in these particular contexts.  (8), the civil law didn't start anything, compared to the criminal law, and what is at stake is the contrast between the two in terms of what each allows. Similarly, in (7), there is no clear sense in which ingression/inception are relevant since the verb neheras 'get ruined' is an achievement predicate.
However, there is no need to resort to an explanation of these data in terms of a semantic weakening or bleaching in the meaning of ba 'come' that renders it more auxiliary-like. Rather, I propose that the source of this reading is due to a combination of the lexical aspect of ba 'come', which is an achievement (Levin & Rappaport Hovav 1992), together with a perfective viewpoint aspect. 4 Importantly, this combination of lexical and viewpoint aspect produces such effects also in regular verbal constructions, and explains why statives are interpreted dynamically (cf. Smith 1991, Homer 2010. Given this explanation in terms of aspectual effect, there is no solid ground for claiming that ba ve-'come and' here differs from its simple use. (2004), Wiklund (2008Wiklund ( , 2009) and Gamliel & Mra'i (2015) draw a link between the aspectual properties of the construction and the special inference of suddenness and surprise stemming from it. Ruling out that the verb is an aspectual one raises the possibility that it is/was on a grammaticalization path to becoming a grammatical category related to the expression of mirativity, bringing about the unexpectedness/unacceptability inference.

Now, de Vos
A clear indication that this is not the case comes from embedding pseudo-coordination under affirmative or negative surprise predicates, as well as other predicates expressing stance/speaker attitude. In these environments, the contrast between the prejacent and the background set of norms is maintained and is orthogonal to the speaker's stance, and to whether the prejacent is perceived as a surprise. Thus, whereas adding a surprise interjection in (18) seems to strengthen or confirm the effect produced by the pseudo-coordination construction, the same inference is available under the negated predicate mafit'a 'surprise' (19) Similarly, embedding under the negated verb metsape 'expect' in example (17) above shows that one can isolate the meaning effect arising in the pseudo-coordination construction related to the relevance of social norms from speaker stance. Here, what the speaker conveys is that, given his behavior/disposition, he does not expect that someone would hate him, since normally (or normatively) people do not hate others for no apparent reason.
To summarize, dismissing claims that ba 'come' is recruited to express aspect or mirativity helps to argue against its transformation into a grammatical category (see Heine, Claudi & Hünnemeyer 1991), and to further establish the claim that grammaticalization was not involved in bringing about this construction, with the inferences it produces. In the last section of the paper, I turn to consider the relevant meaning components that ba 'come' and ba ve-'come and' share, and argue that they are indistinct, including the availability/absence of a motion meaning component. They differ only in the type of complement they allow -prepositional or propositional.

Accounting for the unacceptability/unexpectedness inference.
Here I briefly consider crucial meaning components of spatial ba 'come', which I take to be relevant for the unacceptability/unexpectedness inference arising in pseudo-coordination constructions (see , Abarbanel 2019 for more detailed discussion). I stress the importance of deixis, which is an anchoring not-at-issue implicature indicating the linguistic presence of a contextually given individual (Fillmore 1997, Oshima 2006, Barlew 2015, Sudo 2015, and the availability of a related doxastic modal meaning component, whereby the beliefs of the contextually given individual are relevant (Barlew 2015).
Applied to English spatial, change-of-location come, this means that the contextually given individual -the ANCHOR -is located at the destination denoted by the PP at either the utterance time or at a reference time (Fillmore 1997, Oshima 2006, Barlew 2015, Sudo 2015. The following example adapted from Barlew (2015: 12) indicates the presuppositional nature of the ANCHOR, since it can be accommodated: (20) pagaʃ-ti miʃehu ba-internet etmol. met-1SG someone in.the-internet yesterday. hu amar li ʃe-tramp ba le-pariz they told to.me that-Trump comes to-Paris 'I met someone on the internet yesterday. They told me that Trump is coming to Paris.' ACCOMMODATED INFORMATION: the person the speaker met on the internet yesterday is from Paris (or Paris is their Home Location).
Additionally, as pointed out by Barlew (2015), the anchor is a doxastic agent whose beliefs are relevant in the discourse (cf. Ross 2016); consider the following example, where the location of the ANCHOR is self-conceived, and not an actual one: Given this brief background, I propose the following parallelism between spatial come and pseudo-coordination ba ve-'come and', where the crucial meaning difference is attributed to the type of complement, not the lexical-pragmatic properties of the verbal head itself.
(22) COME + PP, ba 'come': The anchor believes she is located at the goal, and the individual denoted by the subject DP ends up at that location, presupposing that they weren't previously at that location.
(23) COME + vP, ba ve-'come and': The propositional content of the selected vP is added to a set of propositions constituting the beliefs of the anchor regarding norms; previously not being part of this contextual set. The contrast between the proffered content and the belief context of the anchor gives rise to the unacceptability inference, and speaker stance is exclusively due to pragmatics.
One crucial indication for the viability of this analysis comes from ignorance contexts. ba 'come' in either environment -spatial or pseudo-coordination -cannot appear in these contexts. Compare: Note that with halax 'go', no such restriction is observed, for both spatial (Sudo 2015) and for pseudo-coordination uses. Due to lack of space, these examples are not produced here.
As for the motion component, there are some reasons to think that come does not directly encode motion, even in its spatial, change of location use. Namely, it does not contain a path meaning component. It has been noted that ba 'come' behaves overall like an achievement verb (Trommer 1983, Tzuberi 2018, Levin & Rappaport Hovav 1992, see also van Gelderen 2018); that is, it encodes a change-of-state: stating that x was not initially at y but comes to be at y. Therefore, if the lexical entry encodes a path or a motion scale, it should be a simplex path/scale and not a complex one, comprising the final endpoint (cf. Beavers 2011). The following examples illustrate this. First, the final destination PP can be overtly expressed, but even if it is not, and only a source PP is there, still one understands what the final destination is, according to the coordinates of the salient discourse participant (e.g. the speaker). Complex paths are encoded by means of a PP with the relevant semantic properties. Example (27b) clearly indicates a prolonged path with the PP ba-derex ha-'aruka 'the long way', contrary to examples (27a), and in the same way that a manner component is externally added, as in (27c) Thus, the motion component is not due to the core lexical semantics of the verb, but may be contributed by prepositions, and world knowledge stemming from the types of themes available.
Additionally, unlike higi'a 'arrive', in the case of ba 'come', the theme argument has to be 'intrinsically' capable of motion. Thus, other than animates, in the following examples, a vehicle can ba (28a), time periods, or seasons can ba (28b) but not inanimate objects incapable of independent motion (28c).
(28) Modern Hebrew (Tzuberi 2018: p. 8-9) a. ha-otobus ba ba-zman. Lastly, to complete the picture laid out here, complex verb constructions resembling pseudo-coordination have been available for centuries in Semitic languages (cf. Kraus 1987 for Akkadian): they are attested also in Biblical Hebrew (from 10 th century BCE till 200 CE approximately), and Mishnaic Hebrew (200CE). 5 But pseudo-coordination with the unacceptability/unexpectedness inference are witnessed in the 20 th century in Modern Hebrew (Kuzar 2006, Boneh 2019.
I conjecture that this is so because ba 'come' was not necessarily deictic in previous centuries and was used exclusively as a change of location verb meaning enter, arrive or go (qua walk) in previous stages of the language (Polak 2009, Koller 2013, Tzuberi 2018, Rubinstein 2019, Boneh 2019, and in pseudo-coordination constructions, which were probably mono-eventive did not give rise to the meaning available in Modern Hebrew.
The centrality of the deictic-modal meaning component, and its absence in earlier stages of the language, points to meaning gain, not loss, and further substantiate the claim that pseudo-coordination constructions with their special inferences are not emerging constructions, but arise due to a lexical change in V1, in an already available syntactic frame.
To conclude, assuming a stable lexical meaning for ba 'come' constitutes a competing analysis to suggestions of a cross-lingusitically attested grammaticalization process from lexical motion verb to a semi-auxiliary, often suggested in the literature. This paper therefore nuances the array and analysis of complex verb constructions by suggesting that their diversity can be accounted for by a "mere" lexical semantic change in the verb come (possibly also go), not always involving changes pertaining to a full-fledged grammaticalization process. Moreover, an analysis such as the one proposed here helps elucidate the fine machinery of what has elsewhere been labelled as metaphoric meaning (cf. Bourdin 2014, Ross 2016. In other words, the current account provides substance to claims for a metaphorical relation between the two occurrences of ba 'come' (and possibly also halax 'go') since it reveals the similarity in the lexical-pragmatic properties of this lexeme in the simple and complex verb construction environments: the difference between a concrete location and an abstract concept being located in the choice of complement.