The Biblical Hebrew infinitive

Thepaperproposesthatthesamefunctionalcategorieswhichdeterminetheinflection of the Biblical Hebrew finite verb also determine the feature specification of the Biblical Hebrew infinitive. This proposal depends both on demonstrating that the infinitive is a verb, rather than a noun (or a verbal noun), as traditionally assumed, and on showing that the functional categories that embed the infinitive are clausal rather than nominal. The article starts by examining the traditional distinction between the Infinitive Absolute and Infinitive Construct , and makes an argument for a single infinitive, with two allomorphs. The former is a verb marked as [+Mood], while the latter is marked as [–Mood], and both are also specified for two other clausal functional categories: T and Asp/Mod. These two latter categories determine a 4-way classification of finite/infinitival verbs: [+T+Asp/Mod], [+T–Asp/Mod], [–T+Asp/Mod], [–T– Asp/Mod]. This classification determines a concomitant 4-way alternation of attachment options of subject and/or object clitics to the verb: [+subj.cl.+Obj.cl.], [+subj.cl.– Obj.cl.], [–subj.cl.+Obj.cl.], [–subj.cl.–Obj.cl.], and moreover accounts for the distribution of the different verb forms.


Introduction
Biblical Hebrew (BH) verbal forms manifest rich inflection within the finite clause, encoding the functional categories of temporality (T), mood (Mood), grammatical aspect (Asp), and modality (Mod). These categories have been widely discussed in the literature, and their relative role is still under debate (recently Hatav 1997, Joosten 2002, Cook 2006, 2012. In particular, Asp and Mod have proven hard to disentangle in BH, and the present work will reflect this by assuming a composite Asp/Mod category. But what has not so far been proposed is that the same categories regulate the use of the infinitive as well. It is the aim of the present paper to demonstrate the function of these categories within the infinitive clause. I argue that BH has a single infinitive category, which is specified for different combinations of T, Mood, and Asp/Mod, giving rise, in addition to the finite (Fin) construction, to various infinitival constructions: Nom-inf, Poss-inf, and PRO-inf. The examples in (1) illustrate, using the same verb remember, the Fin and infinitival constructions in their typical functions. The Fin construction is a clause in the indicative mood, or in a variety of irrealis moods (imperative / jussive / cohortative), and Nom-inf is an irrealis root clause. Poss-inf and PRO-inf are embedded clauses with a variety of functions, and their distribution will be discussed in detail below.1 Poss-inf often functions as a temporal adverbial, and PRO-inf-as a purpose adverbial:2 ( lə-ʔaḇrāhām to-Abraham lə-yiṣħāq to-Isaac u-lə-yaʕăqōḇ and-to-Jacob Remember Your servants, Abraham, Isaac, and Jacob. (Deut 9:27) 1 The distinction between these two constructions, which serves the base of the distinction between the Modern Hebrew Gerund and Infinitive, was presented in Doron (2016). This distinction had not been made before in Biblical studies. 2 Unless stated otherwise, all Biblical translations are from the New King James Version (NKJV).
The fricative allophones of b, g, d, k, p, t are transcribed as the stop allophone with a diacritic (ḇ, ḡ, ḏ, ḵ, p, ṯ The rainbow shall be in the cloud, and I will look on it to remember the everlasting covenant between God and every living creature. (Gen. 9:16) The article is constructed as follows. Section 2 introduces the two allomorphs of the BH infinitive. Section 3 introduces the different infinitival clauses constructed from the infinitive, and their different morphosyntax and distribution. Crucially, the functional categories specifying the infinitive are clausal rather than nominal: an overt or implicit subject Determiner Phrase (DP) is case; cs-Construct State (morphological marking of the possessee noun); f-Feminine; ill-Illative case; impr-Imperative; inf-Infinitive Construct; infabs-Infinitive Absolute; juss-Jussive; m-Masculine; mod-Modal; neg-Negation; p-Plural; poss-Possessive case; ptc-Participle; q-Question particle; s-Singular. doron Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 119-150 present. Section 4 further argues for the clausal rather than nominal nature of these constructions, by showing that the infinitive is a verb rather than a noun or a verbal noun-which was the category assigned to it by the traditional literature of the Bible. Section 5 briefly mentions the rise of nominalizations which eventually replaced the Poss-inf structure in post-Biblical Hebrew. Section 6 concludes.

One infinitive category, two allomorphs
The theoretical issues raised by the BH infinitive have not yet been tackled. Existing literature assumes two separate infinitives, traditionally called the Infinitive Absolute and the Infinitive Construct. The two are distinguished by their form and distribution, and are described as two separate categories, treated separately in separate chapters of the traditional grammar books of the Bible.3 Historically, the Infinitive Absolute is the original infinitive, also found in Akkadian (Blau 1979: §30), while the Infinitive Construct has been claimed to originate in a different Proto-Semitic form, related to the imperfective (Bauer and Leander 1922: §43). Yet, synchronically, I would like to propose that the two are actually syntactically conditioned allomorphs of a single infinitive category of BH.4 Derivationally, the Infinitive Absolute is the basic allomorph, as it is found in all the seven Hebrew verbal templates, whereas the Infinitive Construct is only derived in the active and middle templates, but not in the passive.5 The derivations are shown in the following table:6 Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 119-150 (2) The basic allomorph, the Infinitive Absolute, serves as the citation form of the verb, and in adverbial uses (typically bare of arguments). The adverbial infinitive either directly modifies the inflected verb (3a-b), as described in Callaham 2014, Hatav 2017, and references therein, or it modifies the VP (3c-d).
(3) a. For the purposes of the present article, I will mostly ignore the adverbial use (3), where the infinitive is "bare" of any functional category, and hence is not clausal and does not introduce a subject.7 I will only be interested in the uses of the infinitive which involve clausal constructions with functional categories, and hence a subject. The point of the present article is to show that there are two types of such constructions, one classified together with finite clauses as having conversational force, and the other-as lacking such force. A second dimension classifies clauses as having temporal anchoring, and others as lacking it. We will see the implications of this classification in the next section: Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 119-150

Two types of infinitival clauses
The present section discusses the two different uses of the infinitive which involve clausal constructions, i.e. the verb together with all its arguments and functional categories.

[+Mood] infinitival clauses
The first construction is a clause with imperative force (including notional jussive and cohortative). The allomorph of the infinitive in this construction is the Infinitive Absolute.8 According to the analysis proposed here, this is due to the fact that the only functional category specified in this construction is Mood, with an [-finite] value interpreted as imperative force.9 Since the T and Asp/Mod categories in the clause are unspecified, there is no inflection to alter the citation form of the infinitive, nor, as we discuss below, to provide an attachment site for subject and object clitics. And as there is no temporal anchoring of the verb to the speech act, these sentences tend to be generic in interpretation unlike the discourse-bound interpretation of the finite imperative (1aii).10 I will call this construction Nom-inf, since it includes a nominative subject, either a null pro (an addressee-oriented logophoric pronoun according to Portner's 2004 analysis of the imperative), as in (4a), and also (1b) above, or a lexical DP, as in (4b-f), an option found cross-linguistically in imperative clauses (Mauck et al. 2005). As is to be expected of imperative clauses, they are typically root clauses (Palmer 2001). This might be true even when introduced by verba dicendi as in (4d), since practically all reported speech in the Bible is direct quotation, constructed as a conjunction of root clauses, which is possible for imperatives as well (Maier 2010 (12)) 8 The same is true in Arabic, where the qatāli form which corresponds to the Infinitive Absolute also serves as an imperative (Wright 1874: Vol. 1, p. 62). 9 As is known from the literature (Portner 1997 and references therein), Mood is the category which determines the conversational force of a root clause (Indicative, Imperative, etc.) 10 This contrast is reminiscent of the contrast between the generic lō+Modal negation and the eventive ʔal+Jussive negation among Fin clauses.

[-Mood] infinitival clauses
The infinitive allomorph in the second type of construction is the Infinitive Construct. This allomorph allows the attachment of pronominal clitics, something that is strictly disallowed in the Nom-inf construction, built with the Infinitive Absolute allomorph. As we will see below, this difference is due to the fact that subject and object clitics attach to the relevant functional categories-T and Asp/Mod respectively-which are present in the second type of construction but not in the Nom-inf construction.
The first subtype, familiar from other languages, has a null pronominal anaphor subject (PRO), typically controlled by another DP in the linguistic context. This is the PRO-inf construction. We will see below that it does not have temporal specification, i.e. its T is unspecified, yet it does have Asp/Mod specification. As it is not specified for T, or Mood, the subject is not assigned case, and is hence PRO. As it is specified for Asp/Mod, but not T, it allows object but not subject clitics.
The second subtype, Poss-inf, has an overt subject with possessive case.11 I will argue that this construction is temporal and includes specification of the functional category T, though not of the category Asp/Mod. As it has T specification, but not Asp/Mod, it allows subject but not object clitics. It is distinguished from finite clauses, which have both T and Asp/Mod specification (and hence both subject and object clitics). I assume that it is non-finite T which assigns possessive case to the subject, in parallel to the non-finite -ing functional category which assigns accusative case to the subject of Acc-ing gerunds in English according to Reuland's 1983 analysis.12 In the following examples of PRO-inf and Poss-inf, notice the Infinitive Construct allomorphs rəʔōṯ 'see' and šūḇ 'return' in (6) and (7), which differ from the corresponding Infinitive Absolute allomorphs rāʔō and šōḇ of the same verbs in (3) above. 11 The possessive case is a marked case of the subject in other languages as well, such as Alaskan Yup'ik (Abney 1987:28), Finnish (Kiparsky 2001), Ladakhi, Lak, Niue (Lander 2011: 590), Tagalog (Aldridge 2006, Collins 2017, Tzutujil Maya (Abney 1987:31), and others. 12 It has often been noticed that the BH Infinitive Construct subsumes properties of both infinitives and gerunds in other languages. PRO-inf subsumes both the English infinitive and the PRO-ing gerund. Poss-inf parallels the English Acc-ing gerund, despite the morphological difference between accusative and genitive. Poss-inf does not parallel the English Poss-ing, which is a nominal rather than a clausal construction (Pires 2001(Pires , 2006(Pires , 2007Moulton 2004 The possessive case of the Poss-inf subject is overtly marked for pronominal subjects, in particular the 1st person singular and the 3rd person masculine singular, where the possessive marking differs from accusative marking of the corresponding object clitics in the PRO-inf construction. Thus, the 1st person object clitic -ēnī in (8a) differs in form from the 1st person subject clitic -ī in (8b), and the 3rd person object clitic -ēhū in (9a) differs in form from the 3rd person subject clitic -ō in (9b) The two constructions contrast sharply in distribution. All the (b) examples in (6)-(9) above are temporal adverbials, and none of the (a) examples are. This is not an accident, as it is the case in general that temporal prepositions only take Poss-inf complements, never PRO-inf complements. I attribute this to fact that Poss-inf clauses include T specification in their structure, whereas PRO-inf clauses do no. Thus only the former can serve as the Specifier of the main clause T head (Cinque 1999). PRO-inf clauses function as purpose clauses, as in (6a)  Purpose clauses are part of infinitival clauses which "are a group which displays a characteristic future-oriented, irrealis semantics" (Portner 1997: 183). Yet, as argued by Wurmbrand (2001Wurmbrand ( , 2014, the seeming temporal relation of the infinitival clause to the main clause is not due to T but to Mod, which determines the inherent future orientation of purposes. 14 Purpose clauses are distinct from rationale clauses (Jones 1985, Verstraete 2008, which can be expressed by the Poss-inf construction. The latter describes a result event, as in (i) below, not necessarily the outcome and agent's intentions, unlike the intentional/modal characterization of purpose clauses: V pro1 Aspectual and modal verbs in the (11b) structure are control verbs expressing root modality (Ability, Deontic). When the same verbs modify the aspectual and modal dimension of a state/event which is not determined by the actions or abilities of an agent, their modality is interpreted as alethic (metaphysical), they do not have an agent, and function as raising verbs (Hacquard 2011). The following examples describe the beginning (a), repetition (b), possibility (c) of an event/state, independently of an agent. The non-agentive subject of the complement raises from the infinitival clause, leaving a trace (t) which is an empty category different from PRO, yet like PRO is restricted to case-less positions: (12) a.  (12) is in demonstrating the clausal nature of infinitives with null subjects.

3.3
The morpho-syntax of the various verb forms Unlike the Nom-inf construction, which is restricted to root clauses, the PROinf and Poss-inf constructions are embedded clauses, i.e. never have conversational force of their own. This fact has been encoded by their including no specification of the functional category Mood. Yet they can acquire conversational force and function as main clauses through a fronted wh-element such as ma 'what' in (14) Another morphosyntactic distinction has been attributed above to the categories T and Asp/Mod-i.e. the variation between the different infinitive constructions, and the contrast between them and the Fin construction, in allowing the cliticization of subject and object pronouns as part of the morphology of the verb. The following table summarizes the allowed cliticization options in the different constructions, and examples are given in (16)-(17) below: (15 )   table 3 Distribution of clitics + Subj cl. − Subj cl.

+ Obj cl. Fin PRO-inf − Obj cl. Poss-inf Nom-inf
The verb in the PRO-inf construction can have object clitics (16b, 17b) but clearly not subject clitics, since, for case reasons, it does not have an overt subject of any kind. On the other hand, the verb in the Poss-inf construction can have an overt object. But crucially-not in the form of a clitic (16c, 17c). This is surprising, since both subject and object clitics appear with a Fin verb (16a, 17a): The ungrammaticality in the (c) examples above is not due to "heaviness" of two combined clitics, since even if the subject is not a pronominal clitic but a full lexical item, an object clitic is impossible in the Poss-inf construction:  The ban against an object clitic in the Poss-inf construction is thus not morphophonological but morphosyntactic. I have assumed that it is the functional category Asp/Mod which licenses the object clitic, both in the Fin clause and in the PRO-inf clause. The Asp/Mod category is unspecified in the Poss-inf construction, hence the lack of object clitics. As has often been remarked in the literature, object clitics attach to inflection which is characteristically verbal (e.g. in Romance, Cardinaletti and Shlonsky 2004).17 I summarize in (21) the morpho-syntactic characteristics of the different finite and infinitival clauses, where the relevant functional categories are ordered by the hierarchy in (20): 17 Indeed the participle, which is inflected as a noun, mostly takes genitive marked object clitics: Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020)

The clausal nature of the infinitive construction
The Hebrew grammatical tradition, ever since the Rabbinic and Karaite grammars of the middle ages, views the infinitive absolute as verbal, and the infinitive construct as nominal (Eldar 2014, Gaash 2018. The European grammatical tradition views both infinitives as mixed nominal/verbal categories (Gesenius 1910, Joüon 1923, Waltke and O'Connor 1990, Arnold and Choi 2003. But, as already argued in Doron (2018), a careful analysis shows that the infinitive is verbal, and has no nominal properties whatsoever. This claim can actually be substantiated on two counts. First, the lexical category of the infinitive is V rather than N, not even a deverbal N. Second, the functional categories specifying V are clausal rather than nominal-similarly to what has been shown by (i) mōṣʔ-ī find.ptc.ms-poss.1s (ii) məpālləṭ-ī deliver.ptc.ms-poss.1s (iii) məśanʔ-ī hate.ptc.ms-poss.1s (iv) šōlħ-ī send.ptc.ms-poss.1s 'anyone who finds me ' (Gen. 4:14); 'He delivers me ' (Ps. 18:49); 'he who hates me' (Job 31:29); 'He who sent me ' (2Sam. 24:13) Yet the participle exhibits noun/verb duality, and there are also a few cases where it heads a finite clause with accusative object clitics: ‫ָה‬ ‫ֵא‬ ‫ל‬  Pires (2001Pires ( ), (2006Pires ( ), (2007 for the English PRO-ing and Acc-ing gerunds, i.e. that they are clausal rather than nominal.18 First, the infinitive assigns accusative case to its direct object, as could be seen in all the examples above, where the infinitive had a direct object.19 Moreover, object clitics attached to the infinitive are always accusative rather than genitive. In the case of nominal forms, such as the participle, one mostly finds genitive object clitics (fn. 18).
Second, the infinitive has no nominal morphological inflection of gender, number, or definiteness.20 The infinitive is case marked in a few examples by the accusative ʔɛṯ, as in (22a), but so are Fin CPs as in (22b)  bam-midbār in.the-desert Remember! Do not forget how you provoked the Lord your God to wrath in the wilderness (Deut. 9:7) 18 In English, Poss-ing gerunds are nominal. 19 Modern Hebrew allows nominalized verbs to assign accusative case as well, which is a marked option crosslinguistically. This phenomenon originates in Medieval Hebrew under Arabic influence (Blau 1990, Goshen-Gottshtein 1951 Third, the infinitive is not modified by adjectives but by adverbs, such as the adverbs hēṭēḇ 'well' , ʕōḏ 'more' , and mahēr 'at once' in (23) Fourth, despite the genitive case marking of its subject, the infinitive in the Poss-inf construction is not a noun. It does not head a construct state phrase. Unlike the nominal construct where the construct state (CS) noun must be absolutely adjacent to its complement, the same is not true of the infinitive in the Poss-inf construction. No adjacency required. The subject of the infinitive is separated from the verb in many examples, something which never happens in a construct. This is exemplified in the following examples where the subject is separated from the infinitive verb hakkoṯ in (24a) by the accusative pronoun ʔōṯō, similarly in the other examples in (24) Finally, negation is found with infinitival clauses, and it can be shown to take scope over the entire clause rather than just modifying the infinitival head. Only clausal scope can give the correct reading in (30). Sacrificing to the Lord is the purpose of sending off the people, not the purpose of not-sending off the people. Therefore, negation attaches to the full clause letting the people go to sacrifice to the Lord rather than to the head letting go. In Modern Hebrew, the negative biltī has grammaticalized into a prefix which attaches to lexical items, in particular adjectives. Late Biblical Hebrew also saw a blurring of the distinction between the two allomorphs of the infinitive (Absolute and Construct), as reported by Fassberg (2007), Morrison (2013), Mor (2015, to appear), and others. Rabbinic Hebrew has a single infinitive allomorph, different from both Biblical allomorphs, and its distribution is that of the Biblical PRO-inf.
It seems reasonable to assume that it was the collapse of the Mood specification in the verbal system of post-Biblical Hebrew, which launched the demise of the Nom-inf construction. As to the replacement of Poss-inf by event nominalization, this is not a trivial development. I conjecture it could have been due to the introduction into Hebrew of the category D, which could have replaced the category T as the head of propositional event complement/adjuncts, favouring combination with event nouns rather than infinitive verbs. This will hopefully be the topic of further research.

Conclusion
The article has shown how the morphosyntax of the different infinitival clauses determines their distribution. Nom-inf clauses are root clauses with irrealis Mood, hence have the conversational force of imperatives. PRO-inf and Possinf clauses are not specified for Mood, and thus have no conversational force. They therefore must be embedded clauses. The lack of T specification, in addition to the lack of Mood specification, determines that the PRO-inf clause cannot be interpreted as an independent proposition, but is rather interpreted as part of the event denoted by the main clause, since it depends for its temporal anchoring on the temporal specification of the main clause. The Asp/Mod specification of the PRO-inf construction allows it to function as complement of aspectual and modal verbs, and as adjunct to Mod/Asp heads, i.e. as purpose clauses. The Poss-inf clause, on the other hand, contains a specification of T, and hence denotes a separate proposition from the one denoted by the main clause. Accordingly, it functions as a complement of propositional attitude verbs or a temporal / rationale / result adjunct. The specification of T and/or Asp/Mod in an infinitival clause has also been shown to explain the various possibilities of subject and object cliticization in the various clauses.
Brill's Journal of Afroasiatic Languages and Linguistics 12 (2020) 119-150 In sum, the paper has shown that the same functional categories which determine the inflection of the Biblical Hebrew finite verb also determine the feature specification of the Biblical Hebrew infinitive.