Arabic PPs in a Rooted Lexicon

Abstract

We motivate a ‘rooted’ PP shell analysis of Arabic prepositional phrases, which takes into account the prepositional dual life, as a lexical root item and as a vocabulary word, projecting a lexical √P headed by the P root, and a functional pP headed by p, the syntactic case assigner. Moreover, PlaceP and PathP projections are motivated by differentiating locative and directional PPs, and AxPartPs represent the structure of adverbial spatial nouns (đ̣uruuf). It is shown that alternative analyses using a single source (or projection) of PPs are inadequate in dealing with prepositional polysemies, and their morpho-syntactic alternations or variations. A bifurcation analysis instead (distinguishing root syntax from category syntax) is motivated and implemented along the lines of distributive models of word formation, making use of the simplest composition operation Merge in both syntax and the lexicon.

1. Introduction

The category of preposition is often poorly defined, both traditionally and within generative grammar. In Arabic traditional grammar, prepositions form a closed set that includes fii ‘in’, bi- ‘with’, min ‘from’, Ɂilaa ‘to’, l- ‘for’, ʕalaa ‘on, ʕan ‘about’, and ka- ‘similar’. Adverbial expressions included in PPs semantically convey similar relational meanings, although they are morpho-syntactically nominal, rather than prepositional. Their list is quite open; it includes fawqa ‘above’, taḥta ‘below’, ʔamaama ‘in front’, xalfa ‘back’, ḥawla ‘around’, ṣawba ‘towards’, etc. They are traditionally thought of as đ̣uruuf ‘adverbials’, rather than ḥuruuf ‘prepositions’. In fact, they are types of ‘weak nouns’, analyzed by Svenonius (2006, 2010), after Landau and Jackendoff (1993), as axial parts (AxPartP; see also Herskovits 1986). Because these prepositions and adverbials have a dual functional and lexical nature (or “mixed signs”, as described by Weinreich 1963, p. 150), their categorical status has been a matter of debate. They are tentatively analyzed as (a) lexical, like other lexical categories (nouns, verbs, and adjectives; Chomsky 1970, 1995; Jackendoff 1977); (b) functional, like other functional categories (determiners, auxiliaries, modals, and complementizers; Baker 2003; Grimshaw 1991); (c) both lexical and functional (Svenonius 2010; Wood and Marantz 2017); or (d) semi-functional (Ryding 2005; Saeed 2014; Al-Humari 2015).

This article argues for a specific treatment of prepositions and PPs. Prepositions are cartographically projected as in Svenonius’ (2010, 2012) work, but more specifically generated as distributed words, in ‘rooted’ PPs, as in Wood and Marantz’s (2017) DM analysis of English; see also Deacon (2011); for Arabic, see Fassi Fehri and Aamiri (2021). More like verbs in the vP complex, the preposition in a pP projection has a dual life; as a lexical root heading the prepositional √P, and as a category head p of pP. ‘Small’ p assigns genitive case, in parallel to v, which assigns accusative. Moreover, the analysis adopts a sort of bifurcation hypothesis of root/category syntax and semantics, as represented in (1a), and mirrors a form of PP shell structure (à la Larson 1988, 2014), as in (1b):

(1)

Both structures represent the dual facets of prepositions: as ‘lexical’ or √, and as ‘functional’ or categorized.

Cross-linguistically, spatial prepositions are either locative or directional Ps. Based on its semantics and its syntax, a particular P is merged as either Place or Path. Locational Ps are Place heads, where Place is P_loc, and directional Ps are Path heads, where Path is P_dir, as in (2a) and (2b) (see Jackendoff 1983; Wunderlich 1991; Zwarts 1997, 2005; Zwarts and Winter 2000; Kracht 2002; Koopman 2000; Svenonius 2010; den Dikken 2010; Gehrke 2008, among others):

(2)

Furthermore, axial adverbials that are selected by Place are identified as AxPart heads, as in (3); Svenonius (2006):

(3)

By investigating the various morpho-syntactic and semantic facets of spatio-temporal expressions headed by prepositions or AxPart weak nouns in Arabic, the article aims at establishing that PP has a dual projection (a favored hypothesis over single projection analyses), and that P is ‘rooted’ in the lexicon, i.e., having a lexical root (like nouns, verbs, or adjectives), rather than just being only functional.1 The article aims also at exploring the practical use of ‘roots’ as the irreducible syntactic units in prepositional word formation, using the simplest combinatory Merge operation.2 Section 2 discusses the structures and meanings of PPs in the space domain. Section 3 provides evidence for the prepositional √P (root phrase) as distinct from pP (functional phrase). Section 4 investigates the meaning specifications of prepositional roots that motivate the existence of √RootP as a separate projection. It focusses on complex polysemies, variation, and syntactic alternations. Section 5 concludes.

2. Structures and Meanings of PPs

2.1. The Basic Distinction of Locatives: Places and Paths

Cross-linguistically, spatial PPs have been shown to distinguish structurally and semantically between two locatives: Places and Paths, projecting as PlaceP and PathP. The locative/directional distinction is often associated with this dichotomy (see Jackendoff 1983; Wunderlich 1991; Zwarts 1997; Zwarts and Winter 2000; Kracht 2002; Koopman 2000; Svenonius 2010; den Dikken 2010; Gehrke 2008; Saeed 2014, among many others).

We show that locatives in Arabic can express three meaning subtypes: (a) locational, (b) directional, and (c) motional.

2.1.1. Locational versus Directional PPs

The locational Places occur in verbless (copulative) sentences, inducing a locative meaning, as illustrated in (4) and represented in (5).

(4)	a.	Ɂanaa	fii	l-bayt-i
		I	in	the-house-gen
	b.	Ɂanaa	bi-l-bet	(Levantine Arabic)
		I	in-the-house
		‘I am in the house.’

(5)

The prepositions fii ‘in’ and bi- ‘in’, are Place heads (Fassi Fehri 2021; Saeed 2014). They locate a Figure (Ɂanaa ‘I’) in (4a) and (4b) in relation to a Ground (l-bayt-i ‘the house’), where there is no verb that can locate the Figure with respect to the Ground. The fact that the sentences in (4) are verbless suggests that the locational meaning has its source in the PP, and not in any verbal predicate (compare with the collocation in (6) below). Accordingly, the prepositions fii ‘in’ and bi- ‘in’ must be bearing a semantically interpretable feature [loc(ational)].

On the other hand, directional PPs or Paths easily collocate with motion verbs inducing directional meanings, as illustrated in (6) and represented in (7):

(6)	a.	Ɂ-aḏhabu	ʔilaa	l-bayt-i
		I-go	to	the-house-gen
		‘I go to the house.’
	b.	mašay-tu	ḥattaa	l-ġurfat-i
		walk-I	till	the-room-gen
		‘I walked till the room’

(7)

The prepositions ʔilaa ‘to’ and ḥattaa ‘till’ are Path heads that are endowed with a [dir(ectional)] feature. Unlike locational PPs, directional PPs are not found in verbless locative contexts. They are ill-formed in such contexts, as in (8a) and (8b):

(8)	a.	*Ɂanaa	ḥattaa	l-ġurfat-i	*directional
		I	till	the-room-gen
		Intended meaning: ‘I am (heading) up to the room.’
	b.	*Ɂanaa	ʔilaa	l-bayt-i	*directional
		I	to	the-house-gen
		Intended meaning: ‘I am (heading) up to the house.’

In such contexts, they allow only a locative meaning:

(9)	l-qalam-u	ʔilaa	jaanib-i	l-kitaab-i	locative
	the-pen-nom	to	next-gen	the-book-gen
	‘The pen is next to the book.’

With stative verbs, the directional reading is not licensed. Clearly, directional PPs are not used with stative verbs like waqafa ‘stood’, as the ungrammaticality of (10a) suggests, while locative PPs collocate easily with the stative verb, as in (10c). The PP in (10b) has only a locative interpretation (due to its unambiguous status):

(10)	a.	*waqaf-a	ḥattaa	l-baab-i		*directional
		stood-3	till	the-door-gen
		Intended to mean: ‘He stood up (heading) to the door.’
	b.	waqaf-a	ʔilaa	jaanib-i	l-baab-i	locative
		stood-3	to	next-gen	the-door-gen
		‘He stood up next to the door.’
	c.	waqaf-a	fii	l-baab-i		locative
		stood-3	in	the-door-gen
		‘He stood up at the door.’

It is generally agreed that the structure of PPs involves two distinct head projections: Path and Place. Path is outside Place, and when both co-occur, Path is normally placed higher than Place (Jackendoff 1983; Koopman 2000; van Riemsdijk and Huybregts 2002; den Dikken 2010, etc.).3 In (9), ʔilaa is heading a complex Place P, the structure of which is (11a), and in (7) it heads a complex PathP, represented as (11b):

(11)

2.1.2. AxPartPs

Both Path P and Place P can be more complex when they include adverbial spatial nouns đ̣uruuf or AxParts, which are not true ḥuruuf prepositions. Among these adverbial spatial nouns, we can list the following: jaaniba ‘beside’, xaarija ‘outside’, daaxila ‘inside’, ʔamaama ‘in front’, xalfa ‘back’, taḥta ‘below’, fawqa ‘above’, quddama ‘in front’, bayna ‘between’, ṣawba ‘towards’, wasaṭa ‘middle of’, ʕabra ‘through’, etc. Their English counterparts have been analyzed as Axparts, heading AxPartPs (Jackendoff 1983; Herskovits 1986; Svenonius 2010). We alternatively treat them as simply (spatial) nPs of some specific ‘weak’ sort (Matushansky and Zwarts 2019).4 These spatial nPs can be semantically parallel to prepositions conveying locational, directional, or motional meanings, rather than being referential like ordinary nouns. However, unlike prepositions, they categorically function as nouns. They receive case, rather than assign it, and they can be pluralized and modified by an adjective (see Fassi Fehri and Aamiri 2021 for more detail about weak spatial nouns). They can be included lower than Ps in Places or Paths. In (9) above, ʔilaa is heading a complex Place P, which embeds nP, the structure of which is (11a). In (12a) below, a Path P embeds nP as represented in (12b):

(12)	a.	y-aḥuum-u	ṭ-ṭaaɁir-u	min	ḥawl-i	š-šajarat-i
		3-roam-ind	the-bird-nom	from	around-gen	the-tree-gen
		‘The bird is roaming around the tree.’

As a noun, ḥawl-i ‘around’ is assigned a genitive case by the preposition min ‘from’.

As explained earlier, the prepositions Ɂilaa ‘to’ and min ‘from’ are (ambiguously) heading a PlaceP or PathP. Prepositions can also remain optionally null (or silent), as in (13a) and (13b), but their interpretation is recovered:

(13)	a.	l-qalam-u	jaanib-a	l-kitaab-i
		the-pen-nom	next-acc	the-book-gen
		‘The pen is next to the book.’
	b.	y-aḥuum-u	ṭ-ṭaaɁir-u   ḥawl-a	š-šajarat-i
		3-roam-ind	the-bird-nom  around-acc	the-tree-gen
		‘The bird is roaming around the tree.’

Because the adverbial constructions in (13a) and (13b) are the null-P counterparts of (9) and (12a), respectively, they are argued to be headed by a null Place and Path:5

(14)

As a result of the null P, the spatial n receives a sort of oblique ‘accusative’ case (by default), instead of the structural genitive assigned by the overt preposition. The accusative case is in fact assigned at an extra structural layer, KP, dominating nP. The spatial n is not an ordinary n whose root has a referential interpretation. It has a ‘lighter’ lexical content, and it depends on (the silent abstract) P for its interpretation. Adopting Nchare and Terzi’s (2014) silent P analysis, Fassi Fehri and Alrawi (2021) assume that the spatial n undergoes head movement to P according to the requirement imposed by the Edge of XP, developed by Collins (2007), which goes back to the Doubly Filled Comp Filter (Chomsky and Lasnik 1977):6

(15)	a.	Edge(X) must be phonetically overt.
	b.	The condition in (a) applies in a minimal way, so that either the head or the specifier, but not both, are spelled out overtly.
		(Nchare and Terzi 2014, p. 694)

Accordingly, the locative or directional n head merges through K into Place or Path, as in (16):

(16)

It is worth noting that not all directional spatial nouns are headed by a P. For example, ʕabr-a ‘through’, as illustrated in (17), does not allow an alternation with the overt preposition:

(17)	a.	saafar-tu	ʕabr-a      l-muḥiiṭ-i
		traveled-I	through-acc   the-ocean-gen
		‘I traveled through the ocean.’
	b.	*saafar-tu	min	ʕabr-i	l-muḥiiṭ-i
		traveled-I	from	through-gen	the-ocean-gen
		Intended meaning: ‘I traveled from through the ocean.’

2.1.3. Motional PPs

Spatial nouns, such as ṣawb-a ‘towards’, naḥw-a ‘towards’, or tujaah-a ‘towards’, that express motion Paths as illustrated in (18), constitute a distinct class from both locational and directional nPs. Their structure is arguably more complex than directional Paths as represented in (19):

(18)	a.	ḏahaba	s-sahm-u	ṣawb-a	r-rajul-i
		went	the-arrow-gen	towards-acc	the-man-gen
		‘The arrow went towards the man.’
	b.	taḥarrak-tu	naḥw-a		š-šamaal-i
		moved-I	towards-acc		the-north-gen
		‘I moved towards the north.’

(19)

The motional n forms a distinct category that usually collocates with motion verbs, as exemplified in (18). Without such verbs, the nominal sentence is ungrammatical:

(20)	a.	*s-sahm-u	naḥw-a	r-rajul-i
		the-arrow-nom	towards-acc	the-man-gen
		Intended meaning: ‘The arrow (went) towards the man.’
	b.	*ʔanaa	tujaah-a	š-šamaal-i
		I	towards-acc	the-north-gen
		Intended meaning: ‘I am (heading) towards the north.’

Another property that sets these prepositional expressions apart from both locative and directional spatial nouns is that they do not collocate with stative verbs such as đ̣alla ‘remained’, as illustrated in (21).

(21)	a.	*đ̣alla	s-sahm-u	ṣawb-a	r-rajul-i	*motional
		remained	the-arrow-nom	towards-acc the-man-gen
		Intended meaning: ‘The arrow remained towards the man.’
	b.	đ̣alla	s-siyaaj-u	ḥawl-a	l-bayt-i	directional-static
		remained   the-fence-nom    around-acc			the-house-gen
		‘The fence remained around the house.’
	c.	đ̣alla	s-sahm-u	quddam-a	r-rajul-i	locative-static
		remained the-arrow-nom     in front of-acc  the-man-gen
		‘The arrow remained in front of the man.’

A third property is that motional spatial nouns (or Axpart) do not collocate with change-of-state verbs like staqarra ‘settled’, Ɂaṣbaḥa ‘became’, etc., which are dynamic but not motional, as illustrated in (22), unlike both locational and directional PPs/nPs, which can freely collocate with change-of-state verbs, acquiring a dynamic sense, as illustrated in (23) and (24):

(22)	*staqarra	s-sahm-u	ṣawb-a	l-masjid-i	*motional
	settled	the-arrow-nom	towards-acc	the-mosque-gen
	Intended meaning: ‘The arrow settled towards the mosque.’

(23)	staqarra	l-ʕaduww-u	ḥawl-a	l-qalʕat-i
	settled	the-enemy-nom	around-acc	the-castle-gen
	‘The enemy settled around the castle.’             dynamic; non-motional

(24)	Ɂaṣbaha	r-rajul-u	Ɂamaam-a	l-baab-i	dynamic; non-motional
	became	the-man-nom	in front of-acc the-door-gen
	‘The man became in front of the door.’

A fourth property is that motional expressions are strictly motional Paths that are only compatible with motional dynamic contexts, but incompatible with static contexts, as illustrated above. They contrast with both directional and locative PPs, which can be either static or directional dynamic depending on the context, as illustrated in (25) and (26):

(25)	a.	saafar-tu	ʕabr-a	l-muḥiiṭ-i	directional-dynamic
		traveled-I	through-acc	the-ocean-gen
		‘I traveled through the ocean.’
	b.	ṭ-ṭariiq-u	ʕabr-a	l-muḥiiṭ-i	ṭuul-u-hu
		the-route-nom	across-acc	the-ocean-gen	length-nom-its
		ʕašra-t-u	kiluumitr-aat-in	directional-static
		ten-fem-nom	kilometer-fem.pl-gen
		‘The route across the ocean is about ten thousand kilometer long.’

(26)	a.	ṭ-ṭuruq-u	fii	jaddat-a	waasiʕat-un	locative-static
		the-roads-nom	in	Jeddah-gen	wide-nom
		‘The roads in Jeddah are wide.’
	b.	fa-siiruu	fii	l-Ɂarḍ-i		locative-dynamic
		so-travel	in	the-earth-gen
		‘So travel through the earth.’ (Quran, 3)

So, both locatives and directionals can be either dynamic or static. But motional nouns are endowed with a [motion] feature that locational or directional Ps or ns are not endowed with. One piece of evidence comes from the fact that motional spatial nouns express a motional reading when collocating with a directional but not (necessarily) motional verb such as ittajaha ‘directed’ (27a). This reading is not available with a locative such as ʔamaam-a ‘in front of’ or a directional spatial noun such as ʕabr-a ‘across’ (as in 27b,c):

(27)	a.	ittajaha	s-sahm-u	naḥw-a	r-rajul-i	motional
		Directed	the-arrow-nom	towards-acc	the-man-gen
		‘The arrow was directed towards the man.’
	b.	*ittajaha	s-sahm-u	ʔamaam-a	r-rajul-i	*locative
		directed   the-arrow-nom		in front of-acc the-man-gen
		‘The arrow was directed in front of the man.’
	c.	*ittajaha	s-sahm-u	ʕabr-a	r-rajul-i	*directional
		directed	the-arrow-nom	across-acc	the-man-gen
		‘The arrow was directed across the man.’

Another piece of evidence can be constructed as follows. Verbs derived from the same base roots of these adverbial nouns are motional, as illustrated in (28), suggesting that the motion feature is akin to the base root:

(28)	a.	ʔaṣaaba	s-sahm-u	r-rajul-a
		hit	the-arrow-nom	the-man-acc
		‘The arrow hit the man.’
	b.	naḥaw-tu		naḥw-a-ka
		leaned-I		towards-acc-you
		‘I leaned towards you.’

On the other hand, no motion verbs can be derived from non-motional roots, be they locative nouns such as ʔamaam-a ‘in front of’, xalf-a ‘behind’, and quddam-a ‘in front of’, or directional nouns such as ḥawl-a ‘around’, xilaal-a ‘through’, and ʕabr-a ‘through’, because their roots, we suggest, are not marked for motion, as illustrated in (29) and (30):

(29)	a.	*ʔamaama	s-sahm-u	r-rajul-a
		went in from of	the-arrow-nom	the-man-acc
		Intended meaning: ‘the arrow went in front of the man’.
	b.	*ʔaxlafa	s-sahm-u	r-rajul-a
		went behind	the-arrow-nom	the-man-acc
		Intended meaning: ‘The arrow went behind the man.’

(30)	a.	*ʔaʕbara	s-sahm-u	r-rajul-a
		went across	the-arrow-nom	the-man-acc
		Intended meaning: ‘The arrow went across the man.’
	b.	*ʔaḥwala	ṭ-ṭaaɁir-u	š-šajarat-a
		went around	the-bird-nom	the-tree-acc
		Intended meaning: ‘The bird went around the tree.’

We propose, therefore, that [motion] is a feature of the root of motional spatial objects or events. This feature is not available for the locational or directional roots, which have [loc] and [dir] features, as represented in (31):

(31)

At the same time, it is reasonable to think that Place and Path meanings (or ontologies) are associated with acategorical roots, and that [loc], [dir], or [motion] are content features of these roots in the root syntax of PPs. By contrast, the categorizing head p is found in a separate projection in category syntax, that is pP.

2.2. Sorts of Domains

Paths, Places, or Individuals are entities of different sorts that are cognitively founded and motivated (Svenonius 2012). The Place domain describes a region, which is modelled as an entity of sort ℓ (for ‘location’), the Path domain provides a description of sort ℘, and the determiner domain D provides a description of sort d (for individual). Between these sortal domains lie p and K, which are transitional relational heads between the domain of sort ℘, the domain of sort ℓ, and that of sort d, as schematized in (32):

(32)

3. Evidence for Rooted PPs

In this section, we build a number of arguments that give further support to the dual nature of PP as both lexical (or root based) and functional (categorized).

3.1. The Multiple Category Argument

As observed earlier, roots of prepositional phrases, (expressing Places or Paths), just like roots of other categories, are acategorial. When Places and Paths surface as categories, they do so as p, n, a, or v. Axpart configurations also take part in this behavior. For example, ʕalaa ‘on’ functions as a ‘preposition’ in (33a), as a ‘nominal’ in (33b) or (33c), as a comparative adjective in (33d), and as a (locative) verb in (33e):

(33)	a.	al-kitaab-u	ʕalaa     ṭ-ṭaawilat-i
		the-book-nom	on      the-table-gen
		‘The book is on the table’.
	b.	Ɂazal-tu	n-naḥlat-a	min   ʕalaa	đ̣ahr-i-ka
		removed-I	the-bee-acc	from   on	back-gen-your
		‘I took out the bee from above your back.’
	c.	nazala	l-maaɁ-u	min   Ɂaʕlaa
		came down	the-water-nom	from   above.gen
		‘The water went down from above.’
	d.	l-jabal-u	Ɂaʕlaa   min	xams-imiʔat-i
		the-mountain-nom	higher   than	five-hundred-gen
		mitr-in
		meters-gen
		‘The mountain is higher than five hundred meters.’
	e.	ʕala-t	l-miyaah-u	l-jabal-a
		went.high-f	the-water-nom	the-mountain-acc
		‘The water went up the mountain.’

It is reasonable to think that these words share the same root as the preposition ʕalaa ‘on’, assuming that the latter is born as an acategorial Place in √RootP, and later categorized as p, n, a, or v; these options are simply represented as in (34):7

(34)

3.2. Multiple Embeddings of AxPart Roots

Similar results obtain with adverbial Axparts. Thus, fawqa ‘above’, xalfa ‘back’, xaarija ‘outside’, daaxila ‘inside’, ʕabra ‘through’, etc., can function as verbs in (35), nouns in (36), and adjectives in (37):

(35)	a.	kaana	ʕadad-u-hum	y-afuuq-u	l-xamsiin-a
		was	number-nom-their	3-exceeded-ind	the-fifty.acc
		‘Their number was exceeding fifty.’
	b.	tafawwaqa	muḥammad-un	ʕalaa	Ɂax-ii-hi
		outperformed	Mohammed-nom	on	brother-gen-his
		‘Mohammed outperformed his brother.’
	c.	xalafa	min	baʕd-i-him	xalf-un
		came after     from	after-gen-them	successors
		ʔaḍaaʕ-uu   ṣ-ṣalaat-a neglected-they  the-prayer-acc
		‘There came after them successors who neglected prayer.’ (Qur’an, 16)
	d.	xaraja	r-rajul-u	min	bayt-i-hi
		left	the-man-nom	from	house-gen-his
		‘The man left his house.’
	e.	daxala	r-rajul-u	bayt-a-hu
		entered	the-man-nom    house-acc-his
		‘The man entered his house.’
	f.	ʕabar-tu	n-nahr-a
		crossed-I	the-river-acc
		‘I crossed the river.’

(36)	a.	waqaf-tu     fii		l-xalf-i
		stood-I     in		the-back-gen
		‘I stood up in the back.’
	b.	šaraʕ-tu     fii		l-xuruuj-i
		started-I     in		the-leaving-gen
		‘I started to leave.’
	c.	l-duxuul-u		min		hunaa
		the-entering-nom		from		here
		‘The entrance is from here.’
	d.	nu-ḥaawilu	l-ʕubuur-a	Ɂilaa	ḍ-ḍiffat-i	l-ʔuxraa
		we-try	the-crossing-acc	to	the-side-gen	the-other.gen
		‘We try to cross to the other side.’

(37)	a.	muḥammad-un			ṭaalib-un   mutafawwiq-un
		Mohammed-nom			student-nom   top-nom
		‘Mohammed is a top student.’
	b.	staxdim			l-baab-a     l-xalfiyy-a
		use			the-door-acc   the-back.acc
		‘Use the back door.’
	c.	laday-ya	ʔalam-un	fii	l-juzʔ-i	d-daaxil-i    min
		to-me	pain-nom	in	the-part-gen	the-inner-gen of
		l-ʔuḏn-i
		the-ear-gen
		‘I have a pain in the inner part of the ear.’
	d.	ntađ̣ar-tu-ka	fii		l-qaaʕat-i	l-xaarijiyyat-i
		waited-I-you	in		the-room-gen	the-outer-gen
		‘I waited for you in the outer room.’

As seen in (35), (36), and (37), these words seem to share the same root with fawq-a ‘above’, xalf-a ‘back’, xaarij-a ‘outside’, daaxil-a ‘inside’, and ʕabr-a ‘through’. Their structures can be simply represented as in (38):

(38)

3.3. Alternating Syntactic and Semantic PP Subtypes

PPs alternate as semantico-aspectual morpho-syntactic types (static/dynamic, loc/dir, Place/Path), in verbless-aspectless (copulative) or verbal (aspectuo-temporal) sentences. They can be classified into three syntactico-semantic subtypes: (a) Loc only, such as fii ‘in’; (b) Dir only, such as sawb-a ‘towards’; or (c) +Dir, +Loc, such as Ɂilaa ‘to’. In fact, sawb-a is also motional, as explained above, a fourth subtype. Fii ‘in’ occurs in copulative sentences, inducing a locative meaning, as illustrated in (4a) (repeated here in (39a)), unlike sawb-a ‘towards’, which is incompatible with static copulative sentences, as illustrated in (39b):

(39)	a.	Ɂanaa   fii	l-bayt-i
		I     in	the-house-gen
		‘I am in the house.’
	b.	*s-sahm-u	ṣawb-a	r-rajul-i
		the-arrow-nom	towards-acc	the-man-gen
		‘The arrow towards the man.’

As a matter of fact, ṣawb-a ‘towards’ requires a dynamic motional context, and it normally collocates with motional verbs, as illustrated in (40).

(40)	taḥarraka	s-sahm-u	ṣawb-a	r-rajul-i
	moved	the-arrow-nom	towards-acc	the-man-gen
	‘The arrow moved towards the man.’

The roots of fii ‘in’ and sawb-a ‘towards’ are specified for either only Loc or only Dir or Mot(ion) features. But unlike fii and sawb-a, Ɂilaa ‘to’ has a root that can be +Dir or +Loc. The duality of √Ɂilaa ‘to’ has been already introduced in (11a) as PlaceP and in (7) as PathP. It allows the preposition Ɂilaa ‘to’ to occur in both static copulative contexts, as in (9) above, inducing a locative meaning, and dynamic contexts (6a), inducing a directional meaning, hence the two structures in (41):

(41)

3.4. Conflating Prepositions in Distinct Event Classes or Roles

In the verbal domain, when PPs are part of the semantico-syntactic structure of vPs, syntactic alternations are more clearly associated with semantic correlates: containment, contact8, (±central) coincidence, etc.9

Consider the contact transitive/PP alternation associated with ‘bi’:

(42)	a.	Ɂaxađ-tu	bi-yad-i-hi	b.	Ɂaxađ-tu	yad-a-hu
		took-I	at-hand-gen-his		took-I	hand-acc-his
		‘I took (at) his hand.’			‘I took his hand.’

In (42a), bi- ‘at’ collocates with the verb Ɂaxađ ‘took’, creating a meaning of contact. This meaning is preserved in the null counterpart (42b), when bi- is covert. The difference then between (42a) and (42b) is basically in terms of case assignment. The object yad ‘hand’ receives a genitive case in (42a) but an accusative case in (42b). That is to say, the meaning connected with the root is available in both (42a) and (42b), but the genitive case assigning head is available in (42a) but not in (42b). This motivates a separation between the root head (√) that expresses the contact meaning and the functional head p that assigns genitive to the complement. The structures of (42a) and (42b) can be tentatively represented in (43a) and (43b), respectively, where p is either overt or covert:

(43)

Another syntactic alternation is related to the containment alternation associated with fii and bi-. While both fii and bi- seem to relate the same arguments, the roles assigned to the arguments change depending on which preposition is used. Fii locates a containee Figure with respect to a container Ground. But bi- entails that the container is the Figure and the containee the Ground, the position of which coincides with that of the Figure, reversing somehow the position of the roles. The alternating contrast is illustrated by the pair of constructions in (44):

(44)	a.	štaʕala	š-šayb-u     fii	r-raɁs-i
		filled	the white hair-nom  in         the-head-gen
		‘The white hair filled the head.’
	b.	štaʕala	r-raɁs-u    bi-š-šayb-i
		filled	the-head-nom   with-the-white hair-gen
		‘The head filled with white hair.’
		(Fassi Fehri 1986; 2021, p. 184)

In (44a), the DP complement of fii ‘in’ (r-raɁs-i ‘the head’) is a G container, whereas the complement of bi- ‘with’ (š-šayb-i ‘the white hair’) is an F containee. At the same time, the subject (or external argument) is an F containee with fii, and a G container with bi-, or so it seems. These contrasts can be represented as in (45a,b), respectively:

(45)

A similar container/containee placement alternation is also manifest in the following pair of sentences:

(46)	a.	malaɁ-tu	l-jarrat-a		bi-z-zayt-i
		filled-I	the-jar-acc		with-the-oil-gen
		‘I filled the jar with oil.’
	b.	malaɁ-tu	z-zayt-a	fii	l-jarrat-i
		filled-I	the-oil-nom	in	the-jar-gen
		‘I filled the oil in the jar’.

In both (46a) and (46b), l-jarrat ‘the jar’ is a container, and z-zayt ‘the oil’ is the containee. The difference is that bi- ‘with’ relates the container to the containee and fii ‘in’ does the opposite. Both prepositions collocate with the same verb as represented in (47a) and (47b). The complement of fii, the container, receives a ‘locative’ G theta role in (47b), whereas the complement of bi-, the containee, appears to play a sort of Means/Instrument role (see below for more discussion):

(47)

It is reasonable to think that the meaning associated with the prepositional root is what determines the thematic roles of its arguments. √bi- carries the meaning of means that makes its complement z-zayt-i ‘the oil’ a tool which the container, l-jarrat-a ‘the jar’, is filled with, whereas √fii has a LOC containment meaning that makes its complement l-jarrat-i ‘the jar’ a true container for the Figure containee, z-zayt-a ‘the oil’.

The relation between the Figure and the Ground can also be analyzed in terms of (± central) coincidence, as proposed by Hale and Keyser (1998). It is important to note that fii establishes a ‘central coincidence’ relation, while bi- establishes a ‘non-central coincidence’ relation, as will be explained in the next section.

4. Complex Polysemies and Syntactic Alternations

PPs are found in more complex polysemies, structures, and dialectal variations. In our system, these phenomena are taken care of by distinct structures, depending on constructions that may or may not be related morpho-syntactically, in association with their semantics. Cognitivists, by contrast, rely on hierarchically ordered networks associated with words (Jan 2018; Lentzner 1977; Esseesy 2010). We claim that this hierarchically ordered ‘word approach’ is inadequate, given that they do not assume a tight correlation between syntactic (or morpho-syntactic) structure and semantics, do not make meaning depending on constructions (but on words), and hence do not make clear predictions about form–meaning correlations built in the syntax of languages, or about empirically motivated lexical classes or subtypes behind dialectal variation, as we will explain below.10

4.1. Meanings of fii and bi-

In traditional grammar, each P is polysemous, although meanings are ordered, starting with basic or primary meanings (Ibn Hišaam 1985; Al-Muraadii 1992, among others). Fii has a number of meanings that are instantiations of the ‘lexical’ root P. Chiefly, among those meanings is ‘locative and temporal circumstantiality’, or more specifically, containment htiwaaʔ, as illustrated in (48):

(48)	a.	zayd-un fii	d-daar-i
		Zayd-nom  in	the-house-gen
		‘Zayd is in the house.’
	b.	sa-ʔuṯrii    fii	biḍʕ-i		siniina
		fut-be rich in	few-gen		years.gen
		‘I will be rich in few years.’

Another meaning is causation taʕliil, as in (49).

(49)	qutila	kulayb-un	fii	naaqat-in
	killed.pass	Kulayb-nom	in	cow-gen
	‘Kulayb was killed for (steeling/losing) a cow.’			(Jan 2018, p. 133)

A third meaning is accompaniment musaahaba:

(50)	xaraja	l-ʔamiir-u	fii mawkib-in	ḥaafil-in
	went out	the-prince-nom	in  procession-gen	festive-gen
	‘The prince went out in a festive procession.’   (Jan 2018, p.133)

A fourth meaning is superposition stiʕlaaʔ:

(51)	ṣalaba-hum	fii juḏuuʕ-i	n-naxl-i
	crucified-them  in  trunks-gen		the-palm trees-gen
	‘He crucified them on the trunks of palm trees.’   (Jan 2018, p. 134 )

In modern grammars, most Arabic linguists converge on the idea that fii has a broad locative meaning, and a more specific meaning of containment (Fassi Fehri 2021; Lentzner 1977) that corresponds to those of the English preposition ‘in’, or more extensively ‘at’ or ‘on’ as illustrated in (52), (53), and (54), respectively (see also Jan 2018; Esseesy 2010):

(52)	l-jamaaʕat-u	fii	l-masjid-i
	The-group-nom	in	the-mosque-gen
	‘The group is in the mosque.’
(53)	y-atajawwalu	fii	š-šaariʕ-i
	3-walk	in	the-avenue-gen
	‘He walks in the avenue.’
(54)	l-qimmat-u	fii	y-ulyuu
	the-summit-nom	in	July.gen
	‘The summit is in July’

The locative meaning of fii can be partial containment (rather than total), in which the Ground partially contains the Figure, as in (55):

(55)	a.	l-ʕuṣfuur-u	fii	š-šajarat-i
		the-bird-nom	in	the-tree-gen
		‘The bird is in the tree.’
	b.	l-wardat-u	fii	l-mizhariyyat-i
		the-flower-nom	in	the-vase-gen
		‘The flower is in the vase.’

The bird in (55a) is not contained by the tree but it is located in one branch of the tree, and only part of its body is attached to the branch. Similarly, in (55b), only part of the flower is located inside the vase (Fassi Fehri and Aamiri 2021; Esseesy 2010).

Moreover, localization can be broad or abstract, where no physical or concrete containment can be inferred, as in (56):

(56)	ṭanjat-u	fii	šamaal-i	r-ribaaṭ-i
	Tangier-nom	in	north-gen  Rabat-gen
	‘Tangier is to the north of Rabat.’

In (56), there is no physical containment relation between Tangier and Rabat.

In the time domain, fii situates the Figure in a certain Ground moment in time and means specifically central coincidence in Tense and Aspect. It corresponds to English ‘at’ or ‘in’, which means that the Figure and the Ground either coincide (i.e., overlap in the position in time), as in (57a), or that the Ground is a container inside of which the Figure is located, as in (57b).

(57)	a.	mawʕid-u-na	fii	l-xaamisat-i
		appointment-nom-our  in		the-five-gen
		‘Our appointment is at 5 o’clock.’
	b.	naḥnu	fii	l-masaaʔ-i
		we	in	the-evening-gen
		‘We are in the evening’.

What is of interest, in particular, is to see whether the primary meanings of fii- can be confused with those of other close prepositions such as bi-. As we will see, their meanings can be contrasted, and the answer appears to be negative.

Bi- ‘at’ is the closest preposition to fii ‘in’. But according to traditional grammar, its primary meaning is contact (ʔilsaaq), compared to fii, which is primarily dedicated to containment or inclusion. Prototypically, constructions include first the following verbal complex, where a contact is established between me and the man’s hand in (58a), or me and the man in (58b):

(58)	a.	Ɂaxađ-tu	bi-yad-i	r-rajul-i
		took-I	at-hand-gen	the-man-gen
		‘I took the man’s hand.’
	b.	ltaqayt-tu	bi-r-rajul-i
		met-I	at-the-man-gen
		‘I met with the man.’

In all these contexts, fii cannot be used instead of bi- in the standard variety of Arabic. Other typical meanings of bi- include instrument/means, or manner, as in the following pair, respectively:

(59)	a.	ṭaʕana-hu	bi-sikkiin-in
		stabbed.3-him	with-knife-gen
		‘He stabbed him with a knife.’
	b.	ṭaʕana-hu	bi-surʕat-in
		stabbed.3-him	with-quickness-gen
		‘He stabbed him quickly.’

What is problematic, in traditional descriptions, however, is that they make no room for the general locational meaning that bi- can express, which entails that the Figure and the Ground can coincide spatially (or temporally), as in the following example:

(60)	ʔanaa	bi-baab-i	manzil-i-ka
	I	at-door-gen	house-gen-your
	‘I am by/at your house door.’

Clearly, the preposition in (60) contributes to locate F with respect to G, but without inducing any containment or inclusion, as indicated by the fact that the introduction of fii is not an option here. Moreover, the translation of bi- should be ‘at’ rather than ‘in’. This difference clarifies the distinctive interpretation of the PP in (61a), compared to (61b):

(61)	a.	ʔinnaka	bi-l-waadii	l-muqaddas-i,	ṭawaa
		you	at-the-valley.gen	the-sacred-gen	Tawa
		‘You are at the sacred valley, Tawa.’  (Quran, 16)
	b.	ʔanta	fii	l-waadii
		you	in	the-valley.gen
		‘You are in the valley.’

We propose that contact or non-central coincidence is the specific meaning of the spatial bi- ‘at’. Hale (1986) claims that two essential distinct local cases in Walbiri can be defined along the coincidence notion, characterizing the role of F (or theme) with respect to G (or Place), or spatial–temporal relations more generally; see Demirdache and Uribe-Etxebarria (2014) for the (adapted) following formulation that extends to temporality:

(62)	a.	[+ central coincidence; +CC]: F within G (fii)
		Location, trajectory, or linear arrangement of F centrally coincides with G.
	b.	[- central coincidence; -CC]: F not within G: ‘terminal’, ‘initial’, etc. (bi-)
		Location, trajectory, or linear arrangement of F does not centrally coincide with G.

In (61a), the location of the interlocutor coincides with that of the valley, without inducing any containment or inclusion, contrary to (61b), that is, without inducing central coincidence. Fii ‘in’, on the other hand, locates the Figure in a position interior and within the boundaries of the Ground, a containment or central coincidence relation. What differentiates bi- ‘at’ from fii ‘in’ is, we propose, the meaning connected with the root. The root of bi- ‘at’ is distinct from that of fii ‘in’:

(63)

Therefore, in Standard Arabic, bi- ‘at’ does not normally replace fii ‘in’, although it so often does in dialects, a variation that we return to in the next subsection.

4.2. Lexical Variation as Distinct Vocabulary

Lexical variation can elegantly be associated with a distributed model in the standard variety or between other varieties. For example, in Mashriqi dialects, bi- ‘at’ replaces fii in denoting containment (in addition to central coincidence), as in (66a,b) below, from Levantine Arabic:

(64)	a.	Ɂana	bi-l-bet	b.	Ɂana	bi-l-bab
		I	in-the-house		I	at-the-door
		‘I am in the house.’			‘I am at the door.’

In Maghribi dialects, on the other hand, fii is often generalized, instead of bi (both are mostly reduced to f- and b-), as in Moroccan Arabic (65):

(65)	a.	Ɂana	f-d-dar	b.	Ɂana	f-l-bab
		I	in-the-house		I	at-the-door
		‘I am in the house.’			‘I am at the door.’

Bi- in Mashriqi dialects replaces fii- in denoting containment in time:

(66)	n-tlaaqa	bi-l-masaa/bi-l-lail
	we-meet	in-the-evening
	‘We meet in the evening/at night.’
	(Fassi Fehri and Aamiri 2021, p. 160)

In Maghribi dialects, on the other hand, f- is used instead of b-, as in Moroccan Arabic (67):

(67)	n-tlaaqaw	f-l-ʕašiia/f-l-liil
	we-meet	in-the-evening
	‘We meet in the evening/at night.’

With perception or contact verbs, Maghribi dialects, as in the (b) examples, more often use b-, while Mashriqi dialects use f-, as in the (a) examples:

(68)	a.	ḥasse-t	fiik	b.	ḥssee-t	bi-k
		felt.I	in-you		felt.I	at-you
		‘I felt you’

(69)	a.	taṣal-t	fi-k	b.	taṣal-t	bi-k
		contacted-I	in-you		contacted-I	at-you
		‘I contacted you.’

(70)	a.	raḥḥab	fiy-ya	b.	raḥḥab	biy-ya
		welcomed	in-me		welcomed	at-me
		‘He welcomed me.’

To express manner, b- is mostly used in these dialects, rather than f-:

(71)	a.	ta-n-tkellem	b-š-šweyya	(Maghribi)
		prog-I-talk	with-slow
		‘I talk slowly.’
	b.	be-n-t-kallam	b-šwayaš	(Mashriqi)
		prog-we-talk	with-slow
		‘We talk slowly.’

4.3. Morpho-Syntactic Alternations

Morpho-syntactic alternations play a major role in revealing the lexical side of the prepositions. Among these alternations are Places/Paths, transitivity, causativity, or other prepositional alternations.

4.3.1. Place and Path Alternations

Fii ‘in’ alternates between being a Place head in space and state as in (72), or a Path head when it collocates with a motion verb such as daxal ‘enter’, indicating a change of state as in (73), represented in (74):

(72)	a.	r-rajul-u	fii	l-manzil-i
		the-man-nom	in	the-house-gen
		‘The man is in the house.’
	b.	r-rajul-u	fii	ġaybubat-i-n
		the-man-nom	in	coma-gen-n
		‘The man is in a coma.’

(73)	daxala	r-rajul-u	fii	ġaybubat-i-n
	entered	the-man-nom	in	coma-gen-n
	‘The man entered into a coma.’

(74)

The motion event can only select a PathP. When fii collocates with a motion event, it denotes a directional Path, having basically the meaning of the directional Path ʔilaa, rather than a pure locational Place.

4.3.2. Transitivity and Causative Alternations

Constructions with ʔilaa ‘to’ seem to exhibit a prepositional/accusative alternation in which ʔilaa ‘to’ is either overt or hidden:

(75)	a.	ṣaʕada	r-rajul-u	ʔilaa   l-jabal-i
		climb	the-man-nom	to   the-mountain-gen
		‘The man climbed (to) the mountain.’
	b.	ṣaʕada	r-rajul-u	l-jabal-a
		climb	the-man-nom	the-mountain-acc
		‘The man climbed the mountain.’

ʔilaa expresses a Goal–Path meaning here. This meaning is available in both (75a) and (75b), which suggests the presence of a Path root P in both constructions. If so, then the difference appears to be only in the case assignor. The genitive case assignor, p, is available in (75a) but not in (75b). The absence of the structural case assigner, p, leaves l-jabal-a ‘the mountain’ in (75b) without case. It is then assigned a sort of oblique ‘accusative’ by default. The two alternations can be represented as follows:

(76)

Consider now causative alternations. Prepositional/accusative alternations also occur in causative constructions in which the preposition has an overt/null alternation, as illustrated in (77):

(77)	a.	ʔ-ahday-tu	kitaab-an	li-zayd-in
		caus.gave-I	book-acc	to-Zayd-gen
		‘I gave a book to Zayd.’
	b.	ʔ-ahday-tu	zayd-an	kitaab-an
		caus.gave-I	Zayd-acc	book-acc
		‘I gave Zayd a book.’

In both (77a) and (77b), Zayd is the dative or the Recipient of the action. The dative interpretation is assigned by the Goal root √li-. To put it differently, the dative root √li- is assumed to be the source of the dative interpretation, and it is available for both (77a) and (77b). The two constructions then differ only in the source of case assignment. The prepositional p head assigns the genitive case in (77a). But in order for Zayd to receive (structural dative) case in (77b), it moves to a position where it is assigned accusative by v. The examples in (77a) and (77b) are represented in (78a) and (78b), respectively:11

(78)

If (78a) is roughly the basic structure for both constructions in (77), then the dative has to move higher in (78b), adjoining to the pP, to get structural case there.

A further example of an overt/null alternation is illustrated in (79) and represented in (80):12

(79)	a.	kasaw-tu	l-walad-a	bi-ṯ-ṯawb-i   (tool/instrument)
		wore-I	the-child-acc	with-the-cloth-gen
		‘I have worn the child with the cloth.’
	b.	kasaw-tu	l-walad-a	ṯ-ṯawb-a
		wore-I	the-child-acc	the-cloth-acc
		‘I made the child wear the cloth.’       (Fassi Fehri 2021, p. 183)

(80)

In both (80a) and (80b), ṯ-ṯawb ‘the cloth’ is the tool or instrument used for the action of wearing. In (80a), the causative verb kasaa ‘made him wear’ makes use of the overt preposition bi-, in expressing the Instrument role. The preposition bi- is dominated by the functional p, and it assigns the complement ṯ-ṯawb a genitive case. In (80b), the preposition is silent, and ṯ-ṯawb ‘the cloth’ receives an accusative case via v.

4.3.3. Distinct Role Alternations

Prepositions can have closely related meanings that can relate the same arguments (Figure and Ground) but with different relations, as discussed in Section 3.4 above. Consider further alternations of fii ‘in’ and bi- ‘by’ in (81):

(81)	a.	ʔaṯṯara	l-ḥaadiṯ-u	fii r-rajul-i
		affected	the-accident-nom	in  the-man-gen
		‘The accident affected the man.’
	b.	ta-ʔaṯṯara	r-rajul-u	bi-l-ḥaadiṯ-i
		reflex-affected	the-man-nom	by-the-accident-gen
		‘The man got affected by the accident.’

In (81a), l-ḥaadiṯ-u ‘the accident’ is the Figure or the Cause and r-rajul-i ‘the man’ is the complement Ground or the entity affected, as represented in (82a). On the other hand, r-rajul-u ‘the man’ in (81b), which is the Ground, is in the Specifier position, and the Figure l-ḥaadiṯ-i ‘the accident’ is the complement, as represented in (82b). It is true that the selection of the preposition depends on the form of verb, but the relation between the Figure and the Ground is determined by the preposition (or, more specifically, the prepositional root).

(82)

These alternations suggest the presence of the lexical meaning associated with the P root, regardless of its function as a case assignor.

5. Conclusions

In this article, we have argued for a dual PP structure: a (lexical) √P and a (functional) categorizing pP. One complexity of the traditional one-projection structure (i.e., PlaceP and PathP) is that not all Ps that are semantically locational Places or directional Paths function as prepositions. Adverbial AxPart phrases (đ̣uruuf) semantically denote location, direction, or motion, but function syntactically as nouns or nPs. We provided evidence that prepositions have a dual life based on the multiple category argument, the syntactic and semantic subtypes of the prepositional roots, and the conflation of prepositions in distinct event classes and roles, resulting in overt/null prepositional/accusative alternations. Different categories of v, n, a, or p can be derived from the same prepositional root, which carries the meaning of three semantic subtypes: locational, directional, and motional. It is proposed that the meaning content of the PP survives even when the preposition is silent. Further complex polysemies, variations, and syntactic alternations have been provided to add further support to the distinct lexical and functional projections of the prepositions. Lexical variation between fii and bi is accounted for as distinct vocabulary instantiations of the same contentful root. Such considerations can hardly be accounted for if the lexical root (‘big P’) and the case assignor (‘small p’) are fused into one projection. The arguments provided throughout the article aim at establishing the superiority of the dual distributed treatment of the PP structure, compared to single projection approaches. If true, then the duality of root and category specifications of DM generalizes to PPs, in addition to other more studied verb, noun, or adjective roots and categories, already well-integrated into the system.