The analysis in this paper is part of an ongoing doctoral dissertation on the morphology of loanwords. I am grateful to Karim Bensoukas for his constructive feedback and insightful comments, which greatly improved the manuscript. Special thanks to Fatima El Hamdi and Badr Eddine El Asri Traiki for their assistance with and discussion of this research. I have also benefited from the comments of two anonymous Asinag reviewers. All remaining errors are my own.

1. Introduction

2This article provides an account for backformation in loanwords in Moroccan Arabic. It argues that the operation consists in forming a verb by extracting a consonantal root from an integrated loan-noun. We claim that this process provides evidence for the consonantal root as a base of derivation. The paper, hence, contributes to the ongoing debate as to whether Semitic morphology is based on the consonantal root, the word, or both (Ussishkin, 1999, 2006; Prunet, 2006; Bat- El, 1994, 2003; Ratcliffe, 1997 and others). The main claim of this paper is that the root is the basic unit in morphological/lexical representation.

3This paper is divided into five sections. After the introduction, we present in section 2 the different views with regard to the notion of the root, particularly the consonantal root. Section 3 presents data from loan verbs derived from French loan nouns based on the process of backformation via root-extraction. We claim in section 4 that the extracted consonantal root serves as a base of derivation for other morphological categories (the medio‑passive, the passive participle, and the deverbal noun), in analogy with MA respective patterns. We also examine internal gemination and claim it to provide further evidence in favor of the extraction of the consonantal root and its serving as the base of derivation in MA. By way of illustration, we provide examples from causatives derived from loan verbs. We recapitulate the findings in section 5.

2. Root-based vs. word-based approaches to Semitic morphology

4Prunet (2006) defines the “root” as the morphological unit that is left once all the affixes have been removed from the word. According to Arad (2003), roots are the atomic non-decomposable lexical elements of a language, serving as the building blocks of more complex words. If the root consists of consonants only, it is called a “consonantal root”, while if it also contains vowels, then it is referred to as a “syllabic root” (Arad, 2003; Prunet, 2006). For the purposes of this paper, the focus will be on the consonantal root.

5The consonantal root is a widely controversial concept especially in Semitic languages. Prunet (2006) provides a critical review of different theoretical and psycholinguistic works, including Prunet et al. (2000) and Davis and Zawaydeh (2001). The findings reveal that, in the case of speech errors and language games, only root consonants are reversed. Similarly, data from hypocoristics (nicknames) show that the consonants in the full name and in the hypocoristic form are the same three consonants of the lexical root (Davis and Zawaydeh, 2001). The previous research provides strong evidence for the existence of an abstract discontinuous consonantal root and, thus, of an unpronounceable entity (Prunet et al., 2000; Prunet, 2006; Schluter, 2013; Diab and Marton, 2014).

6A unique feature of Semitic languages such as Arabic and Hebrew is the combination of roots and patterns, a type known as root-and-template morphology (McCarthy, 1981, for example). In this model, words are formed when roots are combined with a template and vocalic morphemes. In some languages, as is the case in MA, there is only a root and a template with epenthetic vowels (ə in MA). The consonantal root is also considered as a morpheme carrying the meaning of the words containing this root (Prunet, 2006). For instance, the three radicals k t b in Arabic and Hebrew constitute a root related to writing. The root in Arabic is mostly triconsonantal but can also be biconsonantal or quadriconsonantal.

7In opposition to these root-based approaches, scholars like Bat-El (1994, 2003), Ratcliffe (1997), and Ussishkin (1999) argue for a word-based view of Hebrew morphology that does away with the consonantal root. They analyze patterns of denominal verb formation, showing that only a word-based account can explain otherwise mysterious phonological regularities. They question the role of the consonantal root in the derivation and suggest word formation from existing words (Bat-El, 1994, 2003; Ussishkin, 1999).

8In this paper, we argue that the consonantal root is the base of derivation in MA on the basis of evidence provided from verbs derived through backformation from loan nouns in MA.

3. Backformation from loan-nouns as root extraction

10According to Mattiello (2013), creating new words via backformation may be the outcome of reanalysis of supposedly complex foreign forms in analogy with the native inflectional patterns. For instance, Old Northern French cherise (Modern French cerise) has been reanalyzed in English as a singular form cherry in addition to the plural suffix –s, following the plural formation pattern in English.

11In (1) below, we present cases of backformation in MA based on French loan- nouns, and derivation on verbs therein.

(1) Loan- Noun Origin/ Gloss		Root	Verb	Gloss
kustim <costume	‘suit’	√kstm	kəstəm	‘wear a suit’
garṣun <garçon	‘waiter’	√grṣn	gərṣən	‘serve as waiter’
grafaṭa <cravate	‘tie’	√grfṭ	gərfəṭ	‘wear a tie’
randifu <rendez-vous	‘appointment’	√rndf	rəndəf	‘have an appointment’
ṣiʃwar <sechoir	‘hair dryer’	√ṣʃwr	ṣəʃwər	‘use a hair dryer’

12We notice that, although the loan nouns contain both vowels and consonants, only the consonants are extracted for the derivation of verbs, in analogy with existing native verbs such as √ṭṛʒm ‘translate’, √fṛɡʕ ‘explode’, and √krkb ‘roll’. Marchand (1969) stresses the role of analogy as a supportive element to the formation of a verb where the noun exists but the verb does not (cited in Staskova, 2013).

(adapted from Boudlal, 2001)

15We claim that, after extraction of the root from loan-nouns, other morphological categories will be derived based on the same root. We will notice that schwa epenthesis is applied in the same fashion as in native MA words.

4. The extracted consonantal root as a base of derivation

4.1. Normal affixation

16In this section, we show that once the root is extracted from loan-nouns via backformation, it serves as a base of derivation for other morphological categories, among which are the medio-passive, the passive participle and the deverbal noun.

17First, the medio-passive in MA is derived by prefixing tt- to a verb root. The phonetic form is completed by epenthesizing schwa (s) in appropriate places. Examples of MA native medio-passives are tt-xərbəq ‘got disorganized’, tt-fəṛɡəʕ ‘exploded’, tt-mənḍər ‘enjoyed beautiful views’, and tt-fǝrtǝt ‘turned into small pieces’. The following set shows how medio-passives are obtained in a similar fashion on the basis of consonantal roots extracted from French nouns:

(3)	Root	Medio-passive	Origin/	Gloss
	√ʔntk	tt-Ɂəntək	antique	‘elegant/dressed up’
	√ɡrfṭ	tt-ɡərfəṭ	cravatte	‘tie’
	√kstm	tt-kəstəm	costume	‘suit’
	√rndf	tt-rəndəf	rendez-vous	‘appointment’
	√ṣʃwr	tt-ṣəʃwər	sechoir	‘hairdryer’

18Second, the passive participle (PP) in MA is obtained through the prefixation of the morpheme m- to a verb root. Similar to the medio-passive data above, schwa is epenthesized to complete the phonetic shape of the loans. The following set shows how passive participles are derived in analogy with native PP, on the basis of the consonantal roots extracted from French nouns. Examples of native passive participles include m-kərfəṣ ‘in a deteriorated state’, m-bəṛqəʕ ‘stained’, m-xənzər ‘frowning’, and m-xərbəq ‘messy’:

(4)	Root	Passive participle	Origin	/Gloss
	√ʔntk	m-Ɂəntək	antique	‘elegant/dressed up’
	√ɡrfṭ	m-ɡərfəṭ	cravatte	‘tie’
	√kstm	m-kəstəm	costume	‘suit’
	√rndf	m-rəndəf	rendez-vous	‘appointment’
	√ṣʃwr	m-ṣəʃwər	sechoir	‘hairdryer’

19Third, the deverbal noun of quadrilateral verbs in MA is usually formed by the insertion of -i- between the last two consonants of the derived quadrilateral form. Some examples from MA deverbal nouns include tfəṛɡiʕ derived from the root √fṛɡʕ ‘explode’ and tfərkil derived from √frkl ‘to writhe, to flounder’ (Harrell, 1968).

(5)	Root	Deverbal noun	Origin	/Gloss
	√ʔntk	tɁəntik	antique	‘elegant/dressed up’
	√ɡrfṭ	tɡərfiṭ	cravatte	‘tie’
	√ɡrṣn	tɡərṣin	garçon	‘waiter’
	√kstm	tkəstim	costume	‘suit’
	√rndf	trəndif	rendez-vous	‘appointment’
	√ṣʃwr	tṣəʃwir	sechoir	‘hairdryer’

20It can be argued that the above data present very simple examples of backformation in the sense that once the consonant is extracted, it serves as a base of derivation for various morphological categories in a quite straightforward fashion. The next sections provide more intricate cases in which the extracted root plays a crucial role.

4.2. Root-and-pattern morphology

22We will go back to our data, which is divided into four sets: (6i), (6ii), (6iii) and (6iv), including the above mentioned categories, namely the medio-passive, the passive participle, and the deverbal noun. These data sets illustrate the productivity of the extracted roots.

(6i)	Root	Medio-passive	Origin	/Gloss
	√ḅɡṣ	tt-ḅuɡəṣ	beau gosse	‘handsome’
	√ṣfʒ	tt-ṣufəʒ	sauvage	‘savage’
	√ʃmr	tt-ʃumər	chomeur	‘jobless’
	√bdr	tt-budər	poutre	‘a physically strong person’
	√sɡr	tt-suɡər	seguro (Spanish)	‘to ensure; to be ensured’
	√kfr	tt-kufər	coup franc	‘free kick’

23In (6i), the medio-passive is formed in analogy with native medio-passives derived from triconsonantal roots such as √bhl tt-buhəl stated above.

24Similarly, in (6ii) the passive participle of the loan verbs is expressed as the native m-kusəl derived from the root √ksl.

(6ii)	Root	Passive participle	Origin	/ Gloss
	√ḅɡṣ	m-ḅuɡəṣ	beau gosse	‘handsome’
	√ṣfʒ	m-ṣufəʒ	sauvage	‘savage’
	√ʃmr	m-ʃumər	chomeur	‘jobless’
	√bdr	m-budər	poutre	‘physically strong person’
	√sɡr	m-suɡər	seguru (Spanish)	‘to ensure; ensured’
	√kfr	m-kufər	coup franc	‘free kick’

25The deverbal noun in (6iii) is formed in analogy with MA pattern in tkusil and tbuhil.

(6iii)	Root	Deverbal noun	Origin	/ Gloss
	√ḅɡṣ	tḅuɡiṣ	beau gosse	‘handsome’
	√ṣfʒ	tsufiʒ	sauvage	‘savage’
	√ʃmr	tʃumir	chomeur	‘jobless’
	√bdr	tbudir	poutre	‘physically strong person’
	√sɡr	tsuɡir	seguru	‘to ensure; to be ensured’
	√kfr	tkufir	coup franc	‘free kick’

26In (6i), (6ii) and (6iii), we listed the by now familiar French loanwords. In (6iv) the triconsonantal root √srt is derived from the Amazigh loan noun t-a-s-aru-t ‘key’.

(6iv)	Root	Medio-passive	Passive participle	Origin/ Gloss
	√srt	t-surət	m-surət (a)	tasarut (Amazigh) ‘key’

28It is worth noting that in the derivation of surət, sarut comes in analogy with native words such as fusəx ‘to burn incense to remove evil deed’, from fasux ‘incense’ (see also tfusix ‘use of incense’) and musəq ‘make someone listen to music’, derived from musiqa ‘music’ and related to tmusiq ‘listening to music’. The phenomenon witnessed in the derivation and integration of the Amazigh loan surət, as a basic form regardless of the attached affixes, is known in historical linguistics as folk etymology, whereby speakers change the form of a loanword, as a result of mistaken assumptions, to adapt it to a more familiar form through popular usage (Campbell, 1998). Consequently, nonexistent underlying bases are extracted and added to the lexical stock via the process of backformation (Beard, 1995).

29The facts in (6) provide evidence for two major characteristics: first, the productivity of backformation as a word formation process and second, the existence of the consonantal root as the basic unit carrying meaning and serving as the base of derivation. Yet, some back-formed words can show their grammatical limitation in the sense that some of their forms are not accepted by the native speakers while others are (Katamba, 1994 cited in Staskova, 2013).

30In the next section we examine a non-concatenative process, internal morphological gemination, and claim it to provide further evidence in favor of the extraction of the consonantal root and serve as the base of derivation in MA.

4.3. The extracted consonantal root and internal morphological gemination

32Causatives in MA are formed by doubling (geminating) the second segment of the base form (Boudlal, 2001). Determining the base of derivation of causatives in MA has been one of the intriguing questions for different scholars (McCarthy and Prince, 1986, 1988, 1990; Lombardi and McCarthy, 1991; Bennis, 1992; Boudlal, 2001). Bennis (1992) argued that the stem is the base of derivation of causatives in MA. He claims that the causative is formed through the affixation of a bimoraic syllable to a circumscribed prosodic domain, namely the minimal base. The latter can be of the type CəC or CV. In MA, every morphological causative verb takes the form of one of these two patterns: CəCCəC for sound roots or CəCCV for weak roots as illustrated in (7) and (8) below.

(7) Sound roots (CəCCəC)
Root	Gloss	Causative
√ktb	‘write’	kəttəb
√brd	‘cold’	bərrəd
√nʕs	‘sleep’	nəʕʕəs
√ṣḅṛ	‘be patient’	ṣəḅḅə
√fṛħ	‘be happy’	fəṛṛəħ

(8) Weak roots (CəCCV)
Root	Gloss	Causative
√qṛ(v)	‘read’	qəṛṛi
√bk(v)	‘cry’	bəkki
√d(v)x	‘feel dizzy’	dəwwəx
√d(v)b	‘melt’	dəwwəb
√f(v)q	‘wake up’	fəjjəq

34In the present paper, we argue for a root-based approach of causative derivation in MA. We provide data from causatives derived from loan verbs as evidence. The first type of causatives derived from loans consists of sound roots. Sound roots are defined as roots that are composed of three consonants not including glides, for example √qtl ‘kill’ and √ḍrb ‘hit’ (Bennis 1992).

35In (9) above, the causative forms məzzək and məṇṇək follow the same pattern of causative derivation in MA (CəCCəC). They are derived respectively from the sound roots √mzk and √mṇk. The same roots are derived from the French loan nouns musique ‘music’ and manque ‘missing something’ via the process of backformation. McCarthy (1981) suggests that the link between a ‘derived verb’ and its derivational source is the root. An example adapted from Bennis (1992) is provided in (10):

(10)	Derivational source		root		derived verb
	hḍər ‘talk’		hḍr		həḍḍər ‘make someone talk’

36The second type of causatives derived through backformation consists of verbs derived on the basis of weak roots. Bennis (1992) defines weak roots as roots which contain a median or final vocalic element alternating with a glide. Examples are in (11):

(11)	Root	Gloss	Causative	Origin	/Gloss
	√buṭ	‘kick’	ḅəwwəṭ	botte	‘kick’
	√duʃ	‘shower’	dəwwəʃ	douche	‘shower’
	√ʃik	‘elegant’	ʃəjjək	chique	‘elegant’

37According to Bennis (1992), MA weak roots with median glides, as well as final ones, may be realized as vowels as in:

(12)	dwz	‘to pass’	daz   /   duz	duwwəz	‘make pass’
	ʒrj	‘to run’	ʒra   /   ʒri	ʒərraj	‘runner’

38This hypothesis suggests that the vowels appearing in the surface form are derived from the corresponding underlying glides. Verbs like ʃuf ‘see’ and bki ‘cry’ are respectively derived from the roots √ʃwf and √bkj. A counter hypothesis suggested by one of the reviewers of this paper holds that vowels can be underlying and they change into the corresponding glides to satisfy syllable structure constraints. Our position supports Bennis’s idea; hence, we assume that the surface vowels are underlyingly glides. In (11) above, we suggest, following Bennis (1992), that the hollow roots √buṭ, √duʃ and √ʃik are underlyingly √bwṭ, √dwʃ, and √ʃjk, and they follow the same pattern of causative derivation in MA (CəCCəC). Also, schwa epenthesis applied to break the consonantal cluster disallowed in MA stands as evidence for the consonantal status of the root.

5. Conclusion

39This article provides an account for backformation in loanwords in Moroccan Arabic. It argues that the operation consists in forming a verb by extracting a consonantal root from an integrated loan-noun. We presented data which showed that the extracted verbal root serves as a base for the derivation of other morphological categories (the medio-passive, the passive participle, and the deverbal noun), in analogy with the native patterns of derivation. The data revealed that although the loan nouns contain both vowels and consonants, only the consonants are extracted for the derivation of verbs, in analogy with existing native verbs. Also, the non-Semitic origin of the loan nouns stands as further evidence for a process of root extraction since only the consonants of the word were extracted for the verb derivation. We then examined data from causatives formed through internal morphological gemination of the consonantal root extracted from loan- nouns through backformation. For both sound roots and weak roots, the causative was derived following the MA causative pattern CəCCəC. The internal gemination of the glides -ww- and -jj- provides another argument in favor of the consonantal root in MA, given the fact that vowels cannot be geminated. The insertion of schwa supports the same argument since it serves to break consonantal strings banned in MA. In sum, the presentation and analysis of the loan data extracted via backformation proves, on the one hand, that the root is consonantal in MA, and that it is the base of derivation, on the other. The paper, hence, contributes to the ongoing debate as to whether Semitic morphology is based on the consonantal root, the word, or both. Backformation as root extraction provides strong evidence for a root-based morphology in MA.