1 Introduction

This study deals with the treatment of initial geminates as non-actual surface forms in light of Parallel OT as an analytical framework. It focuses on the case of Qassimi Arabic (QA), a Najdi dialect spoken in the Al-Qassim region in central Saudi Arabia. Before stepping into any further analysis, it is crucial to shed light on the properties of gemination in this dialect for readers who are not familiar with QA or with Arabic in general. Two types of geminates are found in this dialect and are differentiated based on distribution: final and medial. First, final gemination is shown in the following examples:

1. (1)Underlying geminates in QA (Alqahtani 2014; Alrashed 2018):

The examples above show that QA final geminates do not result from assimilation or from any morphological derivation. On a moraic approach, according to Hayes (1989), final geminates in monosyllabic words are dominated by one mora, as in the following representation of /biss/ ‘cat’. (Note that a geminate is represented by a colon after a consonant):¹

(2)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

The same final geminate undergoes degemination when a consonant-initial suffix attached to it (Prochazka 1988), as follows:

1. (3)Degemination of a geminate when consonant-initial suffixes attached to it (Prochazka 1988):

The examples above show that geminates in non-final position are prone to degemination when attached to consonant-initial suffixes in QA, as in the following representation:

(4)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

The same type of geminate is resyllabified into different syllables when attached to a vowel-initial suffix, which results in medial gemination (Alqahtani 2014), as shown in the following examples:

1. (5)Medial gemination with vowel-initial suffixes (Alqahtani 2014):

Two ideas can be drawn from the examples above. First, the second member of a geminate is resyllabified as the onset of the following syllable in order to avoid onsetless syllables, since QA disallows vowel-initial syllables (Al Motairi 2015; Alrashed 2018). Second, since members of the same geminate belong to different syllables, of which the first member is the coda of a non-final syllable while the second member is resyllabified as the onset of the following syllable, as noted, they are no longer dominated together by one mora. Instead, the first member bears the mora, as shown in the following representation:

(6)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

Medial gemination in QA is derived from verb pattern II /faʕʕal/ (Prochazka 1988), as shown in the following examples:

1. (7)Medial gemination derived from verb pattern II /faʕʕal/ (Prochazka 1988):

In light of the above examples, items exhibiting medial gemination, which derive from verb pattern II, are allocated to different syllables and are not dominated together by one mora, as shown in the representation below:²

(8)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

A third strategy for creating medial gemination in QA is through the assimilation of the lateral /l/ in the definite article /ʔal-/ to stem-initial coronals. Youssef (2013) states that medial gemination in the urban-Bedouin dichotomy,³ including QA, as well as in other modern Arabic dialects also results from the assimilation of the lateral /l/ in the definite article /ʔal-/ to stem-initial coronals.⁴ Consider the following examples:

1. (9)Medial gemination derived from the assimilation of the lateral /l/ in the definite article /ʔal-/ to stem-initial coronals (Youssef 2013):

In the examples above, the medial geminates result from the assimilation of the lateral /l/ in the definite article /ʔal-/ to the initial-stem consonants of the [+coronal] feature even though they belong to different syllables and consequently share no mora, as shown in the representation below:⁵

(10)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

This paper aims to interpret the initial geminations in QA as non-actual surface forms, employing the Optimality Theory (OT) model of Parallelism. To achieve this aim, there are two questions that should be addressed: How are initial geminates treated as non-actual surface forms in QA? How can the OT model of Parallelism account for the treatment of initial geminates as non-actual surface forms in QA? The significance of this study is that it demonstrates the conformity of QA to the Obligatory Contour Principle (OCP) and the Strict Layer Hypothesis (SLH) with reference to the treatment of initial geminates as non-actual surface forms, and it shows how this process is accounted for by derivational feeding order as a transparent rule interaction and can be explained using Parallel OT. The next section is devoted to previous studies on gemination in modern Arabic dialects.

2 Qassimi Arabic (QA)

Before examining the phonology of QA in detail, we explain where and by whom this dialect is spoken. While QA is linguistically of a Bedouin type, only the sedentary population speaks it (Al Motairi 2015). Ingham (1994), who works on Najdi Arabic, designates ‘[the] speech of [the] sedentary population' spoken in the Central Najd and Al-Qassim regions and refers to ‘[the] speech of [the] main Bedouin tribes of those regions' (p. 4). Al Motairi (2015) observes that the term Qassimi dialect is exclusively used to refer to the speech of the sedentary population rather than that of the main Bedouin tribes in Al-Qassim. Based on this distinction, the current study concerns itself with the speech of the sedentary population in Al-Qassim.

The Al-Qassim region includes Buraydah, the official capital of the region; Unayzah, the region's second-largest city; and nearby towns, including Almethnab, Al-Bukhairiyah, Baday'a, Riyadh Al-Khabra, Al-Khabra, and Nabhaniya, as shown on the map below (Fig. 1 Al-Qassim region map. This map is cited from the following location: https://www.semanticscholar.org/paper/A-re-examination-of-the-Salicornias-(Amaranthaceae)-Al-Turki-Swarupanandan/2760fc3a663c748a527852e3dca2404551c22c9e/figure/3):

View Full Size

Al-Qassim region map

Citation: Acta Linguistica Academica 67, 3; 10.1556/2062.2020.00018

• Download Figure
• Download figure as PowerPoint slide

The phonology of this dialect is outlined in the following two subsections. The first subsection presents the sound system of QA.

2.1 The consonant and vowel inventory of QA

The inventory of consonants in QA is shown in Table 1 below by manner and place of articulation.

Table 1.

Manner and place of articulation of consonants in QA (Al Turki 2009; Al Motairi 2015)

	Bilabial	Labio-dental	Dentals	Emphatic-dental	Alveolar	Emphatic Alveolar	Alveo-palatal	Palatal	Velar	Uvular	Pharyngeal	Glottal
Plosive	b				t d	tˤ			k ɡ	q		ʔ
Fricative		f	Ө ð	ðˤ	s z	sˤ	ʃ			χ ʁ	ħ ʕ	h
Affricate							ʤ
Nasals	m				n
Lateral					l
Trill					r
Glides	w							j

Al Motairi (2015) states that the short vowels in this dialect are /i/, /a/, and /u/, while the basic long vowels are /iː/, /aː/, and /uː/. In addition, novel long vowels /eː/ and /oː/ are found in non-final position, reflecting the classical and modern Standard Arabic diphthongs /aw/ and /aj/, according to Al Motairi (2015). Both long and short vowels are shown on the vowel chart below (Fig. 2 Vowel chart in QA (Al Motairi 2015)):

View Full Size

Vowel chart in QA

Citation: Acta Linguistica Academica 67, 3; 10.1556/2062.2020.00018

• Download Figure
• Download figure as PowerPoint slide

Additionally, the diphthongs /aw/ and /aj/ are restricted to final open unstressed syllables in QA (Al Motairi 2015), as shown in the examples below:

1. (11)Diphthongs in QA (Al Motairi 2015):

a.	/'qaː.law/ →	['ɡaː.law]	‘they (plural masculine) said’
b.	/'ʔir.qaj/ →	['ʔir.ɡaj]	‘go upstairs!’ (feminine)

This subsection has set out the consonant and vowel inventory of QA, a dialect with 28 consonants. QA has ten vowels in total: three short vowels along with their three long counterparts, two new long vowels, and two diphthongs. The following subsection presents the permitted syllable types, with special attention to their occurrence in initial positions, along with the generalization of obligatory onsets and optional codas.

3 Previous studies

Geminates have been the subject of previous investigations in the phonology of modern Arabic dialects such as Cairene Arabic (Archibald 2003; Haddad 2005), Baghdadi Arabic (Abu-Haidar 1991; Davis & Ragheb 2014), the Al-hasa dialect (Aljumah 2008), Hadhrami Arabic (Bamakhramah 2009), Kuwaiti Arabic (Al-Qenaie 2011), Najdi Arabic (Abboud 1979; Al-Sweel 1987; Prochazka 1988; Ingham 1994; Alghmaiz 2013), Syrian Arabic (Mitchell 1993; Kiparsky 2003), Iraqi Arabic (Erwin 1963; Ghalib 1984; Kiparsky 2003; Watson 2007; Youssef 2013; Davis & Ragheb 2014; Abdul Sattar 2015), Libyan Arabic (Harrama 1993; Elramli 2012), and Moroccan Arabic (Muller 2001; Boudlal 2001; Kiparsky 2003; Watson 2007; Davis & Ragheb 2014). Based on their syllable structures and gemination types, the modern Arabic dialects listed above can be categorized into four groups. The first group consists of Cairene Arabic and the Al-hasa dialect; it has medial and final geminates, while initial geminates are prohibited due to restrictions on initial consonant clusters. The second group, containing Hadhrami Arabic (Bamakhramah 2009) and Baghdadi Arabic (Abu-Haidar 1991; Davis & Ragheb 2014), allows word-medial and word-final geminates even though final consonant clusters are not permitted in either dialect. Moreover, initial geminates are inhibited in Hadhrami Arabic (Bamakhramah 2009) and Baghdadi Arabic (Abu-Haidar 1991; Davis & Ragheb 2014); nevertheless, there is no restriction on initial consonant clusters. The third group, including Kuwaiti Arabic and Najdi Arabic, permits medial and final geminates, whereas initial geminates are prohibited even though there is no restriction on initial consonant clusters. The fourth group, comprising Iraqi Arabic, Syrian Arabic, Libyan Arabic, and Moroccan Arabic, features initial, medial, and final geminates and has no restrictions on initial consonant clusters. These groups are shown in the following table:⁶

1. (12)

With regard to the table above, considering the first group, Archibald (2003) and Haddad (2005) agree on the avoidance of initial geminates in Cairene Arabic since this dialect retains the specific phonological feature from classic Arabic of banning any complex onset. To put it simply, a restriction on Cairene syllable structure prevents this dialect from allowing word-initial clusters or initial geminates. Aljumah (2008) states that complex onsets and initial geminates are not allowed in the Al-Ahsa dialect of eastern Saudi Arabia for the same reason.

The second group, covering Hadhrami Arabic and Baghdadi Arabic, have shown the complexity in the onset only. Medial gemination is found in both dialects as well as final geminates even though final consonant clusters are not found in these dialects. Initial gemination is not reported in these dialects even though there is no restriction on initial consonant clusters in either dialect (Bamakhramah 2009; Abu-Haidar 1991; Davis & Ragheb 2014).

The third group, including Kuwaiti Arabic and Najdi Arabic, have (as shown above) syllables with complexity in the onset as well as in the coda position, meaning there is no restriction on initial and final consonant clusters; complex onsets and codas are found in some syllable types. However, when it comes to initial geminates, a different story unfolds. Al-Qenaie (2011) reports that medial and final geminates are found in Kuwaiti Arabic. In Najdi Arabic, scholars including Abboud (1975), Al-Sweel (1987), Prochazka (1988), and Ingham (1994) agree that both medial and final geminates are available. However, these scholars present only descriptive reports and do not explain why initial geminates do not exist in Najdi Arabic. In contrast, Alghmaiz (2013) states that Najdi Arabic speakers deliberately utilize prosthesis, that is, the insertion of a glottal stop and vowel, to avoid initial geminates rather than using internal epenthesis; for example, /t-darris/→ /ddar.ris/→ [ʔid.dar.ris] ‘you (masculine) teach.’ However, the question of why initial geminates are not tolerated in Najdi Arabic remains and falls within the scope of this study, since QA is one of the Najdi dialects.

As for the fourth group of Modern Arabic dialects—covering Syrian Arabic, Iraqi Arabic, Libyan Arabic, and Moroccan Arabic dialects—there is no restriction on initial and final consonant clusters as reflected in the existence of complex onsets and codas in some syllable types and the permission granted to entire types of gemination, especially initial geminates. For example, Mitchell (1993) and Kiparsky (2003) note that initial geminates in Syrian Arabic are the result of prefixing /n-/ to initial stem consonants with the features [+nasal, +sonorant], e.g., /n-‍midd/→[mmidd] ‘we extend.’ Furthermore, according to Kiparsky (2003), initial geminates in Syrian Arabic can be derived from the assimilation of the prefixes /b-/ and /l-/ to initial stem consonants, as follows:

Similarly, initial geminates in Iraqi Arabic are derived from initial affixation or assimilation of the definite article to the initial stem consonant according to Erwin (1963), Ghalib (1984), Kiparsky (2003), Watson (2007), Youssef (2013), Davis and Ragheb (2014), and Abdul Sattar (2015). Consider the following examples:

1. (14)Initial geminates resulting from initial affixation (Ghalib 1984; Kiparsky 2003):

1. (15)Initial geminates resulting from the assimilation of the definite article /ʔal-/ to initial stem consonants (Ghalib 1984; Kiparsky 2003):

In Al-Jabal Libyan Arabic, Harrama (1993) notes that the prefix /ti-/ is underlying and that initial geminates are derived from two processes: the first process is peculiar in causing the syncope of a vowel in a prefix to prepare an environment for the assimilation of the prefix /t-/ to an initial stem consonant, as shown below:

1. (16)Initial geminates in the Al-Jabal dialect (Libya) (Harrama 1993) as the result of prefix /t/ assimilation to coronal as initial stem consonants:

   a.	/ti+diff/→ /tdiff/→[ddiff] ‘you (masculine) push’
   b.	/ti+tˤiːr/→ /tˤtˤiːr/→ [tˤtˤiːr] ‘it flies’

Initial gemination in the Al-Jabal dialect, according to Harrama (1993), also results from the assimilation of the prefix /n-/ to the sonorants /l/ and /r/ as initial stem consonants:

1. (17)Initial geminates in the Al-Jabal dialect (Libya) (Harrama 1993) as the result of prefix /n/ assimilation to coronal sonorous initial stem consonants:

In Misrata Libyan Arabic, Elramli (2012) indicates that initial geminates are the result of the assimilation of the prefix /t-/ to initial stem consonants with the features [+voice, +coronal, +obstruent], e.g., /t+diːr/→ [ddiːr] ‘you do.’

Muller (2001) shows how initial geminates in Moroccan Arabic, which are permitted, are formed by the complete assimilation of the definite article /l-/ to the following coronal. Consider the following examples:

Likewise, Boudlal (2001), who works on the Moroccan vernacular known as Casablanca Moroccan Arabic, found that it exhibits initial geminates through the assimilation of the definite article to the following coronal consonants. Kiparsky (2003) and Watson (2007) draw upon Muller's (2001) and Boudlal's (2001) work on Moroccan Arabic to investigate syllabification and moracity in this dialect, as well as in other dialects. They agree that initial geminates in Moroccan Arabic derive from assimilation.

Davis and Ragheb (2014) follow Boudlal (2001), Muller (2001), Kiparsky (2003), and Watson (2007) in their approach to initial geminates in Moroccan Arabic. Based on Kiparsky's (2003) work, Davis and Ragheb (2014) put forward two characteristics of initial geminates: first, since initial geminates are derived from assimilation, stress is not attracted onto the syllable containing them, unlike final geminates. Second, Arabic initial geminates are represented by Kiparsky (2003) as solely having moras that are directly linked to the prosodic word node, not the syllable node. In accordance with Kiparsky (2003), Davis and Ragheb (2014) note that initial geminates would add an extra mora to the word, not the syllable.

However, none of the scholars above has addressed the case of initial gemination not surfacing in QA. Therefore, this gap is analysed through the lens of Parallel OT in the present study. Furthermore, the gap in the research that this study fills extends to the role of conformity to the OCP and the SLH in QA in eliminating initial gemination, and to exploring the nature of its phonological derivation using Parallel OT as discussed in the next section.

4 Data collection and analysis

The research method in this study consisted of three main procedures. First, the study relied primarily on data harvested from the literature. Second, I imposed my own Najdi Arabic native-speaker judgements on the data. Third, I interviewed and consulted 15 QA native speakers to verify observations about the QA data.

Initial geminations, which are not the actual surface forms in QA, are the result of the assimilation of the prefix /t-/, with the features of [+coronal, −sonorant, +obstruent], to stem-initial consonants of verbs derived from patterns II and III, as well as hollow verbs with the same features.^7,8 This type of gemination is avoided by prosthesis, as there is a restriction on geminates such that initial ones are illegal, even though QA is one of the dialects that allows complex onsets. Consider the following examples:

1. (19)

   1. Verb pattern II (C₁VC₂C₂VC₃):

2. Verb pattern III (C₁VVC₂VC₃):

3. Hollow verbs (C₁VVC₂):

As shown in the examples above, the assimilation of the prefix /t-/, with the features of [+coronal, −sonorant, +obstruent], to the following stem-initial consonants with the same features aims to avoid a violation of the OCP, and prosthesis is utilized to conform to the SLH. Before further analysis, it is important to undertake a basic discussion of the background to the OCP and SLH. McCarthy (1986) first provided evidence for the OCP as a constraint ‘on the organization of non-prosodic or segmental phonology' (p. 208):

1. (20)Obligatory Contour Principle (OCP) (McCarthy 1986, 208):

At the melodic level, adjacent identical elements are prohibited.

The linearization of the morphological tier by TC yields OCP violation in a particular morphological domain. Therefore, it is important to delete the root node of the left matrix (i.e., the matrix to which the prefix /t-/ is linked) in order to avoid such an OCP violation. Consequently, the vacancy, which results from the deletion of the leftmost matrix, would yield an opportunity for the root node of the contiguous coronal to spread from right to left in order to fill this vacancy, as shown in the representations below:⁹

(21)

a. OCP violation!	b. Avoidance of OCP violation

The processes effect the deletion of the leftmost root node followed by the spread of the right root node to the left, resulting in a geminate sequence. With regard to the schematic representation in (21a), two adjacent non-identical segments are [+coronal], which violates the OCP as a result of TC.

Although QA has no restrictions on initial consonant clusters as seen in 2.2, initial geminates do not not surface in this dialect. This statement leads us to invoke perceptual difficulty as the reason behind the treatment of initial geminates as non-actual surface forms. In fact, Flemming (2004) addresses the difference between perceptual difficulty and articulatory difficulty; hence, articulatory difficulty is considered a property of an individual sound in a particular context since it is related to the effort involved in producing that sound. On the other hand, he infers that perceptual difficulty arises in correctly categorizing sounds. Flemming's (2004) finding is aligned with Galea's (2016) with reference to the perception of initial geminates in Maltese. Galea (2016) reports that his participants perceive word-initial geminates as word-initial singletons, i.e., they could not distinguish between word-initial geminates and word-initial singletons. To put it simply, Galea (2016) observes that word-initial gemination is perceived as being perceptually similar to a word-initial singleton without the preceding vocalic insertion in Maltese. He concludes that initial gemination does not exist in Maltese in either production or perception; hence, word-initial geminates in production always require preceding vocalic insertion, and word-initial geminates without preceding vocalic insertions are perceived as singletons in perception. Similarly, initial geminates require preceding prosthesis in order to be distinguished from singletons by QA speakers while the same type of gemination is perceived as a singleton without prosthesis by those speakers.

To reiterate, the initial geminates in (19) are not tolerated in QA, since the first members are assigned to semisyllables, that is, unsyllabified moraic consonants directly linked to the prosodic word rather than the syllable node in the prosodic hierarchy, as shown below:

1. (22)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

The peripheral member of an initial geminate in the representation above fails to conform to the SLH (Selkirk 1984; Nespor & Vogel 1986; Itô 1986), which ‘requires that every non-highest prosodic element be in its entirety a constituent belonging to the next highest category on the prosodic hierarchy' (Rakhieh 2009, 175). According to Selkirk (1984) and Nespor and Vogel (1986), the prosodic hierarchy is a theory in which words and phrases may be parsed into prosodic constituents that form the domains of rule application.

(23) Prosodic hierarchy:¹⁰

The avoidance of this type of gemination motivates prosthesis in order to affiliate the moraic consonant, a peripheral member of an initial geminate as an initial semisyllable, to the syllable node and to comply with the SLH. Consider the following representations:

1. (24)

View Full Size

• Download Figure
• Download figure as PowerPoint slide

Three things must happen to arrive at the surface forms in (19). These happen in parallel, but each one feeds the next, as shown below. The assimilation of the prefix /t-/ to initial stem consonants with the same features, i.e., [+coronal, −sonorant, +obstruent], which is done to avoid violation of the OCP, feeds the insertion of [i] to avoid an initial semisyllable, meeting the SLH requirement, and the insertion of [i], in turn, initially feeds the insertion of the epenthetic glottal to avoid onsetless syllables, conforming with the Onset Principle (Itô 1989). This form of transparent rule interaction is known as feeding. This type of phonological derivation with feeding is shown below:

1. (25)Feeding order:

The feeding order above, as a transparent rule interaction, is accounted for using Parallel OT. This model of OT can easily account for transparent rule interactions, including feeding (Rakhieh 2009). Consider the following constraints:

1. (26)Relevant constraints:

2. b. O-CONTIG (CONTIGUITY-IO) (McCarthy & Prince 1995):

   The portion of S2 standing in correspondence forms a contiguous string. (No internal insertion).

3. c. MAX (McCarthy & Prince 1995):

   Every segment of S1 has a correspondent in S2. (No deletion).

4. d. OCP [+coronal, −sonorant, +obstruent]:

   Adjacent [+coronal, −sonorant, +obstruent] consonants are prohibited unless they are identical.

5. e. ONSET (ONS) (Prince & Smolensky 1993):

   Syllables must have onsets.

6. f. IDENT (McCarthy & Prince 1995):

   If a segment is specified as [αF] in the input, then it must be specified as [αF] in the output.

In light of the above constraints, the tableau below is intended to determine the optimal candidate for the input /t-dʒah.hiz/:

(27)

Candidate (c), as the wrong output, has been chosen as optimal since it avoids the violation of most constraints above compared to the rest of the candidates.¹¹ Candidates (a) and (g) violate the OCP [+coronal, −sonorant, +obstruent] and are therefore excluded. The internal epenthesis is permitted by candidates (b) and (h) to comply with the OCP [+coronal, −sonorant, +obstruent] constraint which, on the other hand, incurs the violation of the O-CONTIG constraint. Therefore, both candidates are eliminated. Candidate (d) conforms to the OCP [+coronal, −sonorant, +⁠obstruent] constraint through the deletion of the prefix /t-/, which leads to the violation of the MAX constraint. The initial epenthetic vowel [i] in candidate (f) complies with the OCP [+coronal, −sonorant, +obstruent], O-CONTIG, and MAX constraints, but it yields an onsetless syllable which violates the ONS constraint. Candidate (e), as the desired output, violates the DEP constraint twice and is thus not eligible to be optimal. Therefore, there should be a constraint which helps to eliminate candidate (c) and determine candidate (e) as optimal. Paying attention to candidate (e), there is an initial semisyllable, i.e., an unsyllabified moraic consonant, which violates the SLH. Based on this observation, the OT constraint below militates against moraic and non-moraic segments that are not affiliated to the syllable node as follows:

1. (28)*C^unsyll (McCarthy 2008):

This constraint prohibits unsyllabified consonants.

The above constraint outranks DEP in order to identify candidate (e) as optimal in the next tableau:

1. (29)

The tableau above shows that the unsyllabified moraic consonant against the SLH in candidate (c) violates the *C^unsyll constraint. As a result, this candidate fails to be optimal, while the same constraint helps to distinguish candidate (e) as optimal.

What this section shows is that initial geminates, which are not the actual surface forms in QA, originate from the assimilation of the prefix /t-/ to stem-initial consonants of verbs derived from the verb patterns II and III, as well as hollow verbs. This assimilation arises to avoid violation of the OCP. However, this type of gemination yields an initial semisyllable—an unsyllabified moraic consonant—which violates the SLH. Therefore, prosthesis is used to affiliate an initial semisyllable to the syllable node. Put simply, this process is accomplished by the feeding relationship order. First, initial gemination, which results from the assimilation of the prefix /t-/ to initial stem consonants of the same features, i.e., [+coronal, −sonorant, +obstruent], facilitates the application of the prosthetic vowel [i], providing an environment for the application of the prosthetic glottal stop, that is, for glottal epenthesis. Thus, Parallel OT is shown in this section to be capable of accounting for the feeding relationship as a transparent rule interaction.

5 Conclusion

This research has addressed initial geminates in QA, treating them as non-actual surface forms derivationally created by the assimilation of the prefix /t-/ to stem-initial consonants found in verbs of patterns II and III and hollow verbs. Conversely, initial geminates are prohibited in this dialect, since they result in a semisyllable as a moraic consonant, which violates the SLH. As a result, prosthesis is used to affiliate initial semisyllables to the syllable node to accomplish conformity to the SLH. In other words, based on the treatment of initial geminates as non-actual surface forms in QA, conformity to the OCP and SLH in QA appears to be achieved by a feeding relationship. First, the assimilation of the prefix /t-/ to the following stem-initial consonants with the same features, i.e., [+coronal, −sonorant, +obstruent], results in an initial geminate with a peripheral semisyllable that motivates the prosthetic vowel [i] in order to comply with the SLH. An onsetless syllable, which results from the prosthetic vowel [i], motivates the prosthetic glottal stop in order to abide by the Onset Principle (Itô 1989). Thus, Parallel OT is shown here to be capable of accounting for the feeding relationship addressed in this paper.