Word Stress in Qassimi Arabic: A Constraint-Based Analysis

This study aimed to produce a formalism of word stress in Qassimi Arabic (QA), which is a sub-dialect of Najdi Arabic (NA), using a constraint-based approach. To this end, this paper investigated two main topics: The first topic explored word stress in QA. Word stress in QA, as well as in NA, is predictable; it can be determined by syllable weight and position. However, two cases do not conform to such straightforward stress rules. These cases are represented by the words: [ʔal.ʕa.sʕir] ‘afternoon’ and [ʔa.ʕa.rif] ‘I know’. Derivational analysis of these exceptions shows the importance of relating the surface structures of such forms to their underlying representations. The second topic aimed to make a formalism for stress patterns in QA using optimality theory (OT). Thus, QA word stress rules and their exceptions are translated into conflicting constraints that are ranked relative to one another by the use of constraint-relation tableaux. This ranking eventually produced the following constraint-relation hierarchy: Lx≈Pr, SYLLABLE-INTEGRITY, TROCHAIC, FAITH-PK >> NONFINAL >> *[ʔa. >> FTBIN-μ, WSP, ALL-FEET-RIGHT >> MAIN-RIGHT, PARSE-σ.


Introduction
OT, as introduced by Prince and Smolensky (2004) and McCarthy and Prince (1993), is a recent grammatical framework that considers surface forms to determine the "optimal" or most harmonic form that best satisfies a set of constraints. Like in previous derivational models--Generative Grammar (Chomsky, 1965) and (Chomsky & Halle, 1968): Principles and Parameters (Hayes, 1980)--there are a set of constraints that need to be satisfied for a particular structure to be considered well-formed. However, the substantial difference in this approach lies in that these constraints can be violated while we can still determine an output form among other forms that is well-formed or optimal.
One of the fundamental tenets of OT is that constraints intrinsically clash with one another. It then follows that to obtain the actual surface form, one of these constraints will be inevitably violated while the other is satisfied. Nevertheless, it is crucially important to decide on the constraint where the violation of which is less harmful, i.e., violating some constraints is more "serious" and could lead to ungrammaticality compared to violating other constraints. For this reason, constraints must be ranked hierarchically: higher-ranked constraints take priority over lower-ranked constraints. Eventually, a constraint-relation hierarchy is produced to define some particular linguistic feature. 1) Stress falls on one of the last three syllables. Stress falls on the ultimate syllable if it is superheavy.
2) Otherwise, stress falls on the penultimate syllable if it is heavy.
3) Otherwise, stress falls on the ante-penultimate. 4) In disyllabic words: stress falls on the ultimate syllable if it is superheavy, otherwise stress falls on the first syllable. 5) Monosyllabic words must receive stress.
Nonetheless, there are two exceptions to these rules that are worth investigating. The first is found in most NA dialects, while the second is a particularity of QA. These exceptions are illustrated next.
3.1 Case 1: [ʔal.ʕa.sˤir] Consider the following set of words: This set has the structure CVC.CV.CVC, which, according to NASRs should have antepenultimate stress since the ultimate is not superheavy and the penult is light. However, stress lodges on the penult. It is clear that this whole set is prefixed with the definite article 'ʔal'. If we try to morphologically formalize a rule for this set by assuming that forms prefixed with the definite article 'ʔal' have penultimate stress, then how can we account for stress in the following forms? These forms have initial stress conforming to NASRs. The fact that there are identical ʔal.CV.CVC structures but alternating stress positions makes it inappropriate to make generalizations at the surface level. Moreover, since the definite article 'ʔal' is stressed in some cases and destressed in others, we might have to resort to phonological rather than morphological derivation. Let us now consider the underlying forms from which the forms in (3.1) are derived: a. /ʔal.ʕasˤr/ b. /ʔal.wazn/ c. /ʔal.ʕumr/ d. /ʔal.bazr/ The underlying forms are structured CVC.CVCC where the final superheavy syllable has undergone epenthesis because the final coda cluster does not comply with the Sonority Sequencing Principle (Roca, 1994), i.e., the first consonant is less sonorous than the second consonant. The resulting structure is, therefore, CVC.CV.CVC. The question that arises now is how underlying forms can serve in deciding the stress positions of surface forms? It is already determined that the underlying CVC.CVCC has final stress. What happened here is that epenthesis has taken place, and the superheavy syllable CVCC is broken by epenthesis into CV.CVC. Nonetheless, the stressed vowel in the underlying form seems to preserve stress in the surface form. The first analysis of this was by Al-Mozainy (1981) in his analysis of BHA. Consider the steps from which the surface form [ʔal.ʕa.sˤur] in BHA is derived (Al-Mozainy, 1981, p. 137 To conclude, despite the structural change that such forms have undergone via epenthesis, it appears that the stressed vowel in the underlying form preserves stress during the process of derivation.

Case 2: [ʔa.ʕa.rif]
Consider the following set of words: ( According to NASRs, the stress in CV.CV.CVC structures should fall on the antepenultimate syllable since neither the ultimate is superheavy nor the penult is heavy. On the contrary, the word set in (3.4) receives stress in the light penult. Furthermore, the syllable structure of this set is CV.CV.CVC, which is not preferable as it consists of successive light syllables (Alqahtani, 2014), and the first vowel should have undergone tri-syllabic elision. Therefore, this set contradicts stress rules as well as syllabification 'vowel deletion' rules. Now, two questions regarding stress and syllabification of these structures must be addressed: How this syllable sequence is permitted in these words? And why does the penultimate syllable receive stress? First, let us consider the underlying forms from which these words are derived: To answer the first question, we notice that the underlying forms represent the syllable structure CVC.CVC. However, guttural consonants are not allowed in a coda position, and therefore, a resyllabification process takes place: Abboud (1979) has demonstrated that the initial syllables of coda guttural undergo two resyllabification processes. Consider, for instance, how the surface form [lħa.mih] 'meat' is derived: /laħ.mih/ /la.ħa.mih/ low vowel insertion after the guttural "epenthesis" /lħa.mih/ initial vowel deletion [lħa.mih] This process represents that surface forms like [lħa.mih] appear with the initial cluster due to vowel deletion. However, we notice that the surface forms in (3.4) do not have this initial consonant cluster, which suggests that another process has also taken place underlyingly to prevent a complex onset. Note that each word in the list has the structure /ʔaC/ as its first syllable. This indicates that the initial consonant cluster of initial syllables of the form /ʔC/ is not allowed. In fact, Al-Mozainy (1981) has introduced the rule of initial vowel epenthesis: This shows that after the application of stress rule '3', the underlyingly stressed vowel is deleted due to LVD. After that, stress is reassigned by the next stress rule (stress rule '4'). Finally, a vowel is epenthesized in the environment #ʔ_C (Oh, 1998, p. 19).
It appears that the analysis of Oh (1998) for such forms is satisfactory for two reasons: First that the underlying forms have identical syllable structures to the surface forms, and second, that the stress rules are applied in their expected order. Thus, if we adopt the analysis of Oh (1998), we would rather assume that these forms are assigned stress on the surface. Moreover, word stress in such forms is phonological rather than morphological: in (3.4) /ʔa /is a first-person prefix, while in (3.6) it is part of the stem.

OT Analysis of QA Stress
Based on the above analysis of QA stress, in this section, we will attempt to provide a formalism for QA stress using the OT framework. In other words, NASRs, as well as exceptions to these rules, will be translated into a constraint-relation hierarchy.

Data
In the following dataset, syllabification and stress positioning are assigned by the author, who is a native speaker of QA. Furthermore, two native speakers of QA who also study linguistics have verified them. In addition, 10 educated native speakers who were born and brought up in Qassim region were asked to produce words from SA to determine how QA speakers realize prosodic features that cannot be decided through QA word structures.

Constraints
Here, we list the most related prosodic constraints that have been referred to in the literature: WSP: Heavy syllables must be stressed (Prince, 1990).
MAIN-RIGHT/LEFT: Align the head-foot with the word, on the right/left edge (Tesar, 1996).
IAMBIC/TROCHAIC: Align the head-syllable with its foot on the right/left edge (Tesar, 1996).

SYLLABLE-INTEGRITY:
The contents of a syllable may not be divided between two feet (Prince, 1976).
For clarity, the related constraints are divided into subcategories. Then, the constraints of one subcategory will be discussed before we proceed into the other subcategory; this gradually builds up our hierarchy as we proceed from one constraint to another. Head categories are taken from Al-Mohanna (2004)  Candidates 'b' and 'c' forfeit optimality since they incur a serious violation to SYLLABLE-INTEGRITY. Such violation becomes also disastrous to foot formation as foot boundaries must coincide with syllable boundaries. Thus, we tentatively assume that SYLLABLE-INTEGRITY is undominated, and therefore, is housed in the undominated stratum: Stratum 1: Lx≈Pr, SYLLABLE-INTEGRITY

Boundedness
"PARSE" constraints were first introduced by McCarthy and Prince (1993a) who demonstrated that when a given element is parsed, it is dominated by an appropriate node in the prosodic tree. Our attention is confined to the constraint that enforces the parsing of syllables into feet, i.e., PARSE-σ. Moreover, the focus will be on foot binarity as an essential foot form in bounded systems, represented by FTBIN.

PARSE-σ
PARSE-σ requires every syllable in PrWd to be part of a foot. That is, if this constraint is undominated, all underlying syllables of the PrWd would be exhaustively parsed into feet. Consider how PARSE-σ evaluates the candidate outputs for /mis.taʕ.mal/ in the following table: [mis.(taʕ).mal] ** d.
[mis.taʕ.mal] !*** Note that the number of violations for candidate 'd' becomes fatal since it renders the entire word structure unparsed, which entails that the word has no prosody at all, thus also incurring a serious violation of Lx≈Pr. Note also that we do not assign the pointing hand (that signals optimality) to candidate 'a' because this output is the potential rather than the actual surface form, as we will see, PARSE-σ must be dominated by NONFINAL to disallow the parsing of final non-superheavy syllables.

FOOT BINARITY (FTBIN)
Feet are binary under syllabic or moraic analysis (McCarthy & Prince, 1993a). Prince (1980), Prince and Smolensky (2004), Hayes (1980), McCarthy and Prince (1986, 1990, 1993a, 1993, and others highlighted that an unmarked foot structure represents a binary foot, thus, in metrical phonology, feet are strictly and maximally branched binarily. Under this restriction, the minimal foot consists of two syllables when FTBIN is subjected to syllabic analysis or two moras under moraic analysis. Therefore, for our hierarchy to yield a correct stress pattern, it is crucially important that we decide whether QA feet are constructed over syllables or moras.
Assuming, first, that feet are constructed over syllables in an undiscriminated weight (see foot inventory in Hayes (1995)), and given that feet have initial prominence, i.e., trochaic, this syllabic trochee can successfully account for the stress in tri-syllabic words with light penultimate (Angled brackets "<>" indicate extrametricality of the final syllable:  In addition Ranking 2 In addition we tentativ look like th Hierarchy Table 7 de candidate Table 7. Lx The questi final nonconsider T [(mak c. [mak.
to be the opti in fact, problem the constraint herefore, since rse ranking: FT It is also monomora dominance  As stress is placed on the antepenultimate syllable; the feet are, therefore, parsed binarily from right to left, since if parsing was performed the other way round, i.e., from left to right, these words would have a different stress pattern: (4.11) a. *(mam).(la.ka).tu.<na> b. *(mak).(ta.ba).tu.<ki> c. *(ʔad).(wi.ya).tu.<ka> Thus, this can be taken as evidence that ALL-FEET-RIGHT dominates ALL-FEET-LEFT. Thus, we have ranking '13': Ranking 13: ALL-FEET-RIGHT >> ALL-FEET-LEFT Nonetheless, note that the right-most foot in the previous word set incurs one violation to ALL-FEET-RIGHT since they do not occur at word right-most edge; this is due to the undominated constraint NONFINAL: Table 18. NONFINAL >> ALL-FEET-RIGHT input NONFINAL ALL-FEET-RIGHT a.
(ʔad).wi.(ya.tu).ka *,**** It is more important then, to leave the final syllable unparsed than to align feet at the right edge. ALL-FEET-RIGHT is, therefore, made subordinate: Ranking 14: NONFINAL >> ALL-FEET-RIGHT Therefore ALL-FEET-RIGHT is set directly below NONFINAL in our hierarchy: The notion of extrametricality (Liberman & Prince, 1977) indicates that a particular prosodic constituent is designated as invisible for purposes of stress rule application. This invisibility entails that this constituent is not parsed into a foot, or in other words, it cannot form a prosodic head. In OT terms, extrametricality is translated into two constraints, namely, NONFINAL and NONINITIAL. Usually, the unparsed constituent is the one at the word's right-most edge, therefore, the unmarked phonological structure corresponds to the constraint NONFINAL: 'a' and 'b' are disyllabic words that have initial stress due to the domination of NONFINAL. On the other hand, 'c' is a tri-syllabic word with initial stress due to the interaction of NONFINAL and FTBIN-µ. This implies that NONINITIAL is a constraint with a less significant effect, and therefore, is dominated by NONFINAL and FTBIN-µ: Ranking 15: NONFINAL >> FTBIN-µ >> NONINITIAL However, to the best of the researcher's knowledge, there is only one exceptional case where stress is, contrary to expectations, strictly banned from falling on the initial syllable. This case has been elaborated in (Subsection 3.2). In OT terms, we introduce here an argument for this exceptional case to NASRs and consider how NONINITIAL affects its stress pattern.

NONINITIAL
In Section (3.2) we posit that some surface forms with a CV.CV.CVC structure do exist, however, with noninitial prominence. As pointed out above, there are two underlying forms for such a structure: /ʔaC.CVC/: /ʔaχ.dim/ ijel.ccsenet. and /CV.C processes. preservatio preserved these struc structure t they begin attempt to *[ʔa.

Initial ligh
By stating this constr that TROC have the f initial sylla  The argument presented above provides evidence that NONFINAL is applied at the syllable level rather than the consonant level. Nonetheless, under this assumption, stressed final superheavy syllables appear to contradict this argument, since if NONFINAL is undominated, final superheavies would never be able to attract stress: In terms of nonfinality, previous work has attempted to find a way to distinguish between parsing final superheavy and non-superheavy syllables in terms of constraint relativization: in Al-Jarrah (2002), for example, NONFINAL is relativized into NONFINAL(σ µµ ), which bans stressing syllables of two mora or less, and NONFINAL(σ µµµ ), which bans stressing syllables of three moras. By this relativization, the following domination relation is defined: On the one hand, this means that since NONFINAL(σ µµ ) dominates PARSE-σ, any final non-superheavy syllable is not parsed. On the other hand, since PARSE-σ dominates NONFINAL(σ µµµ ), it is indicated that the ban against stressing final trimoraic syllables is violated, so that final superheavy syllables are parsed.
Another attempt by AlDweikat (2013) who, instead of relativizing NONFINAL, relativized PARSE-σ into undominated PARSE-σ µµµ that forces parsing any trimoraic syllable at any word position, and a lower-ranked PARSE-σ µ(µ) which is dominated by NONFINAL to ban stressing final non-superheavy syllables. Thus, the following domination relation is established: The question that promptly arises here is do we have to relativize NONFINAL as in Al-Jarrah (2002), so that the violable NONFINAL(σ µµµ ) would allow final stress to superheavies--or do we have to relativize PARSE-σ as in AlDweikat (2013)

Conclusion
This study aimed to employ OT to account for stress patterns in QA. As a subdialect of NA, NA stress patterns were presented, and two exceptional cases were illustrated: the first case, represented in forms like [ʔal.ʕa.sˤir], the analysis instantiates the significance of the finding that stress in underlying forms is preserved. The second case, represented in forms like [ʔa.ʕa.rif], demonstrated that the initial syllables of the form [ʔa. in tri-syllabic words are not allowed to host stress.
After elaborating on stress patterns in QA, the constraints that are most related to QA word stress were stated, and some definitional decisions were made before proceeding to the constraint-ranking process. For instance, it was demonstrated that feet are binary under moraic rather than syllabic analysis. Also, extrametricality was interpreted at the syllabic rather than the consonantal level. Furthermore, the final superheavy syllable consists of two syllables: a biomoraic syllable plus a degenerate one, the latter, as the peripheral element is that which is deemed extrametrical. By such analysis, any relativization to NONFINAL and PARSE-σ can be avoided.
Then, in terms of constraint interaction, constraints are ranked hierarchically one after another until the following constraint-relation hierarchy was produced: Hierarchy 10