Subdivisions of Faifi Arabic: a perceptual dialectology approach

The study addresses the under-documented status of Faifi Arabic, a cluster of dialects spoken in the Faifa Mountains in southwestern Saudi Arabia. Despite distinctive phonological, morphological, and syntactic features and historical isolation, Faifi Arabic has received limited scholarly attention compared with other Arabic dialects. Prior work often divided Faifi Arabic into upper and lower (and sometimes western/central/eastern) varieties based on elevation or selected features, but conflicting descriptions suggest a more complex internal structure. The purpose of this study is to identify and classify subvarieties of Faifi Arabic using a perceptual dialectology experiment that elicits native speaker perceptions of key (morpho)phonological features and their geographic distribution. The research is important for clarifying contradictory accounts in the literature, advancing Arabic dialectology, and illuminating how local tribal geography may relate to dialect boundaries.

The paper situates its work within two strands of literature: (1) descriptive studies of Faifi Arabic and (2) perceptual dialectology. Prior Faifi studies reported heterogeneous realizations for several features: definite morpheme /m-/ as [ʔim-] vs [m-]; allophones of /dʒ/ variously described as [dz], [z], [ʒ]; emphatic fricatives /ð/ (cognate with MSA /ð/ and /d/) realized as [θ], [ð], [δ-like], [f], with conditioning by position (root-initial vs non-root-initial) and vowel context; and reported regional partitions such as upper/lower or western/central/eastern (e.g., Alfaifi 2009, 2016, 2020; Alfaife 2018; Alfaifi & Behnstedt 2010; Davis & Alfaifi 2019; Alfaifi & Davis 2021, 2022). These varying descriptions point to multiple subdialects rather than a simple binary/ternary split. In perceptual dialectology, research maps folk knowledge of dialect regions, correctness/pleasantness judgments, and perceived distances (Preston 1989, 1999; Long & Preston 2002; Cramer & Montgomery 2016). In the Arabic-speaking world, perceptual dialectology has been used to locate perceived dialect areas and attitudes (Theodoropoulou & Tyler 2014; Hachimi 2015; Abdel-Rahman 2016; Al-Rojaie 2021a,b). The current study applies a feature-based perceptual approach to reconcile prior conflicting descriptions and to assess speakers’ awareness of diagnostic features in Faifi Arabic.

Design: A perceptual dialectology experiment with two tasks: (1) a feature-grouping task using 60 audio-recorded words to group features into perceived dialect categories (columns labeled Dialect 1–5), and (2) a mapping task where respondents placed group numbers onto a 2D map of the Faifa Mountains. Sampling: 30 participants (age 20–64; mean ≈ 43.6; median 45), all native speakers of Faifi Arabic and current residents of the Faifa Mountains, recruited via local community networks to cover diverse areas. Materials:

• Feature-grouping form: contained transcriptions, brief descriptions of each audio file, and five columns for grouping features into dialects. Example items explicitly highlighted target contrasts (e.g., [f] instead of [ḍ] in the word ‘white’).
• Map: a 2D map of the Faifa Mountains divided into 122 blank squares, with major roads and three macro-regions marked for orientation.
• Audio: 60 items recorded by a single native-speaker linguist to control for non-target variation; validated for accuracy by three additional native speakers. Item sets: 1–8 emphatic /ḍ/ cognates with realizations [ð], [θ], [q], [f]; 9–16 emphatic /ð/ cognates with the same alternatives; 17–28 definite article /m-/ allomorphy ([ʔim-] vs [m-]); 29–38 high short vowel deletion in initial open unstressed syllables; 39–56 /dʒ/ and reported allophones [dz], [z], [ʒ], with and without phoneme /z/ in the word; 57–60 /k/ as [k] vs [ʃ]. Some later items required additional explanation during sessions due to subtler perceived differences. Procedure:

1. Orientation: respondents listened to all 60 files to familiarize with features.
2. Feature grouping: each item assigned to one or more dialect columns (1–5) or left unassigned if deemed absent; repeated for all items.
3. Mapping: respondents labeled squares on the 122-cell map with the dialect group numbers derived from step 2; multiple group numbers per square were allowed.
4. Post hoc tribal mapping: a 64-year-old lifelong resident consultant labeled tribal regions on the same blank map to relate dialect areas to tribal geography. Analysis:

• Feature-grouping forms were extracted into spreadsheets per participant and merged to identify consistent feature groups. Most respondents recognized at least four groups; two recognized a fifth; one recognized only three.
• Maps: each of the 122 squares was uniquely indexed. For each participant, a 26-column spreadsheet (one per analyzed feature) recorded which group(s) were assigned to each square. All participant sheets were merged to produce counts per square per feature. Rows were reorganized to mirror the map layout, enabling the creation of 26 heat maps to visualize spatial patterns and boundaries.

• Respondents consistently identified four distinct subdialects within the Faifa Mountains, with clear geographic boundaries. The divisions align with western, southern, eastern, and central regions rather than the previously posited upper vs lower (or simple western/central/eastern) split.
• Definite article /m-/ and high vowel deletion: Dialects using [m-] also display deletion of high short vowels in initial open unstressed syllables; these cluster in central, southern, and parts of central-northern areas. Dialects using [ʔim-] retain the high vowel and occur predominantly in western, eastern, and a small central-southern corridor. The phonological pairing suggests a shared preference for avoiding complex onsets in [ʔim-] dialects (epenthesis and vowel retention) versus greater tolerance for onset clusters in [m-] dialects (no epenthesis and vowel deletion).
• /dʒ/ allophones and conditioning by /z/: • With /z/ present in the word: [ʒ] is associated with the eastern region; [dʒ] with the central region; another realization mapped to western and southern regions. Overall, three regional patterns emerged. • Without /z/: [ʒ] remains in the east; [dʒ] predominates elsewhere, with some peripheral occurrences at northern, western, and southern borders. In the western and southern regions, a fricative realization is used for /dʒ/ both with and without /z/; the eastern region consistently uses [ʒ] regardless of /z/.
• /k/ allophones: Respondents showed a categorical pattern tied to the presence of other sibilants: [k] was used when the word contained /ʃ/, while [ʃ] was used when the word contained /s/.
• Emphatic /ð/ (cognate with MSA /ð/ and /d/), with positional effects: • Root-initial /ð/ cognate with MSA /d/: [f] dominates in central and western regions; [ð] in the east; [θ] in the south. • Non-root-initial /ð/ cognate with MSA /d/: central/east favor [ð]; south shows [θ]; west shows [f]. • Non-root-initial /ð/ cognate with MSA /ð/: respondents primarily recognized two realizations, with [ð]-type variants more common in central/west and another [ð]-variant in southern/northern areas.
• Integrating perceptual maps with the consultant’s tribal map suggests a correlation between subdialect divisions and tribal geography. However, closely related tribes can differ linguistically, while more distantly related tribes may share features if geographically proximate. Geographic contiguity appears to be a stronger predictor of dialect similarity than lineage alone.

The perceptual grouping and mapping tasks revealed four geographically bounded Faifi subdialects and clarified why prior studies provided seemingly contradictory descriptions of features such as the definite /m-/ allomorphy, /dʒ/ allophony, and emphatic /ð/ realizations. Rather than a simple binary (upper/lower) or ternary partition, Faifi Arabic shows at least four subdialect regions, each characterized by coherent bundles of phonological traits. The consistent pairing of [ʔim-] with retention of initial high vowels versus [m-] with initial high-vowel deletion indicates underlying phonological strategies aligned with syllable structure preferences. The regional conditioning of /dʒ/ and /ð/ realizations, and the categorical pattern for /k/ relative to co-occurring sibilants, further supports the presence of distinct dialect areas. Comparison with tribal geography suggests that present-day spatial distribution of tribes correlates with dialect divisions more strongly than tribal lineage, highlighting geography and contact as key drivers of phonological patterning. The findings demonstrate that native speaker perceptions can effectively resolve literature inconsistencies by revealing multiple coexisting subdialects and by pinpointing their perceived geographic extents.

The study introduces a feature-based perceptual dialectology approach to classify Faifi Arabic into four subdialects with clear geographic boundaries across the Faifa Mountains. It reconciles conflicting descriptions in prior literature by showing that reported feature differences reflect distinct subdialects rather than a single uniform system. The work contributes methodological value by combining controlled audio stimuli, feature-grouping forms, and perceptual maps, and by relating perceptual dialect areas to tribal geography. Future research should expand the participant pool (including women), diversify speakers in recordings, increase lexical items per feature (avoiding recent loanwords where possible), consult multiple geographical/tribal experts, and incorporate ethnographically informed interviews to probe why perceptions differ across individuals. Such extensions will help refine the mapping of subdialects and deepen understanding of how geography, social history, and contact shape variation in Faifi Arabic.

• Sample size limited to 30 participants from a relatively small region; absence of female participants due to local cultural constraints.
• Audio stimuli recorded by a single speaker; some features tested with only one lexical item, potentially limiting generalizability.
• Possible influence of loanwords (e.g., mahfaða ‘wallet’) on perceived realizations, which might not reflect local lexicon patterns.
• Reliance on a single consultant for tribal geography; broader triangulation with multiple consultants and ethnographic methods was beyond scope.
• The study did not probe reasons for individual perceptual differences (e.g., age, education, mobility); future work should include interviews and a more diverse sample.