This section summarizes the major findings concerning the syllable shapes and numbers across age groups. A further sub-section is devoted to explaining the phonological repair processes (prosodic and melodic) involved in the production and development of syllable structures produced by typical JA children.

The JA children adhered to the universally structured syllable that started as CV and at a later stage became CVC, as opposed to what they are exposed to in their input CVCC structure (Daana, 2009). They produced a variety of syllable types to build their own words. The older age group produced the least-used syllable shape CV. Intriguingly, monosyllabic words take CVV, CVVC, or CVG forms across all age bands (the oldest age group also has CCVVC, and CCVCC). This confirms that JA children develop the ability to maintain the bimoraic weight parameter, which rises across age bands. This dictates that the moraic weight of a prosodic word in JA must minimally comprise two moras (for details on syllable weight in JA, refer to Huneety & Mashaqba, 2016; Mashaqba & Huneety, 2018).

In terms of syllable shape across age groups, the oldest children were able to produce a greater variety of open and closed syllable shapes than the younger groups. Children in the youngest age group, on the other hand, used CV, CVC and CVV syllable shapes in the majority of their productions, similar to the pattern of English-speaking children, as these syllable shapes are usually found in children's late babbling stage and early word production (Locke, 1983).

According to the corpus of the present study, the first attempt to produce a final coda cluster (other than geminate consonants) in a monosyllabic word was registered by one child in the oldest age group (Table 1). This observation indicates that children are able to produce coda clusters at an early age. However, our results do not agree with those of French (1989), who reported that his son produced the first consonant cluster word at age 1;10 and five other words by 2;2. The emergence of the production of clusters in mono-syllabic words in the oldest age group might mark a shift from cluster reduction to production and might be setting the stage for the production of syllable clusters in multisyllabic words. This would explain why the least-used syllable shapes (0.2%) were those containing consonant clusters like CCVCC, CCV and CCVV. This result supports the assumption that consonant clusters emerge in older children. This important result suggests that children can significantly increase the production of consonant clusters after the age of 2;7, with maturity of their speech mechanisms and continuous development of the vocal tract. Future studies on older Arabic-speaking children might consider looking at the trajectory of consonant cluster production in children beyond the age of 3. For example, Ammani Arabic children acquire consonant clusters in both positions at 3 (Daana, 2009).

In adult-speech of JA, clusters are usually produced frequently, but the children of this age in the current study could not produce them correctly. Consequently, they often deleted a sound to simplify the cluster. For example, word-initial CCV and word-final CVCC syllables were simplified to CV and CVC respectively. Although Dyson and Amayreh (2000) did not provide evidence of cluster reduction in their study, this might be explained by limitations in their elicitation protocol; the picture-description task which elicited single words did not contain consonant clusters. Nonetheless, they did find consonant sequence reduction. Al-Tamimi et al. (2011) found no evidence for cluster reduction in their sample of children with cleft lip and palate; this may be because the data scoring protocol they used in their study relied on the occurrence of at least 20% of a phonological process for consideration in the data analysis. Another explanation of this discrepancy might be the severe limitations in the speech of their children with cleft lip and palate. However, our data is in agreement with the findings of Grunwell (1987), who found that cluster reduction is noticeably reduced after the age of 3. It should be noted that children tend to produce the second consonant in the cluster when it occurs in the coda, regardless of sonority; for example /kalb.na/ ‘our dog’ to [kab.na] and /ʔ ird.na/ ‘our monkey’ to [ʔ id.na]. Similar findings were reported in Cairene Arabic (Ragheb & Davis, 2010). However, these findings disagree with those of Abdoh (2011), which can be attributed to the limited data in the speech of her participants (i.e., only one example of this process was found in her sample).

As for the phonological processes employed by children, the acquisition of clusters is one of the longest-lasting aspects of speech acquisition by normally developing children (Mcleod et al., 2001, p. 99). Children as young as 2 produce some consonant clusters correctly (e.g., Preisser et al., 1988; Stoel-Gammon, 1987; Watson & Scukanec, 1997). Naturally, children tend to delete a consonant in consonant clusters; such a conclusion is also attested to in previous investigations (e.g., Bernhardt & Stemberger, 1998; Ingram, 1974, 1989; Kiparsky & Menn, 1977; Olmsted, 1971; Smith, 1973; Vihman, 1980). The acquisition of consonant clusters has been investigated in different positions of syllables; word-initial clusters are acquired at 2, before clusters that occur word-finally. The latter are acquired at 2;6 (Kehoe & Steol-Gammon, 2001). Given that syllable weight is maximally bimoraic in JA where a syllable cannot include more than two moras (Huneety & Mashaqba, 2016; Mashaqba & Huneety, 2018), and given that final codas and initial onsets are moraically weightless (extrametrical), then children are expected to be very sensitive to the prosodic weight at an early stage in their language development; they reduce the syllable structure via deleting the peripheral weightless elements and keeping the core elements which count to the prosodic weight. Recall that V has one mora, VV has two moras, geminates have one mora, and onset singleton consonants and word-final consonants are weightless (Hayes, 1995).

The results here corroborate the findings of Ragheb & Davis (2010), who provided several examples in support of final cluster reduction regardless of sonority. In the onset clusters, the order of the consonants has no influence on cluster reduction as children produced obstruents and deleted sonorants; for example /bri:ʔ/ ‘jug’ was produced as [bi:ʔ], /mħ am.mad/ ‘Mohammad’ as [ħ am. mad], and /mba:.riħ/ ‘yesterday’ as [ba:.liħ]. The children also produced voiceless and deleted voiced obstruents when they occurred in the onset, as in /btiʒ.li/ ‘she is washing’ which was produced as [tiz.li]. Recall that the voiceless /t/ is acquired before the voiced /b/, and /ħ/ and /b/ are acquired before /m/ in the cluster. These results seem to be similar to those of Watson & Scukanec (1997). For more detail on patterns of word-initial cluster reduction see Locke (1983, pp. 68–71).

In terms of deletion (onset/coda/syllable), although the frequency of the occurrence of onset deletion is very low in the younger three age groups, their suppression in 3-year-old children seems to mark an increase in their multisyllabic word production. As for weak syllable deletion, it appears to last longer than the onset/coda deletion, confirming the work of Grunwell (1987) and Ingram (1989). These results indicate that this process is used by children more in the early stages because they cannot produce superheavy syllables and multisyllabic words easily like adults. The CVVC > CVV (coda deletion) process was seen in the speech of children in the youngest age group, perhaps because of their limited ability to produce a more complex syllable structure; recall that final C is extrametrical. In the older groups, children were able to produce multisyllabic words, which led them to use other phonological processes like cluster reduction.

Syllable epenthesis rarely occurred in the oldest age group. This result is in line with Kehoe (1995). Studies on Arabic-speaking children considered epenthesis from a segmental perspective as they documented it in cluster word-final position (Ayyad, 2009; Dyson & Amayreh, 2000), but they provided no evidence of the syllable epenthesis that was found here. The current study found syllable epenthesis in word-medial position, which is in agreement with the finding for English-speaking children (Kehoe, 1995); however, contrary to English which has epenthesis in word-final position, Arabic also showed epenthesis in word-initial position. Even though use of epenthesis was rare in our data, its use was observed in the older group, meaning that some children might be using it as a trochaic template to compensate for the demands on their phonological systems to produce multisyllabic words (cf., Kehoe, 1994). Basically, the purpose of simple syllable insertion is to facilitate meaningful acquisition of words, where children (whether deliberately or otherwise) develop their own ability to articulate multisyllabic words without resorting to complex structures (cf., Lleó, 1990; Moskowitz, 1973). Syllable insertion can also be regarded as a successful strategy in repairing the prosodic word structure if the child is unable to appropriately produce the other syllables of the word (cf., Ingram, 1974, p. 54). Future research on children with speech-sound disorders in Arabic might consider investigating the use of syllable epenthesis, as it might provide prognostic signs in those trying to compensate for their disordered phonological system in an attempt to approximate the adult system.

Assimilation observed in the speech of the youngest two groups is in agreement with that reported for English-speaking children. It is believed that these assimilatory processes are part of the normal phonological development. Where they persist beyond the age of 3, they might be a sign of a speech-sound disorder (Grunwell, 1987). This warrants more caution and testing of Arabic-speaking children exhibiting this process on the part of speech language pathologists. However, this process is not seen in all typically developing children (Grunwell, 1987) as is the case here for the Arabic-speaking children participating in the study.

In agreement with Dyson & Amayreh's (2000) findings, gemination is preserved even when segments have undergone substitution processes. Gemination was also maintained even though the word underwent a syllable structure deletion process. Gemination was preserved even when segments were simplified (cf., Dyson & Amayreh, 2000), but with the doubling of the consonant sequence. Intrigingly, JA children resort to gemination (and assimilation) to rescue the bimoriac minimality forced by the prosodic parameters of Arabic phonology. Adding a new segment grants extra moraic weight to the phonological word. If the child produces bis alone (instead of biss ‘cat’), the word structure loses its bimoraic property, which is a basic principle of adults' language. Given that children's articulatory system is not yet fully developed, this is a simple way to use extra elements that belong to the first produced segments.

Use of onomatopoeia by the youngest group seems to be a sign of their limited phonological systems, which are still unable to produce certain set of consonants, complex word or syllable structures; this is obvious from their frequent use of monosyllabic (CVG and CVVC) words. One explanation of the use of onomatopoeic words might be their salient features, i.e., clear prosodic features; when produced by mothers, they might help children learn them as frequent early productions (Laing, 2014). However, this process had completely disappeared in the three older groups, which also means that the children's abilities to produce complex word and syllable structures were developing. Consequently, the persistence of onomatopoeia may provide clinicians and researchers who study child language development in Arabic with an early indicator to warrant indepth testing of children's phonological systems, in order to rule out the presence of speech sound disorders as most of the target words which underwent change to onomatopoeia also underwent syllable structure processes (i.e., reduction to one-syllable words).

In conclusion, this research examined development of the syllable, word syllable structure and number, and notable phonological processes among JA-speaking children aged 1 to 3. Children acquire syllables by imitating adults' production of different words. They tend to apply phonological processes to repair syllables to adult-like shapes and to compensate for the loss of a syllable or its parts. Ten processes were investigated: assimilation, cluster reduction, syllable deletion, changing from closed to open syllables (CVVC > CVV), gemination, metathesis, onomatopoeia, vowel lengthening, vowel shortening, and syllable epenthesis. Analysis of the data shows that the most frequent syllable shape used was CVC and the least were those with consonant clusters. In line with Rvachew, Nowak, & Cloutier (2004, p. 251), it was shown that the progressive increases of phoneme perception are mainly linked to improvements in the accuracy and stability of syllable and word articulation, as the children adapt their speech production to approach the phonetic and prosodic characteristics of native language speech-sound features. The corpus of the present work proves that children try to develop the metrification system in Arabic at an early stage. They are aware of the inherit weight assigned to segments in the core and peripheral positions and of the coda non-final positions assigned by the Weight by Position rule. The children resort-ed to multiple repair processes to preserve that mora instead of deleting it when encountering an articulatory problem such as gemination, assimilation, and syllable insertion.

Word syllable shapes were investigated in this study, the most frequently used shape being bisyllabic words. Trisyllabic word usage increased as children grew older, but quadrisyllabic words were not produced by any of the children in the four age groups recruited in this study. Children's first words usually target adult words of simple prosodic structure (minimally bimoraic), and one- or two-syllable word forms with open syllables, no clusters and core consonants (stops, nasals, glottals and glides). Despite the documented similarities across children in terms of the sound segment and syllabic types produced, there is often a wide range of differences in the target words that are attempted. This discrepancy re-lates to patterns of preference for, and avoidance of, particular sounds or sound classes or certain syllabic shapes (Ferguson & Farwell, 1975; Vihman, 2013). The study can promote future research on JA that might use larger samples of children.

Finally, this study might provide researchers and clinicians (speech language pathologists) with valuable information on the development of the early syllable and word structure in Arabic-speaking children, which might aid in the early identification of at-risk children and in the best choice of therapy targets in terms of developmental syllable structure.