Type versus token frequency

This section will present similarities and differences between type and token frequency analysis. Table 5.9 shows the occurrence frequency of KA

consonants in target words produced by all groups of children in the current study, grouped according to production manner. From table 5.9, it can be seen that there is general agreement between those two kinds frequency

calculations. Except for stops and fricatives, the frequency of all other less frequently occurring consonants seems to be in agreement. Under the assumption that both type and token frequencies are normally distributed (based on data presented in the current study), then the t-test (t = 0.824 two tailed). Therefore, the difference between type and token frequency is not significant. This agreement could possibly reflect the child’s tendency to select familiar or relatively easy to produce segments (or words containing segments), that are stored in their lexical repertoire (Vihman, 1996, 2013). The differences between type and token frequencies are to be highlighted in the following tables.

140

Token Count

Token

Frequency Type count

Type Frequency

Stops 14,116 29% 2,870 31%

Nasals 7,688 16% 1,281 14%

Trill/tap 2,644 5% 645 7%

Fricatives 14,923 31% 2,296 25%

Approximants 2,713 6% 816 9%

Laterals 3,073 6% 816 9%

Affricates 1,162 2% 224 2%

Emphatics 2,076 4% 381 4%

Total 48,395 100% 36,710 100%

Table 5.9: Type and token frequencies of consonants occurring in KA spontaneous child speech.

However, detailed analysis of frequent consonants (Table 5.10) reveals some discrepancy between type and token frequencies. For example, token

frequency analysis shows that the voiceless glottal fricative (/h/) is the most frequent consonant followed by the voiced alveolar nasal (/n/) and the voiced bilabial stop (/b/). Type frequency, on the other hand, shows that /l/ is the most frequently used consonants, followed by /n/ and /b/. Comparably, earlier studies of the development of Arabic phonology found that /n/ and /b/ occur in high frequency (e.g. Amayreh and Dyson, 2000; Saleh et al., 2007). The token frequency of /h/ could possibly result from the frequent use of the Arabic pronouns /haː.ð4/ ‘this-masculine’ and /haː.ði/ ‘this-feminine’ by either the parent or the child during the recording sessions in an attempt to encourage the child to name as many objects or pictures. This again was reflected in the frequency of the voiced dental fricative /ð/ which has token frequency of 5.6%

and lower type frequency of 0.5%.

141

Least frequent ⟵⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⟶ Most Frequent

Frequency Type frequency Token frequency

>5%

Table 5.10: Type and token consonant frequency in KA child speech

Both type and token frequency analysis shows that the least frequently used consonants are mostly non-Arabic consonants (e.g. /v, p, ɹ, ŋ/) except for /dˤ/

and /ʒ/. The English velar nasal consonant /ŋ/ was produced by only two children out of 70 samples. The two children (cousins) were recently introduced to the cartoon character King Kong. They produced the word /kiŋɡ.koŋɡ/

repeatedly, hence the /ŋ/ was produced 113 times collectively, which had great impact on its frequency rank. However, this consonant was excluded from the production accuracy calculations as it did not meet the criteria of being

produced by five children within an age group. Another less frequent KA consonant is the voiced alveolar emphatic stop /dˤ/, which occurs frequently in MSA; however, in the KA dialect it is often realised as [ðˤ] as in the the following examples:

(1) MSA:

a. /dˤif.d4ʕ/ ‘frog’

b. /dˤa.ɾi/ ‘Dhari’ (Arabic name) (2) KA:

142 a. [ðˤif.d4ʕ] ‘frog’

b. [ðˤa.ɾi] ‘Dhari’ (Arabic name)

Likewise, the voiced post-alveolar fricative /ʒ/ occurs in dialects of Arabic other than KA, such as Jordanian and Lebanese. The consonant /ʒ/ was observed in the speech of one child in the sample. However, it was produced twice only by the child and therefore did not affect its frequency (less than 0.01%). The KA equivalent of the /ʒ/ fricative is the affricate /ʤ/. One possible explanation is that a member of the child’s family speaks other dialects of Arabic (extended family), from which the child acquired the sound.

Consonant occurrence frequency was also calculated for each word position.

Table 5.11 presents type and token frequencies of target consonant occurrence in word-initial –medial, and -final positions and the overall production accuracy percentages for each consonantal group (figures 5.5, 5.6 and 5.7).

Consonant' Frequency'

Word'Initial' Word'Medial' Word'Final' Total'

Words' Type'' Token'' PCC'' Type'' Token'' PCC'' Type'' Token'' PCC''

Stops' 43%' 38%' 88%' 24%' 21%' 85%' 28%' 34%' 86%' Total'Types' '2,806''(35)^*' '4,434'(39)^*' '1,472'(35)^*' '8,712'' Total'Tokens'

Table 5.11: The frequency of target consonant occurrence in word-initial, word-medial and word-final positions.

Table 5.11 shows relative agreement between type and token frequencies in all word positions. For instance, stops were found to be produced in word initial position with type frequency of 43% and token frequency of 38%; followed by nasals and fricatives. The frequency of consonantal groups is calculated in all word positions (table 5.10). Figures 5.5, 5.6 and 5.7 show the occurrence frequency of KA consonants in all word positions.

143

Figure 5.5: Word-Initial Consonant Occurrence Frequency

Figure 5.6: Word-Medial Consonant Occurrence Frequency

Figure 5.7: Word-Final Consonant Occurrence Frequency

24%! 27%!

Emphatics Approximants !Laterals

Trill/tap Affricates Type Frequency Token Frequency

28%!

Emphatics Approximants !Laterals

Trill/tap Affricates Type Frequency Token Frequency

43%!

Emphatics Approximants !Laterals

Trill/tap Affricates Type Frequency Token Frequency

144

There is a general agreement between type and token occurrence frequencies in all word positions. Where stops seem to be most frequent at word edges (initial and final); liquids seem to be more frequently targeted in medial and final than in initial position; fricatives are targeted more frequently in initial and medial than final position.

Statistical measurements were not obtained here for two reasons: (1) the normal distribution of KA consonants in adult speech is unknown; (2) and the assumption of normal distribution could not be based on data collected for the current study because of the limited number of participitants in the current study which may not be representitive of all KA consonants in adult and child speech.

Comparing the results of token and type frequency of word structures of target words reveals a child preference to use of certain word structures.

The most frequently targetted shape is CVC.CVC (type frequency 6%); where this word shape occurs less frequently in token frequency analysis (4%).

Likewise, token frequency analysis showed that CVV.CV word structure occurs most frequently (19%); but it was used less frequently as its type frequency is considered (5%) (see tables 5.5 and 5.3). The observed difference between type and token could possibly mean that KA speaking children show preference to some word shapes which they tend to use more frequently.

This preference for particular word stuctures was reflected in the stress patterns used by KA speaking children. Table 5.12 shows type and token frequency of different stress patterns. Stressed syllables are marked ‘S’ and weak syllables are marked ‘w’. Type frequency analysis of stress patterns shows equal occurrence of wS and Sw patterns; where as token frequency showed higher tendency to use the Sw pattern. Other less frequently used patterns were found to occur in similar order in both type and token frequency (e.g. wSw, wwS, wwSw).

145

Type Stress Pattern Count Frequency Token Stress Pattern Count Frequency

wS 770 31% Sw 8,230 54%

Sw 762 31% wS 4,064 27%

wSw 697 28% wSw 2,240 15%

wwS 144 6% wwS 331 2%

wwSw 88 4% wwSw 191 1%

wwwS 13 1% wwwS 24 0%

Sww 8 0% Sww 20 0%

wwwSw 4 0% wSww 7 0%

wSww 3 0% wwwSw 4 0%

wwwwS 1 0% wwwwS 1 0%

Total^* 2,490 15,112

* Excluding monosyllabic words.

Table 5.12: Type and token frequencies of target word stress patterns in KA.

The difference between type and token frequency can possibly reflect the child’s preference for certain word structures. For example, if a child was tested using a word list that reflects the occurrence of certain structure in the

language, the child preference may not be appreciated which result in higher error rate. Whereas if those tendencies were taken into account, the

assessment may show rather faithful phonological profile of the child.

The child’s preference for certain word structures could possibly reflect ease of production, where the child chooses lexical items with less complex structures or may be influenced by the occurrence frequently in child directed speech. For instance, there was evident tendency to target words containing geminates less frequently in tokens than in types. Type frequency of words containing

geminates was 16%, where as token frequency was only 10%; however, conclusions may not be made within the scope of the current study, thus further detailed analysis is needed.

The following section will show analysis results of consonant production accuracy and error patterns in all word positions.

146 5.3. Consonant acquisition in KA

The section presents the production accuracy of each consonant. The first part reports consonant production accuracy in relation to the frequency of its

occurrence. The second part reports consonant production accuracy of each age group.