TRANSCRIPTION, CODING AND ANALYSIS

JC is mainly an oral language. Many of the lexemes are English based but their phonology is quite different. For all transcriptions and examples reproduced in this dissertation, the JLU

42

(Jamaica Language Unit) modified Cassidy-LePage orthography was employed. This is a system that represents the sounds of JC as faithfully as possible, without relying on the spelling conventions of English. This system has no silent letters and there is a one-to-one mapping of sound to symbol therefore each letter or letter combination is always pronounced the same way (Jumieka Langwij).

All data was transcribed in CHAT (Codes for the Human Analysis of Transcripts) format, following the standard guidelines of the CHILDES (Child Language Data Exchange System) Database. However, for clarity, most of the codes have not been used. The transcription procedure proved to be very time consuming as it took approximately 10 hours to manually transcribe one hour of data. For transcribing the data, the recordings were transferred from the recording device to the computer. High quality earphones were used and the researchers transcribe the exact production as uttered. In most instances, due to surrounding noise including overlapping speech, recordings had to be repeatedly listened to in order to ensure accurate transcription. Inaudible speech was transcribed as xxx.

While the joint recording sessions were transcribed by the main researcher, the recordings of a particular child in the parallel sessions were transcribed by the researcher who had conducted the recordings. In this way all the paralinguistic details not included in the audio recordings, but which could have impacted the interpretation of a particular utterance could be included in the transcriptions.

A backed-up copy of all recordings and transcriptions is stored on the University of Geneva Database for safekeeping. Moreover, the database will be transferred to CHILDES for public access.

43 4.4.1 Coding

Coding entails analyzing the transcriptions and making notes of the grammatical and syntatic phenomena. This is done by creating a one-to-one correspondence between utterances and standard morpho-syntatic codes on a word-by-word basis. Based on the time frame for the completion of the research project, a decision was taken to initially code only the utterances of the target children. I started coding the first set of transcriptions, however, it proved unfeasible and as such four final-year students from the University of the West Indies were subsequently employed to complete the coding of the data, under my guidance. The approximate time to code one transcription was 6 hours. This time however is not fixed as it depends largely on the number and length of child utterances in the transcription.

In order to select competent students for employment for this task, a class of Creole Linguistics final year students was trained in doing transcriptions and coding data in the CHAT format. The four most outstanding students in this task were subsequently employed to undertake the morpho-syntactic coding of the corpus.

Based on the non-standard conventions in transcribing JC, coding of the data had to be done manually. A list of codes was developed for conveying the morpho-syntactic relations (See Appendix 2). Despite this comprehensive list, coding of the data did not prove to be unproblematic as there are some instances where a particular lexical item could lead to different interpretations or yield different codes in the same context. To deal with these occurrences, native speakers’ judgments were employed where applicable, or the word in question coded as unknown.

44

In dealing with other issues regarding the coding of utterances, we adopted the method employed in comparative syntax, assuming on grounds of continuity and uniformity, that child language approximates adult grammar (in line with Bates et.al. 1994, Gleitman et.al.

1999, among others). We acknowledge that this is not necessarily true in general but it is the necessary initial assumption for comparative research. The data was also coded for null elements in the grammar.

Several meetings were conducted with the coding team to resolve all issues. Two recordings were coded per child for the period starting January 2012 to March 2013 and one for each child in April 2013. The decision to start coding the data for analysis as of January 2012 and to include only two of the three recording sessions for the period January to March 2012 was mainly based on the following:

1. the maximal use of funds available

2. standardizing the quantificational dimension of the analysis 3. the final selection and confirmation of all the research participants

4. the initial two months involved familiarization of the participants with the researchers thereby maximizing their language production levels.

All completed coding was duly checked for verification of accurateness, and for inclusion of additional details as required for the analysis. A total of 186 transcripts, comprising more than 80,000 meaningful child utterances (not including fillers e.g. mh) were coded and subsequently became the basis of analysis for this dissertation.

45 4.4.2 Analysis

The coding scheme provides for the description and analysis of the data in terms of syntactic structures. The analysis of the production data was based mainly on age and developmental stages (in line with Radford 1990). Nonetheless, where necessary the participants’ production was classified and compared by their mean length of utterance (MLU) as this method is widely considered to measure the level of language development (see Brown 1973, Miller 1981, Miller and Chapman 1981). Since JC is an isolating language, lacking morphologically complex words, MLU was calculated on a word basis. Single word utterances of fillers such as mmh were not included in the count.

Some utterances were excluded from the data analysis, these include:

- utterances in which any unintelligible portions (coded an UNK) could be critical for the analysis

- utterances where the meaning was unclear based on the context of the discourse

- the child’s stuttering or self-repetitions without the production of contentful utterances in-between

- repetitions of memorized materials, e.g. songs and nursery rhymes - immediate repetitions of adult’s exact utterance

Various sections of the analysis are based on automatic computing of the morpho-syntactic coding using, but not limited to, CLAN, NotePad++ and Excel. Nonetheless, manual analysis was inevitable for certain computations. Of course, in order to conduct cumulative analysis of the group as a whole, and for developmental comparisons, the exact age of each child could not be used. I therefore grouped the participants by approximate age. For example, a child

46

who was aged 2 years, 6 months and 4 days (2;6,4) was categorized as 2 years and 6 months old (2;6.0) and one who was 2 years, 6 months and 20 days (2;6,20) categorized as 2 years, 6.5 months old (2;6.5).

The analysis provided in this dissertation is generally wide-scoped, covering a range of syntactic phenomena in a comparative manner. Given our present knowledge of the target system, as alluded to in the previous chapter, the syntax of adult JC is relatively well documented, thereby facilitating comparisons between the child system and the adult grammar. Additionally, the enormous published theoretical and descriptive work in non-creole child language acquisition enables comparative analyses among the syntactic systems of children in various languages. The analysis of the data therefore seeks to explain not only the comprehensive development of early JC, but to add to our understanding of the general nature of language acquisition.

47 CHAPTER 5