THE INVESTIGATION: SCOPE, PROCEDURES AND HYPOTHESES
13 Again, the issue over whether or not these are misperceptions arises In the case o f insertions it seems likely that Ss are "hearing things' that aren't there but for omissions it could be either failure
however, it is impossible to know whether the student who writes EVS [i] for
the second primary cardinal vowel, [e], is doing so because they have symbolization problems or whether because they genuinely believe that they
have heard [i]. Similarly, when a student represents [in] as a labialized voiced
retroflex lateral [j^j, we cannot know without asking whether they wrote it as
a function o f ignorance and/or confusion or because this was the sound they genuinely thought they heard. Partly for this reason, but also because the symbol is outside the vowel symbol range altogether and because consonants are not readily assessable in terms simply of backness, openness and roundness, consonant symbols used to represent vowels are treated as a type of MFS-error.
So, by establishing these three categories of error (and not addressing the issues specifically and only as problems of a perceptual nature when we do not have the knowledge to hand to confirm this) it is at least possible to establish a systematic approach to handling responses of the TES kind. TES involve a mixture of response-types for which the title transcription errors' is simply a convenience allowing them to be grouped separately from errors of the strictly classificatory type.
Another approach here might have been to treat all the vowel responses not as
Cardinal Vowels, but as TPA vowels' and to deconstruct every wrong response in B, O and R terms. However, although students were trained using the IPA
alphabet chart (and, indeed, had access to this during testing) they were specifically taught general phonetic Cardinal Vowels on the one hand and the phonemic symbols for representation of English vowels on the other. The two were never mixed. The symbol lists in question are reproduced in APPENDIX
5 Symbol Sheets. To employ a simple, holistic, IPA-based approach to the evaluation of their responses would not reflect this and it was therefore felt to be less appropriate - rather like asking someone to provide imperial
measurements and then checking the values using a metric rule. It was decided, therefore, to employ the same values and categories throughout, both in the teaching/learning context and in the evaluation.
In the case of consonants, however, a number of different issues arise. The 'English symbol' vs 'IPA symbol' distinction just described for vowels (that is 'English symbol' vs 'Cardinal Vowel symbol') breaks down because English consonant symbols cannot be factored out from the IPA array in the way in which it was felt the vowel symbols could be. Nonetheless, there are a considerable number of recognised instances where representation of a particularly phonetic value can be interpreted by the reader as having been influenced by English transcription habits. The representation of [ç] as [hj] or
[m] as [hw] would be cases in point and even failure to invert the r-symbol for
a voiced postalveolar approximant could be attributed to this categoiy rather than being regarded as a straightforward perceptual error of place and manner. However, because none o f these possibilities can be categorically and unarguably identified as English (unlike [ou] or [æ], for example) the potential
ECS (English Consonant Symbol) category was rejected.
The two remaining TES-types (MFS and Z) pose no special problems for consonants.
S.2.4.4 Evaluating diacritics
A fourth general parameter called diacritic (D) was also available (that is, it
could be used in the case of inclusion or omission of nasalization, length, aspiration, etc., should the need arise). With regard to vowels, however, since length of stimulus is not overtly trained or assessed and nasalization was poorly
represented in the assessment materials^"* and since both are in any case independent of the judgement o f basic vowel quality, the diacritic parameter was not invoked (the column-heading will be seen to be shaded out in the vowel
version of the specimen forms included in APPENDIX 4 Preliminary Analysis Forms). For consonants, however, the status of D is different. The ejective diacritic, for example, overt marking of aspiration, velarization, labialization, 'dental', etc., are all critical differences in terms of recognition of basic consonant quality and the D category has therefore a larger part to play. However, in that the contribution o f a diacritic is often independent of the voice, place and manner features, not all diacritics can be evaluated the same way. Diacritic errors resulting in a difference of voice, place or manner are therefore analysed as voice, place or manner errors and incur an appropriate V-, P- or M- error penalty. Others are treated as a fourth type of TES. A full discussion of the problems encountered with diacritics and the solutions adopted is included in APPENDIX 3.
5.2.4.5 Closing remarks
After preliminary evaluation as described above, further detailed analyses of aspects o f the results were carried out within the parameters identified. All findings were initially logged on specially designed preliminary analysis forms (specimens of which are included in APPENDIX 4) before being transferred to spreadsheets.
In conclusion here, it is also worth observing that the range of variation in the consonant category is vast by comparison with vowels. The number of vowels under investigation is 12 and given that they are only being treated at a very
14 It is my considered opinion that the training materials (see APPENDIX 1) could also be criticised