MEDELLÍN, CIUDAD DE MIGRANTES, CIUDAD RECEPTORA

The laryngograph recordings can be displayed as waveforms and spectrograms in both packages; Fo contour, Lx waveform and closed quotient contour (Qx) can be derived from the laryngograph trace and displayed simultaneously in the SPG. The packages provide the optimum display for clarity of measurement. As each single vocal pulse can be clearly demonstrated in the Lx waveform, the duration of the disyllabic words is taken from the first vocal pulse to the last. All tokens are digitised at a 10 kHz sampling rate.

2.1.2.4.1. Segmentation

Time points of syllable boundaries are determined by inspection of the speech waveform, the wideband spectrogram, and the Qx and by listening to the signals. The segmentation for intervocalic nasals is relatively straightforward, since the velum opening of the nasals shows an abrupt change between the offglide movement from the preceding vowel to the steady formant pattern, then to rapid onglide movement of the following vowel in the wideband spectrogram. Also the speech waveform can clearly reveal the boundaries between the nasal and its neighbouring vowels - vowels have higher amplitude than the nasals. The starting point of a lateral approximant /I-/ as the initial of a syllable is defined by observing the abrupt stop of the off-glide from a vowel of the preceding syllable to the steady formants and by the noticeably sudden change of the waveform pattern. The F2 movement provides a good cue for defining the boundary of an intervocalic glide /w/. The region in which the F2 slips down and meets with F1 is considered as the beginning of an initial /w-/. The region in which the F2 shows the start of a rapid rise is considered as the end of an offglide /-w/ and the beginning of a high vowel. The segmentation for the semi-vowel /j/ seems to be problematic. However, /j/ is stable in formants, the frequency of its F3 is much higher than that for other vowels. The energy of its F3 is much greater than that of its other formants, especially at the beginning or soon after the beginning of the F3. Thus, a steady state of high frequency and high energy of F3 serves as the starting point of /j/. When an onset /j-/ follows a coda /-j/ or a vowel /-i/, i.e., in sequences of /-i/ # /j-/ and /-j/ # /j-/, its formants still enjoy their steady state, but the energy is weaker compared with its predecessors), /-i:/ is a

full vowel which has steady formants and usually has high energy in all formants during voicing. The F2 of /i:/ is rather concrete in comparison with /j/ whose F2 is rather faded. The formants of a coda /-j/ are always affected by its preceding main vowel. F2 especially performs a rapid rising from a preceding open or back vowel to the /-j/. /j/ is fully voiced but has less friction in an onset position than in a coda position. Thus, the starting point of steady formants of /j-/ where it shows a relatively weak intensity - is considered as the syllable boundary between the coda /-j/ and the onset /j-/. Apart from the conventional segmentation method, the Qx is useful because it demonstrates the increase in the closed quotient of the glottal cycle associated with vocal fold abduction. Generally speaking, the higher the Qx, the lower the frequency - i.e., the Qx contour and the Fo contour are in

inverse direction. Any turning point in the Fo contour more or less accompanies (but not necessarily) a turning sign in the Qx contour at the same time. Changing in segments is also signalled in the Qx contour by increasing or decreasing the closed quotient. The first Fo

value is taken as the tonal onset of the first syllable and the last Fo value is taken as the offset of the last syllable. It is not abnormal to see the last few flaps of the vocal folds drop the Fo contour. This signals that the vibration of the vocal folds is weakened when

approaching the end of the vibrations. In this case, the last Fo value is not necessarily

reliable, so I use the Fo value just before these drops (usually it sits approximately 2 0ms

before the end). When the same phenomenon occurs at the beginning of the Fo contour, i.e.,

the vocal folds have one or two weak vibrations at the beginning, and this would lead the Fo

contour to show a weak start, the first Fo value after the weak start is used. It is observed

that the Fo contour is not necessarily a smooth contour. It has a few turning points. The turning points are observed to occur especially more or less a quarter of the duration of the syllable after the tonal onset or before the tonal offset. Therefore, it was decided to take the Fo value at three points apart from the onset and offset: a quarter, half and three quarters of the total duration of a syllable. The values of these three points plus the values of the onset and offset are taken into computation to determine the Fo contours. It is not uncommon to

produce a creaky voice in low frequency words in which the Fo contours become aperiodical.

A handful of tokens which result from this and have a long period of aperiodical Fo contour were excluded from the investigation together with some tokens judged to be unrepresentative. Subsequently many tokens had to be discarded in order to obtain an equal number of tokens for each set of the two-tone sequences. 648 tokens in total were involved;