Aspects of Assimilation, session 2
Measuring Voice Assimilation

LOT Summer School 1998, Tuesday 23 June 1998.

Rules for Dutch voice assimilation

(after Slis 1985: 123ff.; Rietveld & Van Heuven 1997: 184):
Regressive (or anticipatory) Voice Assimilation:

C1	C2	-->	C1	C2
[obs]	[plos]		[obs]	[plos]
	[+voice]	[+voice][+voice]

Progressive (or perseveratory) Voice Assimilation:

C1	C2	-->	C1	C2
[obs]	[fric]		[obs]	[fric]
[-voice]		[-voice][-voice]


For the following assignment, (Dutch) speech materials have been constructed in which voicing assimilation is likely to occur. (In some cases, I have also made a corresponding utterance in which the 'assimilated' segment is present in the 'underlying' for m).

The necessary audio files, in *.aifc format, can be found in the archive ~quene/data/lotzomer/guus/aa2.tar. Copy or download this archive to a personal directory, and extract its content with
tar xovf aa2.tar *
See Table 2.1 below for a list of the files you will need. The files with extension *.TextGrid contain hand-edited segment boundaries.

1. V~UV via pitch detection

Determine the Voiced or UnVoiced status of the relevant speech segments, in all files with a viable context for assimilation. Use a pitch detection algorithm of your choice (e.g. in Praat: Sound to Pitch (ac)). Set all analysis parameters to thei r default values.
Is the resulting V~UV classification constant for all analysis frames in the speech segment? If it is not, then the V~UV status for the whole segment should be determined by the value of the majority of frames.
Mark the resulting voicing status in Table 2.1 below, in the column labeled V~UV, using plus and minus symbols.

2. Inventory of acoustic correlates of [voice]

Together, list all the acoustic correlates mentioned by Slis & Cohen (1969). What is the direction of the difference, for each parameter? Why? Discuss with your fellow students! (You may be both puzzled and inspired by Figure 18 of Slis & Cohen 1969, p. 1 50).

3. Measure one correlate

Each sub-group has the task of measuring one additional parameter, a phonetic correlate of the voiced~voiceless distinction, as described by Slis & Cohen (1969) and others.
Measure 'your' parameter in all segments in all the speech files listed in Table 2.1.
Next, interpret these measurements as a binary distinction, perhaps with the help of Slis & Cohen, Figures 15 and 16, pp.140-141 (although these Figures refer to VCV rather than VCCV contexts).
Mark the voicing status for each segment in Table 2.1 below, again using plus and minus symbols.

Table 2.1: Results of V~UV and one other acoustic parameter
Regressive Voice Assimilation
viable unviable
file V~UV parameter file V~UV parameter
tb_vaatbak tp_maatpak
kb_bakboord kp_bakpoort
fb_hefboom fp_stofpak
sb_asbak sp_prijspand
pd_stropdas pt_toptien
kd_zakdoen kt_boktor
fd_afdoen ft_aftands
sd_misdaad st_mistaak
Progressive Voice Assimilation
viable unviable
file V~UV parameter file V~UV parameter
sd_misdaad st_mistaak
pv_opvallend --
tv_potvis --
kv_ziekvee kf_ziekfee
sv_huisvuil sf_visfuik
pz_opzoeken ps_opsokken
tz_uitzonderen ts_uitsolderen
kz_bakzeil --
fz_stofzak fs_hofsap

4. Combine all classifications

Results from all sub-groups are combined in a single master table, edited by one participant. This table should contain plus and minus symbols based on all parameters, as measured in section 3 above.
The resulting degree of voicing for each segment is established, based on the proportion of Voiced votes (plusses). Finally, the simple majority of Voiced or Unvoiced votes determines the cla ssification of each segment as either Voiced or UnVoiced.
Both degree of voicing and final classification are noted in the master table.

5. Partial or complete assimilation?

Compare the neutralization cases, e.g. huisvuil ~ visfuik. How are the latter ('underlying' unvoiced) and former ('assimilated' unvoiced) variants represented in the master table? Is this correct? Do you hear a difference between the variants, with respect to the intervocalic consonants? Describe any difference y ou hear or see, as accurately as you can.
BREAK around 14:15 (before 14:30 !)
During the break, the master table is copied for all participants.

6. Gestural score

Take a good look at the master table.
For a selection of cases, make a gestural score, as exemplified by Browman & Goldstein (1992). What are the relevant articulators in the selected cases? What are the tract variables?
6.1. Make a schematic score of the 'input' gestures (block style).
6.2. Make an educated prediction of the 'output' tract variables, resulting from the underlying gestures. Perhaps it is easier to start with the latter task, and to derive the input gestures from these.
Mark problematic cases. Why are these problematic?
Last updated on June 19, 1998, by Hugo Quené.