HQSYN16

In general, it is important to realize that the voicing in /ř/ is the least stable of all phonologically voiced obstruents (Skarnitzl, 2011, p. 215). It tends to get lost even between vowels (words like pořád) and between voiced sounds (v řadě, bříza, pařba). It is quite likely, therefore, that phonetic voicing may have been absent in contexts marked as voiced. Insertion of such a (partially) devoiced /ř/ into a voiceless context thus would not be disturbing.

Specific notes to the individual realizations:

sentence #3 - ..., ohnivzdorné bundy a pep*ř*ové kanony.
- total score: - 0.14
- [ř] score: - 0.43
- baseline is voiceless, neither Ř has a disturbing effect
- there is no obvious reason for such a bias in the evaluation

sentence #11 - Vep*ř*ové je témě*ř* tabu.
- total score: - 0.29
- [ř] score: - 0.14
- the first baseline Ř is clearly from a voiced context, the second is practically voiceless
- the timing is off in both the baseline and noSubsts versions, but baseline sounds worse, there is no obvious explanation for the bias

sentence #12 - a aby nast*ř*ádal na sibi*ř*ský výlet, ...
- total score: - 1.0
- [ř] score: - 0.43
- the first baseline Ř is voiced towards the end, the second is voiceless
- there is a huge quality change in the [a:] in nastř+á+dal which may have caused the biased evaluation_

sentence #16 - zdravotnické pot*ř*eby a léka*ř*ské služby.
- total score: - 0.43
- [ř] score: - 0.14
- both baseline Ř sounds are voiceless and not disturbing
- the bias may be due to timing which is off in noSubsts, especially in potřeby

sentence #22 - A mod*ř* nehybně odplouvá tak jako řeka, ...
- total score: - 0.71
- [ř] score: - 0.29
- baseline is fully voiced, which corresponds to Moravian assimilation (voicing before sonorants), noSubsts is voiceless and correct
- since this is quite a rare word, the spelling and expectation of [d] may have prevailed, but this is just speculation

sentence #25 - Vě*ř*te mi, ...
- total score: - 0.29
- [ř] score: 0.29 was evaluated as better in noSubsts, problem may be somewhere else
- the voiced context of the baseline Ř is audible on closer analysis, but it is phonetically voiceless

sentence #27 - , podivil se signatá*ř* Martin Mejst*ř*ík.
- total score: - 0.57
- [ř] score: 0.43 was evaluated as significantly better in noSubsts, problem may be somewhere else
- both baseline Ř sounds are voiceless and not disturbing

sentence #29 - Množme se a spo*ř*me.
- total score: - 1.0
- [ř] score: - 0.14
- In this item, the voicing is reversed! The noSubsts version is fully voiced, i.e. with Moravian assimilation, the baseline is voiceless = correct, and the score is thus logical.

sentence #30 - Hrát proti mladší soupe*ř*ce, ...
- total score: - 1.0
- [ř] score: - 0.57
- the baseline Ř is voiceless, the timing is very much off in the [e] before Ř in the noSubsts version; in other words, both Ř sounds are ok, the biased evaluation is probably due to the context

sentence #35 - ..., u*ř*knutí a podobně
- total score: - 0.43
- [ř] score: 0.29 was evaluated as better in noSubsts, problem may be somewhere else
- both Ř sounds are voiceless and although the voiced context in baseline is audible upon closer analysis, it is not disturbing

sentence #40 - ..., vysvětlil jeho ošet*ř*ující léka*ř* Ilja Kotík.
- total score: - 0.14
- [ř] score: - 0.14
- all Ř sounds are voiceless and not disturbing

sentence #41 - Hovo*ř*te o t*ř*ech týdnech.
- total score: - 0.57
- [ř] score: 0.29 was evaluated as better in noSubsts, problem may be somewhere else
- all Ř sounds are voiceless and not disturbing

sentence #42 - Pracovali témě*ř* výhradně u pošty
- total score: - 0.29
- [ř] score: - 0.14
- both Ř sounds are voiceless and not disturbing