I came across the following phrase in a description of German pronunciation:
The basic rule is that voiceless obstruents are always voiceless, but voiced obstruents are not always voiced.
...which makes absolutely no sense to me. (In general, "X is not always X", is equivalent to "sometimes, X is not X", but "X is not X" is always nonsense.)
What could the excerpt possibly mean?
The basic rule is that players from Shirtless Team are always shirtless, but players from Shirt Team don't always wear shirts. Sometimes, at the end of the day, it gets too hot and they take off their shirts. They are still players from Shirt Team, but right now they’re shirtless, and look just like players of the other team.
Consequently, if you see a player wearing a shirt, you know they’re from Shirt Team; but if you see a shirtless player, you can’t be sure. They may be from either team. You have to rely on some other information (e.g. which side of the field are they playing on).
Example: The word Tage, “days“, is made from the phonemes /ˈtɑːɡə/. The third one is a voiced obstruent, /ɡ/. You’ve seen that voiced obstruents can be (=sometimes are) voiceless; in German this happens, for example, in word-final position. So take away the plural -e, making Tag “day”, and now it’s pronounced [tɑːk]. Left alone at the end of the word, the /ɡ/ “takes off its shirt”, so to speak. But it’s still a Team Voiced player. How can you tell? Well just add a vowel and you’ll see its voiced form.
Compare the word Volk /fɔlk/ “folk; people”. Its plural form is Völker, /ˈfœlkɐ/. Between /fɔlk/ and /ˈfœlkɐ/, the /k/ never changes shape. That’s why it makes sense to say that Voiced consonants can be voiceless (sometimes), but Voiceless consonants are never voiced.
(What I am trying to convey with capital letters and with the "team" metaphor is the difference between phonemes and phonetic realization, as pointed by Jeremy Needle above.)
This quote is from C. Hall Modern German Pronunciation: An Introduction for Speakers of English, p. 27. He is referring to the well-known fact that German b d g are not fully voiced (not voiced at all) except between phonetically voiced segments. He gives examples like "habe", with a graphic indicating that "b" is voiced, and compares that to "Bad" where the initial segment is half-voiced (voiced on the half before the vowel) – this he writes as "b̥". He says "When using the term 'voiced', then, we musty be aware that these sounds are often not fully voiced". Also note that he transcribed "Bad" as [b̥atʰ], where traditional "final devoicing" is phonetically realized as aspiration.
There are two approaches to the German obstruents, one where "bdg" are voiced and "ptk" are voiceless, and the other where "bdg" are unaspirated and "ptk" are aspirated" ([+spread glottis]). This paper by Jessen & Ringen summarizes the controversy, and provides evidence in support of the aspiration theory over the voicing theory. The argument is, essentially, that phonetic voicing when it exists can be attributed to passive voicing (semi-automatic acquisition of vocal fold vibrations when an unaspirated sound is surrounded by voiced segments, such as vowels), and otherwise "bdg" are phonetically not voiced. This same analysis has also been given to English, given that for many speakers, initial [bdg] are voiceless unaspirates, although they are still distinct from [ptk] as they occur after [s]. Hall's use of scare quotes around "voiced" indicates that he is aware of and does use the term "voiced" because it is traditional, but he does not actually believe that "bdg" are voiced.
