Here's a question I posed to a prominent researcher in French phonology during my undergrad. We didn't spend a ton of time on it, but we couldn't come up with a satisfactory solution. Now that I've discovered this StackExchange, I thought I'd bring it up again...

[ps] pr [bz]?

The English and French words "observe" both derive from Latin, the English through the French.

Some dictionaries give /ɔbˈsɛʀv/ for the French pronunciation and /əbˈsɜːv/ for the English; others give something more in line the common realizations [ɔpˈsɛʀv] and [əbˈzɜːv] respectively. (The realization of that last vowel is not material here.)

This perplexing behaviour is not isolated; a similar situation obtains for pairs like "absurd/absurde", where English tends towards [bz] and French towards [ps].

The Latin, we shall presume, was at some point realized [bs] or at least represented /bs/. Either way, the ensuing assimilation evidently went in two different directions.

Which segment/feature exerts influence over the other?

If the stop is more salient than the fricative, or vice versa, why did opposite developments take place?

If the onset is more salient than the coda, or vice versa, why did opposite developments take place?

If one cluster is otherwise more likely than the other, why did opposite developments take place?

To state it plainly: How are French and English treating these cases differently structurally?

Possible explanations

It would obviously be helpful to know what state the French word was in when it entered English... but in the absence of that information, my ideas so far are as follows:

  1. In French phonotactics, /ps/ is a legal onset cluster, as are other /p/-initial clusters. This is not the case in English. Perhaps the syllable actually breaks or broke before the /b/ and not between /b/ and /s/. Still trying to think of a coherent reason this would produce the data. Perhaps such an argument would go something like this:

    • Considering an isolated syllable segment, [s] causes devoicing assimilation of [b]
    • But in a whole word context, intervocalic lenition produces [bz]
    • If [b] and [s] are closely enough attached by sharing an onset, -voice feature holds out, otherwise +voice holds out
    • This claim may be backed up by the observation that in French orthography, a lone intervocalic "s" generally corresponds to /z/ but in a cluster it is /s/
  2. Old and Middle French had final consonant devoicing due to Frankish influence. I don't know if English did (it would be surprising if not given its Germanic origins and surviving pairs like "half/halves"), but perhaps long-distance assimilation with a devoiced [f] could have influenced the earlier cluster in French and not English...

  3. The given English representation corresponds to most grammatical inflections. The French word corresponds to a particular inflection of "observer" /ɔpsɛʀve/. In that inflection and many others, there is an extra final syllable that carries the stress. Perhaps a stressed syllable's onset is stronger than that of an unstressed syllable. Unfortunately, this line of thinking produces the opposite outcome: the [s] should be strengthened in English and weaker in French. Moreover, other inflections and derivations where the stress shifts (e.g. "observation" in English) don't show any particular difference, to my knowledge. And this assumes a lot about historical stress patterning that I don't know enough about.

Revised explanation

Edit: Here's a revised theory given the commenter's note below that it was thought to already be [ps] in the Latin that would have given rise to French. If we combine the first and third options, we might get something like:

  1. [s] overpowered [b] naturally through, for example, onset strength;
  2. But in variations of the word where the second syllable is stressed, the longer vowel contributed enough intervocalic lenition to turn the whole cluster into [bz];
  3. Such a stress pattern obtained in English but not in Latin ("observare", third syllable stressed) or in the standard French inflection ("observer", third syllable stressed).

Unfortunately, I still don't know enough about the stress patterns of much earlier stages of English and French; if anyone can verify that, we might have a plausible account.

It's been a couple of years since I finished my undergrad in experimental linguistics (and I haven't pursued it at a graduate level despite maintaining it as a hobby), so if any of this indicates the obvious need for further education or a refresher on the basics, I don't mind if you point it out. :)

  • Intuitively I think (1) makes sense, but is /ps/ as an unambiguous onset really found in words other than Greek ones like psychologie? Not that an onset can't be more common in word-medial position than in word-initial position, just a passing thought I had :P Mar 9, 2017 at 5:05
  • 1
    I believe Latin is thought to have devoiced /b/ to [p] in such situations. The reason for thinking so that I remember is that /bs/ also occured word-finally, in words like "urbs," and in this position "g" at least was devoiced, which was marked in writing in words like rex-regem; since "g" and "b" are both the same type of consonant it would make sense if "b" were symmetrically devoiced. Something like that. This doesn't answer your main question, but I thought it was relevant. Mar 9, 2017 at 5:11
  • Thanks for these points; synthesized a possible new explanation from them. :) Mar 9, 2017 at 14:57

1 Answer 1


I believe that Chomsky and Halle's SPE theory predicts progressive assimilation of voice.

For English morphologically simple forms, the only way you can have a weak word initial syllable followed immediately by a stressed syllable (as in "Moˌnonga'hela") is when the initial syllable has a secondary stress (as in "ˌMon'tana"). However, in words with Latinate prefixes (attached with an "=" boundary), we can have unstressed closed weak syllables before a stressed syllable, as in "con='sider".

Because "observe" begins with an unstressed weak syllable followed immediately by a stressed syllable, we know that the "ob-" is a Latinate prefix:

/ob=sɜːv/ => [əbˈsɜːv]

But a phonological peculiarity of Latinate prefixes is that "s" becomes voiced [z] in a voicing environment after a Latinate prefix. Examples are "de=sign", "re=sign" (compare "consign" with voiceless [s]), "pre=serve".

So this voicing assimilation might be expected in "ob=serve".

/ob=sɜːv/ => [əbzɜːv]

(I don't recall the details of the voicing rule given in SPE, so I don't know why it doesn't apply after "con-".)

  • Nice reference to SPE, though synchronic rules always raise questions of diachronic origin, I think -- "So English breaks it down like this. Why did a different pattern develop in French?" :) But on both sides, those other cases of [z] are very compelling points for the intervocalic voicing environment idea; consider also "dis=service" for another [s], and similarly "obstruct", "obstacle" with more voiceless company even after [b]. (You'd think [n] would have a similar effect to [b], though it's at least consistent: "consider", "consist", "consult"...) Mar 10, 2017 at 2:46
  • ...which effect might combine with the point about French considering [bs]/[ps] a more closely cohering cluster than does English, to take a step towards an explanation of the difference? If the [s] devoiced the preceding consonant in a rule that requires them to cohere more closely and applies before the voicing environment, perhaps... Mar 10, 2017 at 2:49
  • Just to revisit the [n] problem, "conserve" vs. "observe" is pretty difficult: why would [n] be less voiced than [b]? But to add some substance to your SPE account, I notice "obsidian". Now I see etymologically that this is one morpheme. This could suggest that the [b] and the [s] are allowed to cohere more closely and hence lead to a similar situation that French already has in "observer" but which is disallowed in English observe" thanks to the inherent "ob=serve" divide. (If this morphological rule is taken into account, it also allows us to dismiss the French [ps] cluster.) Mar 10, 2017 at 2:57
  • Some more data to consider: "observation" where [ps] is at least more possible than in "observe"; "absolute" seemingly [bs] but possibly [ps] (in any case not [bz]); "absolve" unambiguously [bz]! Mar 10, 2017 at 3:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.