I learned from the comment to the answer to this old question that Latin has lost the consonant S before voiced consonants. In the linked post this was used to explain the observed pattern that the prefix ex- becomes ē- before voiced consonants.

Does this effect work also across word boundaries? That is, if a word ends in S and the next word begins with a voiced consonant, does something happen to the S? Would it be lost in either pronunciation or spelling, or would such situations perhaps be avoided?

My guess (and only a guess!) is that such loss of S was no longer active by the classical era, and the Romans would not have observed any difficulty having S followed by a voiced consonant. A quick glance at the Aeneid shows no sign of avoiding this situation. I would like to know if we know something about the matter. For example, if we have some evidence that the mentioned law became inactive at some known time or that laws of this kind do typically not work across word boundaries, that would reasonably answer my question.

  • 1
    One way to examine this is in line with the previous question to look at the choice between e and ex (the preposition, not the prefix) before a voiced consonant. A quick PHI search finds lots of examples of ex before a voiced consonant, which would be evidence that the s-loss rule did not apply across even weak phonological boundaries.
    – TKR
    Apr 7, 2017 at 1:48
  • 2
    Why “ex nihilo” instead of “e nihilo”? is relevant to this question as well, I think.
    – Asteroides
    Apr 7, 2017 at 4:14
  • @TKR That's a good observation. Would you like to add it as an answer?
    – Joonas Ilmavirta
    Apr 7, 2017 at 15:44

2 Answers 2


Alongside the prefix ex- ~ ē- there is the preposition ex ~ ē, which offers a way of testing whether the same s-loss occurs before a voiced consonant at a word boundary. If it consistently does, we should expect to find ē before voiced consonants and ex otherwise.

This is not the case, as the question linked to by sumelic shows: there are phrases like ex nihilō, ex lege, where ex appears before voiced consonants. Searching PHI for e.g. #ex r yields a large number of results, and likewise with other voiced consonants.

Of course it's possible that the effect did occur across word boundaries but not consistently -- digging deeper into the statistics of various combinations might shed some light on that. But as far as it goes, this does suggest that s-loss did not operate even across the weak phonological boundary that separates a preposition from its object, and therefore a fortiori probably did not operate across other word boundaries.

(To the question implied in your last sentence, such phonological rules certainly can operate across word boundaries, so the fact that this one apparently doesn't is not due to any kind of linguistic universal.)


In early classical poetry, a final /s/ did not necessarily make position in Vs CV. This is common enough in Plautus, especially when combined with Brevis Brevians, and seems to diminish in frequency with time. It is still present in Lucilius, and the very last instance is in Catullus. I cannot run an analysis of the frequency of voiced vs unvoiced consonants, because I don't have a scanned corpus of Plautus and because dramatic meters aren't my strong suit. All examples below are with voiced consonants or unvoiced fricatives, but of course a sample size of 5 instances is not exactly the strongest evidence.

Some examples:

A senarius in Plautus, Casina 137

sine ama|bo ted | ama|ri, meŭ̯s | festŭ̱s | dies

3 consecutive hexameters in Lucilius 4.149-151 Marx

Aeser|ninu̯s fu|it Flaccorum munere quidam

Samnis, | spurcu̯s ho|mo, vi|ta illa | dignu̯s lo|coque:

Cum Pacideiano componitur, | optimu̯s | multo

A pentameter in Catullus CXVI, 8

at fix|ŭ̱s nos|trī̱s || tu dabĭ̯s | supplici|um.

  • +1 Thanks! Within a word the loss of S would be compensated by vowel length. I don't know if this shoul be expected across word boundaries.
    – Joonas Ilmavirta
    Apr 7, 2017 at 15:01
  • @JoonasIlmavirta I hadn't thought of the linguistic implications of this, I was just trying to corroborate your guess with some attested data. But it is a very good point that this may be at odds with moraic theory - although I do mention, coincidentally really, that the moraic process Brevis Brevians may have an inflluence on the occurrence of the phenomenon.
    – blagae
    Apr 7, 2017 at 18:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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