The first question about this topic would be: Why would a word-final consonant have to be syllabified in an onset, and not in a normal post-nuclear complement (Coda) position. After all, we have this position in word internally, and this Coda is so important as it differs some languages to others called "CV languages." First of all, Coda is an old term, back to the time that all consonants which occur after a nucleus could be simply attached to the rhyme in the form: (1), where C could even accommodate 2 consonants when N is neither a long vowel nor a heavy diphthong. Nowadays, Coda is more accurately called complement, to accentuate the fact that it is not a constituent, while onset and nuclear are. Why can't a word end in a consonant? If we observe the way languages behave, so many exceptions seem to occur in the word-final "Coda", every rules about how it should normally behaves is so frequently broken that leads us to the question whether this "Coda" could be defined as such. 1- The case of vowel shortening rule. Basically, long vowels are shortened in a closed syllable (Kaye).
And here are some examples to illustrate this proposition. Ex: French, chat [Sa: ] and chatter [Sat]Yawelmani, [sa: pit] and [sap nit]In both cases above, the vowels are shortened to accommodate a consonant in its complement position, because we know that there is no long vowel or heavy diphthong in a branching rhyme. But in cases when the consonant which occurs after long vowel is also situate at the end of the word, this rule is not observe. Ex: French, vert [ve: r] and verdure [verd u: r] English, keep [ki: p], and green [gri: n] (2) Those examples shows violation of the above rule stated where no coda could be accommodated into a rhyme with long vowel or heavy diphthong.
2- The case of word-finally consonant cluster. Words in English like kept, child, find... pose several problems in phonological analysis. First problem would be the rule about "no branching coda in a branching rhyme", the second would be the nature of consonant clusters; consonant like pt, ld, rt... are not the normal consonant cluster so-called well-formed cluster, the well-formed cluster in a language could be easily spotted in a branching onset. If we could not find them at the beginning of any English word, there is a big chance that they are not a good cluster, so we can separate them into two different syllable if found in middle of a word, the natural order of two consonant occurring next to each other must be respected, re-syllabification is not possible.
But in word finally, where could the second half of this ill-formed cluster position itself? The easiest solution is to stick it with the rest, giving us something like this: (3) To sum up those affirmations, the early Coda licensing would be the fact that the coda position in a syllable could behave differently if it occurs in word finally: - No long vowel in a branching rhyme with the exception of word-finally coda. - No branching coda in a branching rhyme with the exception of word-finally coda. - Re-syllabification is not possible due to respect of well-formed consonant cluster, with the exception of word-finally coda. 3- Explanation... Early explanation for this phenomenon had been introduced in the 70 s by Halle&Vergniaud when analyzing the Germanic languages. They propose a new extra-rhythm al constituent called the appendix to explain a different problem, the skewing of consonant distribution in Germanic languages in word-final position...
In the 80 s, Charente use this notion of appendix to assume that a final consonant is not situated in the rhyme, but in the appendix, therefore there was no violation of rules. Ex: vert, and vert e in French, (4) So, until late the 80 s, a word could end with a consonant... However, in 1990, Kaye has gone further to understand this phenomenon, as he thinks that the notion of appendix, while it satisfied the need explaining how vowel shorten in some languages, the 90 s came with fuller theory of phonological government, this analysis is no longer sufficient. In the 90 s, a theory about phonological strings is proposed, called phonological government, after Kay, "it defines under what condition two phonological positions may be viewed as adjacent." And that those positions are linked in a relation called Licensing, which obey a certain principle. - All phonological positions within a domain except one must be licensed, and this unlicensed position is the domain's head. - Government is one form of license, parameter is another form- Condition for A to governs B is: a.
A and B strictly adjacent (no position could intervene between them on their projection) b. A and B must obey a strict given direction (constituent government: left to right; trans constituent government: right to left). By deduction, we could deduct two more theorems: i. Bin arity theorem: all syllabic constituents are maximally binary (otherwise it would violate the strict directionality or strict adjacency), and ii. The head of a domain is always a nucleus (due to its far-right position). Within a constituent, the governor need to be more complicated in its composition than its governed in other to justify its position, a good governor is often a stop or a fricative consonant.
With those new rule, words like [ki: p], [kept], [ve: r], or [vert] have to be syllabify as: (5) It's no longer possible to leave i: and p in the same rhyme, therefore p has to take the place of an onset of the next syllable. (6) As we have the choice of separating or attaching pt and rt together, if we would want respect the "good" order of a consonant cluster, we will separate them, the last consonant then becomes onset for the next syllable, and as an onset always need a nuclear, we would say that this nucleus is licensed to be empty by parameter; to differentiate with some languages like Japanese or Italian which do not permit this position to be empty, where a phonetic interpretation would then be realised. In the government phonology, "r" and "t" would also be separated because: - If occur together in a branching onset, then following the constituent government rule, "r" would be expected to govern "t", but due their respective composition, this is not possible as "t" is a more complicated consonant than "r", and is in fact expect to govern. - When separated, they belong to different constituent of the word, and as the direction of trans constituent government is the mirror image of the constituent government, this hypothesis fits perfectly. Using government phonological could justify why words like green and keep could syllabify, but does this theory expand to the rest of them, making that rule universal? What prevent words like "cotton" from ending with "n"? What push people to propose that words like "cat" is actually bi-syllabic? While observing how English and French behave, we could account some very interesting facts that could tell us more stories about this issue. 1- Suffixes.
Nearly all suffixes in English are in the form V-, for example ing, y, able... and many more. cut to cutting, fish to fishy (7) Those data lead us to believe that the last consonant does not belong to the first syllable, if we insist in leaving it there in words like [cut] or [let], it's by pure convenient, because phonologists do not believe in re-syllabification, and letters like t, d, k... are not good for the Coda position neither. We observe the same phenomenon in French. Route to, citron to citron elle So we could see now that this Coda position at word-finally is not a very stable position, unless syllabified in an independent onset, followed by an empty nucleus, ready for any linguistic morphology, and as in phonology, we aim to make rules, not mere observation, we would propose that this is universally true, and that all words must end with a nuclear, which could be licensed to be empty by parameter in certain languages.
But could Coda be tolerated in word-internal? 2- Prefixes. Prefixes in languages often take the form of -C, like un, im, il, sup, sur... In English we could observe words like true and untrue, legal and illegal In French, we could look at engager and des engager, re elle and sure elle. Those date affirm that the position of complement (Coda) is still possible within word-internal, but it seems to depends a lot on the onset of the next syllable, suffixes tend to adapt with the next onset, we could even push it further by quoting Jonathan Kay that: "post-nuclear (coda) position must be licensed by a following onset." So for the moment, this position is still believed to exist within word-internal.