Quenya did not tolerate any combinations of voiceless stops, aspirates or nasals before liquids. Many of these were transposed by Quenya metatheses, but any surviving combinations were further altered. One common alteration was syllabification of the liquid, especially for l. In the case of voiceless stop + l, the usual pattern was that the l became syllabic and developed a vowel of the same quality as the preceding vowel: ḷ > al following a, ḷ > il following i, etc. Tolkien described this change in both the Outline of Phonetic Development (OP1) from the 1940s and the Outline of Phonology (OP2) from the 1950s:

On the early sporadic changes by metathesis: tr, tl, kl, ml etc. > rt, lt, lk, lm etc.: see General Account, p. 26. Where the original sequence was maintained before l a vocalic element was very early developed, normally of the same quality as the vowel preceding the stop: thus makla > makal. The [l] was thus early removed from the unvoicing influence of p, t, k (OP1: PE19/40).

On the early sporadic changes by metathesis: tr, tl, kl, kr, ml, &c. > rt, lt, lk, rk, lm &c.: see the General Description. Where the original order of sequence was maintained, l (but not r) became at an early period syllabic, and ḷ, ǝl then developed a full vowel of the qualities a, i, u according to the quality of the vowel preceding the stop: thus makla > makal(a). The l was thus early removed from the unvoicing influence of preceding voiceless stops and aspirates (OP2: PE19/83).

It is not clear what the developments were when the preceding vowel was e or o; perhaps the result was ḷ > al, but just as easily it could be il/ul. In this ancient period, r did not normally become syllabic after voiceless stops:

The vocalic element was not however developed before r at an early period (OP1: PE19/40).

At this (prehistoric) period r did not become syllabic, possibly because it was early more susceptible of unvoicing than l (OP2: PE19/83).

The normal development was that [r] became [s] after voiceless stops and aspirates. However, there was one case where medial r could become syllabic in the ancient period, when it was at the end of a cluster of three consonants:

In a few old formations r, l followed after consonant groups. Here syllabification was early and the development of tr > ts seldom occurred. So √KWET: kwentrō “reciter” > Q quentaro, beside †quenso “minstrel” < kwentso, but tañklă “fastener, brooch” > tañkal(ă), Q tañkal (OP1: PE19/40).

In a few surviving formations of archaic pattern asyllabic r, l had followed after consonant-groups (usually nasal + stop). Here syllabification of r > ar was probably older and occurrences of the unvoicing to s are rare. Thus from base √KWET: kwentrō “speaker” > Q quentaro, beside the archaic form quenso, specialized in the sense “minstrel, reciter”. Syllabification of l always takes place, as in √TAK: tanklă “fastener, brooch” > Q tañkal “brooch” (OP2: PE19/83).

For these three-consonant clusters, metathesis could not occur, and syllabification was the normal pattern with ṛ > ar. These clusters were apparently quite rare, however: the word Q. quentaro mentioned above is the only published example. Such clusters also provide the best-known examples of ancient l-syllabification (such as Q. tankal), again because metathesis was inhibited.

With l, similar syllabification occurred after aspirates:

As after p, t, k the development of l after spirants (aspirates) to syllabic ḷ was early enough to prevent unvoicing, and the normal products are f-l, þ-l, h-l, with intervening vowel taking the quality of the vowel preceding the spirant (OP1: PE19/44).

The groups phr, phl; thr, thl; khr, khl ... As after voiceless stops, the development of l to syllabic ḷ was early enough to prevent the unvoicing of l, and the normal products were f-l, þ-l, h-l, with intervening vowels. On the quality of these vowels see above: C 1 (iv) (OP2: PE19/88).

With r after aspirates, the usual pattern in the ancient period was unvoicing and development to s, as mentioned above. Syllabification also occurred for those combinations of ml, mr that did not undergo metathesis, with results similar to those of syllabification after stops:

Where the original sequence was maintained, after m l became syllabic (as after stops), and the resultant vowel took its quality from the preceding vowel: amla > amala; imla > imila etc. ... mr where not metathesized probably remained later than ml, and the ultimate product was mar (OP1: PE19/47)

mr, ml became syllabic and produced mar, mul. Observe that the vowel generated between m-l was usually u irrespective of the neighbouring vowels: thus imla > imula (OP2: PE19/98).

In the 1930s, Tolkien felt that the syllabification of l after m followed the same pattern as after stops and aspirates, with the resulting vowel depending on the quality of the preceding vowel, but in the 1950s he decided the resulting vowel was always u. There are no examples of ml/mr syllabification other than the ones given above; it seems metathesis to lm, rm was more common (even there the only published example is Q. holmo). Combinations of l, r after n did not become syllabic, since [nl], [nr] became [ll], [rr]. Syllabification of l, r also did not occur after voiced stops either: these combination followed the more usual pattern whereby voiced stops became spirants (PE19/46, 94).

As a last case, syllabification could also occur when a consonant cluster ending with l, r became final after the loss of short vowels in final syllables. Tolkien only discussed these word-final syllabifications in the Outline of Phonetic Development (OP1) from the 1940s, because his revisions in the 1950s (OP2) stopped short of that point:

Syllabic ḷ, ṛ were only produced by loss of ǝ (< a, e, o). ṛ produced ar. ḷ produced il/al/ul according to the surroundings of which the preceding vowel was the strongest influence. Initial ḷ only occurred as reduced form of negative prefix la. It yielded usually a [al-]. On the special vocalic development in Q. final syllables see below (OP1: PE19/56).

Final syllables [rough notes] ... When inflexional in significance stressed[?] ending[?] in CE. weakening[?] of a, e, o > ǝ {(> nil)} after or before continuant or nasal ... after l, r: la > ḷ > il/al/ul. ra > ṛ > ar ... after l, r: la > ḷ > il/al/ul. ra > ṛ > ar ... [before] l, r: ḷ, ṛ > al/il/ul, ṛ (OP1: PE19/66).

-lǝ > ḷ > il/ul/al ... -rǝ > ṛ > ar (OP1: PE19/67, summary of the preceding according to note #173).

Thus it seems final syllabic l, r had basically the same developments as medially: ṛ > ar and ḷ > al/il/ul depending on the preceding vowel. The developments after e/o were not discussed, but based on examples like Q. hecil (< hek-l(a)?), probably they also developed to -il/-ul; hat-tip to Parmandil for this suggestion. There is also one example where initial ḷ became syllabic: in the negative prefix ᴹ✶ḷ- > ᴹQ. al- (but see Conceptual Development below). Aside from the maklă > makal example given above (OP1: PE19/39), the clearest examples of these final developments can be seen with the primitive agental suffix ✶-rŏ (WJ/371) which frequently becomes -ar in Quenya after another consonant:

• √IS [> ?istrŏ > istṛ] > Q. Istar “Wizard, One Who Knows” (UT/388).
• √MAK [> ?makrŏ > makṛ] > Q. macar “swordsman; forger” (VT39/11; VT41/10).
• √SAN [> ?sanrŏ > sanṛ] > Q. sanar “mind, (lit.) thinker” (VT41/16).

The challenge here is that it is often not possible to distinguish these from cases where the ar existed primitively. The above examples could instead be from primitive *ista-rŏ, *maka-rŏ, *sana-rŏ; compare with ✶satarŏ > Q. satar “companion” (PE17/183).

To summarize, in cases where metathesis did not occur:

• l became syllabic ḷ medially after voiceless stops, aspirates and m, and developed into il/al/ul depending on the quality of the preceding vowel (front, mid, back), with the exception of mḷ which always became mul.
• r became syllabic ṛ after m and at the end of (rare) three-consonant clusters; in all cases syllabic ṛ > ar.
• Where l, r became syllabic finally after vowel losses in final syllables, they developed vowels following the same patterns as above.

Tolkien also described a “second wave” of syllabifications for r, l well after the ancient period but with very similar results; these “second wave” changes are discussed in another entry.

Conceptual Development: In Tolkien’s early conception Qenya from the 1910s and 1920s, syllabic ḷ, ṛ were among the vowels of Primitive Elvish, so that any syllabifications of r, l are hard to distinguish from original primitive ḷ, ṛ. He described the Early Qenya development of syllabic ḷ, ṛ in a table appearing in the Qenya Phonology of the 1910s (PE12/10):

[Before]	q	k	k̂	t	p
ḷ short	ul	il	il	il	ul
ṛ short	ar	ar	ir	ar	ar

In the chart above, k̂ represents the palatal series, which elsewhere in this document Tolkien represented with ty or c (IPA [c]). Tolkien also described the development of final syllabic -ḷ, -ṛ (PE12/11):

• -ḷ appears as -il/ul, according to preceding consonant.
• -ṛ appears as -ar, irrespective of preceding consonant.

Regarding the development of final -ḷ, Tolkien first wrote: “-ḷ appears as -il, irrespective of preceding consonant” (PE12/11, note #64). The development of syllabic ṛ was basically the same as Tolkien’s later conception: ṛ > ar, except before palatals, where it became ir. In the case of ḷ, however, the development was usually to il, becoming ul before and after labials like [p] and labialized velars like [kʷ]. Based on the examples in the contemporaneous Qenya Lexicon, it seems that Early Qenya could also tolerate final syllabic ḷ without any vowel when it followed a t, such as: ᴱQ. etl, patl, sutl.

The idea that syllabic ḷ developed into il lingered for some time. This was the usual development in The Etymologies of the 1930s:

• ᴹ✶tekla [> *tekḷ] > ᴹQ. tekil “pen” (Ety/TEK).
• ᴹ✶tankla [> *tankḷ] > ᴹQ. tankil “brooch” (Ety/TAK).
• ᴹ✶ḷ- > ᴹQ. il- “no, not” il- (Ety/LA).
• ᴹ✶makla [> makḷ] > ᴹQ. makil “sword” (Ety/MAK).

For the second and third examples, compare with tankal and al- above, where ḷ > al under the influence of the base vowel a. The fourth example might also have been revised to maklă > makal (OP1: PE19/39), but that one is harder to tell: it could be a contrived example instead. In the 1950s and 60s Tolkien restored Q. macil “sword” (VT39/11; VT41/10), but in the later conceptual stages its Sindarin equivalent was S. megil (vs. N. magol), indicating the il was in the primitive form of the word and not the result of syllabification.

The earlier phonetic developments ḷ > il (and sometimes ḷ > ul) can be seen in rejected notes in the first layer of the Outline of Phonetic Development from the 1940s:

It must be noted the normal articulation of l in Q. was high and without back-hollowing — hence its usual vocalization as il (OP1: PE19/40 note #66).

tl, kl > ty or to syllabic til, kil. But note after p the palatal character of l was not present and it could not develop to py. The development of pl was thus to syllabic pul (except sporadic, mostly ancient, case of metathesis to lp). Cf. ml > lm or mul (OP1: PE19/43 note #82).

These rejected notes also indicate that sometimes tl > ty (as tr > ts) but this seems to be a transient idea. In the final revisions of the 1930s, it seems that Tolkien abandoned ḷ > il/ul and replaced it with the phonetic development discussed above whereby the resulting vowel was derived from the quality of the vowel appearing earlier in the word. This remained the case in the 1950s, except that Tolkien restored the ḷ > ul development in the specific case of mḷ > mul.

See AQ. [r], [l] often became syllabic in clusters for discussion.

See AQ. [r], [l] often became syllabic in clusters for discussion.

See AQ. [r], [l] often became syllabic in clusters for discussion.