If we really want to understand Primitive Elvish roots, we first need to understand what sounds were used in the Primitive Elvish language. Tolkien like to arrange his sound systems into tables of “series” and “grades”. The series were organized in columns by the area in the mouth where the sound was formed (front, middle, back; labials, dentals, velars) and the grades in rows that indicate the method of articulation (stops, nasals, etc.).
Tolkien’s sound system for Primitive Elvish was more or less established by the middle of the 1930s, with some minor variations discussed below. You can see the following table appear in roughly the same form in various places from that period forward, for example in Parma Eldalamberon #18, pp. 59 and 82:
|5.||w||l, r||y, ʒ|
The first series, which Tolkien generally called the “p-series”, were labial sounds formed closed to the front of the mouth using the lips. The second series, the “t-series”, were dental sounds formed in the middle of the mouth with the tongue placed just behind the teeth. The third “k-series” were velar sounds with the tongue placed in the back of the mouth.
The first grade (p, t, k) were voiceless stops, formed by interrupting the flow of breath. The second grade (ph, th, kh) were similar, but with extra aspiration, and “h” sound added after the stop. I usually refer to this grade as the “aspirates”. The third grade (b, d, g) are voiced stops, similar to the voiceless stops but with extra resonance in the vocal cords of the throat.
The fourth grade (m, n, ñ) were the nasals formed by routing air through the nasal cavity rather than the mouth. The fifth grade Tolkien simply called “oral continuants” for various consonant sounds formed by more-or-less open air flow in the mouth. The sixth grade was a bit unusual, in that it contained only a single sound, the sibilant “s”.
The same table using IPA notation for the consonants would be:
|5.||[w]||[l], [r]||[j] (“y”), [ɣ] (“ʒ”)|
Tolkien often grouped the oral continuants y, ʒ, w ([j], [ɣ], [w]) together, and in fact he usually put “w” in both the p-series and the k-series, because it was often combined with velars as kw, gw and ñw ([kʷ], [gʷ], [ŋʷ]). The sound ʒ ([ɣ]) is one that is hard for English-speakers to pronounce. It is a voiced velar spirant, a sound that doesn’t exist in English at all. It sounds like the “ch” at the end of the German name “Bach”, but with extra resonance in the vocal cords. Fortunately for us, this sound disappeared in the Elvish child-languages, so its pronunciation is mainly of theoretical interest.
The primitive sounds ʒ ([ɣ]) and s actually existed in two variations, a “voiced” and “unvoiced” variant depending on whether they were adjacent to a voiced consonant (voiced stops, nasals) or unvoiced consonant (voiceless stops and aspirates). The voiceless variant of ʒ was [x] (like “ch” in Bach above) and the voiced variant of s was [z]. From the perspective of the speakers of Primitive Elvish, these were the “same” sound that simply manifested in different ways depending on context, but these two manifestations matter because they developed in different ways in the child languages.
Speaking of which, it’s worth talking about how these sounds developed in the child languages.
Voiceless Stops: The voiceless stops p, t, k survived unchanged at the beginning of words in both Quenya and Sindarin. In Quenya they usually survived in the middle of words as well, but in Sindarin they usually went through the same sound changes as in “soft mutation” in the middle words, become b, d, g. This phonetic effect is, in fact, how Sindarin’s soft-mutation rules arose.
Aspirates: In both Quenya and Sindarin, the aspirates ph, th, kh developed into voiceless spirants f, þ, ch ([f], [θ], [x]). Somewhat confusingly, Tolkien often used the combination “th” to represent the voiceless dental spirant þ (the sound at the beginning of English “thin”) in the child languages, and sometimes used “ph” in Sindarin for “f” in the middle or end of words. The letter combinations “ph/th” are always aspirates in Primitive Elvish, but are voiceless spirants in the child languages.
In both Quenya and Sindarin, the sound “ch” ([x]) further developed to “h” at the beginning of words. In Quenya, this change also occurred in the middle of words, with the [x] sound surviving only in a few combinations, notably before before the dental “t”. Tolkien spelled this combination “ht” in Quenya, such as Q. ehtele [extele] “spring, issue of water”.
Late in Quenya’s history, the “th” sound became “s”. Many of the “s” you see in the middle of words in Quenya were originally the aspirate “th” in Primitive Elvish, such as in Q. hísë “mist”. In my Eldamo dictionary, Quenya words where an “s” was originally a “th” are marked with a [þ] sign; this matters because the original sound is retained in tengwar spelling even it isn’t in pronunciation. Some Elvish scholars (and some Neo-Quenya writers) restore this sound in pronunciation as well, since it is a sound the Noldor relearned when they met the Sindar.
Voiced Stops: In Sindarin, the voiced stops b, d, g survive at the beginning of words, but in the middle of words underwent “soft mutation” to the voiced spirants bh, dh, gh ([β], [ð], [ɣ]). In the case of “bh” it further developed to “v”, but in the case of “gh” ([ɣ]) the sound generally vanished, just as g vanishes (marked ’) in soft-mutation: S. galadh “tree” and i ’aladh “the tree”.
In Quenya, the primitive g vanished in both the beginning and middle of words, surviving only rarely in the middle of words in combinations with other consonants, notably “ng” and “nc” ([ŋg], [ŋk]). The primitive b became v in both the beginning and middle of words as well. The primitive d went through some peculiar changes in Quenya. At the beginning of words, it became l. In the middle of words, it survived only in combinations like “nd, ld, rd”. In isolation it became a voiced spirant dh [ð], which further weakened to [z] and ultimately became [r].
Nasals: At the beginning of words, the nasals n and m survived in both Quenya and Sindarin, and ñ survived at the beginning of words in Quenya at first, but in another late phonetic development it eventually came to be pronounced the same way as “n”. The ñ spelling was retained in tengwar, however, and in my Eldamo dictionary Quenya words where “n” was originally pronounced “ñ” are marked [ñ]. As with [þ], the ñ-sound is restored by some Elvish scholars and Neo-Quenya writers.
In the middle of words, the ñ-sound usually vanished in both Quenya and Sindarin, surviving only in combinations like “ng”, “nc” ([ŋg], [ŋk]). In Sindarin, the sound vanished initially as well, though it reappears sometimes as part of soft and nasal mutations. In Sindarin n survived unchanged medially, but m usually underwent soft-mutation in the middle of words, first to “mh” (nasalized [ṽ]) and ultimately to [v].
Oral Continuants: The oral continuant [w] usually became [v] in Quenya, both initially and medially, though it could survive in combinations like “lw” and “rw”. There are a few example of it surviving medially and initially as well, but these are all cases of an ancient combination [gw], where the [g] was lost leaving only the [w]. Tolkien sometimes said these [w] from [gw] should also eventually evolved to [v] in Quenya (this sound change occurred after any original primitive [w] had already become [v]), so these remnant “w” may be archaic forms in words like Q. wilwarin “butterfly” and awalda “moved, stirred, excited”. Some Neo-Quenya authors update them to “v”, especially medially: vilwarin (or even vilvarin) and avalda.
In Sindarin, w often survived medially but became [gw] initially.
In Quenya primitive y ([j]) usually survived, but in Sindarin it became the vowel [i] medially, surviving only at the beginning of words. Since Sindarin uses the symbol “y” for a vowel, the English y-sound is represented initially in Sindarin as “i”. For example S. iant [jant] “bridge” would likely be written (and pronounced) “yant” if it were an English word. In the middle of words, Sindarin “i” is always a vowel.
In both Quenya and Sindarin, primitive ʒ ([ɣ]) vanished medially with various affects on surrounding vowels. In Sindarin it vanished at the beginning of words as well, but in Quenya an initial ʒ became “h” (see further below).
Both r and l survived more or less unchanged in both Quenya and Sindarin, but in Noldorin (the precursor to Sindarin from the 1930s and 1940s) any initial primitive r and l were unvoiced as “lh” or “rh”. Thus, Tolkien had N. rham “wall” but S. ram “wall” from the primitive root √RAMBA (Q. ramba). An unvoiced initial “rh” and “lh” sound did appear in Sindarin, but only as a development from initial sound combinations like primitive khl/khr and sl/sr. Most Neo-Sindarin authors update Noldorin forms with initial rh/lh to simple r/l in Neo-Sindarin, but you have to be careful. Sometimes Tolkien retained the Noldorin form of words and revised the Primitive Elvish instead. A good example of this is N. and S. lhaw “ears” (with irregular singular form lheweg “ear”), from the root ᴹ√LAS “listen” in the 1930s but from the root √SLAS in the 1950s. You can see this more clearly in the revised Quenya forms: ᴹQ. lár “ear” but later Q. hlas “ear”.
Sibilant: The sibilant [s] survived at the beginning of words in both Quenya and Sindarin, but in Sindarin it underwent soft mutation medially to [h], surviving only in combinations like [st]. In Quenya it went through a series changes similar to medial [d], generally developing into [z] and then [r] and surviving only in combinations like [sk/st] and [ks/ts]. This means medial Quenya [r] could have been derived from primitive [r], [s] or [d], and a medial Quenya [s] in isolation was actually derived from a primitive [tʰ]!
To summarize, here is a chart of how primitive sounds developed in Quenya and Sindarin. Where sounds appear with a “/” they distinguish initial and medial developments, and a [’] represents a vanished sound. First, here is a repeat of the Primitive Elvish sounds:
|5.||w||l, r||y, ʒ|
Here is how they developed in Quenya:
|2.||f||s (< þ)||h|
|3.||v (< b)||l/r (< d)||[’] (< g)|
|4.||m||n||n/[’] (< ñ)|
|5.||v (< w)||l, r||y, [’] (< ʒ)|
|6.||r (< s)|
And in Sindarin:
|5.||gw/w||l, r||y (“i”, IPA [j])/i, [’]|
Note that the resemblance of the soft-mutation rules to the initial/medial development of primitive consonants in Sindarin is not a coincidence: these phonetic changes are the major reason these soft-mutation rules came to be.
Conceptual Developments: Starting in the mid-1930s, Tolkien’s idea for the phonetic inventory of Primitive Elvish were remarkably stable. Early on, he considered allowing a variation [v] of [w] as part of Primitive Elvish (a remnant of his ideas from the 1910s and 1920s) but he soon dropped this idea.
The only major change was that sometime in mid-to-late 1950s he decided that the sound ʒ [ɣ] was actually h in Primitive Elvish (a weak voiceless spirant [x]). He still had voiced and unvoiced variants of this sound ([ɣ] and [x]); the only change was that the consonant in isolation was pronounce [x] instead of [ɣ]. This sound had all the same developments as ʒ: vanishing in Sindarin and surviving only initially in Quenya as “h”. He updated some roots such as √ƷAN “extend” to √HAN “increase”. This change has only theoretical implications, since the derived words in both Quenya and Sindarin remain unchanged.
The above discussion only addresses basic consonants and their sound changes in isolation. Combinations of consonants developed in various other ways, but that’s way too much to cover a single article. The phonetic inventory of Early Primitive Elvish from the 1910s and 1920s was also quite different. But those are topics for another day.