Historical Phonology of Goldogrin

Roman Rausch

Oct. 8th 2007

And thus said Littleheart the Gong-warden once upon a time: ”Gnome-speech,” said he, ”is enough for me — did not that one Eärendel and Tuor and Bronweg my father (that mincingly ye miscall Voronwë) speak it and no other?”
The Book of Lost Tales 1, p.44




Tolkien’s earliest linguistic creation includes two huge dictionaries of the main languages – the Qenya Lexicon (QL) and the Gnomish Lexicon (GL). The QL begins with a detailed treatment of its historical phonology – the sounds of the common ancestor Common Eldarin and their evolution towards Qenya. There is no grammar of Qenya given. On the other hand, the GL stars with a grammatical introduction where only a few remarks on phonology are made.

The aim of this article is therefore to give an overview of the sound changes in the evolution from Common Eldarin to Goldogrin. References to the GL will be given simply by the page number in brackets.

1  The vowels

There is an overview of the Goldogrin vowels and some developments in the chart Gnome Vowels (GV, PE15:13). However, as the editors already point out, not all the developments shown there actually match those of the Lexicon, so that this chart must have been written earlier. Some important notes are also given in the introductory grammar of the GL.

1.1  Long vowels

Page 14 gives ā, ē, ō, ī, ū, > ō, ī, ū, ai, au and this can be verified in the Lexicon itself, here are some examples:

§ ā > ô [GV: â > â]

§ ē > î [GV: ê > î]

§ ī > ai [GV: î > î]

§ ō > û [GV: ô > û]

§ ū > au [GV: û > û]

This change is prevented whenever a labial follows:

All these developments can be also immediately comprehended by a look at the past tenses of verbs that are often formed by vowel-lengthening and addition of -i. They have already been discussed in [1].

According to the grammar long vowels originally did not change in monosyllabic words, but the changed forms were reintroduced in most cases by analogy from declensions (where an addition of syllables was made) (14). However, it is not easy to come up with clear examples. Monosyllabic words with unchanged long vowels could have had short vowels etymologically, lengthened only in recent times. Words where the long vowel has undergone change could have had a final vowel that has fallen away (as in the above examples).

Limiting oneself to words with a final vowel (clearly without older affixes) one can find unchanged forms:

And some changed forms:

The fact that mar and môr probably conceal two roots and a confusion (56) seems to reflect the idea that monosyllabic words remained unchanged. Otherwise Môr ’the actual Earth 1) the whole of the lands, opposed to the seas 2) the earth, opposed to hell or heaven’ beside Mar ’Earth, ground, soil’ could be explained by a lengthened form *mār without the need of a second root.
Compare also the glosses:

1.2  Diphthongs or combinations of two vowels

GV mentions two developments of diphthongs – in accented and unaccented position. The unaccented changes are more closely treated in 1.4 and 1.5 below. Here are the developments which can be extracted from the GL proper compared to those from GV:

§ Several times we see au > ô/o, although au is otherwise usually retained [GV: au > au]

In the first case au > o might be due to simplification in front of a cluster. Compare ai > e and see 1.3.

§ iu̯ > io; once iw is retained [GV: iu > iu]

§ eu > io, iw [GV: eu > iu]

§ ou̯ > ô/o [GV: ou > ô]

§ several times ai > ê/e, although ai remains unchanged most of the time [GV: ai > ai]

Bailmoth > Belmoth might be simplification before a consonant cluster here as well. Compare au > o above and also see 1.3.

§ oi, ōı̯ > ui [GV: oi > ui]

§ ei > e/ê or ai [GV: ei > ê]

§ ı̯e > e, ı̯ē > î (perhaps simply by dropping ı̯?)

§ ao > o

1.3  Final vowels

Goldogrin words usually end either in a consonant or in the vowels -a, -i. Final -u is quite rare (for instance duru ’wood, a pole, beam or log’ (31), gurthu ’death’ (43), gadu ’joined, connected’ (36)).
Final -e is apparently not usually allowed at all except in monosyllables like le (53), the archaic form of li; or ’mark, line — track — path’ < teʒ́e (69). Final -o occurs also in monosyllables or in the combination -io.

GV tells us that the vowels vanished all finally; but in the Lexicon this seems to be true only for short ones. Final long vowels change as following:

A note in the grammar assists that phonologically, -u, -i only refer to -ū, -ī (14). However, evidence for > -i can be found: Above all, there is the concrete example Ou̯lē > Ôli (62). But this corresponds to Q. Aule (i) (QL:34) – the bracketed i indicates that it is declined Auli- and should thus derive from *Auli < *Ou̯lī. In fact, the Goldogrin word was changed from Ôla. There is also ı̯ōı̯ē > *gōi > gui, past tense of gôtha- ’possess, have, hold’ (42) indicating that the past tense suffix -i derives from -ı̯ē (compare Q. past tense -ie, -ye < i̯ie in PE14:56).

Looking for a solution we may consult GV: æ̂ > ê and ê > î is distinguished there. And from the example Ŏrŏmǣ > Orma (63) we might deduce that such a difference in quality between long open ǣ and long closed ē led to:

If so, the pair æ/e must have been developed at a time later than Primitive Eldarin, because it is not on the list of its vowels in QL:3. Furthermore it would be not the same conception Tolkien had when he wrote the grammar on page 14.


1.4  Unaccented variations

One important point we can learn from GV is that one has to distinguish accented and unaccented development (the latter is there given for diphthongs).
QL:4 tells us that in Primitive Eldarin the accent was free and was in closest connection with etymology and morphology playing a considerable part in grammatical usages and in word formation.
In plain text this could mean that in compound formation the accent remained on one of the respective parts.

From QL:3 we learn that some slight assimilative alternations and some later dissimilative seem to regulate variance between o – u : e – i and occasionally e – a – o, in unaccented syllables especially in case of agglomeration of short syllables.

The first two variations can be found among others in Goldogrin:

§ o | u

§ e | i

More specifically we are told that e > i before before accented original -i, after one consonant (38). This is attested for (indicating stress with a grave accent):

Two other examples seem to follow the same pattern, or perhaps the more general e-i-variance:

§ wa, a | o

Tolkien states that unaccented wa (> *wo) > o, especially the prefix gwa- ’together, in one, etc.’ (40) therefore often becomes go- (and sometimes weakens further to g-):

There is no explanation given why gwamod is more correct than gomod, but we may assume that it is because gwamod is stressed on the first syllable and wa should properly remain unchanged. Gomod might be explained by analogical levelling, assimilation or a later stress shift – note that stress is explicitly marked in gomód (41).
See also 1.5 for a, u > o in the ultimate syllable.

§ ai | e

While the diphthong ai is unchanged most of the time, ai > e/ê is sometimes seen (1.2). According to GV ai gives e/i when unaccented and this could be the explanation for some examples:

§ au | o

Also according to GV au unaccented > o/u. This may be the explanation for the variation au/o in the ultimate syllable in some examples:

§ ui | wi

According to GV unaccented oi, iu > we/wi. The variation ui/wi can be found in the ultimate syllable in several forms throughout the Lexicon:

§ ə

The schwa occurs in some examples:

The indication of stress in əƀánnā̆ here suggests that ə vanishes in unstressed position with ı̯ə- > i- (compare *ı̯i > i-, 2.1).

§ long/short vowels

Long vowels are presumably shortened in unstressed position. For instance, ’av. away, off, as excl. begone!’ has the unaccented vowel relengthened. Also:

1.5  Unaccented a, u in the ultimate closed syllable

In a context with declensions Tolkien states that universal[ly] -an unaccented > on (14). Later on we get to know that adjectives in -a, -i, -u change to -on, -in, -on with addition of pl. -n (15). Hence it appears to be a general rule (albeit with a couple of exceptions as it will be seen) that short a, u change > o in the ultimate, closed (and unaccented) syllable:

Note that there is no change in comar. Tolkien explains it:
Comar is direct from nom. + r. The rest [comon, comar] are generalized from -ā words. The only form true phonologically and still common that is not derived from the far commoner -ā words is the -oth plural of -o words as goldoth.

Hence, one could summarize it as following: In an open syllable the vowels change -ā, -ē, -ū > -a, -a, -u. In a closed syllable in unstressed position the development seems to go -āC > *-ōC > -oC; -ēC > -īC > -iC; -ūC > *-auC > -oC. The resulting variation of -a(C), with -oC is then by analogy extended to a from ē as in coma < *kwāmē (Q. qāme), genitive comon. Comar is formed by a different analogy – to the nominative singular form.

In the Lexicon itself we find that a, u mostly become o in the unstressed closed syllable. It is not alway clear whether it is due to them having been long:

Especially noted should be the agentive suffix -or correspondung to Q. -ar, the frequent noun ending -os corresponding to Q. -as(se); another noun ending -oth, and the adjectival ending -og. When the stem vowel is e or i, endings like -eth, -ith; -eg, -ig are often found instead. See Appendix A on derivational suffixes.
The participial/adjectival ending -ol probably also belongs here, since all verbs end in -a or rarely in -u.

Some exceptions or variations are encountered:

Otherwise this change apparently does not happen in compounds where the second part is a recognized monosylliabic word, like goldobar, goldomar Gnomeland (41) with bar ’home, -ham’ (22), narcolas ’The fall of the leaf, Autumn’ (59) with lass ’a leaf’ (52). surfang ’a long-beard’ (68) with fang ’a long beard’ (34), udrug ’intractable, fierce, untamed’ (73) with trug ’easy to handle – convenient, tractable, docile’ (71) and so on.
Some variation seems to be found with -gar ’place’, however:

2  The consonants

2.1  Initial combinations

Initial b-, d-, g- (gw-) sometimes correspond to Qenya unvoiced p-, t-, k- (q-):

Probably original b-, d-, g- were devoiced in Qenya or there were originally variants of roots with a voiced and unvoiced stop in existence. For instance, beside dal, dalech, daleg etc. Goldogrin also derives words with t-, as tâl ’foot’ taltha ’foot (of things), base, pedestal, pediment’, taleg, taloth ’a (foot)path’ (68,69). In Qenya TALA yields a set of words having to do with ’foot’ and another one connected with ’weight, burden’, probably due to a blending of *TALA and *DALA.
Tolkien explains the correspondence of gwin/qin by a derivation qiu̯i > g’w, not gu̯- (45), the latter would have given ’w- in Qenya (as ’wendi ’maiden’ < GWENE (QL:103)). Hence, Qenya devoices initial g- in this case. On the other hand G. gu̯or- corresponds to Q. kor- (47) and G. gwar- to QARA without an explanation.

Note also that b- usually yields Qenya v-. Finally, GLS list kalda > gâl, kāle > gaul, where the initial consonant becomes voiced (or permanently lenited?) in Goldogrin.

§ dı̯-/ı̯- > g-, but *ı̯i- > i- or gi-; once ı̯ə- > i-:
Hence, no initial ı̯- occurs in Goldogrin. Compare QL:15 for the variation j/d͡y (the latter probably being a voiced palatal stop) in the primitive language.

§ sı̯- > χ̑(ı̯)- > h- merging with χ- > h-:

Note that ı̯ may become syllabic, and then s- is retained. With a different contraction of the root SAYAP > *SAY’P rather than S’YAP no combination sı̯ is created:

§ gu̯-, ŋw-, u̯- > gw-

It is apparently such a favoured combination that initial ui- becomes gwi- by analogy: uivrin > gwivren ’barren’ (both forms in existence) (46,74).

§ l-, r- > gl-, gr- and r- > dr- is a root strengthener, only in nouns, adjs., and denom[inative] verbs (40)

§ nr-, mr-, vr- > dr-, br-, br- and similarly nl-, ml-, vl- > gl-, bl-, bl- (symmetric except for no **dl-)

2.2  Sonorant + voiceless stop

The sonorants r, l, n may cause spirantization of the following unvoiced stops: lp, lt, lc > lf, lth, lch etc. In other cases, the respective combinations remain unchanged.
In final position the spirantization is often prevented or else the spirants are restopped just as original spirants, (for example √lef- > lemp(2) ’a half’ beside lemfin, lemfa ’halved, in half’ (53) or mort (-th·) ’1) warrior, champion, hero’ (58)). A rarer possibility is lenition.

§ l with medial spirantization:

§ l without medial spirantization:

§ l with final restopping:

§ l without final restopping:

§ r with medial spirantization:

§ r without medial spirantization:

§ r with final restopping:

§ r without final restopping:

§ n (m) with medial spirantization; sometimes ngh for nch:

§ n without medial spirantization:

§ n with final restopping:

§ n without final restopping:

-not attested-

This is not an exhaustive list, just some representative examples – these combinations are very common. Some quantitative results can still be given: Final stops are much more common, final spirants appear only as -rf, -rch, -lf (-lph) and never in a combination with n (**-mf, **-nth, **-nch).
Medially, the combinations without spirants are just about as common as those with spirants, but seem to appear more often in the second half of the Lexicon. Emendations, also in the first half, were carried out in both directions but mostly in favour of the spirants:

§ falchos/-or << falcos/falcor (33), famfa << fampa (33), ganthi << ganti (36). gwanthi << gwanti (46). halcha- << halca- (47). helchor << helcor (48), Inthavros << Intavros (51), lenchos << lencos (53)

§ donti << donthi (30)

This has also been done for verbs ending in -ta, as faltha << falta (33), glantha- < glanta (39). gwintha- << gwinta- (46), hantha, hanthi << hanta-, hanti (48). But we cannot be sure what was intended here, as there are evidently two different verb suffixes -ta and -tha (see Appendix A), so Tolkien maybe just switched one for another.
Perhaps he hesitated about this part of Goldogrin phonology and went through the Lexicon altering the forms but maybe changed his mind again and abandoned it. At least the two plurals tenthin and tentin show how two variants may coexist; and also the entry gwinta- ’see’, past tense gwanthi (earlier gwinta-, p.t. gwanti).
Of particular interest is the entry:

§ bartha- ’change, exchange, alter’ {later barta-} (22)

The deleted note indicates that Tolkien at one point imagined a later medial restopping.

2.3  Sonorant + voiced stop or spirant

§ The usual change of nd is to -nn- medially, although unchanged -nd- is also found. In final position there is a variation between -nd, -nn, and -n. Similarly -nđ- > -nn- is usual with the same variation finally:




There is no further assimilation whenever there is a compound:

Especially the last comment suggests that nd usually remains only at a boundary of a compound and shifts to nn whenever it is a combination within one root. Gonnolin would be formed with gonn ’great stone, rock’ (*gondo- > *gonno-) and the suffix -(g)lin ’sound, voice, utterance’. But even so, this conception was not carried out consequently, as nandin, nandir, nandor, indos show.

§ Medial -mb- behaves just the opposite way of -nd-, it seems to be more often retained. A couple of entries have -mb- > -mm- that was externally changed from -mb-, perhaps an unfinished attempt of revision?
In final position -mb > -m. There is no variation as in the case of -nd, on the contrary – the single consonant is reintroduced medially by analogy. Also mƀ > mb:

Note, however, that it is difficult – in fact practically impossible – to decide whether a root originally had a voiced spirant or a voiced stop, as both Qenya and Goldogrin turn them into spirants. For example, both *gweđ- and *gwed- would yield gwedh- in Goldogrin and ’wer- in Qenya (perhaps d > dh is a Common Eldarin change). Qenya re-stops the spirants after sonorants: nđ, > nd, and mƀ > mb (QL:24) (beside rđ, lđ, zđ > rd, ld, rd and rƀ, lƀ, zƀ > rb, lb, rb) and Goldogrin probably does the same – it can certainly be analyzed this way. But it might also be that the original stop reappears.

§ Both mb and nd often become followed by l, r, w. The observed changes are mbl, mbr, mbw > ml, br/mbr, bw and ndl, ndr, ndw > ngl, ndr/nr, ndw:

§ Both medially and finally ld > ll > l usually with lengthening of the previous vowel. Since the lengthened vowel does not suffer any further change (so that â remains â etc.), it must be a development at least contemporary with the vowel shifts:

§ There is no vowel lengthening in:

§ Where -ld- occurs, it is from -lđ- (or survives in the combination -ldr-):

§ -rđ- > -rth- once:

But otherwise -rdh- is retained in gardhin ’pl. places’ (37), curdhu ’sin, wickedness, evil’ (28) and so on.

2.4  Miscellaneous changes

§ Internal lenition is the most basic development, unvoiced stops become voiced after a vowel p > b, t > d, k > g:

§ dl, đl > gl:

§ tl- > cl- initially, but -tl- > -gl- medially:

§ Primitive Eldarin had t͡y (probably a palatal stop [c]) and c (probably a palatalized k [kʲ]) as a modification of k (QL:15), this leads to a variation t-c ([t]-[k]) in Goldogrin. There is another a [t]-[k] variation independent of this (called euphonic dissimilative interchange by Tolkien (QL:22)), e.g. Q. TELPE/TELEPE, G. celeb (QL:91) or √kail(i)k beside tail(i)k (25).

§ þ, đ > s before t, c, b, p:

§ geminated tt, kk, pp become single spirants th (þ), ch (χ), f:

§ voiced stop + h leads to voiceless spirants d-h > th, b-h > f, probably also *g-h > *ch:

§ hl > thl, probably also *hr > *thr:

2.5  Combinations with s

§ sr, sl, sw > thr, thl, thw medially; but initially thw- > fw- and sometimes thl- > fl-; but thr- > fr- is said to be dialectal:

However, sw > f in:

The combination -str- is retained:

The variation in hestril, hethril ’sister’, hestron, hethron ’brother’, hestri, hethri ’consanguinity’ (48-49) is probably due to a derivation from heth(1) ’brother or sister ’ beside hest ’(1) but properly, consanguinity, the relationship of brother and sister, or brother and brother, etc. (2) a brother or sister’ (< *heth-t-).

§ sƀ > sf or thw; but sm, sf > f:

§ ts > th or ss, ks > ch; presumably also *ps > *f, but this does not seem to be attested:

§ sometimes: ss > th:


The last examples show the common noun ending -os which seems to correspond to Q. -as(se). But there is also a very similar ending -oth. So maybe there is in fact a variation ss > -th or -s; comparable with ts > th/ss.

2.6  Combinations with χ, ʒ, vocalization

§ The velar spirants χ, ʒ form a diphthong with the preceding vowel if followed by the dentals s, t, þ, ð (with t > þ). If the preceding vowel is e, the formed diphthong is ai. This points towards a development eχC, eʒC > *ei > ai, but note that ei > ê/e is otherwise attested (1.2).

Combinations like kþ, gþ also change > χþ, ʒþ merging with original χ, ʒ. Since cth is a permitted (and in fact favoured) cluster, it remains medially in polysyllabic words. Some analogical levelling can be found here:

We may also assume that the change is likewise carried out medially when further consonants follow (*VkþC > ViþC), i.e. to prevent three-consonant clusters:

With the consonant r the development is χr > thr instead:

Also apparently *ʒl > chl:

Whenever χ, ʒ happen to stand in final position, the two possible results are the formation of a diphthong once again, or contraction with lengthening of the preceding vowel. Perhaps it depends on the affix or the vowel itself – the data set it too small to tell:

If χ, ʒ appear between two vowels the formation of a diphthong is also possible:

In other examples ʒ seems to be lost:

§ Medially sg > *sg/*sʒ > sg/si seems to occur, which is probably the same for consonants other than s:

§ however nχ > nth:

2.7  Combinations with w, vocalization of f, v

§ In medial and final position q changes > p and is then treated according to the environment (lenition > b or spirantization > f):

Because of unaccented *kwa > *kwo > *ko this change is apparently prevented, for instance in laigos ’verdure, greenness’ (Q. laiqasse) with lenition k > g instead; or in narcolas < *narkwa-lass. In final position only -kw > -c is occasionally found – see the variations usc, usg above.

Note that unlike later Noldorin and Sindarin, q > p is a medial and final development only, cw- and gw- being retained initially. There are also some examples with -q- > -gw-:

Perhaps these are formed with -weth rather than -eth (see Appendix A), so that q+w (labialized k [kʷ] + glide) > gw.

A development symmetrical to this is chw > f/fw, although the velar spirant may also become ı̯ according to 2.6:

§ final -nw > -m and -lw > -lb; -tw > -du

Presumably final -w always becomes -p with further change, unless it follows a vowel as in gwiw ’young’ (42), thus: -nw > *-np > *-nb > *-mb > m. Further evidence for this is for example the adjective fembrin from fem, restoring original mb (it was changed to fenwed).
Note also gaiw ’pregnant’ < gīwā̆, changed from gaib (37). According to this erw ’only (av.)’ (33), garw ’sown-field, aj. tilled’ (38), glarw ’bright’ (39) would be also pronounced *erb, *garb, *glarb.
Note moreover that -nw- is preserved in medial position, as for instance in tinwin ’a small star’ (70).
Comparing with gadu it seems that -w > -p did not appear after d, so that -w > -u is found instead. But compare deleted taru ’horned’, cogante of Q. tarwa (69). Also note that celu ’rill, stream, runlet’ pl. celwin (25) is rather < *kelū rather than *kelwa, since the Qenya cognate is also kelu ’stream’ (QL:46).

Finally it should be pointed out that there is already some variation n-m and q-p in the primitive language, as the Qenya roots GWINI/GWIMI or LIQI/LIPI show. In later development of all dialects q (1) and p (5) have been much interchanged (QL:15).

§ A common pattern is the vocalization apC, abC, aƀC, awC, afC > auC. So always in monosyllables; otherwise a preservation is possible when a permitted cluster is already formed, as -vr-, -fr-, -pth-. We may assume a spirantization abC, apC > aƀC, afC and a subsequent merging with original aƀC, afC. The close symmetry to the vocalization χ, ʒ > ı̯ described in 2.6 is apparent.

Note that the verb drautha- ’to weary, tire out’ is formed directly < drauth-, i.e. it is an analogical formation. The historical form would have been *draptha- with retained medial -pþ-.
It should also be pointed out that Tolkien gives primitive elements with lenition already carried out. It can be often seen by comparison, so sab- (67) must be from *sap- because of Q. SAPA ’dig, excavate’ (QL:82). However, both Qenya and Goldogrin spirantize b > v, so that a distinguishing is difficult in such a case.

Note that -wr- > -vr- in tavros and also in so u̯- > sovri ’cleansing’ (68), but -wr- > -br- is found in:

§ Between vowels otherwise -w- > -v-, -f when final:

An exception is:

§ wo > o:

2.8  Final devoicing, dissimilation

§ The sound dh is devoiced to th whenever it comes to stand finally:

Final devoicing is a common phenomenon in languages, but not always reflected in the transcription. Since it is reflected in Goldogrin, we can assume that other final voiced consonants, as -g or -b have to be pronounced as such.

A bit difficult is final -f. For, instance one finds duif (v-) ’stream’ (31), gwef ’louse’, pl. gwevin (45), gôf ’fruit, esp. of trees’, pl. gôvin (40) and so on. It coul be devoiced, but already at that time Tolkien experimented with Welsh orthography, for instance in the transcription Tynufiel = Tinúviel (LT2:41), where f represents voiced [v]. And in fact, we find both Ifon (18) and Ivon (52) as names corresponding to Q. Yavanna where the historical sound is ƀ (bilabial [v]). Also, [f] is once transcribed by the digraph ph, in calph ’a bucket’ (25).
So it seems likely that f actually represents [v], at least under certain circumstances.

§ A dissimilation th > s or th > f is sometimes found when an adjacent syllable contains th. In the attested examples the changed sound is initial or final:

3  Development of syllabic consonants

There is a multitude of monosyllabic bases with the structure C + syllabic ṛ/ḷ/ṇ + C. These ṛ/ḷ/ṇ become non-syllabic and may form a cluster with either of the surrounding consonants. The main observed results are CliC/CriC and CalC/CarC.
Where an initial cluster Cl-/Cr- is impossible, the pattern is always Calc/CarC or CilC (CirC is not attested). This is especially true for – no initial clusters of the shape Cn- exist in Goldogrin, so the only results are CinC, CanC.
The combination CwṛC may also result in CurC, with a rounded vowel under influence of w.

§ -li-, -ri-

§ -il-, -in-

§ -al-, -ar-, -an-

§ -ur-

Note that thlib- has the past tense thlaibi ’for salpi*’ (sic), thlid- has thlinti ’for salti*’, Briga has braigi ’for *barchi, drib- has draibi, flig- has flaigi or flinchi ’for *falci, clib- has claibi ’for talpi*’ and crib- has crimpi. These are all analogical developments. The past tense is often formed by lengthening of the root vowel, thus *sḹp-ı̯ē> salpi, *sḹt-ı̯ē > salti, *vṝk-ı̯ē > barchi, nṝq-ı̯ē > *narbi, *fḹk-ı̯ē > falci, *tḹp-ı̯ē > talpi, *kṝp-ı̯ē > *carpi would have been the historical forms implying that long syllabic consonants always result in alC, arC (and this is confirmed by other examples). The forms thaibi, braigi, draibi, flaigi, claibi are formed from long ī > ai as in the case of other verbs with the historical root vowel i. But (g)intha- and ilt- (among others) apparently show the historical past tense ganthi < *yṇ̄t-ı̯ē and galti < *yḹt-ı̯ē. [1]

There are some examples of words beginning in a syllabic consonant. The observed patterns are ḷC, ṛC, ṇC > ilC, riC, inC with a short sound (again no irC) and ṝC, ṇ̄C > arC, anC with a long sound. When a labial like m, b, w (or q) follows, the result is ulC, unC rather than ilC, inC. The nasal is assimilated to the following consonant:

Some words in Goldogrin are derived by adding syllabic -ḷ which then has several possibilities break: -ḷ > -il, -el, -li (rarely -ḷ seems to remain unchanged).
The ending -li < *-ḷ can be distinguished from the usual -li < *-lē or *-lī if there is a simultaneous root fortification (like lengthening or nasal infixion). As far as it can be determined, such a fortification is accompanied by vocalic suffixes only, and -ḷ is treated as one.

The etymological note associated with Danigwethl implies that -l in such a position is still syllabic. Probably the same is true for -r in words like gadr ’joint, link of chain’ (36) and for in tathn ’number’ (69).
It is also eye-catching that the adjectival ending -n often comes in pair with an equal ending -on (for instance barn, baron ’tilled, inhabited (21)). One may suspect that both are variations of a common ending -na becoming syllabic after the loss of the final vowel: -Cna > -Cṇ > -n, -on. Hence also:

Compare these developments with the Qenya changes in QL:10-11. They happened late in the Cor-Eldarin period, so that the syllabic consonants were broken in Qenya and Goldogrin independently of each other.
In Qenya the outcome depends on the articulation of the following consonant, so short gives ul before labials. In Goldogrin this seems to be the case for initial as well, but not for the CḷC structure. So tḷp- > clib- whereas Qenya has tulpu- (QL:93). There is no breaking into CriC, CliC etc. in Qenya, as it does not permit initial clusters.

Attempting to draw a tentative overview table similar to the one Tolkien does for Qenya in QL:10, it might look like this:

 initialinitial before labialsmedialfinal
short nasalsinumin-n (-ṇ?), -on
long nasalsan*aman? probably not occurring
ilulli, il, al-ḷ, -il, -el, -li
al -all-? probably not occurring
ri*urri, ar?
ar -all-? probably not occurring

Perhaps medial -al- actually derives from long – it matches with Q. -al- (which is according to QL:10 always < ) at almost all instances. Note also the complete absence of -ir- which in Qenya appears only before (a fronted k, i.e. [kʲ]).

Tolkien also explains that initial q, p, ty influenced the timbre of the following sonants, so that qntā́ > cunta ’full’, but qṇ́tā > qanta ’whole, all’ (with an instead of un). This might also apply to Goldogrin and explain why we find qṛđ > *qṛđū́ > curdhu ’sin, wickedness, evil’, but qṛđ > *qṛ́đa > cwarth ’evil, bad, wicked’ (in Qenya curdu and qarda).

4  Phonology in change – the Gnomish Lexicon Slips

Ten slips are tucked into the back of the Gnomish Lexicon, postdating it by at least a year. They contain a word list similar to the GL, beginning with the letter A and some additional notes. The fact that they indeed postdate it can be immediately seen from the content – a remarkable shift can be observed in the phonology underlying the entries, especially regarding the vowels which become much closer to Welsh.

The main differing points are:

§ A Welsh-style transcription is often employed:

§ I-affection is introduced, the vowel i causes preceding a, ū > e, ŷ, so that the new vowel y [y] appears. But it seems that only the immediately preceding syllable is affected:

§ The plural can now also be formed by i-affection, see especially GLS:116:

§ The diphthong ei now appears, particularly via *egþ- > *eʒþ > eith, as gweg, pl. gweith rather than earlier gwaith; also by i-infixion alyan > eilian ’branch’.

§ Now â > au (aw) and au > o:

§ A-affection may now cause preceding u > o:

§ A seemingly new past tense formation appears, with the ablaut a, u > ai, ŷ beside a > aw. Perhaps it could be explained by a lengthening of the vowel and its subsequent affection triggered by the following -i (compare bawr, pl. bair):

§ Initial unvoiced lh- appears in lham *’language’ (GLS:106).

§ The hesitation concerning the development of the combinations ’sonorant + voiceless stop’ (2.3) has perhaps been resolved in favour of spirantization, although there are too few examples to be sure:

§ The change a > o in the final syllable has perhaps been rejected:

Quite interestingly the sheets fill eight pages as presented in the issue of Parma Eldalamberon, three of which deal with the initial letter A; then it goes on faster finishing up with ŷ. Obviously Tolkien had the GL at hand and began to rewrite the entries with the new structure of the language in mind. The new phonological details have set the course and the fundament for the following Noldorin.

5  Comparison with later sources

Some final remarks and observations:

§ ī > ai, ū > au

The correspondence i-ai and a-au still remains in later Noldorin and Sindarin. For instance, N. taur ’vast, mighty, overwhelming, awful’ is derived from TUR- just like G. taura ’powerful’ from tur-. But it is not quite the same – while Goldogrin breaks long ī, ū into the diphthongs ai, au, the later mechanism is an infixion of a at an early period.
The difference can be seen by the correspondence with Q(u)enya – in the former case it retains ī, ū, in the latter ai, au are found in both languages.

§ sonorant + voiceless stop

As mentioned in 2.3, Tolkien hesitated about the development of stops after sonorants – they were either retained or spirantized. In fact, this matter was still not settled even many years later. In the Noldorin of The Etymologies ’nasal + voiceless stop’ yields a geminated nasal medially. Several late conceptions can be found in VT42:27 and PE17:131-134. A dialectal distinction is made there, the northern dialect of Sindarin retains the stops after nasals and l or goes only as far as spirantization, but otherwise the combinations ’l/nasal + voiceless spirant’ become ’long unvoiced l/nasal’, which are later voiced under certain conditions, according to one conception.

§ -nw > -m

This change can still be traced in Early Noldorin:

In later Sindarin it becomes dialectal:
In the Northern dialect, however, in final position only, C.E. tw > dw, dw > ðw, thw > þw, nw became b, v, f, m. (VT41:8)

But the Goldogrin personal pronoun im ’I’ from im len ’I have or am come’ (53) and the prefix ni· (PE13:97) are probably related just in the same way: *inwi > im. This pronoun also appears much later in The Lord of the Rings in the phrase im Narvi hain echant ’I, Narvi, made them’ before Tolkien became dissatisfied with it and reinterpreted im as a reflexive pronoun ’same one, self’ (VT47:38).

§ thr-, thl- | fr-, fl-

Interestingly, some words in Gothic show fl- > þl-, for instance þlauhs ’flight’ (Gothic Bible, Mark 13:18) while others do not, as flodus ’flood’ (Luke 6:49) – for a not very clear reason. A similar variation can be found in the Noldorin of The Etymologies, for example SLUS-, SRUS- > thloss, floss, thross ’a whisper or rustling sound’.
As it turns out, a dialectal variation fr-/thr- (see 2.5) occurs already in Goldogrin and may be a point where Tolkien was influenced by Gothic. In fact, he originally intended to use the notebook of the Qenya Lexicon for some work on Gothic (QL:x).

A  Goldogrin derivational endings

Some words in Goldogrin are derived by the means of n-infixion, probably with addition of a vowel which usually fell away:

Another rarer method is the doubling of the final consonant and addition of a vowel ending

And a third method is the lengthening of the root vowel, and probably also the addition of a vowel ending. While the final vowel usually fell away, the former lengthening can be seen by the change of the long vowel:

Beside that there is a huge number of derivational endings which are in the following listed alphabetically. Masculine and feminine endings are discussed separately in Appendix B for convenience.
A lot has been already discussed in [2].

§ -a < *-ā

This is a not very common noun ending; it seems only to be employed in monosyllabic roots with the root vowel a, and no root vowel lengthening occurs.

It is also employed to derive denominative verbs:

§ -c, -g < -kı̆, -kā̆

This is a noun and adjectival ending, especially favoured after diphthongs, and otherwise appended to roots ending in the dentals -r, -l, -n, -s, -þ (not dental stops). As an adjectival ending it often has the sense of a past participle.

§ -ca

This is a verbal suffix in:

§ -chi, chin(t)

This is mentioned as a diminutive suffix (25).

§ -eg

See -og, -eg, -ig

§ -el

See -ḷ (part 3).

§ -eth

See -os, -oth, -eth, -ith

§ -g

See -c, -g

§ -i < *-ī, (*-ē?)

This is both a rare noun and an adjectival ending:

§ -iel

This rare noun ending might have a kind of reverential or augmentative function. It is also a feminine suffix, see Appendix B.

§ -ig

See -og, -eg, -ig

§ -il, ril < *-(r)illĕ/*-(r)il

This is a rare noun ending obviously denoting an tool or instrument.

§ -in < *-inā̆

This is a frequent adjectival ending, see also the variant -rin. It is often favoured when the root vowel is i. Some examples are in fact past participles, usually with corresponding verbs. Substance or material is often denoted by this ending as well. Compare also -on, -ṇ

The ending also appears in the derivation of a few nouns:

§ -inc, -ing < *-inkĕ, *-ingĕ

This is a diminutive suffix:

§ -iol

See -ol, -rol, -iol, -riol

§ -ion

This is apparently a variant of -on, -n, also forming nouns and adjectives:

§ -ith

See -os, -oth, -eth, -ith

§ -ḷ (archaic)

The role of this noun suffix is probably very similar to -li, but it develops several possibilities and is not always distinguishable from -li. See part 3.

§ -la < *-lā, (*-lē?)

This is both a noun and an adjectival ending. As a noun ending it has perhaps the sense ’object’ or ’being’:

It is a verbal ending e.g. in:

§ -li, -l < *-lē̆/*-lī̆

This is a common abstract noun ending, similar to -ri, -wi, but often denotes concrete objects (more specifically instruments associated with an action) similar to -m. It seems never to be used after l, n.

It also seems to be a diminutive suffix (some forms could likewise contain -thli, compare below):

§ -m, -ma < *-mā, *-mē (or *-mǣ, see 1.3)

This is both a noun ending (usually after r, l) and rare adjectival ending. As a noun it often denotes concrete objects:

§ -n

See -on, -n

§ -na

This is a rare verbal ending:

§ -og, -eg, -ig < *-Vkā̆

This is a very common adjectival ending that often denotes the material of things or the mood of persons. The root vowels a, u, e, i would appear as o, o, e, i in the ultimate syllable. But looking at the examples it can be verified that the vowels in the final syllables do not always descend from root vowels:

§ -ol, -rol or -iol, -riol

These suffixes usually denote deverbal adjectives that are in fact present participles. But also without corresponding verbs they often describe a present condition. In other cases they are interchangeable with -rin, -in.

The form -r(i)ol is probably a variation after suitable consonants or between vowels similar to -in, -rin:

§ -on, -n < *-ṇ < *-nā̆, *-nĕ

This is a pair of often interchangeable adjectival endings and probably with a common origin (-n usually only after r). They are similar in role to -in, -rin and often form past participles.

But both also function as a noun suffix (-on is also a male suffix, see Appendix B):

§ -or

This is a noun ending, often describing localities:

It is also a male agentive suffix, see Appendix B.

§ -os, -oth, -eth, -ith < *-as, *-ass, *-assĕ/*-attĕ (or the same with vowels o, e, i)

These very common noun endings are often interchangeable and may show the variation -ss/-th (else -os > *-Vsse, -oth < *-Vtte). They seem to have a very large variety of meanings.

The endings -oth, -os are also used to derive deverbal nouns:

§ -oth

See -os, -oth, -eth, -ith

§ -ra, -r < *-rā, *-rā̆

This is a general adjectival ending. Similarly to adj. -a it is often (but not exclusively) employed for monosyllabic stems with the root vowel a.

It is a verbal suffix e.g. in:

§ -ri < *-rē, *-rī?

This is a common noun ending, similar to -li, -wi, but far more frequent. It often has an abstract meaning or denotes mass nouns like ’clothes’ or ’food’ – in this sense it mighty be translated as ’collection of things’. It is also used to derive deadjectival nouns describing the corresponding state. In other examples it seems to be deverbal. It seems never to be used after n, l where -thi, -wi, -weth are employed instead.

§ -rin

This is apparently a variant of -in after th, d, v, g, s and diphthongs; as well often functioning as a past participle. A difference to -in is that when applied to substances, -rin means ’like the substance’ and -in ’consisting out of the substance’

§ -riol

See -ol, -rol, -iol, -riol

§ -rol

See -ol, -rol, -iol, -riol

§ -s < *-sā̆, *-sĕ or *-ssĕ

This is both a noun and an adjectival ending, possibly favoured after s and original χ.

§ -st < *-stā

This is a noun ending, usually formed from roots ending in a vowel:

It also refers to habitation, but in this case is probably appended -ost ’enclosure, yard – town’:

§ -t

This is a noun ending for roots ending in s, r, đ, þ. Sometimes it denotes a single, finished action.

This is also an old dual suffix:

§ -ta

This is a common verbal ending:

§ -th < *-þV (perhaps also < *-ttĕ)

This is a noun ending, used with roots ending in -k, -g, -χ or a vowel (see 2.6 and compare -os, -oth, -eth, -ith; note that *-ktV, *-gtV, *-χtV would also yield -ith).

§ -tha

This is clearly an adjectival ending in sitha(2) ’this’ < si(n)-, but the other glosses are here tentatively at place, since we might just as well be dealing with the ending -ta (see 2.3). As a noun ending it denotes a ’thing with this property’ – paltha for instance may refer to three different flat objects. As an adjectival ending applied to colour stems it is used to derive similar shades, like ’pink’ from ’red’.

It is also a very common verb suffix, for instance:

§ -thi

This is a rare noun ending, which might as well be *-ti (see 2.3):

It is an adverbial suffix in:

§ -thli < -þlī

This is a diminutive/affectionate ending, but not always distinguishable from -li (cf. above).

§ -wa < -wā; primitive -wa, *-wĕ/*-wı̆

While -wā still appears as -wa; the adjectival ending -wā̆ and the noun ending *-wĕ/*-wı̆ can now only be seen by the sound shifts it created:

§ -thod

This is a very rare noun ending, possibly an extension of -tha.

§ -wed

This is a common adjectival ending, it seems that A-wed could be translated as ’having A’ or perhaps ’A-ful’.

§ -wen

This ending forms nouns, perhaps with a sense ’the act of’, ’the place for’ or ’the condition of’ (cf. -weth for the last):

Perhaps due to a coincidence with gwennin ’girl’, suffixed *-(g)wen(n) some of such nouns are personified:

§ -weth

In most cases this seems to be an abstract ending similar to -wen usually denoting ’the state of’ (English -ness, -hood, -ship). Compare -ri.

§ -wi < *-wē/*-wī?

This is a noun ending, similar to -li, -ri, but much rarer (probably preferred to them after l, r, n):

It has merged with the original -wi, a dual suffix (cf. initial gwi- ’twi-, bi, di-, etc.’):

§ -win

This is a very rare noun ending:

B  Masculine and feminine suffixes

Since Goldogrin has a huge variety of male and female endings, it is more convenient to list them separately. They often come in pairs:

§ m. -weg, f. -win

§ m. -os, f. -(th)ir

§ m. -n, -s, f. -il, -(n)ir

§ m. -g, f. -s

§ -*se, *-no (?)

§ m. -on, f. -el

§ m. -rog, -ron, f. -ril, -ros

§ m. -ren, f. -res – this is probably suffixed ren(d). ress for ’male/female relative’

§ m. -g, f. -lin

§ m. primitive -u, f. -is (?). -in

Note that the ending -os is masculine in hethos, but feminine in sacthos.
There is sometimes a separate ending for the female part only:

A number of suffixes appears without a pair:

The ending -ir seems to be genderless in:


Patrick Wynne. The Goldogrin Past Tense http://www.elvish.org/Tengwestie/articles/Wynne/goldpat.phtml
Thorsten Renk. The Goldogrin Grammar – An Introduction http://www.phy.duke.edu/~trenk/elvish/goldogrin/grammar.html

update: Jan 26th 2008 — contents and hyperlinks added


This document was translated from LATEX by HEVEA.