Introduction
Scholars of Indo-European historical linguistics have long found it convenient to refer to well known sound changes by the name of the researcher who first noticed the correspondences the sound change accounts for. Because of the proven utility of such named sound laws in Indo-European linguistics, the explicit listing and naming of sound laws in the Tibeto-Burman family could be expected to bring similar benefits.Footnote 1
Here I present those sound changes which are widely accepted to have occurred between the Tibeto-Burman Ursprache and Old Tibetan. I name each law after the first researcher known to me to have described it. The first four proposals, those of Shafer, von Koerber, Walleser, and de Jong, concentrate more on the interpretation of the Tibetan script than actual sound changes from pre-Tibetan to Tibetan, and thus are labelled ‘rules’ rather than ‘laws’. The sound laws have been ordered such that later laws can employ the results of earlier ones as evidence.
1 Shafer's rule, the sub-phonemic status of aspiration
The Tibetan script distinguishes the unaspirated consonant series k, c, t, p, ts from the aspirated consonant series kh, ch, th, ph, tsh. The distribution between voiceless aspirated and voiceless unaspirated stops in Written Tibetan is nearly complementary. Only as a simple Anlaut are the two series distinctive. Robert Shafer appears to be the first to have put forth the conjecture that aspiration in Tibetan was originally non-distinctive (1950/51: pp. 722–723). He did not however give an explanation of counter examples.
If one takes into account the phonological word and not just the syllable the distinctiveness of aspiration in Anlaut position can itself be considerably reduced. By far the majority of occurrences of unaspirated voiceless initials in Old Tibetan are word internal, either derivational suffixes or the second element of a compound. Aspiration occurs word initially, and can be seen as a non-phonemic super-segmental attribute of the phonological word (Hill Reference Hill2007). Even so, there are a small number of exceptions to these generalisations, in particular ci ‘what’ and kun ‘all’. Such cases require further study.
2 Walleser's rule, the difference between <gy> and <g.y>
The Old Tibetan orthographic distinction of <gy> and <g.y> represents the phonetic distinction of [gʲ] and [gj]. It is necessary to analyze the letter <y> into two phonemes: the glide /y/ when it is written as a miṅ-gźi
and a phonemic feature of palatalisation /
/ when written as a ya-btags
. This analysis was first proposed by Max Walleser (Reference Walleser1926: p. 9). Other strategies are available to distinguish <gy> from <g.y>, but in addition to the arguments in favour of Walleser's view presented in Hill (‘forthcoming’), this rule facilitates the elegant description of other sound changes, in particular Li's second Law.Footnote 2
3 von Koerber's rule, phonemic status of the palatals
Noticing that that the character <y> standing for the phonemic feature of palatalisation /
/ cannot follow a dental or a palatal *<dy>, *<cy>, but can follow velars and labials <ky>, <by>, the palatals may be analyzed as palatalized dentals, i.e. /t
/ for <c>, /n
/ for <ñ>, /d
/ for <j>, /s
/ for <ś>, and /z
/ for <ź>. The fact that Indic alphabets have a palatal series probably inclined the originators of the Tibetan script to choose to spell these sounds as single consonants rather than as <ty> and <dy> (von Koerber Reference von Koerber1935: pp. 121, §69).
Von Koerber describes the identification of the palatals as palatalised dentals as a synchronic fact (1935: pp. 120–121). He has been followed in this by numerous scholars (e.g. Miller Reference Miller1956: p. 348 note 2, Kjellin Reference Kjellin1975). Gong and Beyer however accept the palatals as phonemes but propose that they originate as a merger of the palatalised dentals and palatalised dental affricates (Gong [1977]2002: p. 388, Beyer Reference Beyer1992: pp. 81–84). Velar and labial stops can be followed by both an orthographic <r> and an orthographic <y>. Dentals are followed by <r> only. The palatals and affricates are followed by neither <r> or <y>. The inability of the dental affricates to directly precede <r> would appear to weaken the supposition that palatalised dental affricates are one origin of the palatals. Analyzing the palatals as palatalised dental stops restores the dental stops to the distribution of the labials and velars, but analyzing the palatals as palatalised dental affricates still leaves the dental affricates with a distribution deficient with respect to the other consonants. I therefore reject Gong and Beyer's speculation that one origin of the palatals is as palatalised dental affricates. As for Gong and Beyer's suggestion that the palatalised dentals are the origin of the palatals, there is no need to propose this as a historical change rather than a synchronic analysis.
4 de Jong's rule, spelling conventions before laterals and rhotics
Old Tibetan phonemically distinguishes both voiced and voiceless laterals and voiced and voiceless rhotics. The voiceless lateral /
/ as a simple initial is spelled as <lh>, the voiceless rhotic /
/ as <hr>. The prefix /g-/ is written as <k-> before voiceless laterals and rhotics. The prefix /s-/ may be written as <z> before a voiced lateral, as seen in the pairs lo ‘a report’ zlo ‘report, say’ and log ‘turn around’, zlog ‘turn something around’ (Hahn Reference Hahn and Helmut1999). In other cases no distinction in spelling is made between the voiced and voiceless lateral. The spelling <sl> can like <zl> represent /sl/ as in the pairs of verbs laṅ, laṅs ‘rise’ and sloṅ, bslaṅs, bslaṅ, sloṅs ‘raise’, ldob, lobs ‘learn’ and slob, bslabs, bslab, slobs ‘teach’ (Li Reference Li1933: pp. 139–140), but <sl> can also represent /s
/ as in the following denominative verb and noun pairs: slad, bslad ‘mix, adulterate’: lhad ‘an alloy’; slan ‘to mend’: lhan-pa ‘a patch’; sle ‘to braid’: lhe ‘a braid’ (Li Reference Li1933: pp.139–140). The spelling <bl> can thus represent /bl/ or /b
/. The clearest examples of <bl> reflecting /b
/ are the past stem of the verbal root √
ag ‘read’, spelled (present) klog, (past) blags, (future) klag, (imperative) lhogs with the phonemic interpretations /g
og/, /b
ags/, /g
ag/, and /
ogs/ (de Jong Reference de Jong1973), and the past stem of the verbal root √
ub ‘bedeck’ spelled klub, blubs, -, - with the phonemic interpretations /g
ub/, /b
ubs/, -, - (Eimer Reference Eimer1987). In some Dunhuang manuscripts the past stem of √
ag ‘read’ is even spelled <plags>, <phlags>, and <plhags> (de Jong Reference de Jong1973). Many examples could be pointed to where <bl> reflects /bl/; one such example is √laṅ ‘take’ with the stems len, blaṅs, blaṅ, loṅs, whose orthographic and phonemic interpretations coincide.
Although Li can be credited with the realisation that the spellings <l> and <lh> are a pair in the same way as <b> and <p>, but not in the way of <p> and <ph> (Li Reference Li1933: pp. 139–140), the extention of this realisation to clusters can be credited to Pulleyblank.
Pulleyblank has pointed out, in correspondence, that the orthographic distinction lh- versus l- might be used to account for kl- versus gl-, the former being interpreted as *glh-. The absence of an initial *pl- to match bl- (and of *rlh-, possibly, to match rl-), tells against this interpretation; (Sprigg Reference Sprigg1972: p. 552 note 10).
De Jong came to the same realisation independently of Pulleyblank, and demonstrated with philological evidence that there is an initial pl- to match bl-. The works of Eimer (Reference Eimer1987) and Hahn (Reference Hahn and Helmut1999) depend on de Jong and are unaware of the earlier contributions of Li and Pulleyblank. Although dubbing this rule ‘Li's rule’ or ‘Pulleyblank's rule’ is perhaps justified, I find it most appropriately named ‘de Jong's rule.’
5 Sa-skya Paṇḍita's Law, *g- > d- before graves, *d- > g- before acutes
As a synchronic fact d- and g- are in complementary distribution as the initial of a cluster with an obstruent in Old Tibetan. Before grave consonants (labials and velars) d- appears, and before acute consonants (dentals and palatals) g- appears. This synchronic fact is however not a rule for the phonemic analysis of Old Tibetan like the proceeding four, but has historical significance. Jacques (Reference Jacques2001) demonstrates with comparative evidence that originally separate *d- and *g- have fallen together. Before the consonant -r- they remain distinct even in Tibetan. Although Jacques’ 2001 presentation remains unpublished, he presents some of the relevant comparative data in Jacques (Reference Jacques2008: pp. 53–54). These data show that a morphological *g- prefix in animal names can be distinguished from a morphological prefix *d- in body parts.
A comparison with Rgyalroṅic cognates demonstrates that the animal prefix was originally velar (or even uvular) and not dental.

Since dr- is just as possible as gr- in Old Tibetan, the velar nature of this prefix is also confirmed by Tibetan grog-ma ‘ant’ (instead of *drog-ma).
In the case of the body part prefix d- Jacques does not present direct Tibetan cognates in the Rgyalroṅic languages, but the Japhug Rgyalroṅ words tɯ-mtshi ‘liver’ and tɯ-mke ‘neck’ demonstrate that in Tibetan words such as gdoṅ ‘face’, dbu ‘head’ and dpuṅ ‘back’ there is a morphological prefix, and it was originally a dental (Jacques Reference Jacques2008: 57).
The intellectual history of this law is difficult to trace. The realisation that the complementary distribution of d- and g- before obstruents implies a set of sound changes from pre-Tibetan to Tibetan appears to originate only with Jacques (Reference Jacques2001). Li acknowledges the fact that d- and g- are in complementary distribution, when he writes that their “notorious compensatory behaviour has made many people suspect them of a single origin” (1933: p. 136). Although a number of earlier scholars do correctly describe the distribution of d- and g- (Schmidt Reference Schmidt1839: p. 18; Schiefner Reference Schiefner1852: p. 328; Foucaux Reference Foucaux1858: pp. 106–107), I have been unable to find a scholar earlier than Li who makes explicit their complementary distribution. This may well be linked with the fact that the importance of complementary distribution as a linguistic idea only reached prominence around or after the time of Li's writing.
The correct description of d- and g- has its origins in the Tibetan grammatical tradition, and is discussed in the sūtras attributed to Thonmi Saṃbhoṭa. Dbus-po blo-gsal (thirteenth century) is the author of the earliest known commentary on these texts (cf. Mimaki Reference Mimaki1990, Reference Mimaki1992), which suggests that far from being eighth-century texts, they only became available in the twelfth or thirteenth century. Sa-skya Paṇḍita Kun-dgaḥ Rgyal-mtshan (1182–1251) appears to have written the earliest known statement of these distributions, in an account that acknowledges no awareness of Thonmi Saṁbhoṭa (Miller Reference Miller1993: p. 137). The most appropriate name for this law is then ‘Sa-skya Paṇḍita's Law.’
6 Houghton's Law, *ṅ
>ñ
In a number of examples a velar nasal (ṅ) in Burmese corresponds to a palatal nasal (ñ) in Tibetan. Combined with the observation that the velar nasal is never palatalised in Tibetan *<ṅy>/ṅ
/, these correspondences suggest a sound change pre-Tibetan *ṅ
> Old Tibetan ñ.

Houghton suggested the first two of these examples (Reference Hill1898: 52), and the third was added by Benedict (Reference Benedict1939: p. 228 note 26). Benedict was the first to make explicit that such a correspondences suggests a reconstruction */ṅ
/. However, since Houghton brought forth the comparisons that lead naturally to such a suggestion, and because I would prefer to reserve the moniker ‘Benedict's Law’ for the change l
>ź, I have chosen to dub this sound change in honour of Houghton.
If Chinese comparisons are added to those with Burmese the chart provided above can be augmented as follows.

7 Benedict's Law, *l
> ź
Benedict does not specifically discuss this sound change, but proposes it by way of comparing Burmese liy ‘four’ with Tibetan bźi ‘four’, which he reconstructs *bli (1939: p. 215).

The two examples ‘four’ and ‘field’ appear to be the only well agreed on examples. Schuessler suggests the comparison ‘ground’ (2007: p. 299). Jacques draws attention to Japhug Rgyalroṅ tɤ-lu ‘milk’ which suggests Tibetan źo <*l
o ‘yoghurt’ (2008: p.128). One might also consider comparisons among Tibetan gźu ‘bow’, Burmese liy ‘bow’ and Old Chinese 矢 syijX<*
ijʔ (0560a) ‘arrow’ but the Tibetan vowel is wrong. There are also grounds internal to Tibetan for such a reconstruction (Gong [Reference Gong1977] 2002: pp. 391–392).

8 Chang's Law, assimilation of b- before nasals
Betty Chang (Reference Chang1971: p. 738) discovered that cluster initial b- assimilates to the labial nasal m before nasals.
*bn > mn, e.g. √nan ‘suppress’, past *bnans > mnand
*bñ > mñ, e.g. √ñan ‘listen’, past *bñans > mñand
The seemingly anomalous m- in the past stem becomes thereby a subcase of the nearly ubiquitous b- prefix of the past stem.
9 Coblin's Law, loss of prefixes
Prefixes are lost when the resulting cluster is not phonotactically possible (Coblin Reference Coblin1976). This law greatly facilitates the internal reconstruction of the Tibetan verbal system.
*ḥrk > rk, e.g. √rkam ‘long for’, present *ḥrkam > rkam
*ḥrṅa > rṅa, e.g. √rṅa ‘mow’, present *ḥrṅa > rṅa
*gzl > zl, e.g. √zla ‘say, speak’, present *gzlo > zlo
*gsk > sk, e.g. √skaṅ ‘fulfill’, present *gskoṅ > skoṅ
*bb > b, e.g. √bya ‘do’, past *bbyas > byas
*bp > p, e.g. √pyag ‘bow’, past *bpyags, > phyags
In each such case the positing of a lost prefix resolves some anomaly in a verb's paradigm, and renders the verb in question an example of a paradigm type which is otherwise well attested. Here is not the place to discuss these proposals in detail.Footnote 3 One example may however prove illustrative. Some verbs have an ‘o’ vowel in their present stem, but not in the past or future; one such verb is skoṅ, bskaṅs, bskaṅ, skoṅs ‘fulfill’. Of those verbs which have such an ‘o’ in the present, those where a g- prefix in the present stem is phonotactically possible have such a prefix, e.g. gsod, bsad, gsad, sod ‘kill’. Reconstructing the present stem of skoṅ to *gskoṅ < *gskaṅ not only accounts for the presence of the ‘o’ vowel in both skoṅ and gsod, but also for the lack of a g- in the former and its presence in the latter.
10 Li Fang-Kuei's first Law, epenthesis after ḥ
When an ḥ precedes a fricative, lateral, or r, a dental stop is inserted between ḥ and the following consonant (Li Reference Li1933: p. 149).
*ḥs > ḥts, e.g. √so ‘nourish’, present *ḥso > ḥtsho
*ḥś > ḥc ( = ḥtś), e.g. √śad ‘explain’, present *ḥśad >ḥchad
*ḥz > ḥdz, e.g. √zug ‘plant’, present *ḥzugd > ḥdzugs
*ḥź > ḥj ( = ḥdź), e.g. źo ‘milk’, present *ḥźo > ḥjo
*ḥr > ḥdr, e.g. √ri ‘write’, present *ḥri > ḥdri
The effect of this sound change is more complicated before laterals. Simon proposes the change *dl- > ld based on groups of related words such as ldum-po and zlum-po ‘round’, ldog-pa, log-pa ‘reverse (intrans.)’ and zlog-pa ‘reverse (trans.), ldoṅ-pa and loṅ-ba ‘be blind’, ldaṅ-ba, laṅs ‘rise’ and slaṅ-ba ‘raise’ (Simon Reference Simon1929: p. 187).Footnote 4 The results of Li's first Law is followed by this metathesis proposed by Simon, and then the cluster is simplified following Coblin's Law.
*ḥl > *ḥdl > *ḥld >ld, e.g √lad ‘chew’, present *ḥlad > *ḥdlad > ldad
*ḥ
> *ḥt
> *ḥ
t >lt, e.g. √
uṅ ‘fall’, present *ḥ
uṅ > *ḥt
uṅ > ltuṅ
11 Li Fang-Kuei's second Law: *ry- > rgy-
Li Fang-Kuei proposes the change *ry > rg
in order to make Tibetan <brgyaḥ> /brg
aḥ/ ‘hundred’ and <brgyad> /brg
ad/ ‘eight’ more closely parallel Old Chinese 百 paek <*pˤrak (0781a) ‘hundred’ and 八 peat <*pˤret (0281a) ‘eight’ (Li Reference Li, Egerod and Glahn1959: p. 59).Footnote 5 Three further examples of this correspondence are available. With no mention of an earlier authority Schuessler proposes the comparison of Tibetan rgyu <*ryu ‘flow’ with Old Chinese 流 ljuw < *[r]u (1104a) (2003: p. 238, 2007: p. 362) and Tibetan rgyud<*ryud ‘continuum’ with 繂 率 lwit < *[r]ut (498a) ‘rope’ (2003: p. 238).Footnote 6 Pulleyblank in a different context proposes a second relevant pair of words (1962: p. 215): Old Tibetan brgyal < *bryal ‘sink down, faint’ and ḥo-brgyal ‘fatigue, weariness’ and Old Chinese 罷 疲 bje < *[b]raj ‘fatigue’. Schuessler (Reference Schuessler2007: p. 512) however doubts the validity of this comparison.Footnote 7

Since Li did not recognise the difference between /y/ and /
/ his reconstruction can be understood as either */ry/ or */r
/ in Pre-Tibetan. Old Burmese ry- corresponds to both Old Tibetan rgy- (Old Burmese ryā ‘hundred’ and Old Tibetan brgya ‘hundred’) and Old Tibetan ź- (Old Burmese ryak ‘day’, Old Tibetan źag ‘day’). It is probably judicious to reconstruct Old Tibetan ź < pre-Tibetan *r
and Old Tibetan rgy< pre-Tibetan *ry, because a change *r
> ź is parallel to Benedict's Law *l
> ź.

Three instances of the change *ry>rgy require no comparison to other Tibeto-Burman languages. The place name Uḍḍiyana in Tibetan becomes U-rgyan or O-rgyan. Middle Indic languages regularly loose a final -a; this yields *Uḍḍiyan. If ḍ were pronounced as a rhotic, or heard as one by Tibetan ears this gives *Uriyan, and such a pronunciation was nativised following this sound change to U-rgyan (cf. Jacques and Chen Reference Jacques and Chen2010: p. 71 note 7). Also, in the Dunhuang document PT 1047 there is vacillation between ryags and rgyags as the spelling of a word for divination board. This vacillation probably indicates a process of nativising a non-Tibetan word ryags. Similarly, the name of the Źaṅźuṅ emperor found in the Dunhuang documents PT 1047 and PT 1287 as Lig myi rya or Lig myi rhya in the later Rgyal rabs bon kyi ḥbyuṅ gnas appears as Lig mi rgya (Uray Reference Uray1968: pp. 293–294). Finally, Tauscher mentions that in the Gondhla Proto-Kanjur there is regular confusion between the spellings rya and rgya (2008: p. xxxvi).
12 Simon's Law, *mr > ḥbr
Simon proposed the sound change pre-Tibetan *mr > Tibetan ḥbr (1929: pp. 187, 197 §86). Five of the examples Simon presented can no longer be accepted:
1) Simon compares bro ‘dance’ with 舞 mjuX < *m(r)aʔ (0103g) ‘dance’. The absence of an initial ḥ in the Tibetan and lack of a vowel correspondence suggests that this comparison is false.
2) Simon compares Tibetan brod ‘taste’ with Chinese 味 mjɨjH < *m[ə][t]-s (0531g) ‘taste’. The lack of an initial ḥ in the Tibetan and a medial -r- in the Chinese invalidates this comparison.
3) Simon compares Tibetan sbrul ‘snake’ with 閩 mǐn ‘an ethnonym’ on the mistaken belief that the later means ‘snake’ (Schuessler Reference Schuessler2007: p. 386).
4) Simon compares Tibetan ḥbum ‘100,000’ with Old Chinese 萬 *mjonH < *mans (0267a) ‘10,000’; both Tibetan and Chinese lack medial -r-.
5) Simon compared Tibetan ḥbras with Old Chinese 米 mejX < *mˤijʔ (0598a) ‘rice’ rather than 糲 ljejH < *mə-rˤat (0340g).
Gong (Reference Gong1995, #368) compares 妊 姙 nyimH <*n[ə]m-s (0667i,k) ‘pregnant’ with sbrum ‘pregnant’ which he reconstructs *smrum. No 諧 聲 xiesheng contacts suggest an m- in the series GSR 667. Gong appears to be following the suggestion of Pulleyblank (Reference Pulleyblank1979: p. 36) that based on the transcription 任 那 for Mimana (a fifth century polity, which was a member of the Kaya 加 耶 federation on the Korean peninsula) that this 諧 聲 xiesheng series once had initial *m-. The evidence for reading 任 那 as Mimana comes from the 日 本 書 紀 Nihonshoki, where in the record of 垂 仁 Suinin it is also spelled 彌 摩 那 (Kojima et al. Reference Kojima1994: p. 295). Sagart argues that 妊 姙 nyimH <*n[ə]m-s ‘pregnant’ (0667i,k) is etymologically derived from 任 nyim < *n[ə]m (0667f) ‘to carry’. The semantics are thus not favourable to Gong's suggestion. Sagart also proposes an etymological connection with 南 nom < *nˤ[ə]m (0650a) ‘south’, which argues against the m- initial proposed by Pulleyblank (Sagart Reference Sagart1988). Jacques (Reference Jacques2003: p. 124) citing Pan (Reference Pan2000: pp. 240–241) instead compares Tibetan sbrum ‘pregnant’ with 孕 yingH < *l[i]ŋ-s (*m.rəm-s is also a possible reconstruction).
The following table presents the evidence in favour of Simon's proposal known to me.

I find the evidence of this sound change compelling but not yet convincing. Before such a correspondence can be wholeheartedly embraced the Tibetan words containing the cluster smr-, such as smra ‘say’, ṅur-smrig ‘saffron’, smre ‘suffering’, smreg ‘root, remainder’, smraṅ, smreṅ ‘(ritually) say’ must be explained.
I have previously argued that ḥ- represented a voiced velar fricative in Old Tibetan and not a nasal (Hill Reference Hill2005: pp. 126–127, Reference Hill2009a: pp. 127–131). If the sound change *mr-> ḥbr- were valid this would suggest that ḥbr- had the pronunciation [mbr], known from Common Tibetan, already in the earliest Old Tibetan. The arguments presented in Hill (Reference Hill2005: pp. 126–127) against interpreting a pre-consonantal ḥ as a homorganic nasal are therefore also arguments against the sound change *mr-> ḥbr-.
Coblin on the basis of the comparison of Old Tibetan rmaṅ ‘horse’ and Old Burmese mraṅ ‘horse’ contrastingly suggests a change *mr>rm (1974). Because the Ursprache of these languages almost certainly pre-dates the domestication of the horse, I believe that this correspondence is likely characteristic of Wanderwörter.
13 Sun's Law, fortition of laterals
Sun (Reference Sun1993: p. 334 note 201) appears to be the first scholar to explicitly propose the sound change *ml- > md.Footnote 12 He connects this change to Li's first Law before laterals *ḥl>ld without elaboration. The following two tables present the relevant data known to me.
Internal Tibetan Evidence

Gong instead proposes that the explanation for the -d- versus -ź- in these last two examples is a palatalisation used morphologically to form honorifics (1977[2002]: p. 390). His proposal would yield the reconstructions bźugs ‘stay’<*bd
ugs and bźed ‘desire’<*bd
ed.Footnote 14 One may suppose that Gong would reject the comparison of bźugs ‘stay’ with lugs ‘way’.
Comparative Evidence

A parallel change *m
> mth could be suggested. The only possible instance known to me is mthil <*m
il ‘bottom ‘floor’ which Gong (Reference Gong1980, #79; Reference Gong1995, #169) compares to Old Burmese mliy ‘earth, ground’. However, Schuessler instead compares Burmese mliy with Tibetan gźi ‘ground’ (2007: p. 299). Since the change *l
>ź is well established and the semantics are more similar, this comparison is superior.
The fact that there is evidence for an original l- both in the cluster md- and in the cluster ḥd indicates that the single change *ml>md, is not explanatorily sufficient. However, one cannot propose the sound change *ḥl>ḥd because *ḥl is already used as the input of Li's first Law *ḥl>ld. How to reconstruct cases of ḥd- where there is evidence for an original l- is a problem requiring further attention.
14 Laufer's Law, *ʷa, *ʷə > o
Laufer noted that where Burmese has -wa Tibetan often has -o (1898/1899: part III, p. 224; 1976: p. 120). The first two examples are Laufer's, the next four from Sagart (Reference Sagart2006: p. 211), and the last from Benedict (Reference Benedict1972: p. 106).

Noting that all Old Tibetan words which have the diphthong -wa- are open syllables (grwa ‘corner,’ bcwa ‘ten,’ phywa ‘a class of gods,’ rtswa ‘grass,’ ḥwa ‘fox,’ źwa ‘hat,’ rwa ‘horn’) and the pair ḥo-dom ‘fox tail pendant’ and ḥwa ‘fox’, I previously suggested that Laufer's Law did not apply to open syllables (Hill Reference Hill2006: pp. 88–90). It is clear from the above examples that some Tibetan words with open syllables are valid examples of this correspondence and my explanation must be caste aside. Guillaume Jacques suggests that instances of -wa in Tibetan open syllables should be reconstructed as *uba in pre-Tibetan (2009). I accept this explanation.
The previous research treating this correspondence (Laufer Reference Laufer1898/1899, Gong 2002[Reference Gong1980], Matisoff Reference Matisoff2003, Hill Reference Hill2006, Jacques Reference Jacques2009) appears unaware that Inlaut Written Burmese -wa- originates from -o- in early Old Burmese (cf. Ba Shin Reference Ba Shin1962: pp. 27–28 and pp. 38–39, Maung Wun Reference Maung1975: p. 89, Nishida Reference Nishida1972: pp. 258, Dempsey Reference Dempsey2001: pp. 222–225). Because the correspondence of Written Burmese Inalut -wa- (Old Burmese -o-) with Tibetan -o- is a retention in Tibetan, the rubric ‘Laufer's Law’ should not be applied to such instances and instead be reserved for those cases where Burmese Anlaut wa- corresponds to Old Tibetan -o-. Laurent Sagart points out that in such cases the Tibetan cognate begins with g- (2006: p. 211).Footnote 17

Gong further notes that Tibetan -o- often corresponds to Old Chinese lavio-velars (and presumably labio-uvulars if he recognised them). He reconstructs *ʷa and *ʷə as sources of Tibetan o (2002[1980]: p. 24).

Because Burmese wa ‘tuber’ (compared to Chinese 芋 *[ɢ]ʷ(r)as and Tibetan gro-ma) has Anlaut wa- and the Tibetan examples among Gong's comparisons all begin with velars or uvulars one can combine Sagart and Gong's observations. If ‘K’ is used to represent a velar or uvular, it becomes possible to more precisely state Laufer's Law as the merger of Tibeto-Burman *Kʷə and *Kʷa as Tibetan Ko.Footnote 19
15 Synchronic mysteries
The rules of Shafer, von Koerber, Walleser, and de Jong sharpen the phonemic interpretation of the Tibetan script. This improved analysis aids in the discovery and the elegant statement of diachronic sound changes. Consequently, in tandem with the restatement of the exceptions to the sound laws presented here, a consideration of remaining problems in the phonetic interpretation of Old Tibetan is in order. The clarification of these issues may in future engender the discovery of, or more elegant statement of other sound laws.
Two issues remain unresolved in the interpretation of the Tibetan script. Old Tibetan has two graphic forms of the vowel which is called gi-gu in Written Tibetan. One of these characters is the same as the Written Tibetan gi-gu
<i>. The other is the mirror image
<ĭ>, and has thus come to be called the gi-gu inversé. Whether this character represents a phonetic reality or not remains controversial. Laufer (Reference Laufer1914: p. 84) believes that these two graphs represent vowels which originally were phonemically contrasting in Old Tibetan, but which quickly began to collapse into one phoneme. Miller (Reference Miller1966, Reference Miller1993: pp. 156–172) argues that the two represent sub-phonemic allophones of /i/. Other authors (e.g. Ulving Reference Ulving1972 and Róna-Tas Reference Róna-Tas, Ihara and Yamaguchi1992: pp. 698–699) have regarded these two characters as meaningless graphic variants. This issue appears to have not received attention since 1993 and remains unresolved.
Old Tibetan sporadically but non-randomly has a -ḥ where this consonant is missing in later forms of Tibetan. Previous researchers have not attended to this question and taken for granted that syllables ending in -ḥ can be treated as open syllables. I see no reason to think a final -ḥ is any more meaningless than a final -b or -g (Hill Reference Hill2005: pp. 115–118), especially when it is kept in mind that those cases where -ḥ was not lost in Written Tibetan have regular reflexes distinct from open syllables in some Tibetan dialects (Hill Reference Hill2009a: pp. 129–131)Footnote 20 and that Chinese sometimes has a final -k corresponding to Tibetan final -ḥ.

The final consonant -ḥ is potentially of great consequence in Tibeto-Burman historical phonology, and deserves more attention than it has received.
16 Diachronic mysteries
Today's exceptions to sound laws are tomorrow's sound laws. Tibeto-Burman historical linguists following in the tradition of Benedict (1992) and Matisoff (Reference Matisoff2003) have been over eager to credit exceptions to ‘allofamic’ variation in the proto-language.Footnote 21 Although such proto-variation probably does exist as examples such as ‘have’ (<Indo-European *kap, cf. Latin capio) and ‘give’ (<Indo-European *gebh, cf. Latin habeo) demonstrate, being satisfied with proto-variation as an explanation of anomalies is to abandon potential progress in the understanding of historical phonology and morphology.
The most valuable contribution of a survey of Tibetan sound laws is to draw new focus on the exceptions to these sound laws. After having surveyed what is known so far about Old Tibetan historical phonology those areas in need of better study merit focus. Exceptions to the respective sound laws presented have been provided above, but it is convenient to assemble them together here. The exceptions to Simon's Law are smra ‘say’, ṅur-smrig ‘saffron’, smre ‘suffering’, smreg ‘root, remainder’, smraṅ, smreṅ ‘(ritually) say’. As exceptions to either Sun's Law or Li's first Law are the words ḥdug ‘stay’, ḥdod ‘desire’ and ḥdom-pa ‘fathom’ which have connections to words with lateral initials but cannot be reconstructed as Li's *ḥl or Sun's *ml.
