On comparative Proto-Mǐn *Dh- and putting conjectural morphology in its place

Recent conjectural morphological (‘word family’) approaches to early Chinese assign the aspirated causative verbs of the Mǐn group to Jerry Norman’s comparatively reconstructed Proto-Mǐn voiced aspirated *Dh-, proposing on this basis that *Dhreflects prefixation of Old Chinese provenance. In this article, I argue that comparative phonological work on Mǐn has never suggested *Dhfor these items. In this case as elsewhere, morphological models can be of use but require grounding in comparative results.


Introduction
comparative reconstructions of the Mǐn proto-language, so difficult to reconcile with the Chinese philological tradition, have long been left aside in studies of early Sinitic. Recent work, notably Baxter & Sagart (2014), moves to prioritize Norman's conclusions. However, a new tension slips in: comparative results are likewise not always a fit for the conjectural morphological (or 'word family') models which inform modern approaches to Old Chinese (OC). Below, I consider this tension by reference to Norman's (1973) Proto-Mǐn (PM) voiced aspirated obstruent onsets bʰ-/dʰ-/dzʰ-/gʰ-(below = *Dʰ-), a key oddity of Norman's PM from the standpoint of mainstream Chinese as well as a key jumping-off point for Baxter & Sagart's (2014) new ideas about the shape and structure of OC words.
In section 2, I present first the generally agreed set of items exhibiting the onset and tonal correspondences indexed by Norman's PM *Dʰ-(section 2.1), followed by a collection of previously recognized and newly identified cases of lower tonal register 'causativizing' aspiration in Mǐn (section 2.2). These two are seen to be disjoint sets, the latter having emerged only relatively recently in particular Mǐn daughter branches. In section 3, I touch on some state-of-the-field implications. Baxter & Sagart's (2014) conflation of these groups represents prioritization of a conjectural morphological or 'word family' model (on which see Baxter & Sagart 1997, Sagart 1999, section 1.1) -but this disposition escapes notice in Fellner & Hill's (2019) otherwise pointed critique of the word family as a theoretical construct, while new OC forms which run counter to comparative evidence bounce on through the literature. Actually, Sagart's program of the 1990's remains of much value and can be usefully applied to PM *Dʰ-, but requires that morphological hypotheses be both explicitly presented as such and everywhere subordinated to comparative conclusions.

Norman's comparative *Dʰ-
In the Mǐn languages, uniquely within Sinitic, tonal categories associated with the historical lower tonal register are cross-cut by two onset correspondence sets mostly involving voiceless unaspirated obstruents T-and voiceless aspirates Tʰ-respectively. The former set is larger and often regarded as typical of Mǐn; the latter comprises only some three dozen items but finds equally regular reflection across the group. In his early comparative work on Proto-Mǐn, Norman addressed this "most important defining feature [of] the Min group" (Norman 1982: 580) using voiced unaspirated obstruents *D-versus voiced aspirates *Dʰ-.
In part due to the peculiarity of such a contrast from the standpoint of mainstream Sinitic, many authors have felt that this Mǐn situation must be due not to a proto-language feature but to later dialect stratification of some kind (e.g., Lǐ & Dèng 2006). While this appears unlikely, my focus here is on the lexical incidence of the split. Table 1 presents Mǐn items consistently exhibiting historical lower tonal register aspiration, i.e., items implicated by Norman's PM *Dʰ-, as completely as possible by reference to work noted to follow. Columns contain forms for 37 etyma from four representative lects belonging to four Mǐn branches, from left to right: Northern Mǐn Díkǒu/DK 迪口 (Akitani 2008), Far Western Mǐn Shàowǔ/SW 邵武 (Norman 1982, Cheng 2001, Eastern Mǐn Hǔbèi/HB 虎浿 (Akitani 2018), and Southern Mǐn Xiàmén/XM 廈門 (Douglas 1873). 1 1 Mǐn is now often regarded as having undergone a primary split into Inland vs. Coastal branches, with both of these consisting in turn of two subbranches: Central vs. bifurcated Northern + Far Western (a.k.a. Shào-Jiāng 邵將) for the former, and Eastern vs. bifurcated Southern + Púxiān 莆仙 for the latter (see, e.g., Kwok 2018a).  Table 1 is meant simply to reflect and collate previous collections including Lǐ Rúlóng (1985, 139-140;40 items), also an attempt at an exhaustive list of words exhibiting historical lower register Tʰ-across Mǐn; Norman's various shorter * Dʰ-lists (e.g., 1982, 555-557;21 items); Akitani's lists of all cases in his Northern and Níngdé 寧德 (Eastern) Mǐn data where modern voiceless aspirated onsets correspond to Middle Chinese voiced D- (2008, 79, 125, 172-173; 2018, 38-39, 129-130, 221); and Norman's (1969) less extensive Northern Mǐn Jiànyáng 建陽 data. 2 Numerous details concerning Table 1 and related material are not immediately relevant to my discussion and are left to an Appendix. These include borderline cases, the problem of Mǐnspecific lexemes, and issues particular to one or another Mǐn branch or lect. For instance, Far Western Mǐn material is relatively uneven: in Table 1 Column 2 where indicated, I have supplemented available Shàowǔ data with forms from Norman (ms.) on Gāotáng/GT 高唐 as well as from Cheng (2001) on Hépíng/HP 和平 and Guāngzé/GZ 光澤. 3 Significantly, Norman's *Dʰ-sets involve intricate tonal correspondences, again considered more carefully in the Appendix. 4 These correspondences, first elucidated by Norman (1982), naturally complicate efforts to avoid reference to a proto-language. My concern in this section, however, is above all the fact of membership or not in the Table  1-type *Dʰ-sets. As far as I can tell, no past study of Mǐn has suggested that the items shown in Table 1, including marginal cases presented in the Appendix, overlap with the modern aspirate-onset words to be considered in section 2.2 below. 5 2 Some meso-level reconstructions have become available -Sūn (2016) on Proto-Northern Mǐn, Akitani (2018) on Proto-Níngdé, Kwok (2018a) on Proto-Southern Mǐn -but here I follow my sources above in citing modern forms. For Proto-Mǐn, very much a work in progress, see Norman (1981); changes affecting that author's later 'Common Mǐn' system are briefly considered to conclude section 2.2. 3 Also, '!' marks anomalous (= mainstream-Chinese-like) tones; see the Appendix. The two studies named were generously shared with me by Shěn Ruìqīng in personal communications of June 2020. From here on I use newly unambiguous 'Western Mǐn' for Norman's (1982) 'Far Western Mǐn' (in some studies 'Shào-Jiāng Mǐn).' Boxes in the table's final column indicate missing or etymologically problematic written forms; see the Appendix. 4 The *Dʰ-sets implicate certain historical upper register tonal categories in Western Mǐn and also, given historical Tone C, in some Eastern and Southern Mǐn varieties: see the Appendix and Table 1 SW as well as HB, where Tone C 'nose', 'seam', 'to poison', and 'tree' are involved. Also striking is that PM *D-≠ *Dʰ-is reflected only tonally in Western Mǐn, with modern voiceless aspirated onsets Tʰ-in both sets. This onset situation initially led Norman (1969, 1 Note 2) to regard Western Mǐn-type lects as 'Gàn-Hakka'; for the definitive demonstration of their Mǐn affiliation, see Norman (1982) and Shěn's (2018)

Mǐn's lower register pseudo-causative aspirates
At issue in this subsection are certain Mǐn historical lower tonal register plain vs. aspirated doublet pairs. Doubleting of this basic kind is common across the group and has various causes, interdialectal borrowing primary among them. However, some colloquial pairs involve distinct but systematically related meanings, with aspirated members seeming to be historical derivatives of plain counterparts via a "generally transitivizing or causative morphological mechanism" (Norman 1991, 340); the possible nature of such a mechanism has been explored more specifically by Kwok (2018b). Examples listed below are among those raised by Lǐ (1985, 142), from Southern Mǐn (SM) Quánzhōu 泉州 (where glosses are my translations), and by Norman (1991, 341), from Eastern Mǐn (EM) Fúzhōu 福州 (where glosses are Norman's). To these I have added a few pairs gleaned from Douglas's (1873) dictionary of colloquial Xiàmén (= Amoy), an SM variety closely related to Quánzhōu. 6 Again, it is a simple matter to consult past presentations of the Table 1-type lower register aspirates and confirm that the aspirateonset alternants addressed just below -here termed pseudocausatives -are nowhere to be found. Instead, these words belong always to Coastal Mǐn or particular of its subbranches, reconstructable neither to PM (whether as *Dʰ-or some other segment) nor to OC. I examine the data more closely for two reasons. First, while projecting the pseudo-causatives to PM based only on Coastal Mǐn attestation is a clear methodological misstep, these items do constitute a vexing open question in morphologically bereft Sinitic: after all, as Norman (1991, 341) states, "to claim that such […] pairs are due to dialect mixture is of course untenable." Second, Norman's (1991) general survey of the Mǐn group, while apparently not utilized by Baxter & Sagart (2014), does anticipate those authors in assigning the pseudo-causatives to a very early era, in particular to pre-PM. 7 Why Norman felt this to be necessary in 1991, despite what he would have known to be lack of Inland Mǐn support for such a status, is a field historical question with some important implications for the future of Mǐn studies. 6 Douglas's (1873) Amoy dictionary, which also includes forms from nearby locales including Quánzhōu, should be scoured systematically, as additional pairs lost in modern lexicons probably remain to be discovered. Below, I often paraphrase Douglas's detailed descriptions and also normalize his Romanization. 7 Norman (1991, 341) writes that "it seems probable that the aspirated/non-aspirated distinction of the voiced stops was once utilized as an important morphological device in the dialect that later became Proto-Mǐn." Baxter & Sagart's (2014) consideration of the problem appears to be independent of Lǐ (1985) and Norman (1991).
(16a) /tiã6/ 定 'still, quiet, steady (adj.)' (16b) /tʰiã6/ 'to take a small amount of food, medicine, etc., for strength or comfort [i.e., to still, to settle]' (Xiàmén; Douglas 1873, 494-495, 552) As for Inland Mǐn, Baxter & Sagart (2014, 126) here cite an NM Jiàn'ōu 建甌 /tʰiaŋ6/ 'fix (a date or time) in advance' from Lǐ & Pān's (1998, 192) Jiàn'ōu dictionary, but this is not to be compared with (16b), meaning we lack grounds for a PM *dʰ-onset 'to settle'. 16 This final pair points up difficulties faced by Baxter & Sagart (2014) in trying to fold together certain lower register plain ~ aspirated Coastal Mǐn pairs with the well-known voiceless ~ voiced (i.e., historical upper vs. lower tonal register) transitive vs. intransitive pairs of general Sinitic. Since the latter are accounted for by the authors via an intransitivizing Old Chinese prefix *N-(e.g., OC intransitive *N.t-> Middle Chinese d-and PM *d-), some closely parallel but contrasting source is needed for any associated Mǐn pseudo-causative. Thus Baxter & Sagart's (2014) OC *m-: given the pair just above, for instance, the idea is *N-tˀeŋ-s (> PM *d-) for the (16a) adjective vs. *m-tˀeŋ-s (> PM *dʰ-) for the (16b) causative. 17 The same applies to their treatment of, e.g., (10)/(11) 'straight' ~ 'make straight'. Aside from the larger methodological problem which has been my primary concern to this point, such a device is odd in requiring that some of the above pairs are ancient independent derivatives of a third root rather than immediate relatives. 18 Suffice to say that I think this unnecessarily complicates a problem which is on the evidence restricted to Coastal Mǐn.
While the matter of how best to explain the pseudo-causative aspirates deserves separate focused study, I nonetheless conclude this section with a brief sketch of a possible solution, followed by reflections on Norman's motivations in 1991.
First: it is surely significant that these Coastal Mǐn pairs appear exclusively in the historical lower tonal register, a fact which presents certain challenges for morphological accounts (with semantic range also a concern). I suspect that a satisfactory treatment will need to revisit Sagart's (1984) consideration of stress-conditioned differential devoicing. The historically anomalous (b) items seen above are, after all, verbal reapplications of historically regular (a) items (various parts of speech) in first position of particular syntactic frames: tonal category, and more directly its phonetic correlates including pitch, is the one feature we know to have conditioned just such a postdevoicing aspirated vs. unaspirated contrast in other Chinese varieties; and (c) the late operation of voicing neutralization, for which see directly to follow.
As for Norman, he preferred in later work to hew as closely as possible to modern reflexes in reconstructed forms, and within his largely unpublished Common Mǐn scheme adjusted his earlier PM voiced *D-/*Dʰ-to lower tonal register *T-/*Tʰ-after the modern situation. 20 This meant that the pairs considered above -which must relate in some way to pan-Sinitic historical voiced onsets -could only be assigned to a still earlier 'pre-PM' stage at which contrastive voicing 19 Sagart (1984) proposed rather the opposite -a historical rule prohibiting breathy phonation in final (i.e., stressed) position -in seeking to account for Mǐn lower register aspiration in general as opposed to the pseudo-causatives in particular, but his basic insight can be preserved. Also, devoicing to aspirates given frames (i) and (ii) may be SM-specific, whereas given (iii) and (iv) may be common to EM and SM. 20 This move gave Norman space for reconstruction of the voiced onsets of NM, a long intractable problem, to earlier plain voiced onsets (i.e., no notional 'softening'). was retained. 21 There are some problems here. First, loss of contrastive voicing in Mǐn cannot be particularly ancient. The pseudo-causatives themselves point to at least partially separate EM vs. SM neutralization processes, for instance, to say nothing of the NM situation. We also have the mid-19th century testimony of Douglas (1873, 572) regarding the SM dental affricates: "[i]n the lower series of tones, ts-sometimes changes to dz-, especially in [Tóng'ān 同安] and [Quánzhōu]." Finally, there are the WM voiceless aspirated reflexes of both PM *D-and *Dʰ-: here Norman's revised Common Mǐn lower register *T-and *Tʰ-, seen as already merged with upper register counterparts, appear to block the way to a comparatively principled account. 22 There are certainly aspects of Norman's earlier Proto-Mǐn framework which also prove problematic. Nonetheless, at present, I anticipate that his work of the 1970's will be the foundation for future progress, and that elements including *Dʰ-for the Table 1 correspondences will stand.

The role of conjectural morphology
At the level of the syllable, modern Sinitic lexicons are largely devoid of analogical relationships /formA/ : 'meaningA' :: /formB/ : 'meaningB' of the kind that constitute morphology, meaning that resemblances between members of an analyst's putatively related pairs are, from the standpoint of native intuition, purely fortuitous. At their core, conjectural morphological approaches to earlier stages of Chinese are classically historical linguistic efforts to recover past tune from this attested noise. We can address Fellner & Hill's (2019) objections to word family approaches in part via paraphrastic treatment of the offending Sino-Tibetanist terminology: allofam > 'candidate cognate'; word family > 'candidate root and derivatives,' etc. There is nothing methodologically untoward about these concepts.
But Fellner & Hill's (2019, 109-110) more important point, one I heartily endorse, concerns the danger of "accepting word families as given before turning to cross linguistic comparison". Given the peculiarity of the pairs in section 2.2 within Sinitic, Baxter & Sagart's (2014) attempted treatment in terms of a historical derivational process is understandable. However, this presumption winds up 21 W. South Coblin (personal communication, August 2020) has kindly consulted Norman's posthumous papers, including his Common Mǐn comparative tables, and not taken note therein of reference to what I have termed the pseudo-causatives. I am not sure if Norman considered these forms outside of his 1991 survey. 22 This problem, considered in some detail in Coblin (2018), may eventually have led Norman to regard Mǐn daughter branches as irreconcilable in some respects. I think Mǐn unity is established beyond doubt in Norman (1973Norman ( , 1982, not least by *Dʰ-. motivating the reconstruction of PM *Dʰ-and OC *m-D-/*m-T-in direct contravention of comparative indications. The ramifications here are worrisome: OC configurations of the same kind are extended to other PM *Dʰ-items, a proto-onset category which is further a key piece of the authors' reconfigured Proto-Mǐn; this Proto-Mǐn serves, in turn, as linchpin of the entire new Old Chinese enterprise. The onus also falls on readers to query these conjectural morphological premises. 23 We might object, for instance, to Sagart's cameo within Fellner & Hill (2019) only as comparativist foil to Matisoff -and some have taken up Baxter & Sagart's (2014) forms far less critically than does Hill (e.g., 2019). In my view, a factor here is the perception of vanguard vs. reactionary 'camps'. It is telling that Baxter & Sagart (2017, 260) regard the central theme of Schuessler's (2015) critical review to be "regretting that [the authors] do not retreat from reconstructing complex onsets". Really, the terms of the debate regarding OC are not that it 'had' vs. 'lacked' consonant clusters, uvular onsets, etc. (and note that the few practitioners who do remain constitutionally opposed to these elements per se stand all too ready to join such a category-erroneous battle.) Meaningful stances, rather, involve this or that view of the most descriptively economical and typologically plausible means of accounting for observed facts. If these means involve clusters and/or uvulars, etc., then so be it.
I am unsure to what extent the study of early Chinese phonology could be purely comparative, even if prospects on this front may be brighter than generally imagined. 24 Certainly, though, comparison must take priority. If we adhere to this principle in the case of PM *Dʰ-, the aspirates of section 2.2 are instantly set to one side. In so doing, the possibility of partially morphological origins for *Dʰ-, far from being lost, becomes much easier to scrutinize.
An example: among Lǐ's (1985) Quánzhōu doublets we find the pair /tui2/ 'to beat' vs. /tʰui2/ 'hammer (n.)' (cf. Xiàmén at Douglas 1873, 532, 569), semantically unlike the section 2.2 pairs and with aspirated member actually reconstructable to PM (see Table 1). On closer inspection, there seem to be a not insubstantial number of PM *Dʰnouns which, in parallel manner, are directly relatable to D-onset 23 On this point see, e.g., Schuessler (2017: 584), who notes that Baxter & Sagart (2014) "assume prefixes based only on morphological ideas" and depend more "on […] speculative etymology than phonology ", or Starostin (2015: 386), who remarks that certain of the authors' etymologies are "not to be trusted [as they are] based on no stricter methodological basis than an intuitive feel for 'word-family' connections." 24 As for true Proto-Chinese, the data is incomplete and the work is far from done: look no further than this paper for the state of PM. Coblin's recent projects (2019, etc.) concern Gàn/Hakka, where again to my mind there is "all to play for." noun/verb pairs of mainline Chinese (where noun and verb are homophones with the sometime exception of tone). Table 2 below gives a cursory account of this situation in terms of PM and Middle Chinese/MC onsets. Whereas the verbs on the right on balance have quite early origins in Sinitic and are in many cases not represented in PM, the seemingly younger nouns on the left have become part of latter-day basic vocabularies: note especially '(planted) tree', a late innovation but now 'tree (n.)' across the family. 25 Also of interest is that we find PM *dʰ : MC ḍ-, not d-, in these items.
Given these indications, we could perhaps entertain the idea that a group of OC *D-onset verbs had prefixed nominal derivatives which proceeded to mainstream Chinese D-but to PM *Dʰ-. Might OC *s-D-or the like be a possibility for this configuration given the sigmatic nominalization attested in a number of Tibeto-Burman languages (Jacques 2019(Jacques , 2020 as well as past proposals for a similar mechanism in early Chinese (Schuessler 2007, 54-55)? 26 This question and implications for 25 Early /ḍ-/ onset 'to hammer' finds support not from the philological tradition but from comparative material: Běijīng /tʃʰuɛiA2/ : Guǎngzhōu /tsʰəɥA2/ : Xiàmén /tuiA2/. 26 The earliest such proposal may be Pulleyblank's (1973) morphologically complex OC *ɦTʰ-to account for Norman's PM *Dʰ-, just the sort of word family thinking later to be revitalized by Sagart. An obvious question regarding an adjusted '*s-D-' for the Table 2 nouns would concern its upper tonal register counterparts. the rest of the genuine PM *Dʰ-words -apparently not in general morphological in origin but suddenly a smaller and decidedly more coherent collection -will merit careful consideration in future.

Comments invited
PiHPh relies on post-publication review of the papers that it publishes. If you have any comments on this piece, please add them to its comments site. You are encouraged to consult this site after reading the paper, as there may be comments from other readers there, and replies from the author. This paper's site is here: https://doi.org/10.2218/pihph.6.2021.5515 Appendix: Table 1 notes Column 1. Here Díkǒu 迪口 town (Jiàn'ōu) represents NM as supported by Akitani's (2008) Díkǒu, Shíbēi 石陂 town (Pǔchéng), and Zhènqián 鎮前 town (Zhènghé) as well as Norman's (1969) Hòushān 後山 (Jiànyáng). (Here and below, names of locales are followed in parentheses by county-level administrative divisions.) Compare also Sūn (2016), featuring a broader base of NM lects.
Díkǒu tonal reflexes associated with PM *Dʰ-given historical tones A B C D are 2 4 7 4 respectively. As is the case across NM, these reflect lower register jiǎ 甲-type tonal categories, where cognates of historical voiced-onset words have modern voiceless onsets, not lower register yǐ 乙-type tonal categories, where such words have modern voiced onsets in some varieties. Given historical Tone A, for instance, PM *Dʰ-is represented by modern Tʰ-in Díkǒu and Hòushān Tone 2 (not Tone 9), Shíbēi Tone 5 (not Tone 2), etc. On Norman's approach, PM *D-tonal results are seen as identical to those of *Dʰ-in NM, but I think that in Tone A, PM *D-> NM A2 yǐ 乙-type tones (e.g., Díkǒu 9 and not 2). This argument concerns the status of Norman's (1973) PM 'softened' onsets *-D-and is not essential here; see Smith (2021).
Some Inland Mǐn varieties have been affected by chain shifts /tʰ-/ > /h-/ (> /x-/); /tsʰ-/ > /tʰ-/. Shàowǔ A is largely unaffected, but Norman's (1982) rural 'Dialect B' and Cheng's (2001) Shàowǔ, as well as some NM varieties including Norman's (1969) Hòushān, show changes of this kind. Finally, both PM *D-and *Dʰ-are reflected as voiceless aspirates in modern WM, giving a resemblance to neighboring Hakka, in which modern lower register Tʰ-: MC D-. This contact situation may have led to a degree of unevenness in the tonal reflections of PM *Dʰ-in WM: see the general-Sinitic-type tones marked '!' in Column 2.