Malone 1971

Linguistic Society of America
Systematic Metathesis in Mandaic

Author(s): Joseph L. Malone
Source: Language, Vol. 47, No. 2 (Jun., 1971), pp. 394-415
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/412088
Accessed: 26-02-2016 12:47 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].
Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.
http://www.jstor.org
This content downloaded from 193.0.65.67 on Fri, 26 Feb 2016 12:47:53 UTC
All use subject to JSTOR Terms and Conditions
SYSTEMATIC METATHESIS IN MANDAIC
JOSEPH L. MALONE
Barnard Collegeand Columbia University
Classical Mandaic manifests two types of systematic metathesis, both of which
have interesting repercussions for Semitic and general linguistics. (1) t of the
reflexive prefix metathesizes with an immediately following sibilant; e.g.
*'itsamika > estdmex'he was reassured'. Though this change must be posited to
have occurred in Common NW Semitic, there is evidence that it RECURRED cen-
turies later in Mandaic. (2) h as third radical metathesizes with an immediately
preceding simple consonant, and then optionally undergoes total assimilation to
that consonant; e.g. *miShit> mehseO' me6?eO'I anointed'. But if these changes
are posited as having been somewhat more general, a natural explanation is
suggested for certain instances of vowel length, not only in Mandaic but perhaps
also in Akkadian. This nexus of developments is strikingly similar to a complex
of changes posited by Kiparsky 1967 for Greek.
1. Classical Mandaic manifests two types of systematic metathesis, both of
which have interesting repercussions for Semitic and general linguistics.' The
first type to be discussed (?2), involving the permutation of reflexive t with an
1 A shorter version of this paper was read to the 179th AOS meeting, New York, 25 March
1969.For comments on that presentation, or on a dittoed version of the same, I am grateful
to William Hallo, Moshe Held, and Erica Reiner. Among several abbreviations used
throughout, perhaps the following should be mentioned. LANGUAGES AND LANGUAGE
GROUPS: P = 'Proto-' (e.g. PNW = 'Proto-Northwestern'), Mod. = 'Modern', Cl. =
'Classical', W = 'West', E = 'East', Jud. = 'Judean', Gal. = 'Galilean', BA = 'Biblical
Aramaic', J/CP = 'Jewish/Christian Palestinian', EA = 'Egyptian Aramaic', BT =
'Babylonian-Talmudic'. Note also OT = 'Old Testament', trad. pron. = 'traditional
pronunciation'. BIBLIOGRAPHICAL REFERENCES: BLA = Bauer & Leander (Aramaic) 1962a;
BLH = Bauer & Leander (Hebrew) 1962b; CH = Chomsky & Halle 1968; CF = Cross &
Freedman 1952; DM = Drower & Macuch 1963; KNS = Kutscher, Naveh, & Shaked 1970;
MSUvS = Moscati, Spitaler, Ullendorff, & von Soden 1964. LINGUISTICSYMBOLOGY: C, V,
S, T, H = resp. 'consonant, vowel, sibilant, dental/alveolar stop, guttural glide' (9 h 9 h;
cf. fn. 2). Radical consonants are given by nv/C, e.g. 1A/? = 'first radical ?'. For roots, the
type form employed is V/pqd and derivatives, e.g. V/Sqd = 'any triradical root with 1st
radical sibilant' (?2.1). Linguistic forms are ordinarily cited in italics, but several other
devices are intermittently employed to focus attention on the level of abstraction: ( ) =
'graphological representation' (except explicit references to entries in DM, which are
boldfaced), notably used for transliterations from NW systems (e.g. Syriac, Hebrew), in
which case note that (b g d k p t) are ambiguous as representing b g d k p t or v zy x f 0
(i.e. the corresponding graphological symbols are pointed neither with dagesh nor raphe in
Heb., and neither with qus?aya nor rukkaxa in Syr.) The symbols (v y t x f 9) imply
explicit pointing with raphe/rukkCxa; (b g d k p t) imply explicit pointing with dagesh/
qub?aya.CP vowel pointing is carried over into the transliterations; see esp. fn. 42. Mand.
transliteration follows Macuch 1965, except that (e) replaces his (') (ayin), for reasons ex-
plained in Malone (1971, fn. 18). Brackets = 'broad phonetic representation' (but with a
special use in ?3, where they contrast with diagonals; see ?3.1). In a few cases, stress is
omitted where not germane to the discussion (e.g. fn. 16). Most aspects of phonological
transcription will be self-evident; but note that vowel length is not indicated for Hebrew,
being considered predictable in terms of stress and syllable structure. Note finally slight
differences in the transcription of post-Biblical Hebrew (e.g. fn. 10).
394
SYSTEMATIC METATHESIS IN MANDAIC 395
immediatelyfollowingsibilant,is ratherwidespreadin Semitic;but only Mandaic

presentsclearevidenceof this metathesishaving occurredTWICE,at an interval
of at least a millenium.The secondtype of metathesis(?3), involvingthe permu-
tation of a 3rd radicallaryngealglide (h, ?)2 with an immediatelyprecedingcon-
sonant, is unambiguouslyattested only in Mandaic;but the formulationoffered
here suggestssolutionsto seeminglyunrelateddiachronicproblems,and presents
striking parallelismwith a similardevelopmentalnexus in Greek.3
2. Examplesof the first type of metathesiswill be drawn where practicable
from the simple reflexiveconjugation(in generalAramaicstudies often labeled
hi6pa9el,ethpeel,and the like), and within that conjugationlargely from the
perfective paradigm (elsewhere called PERFECT, SUFFIXEDCONJUGATION, NOMINAL
[BLAI), but occasionally from the imperfective (IMPERFECT, PREFIXED CONJUGA-
TION, AORIST [BLA]) or participle. BLA (106-9) reconstructs,as the proto-
Aramaic ('Uraramaiiisch') forms for these respective paradigms, *?ftpaqada
(reshaped from **tdpaqada), *ydtpaqidu, and *mdtpaqidu. While the various
aspects of these reconstructions are individually understandable on the basis of
comparative evidence, the forms are likely to be heterogeneous as reconstructed
wholes; e.g. from the point of view of stem ablaut (pf. a-a, impf. and part. a-i),
comparison is invited to the Arab. 8th conjugation (îptaqada, yaptaqidu, mupta-
qidun), but the latter language shows infixation rather than prefixation of
reflexive t. In this paper, the forms posited for PAram. are pf. *h/?itpaqida,
impf. *yitpaqidu, part. *mitpaqidu. These reconstructions jointly imply that
Common Aramaic LACKED PREFIX AND STEM ABLAUT in the simple reflexive from
one paradigm (e.g. pf., impf., part.) to the other, and this holds water not only
for this conjugationbut more generally for all the derived conjugations (cf.,
e.g., BA intensive reflexive pf. hiOpaqqad,impf. yiGpaqqad;Jud. causative pf.
?apqed,impf. yapqed).
No attempt will be madein this paperto followthe developmentof the simple
reflexive (or any other conjugation)farther back than PAram., and even the
matter of the precise taxonomic placement of PAram. may be left in abeyance,
except to note the earliest Aramaic monuments (early 1st millenium B.C.) as
terminus post quem non.4 Since the metathesis seems to have been widespread in
NW Semitic, not merely in PAram. (?2.1), it is quite possible that this change
antedates the formation of Aramaic dialects per se. In that case the specifically
Aramaic stem shapes *h/?itpaqida, *yitpaqidu, *mitpaqidu may not have
developed until after the metathesis, and to that extent the examples of ?2.1
might be considered anachronistic. But detailed reconstruction of pre-Aram.
stems would take us far from our topic, the conclusions of which would hardly
be affected one way or the other.
2
I.e., glides in the sense of CH (302-3). This usage will be uniform throughout, likewise
covering the semivowels y, w and the pharyngeal spirants 9, h.
3A third type of metathesis, which might be systematic (i.e. regularly affecting entire
classes of linguistic items) in Mod. Mand., is briefly mentioned by Macuch (22).
4For general discussion of Aramaic and its relation to other NW languages, see Garbini
1957, 1960.
396 LANGUAGE, VOLUME 47, NUMBER 2 (1971)
2.1. Reflexive prefixal t- metathesizes with an immediately following sibilant,

and assimilates to the latter with respect to voice and pharyngealization (em-
phasis). This process may be posited as having affected those early NW dialects
which had the requisite structural conditions, viz. prefixed t- (PAram., pre-
Hebrew)6 rather than infixed -t- (widespread in most NW dialects).6
6 The only lVz verb attested with reflexive t- in Biblical Hebrew shows not the expected
metathesis tz > zd, but rather total regressive assimilation: *hitzakk > hizzakki 'cleanse
yourselves!' Is. 1:16 (Bergstriasser 1962:1.109; alternatively BLH 197 considers the de-
velopment to have been *hitzakkt > *hizdakk > hizzalkkt). Since hizzakki.i is the only
reflexive l/Vz verb in the entire OT, whereas post-Biblical Heb. shows expected zd (cf. Mod.
Heb. hizdaqq6n'to grow old'), the legitimacy of hizzakkiimight be questioned; in that case
tz > zz (or tz > zd > zz) would be a spurious change, and the metathesis could in all re-
spects be identified with the normal Aramaic development. But there is a way in which
hizzakkti may be accepted as genuine without positing Hebrew-Aramaic differences in the
metathesis, as long as the metathesis antedates the resolution in either dialect of the PSem.
slit fricatives *', *f, * ', *a. The PSem. root underlying hizzakkti is most likely *V/dky (see
the cognates listed in Brockelmann 1928, s.v. (dkiy)), though there seems also to have been
a formally and semantically related PSem. root *v/zky (cf. Arab. V/zky and the forms
cited by DM, s.v. ZKA). But then if the metathesis involved only GROOVED fricativesc,
i.e. the sibilants *s, *z ,*1, *, pre-Heb. *hitSakku (or perhaps *hittâkkayu, BLH 414)
would not meet the structural conditions for the change, but might on the other hand have
resolved the homorganic cluster t&by total regressive assimilation to &&,which later > zz
in accordance with the normal Heb. development of & > z. Total assimilatory resolution of
similar clusters is not uncommon in Semitic; cf. the progressive resolution zt > z5 in the
Arab. 8th conjugation, e.g. ?i&axxara 'to hoard'. This suggestion for Heb. hizzakki is
per se inconclusive, and it might be fruitful to examine the Hebrew development of the
remaining PSem. slit-fricative clusters, *t@,*t5, *tq. In a cursory look through the list of
reflexive 1VS verbs attested in the OT, I discovered NO clear reflexes of *t6 or *t2, and only
two plausible instances suggesting *t0 > ti > it (or perhaps > At > it): hiitobld 'to be
spoiled', if the root v1111< *VOll (cf. Arab. Oalla'to destroy'; but Arab. sulla 'to be con-
sumptive' makes *V/ill a viable alternative); and hiltammir 'to take heed', if cognate to
Syr. temra (see the suggested cognates in Brockelmann 1928, s.v.) On the other hand
there were several clear cases of PSem. *ts > Heb. it, *4 > at, as well of course as *ts > st
and *t > Et.These observations cannot be compelling in default of a thorough etymologi-
cal study, but the patterning uncovered does lend credence to the likelihood of PSem.
*t3 > tS > Heb. zz as a process independent of PSem. *tz > Heb. zd (fortuitously not
documented in the OT). The foregoing explanation might be taken with a grain of salt,
however, in view of BA hizzamminttin'ye agreed', ktib in Daniel 2:9 for qre hizdammintdn,
inasmuch as the former, if it be the genuine one of the two variants, could imply that zz
for expected zd may be common to Hebrew and Aramaic. But zd seems clearly to be the
favored Aramaic reflex elsewhere, including the closely related Egyptian Aramaic (cf.
(?zdhrw) 'take care!', Leander 1966:54), and hence the qre hizdamminmtnis likely to be the
better reading. On the other hand, zz for expected zd, and more generally SS for expected
ST, is sporadically attested throughout Aram. and Heb. texts: e.g. JP cognates to the BA
verb just discussed, l9mizdammdna'to join', but ?izz9mln 'he prepared himself' (Jastrow
1950, s.v. (zman) I), ?istaxdr ?issdxdr 'he was hemmed in' (ibid., s.v. (skar) I);
BT (?yzdryq) (?yzryq) 'it was sprinkled' (Margolis 1910:41);Mand. (tisthit) and (l(e)shit),
resp. masc. and fem. 'let it be spread!' (DM, s.v. SHT), and similar variations ibid., s.vv.
SHP, SKE, SMK, SBA II, ?PL; Heb. ?eltomdm'Iwas appalled' in Dan. 8:27, but tiisommi
'thou destroyest thyself' in Ec. 7:16. It is quite likely that different explanations may lie
behind many of the forms (e.g. Heb. tiisom*m is considered a scribal contamination by
Bergstrasser (2.99d), and at least some of the Aramaic cases may derive from a vanished
n-prefixal conjugation [cf. Heb. nif9al], e.g. *?inzamina > JP ?izzamin); but viewed com-
Examples: PAram. *?itsamira > *îstamira 'it was hidden', *it?abi9a >
*istabi9a 'he was cloyed', *?itGapika> *îGtapika'it was poured', *?itsaliba >
*istaliba 'he was hanged', *?itza9ipa > *?izda9ipa 'he was driven off'; pre-
Heb. reflexive intensive *hitsabbila > Heb. histabbel 'he dragged himself',
*hitsappika > histappdx 'it was poured', *hitgakkira > histakk6r 'he earned',
*hitsaddiqa > histadddq 'he justified himself'. Examples of forms from NW
languages with infixed -t- (MSUvS, 127): Ugaritic (yrth?) 'he washes himself',
Phoenician (tthtpk) 'she is being overthrown', Moabite (?lthm) 'I am fighting'.
Alternatively, the metathesis may date from PSem., in which case the 1V/S
verbs may have provided an analogical model for all other verbs in all dialects
except pre-Heb. and PAram.: schematically, early PSem. *tpqd and *tSqd, then
*tSqd > Stqd by metathesis, and finally in most dialects *tpqd > ptqd by anal-
ogy, resulting in the creation of infixal -t-. In either case it is quite likely that the
metathesis studied in this paper is but part of a wider nexus of very ancient
processes serving in effect to dispel any tS cluster, whether or not t = reflexive.7
It seems possible, in fact, to posit this purely phonological condition even if the
metathesis is hypothesized for the NW stage, as sketched above. This is so be-
cause seeming counter-instances all clearly or probably bespeak developments
later than the metathesis. This is clearly the case for, e.g., the Cl. Mand. variant
etsa # *esta 'nine', since the prothesis whereby etSd < ta?&(this latter form like-
wise evidenced in Cl. Mand.) certainly postdated the metathesis in question;
similarly Targumic variants such as Je0sdr # *?estdr 'sixteen' (syncopated
from earlier *?e69asdr) and borrowings like Syriac me6S&$ *meGta'primum
texturae filum' (a seemingly assimilated variant of me8S& < Greek mitos).8
Post-metathetic developments are less clearly but probably reflected in those
prehensively, they may bespeak incipient and/or residual total assimilatory processes. In
the first instance, cf. Bauer & Leander's postulation of tz > zd > 2Zin explanation of both
Heb. hizzakkti (BLH, 197) and BA ktib hizzamminttn (BLA, 32). In the second instance,
cf. fnn. 7 and 11 below. For Aram. anomalies showing non-metathesized and non-assimilated
tS, see Brockelmann (1961:268);cf. also fn. 6 below. An example from Mandaic is (mitzahar)
instead of expected (mizdahar) (DM, s.v. ZHR I). This form is attested in a very late text
(1204 A.H. = 1826 A.D.), so that we may well be dealing with a scribe's difficulty in con-
trolling archaizing language, aggravated by Mod. Mandaic's lack of t reflexives (Macuch,
247) and perhaps influenced by Arab. conjugations V and VI with prefixed ta-. For a sug-
gested dissimilatory origin of Heb. hiOsotdtno'run ye (fem.) to and fro!' (Jer. 49:3), see
Bergstrasser (1.112).
6 For infixed t in Nabatean (mqtry) 'called', as well as other pertinent anomalies in this
Aramaic dialect, see Cantineau (1930:72-4). For possible t prefixes in Phoenician, see CF
(p. 18, no. 40); in the Amarna letters, see the reference in Garbini (1960:127,fn. 3).
7 All the classical Semitic languages seem to have restrictions on tS clusters, at least
where the data are clear enough to allow phonological interpretation, and the favorite
method of resolution seems to be total regressive assimilation to SS: e.g. Akk. 3V/t +
suffixes beginning with s, *isbatsu > isbassu 'he seized him' (perhaps through an inter-
mediary *i?basOu),*assatsu > assassu 'his wife' (cf. von Soden 1952:30); Geez reflexive
+ 1b/S, *y9tsammay> ydssammay 'he'll be named' (cf. Brockelmann 1961:171); Qoranic
Arab. reflexive t + 1/vS, *yataaddaqu> yaâddaqu 'he justifies himself' (cf. ibid.; con-
trast Heb. hiatadddqabove).
8 Cf. Macuch (230), and note Mod. Mand. ecca (ibid., 231); Jastrow, s.v. (<iyt) IV; Brock-
elmann 1928, s.v. (metso?).
forms of 2x/t3/S verbs where these two radicals are in contact: e.g. with /nt8,
Syr. naOfiih # *naSth 'they tore her' < *natasuha, or Heb. noOi r *noGti'my
uprooting' < *nutvS.9 The non-metathesis in such forms may merely reflect a
later date for pretonic syncope than for the metathesis.'0
As a working hypothesis for this paper, the metathesis will be considered a
PNW development, but acceptance of an earlier date would do no damage to
the points made.1"The question of whether or not the metathesis should be
limited to reflexive t will be returned to briefly in fn. 56.12
9 Smith 1903, s.v. (ntas>; Gray (1936:171).
10For Aramaic, cf. ?2.29below. I have not been able to find Heb. or Aram. forms which
are likely to contain reflexes of PSem. tS in contact. Theoretically suggestive forms would
include the imperfective of simple verbs with 1l/t 2/S (PSem. *yVtSVd-), simple causa-
tive verbs with l/t 2/S (PSem. */h/?VtSVd-), and 2V/t 3V/S nouns of the measure pVqd-
(segolates; PSem. *pVtS-). However, the only instances of such forms I have been able to
uncover to date seem to stem from post-classical sources; e.g. post-Biblical Heb. yitsos
'it will ferment', causative hitsis 'he caused to ferment', kotso 'his olive mortar' (for all
these forms, see Even-Shoshan 1964, s.vv.) The dearth of more ancient forms is unlikely
to be coincidental, and Greenberg 1950 presents evidence that the paucity of VtSq and
/ptS roots dates from an ancient network of dissimilations, of which our metathesis may
be one manifestation.
" The metathesis is explicitly dated to PSem. in Brockelmann (1961:268), and the pos-
sibility of an even earlier (pre-Semitic) date is suggested in Greenberg. MSUvS (63) more
cautiously describes the process as 'very wide-spread in the Semitic languages'. In support
of their theses, all these studies adduce a variety of evidence which, while highly suggestive
when viewed comprehensively, remains circumstantial; Aramaic and Hebrew seem to be
the only classical Semitic languages where quite direct and unambiguous effects of the
metathesis are captured in morphophonemic processes. It might incidentally be considered
that, if a PSem. date is accepted for the metathesis, then this process might be put in a
before-after relationship with the assimilatory processes discussed in fn. 7: schematically,
TS > ST > SS. Then the formsof fn. 7 would be providedwith an intermediary stage of
metathesis, e.g. Arab. *yataaddaqu > *yaetaddaqu > yaâddaqu, and Heb. and Aram.
could be viewed as the only classical Semitic languages not carrying the process to comple-
tion. This view is prima facie very attractive, but fails to explain a great number of sys-
tematic exceptions in many relevant languages; e.g. why isn't the Arab. 10th conjugation
*?issapqada instead of ?istapqada? and similarly in the corresponding conjugations of at
least Geez, Akkadian, and Ugaritic, as well as in cases of infixed t following 1x1S in all
relevant languages, e.g. Arab. 8th conjugation ?istamala # *?i??amala 'he covered him-
self'. It thus seems provisionally more likely that metathesis and total assimilation are
ALTERNATIVE processes effecting the dissolution of tS clusters in the most ancient period
(cf. fn. 7), though in later Hebrew and Aramaic a secondary process ST > SS may be re-
sponsible for at least some of the SS forms discussed in fn. 5.
12 There are phenomena here and there suggesting that the metathesis is more general
than even tS > St with unqualified t. Brockelmann (1961:268) seems to want to relate to
this metathesis the observation that ' ... im Aram. und Arab. ks und ps in Lehnwortern
aus dem Griech. zu sk und sp werden.' I doubt that the two phenomena are to be related
in any specific sense; it would rather seem that the permutations attested in Greek loans
are manifestations of sporadic (lexical) metathesis, and inspection of Greek loans at least
in Aramaic will reveal exceptions galore: e.g. the thirty-odd forms showing Gk. xi > Syr.
ks ?^ *sk in Brockelmann (1928:18-19). Akkadian has a general restriction against ANY TS
cluster, where T = any dental/alveolar stop, including t. Furthermore, at least one instance
of metathesis can be viewed as an implementation of this restriction: *qadsu> qaXdu 'holy'.
But it may be rash to identify this metathesis with the more general interlinguistic process
under study. ds > sd seems to be restricted to this one stem in Akk., and furthermore dX
2.2. Like the metathesis, the changes discussed below can be posited as
Common Aramaic, though some are evidenced beyond Aramaic and others may
postdate the proto-Aramaic stage. The changes are discussed here in approximate
chronological order, but a good deal of such relative dating cannot yet be as-
sured.13
2.21. Apocope of final short vowels may have occurred by the end of the 2nd
millenium B.C.14, and is evidenced not just in Aramaic, but in all later NW
languages. Examples: *?istabi9a > *?istabi9, *?itaxija > *it?ax? 'he was
seized'; cf. also Heb. *histabbila > histabbel.
2.22. The velar spirants x and 7 merge with the pharyngeal spirants (glides)
fi and 9. This change may have been completed before the end of the 2nd mil-
lenium B.C. or shortly thereafter,'5 and is evidenced in all later NW languages.
Examples: *it?axij > *ît?ahiQ,*îtyazil > *?it9azil 'it was woven'; cf. also
Heb. *axaz > ?3ffzj 'he seized'.
2.23. In Aramaic the dental/alveolar slit fricatives 0, c, 8 occlude to stops
t, d, t, while the seemingly homorganic aYchanges to 9 through an intermediary
stage in which it is represented graphologically by (q). These changes may have
occurred around the middle of the 1st millenium B.c.;16 they seem to have been
underway in EA, where the reflex of * is sometimes written (d) but sometimes
(z) (putatively for c),17 and that of *E vacillates between (9) and (q). Examples:
*?itafii > *?itPaIiid,but EA &ahdv(zhb) 'gold' < *;!ahab,though daxdr (dkr)
'male' < *&akar.ls8
2.24. Immediately post-vocalic non-geminate non-pharyngealized stops are
spirantized. This change affected Aramaic after the changes of ?2.23, but may
have affected other NW dialects at an earlier date.'9 Examples: *?iStapik >
seems to be resolved by total assimilation (cf. fn. 7) in other forms, including the homo-
paradigmatic feminine form: *qadsatu > qasiatu. For discussion of these Akk. phenomena,
cf. Reiner (1966:50-1). Finally, the remaining classical Semitic tongues seem to tolerate
dS, indeed within stems cognate to Akk. qaSdu: cf. Heb. hbr qoAiH 'my holy mountain'
(I 2:6), Syr. qudQ 'sacrum' (Brockelmann 1928, s.v.), Jud. [Onkelos] qu$st (Dalman
1960:144); cf. also Arab. qudsun.
13For a recent comprehensive treatment, see Garbini 1960. Note incidentally that the
question of strict chronological ordering may in many instances be inappropriate, since
changes may overlap in time, and in fact manifest a variety of mutual relations in time
and space; cf. Wang 1969 and Malone 1971 (esp. ??3, 4).
14 Cf. Garbini (1960:80-1); Harris (1939:59-60).
15 Garbini (1960:52-3); Harris (40-1, 62-3).

16 Cf. Garbini (1960:34-5),where the suggestion is also made that (q) for the intermediate
reflex of *d may have represented a. C]. Mand. shows (q) in a handful of items with original
*d, but it is interesting to note that one of the commonest of these, (arqa) 'earth' < *âria,
is consistently pronounced [ara] by modern Mandean priests. Macuch (77) explains this as
*arqa > ara by loss of stops (here q) following r under certain prosodic conditions; but his
statement is by no means without exception, and it seems possible that ara might continue
an earlier ?ary4. Other forms with (q) < *d are pronounced with [q], however; cf. the forms
and discussion, ibid. (95-6).
17The question of PSem. *d and its reflexes retains many puzzles; cf. e.g. Garbini (1960:
36, 194-6), Noldeke (1964:43-4), Macuch (66-9).
8 Leander (8-9).
19 Cf. Harris (66-7). Garbini (1960:22-3, 25-6, 39, 49-50) discusses the evidence in some
*it?aiid
*?iStafix, > ?iO?aitj, *îstalib >*istaliv; cf. also Heb. *katab > ko6dv
'he wrote'.
2.25. Stress is shifted toward the end of a word, according to a prosodic pat-
tern which may have been dictated in part by grammatical conditions. Precise
dating of this change is at present impossible:20 it may have been very early, and
causally connected with the apocope of final short vowels (?2.21), but in any
case it is presupposed by the syncope of pretonic open short vowels (?2.29).
It is also not clear whether or to what extent this change is to be identified
across dialects, but the conditions seem to have been slightly different from
Aramaic to Hebrew, at least in the verb.2' For present purposes only this much of
the Aramaic shift need be stated: In strong verbs, stress shifts to the last VC of
the word. Examples: *?ija?ij > *ij?at Wj 'he was seized', *?ji?afiiu >
*i@aW{u 'they were seized'.
'-.4
2.26. Immediately preceding a guttural glide (at least hfi 9) or r as 3rd

radical, short vowels lower to a. The precise conditions for this change are diffi-
cult to ascertain; specifically it is difficult to judge whether seeming exceptions
bespeak non-fulfillment of some integral condition for the change, or rather
indicate later changes effectively restoring the original vocalism.22It is also likely
that we may be dealing with a family of similar changes, related typologically if
not genetically. Specifically, pre-guttural V > a may have occurred both in
PSem. and then later again in PNW, while Vr > ar is pre-eminently Aramaic
though traces of similar patterning are evidenced elsewhere.23The changes are
likely to have been completed in Aramaic by the 2nd half of the 1st millenium
B.c.24Finally, a sporadically replaces i even in default of a following H or r.25
detail and concludes that there is no unimpeachable testimony for spirantization before
the Christian era-except in the case of the bilabials p andb, which in part of the NW
area may have been unconditionally spirantized.
20Cf. Garbini (1960:73-4). It is usually assumed that the accentual system in NW was
similar or identical to that of Cl. Arabic prior to the stress shift, but substantive evidence
is wanting; see the discussion and references,loc. cit.
21For a suggestion as to the Aramaic-Hebrew relationship in part of the noun system,
cf. Malone 1971.
22 Forms such as EA
(?stmy9) 'he was heard' (Leander, 55) and Mand. (etikpir) 'they
were bound' (Noldeke, 15) should perhaps not be interpreted as exceptional ?istami9 and
e96xfer, but rather as ?istamf9 and eOdxfir,where maintenance of the long I is quite regular.
Leander (loc. cit.) suggests that this I may be explainable as contamination from the pas-
sive participle measure poqid; internal-reconstructive Mandaic evidence for positing I may
be seen in the morphophonemic derivations cross-referred in Malone (1967:132); /itibhiirt/
etc.) Cf. also traditional pronunciations of Cl. Mand. such as [mipsiq] for (mipsiq) 'cut off'
(Macuch, 4). Later changes are discussed for CP in Schulthess (1965, ?34a), and for Mod.
Mand. in Macuch (?69). Mand. nouns like (rima) and (bila), cited by Noldeke (16), also
reflect innovatory vocalization; cf. ?3.11 below.
23 For the PSem. development, cf. Brockelmann (1961:194). Specifically PNW develop-
ments are probably reflected in perfectives such as Syr. homd9, Heb. khmd9 as against Arab.
sami9a. For piecemeal evidence of Vr ' ar in Hebrew, see Bergstrasser (1.153).
24 The vocalism of the Aram. cuneiform incantation tablet from Erech implies that this
change had already occurred, at least in the dialect of the tablet; see Gordon (1938:112,
no. 36, 1940:31) for explanation of the one seeming exception.
25 Dalman (267) describes this replacement in the simple reflexive as 'irrige Schreibungen
Examples: *istauv9 > *istavd9, *i,taxih > *iStaxxin 'he was found', *îstamir
> *jistamdr; sporadic replacement *yiO?am > pre-JP *yi?a?d& 'it is poured'.26
2.27. s merged with s in Aramaic during the 2nd half of the 1st millenium
B.C.27 The change had apparently not yet occurred in the Erech tablet, but was
incipiently evidenced in EA, and was underway in BA.28Example: *?istavd9 >
*9istavd9, but EA impf. intensive yistabbd9.29
2.28. i lowers to e. It may be assumed that this is a relatively late change,
increasingly evidenced in later forms of Aramaic, both in absolute number of
morphemes affected and in occurrence under ever wider structural conditions-
i.e., approaching the status of an unconditioned sound change.30The change is
incipiently or not at all attested in the Erech tablet,31is underway in BA (under
certain prosodic conditions),32and is virtually complete in post-Christian dialects
(under certain prosodic conditions in WAram.,33unconditionally in EAram.)34
..., welche durch Wegfall der Mater lectionis [i.e. (y)] im Konsonantentext entstanden'.
But this explanation can hardly obtain for similar Mandaic forms, e.g. (milgaft)(N- (milgit))
'withheld', (etimlak) (- (etimlik)) 'has been placed in charge'; see DM, s.vv. LGT and
MLK I. A similar (or identical) phenomenon in nouns may be that given as change (h3)
in Malone, MS a (?8.1).
26 Syriac (?eOesg5,cf. ?2.32) doesn't show this replacement in this root. Neither does
Mand. este6 (?2.33), though the form (nistad), listed mistakenly in DM (s.v. A?D) as a
reflexive intensive, does betray a-vocalism. A real intensive ought to read *(nitasad).
27 Cf. Garbini (1960:46-7).
28Gordon (1938:113,no. 50), Leander (10-11), BLA (26-7).
29 Written (ystb9); see Leander (56).
30 For another likely Semitic example of a change's losing conditions through time, cf.
Malone 1971, ?5.1; for the general formulation of such an evolution, see the reference cited
ibid. (fn. 56).
31Gordon (1938:112, no. 39).
32 BLA (30-1), and fn. 33 below.
"3The conditions, at least as inferrable for Judean from the supralinearly pointed forms
of the Targum Onkelos (Dalman passim, but esp. 249-91), include primarily STRESSED
CLOSED SYLLABLES (seeming exceptions are explained in Malone, MS). Similar conditions
are apparent for BA also, except that there is vacillation in syllables closed by C# but not
in those closed by CC (thus yiOyahev- yiOyahiv'is given'; but only tdq4ft,not also *tdqlft
'thou wert strong'). In Mod. WAram., especially the dialect of Ma91ula, stressed closed
syllable remains a central condition; cf. Spitaler (1966:2-7, 10, 46-7). The forms Umsa
'Sonne' and siqja 'bewasserbares Gebiet' (ibid., 139) are interesting because their seem-
ingly exceptional vocalism is understandable in the same terms as analogous sibilant forms
in Eastern Aramaic; cf. fn. 34 below. In default of contrary evidence, we may assume for
EA conditions similar to BA; and for Cl. WAram. generally (incl. CP and Galilean JP),
conditions similar to Judean. In this vein, the different assumptions of Schulthess (1965:
20-1) strike me as too hasty.
34For Syriac, cf. Brockelmann (1965:30-1). While Cl. Mand. writing is moot with respect
to i vs. e, and the trad. pron. of Cl. Mand. by Mandean priests evidences quasi-patterning
just short of free variation (cf. Macuch, 102-4), examination of inherited Aramaic forms
in Mod. Mand. reveals a situation virtually identical to that of Cl. Syr., viz. unconditional
lowering of i to e, with a handful of exceptions for the most part involving an abutting
sibilant (and possibly, but by no means conclusively, forms like biyya in fn. 64 below).
The following clear cases are taken from Macuch's Vocabulary of the vernacular (489-526):
ni.md 'breath' (but neGemOa'spirit'), liUana 'tongue', riiana 'prayer', and possibly lissa
'dough' (Syr. laysa). s does not preserve i in all cases, however: e.g. neXemOa(above),
senna 'tooth', senana 'teeth'. ti,3na 'straw' seems anomalous, but influence from Iraqi
Examples: *yiOyahiv > pre-BA *yiQyahiv *yi6yahev 'is given'; *iohafix >
pre-Jud. *iehafex,36 pre-Syr. *eohafex 'was changed'.
2.29. The nearest pretonic open-syllabic short vowel reduces to a immediately
following a heavy syllable (after VCC or VC), or is syncopated immediately
following a light syllable (afterVC).36 The Erech tablet captures in process
this relatively late change, which may then be tentatively dated around the last
quarter of the 1st millenium B.c.37 Examples: WAram. (pre-JP, CP) *?iâmdr >
*?i0?amdr'it was said', EAram. (pre-Syr., Mand., BT) *?eamdr > *?e3amdr.
2.3. At this point we will switch our perspective from Common Aramaic to
the history of individual languages, and specifically study a variety of processes
whereby the sequence '? is effectively dissolved. Though, for sake of convenience,
these processes are treated as post-Common Aramaic, at least one of them
(?2.31) probably predates some of the changes in ?2.2, and may in fact be more
widespread in Semitic than just in Aramaic (cf. also fn. 38). Also, when we say
that, as a result of these processes, the sequence ?âis 'effectively' dissolved, this
means that dissolution ofâ may be no more than a by-product (?2.31), or merely
a special case of a more general process (at least ?2.33). The processes studied
will be called ASSIMILATION,VOCALIZATION, and SYNCOPE (??2.31, 32, 33 resp.)
2.31. ?9 is dissolved in several Aramaic languages by total ASSIMILATION of ?
to the immediately preceding consonant. In our cases this consonant is the
reflexive-prefixal 0- (< *t- by ?2.24), and the resulting 00 occludes to tt.35This
Arab. tibin may not be out of the question. yidlit 'I bore' (ibid., 313) is probably to be
understood as progressive assimilation from y. In sum, it seems quite likely that the distri-
bution of i and e in Cl. Mand. was like that in Cl. Syr., and in default of contrary evidence
we may assume the same for BT. The free variation in the traditional pronunciation of
Mandean priests most likely reflects in part adventitious factors, not least of all reading
pronunciation (Macuch's'abagadical' pronunciation; see his references, 641).
There is evidence for i > e elsewhere in NW Semitic, notably in Hebrew where i is re-
tained in closed pretonic syllables. A parallel lowering, u > o, is likewise widespread in
NW Semitic, though the conditions in the various languages differ slightly from those for
i > e. For Mandaic, cf. ?3.13.
36Dalman (259) also cites (?isfaliyb) 'was crucified', which he explains as pausal length-
ening (257). This may also be interpretable in terms of fn. 22 above.
36 The change might equally well be limited to syncope, particularly if the focus is on
phonemic (or morphophonemic) repercussions. In such a case the statements of ??2.32 and
2.33 must be correspondingly adjusted, as will be specified.
37Cf. Garbini (1960:77-8). BA is pointed to show the reduction (and syncope) as a fait
accompli, but Masoretic anachronism is a possibility. A similar caveat holds for EA, but
for simplicity's sake these changes will be assumed for EA below.
38 Occlusion of slit fricatives is common in the Semitic languages, early and late: cf. the
early Aram. developments sketched in ?2.23, later conditional occlusion in Cl. Aramaic
and Hebrew (cf. Brockelmann 1965:41,55-6; BLA, 43-4; BLH, 210), and further such occlu-
sion, in part unconditional, from classical to modern languages (cf. Macuch, 32-40, and
Mod. Heb. t, d, g < Cl. Heb. 0,8, y). It is quite likely, however, as intimated above, that
assimilation of ? antedates the spirantization of ?2.24, in which case the development
would be *t? > tt, whereby geminate tt would be exempt from subsequent spirantization.
In such a case this assimilation might be related to some of the similar develop-
ments sketched in Brockelmann (1961:158-9). It is particularly inviting to relate the shap-
ing of the Aram. reflexive causative (as Brockelmann does, ibid., sec. ga), i.e. *'it?apqad >
resolution in the simple reflexive seems to be absolute in EA and BT, widespread

in WAram. (both JP and CP),39 and sporadically manifested in Syriac (lexical
selection). There is no clear evidence for such assimilation in Mandaic (but see
fn. 47). Examples: *?j3ie?aSB > EA ?ittaEiiû'they were seized', *yiOe?asdr>
EA yittasdr 'he is constrained',40*eB3andM> BT 7ettah65 'it was closed', *te?dasdr
> BT tettasdr 'she is forbidden',4' *?ie?amdr > Gal. and CP ittamcdr ?iO9arndr
'it was said'; *yij9?slV4 > Gal. yittaSd&'it is poured' but *yiO?asdr > Jud.
yiO?asdr 'he's bound'; *miO?asadr> CP mittasdr ' mi6?asdr 'bound',42 *?eO?afie >
Syr. ?ettahS 'he was seized'.43
2.32. The normal way of dissolving ?a in Syriac is by VOCALIZATIONto the
full vowel e.44This resolution is most likely part of a nexus of developments in-
volving the glides 9, y, and w; and though quite similar processes are evidenced
elsewhere in Aramaic, Syriac seems to be the only language where ?a is vocalized
in the simple reflexive.45Examples: *?eOeaxel > ?eeexgl 'it was eaten', *e?@?a?d >
?eOeeS 'it was poured', *?&e?asdr > ?eOesdr 'he was bound'.46
?ittapqad as reflexive to ?apqid. But if Common Aramaic ?ittapqad is explained in such

terms as a reflex of *?it-?a-pqad, for the reflexive of causatives in *?a-, then on the basis of
position class (specifically, causative prefix following reflexive prefix and immediately pre-
ceding the stem), simplicity would dictate analogous reconstruction for the non-*7a-causa-
tives, of particular interest here the *sa-causatives like *?it-4a-pqad. But in that case the
resulting Common Aramaic pattern ?istapqad could only be understood in terms of meta-
thesis *tX> at, which would be evidence for the permutation under conditions other than
immediately following h/S. Corresponding remarks are in order for languages taking
INFIXED t (fnn. 6 and 11), unless the causative consonant-whatever its nature-system-
atically preceded the reflexive t in PSem. In that case the proto-forms of the Common
Aramaic *sa- and *?a-causatives would be respectively ?iXtapqad and ?i?tapqad, which
latter > ?ittapqad by regressive assimilation as in the Arab. 8th conjugation with lv/?,
e.g. *?iVtaxata > ?ittaxaja 'he occupied' (cf. Brockelmann 1961:177). Note, however,
that forms such as Gal. (?t?pq) and Jud. (mit?apqo?) (Dalman, 97, 297), as well as the
h-causative (?thhsynn) of the Kandahar I inscription (KNS, 134), may be interpreted as
militating against this suggestion. Finally, it seems possible that part or all of the assimila-
tions subsumed here under ?2.31 are in some way relatable to change Dl discussed in ?3.12.
" It is of course possible that the assimilation is much more widespread in WAram. than
the orthography indicates, and especially that spellings with (?) are largely etymological
and/or morphophonemic. See the spellings cited in fn. 42.
4?Spelled (?thd[w]) and (ytsr) respectively (see Leander, 59).
41(?ythyd), (tytsr) (Margolis, 41).
42 Gal. (?yt(?)mr) (Dalman, 303); CP
(<d(?)mr)(Schulthess 1903,s.v. (?mr); Gal. (yit.9ao)
[Tiberian pointing], Jud. (yit?osar) (Dalman, ibid.); CP (mtVsr),(mtsr) (Schulthess 1903,
s.v. (?sr)).
43(?ettheo) (Brockelmann 1965:23, where *t? > tt in this stem is explained as guttural
dissimilation from the following h; but cf. fn. 47 below).
44Under some conditions, to other vowels; cf. Brockelmann (1965:44).
46Various pieces of evidence, including the partially historical orthography (see fn. 46)
which places the vowel point for e BEFORE the (?), suggest that vocalization proceeded in
stages: *?eo?apqdd > *?eEe?pq6d > ?eOepqFd,in which case it may be related to the so-called
'Aufsprengung' in various Aramaic languages and Hebrew (cf. Bergstrasser 1.135-6; a
Syr. example is hexem&afrom and beside hexmOa&, see Brockelmann 1965:11). Several
Aramaic languages show vocalization of ?9 in initial position: regularly in Syriac (ibid.,
44), under certain grammatical conditions in Judean (Dalman, 298), sporadically in BA
(BLA, 67p). Analogous processes involving ya (and wo) are equally widespread; Syriac
2.33. Mandaic ?9 is SYNCOPATED interconsonantally, a fate which is shared at

least by ya.47 Examples: *?ê?axl, *?eO?aS6&,*eOâsdr, *ê^?amdr > pre-Mand.
*?eOxel,*?@e6?, *eOsdr, *?e6mdr.
2.4. At this point we are dealing with specific-language forms in all the
Aramaic dialects examined except for Mandaic,48and it will be noted that the
phonotactics of the 1v/? verbs split into four classes: -VO?aqVdwhere ?9 is re-
tained (JP and CP examples of ?2.31), -VttaqVd in assimilated forms (?2.31),
-VOeqVdin vocalized forms (?2.32), and finally -VOqVdin syncopated forms
(?2.33).
Note now that only in pre-Mand. -VGqVddoes the 2nd radical (q) abut upon
the reflexive prefix (8). Hence in the case of 2V/S verbs it is ONLY IN PRE-MANDAIC
that the conditions for the pristine reflexive-prefixal metathesis of ?2.1 are once
again fulfilled, and it is only in this language that the metathesis in fact recurs,
after a period of a millenium or more.49Examples: *e0eg > ?e8l0& > Mand.
examples involving Aufsprengung may be seen in Brockelmann (1965:45), and most com-
parably in the simple reflexive of 1i/y verbs, e.g. *?eByalb (> *?eOiylS) > ?eOzljd 'he was
born' (ibid., 93). Finally, if a is a spurious phonological entity for Syriac, or an allophone
of zero (cf. Morag 1962:52-3), and the changes are posited at the phonemic level, we can
view Aufsprengung as a straightforward instance of epenthesis (CCC > CeCC), in our
special case followed by loss of ? (*?eo?qed> *êoe?qbd> ?eOeqbd).
46 <?eEe?xel), <?eAe??et), <?eEe?sar) (Smith} B.VV. (?exal), <?e?atS), (?esar)).
> pre-Mand. *?êa5U> Cl. Mand. eOle8S'he was born' (spelled (etlid>,
47E.g. *?eOyal6&
DM, s.v. YLD). In initial position ?9 and ys are normally vocalized, as in Syriac (cf. Ma-
cuch, 295 ff., 308 if.) It is possible, incidentally, that what is here called Mandaic syncope
results from the assimilation of ?2.31 followed by simplification of C1C19> C1, in which
case the first example below, e.g., should be *?eO?9x61> *?ettaxdl> pre-Mand. *?etxel.
This alternative explanation has much to recommend it, including the attestation else-
where in Semitic of both processes (for C?9 > CC7, cf. fn. 38 above; for CC9 > C, cf. BLA
58, BLH 220, and for Mandaic itself Macuch, 4, 144). Cf. also fn. 49 below. In this general
vein, it should be mentioned that some cases of ?i/ett9qi/dd, as explained by assimilation
under ?2.31, might in fact have further simplified to ?i/etqi/dd. The orthography is quite
often moot in this respect, specifically in the case of simple (t) with no diacritic notation
of occlusion and/or gemination: e.g. Syr. (?etmar) 'it was said' (Smith, s.v. (?emar)) or
most of the forms listed in fnn. 40-42 above. Cf. the uncertainty implicit in Dalman (?67.2).
Finally, if a should be discounted in these changes, Mandaic syncope would take the form
C1?C, > C1C2in the first alternative discussed, or C1?C, > C1C1C2> C1C2in the second
(cf. fn. 45).
48Examples listed passim above, whose derivations from PAram. were not completed,
include BA yiey9hi/6v (yiOyahi(y)v, yiOyahev) (BLA, 171), Jud. ?iehafex (?ithapeyk) (Dal-
man, 259), Syr. ?eoh9f6x (?e6hfex) (Smith, s.v. (hpak)) [for all the preceding forms, cf.
?2.28]; Syr. ?estdvd9(?estva9) (ibid., s.v. (sba9)) [?2.27]; EA ?istdxdh (??tkh) (Leander, 55)
[?2.261;BT ?eltdf6x (??tp(y)k) (Margolis, 41), Syr. ?e?talv (?estlev) (Smith, s.v. (slab))
[?2.24]; Syr. ReO9azdl (?eO9zel) (ibid., s.v. (9zal)) [?2.22]; CP ?izda94f (?zd9p) (Schulthess
1903, s.v. (z9p)) [?2.1].
49If the Mand. resolution of ?9 followed the second course mentioned in fn. 47, via.
*?eO?aS6d> *?ettaS6d > ?etS&d,the conditions for this re-metathesis would be IDENTICAL
to those for the PAram. metathesis of ?2.1, since the reflexive prefix would be the stop t
rather than the fricative 0. But there is no a-priori reason to expect such perfect symmetry
between the earlier and later changes, and moreover the precise conditions of the original
metathesis are not known (cf. the discussion in ?2.1); cf. also fn. 53. Note finally that the
observation that the conditions for the metathesis were met 'only in pre-Mandaic' should
PAram. ?itsamira ?it7asira

?2.1 ?ISTAMIRA
?2.21 ?ISTAMIR ?it?asir
?2.24 ?iOasir
?2.25 ?ISTAMfR ?iO?asir
?2.26 ?IsTAMXR ieOasdr
?2.28 ?ESTAMAR ?eOasdr
?2.29 ESToMAR ee? asdr
?2.33 9eOsdr
?2.4 ?ESuAR
Mandaic SSTaMAR fSOAR (fSTAR)
TABLE 1
B?GeZ(or gtez5) 'it was poured', *eOsdr > *?esOdr > Mand. tsear (or dstar) 'he
was bound'.50
It is hence quite conclusivethat Mandaicpresentsevidence of having under-
gone the metathesistwice, first in PAram. (or PNW or even PSem.) for 1-/S
verbs, and much later in pre-Mand.for 2V/S verbs, after the intervening?o had
been lost by the characteristicallyMandaicprocessof syncope. This recurrence
is summedup in Table 1, where small capitals mark the metatheticforms.
2.6. Recently Chafe (1968:131-6) has formulatedthe notion of PERSISTENT
CHANGES, that is, ones which 'remainin effect over a long periodof time during
the history of a language, and ... exert their influence whenever,through the
operation of other changes, their structural descriptionscome to be fulfilled'
(131). It seems primafacie likely that MandaictS metathesismay be considered
an instance of such a persistentchange, wherebythe later specificallyMandaic
metathesis is interpretableas the persistence of the earlier PAram. process
exertingits influencewhen, 'throughthe operationof other changes',i.e. ??2.25,
2.29, and 2.33, the structuraldescription(specifically,t abuttingupon S)61came
to be fulfilled. Chafe's discussionof changes is synchronicallyoriented, i.e. he
restricts his general treatment to morphophonemicrules resulting from dia-
chronic changes; thus the persistent nature of the Mand. metathesis will be
pursuedfromthat point of view, based mainlyon Mialone1967.
not be interpreted as implying that those conditions were SUFFICIENT for the re-metathesis
independent of the language. That is, it is not meant that ANYAramaic language having
verbs with reflexive t (0) immediately preceding 2VS would automatically undergo the
metathesis. Indeed, there may be such cases which did not entail metathesis; cf. the dis-
cussion toward the end of fn. 47.
64 (e?tid), (estar) (DM, s.vv. ASD and ASR). In the derivation *?es@O > di9eQ, we see
the loss of pre-Mand. v (Macuch, 90), and stress shift to the second-last full vowel (Macuch,
137; it is possible however that this stress change postdates Cl. Mand.) As for the uncer-
tainty of S?0eS or dAtet (i.e. occlusion following s), both Mod. Mand. and the trad. pron. of
C1. Mand. leave us in the dark, since the original Aram. stop/fricative distribution has
been altered in a patchwork of ways, and specifically verbal affixes, including reflexive t,
are invariably pronounced as stops (Macuch, 32-40; cf. also the pronunciations listed
below in this note, the verb-affixal examples listed in Malone 1967:280-5, and fn. 64 below).
Mand. forms listed passim above whose derivations from PAram. were not completed are
60xel (etkil), trad. pron. [6txel] (Macuch, 90); lOmar(etmar) [etmar] (ibid.) [for the pre-
ceding forms, cf. ?2.331;6stamar(estmar) [6stamar] (ibid., 125) [cf. ?2.26].
"1The difference between t (?2.1) and 0 (?2.4) is immaterial here; cf. fn. 53.
Chafe expects persistent morphophonemic rules normally to apply at depth

I, i.e. among the last rules before the phonemic or systematic phonetic level.
Provisionally, this expectation is fulfilled for C1. Mandaic. Although the syn-
chronic analog of the later Mand. change (rule P8.2 in Malone 1967:45, 187) is
relegated to what Chafe would call depth VI (my order 8), none of the subsequent
rules (i.e. those of depths V-I = orders 8-13) presupposes application of rule
P8.2, and hence the latter could be shifted to depth I (order 13) with only minor
technical adjustments.52
Chafe finds that some persistent rules apply only at depth I, while others may
apply at one or more additional depths. The Mand. metathesis is clearly of the
latter sort, since there is also a synchronic analog of the PAram. change (?2.1
above), namely rule P2.6 of order 2 (depth XII) which is identical in content to
P8.2.53 Moreover, synchronically there must be two distinct applications of this
metathesis: P2.6 must precede order 7, and P8.2 must follow order 7, for reasons
discussed in Malone (1967:164-5, 187).
Finally, Chafe suggests that 'it must be the case that only rules of a particular
kind-perhaps of a high degree of universality or the like-can belong in (or
enter into) this category [viz., that of persistent rules], but I am unable to say
anything specific on this subject at the present time' (135). This leaves us at an
interesting frontier indeed, for two reasons: (1) while a 'high degree of uni-
versality' may be defensible INTRA-SEMITICALLY for tS metathesis (cf. above,
?2.1 and fnn.),54the phenomenon must also be studied from the point of view of
general linguistic theory;55 (2) the tS metathesis is likely to be in part GRAM-
52 Most specifically, the conventions governing 'linkage' would have to be changed
slightly (but cf. also fn. 56 below). For general discussion of linkage, cf. Malone 1967. Inci-
dentally, linkage is not to be identified with the totally different situation called'linking'
in CH (419-35).
'3 The formal statement of the rule is cast in distinctive features (Malone 1967:41),
and hence the difference between stopt (cf. ?2.1 above) and homorganic fricative 0 (?2.4)
can be ignored in favor of the 'archiphonematic' common denominator of these two con-
sonants. Since Malone 1967 contains no analog of the spirantization of ?2.24 above, this
specific question did not arise at the time, but the pertinent adjustment involves no diffi-
culty.
64 Noldeke's discussion, which however falls just short of explicitly stating that Man-
daic underwenttS metathesistwice, impliesrecognitionof such metathesisas persistent

in Semitic: 'Wie vollstandig hier auch jede vokalische Spur [i.e.?9] geschwundenist, zeigt
der Umstand, dass, wenn der 2. Radical hier ein Zischlaut ist, die Umstellung eintreten
muss, WELCHE IM SEMIT. tBERHAUPT NOTIG 1ST, WO DAS t DES REFLEXIVS UNMITTELBAR
EINEM ZISCHLAUT
VOR ZU STEHENKOMMT' (70; emphasis supplied).
65
Modern Irish seems to show a similar metathesis in foreign borrowings, e.g. loistin
[lo:sti:A]'lodging' [lodzii], cistin 'kitchen', Risteard 'Richard', but not e.g. d'ditseach
'Dutch'. There is also no metathesis when enclitics in s are suffixed to forms ending int,
e.g. duit 'to thee' + se emphatic particle -* duitse 'to thee thyself'. On the other hand, the
phonetic conditions for the metathesis may be wider in Irish than in Semitic, involving the
transposition of ANY stop + s (cf. Brockelmann's suggestion criticized in fn. 12 above),
frequently in foreign borrowings (e.g. Dinneen 1927, s.VV. aspal, coimpl6asc but not
exclusively (ibid., s.vv. feiscint, foisce). Since I have not studied these Irish phe-
nomena in any depth, the preceding discussion can claim no more than suggestive value.
Perhaps also typologically related to the Semitic development is the evolution of Slavic
palatalized t: ty > 1t at
> >St (Zuravlev 1967:590); the approach to this sketched in An-
MATICALLY CONDITIONED, at least for Cl. Mand. rule P8.2.66 In such cases, the
problem of universalistic typologizing of phonetic changes might be complicated
in a variety of ways by attendant grammatical (non-phonetic) conditions.
3. The second type of systematic Mandaic metathesis is partially overt and
observable, partially covert and inferential. Of special interest is the dovetailing
of both types into a promising hypothesis explaining not only seemingly unre-
lated phenomena within Mandaic, but conceivably in Akkadian as well (fn. 66).
Of equal interest is the fact that the hypothesis now to be ventured was suggested
to me by Kiparsky's explanation of seemingly homologous patterning in a non-
Semitic language, Classical Greek.
3.1. Table 2 synoptizes the Mandaic developments to be discussed. In the
first column are pre-Mand. (PAram.) forms,57several of them undergoing the
changes listed in ??2.22, 2.23, 2.29 above, and in ?3.11 below. The next three
columns show the forms undergoing changes B, C, and D (described below), of
which C is the metathesis proper; these letter labels bring the changes into corre-
spondence with similar Greek changes posited by Kiparsky (Table 3 below).
The following column leads to C1. Mand. by the changes of ??2.24, 2.27, 2.28,
of ?3.13 below, and of fn. 50 (above). The last column derives Mod. Mand. by
the changes of ??3.14-3.17, below.68Hypothesized Cl. and Mod. Mand. phonetic
strings are enclosed in / /, orthographic forms in ( ), and pronunciations
provided by Macuch in [ ] with page reference.59Chronologically, it is not
possible to achieve a strictly cogent ordering of all the changes (cf. fn. 13), but
the sequence as stated is tentatively posited. The seeming prematurity of ?2.29
within ?2 may well attest an earlier date for this change than that assumed in
?2, at least for pre-Mandaic.60
3.11. a > i in closed pretonic syllables in Hebrew and Aramaic. The change
is apparently not carried out fully, and spreads irregularly through the lexica of
individual dialects (cf. Brockelmann 1961:146-7 and Malone 1969:199, fn. 18).
3.12. The following changes affect Mandaic:
(B): Pharyngeal glides 9 and fi drop to laryngeal? and h (cf. Macuch, 79-80).
dersen (1969:558) provides a promising path toward general understanding of at least

some such developments.
56 A purely phonological formulation was attempted in Malone 1967, with the result that
rule P8.2 might not have been relegable to depth I if certain non-verb forms had been
morphophonemicized in the way suggested (704, fn. 57). However, this difficulty can be
overcome if the metathesis is restricted to reflexive-prefixal t, which I now feel is, all told,
the better way (perhaps the only feasible way) of analysing the situation SYNCHRONICALLY
for Cl. Mandaic.
67 For etymological particulars, cf. Brockelmann 1928, and DM.
58 Note that it is in general difficult to assign a terminus post quem to changes posited
in the development of Mod. Mand. In fact, ANY of the processes in ??3.14 through 3.17
may have applied in Cl. Mand.
69These forms are listed exactly as by Macuch, i.e. often with stress and ancipital -a
(cf. ?3.15) to be understood.
fi This much is internal-reconstructively clear for Mandaic: syncope must precede 3.11,
e.g. *baAara> bagra > biSra ... besra 'flesh'; 3.11 in turn must precede D, e.g. in class 2
of Table 2; D in turn must precede ?2.24 and ?2.28, e.g. *zir9a and *?uhdaof Table 2.
PRE-1MANDAIC
B C D CLASSICALMANDAIC
(?g2.22, 2.23, (??2.24, 2.27, 2.28, 3.13, fn. 50)
2.29, 3.11)
1. -VCh-
*miXha'oil' misha mihsa (1) missa, /mT?r/ (misa) [meNSa]8
**yarxa 'mouth' *yarha yarha yahra (3) yahra /yahra/ (iahra) [ya/aehra
13, 108
2. -VC9-
**zar9a 'seed' *zir9a zir?a zi?rd (1) zirra /zerra/ (zira) [zerra] 10
**0arya 'door' *tir9a tirâ ti?r& (2) t1tr /tira/ (tira) [tlira] 212
3. -VhC-
*lahma 'bread' lahma (1) lamma
*suhda 'bribe' suhdd (2) iudd /sua/ (suda) [isuda]84

**raxXa'reptile' *rihsa rih9a (3) rihms /rmhsa/ (rihna)
4. -V9C-
*?u91a'handful' (1) sulla /?611a/ (sula) [?olla] 174
**ra9ma 'thunderclap' *ri9nma ri'md (2) rZma /rima/ (rima) [rima] 92
5. -VCh-
*timha 'wonder' tihmd (3) tihma /tehma/ (tihma) [tehma] 8
6. -VC?-
*sin7a 'hatred' si3nd (2) sm,na /sfna/ (sina) [sina] 91
7. -VhC-
**sahra 'moon' *sihra (1) sirra /serra/ (sira) [serra] 55
**nahara 'river' *nahra (3) nahra /nahra/ (nahra) [nahra] 1
8. -V?'C-
*ki?ba 'pain' (1) kibbd /kA bba/ (kiba)*
*bi?ra 'well' (2) bZra /bfra/ (bira) [bIra] 114
TABLE 2
(Asterisks after words refer to discussion in fn. 64.)
(C): A laryngeal glide , h metathesizes with an immediately preceding con-

sonant if the laryngeal is the 3rd radical.6'
(D): There are three possibilities, but it is not clear what factors govern the
choice (cf. ?3.41, below):
(D1): A postvocalic laryngeal glide totally assimilates to an immediately
following consonant, i.e. VHC > VCC.
(D2): A preconsonantal laryngeal glide totally assimilates to an imme-
diately preceding vowel, i.e. VHC > VC.
(D3): Either assimilation is optional where H = h.
3.13. A perusal of Miacuch's Vocabulary of the vernacular (489-526) sug-
gests that inherited Aramaic u > o unconditionally in Mod. Mand. (but not in
the trad. pron. of Cl. Mand., for which see Macuch, 101; cf. [pulhana] in fn. 61
above). This result is much like that for parallel i > e in fn. 34 above, and the
change will be attributed to C1. Mand. for the same sorts of reason.62
3.14. Sporadic assimilatory labialization is rather common in Aramaic of all
periods (Brockelmann 1961:200-1; Macuch, 114-7).
3.15. The nominal suffix -a (status emphaticus) is neutralized as to length in
both Mod. Mand. and in the traditional pronunciation of the Cl. language (cf.
Macuch, 102).
3.16. For Mod. Mand. (but not the trad. pron. of Cl. Mand., for which see
Macuch, 34), examination of Macuch's Vocabulary of the vernacular suggests
that in inherited Mand. forms obb > of3f3;following any other vowel, bb remains
occlusive. 63
3.17. [a] and [ae]may be in free variation in Mod. Mand. (passim in Macuch's
Vocabulary of the vernacular, e.g. gadda s.v. 'fortune' but gaedda s.v. 'hap'),
and hence also in the trad. pron. of Cl. Mand. (cf. also Macuch, 101).64
61Other laryngeals fail to metathesize, e.g. li/h in eOhdmbal$ *eh9dmbal'was spoiled',
2i/h in dvhos'I search', suffixal h in a8r9x6nhon'lead them!' (DM, s.vv. HB,L BH?, DRK),
also miscellanea like (haghag) 'convulsion', whatever the precise phonology (DM, s.v.) An
exceptionally non-metathesizing 3Ah seems to be Cl. polhana 'worship', (pulhana), trad.
pron. [pulhAna],88.
62 The only exception uncovered is [rusuma] 'token', but the authenticity of this form
as Mod. Mand. is highly suspect. Rather, it is likely to be a spelling-pronounced Cl. word

with ritualistic significance (cf. DM, s.v. rus(u)ma, and the points adduced in Malone
1969:196, fn. 10).
63 Examples (retrievable in Macuch, s.vv.) in addition to kofS,jaare to3IS3ar '(to) break',
hoXjet '(to) mix', soj58a'seven', as well as Cl. ?of3,3an'baptize me!' (ibid., 91); but gabbf
'(to) choose', rabba'great', lebba'heart', zabben'(to) sell', yum he,Ssabba'Sunday'. Preceding
o seems to be a necessary condition for this spirantization, and it may ascend to a sufficient
condition if dobba 'bear', in lieu of *do,Ba, bespeaks foreign influence (Cl. Arab. dubbun,
Iraqi dibb).
64 Empty cells in Table 2 reflect the limits of my success so far in locating pertinent
forms. Select examples of the various classes in addition to those presented in the chart,
sometimes prefaced by notes relevant to some facet of the class, here follow. Pre-Mand.
etyma are not provided (cf. in general the references cited in fn. 57), and forms are Classical
unless otherwise indicated. CLASS1: (Dl or 2) (puta) 'opening' [DM, s.v.]; (miMt) 'I
anointed' [DM, s.v. M?A II]; (D3) /m6hsi/ (mihsh) [mehsi] 88 'he measured it'; /?6hwa/
(sihua) [?ehwa] 88 'cry'; Cl. and Mod. /mehla/ (mihla) [mehla] 518 'salt'; Cl. and Mod.
/6hra/ (euhra) [ohraI 514, 518 'path, road'. CLASS2: (Dl) /n6bbi/ (nibia) [nibbi] 91 'he
3.2. I had formulated the overt and observable part of the hypothesis men-
tioned in ?3 long before reading Kiparsky's treatment of Greek. Specifically,
comparative and internal reconstructive evidence assured h-metathesis in the
Mandaic reflexes of such Table 2 forms as are marked for C and D3, i.e. *yarfid
and *timha, and other such forms exemplified in fn. 64. It was also clear that
pre-consonantal h (whether or not metathesized) could be lost, though the con-
ditions under which this compensatorily involved lengthening of the preceding
vowel vs. gemination of the following consonant were (and remain) unclear.
At any rate, this view rendered understandable the absence of h in the reflexes of
such Table 2 forms as *miSh6, *lahmd,*hhda, and *sihra. Also, it was easy to
relate to this development the reflexes of forms with pre-consonantal gutturals
other than h, with the same uncertainty as to the compensatory mechanism, but
viewing the loss as obligatory rather than optional (cf. Table 2, classes 4 and 8).
The situation with PosT-consonantal gutturals, however, seemed problematic
in at least one way: to posit loss with compensatory gemination of the IMME-
DIATELY PRECEDING CONSONANT made direct phonetic sense (e.g. *zar9a >
zerr&, class 2), a number of cases rather showed lengthening of THE NEXT PRE-
CEDING VOWEL, SEPARATED FROM THE VANISHED GUTTURAL BY THE 2ND RADICAL
CONSONANT: e.g., *tir9a > ttrd, *4in?a > stnd (classes 2, 6). Such developments
defied straightforward formulation of a strictly 'compensatory' mechanism.65
3.3. It was at this impasse that Kiparsky 1967 threw a new light on the matter-
Some of his suggestions are synoptized in Table 3. The changes labeled C'
wants'; Cl. /s6bba/ Mod./sog3/a/ (<uba) [sofS,a] 231 'seven' (cf. fn. 63 above); possibly
Mod. /biyya/ [biyya] 178 fn. 63, 501 'egg' < *biy9a (cf. fn. 34 above);/s6bban/ (suban)
[so,3,3an]91 'baptize me!' (cf. fn. 63 above); (D2) /pfqa/
(piqa) [fiqa] 174 'fissure' (for the
[f], cf. Macuch, 37);/simaO/ (simat) [simat] 92 'she heard' ([t] seems to have been gen-
eralized for verb affixes in Mod. Mand.; cf. fn. 50). CLASS3: For[dl in the trad. pron. of
(<uda), see Macuch (35). Cl. (lahma) 'bread' implies D3, hence/lthma/. But the alternate
spelling (lhama) (DM, s.v.) might indicate that the (h) is no more than etymological flotsam.
(lama) is apparently a Mod. spelling (Macuch, 101). The Mod. pronunciations of 'reptile'
with [res-], listed in Macuch (172 and fn. 45), seem anomalous. Other exx.: (Dl) /p6Aanni/
(ptanh)[patanni] 86 'we opened it'; (D2) /sabbaOi/ (sabath) [sabati] 86 'I praised! him'
(re [0] for[t], cf. [simat] above in this note; for
[bl, cf. Malone 1969:200); (D3)/pahhnin/
(ptahnin) [p9tahnin - eftahnin] 86 'we opened'; /qaddah0a/ (qadahta) [qadah0a] 86 'cry'.
CLASS4: (Dl) Cl. and Mod. /b6lla/ (bila) [bella] 33, 506 'lord';
/iw6tton/ (editun) [idetton]
135 'ye knew'; /samatton/ (smatun) [kamatton] 112 'ye heard'; /mmannin/ (smanin)
[samannin] 92 'we heard'; /mammet6/ (mamid) [mammed] 91 'baptizing'; (D2) /sari/ (saria)
[sari] 71 'hair
(pl.)'; /samana/ (samana) [samana] 112 'I hear'; (Dl and 2) Cl. /tamma/
(tama) [tammal 112 'taste', but Mod. perhaps /tamma aam// t [tama] 171, [tamma] 112
(the
[a] of the latter pronunciation may be explainable in terms of Malone 1967:276-7).
CLASS5: (D3) /tahmi/(tah mia) [talmi] 88 'turbid (masc.pl.)' (for [a], cf. Malone 1967:277).
CLASS7: (D3) /behrnan/ (bihrana) [bihrana] 81 'glitter'; /y4hvae/ (yahbat) [yathlat] 97
'she gave'. CLASS8: (kauba) (DM, s.v.), an alternate spelling to (kiba), must postdate
labialization by ?3.14. (Dl) /n6mmar/ (nimar) [nimmar] 91 'will say'; (D2) Cl. and Mod.
/kasa/ (kasa)
[kasa] 215, 494 'bowl'.
66 For the foregoing diachronic observations, cf. Noldeke (57-71), and Macuch (79-92).
For synchronic formulation of the morphophonemic repercussions, cf. Malone (1967:45-6,

187-92), especially the statement and discussion of rules P8.1 and P9.3. For an attempt to
understand forms such idraas in terms other than those proposed in this paper, cf. ibid.
(277).
-VCy- -VCh- -VhC-

Pre-Greek (presuppos- *krinyo 'I judge' *ekrinha 'I judged' *ehmi 'I am'
ing certain changes
from PIE, incl.
**8 > *h under
certain conditions)
(B) y > h under cer- krinho
tain conditions
(C) VCG> VGC,where krihno ekrihna
G = h, y, and C =
most resonants
(D1) VhC > VCC /krinno/ /ekrinna/ /6mmi/
(Lesbian)
(D2) VhC >VC (Attic) /krfno/ /6krina/ /em!/ (etl)
TABiLE3
D1, D2, and to a lesserdegreeB, arestrikinglysimilarto the developmentsalready

posited for Mandaic:for Greek,B posits the laryngealizationof the glide y to h;
for Mandaic, B (?3.12) posits the same laryngealizationof the glides 9 andA.
For Greek, C posits the metathesisof the glides h and y with a certain class of
immediatelyprecedingconsonants;forMandaic, metathesis of h is posited for
any immediatelyprecedingconsonant(?3.2).ForGreek,D involvestwo varieties
of assimilationof the glide h; for Mandaic, D includes just these two varieties
for the glides h and ?.
But the crucial differenceis that Kiparskyhad posited, in C, metathesis of
EITHERGLIDEh ORY, while the Mandaic metathesis had been viewed as LIMITED
TOh. Moreover,Kiparsky'smetathesis provided an illuminatingintermediary
step for such dubiously compensatorychanges as VCh > VC, WHICHWAS
EXACTLY THEMANDAICPROBLEM sketched at the end of ?3.2. This suggestive
parallelisminducedme to review the entire Mandaicsituation, with the results
as set forth in Table 2 and ?3.1,whereit is especiallyto be noted that C now pro-
vides for the metathesis of EITHERLARYNGEAL GLIDE, h OR 9. Now the vowel
lengtheningin such derivationsas *?inja > stna ceases to be problematicas a
compensatoryprocess; rather it is reduced,via change C, to a simple case of
total progressiveassimilation,quite on a par with*bijr? > bird. Moregenerally,
the effect of C on Table 2 is to combinenot only classes 1 with 3, and 5 with 7,
but likewise2 with 4, and 6 with 8.
3.4. A few summaryobservationson this type of Mandaicmetathesis:
3.41. While the choice between Greek Dl and D2 is dialectal (Kiparsky,
623), the factors deciding between Mandaic Dl and D2 are unknownto me.
Dialect mixturemight be involved to some degree,but significantpatterningis
likely to be obscuredby readingpronunciationin the trad. pron. of Cl. texts
(Macuch's'abagadical pronunciation',cf. fn. 34 above). More data on Mod.
Mand. might clarify matters (cf. ibid.) In the case of h, the factorsdecidingbe-
tween Dl and D2 on the one hand and D3 on the other are likewisenot clear,
but there is some evidencefor (1) free alternationwithin the same (verb) stem
(e.g. the formsfor'open' in fn. 64, cf. Class 3), and (2) a certaindegree of para-
digmatic leveling (e.g., in verbs D3 may be obligatory for lx/h and 2V/h; cf.
Malone 1967:190).
3.42. There is some evidence for other origins of the Mandaic alternation
VCC VC, in addition to the assimilations of D1 and D2. Moreover, these other
origins may have precedent both elsewhere in Aramaic and, possibly in part
co-territorially with Mandaic, in Akkadian. It cannot be excluded that at least
some cases attributed to the hypothesis of this paper are in reality to be traced
to these other origins. On the other hand, the hypothesis of this paper may hold
also for Akkadian, at least in part.66
3.43. The Greek metathesis effects the permutation of y in addition to h, e.g.
*banyo > bdyno 'I go'. There is striking if residual evidence that Mandaic, PER-
HAPSALONEOF DOCUMENTED ARAMAICLANGUAGES, underwent a similar change in
forms such as *qanya > Cl. /qayna/, Mod. /q6yna/ 'reed'.67 My attempt to
66 These other origins may involve either or both of the following factors: (1) A general
variation (phonological equivalence) may exist between VCC and PC, more clearly ob-
taining in Akkadian (Reiner, 45-6) and hence perhaps betokening an ancient Mesopotamian
Sprachbund affecting any number of co-territorial languages (Malone 1969:199, fn. 18,
where Mandaic examples not explainable in terms of the present paper's Dl 1 D2 are also
adduced. On the other hand, ALL the examples there cited may reflect 'abagadical' pronun-
ciation as mentioned above). (2) There may be metathesis of length (VC: > V:C, or con-
versely V:C > VC:; i.e. effectively VCC > VC or conversely), under opaque conditions,
but perhaps as a manifestation of the same phenomenon seen in the structure of the Com-
mon Aramaic geminate verb; e.g. *tadiqq(u) > BA taddiq 'she'll crush', contrast Arab.
tudiqqu, Heb. tobeq. (Cf. BLA, 57. Such processes are likely to have continued their mor-
phophonemicviability in Cl. Mand.; cf. Malone 1967,rule P2.4, pp. 41,163-4). In the original
talk underlying this paper (fn. 1), I had proposed that some Akkadian VCC VC might
likewise be understood in terms of B, C, D; e.g., pre-Akk. *zer9u 'seed' (B) > zer?u (C) >
ze?ru (D1l2) > /zerru r zeru/, *milvu 'flood' (C) > mi?lu (D1.2) /millu ' milu/. But
Moshe Held pointed out that forms such as /millu/ occur chronologically LATER than ones
such as /milu/, suggesting that /millu/ < /milu/, not directly < /mi?lu/. Held's observa-
tion is corroborated by Erica Reiner who writes (personal letter, 12 May 1969): 'As far as
we can judge, the development is from VHC to VC, and only in later periods are VCC and
VC in variation, as Professor Held has pointed out.' But though the evidence is thus against
assuming for Akkadian a direct interrelation of changes like Dl and D2, we may still fruit-
fully hypothesize B, C, and D2-a sequence of changes which, it will be noted, explains the
compensatory anomaly of direct VCH > VC here no less than in the Mandaic and Greek
cases examined above; because of the metathesis, the development of forms like /zeru/
and /milu/ converges at the crucial point with such non-problematic derivations as /belu/ <
*be9lu 'lord' and /kasu/ < *ka?su 'bowl'. It may finally be mentioned that the Akk. meta-
thesis, if appropriately formulated, seems to permute ANY glide (spec. y and w in addition
to ?) with an immediately preceding consonant, unless Akk. B merged post-consonantal
*y, *w > ? before the action of C. Examples (see von Soden, 16): *kuryu > /kuru/ 'short-
ness', *minwu > /minu/ 'number'.
67 For general discussion and examples, see Noldeke (24) and Macuch (122-3). All the
examples (Noldeke adduced seven, Macuch added three or so more) involve metathesis of
3VA/with an immediately preceding nasal (m, n) itself immediately preceded by a long or
short a, i.e. aNy > ayN. Kiparsky's Greek metathesis (622-3) is more general, involving
the metathesis of h or y with any immediately preceding resonant except y (i.e. m n r I w)
itself immediately preceded by any vowel. If some such wider conditions obtained also for
Mandaic, the following two cases may also reflect the metathesis, though neither case is
less than shaky: *zary?a> /zayra/ as a by-form for /zerra/ (Table 2) 'seed' (see DM, s.vv.
zaira and ZRA I for the root alternation /zr9 - Vzry); *sirya > siyra = /sfra/ 'refuse'
relate this relic development to the metathesis of C, given the paucity of avail-
able evidence, presents a somewhat questionable reduction.68
3.44. In sum, if my transposition of Kiparsky's Greek findings to Mandaic is
correct, the homology between the two historically independent developmental
nexuses almost certainly speaks for the validity of typological universals of
linguistic change. The precise nature of the processes involved remains to be
studied.69
(see DMI,s.v. sira 3, and cf. JP (seriy) inJastrow, s.v.) In any event, traces of Mand. y-meta-
thesis had by the time of C1. Mand. been all but obliterated, putatively by analogical
leveling; and for each clear token of metathesis, there are any nllmber of non-metathetic
forms of the same morphological structure.
68Specifically, y-metathesis is not transparently reducible to a special case of H-meta-
thesis (C), nor conversely. In keeping as closely as possible to Kiparsky's use of distinctive
features, Mand. C may be formulated as follows:
1-[-voc}] [_voc
c-con
[-coni +conJ - difj
1 2 3 == 132
Condition: 3 = 3rd radical.
Mand. y-metathesis (if interpreted as VNy > VyN) may be formulated as follows:
_- -voc
-voc _i
r il>I+con -con
[-con] -rounded
L-obsJ +dif
1 2 3 = 132
Condition: 3 = 3rd radical.
Amalgamating these, we may get:
n-voc[
-~ ~~' . - /
vo
-con
co
1
(-con]
Uco
Sobs ] adif
1 2 3 = 132
Conditions: (a) if a = plus, then ,8 = minus, regarding specifications of [-obs] and
removal of braces { J; viz., relation between [-voc] and [+con]
must be CONJUNCTIVE to specify N.
(b) 3 = 3rd radical.
The question perhaps boils down to the legitimacy of condition (a), and it may be that
the two metatheses were historically independent processes. It is interesting to note,
finally, that pre-Kiparskyan attempts to explain the Greek y-metathesis (references cited
in Kiparsky, 622, fn. 5) are very much like Brockelmann's early attempt to explain Mandaic
y-metathesis (1961:279).
69 Reference was made in fn. 55 to Irish patterning somewhat reminiscent of Semitic tS
metathesis. Interestingly enough, the Western Munster dialect of Irish shows hR > Rh
(where R- = resonant), i.e. the opposite direction from the otherwise so similar Greek and
Mandaic metatheses: e.g., [sla:h9r] 'to provide'; but when the s is syncopated under suffixa-
tion, [sla:rhi :r1] 'I provide'. Cf. also [Aefhd:r] 'ten people' vs. Connaught dialect
[dehfukr]. In these and all other cases, it is perfectly clear that hR is the historical ante-
cedent; e.g. the latter example < Old Irish dechenbur,cf. Lat. decemvir.Korean seems to
evince a similar but even more general metathesis, hC > Ch (Koutsoudas 1966:64). On the
other hand, Kiparsky (623, fn. 5) cites Finno-Ugric evidence for Rh > hR metathesis, i.e.
in the Greek-Mandaic direction.
414 LANGUAGE,VOLUME47, NUMBER2 (1971)
REFERENCES
ANDERSEN,HENNING.1969. Lenitionin CommonSlavic. Lg. 45.553-74.
BAUER, HANS, and PONTUSLEANDER.1962a. Grammatik des Biblisch-Aramaischen.
Hildesheim.
----. 1962b.HistorischeGrammatikderhebraischenSprachedes Alten Testamentes.
Hildesheim.
BERGSTRASSER,G. 1962.HebraischeGrammatik,I-II. Hildesheim.
BROCKELMANN, CARL.1928. Lexicon Syriacum. 2nd ed. Halle.
--. 1961. Grundriss der vergleichenden Grammatik der semitischen Sprachen, I.
Hildesheim.
--. 1965. Syrische Grammatik. 10th ed. Leipzig.
CANTINEAU, J. 1930. Le Nabateen, I. Paris.
CHAFE,WALLACE L. 1968. The ordering of phonological rules. IJAL 34.115-36.
CHOMSKY, NOAM,and MORRISHALLE.1968. The sound pattern of English. New York.
CROSS,FRANKMOORE,JR., and DAVIDNOEL FREEDMAN. 1952. Early Hebrew orthog-
raphy: a study of the epigraphic evidence. New Haven.
DALMAN,GUSTAF.1960. Grammatik des jiidisch-palaistinischen Aramaisch. Darmstadt.
DINNEEN,PATRICK S. 1927. Focl6ir Gaedhilge agus B6arla [an Irish-English dictionary].
Dublin.
DROWER,E. S., and R. MACUCH.1963. A Mandaic dictionary. Oxford.
EVEN-SHOSHAN, ABRAHAM. 1964. Millon hadas manuqqad u-mosuyyar. 5 vols. Jerusalem,
5724.
GARBINI,GIOVANNI.1957. L'aramaico antico. Atti dell'Accademia Nazionale dei Lincei,
Memorie, serie 8, 7.239-85.
. 1960. Il semitico di nord-ovest. Napoli.
GORDON,CYRUSH. 1938. The Aramaic incantation in cuneiform. Archiv fur Orient-
forschung 12.105-17.
--. 1940. The cuneiform Aramaic incantation. Orientalia 9.29-38.
GRAY,LoUIs H. 1936. Observations on the phonology of the BaPrAAKa4A. American
Journal of Semitic Languages and Literatures 52.171-7.
GREENBERG, JOSEPHH. 1950. The patterning of root morphemes in Semitic. Word
6.162-81.
HARRIS, ZELLIG S. 1939. Development of the Canaanite dialects. New Haven.
JASTROW, MARCUS.1950. Dictionary of Talmud Babli, Yerushalmi, Midrashic literature,
and Targumim, I and II. New York.
KIPARSKY, PAUL. 1967. Sonorant clusters in Greek. Lg. 43.619-35.
KOUTSOUDAS, ANDREAS.1966. Writing transformational grammars. New York.
KUTSCHER, E. Y.; J. NAVEH;and S. SHAKED.1970. Haktovot ha'Aramiyyot sel Aboka
[The Aramaic inscriptions of AMoka].LeFonenu 24.125-36.
LEANDER, PONTUS. 1966. Laut- und Formenlehre des Agyptisch-Aramaischen.
Hildesheim.
MACUCH,RUDoLF.1965. Handbook of Classical and Modern Mandaic. Berlin.
MALONE,JOSEPHL. 1967. A morphologic grammar of the Classical Mandaic verb.
Berkeley: University of California doctoral dissertation (available from University
Microfilms, Ann Arbor).
. 1969. Review of Handbook of Classical and Modern Mandaic, by R. Macuch.
Lg. 45.191-203, 951.
. 1971. Wave theory, rule ordering, and Hebrew-Aramaic segolation. JOAS 91.
44-66.
. MS.Juncture in the Judean Aramaic verb. Paper read to the 180th annual meeting
of the AOS, Baltimore, April 1970.
MARGOLIS, MAX L. 1910. Lehrbuch der Aramaischen Sprache des Babylonischen Tal-
muds. Munchen.
MORAG, SHELOMO. 1962. The vocalization systems of Arabic, Hebrew, and Aramaic.
The Hague.
SYSTEMATICMETATHESISIN MANDAIC 415
MOSCAT, S.; A. SPITALER;

E. ULLENDORP;
and W. VONSODEN.1964. An introduction to
the comparative grammar of the Semitic languages. Ed. by Sabatino Moscati.
Wiesbaden.
THEODOR.
NOLDEKE, 1964. Mandiiische Grammatik. Darmstadt.
ERICA.1966.A linguisticanalysisof Akkadian.The Hague.
REINER,
SCHULTHESS,F. 1903. Lexicon Syropalaestinum. Berlin.
--. 1965. Grammatikdes christlich-palatinischenAramaisch.Hildesheim.
J. PAYNE.1903.A compendiousSyriacdictionary.Oxford.
SMITH,
SPITALER, ANTON. 1966. Grammatik des neuaramaischen Dialekts von Ma'lula (Anti-
libanon).Nendeln,Liechtenstein.
VONSODEN,WOIFBRM.1952. Grundriss der akkadischen Grammatik. Rome.
WANG,WLIAM S-Y. 1969.Competingchangesas a cause of residue.Lg. 45.9-25.
BURAVLEV, V. 1967. Pererazloienie differencial'nixprisnakov kak osnovoj istocnik
fonologiceskixismenenij.Linguisticstudies prented to Andr6Martinet, I, ed. by
AlphonseJuilland(Word23), 588-91.
[Received16 May 1970]

Malone 1971

Uploaded by

Copyright:

Available Formats

Malone 1971

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Malone 1971

Uploaded by

Copyright:

Available Formats

Linguistic Society of America

Systematic Metathesis in Mandaic

immediatelyfollowingsibilant,is ratherwidespreadin Semitic;but only Mandaic

2.1. Reflexive prefixal t- metathesizes with an immediately following sibilant,

elmann 1928, s.v. (metso?).

15 Garbini (1960:52-3); Harris (40-1, 62-3).

2.26. Immediately preceding a guttural glide (at least hfi 9) or r as 3rd

resolution in the simple reflexive seems to be absolute in EA and BT, widespread

?ittapqad as reflexive to ?apqid. But if Common Aramaic ?ittapqad is explained in such

2.33. Mandaic ?9 is SYNCOPATED interconsonantally, a fate which is shared at

PAram. ?itsamira ?it7asira

Chafe expects persistent morphophonemic rules normally to apply at depth

daic underwenttS metathesistwice, impliesrecognitionof such metathesisas persistent

dersen (1969:558) provides a promising path toward general understanding of at least

*suhda 'bribe' suhdd (2) iudd /sua/ (suda) [isuda]84

(C): A laryngeal glide , h metathesizes with an immediately preceding con-

as Mod. Mand. is highly suspect. Rather, it is likely to be a spelling-pronounced Cl. word

For synchronic formulation of the morphophonemic repercussions, cf. Malone (1967:45-6,

-VCy- -VCh- -VhC-

D1, D2, and to a lesserdegreeB, arestrikinglysimilarto the developmentsalready

MOSCAT, S.; A. SPITALER;

You might also like