• Keine Ergebnisse gefunden

Prosodic words

3.4.6. Clause boundaries

In one important sequence in KAI 24 univerbation occurs at a clause boundary:

(130) KAI5 24.14–15

𐤓𐤓𐤏𐤁𐤅·𐤌𐤓𐤓𐤏𐤁𐤋·𐤃𐤁𐤊𐤉𐤋𐤀·𐤌𐤁𐤊𐤔𐤌 ⟵

𐤓𐤁𐤂𐤋·𐤔𐤀·𐤃𐤌𐤑·𐤋𐤏𐤁·𐤔𐤀𐤓·𐤕𐤇𐤔𐤉·𐤆𐤓𐤐𐤔𐤄·𐤕𐤇𐤔𐤉·𐤉𐤌𐤅𐤌𐤁𐤊𐤔𐤌𐤋·𐤃𐤁𐤊𐤉𐤋𐤀·𐤌 mškbm 〈ω〉 ʾl=ykbd 〈ω〉 l=bʿrrym 〈ω〉 wbʿrr 〈λ〉m 〈ω〉

PN not=honour to=PN and=PN

ʾl=ykbd 〈ω〉 l=mškbmwmy 〈ω〉 yšḥt 〈ω〉 h-spr=z 〈ω〉

not=honour to=PN=and=who destroy the-inscription=this yšḥt 〈ω〉 rʾš 〈ω〉 bʿl 〈ω〉 ṣmd 〈ω〉 ʾš 〈ω〉

destroy head DN DN who

l=gbr 〈λ〉

to=PN

‘Let the Mškbm not honour the Bʿrrm, and the Bʿrrm not honour the Mškbm. And whoever destroys this inscription, let Bʿl Ṣmd who is Gbr’s destroy his head.’ (trans.

after Collins 1971, 187)

Collins (1971, 186) versifies this as follows:

mškbm · ʾl ykbd · l bʿrrm w bʿrrm · ʾl ykbd · l mškbm w my · yšḥt · h spr z · yšḥt · rʾš · bʿl · ṣmd · ʾš · l gbr

However, the lack of word divider after l=mškbm perhaps points to the following versification:

mškbm · ʾl ykbd · l bʿrrm

w bʿrrm · ʾl ykbd · l mškbm w my · yšḥt · h spr z ·

yšḥt · rʾš · bʿl · ṣmd · ʾš · l gbr

The poetic effect would be clear: by univerbating l=mškbm with w=my, and thereby keeping w=my in the previous colon with l=mškbm, the writer is able to begin both elements of the next bicolon with yšḥt.

3.4.7. Implications for the target of word division in Phoenician

Although there appears to be some consistency in the treatment of nps in apposition and Noun + Modifier/Determiner phrases, the fact that there is inconsistency in a number of syntagm types points to a lack of isomorphy between graphematic word division and morphosyntactic structure. This is consistent with what we saw in the case of prefix particles (§3.3), where multiconsonantal morphemes are optionally univerbated with the following morpheme.

Morphosyntactic inconsistency is of course negative evidence. This is to say that it constitutes evidence as to what word division does not do, viz. identify morphosyntactic units. It does tell us what word division actually does do. Although we have provided some positive evidence on typological grounds for word division corresponding to actual prosodic words (§3.3.4), it is also desirable to have direct evidence for this. For this we look to two further pieces of evidence: sandhi assimilation (§3.5) and morphosyntactic comparison with maqqef phrases in Tiberian Hebrew (§3.6).

3�5� Sandhi assimilation

Sandhi assimilation is associated with construct phrases headed by bn ‘son (of)’ (§3.4.2) in Phoenician. When bn occurs immediately before 〈ʾ〉, it is written with final 〈n〉: (131) 𐤌𐤓𐤇𐤀𐤍𐤁 bn=ʾḥrm ‘son of ʾḥrm’ (KAI5 1.1)

(132) 𐤋𐤏𐤁𐤋𐤀𐤍𐤁 bn=ʾlbʿl ‘son (of)’ (KAI5 7.2)

By contrast, when written before 〈y〉 or 〈k〉, the 〈-n〉 is not written:

(133) 𐤊𐤋𐤌]𐤇𐤉𐤁 b=yḥ[mlk ‘son of Yeḥawmilk’(KAI5 6.1) (134) 𐤉𐤁𐤋𐤊𐤁 b=klby ‘son of Klby’ (KAI5 8)

I take these latter examples to be representations of sandhi assimilation of /n/.

This is part of a more general phenomenon of the representation of sandhi assimilation in early Northwest Semitic orthography (Steiner 2016, 321–326).5 Outside of Old Byblian, Phoenician examples are restricted to inscriptions from Cyprus, where sandhi assimilation occurs in Preposition + Noun sequences, as well as in construct phrases.

Of sandhi assimilation in Preposition + Noun sequences the following is an example (cf. Friedrich, Röllig & Amadasi Guzzo 1999, §251):

(135) KAI5 33.2

𐤔𐤀𐤉·𐤕𐤔𐤇𐤍𐤌·𐤀𐤍𐤈𐤉𐤅·𐤍𐤕𐤉·𐤔𐤀·𐤆𐤀·𐤕𐤋𐤌[𐤎] ⟵ [s]mlt 〈ω〉 ʾẓ 〈ω〉 ʾš 〈ω〉 ytn 〈ω〉 w=yṭnʾ 〈ω〉

image this which gave and=set_up

m=nḥšt 〈ω〉 yʾš 〈ω〉

from=bronze PN

‘This image which the Yʾš gave and set up out of bronze’ (trans. with ref. to Donner

& Röllig 1968, 51)

5 Cf. the related phenomenon of consonant coalescence (Steiner 2016, 313–321), e.g. 𐤉𐤕𐤊𐤋𐤌mlkty is equivalent to mlk=kty ‘king of Kition’ (KAI 33.2) (cited in Friedrich, Röllig & Amadasi Guzzo 1999, 56, §99a).

This, however, is not attested in the Old Byblian corpus or in KAI 24, and is therefore not treated here.

In the next example, sandhi assimilation of /-m/ occurs in construct chain:

(136) 𐤔𐤉𐤌𐤋𐤕𐤐𐤌𐤊𐤋𐤌𐤃𐤀𐤋𐤅 w=l=ʾdmlkm=ptlmys ‘and to the Lord of Kings Ptolemy’

(KAI5 42.2)

In this example ʾdmlkm is for ʾdn=mlkm [lord=kings] (Steiner 2016, 317, citing Harris 1936, 30).

In Tiberian Hebrew sandhi assimilation is much more restricted than sandhi spirantisation, which operates at the level of the prosodic phrase (§1.4.2.2). Here sandhi assimilation is limited to the boundary between the preposition ן ִמ min ‘from’

and the following morpheme (Steiner 2016, 323) (see also §11.1.4); it does not occur at the boundary of ןב bn ‘son (of)’ and the following morpheme. Nevertheless, insofar as ן ִמ min when assimilated forms a single prosodic word with the following morpheme, sandhi assimilation there too is restricted to the level of the prosodic word.

In both Phoenician and Tiberian Hebrew, therefore, sandhi assimilation appears to operate at the level of the prosodic word.6 Accordingly, the fact that in KAI 6 and 7 we see sandhi assimilation within a univerbated construct phrase is consistent with word division targeting prosodic words.

Corroborating evidence for this claim would come from 〈n〉 occurring at a boundary where we would not expect to find internal sandhi phenomena. Ideally this would be in the same inscription where we also find assimilation, i.e. in KAI 6 or 7, but such a sequence is not attested there. We do, however, have such a sequence in KAI 1:

(137) KAI5 1:1

𐤋𐤏𐤁𐤕[𐤀]·𐤋𐤏𐤐𐤆·𐤍𐤓𐤀 ⟵

ʾrn 〈ω〉 z=pʿl 〈ω〉 [ʾ]tbʿl 〈ω〉

sarcophagus which=made PN

‘Sarcophagus that ʾtbʿl made’ (trans. with ref. to Donner & Röllig 1968, 2)

It remains to account for the alternation between the assimilated and non-assimilated variants given above. As Friedrich, Röllig & Amadasi Guzzo (1999, 56) note, the assimilation occurs ‘außer wenn der folgende Name mit einem Laryngal beginnt’.

This is the same distribution that we saw with ן ִמ min in Tiberian Hebrew (§11.1.4).

The distribution of the morpheme 𐤍𐤁 bn ~ 𐤁 b therefore parallels that of Tiberian

6 Steiner (2016, 317–318) classes these cases as instances of ‘external sandhi’, reserving the term

‘internal sandhi’ for assimilation/coalescence occurring at the ‘morpheme boundary between stem and affix’ (Steiner 2016, 321). For Steiner, then, internal sandhi is sandhi at a morpheme boundary within morphosyntactic words, including between stem and affix, while external sandhi is sandhi at a morpheme boundary between morphosyntactic words. Under the definitions of Zwicky (1985), therefore, it can be argued that sandhi assimilation is a case of ‘internal sandhi’, while spirantisation is both an internal and an external sandhi process.

Hebrew ן ִמ mỉn ~ - ִמ mi- in Tiberian Hebrew (§11.1.4). The lack of assimilation in these contexts could be explained by the weakening of /h/ and /ʾ/ to zero, for which there is evidence even in the inscriptions from Serābît el-khâdim (Steiner 2016, 326–328).

3�6� Comparison of composition and distribution with prosodic words in Tiberian Hebrew

3.6.1. Distribution

External evidence for the semantics of word division in Phoenician comes from distributional comparison with Tiberian Hebrew. In order to assess the suggestion that graphematic words target prosodic words, the distribution of the morphemes making up graphematic words in the inscriptions are directly compared with the distribution of maqqef phrases in Tiberian Hebrew (cf. §1.4.2.6 above). The results of this comparison are given in Table 3.2.7

Table 3.2: Comparison of Tiberian Hebrew morphemes joined by maqqef vs. separated by spaces (Gen 14:1–3; 2Kgs 1:3, 8:16–18) with Phoenician morphemes either univerbated vs. separated by word dividers (KAI 1, 4, 7, 24)

Phoenician Tiberian Hebrew

Freq. % Freq. %

Separated 55 78.57 27 72.97

Univerbated 15 21.43 10 27.03

Total 70 37

The table indicates that the degree to which possibly univerbated morpheme sequences in the Phoenician inscriptions are in fact univerbated parallels the degree to which morpheme sequences that have the potential to be joined by maqqef in Tiberian Hebrew are in fact joined by maqqef. This parallel distribution speaks in favour of the two units corresponding. This is to say that the distributional evidence supports the view that graphematic words in Phoenician correspond to prosodic words in Tiberian Hebrew.