Some Observations on the Construct Marker -a in Classical Ethiopic

This contribution offers a discussion of the Classical Ethiopic use of the marker -a to indicate both the direct object and the construct state of nouns. Following a critical evaluation of the explanation of the phenomenon given by Josef Tropper, a few observations are made here which seem to lend support to the idea that the use of the marker in the construct state is in fact a generalization of the accusative ending.


Introduction
A wellknown phenomenon in Classical Ethiopic (Gǝʿǝz) is the use of the ending a as a marker of both the accusative case and the head noun of construct noun phrases, irrespective of the role of the head noun in the sentence. 1 For instance, in baqadāmi gabra ʾǝgziʾabǝḥer samāya wamǝdra, 'in the beginning God created the sky and the earth' (Gen. 1:1), the a on the two last words marks them as direct objects. This a reflects the Pro-toSemitic accusative singular *a, while the ProtoSemitic endings in the nominative and genitive (*u and *i) were reduced to ǝ and subsequently lost in Ethiopic, producing an opposition between accusative a and nonaccusative Ø. In the construct state, the distinction between accusative and nonaccusative is neutralized. In kama yǝʿqabu fǝnota ʿǝḍa ḥǝywat, 'that they should guard the way to the tree of life ' (Gen. 3:24), there are two nouns in the construct state (fǝnota and ʿǝḍa). The first one is the head of the whole noun phrase and thus the object of yǝʿqabu. In this case, then, the noun would also be marked by a if it was not in the construct state. However, in this sentence this is not the reason for the a. The second word shows that nouns that are not the direct object of a verb have the same ending when they are in the construct state. Compare the following instances: wamanfasa ʾǝgziʾabǝḥer yǝṣellǝl malʿǝlta māy, 'and the spirit of God hovered over the water' (Gen. 1:2), where the noun phrase is the subject, and walak w ǝllu ʾaʿwāfa samāy, 'and to all the birds of the sky' (Gen. 1:30), where the noun phrase is part of a prepositional phrase. Of course, the term for God ('lord of the land/world') is also a construct chain with the first part in a, regardless of its role in the sentence.
This state of affairs has been explained as a case of analogical spread of the accusative ending, for instance by Moscati: 'The ending a of the construct state has possibly arisen from an analogical extension of the accusative morpheme'. 2 However, Tropper challenges this view, arguing that several phenomena in various Semitic languages might indicate that early Semitic did not have nominativeaccusative syntax, but rather an ergativeabsolutive syntactical system, as has also been argued by other scholars. According to this hypothesis, the later accusative form in a was originally an absolutive form that covered anything that was not the subject of a transitive verb, for which the ergative form in u was used (which later became the nominative). The reason for this view is the 'grosse funktionale Bandbreite des aKasus und insbesondere die eigentümliche Tatsache, dass der aKasus auch Nominativfunktion besitzt' 3 (such nonaccusative usage of the acase is interpreted as 'Relikt einer älteren AbsolutivSyntax'). 4 Against this background, Tropper sees the use of a as a construct marker as another relic of the absolutive case from the earliest stage of the language group and sketches two alternative explanations. 5 In the first alternative, the background for the Ethiopic situation was a system where the construct forms had a diptotic inflection (a for the nominative/accusative, or rather absolutive, and i for the genitive). In the second one, there never was any inflection in the construct, the a form covering all functions, including the genitive. In the first explanation, the process leading to the attested Ethiopic usage would include the replacement of the genitive i by a, a process which was facilitated by the reduction undergone by i ( > ǝ). While Tropper tends to favour the second explanation, the crucial point is the same for both approaches: there never was an ending u in the construct state.
While at least some of the phenomena described by Tropper might be relics of an earlier syntactic system where the a ending was used to mark a wider range of functions than expected of an accusative form, it is questionable whether this can serve as an explanation of the a in the Ethiopic construct state. Several aspects of Tropper's explanation are problematic. The intention of this short contribution is to provide a few critical comments on Tropper's approach, thereby lending support to the traditional view of the process as a secondary extension of the accusative ending.

A Critical Evaluation
The first point to note is the curious statement made by Tropper when introducing the phenomenon, according to which '[d]as gleiche Prinzip der Bildung des St.cs. [Status constructus]' 6 is found in Amharic. The same claim is made by Lipiński. 7 Such statements are misleading, since this is clearly not the normal procedure in Amharic, where the order of the elements is the opposite ('the possessor' comes first) and there is no ending, but rather (most often) a prefixed ya on the first element ('the possessor'). 8 The Amharic examples of the classical type are best described as loans from the classical language. 9 In fact, no modern Ethiopic language uses the classical type productively. 10 Following Voigt,11 Tropper criticizes the account of the phenomenon given by Gragg,12 who, in his view, seeks to explain the formal identity of the construct and the accusative through the similar periphrastic constructions used. As is well known, the construct state Xa Y can be replaced by Xpossessive suffix laY, while a verb plus direct object X Ya can be replaced by Xobject suffix laY. Voigt and Tropper point out that the substitution does not occur at the same place, and therefore Tropper states that Gragg's 'Konzeption' is only a 'Scheinlösung'. 13 However, nowhere does Gragg claim that his statements are intended as an explanation of the phenomenon. What he gives is merely a description of the similarities: 'In both constructions, morphological indication of case can be replaced by syntactic 6 Tropper 2000, 201. 7 Lipiński 1997 Cf. Hartmann 1980, 293-315. Curiously, Tropper's reference is Hartmann's book, p. 312. However, on that page Hartmann explicitly states that this type 'gehört nicht zu den produktiven Formen des Amharischen'. 9 Cf. Weninger 2001Weninger , 1765 For a general overview, see the description of the typological differences between Classical Ethiopic and the modern languages in Weninger 2001, specifically concerning the word order of sentences as well as the order of elements in noun phrases. On Tǝgrǝñña, see Leslau 1941, 40-41, who notes that the classical construction is 'conservé uniquement dans les termes religieux et dans quelques noms composés'. 11 Voigt 1999, 218. 12 Gragg 1997, 248-249. 13 Tropper 2000 paraphrase'. 14 Tropper also criticizes Lipiński's approach, claiming that Lipiński holds the view that the construct marker is an independent phenomenon not related to the case marker. 15 However, Lipiński explicitly says (and Tropper quotes this in his note 10) 16 that '[t]his ending [the construct marker] derives from the caseform of the nonactive component of the sentence '. 17 In fact, Lipiński's general perception of the early Semitic case system seems to be rather similar to Tropper's.
The idea that ProtoSemitic had absolutiveergative syntax has been questioned by several scholars. Waltisberg convincingly shows that the arguments adduced are not persuasive and that ProtoSemitic 'wie die belegten Sprachen akkusativisch strukturiert war'. 18 All the phenomena adduced by various scholars in support of the view that the a ending was really an absolutive can in fact be considered 'sekundäre Funktionen des /a/Kasus, die mit der adverbialen Grundfunktion des semitischen Akkusativs konform gehen'. 19 Similarly, Hasselbach rejects the ergative hypothesis and suggests that an early phase of ProtoSemitic ('Archaic ProtoSemitic') was a markednominative language, which is a subtype of nominativeaccusative languages. Instead of using the nominative as the unmarked citation and basic form, for naming and so on, a markednominative language uses the accusative as 'citation form, for naming, outside grammatical constructions (such as the VOC [vocative], numbers, extraposition), and for nominal PREDs [predicates]'. 20 The genitive is considered a secondary development derived from the nisbe ending. 21 Thus, the relics noted by Tropper and other scholars can be seen as relics of this type of syntax, rather than as relics of an ergative system. 22 14 Gragg 1997, 249. 15 Tropper 2000, 206. 16 Tropper 2000, 204, n. 10. 17 Lipiński 1997, 266. 18 Waltisberg 2002, 56. 19 Waltisberg 2002, 34. 20 Hasselbach 2013, 327. 21 Hasselbach 2013 Note for instance that modern (mostly colloquial) Danish uses oblique pronouns in many contexts similar to the 'unexpected' uses of the Semitic accusative, namely, as citation form or in answers to questions, as predicate, and as subject when the pronoun is coordinated with another subject or is followed by a relative sentence or determining adjective (some of these usages are also found in English). The point is that such usage can develop without any implication that it evolved from an earlier ergative system.
Hasselbach also suggests that there might be relics from an even earlier stage of the language with no case endings at all. At this stage, words could end in a 'without a being a case marker'. 23 According to Hasselbach, this system evolved into the markednominative system and further into the classical triptotic, nominativeaccusative system already at the Pro-toSemitic stage. Thus, ProtoSemitic proper (as opposed to the two stages of Archaic ProtoSemitic) exhibited the type of case syntax attested in the languages that preserve case marking. However, the earlier system is reflected in 'vestiges in early PNs [personal names] and loanwords and in certain constructions in which the ACC [accusative] still reflects its prior function as citation form'. 24 While the attempt to reconstruct such stages prior to ProtoSemitic is highly speculative, Hasselbach's conclusion regarding the construct state is of importance in the present context. She argues that the presence of relics of case marking in the construct in the oldest stages of Akkadian 'in addition to the diptotic and partially triptotic declension before pronominal suffixes, indicates that Akkadian originally had a fully productive case inflection on singular nouns in the CSTR [construct state] that was characterized by the same vocalic case markers as those attested on unbound nouns'. 25 Evidence from several other languages also suggests that ProtoSemitic had case marking on nouns in the construct state-directly attested in Ugaritic and Arabic, and more indirectly in the 'linking vowels' before pronoun suffixes in other languages where the case system was otherwise abandoned. There are a few statements in Tropper's article that seem to point in the same direction-and consequently away from his general view. Thus, he mentions the traces of triptotic inflection in 'sehr frühen Sprachstufen des Akkad.' 26 and he notes that the determinative/relative pronoun in old Akkadian sometimes '(noch) triptotisch flektiert wird'. 27 Concerning 'case endings' on prepositions, '[e]s kann kein Zweifel darüber bestehen, dass die Kasusvielfalt des Akkad. einen typologisch älteren Befund widerspiegelt als die im Äth. bezeugte Kasushomogenität'. 28 Similarly, some of his examples of personal names actually have u on their first part, namely in the construct state. 29 The most decisive evidence seems to come from Ethiopic itself. We need to decide whether the use of a on construct nouns is a relic of an earlier phase in the development of Semitic case syntax or a secondary development in a system that used to have a triptotic inflection. The fact that Ethiopic does have case inflection before pronominal suffixes seems to point in the latter direction (nominative/genitive ǝ, accusative a). As Tropper notes, the pronominal state is just a special instance of the construct state, 30 and it might be argued that the protected environment before the suffixes is precisely the place to find relics of an earlier more widespread system. A small group of words ('father' and others) might support this view. Their nonaccusative form before suffixes is u, that is the old nominative in a long form characteristic of these words in several languages (probably caused by contraction with an original root consonant, as Tropper notes). 31 The accusative is ā, but in fact the nominative form is often used where the accusative would be expected. 32 The most likely explanation of the presence of u in these words seems to be that they are relics of an older stage where construct nouns were regularly marked for case. According to Tropper, in both of his two alternative scenarios referred to above, there never was an u ending in the construct state. However, this claim is not really explained. Even if the ergative hypothesis of ProtoSemitic is accepted, the lack of u forms is still not clear. Construct phrases whose head nouns are the subject of a transitive verb should still have u marking. Within the ergative approach, a possible explanation is provided by Lipiński, who claims that there is a connection between construct state and words with a nonactive role in a sentence, which is why a is used: 'The noun determined by another noun can be regarded as a kind of recipient and be considered, therefore, as a nonactive component of the phrase'. 33 While such a relationship between the parts of a construct phrase is surely not present in all instances, it does make sense in some cases (e.g. the house of the king ~ the king owns/controls the house), and from these it could have been extended to all construct phrases. However, if this explanation applies at a hypothetic ergative stage of ProtoSemitic it is doubtful whether it can explain the Ethiopic situation. On the contrary, as has been argued, it seems that the use of a as construct marker is in fact a secondary development 30 Tropper 2000, 202. 31 Tropper 2000 See Dillmann 1907, 357. 33 Lipiński 1997. In Tropper's quotation of Lipiński (Tropper 2000, 204) this part is left out.
from a system with full case marking in the construct state. However, Lipiński's explanation might apply in a nominativeaccusative system. It could be argued that there is a common semantic element shared by a head noun determined by another noun, on the one hand, and an accusative object determined by a verb, on the other hand. However, such speculation seems rather farfetched, and, in any case, another explanation is available, which is much more straightforward. 34

Conclusion
How could the development have taken place? Tropper sees no reason for the generalization of the accusative ending rather than a zero ending as in Akkadian. 35 However, as also noted by Tropper in his first alternative referred to above, the accusative ending was the only one left when both u and i were reduced to ǝ and subsequently lost. Thus, since the genitive ending on the noun following the construct noun was also lost (in contrast to the Akkadian situation, for which reason Tropper's comparison with this language does not carry much weight), the only way of marking a construct phrase was by using the a, simply because this was the only surviving ending. 36 Apparently, some kind of marking of these phrases was needed, possibly for euphonic reasons. 37 Such a euphonic reason might be nothing more than the need for some kind of vowel, possibly derived from the fact that the language had preserved the case vowels (in partially reduced form) before the pronominal suffixes. From the singular the a as a construct marker spread to the plural, probably aided by the widespread use of internal plurals (which would have had the singular case endings to start with). 34 Note that Waltisberg rejects Lipiński's suggestion with the following words: 'was auch immer das bedeuten mag', Waltisberg 2002, 43. However, Waltisberg's own tentative explanation (Waltisberg 2002, 24-25) seems to be more descriptive than explanatory. 35 Tropper 2000, 215. 36 Schematically, Xu Yi > XØ YØ; Xa Yi > Xa YØ; Xi Yi > XØ YØ. X is the construct noun and Y the determining noun. 37 Cf. Stempel 1999, 96. Note that Stempel's short statement on the matter does not take account of the fact that only the accusative ending was preserved. Further, he claims that a reason for the generalization of the accusative might be its higher frequency (as compared to nominative and genitive) but this is not necessarily the case, cf. Tropper 2000, 215.