• Keine Ergebnisse gefunden

Morphological derivation and semantic derivation

Properties of the Arabic language

4.2 Morphological properties

4.2.2 Morphological derivation and semantic derivation

As already mentioned in previous sections, the derivation of Arabic verbs can allows two different directions according to the morphological and semantic properties of their base forms. This is particularly the case with pairs of transitive and intransitive verbs. Thus, the transitive-causative can not only be derived from an intransitive base form, as is the case with the verb pair ˇsah

˙iba ‘pale’ and Pa-ˇshaba ‘CAUS-pale’, but the intransitive can also be derived from an transitive

Transitive Intransitive

halaka ‘perish’ halaka ‘CAUS-perish’

badaPa ‘begin’ badaPa ‘CAUS-begin’

zaada ‘grow’ zaada ‘CAUS-grow’

Table 4.9: Intransitive-transitive conversion

base form by adding an intransitivity affix such as with the verb pair sakaba‘pour’

and Pin-sakaba ‘INTR-pour’ (see (6) and (7)). This fact presents a challenge for compositional semantics under the assumption that the transitive causative is derived from the intransitive unaccusative by adding the appropriate semantic predicate (Chierchia,1989). On the other hand it violates the general principle of diagrammatic iconocity which assumes a correlation between the morphological and the semantic base and derived forms as is attested by derivationally related pairs such as singular and plural, diminutive and non-diminutive, comparative and superlative, etc. (Haspelmath, 1993).

(6) Intransitive-inchoative =⇒Transitive-causative (marked) d¯aaba ‘melt’ =⇒Pa-d

¯aaba ‘CAUS-melt’

(7) Transitive-causative =⇒Intransitive-inchoative (marked) qat˙aca ‘cut’ =⇒Pin-qat

˙aca ‘INTR-cut’

Studies on the derivational behavior of languages as in Nedyalkov & Silnitsky (1973), Haspelmath(1993) and Nichols et al.(2004) have shown that there are specific typological tendencies in the lexicalization of causative and non-causative verb pairs that separate groups of languages. There are at least four groups:

The first group has the intransitive non-causative as base form and the transitive causative as the derived and overtly marked form (Mongolian). These languages mark both verb pairs each with a distinctive marker. The second group adopts the inverse strategy of intransitivization (Russian), and a third group of languages marks neither of the verb pairs (English). The examples (6), (7), (8) and (9) show that all four lexicalization strategies are attested in Arabic, whereby the first two strategies are more frequent and the two last strategies seem to be an accidental product.

(8) Intransitive-inchoative (unmarked) =⇒Transitive-causative (unmarked)

halaka ‘perish’ =⇒halaka ‘cause to perish’

(9) Intransitive-inchoative (marked) =⇒Transtive-causative (marked) Pa-ˇskala ‘became difficult’ =⇒Pa-ˇskala ‘make difficult’

The difficulty in these cases arises when one has to decide which one is the base form and which one is the derived form. Making this decision reveals itself to be particularly important for our work, since diathesis alternations in Arabic often involve a morphological marking and the direction of verb derivation decides on the nature of the alternation. For the causative-non/causative derivation and non-causative-causative, we do not only have one alternation (non-causative/causative alternation) but also a second one (causative/non-causative alternation) (see chap-ter 5).

Doron (2003) resolved this problem by considering transitive verbs as well as intransitive verbs to be directly derived from a basic predicate: the root. A transitive verb is built by combining the root with a morpheme which contributes an additional argument, where in the unmarked case this argument is an agent.

Similarly, intransitive verbs are directly derived from the root by combining the root with a morpheme that blocks the appearance of the agent. This approach, however has two related shortcomings. First, it treats lexically and semantically related verbs as different entries. Transitive verbs and their intransitive counter-parts are derived from a common root in the same way that nouns, adjectives and other parts of speech are derived. This entails that intransitive verbs are related to their transitive counterpart just as nouns are related to verbs. They share the same root, which is semantically not well-formed, since, as we have seen, they share the same meaning components and may participate in the same alternations.

Furthermore, this approach leads one to believe that the lexicon is a construct in which anything appears to be derived directly from root forms and consequently disconnected from each other. Thus, pairs of derivationally related words like verbs and their nominalizations and verbs and their participle forms also lose any connection to each other since they are built by two independent processes. In addition, it gives the impression that the distinction between derivationally derived

transitive and intransitive verbs is a matter of regular morphological marking.

That is, there is always a morpheme or a class of morphemes which is exclusively and unambiguously reserved for marking intransitivity, others are reserved for marking transitivity. The derivational processes responsible for building verbs in Semitic languages entails a combination of abstract roots with minimal semantic content with this class of morphemes.

However, as we have seen in section 4.2.1, one template/marker can fill different functions according to the verb it is combined with. Concomitantly, one mor-phological verb form can be assigned two different meanings, a transitive and an intransitive as is the case with the verb badaPa ‘CAUS-begin/INCH-begin’. This case challenges Doron’s point of view. By applying her approach to such verb pairs, one must conclude that they are directly derived from the root, which means that we have the same process generating two different readings using the same morphological material. Thus, deriving two or three related verb forms from the same root in two or three independent processes results in an unnatural degree of redundancy. It also entails that generating three related verbs like Pinokataba

INCH-write’, kataba ‘write’ and Paktaba ‘cause to write’ takes place in three in-dependent processes. In the case of kataba the feature CaCaCa is responsible for producing a simple transitive verb with two thematic roles: agentand theme assigned to the subject and the object respectively. In the case of the verbPaktaba the feature PaCCaCa introduces the thematic structure<cause, agent, theme>, assigned to subject, object and object2 respectively. The redundancy lies in the fact that two different morphological templates introduce the same semantic information. In our case, the same role (agent) is introduced by two different templates, in the first case by CaCaCaand in the second case byPa-CCaCa, where the information introduced by this one subsumes the information introduced by CaCaCa. This subsumption however is not to be understood as a derivational relation in Doron’s approach (10c).

(10) a. in-kataba-ti l-qis

˙s

˙at-u.

INCH-write-F DEF-story-NOM

‘The story was written.’

b. katab-a l-muPallif-u qis

˙at-an.

wrote DEF-author-NOM story-ACC

‘The author wrote a story.’

c. Pa-ktaba l-h

˙uzn-u l-muPallif-a qas

˙iidat-an.

CAUS-writeDEF-sadness-NOM DEF-author-ACC poem-ACC

‘The sadness makes the author write a poem.’

In contrast, we claim that the derivational behavior of verbs at the level of event decomposition is reflected on the morphological level. We rely onMcCarthy(1979, 1981) in assuming the root carries semantic content. However, we go further in claiming that the semantic content of the root is not primitive, that is, it can be constituted of one semantic predicate or of more than one.

Verbal derivation (and derivation in general) is an incremental process in which only a small number of elements are directly derived from the root and the rest is recursively derived from these elements at the next levels. The forms derived at the first level of the derivation in particular includes the parts of speech e.g. nouns, verbs, adjectives and adverbs, which are themselves subject to other derivational processes. For instance, the reciprocal verb takaataba ‘correspond’ is the result of an incremental derivational process that starts with the root [k][t][b] and involves two intermediary forms: The transitive form kaataba ‘correspond’ is itself derived from the verb kataba‘write’, which in turn is directly derived from the root (10).

The first step of this derivational process always produces one simple verb with simple templates like CaCaCa, CaCiCaor CaCuCa (4.10). These three templates

Template Verb

CaCaCa [s]a[k]a[b]a ‘pour’

CaCiCa [y]a[?]i[s]a ‘despair’

CaCuCa [m]a[l]u[h

˙]a ‘INCH-salt’

Table 4.10: The three templates of simple verbs

can encode both readings (originally attributed by the root): The causative and the non-causative reading (inchoative or reflexive). This means that Arabic allows for a bidirectional derivation on the morphological level: one that derives causative verb form inchoative verbs and one that derives inchoative verbs from causative verbs (11).

Root([k][t][b]) V(kataba)

‘write’

V(Pinkatab)

INTR-write’

V(kaataba)

‘correspond’

V(ta-kaataba)

INTR-correspond’

V(ta-kaatub)

INTR-correspondence’

N(mukaataba)

‘correspondence’

N(kitaab)

‘a book’

ADJ(kitaabii)

‘relative to book’

N(kitaaba)

‘writing’

1

Figure 4.2: A part of the derivation tree of the root [k][t][b]

(11) a. Inchoative =⇒causative ˇsah˙iba‘pale’ =⇒Pa-ˇsh

˙aba ‘CAUS-pale’

b. Causative =⇒Inchoative

sakaba ‘pour’ =⇒Pinsakaba INCH-pour’

The bidirectionality of the derivation of causative and non-causative verbs seems to reflect a general verbal event property. Thus, by analyzing the semantic properties of groups of verbs that show this derivational behavior, one finds considerable differences between them, especially when it comes to the nature of causativity and the obligatory presence of a real or a presupposed agentive causer.

Verbs that are basic intransitive-inchoative and derive a transitive causative have a meaning that excludes the agent and involves an event that occurs spontaneously.

Verbs that are basic transitive-causatives and derive an intransitive-non-causative require an agentive participant who causes the situation (Haspelmath,1993, 90).

Unlike the inchoative base form, the derived intransitive verb presupposes the existence of an agent although it is blocked from appearing at the syntactic and lexical surface. Thus, with the verb d

¯aaba ‘melt’, the event of melting is conceptualized as occurring spontaneously without the intervention of an agent, whereas with the verb Pin-qat

˙acINCH-cut’ the event of cutting cannot occur

spontaneously and without the intervention of an external agentive force. Levin

& Hovav (1995) claim that the first type of verb denotes an “internally caused”

event. This type of verb has one event only and there is no explicit element that controls it. These verbs mark their transitive alternant. The second type denotes an “externally caused” event, which is constituted of two subevents controlled by the agent. These verbs mark their intransitive alternant morphologically.

In this dissertation, we will treat both alternations (inchoative/causative) as distinct. The inchoative/causative alternation has the semantic analysis in (12).

In the inchoative variant no agent role is presupposed and the causative/non-causative alternation generally has the meaning in (13). The non-inchoative derived form presupposes an agentive force. Note that, this representation style is taken form English VerbNet (Kipper Schuler,2005), where the presupposed theme is marked by the interrogative (?). This style will be used in this dissertation as well as in the Arabic VerbNet (See section 9 for more details).

(12) Inchoative/causative alternation

¯daaba ‘melt’ =⇒?a-d

¯aaba ‘TR-melt’

change of state(Patient) =⇒cause(Agent, change of state(Patient)) (13) Causative/inchoative alternation

qat˙aca ‘cut’ =⇒Pin-qat

˙aca ‘INTR-cut’

cause(Agent, change of state(Patient)) =⇒change of state(?Agent, Patient)