Ouhalla - Semitic Relatives

Embed Size (px)

Citation preview

  • 7/29/2019 Ouhalla - Semitic Relatives

    1/13

    288 R E M A R K S A N D R E P L I E S

    Semitic Relatives

    Jamal Ouhalla

    This article evaluates a promotion-based analysis for Semitic relatives

    along the lines of Kayne 1994 and compares it with an alternative

    analysis that does not make use of promotion but shares with Kaynes

    analysis an antisymmetric view of phrase structure. The alternative

    analysis is based on establishing a parametric distinction relating to

    categorial identity of the relative clause such that it is a CP in some

    languages and a DP in others. The first type is found in languages

    where the relative complementizer is the same as the normal comple-

    mentizer of sentential complementation (e.g., Hebrew). The second

    type is found in languages where the relative complementizer is a

    determiner (e.g., Amharic and Arabic). This difference is shown to

    have crucial implications for the structure and derivation of N-initial

    and N-final relatives, as well as for some relevant typological generali-

    zations, including a generalization relating to the phenomenon of (rela-

    tive) clause nominalization.

    Keywords: CP versus DP relative clauses, construct state versus free

    state relatives, N-initial versus N-final relatives, external versus inter-

    nal nominalization

    1 N-Initial Relatives with Two Determiners

    Arabic definite relatives typically have the form [Det-N RC] seen in the Lebanese example (1).The pronominal element attached to the verb of the relative clause does not exclude cooccurrence

    with an (extracted/null) DP object (see Aoun and Choueiri 1997 and Choueiri 2002 for evidence).

    The relative complementizer illi is glossed as relative marker (RM) pending discussion of its

    status.

    (1) 1-baTT-a illi akalnaa-ha . . .

    the-duck-FEM RM we.ate-it

    the duck we ate . . .

    (Haddad and Kenstowicz 1980:144)

    Extending Kaynes (1994) analysis of N-initial relatives to (1) would consist of assigning it the

    complementation structure [DP the [CP RM [TP we.ate-it [duck]]]], and a derivation that raises/

    promotes some projection of the relativized category to Spec,C: [DP the [CP[duck]i [RM [TP we.ate

    -it [e]i]]]]. (Here and throughout, English words are used in representations for conven-

    ience.) However, there is a significant difference between Arabic and English relatives that

    casts doubt on this particular version of the promotion analysis. The difference relates to relatives

    such as (2), often cited as evidence for the complementation structure and the promotion-based

    analysis.

    Linguistic Inquiry, Volume 35, Number 2, Spring 2004288300

    2004 by the Massachusetts Institute of Technology

  • 7/29/2019 Ouhalla - Semitic Relatives

    2/13

    R E M A R K S A N D R E P L I E S 289

    (2) (*1-)paris lli bibba

    the-Paris RM I.love.it

    the Paris that I love

    (Lina Choueiri, personal communication)

    When the head N is a name, a definite article is excluded from the initial position in Arabicrelatives, contrary to what is found in their English counterparts. This suggests that the definite

    article is associated with the relativized category rather than being the head of the outer DP in

    Arabic relatives.

    The difference in (2) appears to be linked to another difference that involves the relative

    complementizer. In English, it is the same as the normal complementizer of sentential complemen-

    tation, as is well known. In Arabic, however, the relative complementizer (ya/i)lli is different

    from the normal complementizer of sentential complementation, which has the completely un-

    related form enno in Lebanese.

    (3) xabbaret-na laila enno l-mmaslin madrabiin

    told-us Laila that the-actors on.strike

    Laila told us that the actors are on strike.

    (Choueiri 2002:209)

    Aoun and Choueiri (1997) argue that the Arabic relative complementizer (ya/i)lli is actually the

    definite article l- the with additional number and gender inflection. This property is more trans-

    parent in Standard Arabic, where the relative complementizer has different inflectional forms

    depending on the number and gender features of the head N it agrees with: for example, (a)lla-

    dhii the-MASC.SG, (a)lla-tii the-FEM.SG, (a)lla-dhaani the-MASC.DUAL, (a)lla-dhiina the-

    MASC.PL (see also Fassi Fehri 1981). Aoun and Choueiri (1997) and Choueiri (2002) provide

    additional evidence for their claim, including the fact that the relative complementizer enters into

    an (in)definiteness agreement relation with the relativized category. When the relativized category

    is indefinite, in which case it lacks an (overt) definite article, the relative complementizer is

    excluded.

    (4) kteeb (*yalli) kyit ann-o laila

    book theAgr talked.she about-it Lailaa book that Laila has talked about

    (Choueiri 2002:211)

    The conclusion that the Arabic relative complementizer is a definite article raises the possibil-

    ity that it is the relative complementizer that functions as the head of the outer DP. On this

    view, Arabic relatives can be assigned the complementation structure and the promotion-based

    derivation outlined in (5), where the relative complementizer theAgr occupies the outer Dposition (see Choueiri 2002 for an alternative version).

    (5) a. [DP theAgr [CP C [TP we.ate-it [DP the duck]]]]b. [DP[DP the duck]i [theAgr [CP C [TP we.ate-it [e]i]]]]

  • 7/29/2019 Ouhalla - Semitic Relatives

    3/13

    290 R E M A R K S A N D R E P L I E S

    Besides the fact that the definite article in initial position is associated with the relativized category,

    the analysis in (5ab) differs from the one briefly entertained above in that the promoted category

    is a DP rather than a smaller projection of the relativized category (see Bianchi 1999, 2000a,b).

    This is necessary to account for the fact that (2) is excluded in Arabic. Moreover, promotion

    targets the Spec position of the outer D rather than Spec,C, which is necessary to derive therequired ordering while excluding (2). It turns out that the analysis in (5ab) has the additional

    advantage of accounting for agreement in (in)definiteness between the relativized DP and the

    relative clause by reducing it to a specifier-head relation.

    Before I outline an alternative analysis for Arabic relatives, a word about the observed

    difference between their relative complementizer and its English counterpart. This difference can

    be found within Semitic and opposes Arabic to Hebrew, where the relative complementizer is

    the same as the normal complementizer of sentential complementation.

    (6) ha-yeled se- rina ohevet oto

    the-boy that Rina loves himthe boy that Rina loves

    (Borer 1984:221)

    (7) amarti le-david se- oto rina ohevet

    said.I to-David that him Rina loves

    I said to David that Rina loves him.

    (Borer 1984:241)

    It seems that there is a genuine parametric difference between languages relating to the categorial

    identity of the relative clause such that it is a CP in Hebrew and English and a DP in Arabic.

    Internally, though, Arabic relative clauses are similar to their Hebrew and English counterpartsin that they have the same properties as normal finite clauses. This can be taken to mean that

    both types of relative clause contain a TP.

    (8) Variation in the categorial identity of the relative clause

    a. CP with a [CP C [TP]] structure in some languages (e.g., Hebrew)

    b. DP with a [DP D [TP]] structure in others (e.g., Arabic)

    The alternative analysis I would like to consider here assumes the traditional structure of

    relatives with the simple but crucial difference that the relative clause is located in a left-branching

    Spec position of N rather than being right-adjoined to NP. The Det-N-initial order is guaranteedby head raising of N to D, widely attested in Arabic noun phrases (see, e.g., Mohammad 1989,

    Ouhalla 1988).

    (9) a. [DP the [NP[DP theAgr [TP we.ate-it]] [N duck]]]b. [DP the [N duck]i [NP[DP theAgr [TP we.ate-it]] [N [e]i]]]

    Example (2) is excluded as explained above, on the grounds that Arabic names are incompatible

    with the definite article. Agreement in (in)definiteness between DN and the relative clausein Spec,N is an instance of the type of (long-distance) agreement relation more familiar from

  • 7/29/2019 Ouhalla - Semitic Relatives

    4/13

    R E M A R K S A N D R E P L I E S 291

    combinations that involve TV and a postverbal subject located in Spec,V found in numerousArabic varieties, including Lebanese (see Benmamoun 2000 and references cited therein).

    The Hebrew relative (6) can be assigned the same structure and derivation, given that Hebrew

    noun phrases exhibit the same range of evidence for N-raising to D (see, e.g., Ritter 1988, 1991,

    Siloni 1994). The difference is that there is no (in)definiteness agreement relation between D

    Nand the relative clause in Spec,N simply because the head of the relative clause, being C in

    Hebrew relatives, does not include the relevant feature.

    Obviously, the antisymmetric representation in (9), with the relative clause as a left-branching

    constituent of NP, is made possible by the fact that Arabic and Hebrew noun phrases involve N-

    raising to D. Whether an analysis along the same lines can be extended to N-initial relatives in

    other languages is a question beyond the scope of this article.

    2 N-Initial Relatives with One Determiner

    Haddad and Kenstowicz (1980) report the existence in Lebanese Arabic of another type of definiterelative illustrated in (10).

    (10) baTT-it illi akalnaa-ha . . .

    duck-FEM theAgr we.ate-itthe duck we ate . . .

    (Haddad and Kenstowicz 1980:144)

    (10) is a paraphrase of (1), but differs in that it includes only one instance of the definite article,

    namely, the one associated with the relative clause (i.e., the relative marker theAgr). Thedefinite article normally associated with the relativized category is missing. Moreover, the relative

    noun phrase has a definite reading even though the definite article of the relativized category ismissing. This property suggests that it is the relative marker that functions as the head of the

    relative noun phrase, and therefore appears to confirm the version of the promotion analysis

    outlined in the previous section (5ab), whereby the relative marker occupies the outer D position

    and the relativized DP raises to the Spec position of the outer D.

    (11) a. [DP theAgr [CP C [TP we.ate-it [DP e [duck]]]]]b. [DP[DP e [duck]]i [theAgr [CP C [TP we.ate-it [e]i]]]]

    According to the analysis outlined in (11ab), the relative in (10) differs from the relative

    in (1) only in that the D of the relativized DP is null, which is consistent with Bianchis (1999)suggestion that the relativized/promoted phrase is a DP and that its D can be null in some lan-

    guages. However, when the two relatives are placed in a broader context that includes other types

    of noun phrases in Arabic, significant differences between them emerge that do not readily reduce

    to whether the D of the relativized DP is overt or null. An appreciation of the relevant properties

    of the two relatives requires discussion of other types of noun phrases, although the discussion

    below is restricted to possessives to minimize complexity.

    Haddad and Kenstowicz (1980) point out that there are close similarities between the type

    of relative in (10) and the type of possessive in (12), called idaafa or construct state.

  • 7/29/2019 Ouhalla - Semitic Relatives

    5/13

    292 R E M A R K S A N D R E P L I E S

    (12) suur-it l-bint

    picture-FEM the-girl

    the girls picture

    (Haddad and Kenstowicz 1980:143)

    First, (12) lacks a definite article associated with the head N, as in (10). Second, the absence ofthis definite article does not prevent the noun phrase as a whole from having a definite reading,

    as in (10). Third, the head N and the possessor that follows it appear to form a single phonological

    word, a prosodic property with structural implications as we will see shortly. To provide back-

    ground information: Arabic feminine nouns are marked with the suffix -at/-it, the consonantal

    member of which, [t], remains silent in word-final positions but is pronounced in word-internal

    positions. The fact that the [t] of the feminine suffix of the head N is pronounced in (12) suggests

    that it is word-internal and that the head N must form a single phonological word with the possessor

    that follows it. Exactly the same situation is found in the relative (10), where the [t] of the feminine

    suffix of the head N is also pronounced, suggesting that it forms a single phonological word withthe relative marker immediately following it.

    To put the distinctive properties of construct state noun phrases in perspective, it is necessary

    to compare them with another type called the free state, found in some Arabic varieties, including

    Moroccan.

    (13) t-teswir-a (lqdima) dyal l-bnt

    the-picture-FEM old of the-girl

    the girls (old) picture

    In (13), the head N carries a definite article of its own independent of the definite article of the

    possessor. Moreover, the head N and the possessor do not form a single phonological word, as

    shown by the fact that an adjective can intervene between the head N and the possessor, which

    is not possible in the construct state possessive (12). In the absence of an intervening adjective,

    the noted property is reflected by the fact that the [t] of the feminine suffix of the head N is silent

    as it usually is in word-final positions. Exactly the same properties characterize the type of relative

    discussed in the previous section. Example (1) is repeated in (14).

    (14) l-baTT-a illi akalnaa-ha . . .

    the-duck-FEM theAgr we.ate-itthe duck we ate . . .

    (Haddad and Kenstowicz 1980:144)

    The conclusion to be drawn from the comparison is that relatives of the type in (10) are a

    species of construct state noun phrase, along with the possessive in (12). On the other hand,

    relatives of the type in (14) are a species of free state noun phrase, along with the possessive in

    (13). The question whether the two types of relatives have the same or different representations

    therefore is part of the more general question whether construct state and free state noun phrases

    have the same or different representations.

  • 7/29/2019 Ouhalla - Semitic Relatives

    6/13

    R E M A R K S A N D R E P L I E S 293

    Construct state and free state noun phrases have attracted a substantial amount of attention

    in the literature on Arabic and Hebrew, and the view that emerges is that their respective distinctive

    properties imply different representations and derivations. For example, Borer (1989) argues at

    length that construct state noun phrases such as the possessive in (12), but not free state noun

    phrases such as (13), are characterized by merger of the possessor with the head N, resulting ina single complex word. Benmamoun (2000) basically endorses this analysis for corresponding

    Arabic possessives and goes further in arguing that the remaining distinctive properties of construct

    state possessives follow precisely from merger between the head N and the possessor. The details

    of the analysis are not crucially relevant to the point of the discussion. What is relevant is that

    both studies stress that merger takes place under structural adjacency, with the clear implication

    that the head N and the possessor are structurally adjacent in construct state noun phrases, where

    merger takes place, but not in free state noun phrases, where merger does not take place. Extended

    to relatives, this reasoning implies that the relative clause is structurally adjacent to the head N

    in construct state relatives such as (10) but not in free state relatives such as (14). Therefore, any

    attempt to assign the two types of relatives a uniform representation where the head of the relative

    clause bears the same structural relation to the relativized N would be inconsistent with the

    observed crucial difference between them.

    To arrive at the desired representations of the two types of relatives, what needs to be done

    is identify the representations of construct and free state possessives and then extend them to

    relatives by substituting the relative clause into the position of the possessor. Adopting the structure

    of the noun phrase and analysis outlined in Ritter 1988, 1991, and numerous other studies, construct

    state possessives such as (12) have the structure and derivation outlined (15).

    (15) [DP[picture]i [NumP[DP the-girl] [Num . . . (Adj) . . . [ NP[e]i]]]]

    The structure includes a Num category intervening between D and NP. The possessor is located

    in Spec,Num (the genitive position), and the head N raises to D (via Num). The resulting structural

    adjacency between the possessor and N in D makes it possible for the possessor to merge with

    N. Free state possessives such as (13) have the different representation outlined in (16), where

    the possessor is located in Spec,N (the lower subject position). The fact that the possessor is not

    structurally adjacent to the head N located in D is what accounts for the absence of merger

    between the two constituents in free state possessives.

    (16) [DP[picture]i [NumP Num . . . (Adj) . . . [ NP[DP the-girl] [N[e]i]]]]

    The derivational difference between (15) and (16) that yields the two different representations is

    that the possessor raises to Spec,Num in (15) but not in (16), where it remains in Spec,N.

    Turning now to relatives, construct state relatives such as (10) have the representation outlined

    in (17) with the relative clause located in the position of the possessor in Spec,Num. The structural

    adjacency between the relative clause and N under D makes it possible for the head of the drelative

    clause to merge with N.

    (17) [DP[N duck]i [NumP[DP theAgr [TP we.ate-it]] [Num [NP[e]i]]]]

  • 7/29/2019 Ouhalla - Semitic Relatives

    7/13

    294 R E M A R K S A N D R E P L I E S

    In comparison, free state relatives such as (14) have the different representation outlined in (18),

    where the relative clause is located in the position of the possessor in Spec,N. (18) is basically

    the representation assigned to (14) in the previous section, except that it includes the additional

    category Num.

    (18) [DP the [N duck]i [NumP Num [NP[DP theAgr [TP we.ate-it]] [N[e]i]]]]To round off this section, a word about the nature of the merger operation that takes place

    in Arabic and Hebrew construct state noun phrases and its absence in free state noun phrases.

    The evidence relating to the pronunciation or not of the consonantal sound [t] of the feminine

    suffix -at in Arabic noun phrases, and indeed also the evidence relating to stress placement in

    their Hebrew counterparts discussed in Borer 1989, while it does not exclude merger in Syntax

    (Borer 1989) or at the level of Morphological Structure (Benmamoun 2000), does not necessarily

    imply it either. As far as this particular kind of evidence is concerned, which is phonological in

    nature, the merger might as well be prosodic and takes place at PF.

    To summarize, the promotion-based analysis assigns construct state and free state relatives

    the same representation, with the difference between them reduced to whether the D of the

    relativized DP is overt or null. However, the two types of relatives exhibit significant differences

    that do not appear to be reducible to whether the D of the relativized DP is overt or null. Rather,

    the differences appear to reflect a structural difference relating to whether the relative clause is

    structurally adjacent to the head N or not. In contrast, the proposed alternative analysis allows

    for the noted difference by virtue of treating the relative clause as a subject of the relative noun

    phrase that can occupy either the higher subject position (Spec,Num), adjacent to N in D, or the

    lower subject position (Spec,N), which is nonadjacent to N in D.

    3 N-Final Relatives

    Amharic relatives are N-final, consistent with the fact that the language is generally head-final,

    where main verbs follow their objects, auxiliaries follow main verbs, determiners follow nouns,

    and so on. The identity of the grammatical marker (GM) ya- on the verb in (19) is discussed in

    detail below.

    (19) lj-u ya-gaddala-w baab

    boy-the GM-killed-the snake

    the snake the boy killed

    (Mullen 1986:386)

    Kayne (1994:9297) bases his version of the promotion-based analysis of N-final relatives

    on Amharic relatives such as (19). The analysis assumes the same head-initial complementation

    structure as for N-initial relatives, where the definite article occupies the outer D position and

    the relativized category is located in the relativized position. The N-final order is derived by

    promotion of the relativized category to the Spec position of C, followed by raising of IP/TP to

    the Spec position of the outer D. The derivation and structures are repeated in (20) with English

    words corresponding to the Amharic example (19).

  • 7/29/2019 Ouhalla - Semitic Relatives

    8/13

    R E M A R K S A N D R E P L I E S 295

    (20) a. [DP the [CP C [IP boy-the [[snake] GM-killed]]]]

    b. [DP the [CP[snake]i C [IP boy-the [[e]i GM-killed]]]]

    c. [DP[IP boy-the [[e]i GM-killed]]j [the [CP[snake]i C [e]j]]]

    The analysis outlined in (20) treats the definite article attached to the verb of the relative

    clause as the head of the whole/outer relative DP. However, this is not the only possibilityavailable, at least in the present context. Another possibility, suggested by the facts of Arabic

    discussed above, is that the definite article in question has the same role as the relative marker

    of Arabic relatives, meaning it is the head of a DP relative clause with a Det-final order consistent

    with the Det-final character of Amharic noun phrases in general. On this particular view, (19)

    will have the representation [DP[DP[TP boy-the GM-killed] the] snake], where the definite article

    is bracketed along with the DP relative clause of which it is the head. In other words, (19) will

    have a representation essentially similar to that of possessives such as (21) with the possessor in

    place of the relative clause: [DP[DP[boy] the] notebook].

    (21) ya-lj-u dabtar

    GM-boy-the notebook

    the boys notebook

    (Mullen 1986:307)

    The possibility that the definite article in (21) may not be a constituent of the possessor is excluded

    by the fact that when the possessor is a name, the definite article does not and cannot appear.

    (22) ya-brhanu wssa

    GM-Birhanu dog

    Birhanus dog(adapted from Mullen 1986:312)

    To be in a position to outline the details of an alternative analysis of Amharic relatives based

    on the conclusion that the definite article attached to the verb is the head of the relative clause

    to its left rather than the head of the outer relative DP, it is necessary first to provide the derivations

    of simpler noun phrases, starting from the simplest type such as dabtar-u notebook-the. Follow-

    ing Kaynes (1994) antisymmetric approach to head-final phrases, these can be assigned the head-

    initial structure in (23) with the Det-final order derived by raising of NumP to Spec,D.

    (23) a. [DP the [NumP Num [NP notebook]]]

    b. [DP[NumP Num [NP notebook]]i [the [e]i]]

    Pursuing the same approach to head-final phrases, relative clauses can be assigned the structure

    and derivation outlined in (24), where the Det-final order is derived by raising of TP to Spec,D.

    (24) a. [DP the [TP[boy-the] GM-killed]]

    b. [DP[TP[boy-the] GM-killed]]i [the [e]i]

    In both (23) and (24), the raised category is not of the DP-type; moreover, it targets a non-Case

    position (Spec,D). It is possible that these movements apply at a postsyntactic level and involve

  • 7/29/2019 Ouhalla - Semitic Relatives

    9/13

    296 R E M A R K S A N D R E P L I E S

    PF considerations having to do with word ordering, including the possibility that the definite

    article is a phrasal clitic on a par with English s. It is worth noting that when the relative clause

    includes an auxiliary in addition to the main verb, the definite article appears attached to the

    auxiliary, the nearest word to it.

    (25) wadKo ya-nabbara-w beetfallen GM-PAST-the house

    the house that had fallen down

    (Mullen 1986:371)

    The remaining types of noun phrases are possessives and relatives. The strategy adopted for

    Arabic in the previous section is to assign them basically the same representation and derivation

    such that the possessor and the relative clause occupy the same position within the larger DP

    that includes them. Starting with possessives, they can be assigned the representation in (26),

    corresponding to example (22), where the possessor occupies Spec,Num. No movements of the

    type assumed for (23) and (24) are required.

    (26) [DP D [NumP[DP GM-Birhanu] [Num [NP dog]]]]

    Substituting the relative clause for the possessor in (26) yields the representation in (27) for

    relatives such as (19).

    (27) [DP D [NumP[DP[TP boy-the GM-killed] the] [Num [NP snake]]]]

    According to the analysis in (27), Amharic relatives are basically similar to Arabic construct

    state relatives in that their relative clause is a DP that occupies Spec,Num (the genitive position).

    The N-initial versus N-final difference reduces to the existence of N-raising to D in Arabic nounphrases and its absence in Amharic noun phrases. It turns out that Amharic relatives include more

    direct evidence that their relative clause is a DP and that it occupies the genitive position within

    the relative noun phrase. The evidence has to do with the prefix ya- that appears on the verb of

    the relative clause noted earlier.

    Earlier analyses of ya-, which include Bach 1970 and Fulass 1972, treat it as a relative

    marker, presumably corresponding to the Comp position of the relative clause (see Mullen 1986

    for discussion). Its appearance in possessives was taken as evidence that they are hidden rela-

    tives insofar as they derive from an underlyingly complex structure that resembles that of

    relatives. Manahlot (1977) cites a third context for the prefix ya- that is not amenable to treatment

    as a relative marker. The context is factive complements such as in (28). CM stands for clause

    marker and is identified in section 4 below in the context of a discussion of the phenomenon of

    clause nominalization.

    (28) ya-kasa-n bet ma-gzat samma-hu

    of-Kasa-OM house CM-buy heard-I

    I heard that Kasa bought a house.

    I heard of Kasas buying a house.

    (Manahlot 1977:123)

  • 7/29/2019 Ouhalla - Semitic Relatives

    10/13

    R E M A R K S A N D R E P L I E S 297

    Manahlot considers in detail an analysis for factive complements that treats them as constituents

    of a noun phrase headed by a null N with the meaning fact. The analysis has the consequence

    that factive complements are included in a context that, at least at the time, was deemed similar

    to that of relative clauses (a complex noun phrase). Besides the fact that complement clauses of

    nouns and relative clauses are not the same thing, Manahlot goes on to point out that the analysishas the significant problem of failing to explain why the prefix ya- appears on the verb in relatives

    (19) but on the subject in factive complements (28).

    The analysis of the prefix ya- compatible with its distribution in all three contexts is one

    that treats it as a genitive Case marker. On this view, its appearance on the possessor in possessives

    is self-evident. Its appearance on the subject of factive complements is not unexpected either

    given that factive complements are known to have nominal properties across languages, including

    English (see the second gloss of (28)). What may be considered unexpected is the appearance of

    ya- on the relative clause in relatives. However, according to the analysis outlined in (27), where

    the relative clause is a DP that occupies the same position as the possessor in possessives, the

    fact that it bears a genitive Case marker is precisely what is expected. The reason ya- appears as

    a prefix on the verb rather than as a phrasal clitic associated with the whole relative clause is

    simply due to its property as a prefix. It is worth noting that ya- consistently appears on the last

    member of the relative clause, so that if the relative clause includes an auxiliary in addition to a

    main verb, as in (25), ya- appears on the auxiliary.

    To summarize, an analysis for Amharic N-final relatives that treats the relative clause as a

    DP that occupies the genitive subject position of the relative noun phrase is not only possible,

    but arguably more desirable. It provides a unified analysis of Amharic noun phrases in general,

    and a consistent account of the prefix ya- in all contexts as a Case morpheme that marks DPs

    located in Spec,Num.

    4 Relatives and (Relative) Clause Nominalization

    Kayne (1994) cites the two typological generalizations in (29) and explains how they both follow

    from an analysis of N-final relatives based on a complementation structure and promotion of the

    relativized category. Given that both properties implicate CP, they follow from the idea that what

    moves to the prenominal position is something smaller than CP, namely, IP (see derivation in

    (20)).

    (29) a. N-final relatives lack relative pronouns.

    b. N-final relatives never display a complementizer that is identical to the normal

    complementizer of sentential complementation. (Kayne 1994:93)

    Properties (29ab) also follow from the alternative analysis outlined above, but with an

    interesting twist. As far as Amharic relatives are concerned, these properties follow from the

    conclusion that the relative clause is a DP with a [ DP D [TP]] structure that lacks CP altogether.

    The interesting twist is the implication that (29ab) are expected to be properties of relatives

    where the relative clause is a DP irrespective of whether they are N-final or N-initial. In other

    words, the proposed alternative analysis predicts that both properties should characterize Arabic

  • 7/29/2019 Ouhalla - Semitic Relatives

    11/13

    298 R E M A R K S A N D R E P L I E S

    relatives, which are N-initial. The prediction is consistent with the facts since Arabic relatives

    lack relative pronouns and, as explained in section 1, do not display the normal complementizer

    of sentential complementation.

    (30) a. Relatives with a DP relative clause lack relative pronouns irrespective of whether

    they are N-final or N-initial.b. Relatives with a DP relative clause do not display a complementizer that is identical

    to the normal complementizer of sentential complementation, irrespective of whether

    they are N-final or N-initial.

    Kayne (1994) discusses another typological generalization about relatives pointed out in

    Keenan 1985 and repeated in (31).

    (31) In prenominal RCs, VREL is almost always some sort of non-finite form. . . . We have

    often above called such verb forms participles. . . . By contrast, postnominal RCs most

    typically present VREL in the form it would have as the main verb of a declarativesentence though this is certainly not always the case. (Keenan 1985:160161)

    (31) implicates the phenomenon of clause nominalizationthat is, processes by which a clause

    is turned into a noun phraselike category. As far as Amharic and Arabic are concerned, there

    are two major strategies, which I will call here internal nominalization and external nominaliza-

    tion. Internal nominalization involves replacement of the finite inflectional category specified for

    tense and nominative Case (INom/TNom) with another inflectional category specified for genitive

    Case and no tense (IGen/NumGen) (see Abney 1987). This inflectional category corresponds to

    special morphology on the verb, giving it a form variously known as the participle, gerund, verbal

    noun, and (in the Arabic tradition) masdar (see Comrie and Thompson 1985). An instance ofinternal nominalization can be seen in Amharic factive complements, where the verb is uninflected

    for tense (and agreement) and instead carries a nominalizing morpheme glossed as CM in

    (28) (see Manahlot 1977 for discussion). External nominalization, on the other hand, involves

    replacement of C(P) with D(P), leaving the rest of the clause untouched, especially its internal

    finite character. Instances of this strategy of nominalization can be seen in Amharic and Arabic

    relative clauses, both of which are DPs (rather than CPs) that contain a finite TP, as concluded

    above.

    (32) Strategies of (relative) clause nominalization

    a. Internal: replacement of TNom with NumGen

    b. External: replacement of C(P) with D(P)

    As far as Amharic factive complements are concerned, it is not clear if internal nominalization

    necessarily involves simultaneous replacement of C(P) with D(P) (i.e., external nominalization).

    Example (28) does not include a definite article associated with the factive complement. However,

    there is evidence from Hebrew relatives of the type in (33) that some instances of internal nominali-

    zation within Semitic also involve external nominalization.

  • 7/29/2019 Ouhalla - Semitic Relatives

    12/13

    R E M A R K S A N D R E P L I E S 299

    (33) hine ha-is ha-xosev rak al kesef

    here the-man the-think only about money

    Here is the man that thinks only about money.

    (Siloni 1995:447)

    Hebrew relatives of the type in (33) are discussed by Siloni (1995), who explains that they differfrom the type in (2) in at least two major respects. First, their relative clause is nonfinite. Second,

    their relative marker is the definite article instead of the complementizer of normal sentential

    complementation. In other words, the relative clause in (33) is a DP relative clause, although it

    differs from its Amharic and Arabic counterparts in that it is internally nonfinite.

    The nonfinite prenominal relative clauses referred to in the first part of generalization (31)

    are presumably instances of internal nominalization. According to the analysis outlined in the

    previous sections, prenominal relative clauses are expected to be nominalizations because they

    occupy a Case position (Spec,Num). Amharic relative clauses are actually an exception to the

    strong tendency described in generalization (31) because they are internally finite. However, theyare not an exception to the generalization emanating from the analysis proposed here that prenomi-

    nal relative clauses that occupy Spec,Num are expected to be nominalized one way (internally)

    or the other (externally).

    Turning now to the second part of generalization (31), postnominal relative clauses are not

    expected to be nominalized because they have the option of occupying Spec,N, a non-Case

    position. In Semitic, nonnominalized postnominal relative clauses are found in Hebrew relatives

    of the type in (2). (31) makes it clear that postnominal relative clauses can also be of the nomi-

    nalized type, bearing in mind that nominalization can also be of the external type. In Semitic,

    nominalized postnominal relative clauses are found in Arabic (external nominalizations) and

    Hebrew (simultaneous internal and external nominalizations). Moreover, they can either occupy

    the lower subject position (Spec,N), as in Arabic free state relatives, or the higher genitive position

    (Spec,Num), as in Arabic construct state relatives. It is not clear which of the two positions the

    nonfinite DP relative clause occupies in the Hebrew relative (33). What the proposed analysis

    predicts is that if a relative clause occupies the genitive position of the relative noun phrase, then

    it is expected to be nominalized.

    To summarize, the alternative analysis of relatives outlined in the previous sections, which

    treats the relative clause as a subject that can occupy either of two subject positions of the relative

    noun phrase, is not only consistent with the typological generalizations in (29)(30) and (31),

    but arguably offers comparatively better insights into the reasons behind them.

    References

    Abney, Steven. 1987. The English noun phrase in its sentential aspect. Doctoral dissertation, MIT, Cambridge,

    Mass.

    Aoun, Joseph, and Lina Choueiri. 1997. Resumption and Last Resort. Ms., University of Southern California,

    Los Angeles.

    Bach, Emmon. 1970. Is Amharic an SOV language? Journal of Ethiopian Studies 7:920.

  • 7/29/2019 Ouhalla - Semitic Relatives

    13/13

    300 R E M A R K S A N D R E P L I E S

    Benmamoun, Elabbas. 2000. The feature structure of functional categories: A comparative study of Arabic

    dialects. Oxford: Oxford University Press.

    Bianchi, Valentina. 1999. Consequences of antisymmetry: Headed relative clauses. Berlin: Mouton de

    Gruyter.

    Bianchi, Valentina. 2000a. The raising analysis of relative clauses: A reply to Borsley. Linguistic Inquiry

    31:123140.Bianchi, Valentina. 2000b. Some issues in the syntax of relative determiners. In The syntax of relative clauses,

    ed. by Artemis Alexiadou, Paul Law, Andre Meinunger, and Chris Wilder, 5382. Amsterdam: John

    Benjamins.

    Borer, Hagit. 1984. Restrictive relatives in Modern Hebrew. Natural Language & Linguistic Theory 2:

    219260.

    Borer, Hagit. 1989. On the morphological parallelism between compounds and constructs. In Morphology

    yearbook, ed. by Geert Booij and Jaap van Marle, 4564. Dordrecht: Foris.

    Choueiri, Lina. 2002. Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic. Doctoral

    dissertation, University of Southern California, Los Angeles.

    Comrie, Bernard, and Sandra A. Thompson. 1985. Lexical nominalization. In Language typology and syntac-

    tic description, ed. by Timothy Shopen, 2:349397. Cambridge: Cambridge University Press.Fassi Fehri, Abdelkader. 1981. Linguistique arabe: Forme et interpretation. Rabat: Publications of the

    Faculty of Letters.

    Fulass, Hailu. 1972. On Amharic relative clauses. Bulletin of the School of Oriental and African Languages

    35:497513.

    Haddad, Ghassan, and Michael Kenstowicz. 1980. A note on the parallels between the definite article and

    the relative clause marker in Arabic. Studies in the Linguistic Sciences 10:141148.

    Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press.

    Keenan, Edward. 1985. Relative clauses. In Language typology and syntactic description, ed. by Timothy

    Shopen, 2:141170. Cambridge: Cambridge University Press.

    Manahlot, Demissie. 1977. Nominal clauses in Amharic. Doctoral dissertation, Georgetown University,

    Washington, D.C.

    Mohammad, Mohammad. 1989. The sentence structure of Arabic. Doctoral dissertation, University of South-

    ern California, Los Angeles.

    Mullen, Dana. 1986. Issues in the morphology and phonology of Amharic: The lexical generation of pronomi-

    nal clitics. Doctoral dissertation, University of Ottawa, Ontario.

    Ouhalla, Jamal. 1988. The syntax of head movement. Doctoral dissertation, University College London.

    Ritter, Elizabeth. 1988. A head-movement approach to construct state noun phrases. Linguistics 26:909929.

    Ritter, Elizabeth. 1991. Two functional categories in noun phrases: Evidence from Hebrew. In Perspectives

    on phrase structure: Heads and licensing, ed. by Susan Rothstein, 3762. Syntax and Semantics

    25. San Diego, Calif.: Academic Press.

    Siloni, Tal. 1994. Noun phrases and nominalizations. Doctoral dissertation, Universite de Geneve.

    Siloni, Tal. 1995. On participial relatives and complementizer D: A case study in Hebrew and French.

    Natural Language & Linguistic Theory 13:445487.

    University College Dublin

    Linguistics Department

    Belfield, Dublin 4

    Ireland

    [email protected]