Rethinking the relationship between SELF-intensifiers and reflexives.

Gast, Volker ; Siemund, Peter

Abstract

Recent studies into the syntax and semantics of intensifying self-forms (e.g. [John HIMSELF] came) have shown that a distinction needs to be drawn between two uses of such expressions: a juxtaposed or adnominal use (cf. above), and a nonjuxtaposed use (e.g. John [came HIMSELF]). This differentiation allows us to reconsider a number of issues relating to the synchronic and diachronic relationship between SELF-intensifiers and reflexive anaphors. Assessing relevant crosslinguistic data against the background of the aforementioned distinction reveals a surprising fact: patterns of "formal relatedness" suggest a particularly strong empirical as well as conceptual tie-up between reflexives and SELF-intensifiers in their nonjuxtaposed rather than adnominal use. This is remarkable because it has generally been assumed that it is always the adnominal SELF-intensifier which gives rise to the development of reflexive markers. In the light o four crosslinguistic findings, we explore the synchronic and diachronic relationship between reflexives and SELF-intensifiers in their nonjuxtaposed use. We argue that the picture of a (unidirectional) development from adnominal SELF-intensifiers to reflexives needs to be modified insofar as reflexive markers often develop from nonjuxtaposed, rather than adnominal, intensifiers. Moreover, reflexive markers often form part of a strategy of SELF-intensification, which entails that the reflexives are older than the resulting intensifiers.

1. Introduction (1)

1.1. SELF-intensifiers and reflexives in a historical perspective

As is well known, there is a particularly strong (synchronic and diachronic) relationship between intensifying self-forms or "emphatic reflexives" on the one hand, and (nonemphatic) reflexives or "reflexive anaphors" on the other. For example, the English self-anaphors have developed from the adjunction of an intensifying self-form to the plain object pronouns in Old and Middle English. This process of reanalysis is well documented and has been described by Konig and Siemund (2000a), van Gelderen (2001), and Keenan (2002), among many others. It is illustrated in (1) and (2):

(1) he hine vncu[??] makede. he him unknown made 'He made himself unknown/unrecognizable.' (Caligula 3302, quoted from van Gelderen 2001: 72)

(2) he makede him-seluen muchel clond. he made him-SELF much pain 'He made for himself much pain.' (Caligula 5839, quoted from van Gelderen 2001: 73)

Similar developments can be observed in many other languages, and it is widely agreed that the development from SELF-intensifiers to reflexive markers as illustrated above is a process of considerable generality (cf. Faltz 1985; Levinson 1987, 1991; Kemmer 1993; Konig and Siemund 2000a, 2000b; Schladt 2000; Heine 2003). It can be represented in the simple formula given in (3).

(3) SELF-intensifier > reflexive

In this article we aim to show that the formula in (3) is not only simple, but also simplifying, insofar as it lumps together a number of similar and possibly interrelated, but clearly different, processes. While carrying out a typological study on the formal encoding of SELF-intensitiers and reflexive anaphors (cf. Konig and Siemund 2000b, 2005; Gast et al. 2003), we found that the relationship between those two types of expressions is in fact much more complex than the formula in (3) suggests. First, the genesis of reflexive markers often involves SELF-intensifiers that occur in a nonadjacent position to the NP they are associated with (e.g. himself in John [did the washing himself]). Such non-juxtaposed intensifiers, which differ both syntactically and semantically from the juxtaposed self-form in (2), and which are encoded using distinct lexical items in many languages, interact with predicate meanings rather than referential meanings. And second, SELF-intensifiers may not only develop into, or give rise to the creation of, reflexive markers; they may also be the target of a semantic change which originates from expressions of reflexivity.

1.2. The empirical basis of the investigation

Our typological survey is based on data that has been collected during a six-year research project on SELF-intensifiers and reflexives, on the basis of both native speaker consultation (field work and questionnaires) and grammatical descriptions. (2) In the course of this project, information was assembled on more than two hundred languages, but given that the issues under investigation, especially the differentiation between two types of intensifiers (juxtaposed vs. nonjuxtaposed), requires fine-grained analyses and a very high standard of reliability, not all of those data were used. Since we want to make at least a weak statistical statement, we have moreover aimed to balance the sample genetically and areally, as far as this was possible. As a consequence, the sample was cut down to 72 languages, and some of the languages used in the text for the purpose of illustration do not appear in it.

The sample, which is listed in Appendix 2, contains languages from 32 families, according to the Ethnologue classification. The four largest families (Afro-Asiatic, Austronesian, Indo-European, and Niger-Congo) are each represented by at least four languages, while all other families are represented by three languages at most. As far as the areal distribution of the sample languages is concerned, we have aimed to achieve a more or less even distribution over five macro-areas: (3) Americas: 16 languages; Africa: 18 languages; Northern Eurasia: 16 languages; Southern Eurasia: 11 languages; Australia/Papua New Guinea/Oceania: 11 languages. Still, we are aware that the degree of areal and genealogical balancing of the sample does not perfectly meet the common standards of large-scale typological investigations as described, for example, by Rijkhoff et al. (1993). We will therefore refrain from applying any sophisticated statistical procedures, providing the relevant figures in terms of absolute numbers instead. We would like to point out that the main objective of this article is not to describe the exact quantitative distribution of linguistic types in the languages of the world, but to show in what way two closely related areas of grammar (SELF-intensification and reflexivity) may be expressed, and how specific patterns of formal overlap may be explained in diachronic terms.

1.3. The structure of the article

Following these introductory remarks, Section 2 will be concerned with establishing a three-way opposition between (i) reflexives, (ii) "adnominal intensifiers," and (iii) "actor-oriented intensifiers." Section 3 presents a discussion of the crosslinguistic relatedness patterns in the domain of reflexivity and intensification. The findings presented there give rise to the assumption that in semantic terms, reflexivity is closely related to actor-oriented intensification and/or vice versa. Section 4 consequently provides a fine-grained analysis of actor-oriented intensification. In Section 5 the semantic relationship between actor-oriented intensifiers and reflexives is approached from the perspective of the hypothesis that what is common to actor-oriented intensification and reflexivity is a high degree of "involvement" in an event. While intuitively plausible, this argument turns out to be inconsistent on closer inspection and consequently needs to be refined. In Section 6, we argue that an explanation of the relationship between reflexives and actor-oriented intensifiers requires a closer look at the various synchronic and diachronic parameters linking both types of expressions. We make a distinction between two different scenarios that may give rise to a formal overlap of actor-oriented intensifiers and reflexives: (i) reflexives may form part of a strategy of intensification ("oblique reflexives"), and (ii) actor-oriented intensifiers may be used to reinforce semantically weak reflexivization strategies, consequently turning into markers of reflexivity themselves. Finally, Section 7 summarizes the results of our study and points out some corollaries of our argument for a general consideration of grammaticalization processes in the domain of reflexivity and intensification.

2. Reflexive pronouns and different types of SELF-intensifiers

2.1. Reflexive pronouns and adnominal intensifiers

If we agree on an informal definition of reflexivity as the co-indexation of two argument positions of a transitive predicate, the function of the self-forms in (4) can plausibly be called a reflexivizing one, whereas this characterization would be inadequate for the self-forms in (5) and (6), whose function is one of giving emphasis or intensification to a preceding nominal constituent.

(4) [John.sub.i] likes [himself.sub.i/*j].

(5) [The president himself] opened the meeting.

(6) John has [mown the lawn himself].

Self-forms such as those in (5) and (6), or their translational equivalents in other languages, have been referred to by a multitude of different labels while in the context of Engl. himself, herself, etc., the terms "emphatic reflexive" or "intensive reflexive" are probably most widespread (Leskosky 1972; Quirk et al. 1985; Kuno 1987). Given the formal identity of emphatic and intensifying self-forms in English, such terminological choices are comprehensible, but as a recent crosslinguistic study shows (Konig and Siemund 2005), the encoding of the reflexivizing function and this particular intensifying function by the same expression is representative of only about 45 percent of the languages of the world. Other languages of this type are Arabic (nafs-), Chinese (ziji), Finnish (itse-), and Turkish (kendi-).

Among the languages that do not encode these two functions by the same expression is German, where the reflexive marker sich is clearly differentiated from the intensifying expression selbst. The examples in (7)-(9) are the direct German translations of those in (4)-(6). It is not possible to exchange Germ. sich and selbst in these examples, that is, *Johann mag selbst, *Der Prasident sich eroffnete die Sitzung, and *Johann hat den Rasen sich gemaht are strictly ungrammatical. The same lexical distinction as in German is made in Abkhaz (xata- vs. -x[??]), Latin (ipse vs. se), Russian (sam vs. sebja), and Somali (naft- vs. is), among other languages.

(7) Johann mag sich / *selbst. John likes REFL SELF 'John likes himself.'

(8) [Der Prasident selbst / *sich] ertffnete die Sitzung. DET president SELF REFL opened the meeting 'The president himself opened the meeting.'

(9) Johann hat [den Rasen selbst / *sich gemaht]. John has DET lawn SELF REEL mown 'John has mown the lawn himself.'

For those languages that draw a formal distinction between reflexivizing and intensifying function, the expressions bearing the intensifying function have inter alia been analyzed as "reinforcing pronouns" (Penning 1875), "personal pronouns" (Poutsma 1916), "adjectives" (Visser 1970), "focus markers" or "focus particles" (Konig 1991; Ferro 1993; Sanchez 1994), and "scalar adverbs" (Primus 1992). The obvious variation in the categorization of these expressions has the following reasons: (i) the expressions do not have the same morphological, morphosyntactic, and distributional properties across different languages--for example, Swedish sjalv- and Spanish mismo/a inflect like adjectives, while German selbst does not inflect; (ii) analysts apply different criteria for the definition of word classes: distributional criteria often lead to a classification of SELF-intensifiers as focus particles--at least if those intensifiers do not inflect--while intensifiers are commonly categorized as nouns or adjectives (or pro-forms thereof) on the basis of morphological considerations; and (iii) even in single languages these expressions frequently straddle the line of established word class boundaries.

In spite of these differences, all of the labels listed above capture important grammatical properties of the expressions at issue: they carry emphatic stress, are sometimes used like pronouns, in some languages inflect like adjectives, in others behave like particles, are in association with a nominal constituent, interact with the referential properties of NPs, etc. What these expressions have in common semantically across different languages is their specific "intensifying" function: they evoke alternatives of a specific type which are paradigmatically opposed to the referent of the NP they relate to. For example, himself in the president himself has the effect of bringing alternative referents into the discussion which are different from, but in some way related to, the president, like the president's wife, the president's spokesman, or the vice-president (cf. Konig 1991; Siemund 2000, 2002; Eckardt 2001; Hole 2002; and Gast 2002 for more details). (4) Since NP-juxtaposed intensifying self-forms are not the primary topic of this article, we will not go into any more detail here, and the reader is referred to the relevant literature for more information. In keeping with Farr (1905), Cantrall (1973), Moravcsik (1972), Edmondson and Plank (1978), as well as the aforementioned publications, we will in the following simply refer to these expressions as "intensifiers" or "SELF-intensifiers," the latter term being motivated by the fact that in using the notion 'intensifier' we are impinging on the territory of degree adverbs like very. The small caps in the prefix SELF- are intended to make it clear that we are dealing with a generalization over expressions from different languages, rather than any particular lexical item. (5)

SELF-intensifiers always stand in a syntactic relation to some nominal (clause-mate) constituent. This relation has been described by notions as different as "coreference" or "binding" (Moravcsik 1972; Cantrall 1973), "apposition" (Hall 1965; Verheijen 1986), "theta identification" (Browning 1993), "complementation" (Eckardt 2001; Hole 2002; Gast 2002), (6) and "interaction" or "association with focus" (Konig 1991; Primus 1992; Ferro 1993; Siemund 2000), depending on the language analyzed, the perspective taken, and the theoretical framework adopted. Since it is not our goal here to contribute to this discussion, we will leave this syntactic relation unspecific and make no attempt at explicating it through our terminology. We will simply say that SELF-intensifiers "interact with" or "are associated with" the NP to whose interpretation they contribute. This NP will be called the "associated NP." (7)

2.2. Adnominal vs. actor-oriented intensifiers

As the English examples in (5) and (6) above, and also the German translations in (8) and (9), show, there are two positional variants of the SELF-intensifier, where one occurs in a juxtaposed position to the NP that it modifies and the other in nonjuxtaposed position embedded in some verbal projection. Attached to this syntactic difference there is a contrast in meaning which can be made explicit by finding suitable paraphrases to the intensifying self-forms (cf. [10] and [11]).

(10) [The president himself] opened the meeting. [approximately equal to] 'The president as opposed to other people related to him (his wife, spokesman, etc.) opened the meeting.'

(11) John has [painted the house himself]. [approximately equal to] 'John did not delegate the painting of the house to someone else.'

In the following we will refer to juxtaposed occurrences of self-forms as in (10) as "adnominal intensifiers," thus providing a simple syntactic characterization. Since such a straightforward syntactic description is not available for the nonjuxtaposed variants of SELF-intensifiers (the exact syntactic position of such intensifiers is an open issue; cf. Siemund 2000; Hole 2002; Gast 2002 for some discussion), a semantic characterization will be used instead. The label to be applied in the following for such occurrences of self-forms is "actor-oriented intensifier" because, as we will show in Section 3, the common denominator of these expressions is that they interact with an NP bearing the semantic (macro-)role of an actor, in the sense of Foley and Van Valin (1984) (cf. also Dowty 1991; Van Valin 1993). (8) Alternative labels that have been used for such uses of SELF-intensifiers include "predicate emphasizers" (Dirven 1973), "non-head-bound intensifiers" (Edmondson and Plank 1978), "adverbial reflexives" (Browning 1993), "agentive intensifiers" or "self-forms" (Kemmer 1995; Hole 2002), and "adverbial intensifiers" (Krnig 1991; Siemund 2000; Gast 2002).

To be sure, providing paraphrases as in (10) and (11)--in addition to the syntactic characterization "juxtaposed" vs. "nonjuxtaposed"--is not enough to motivate the distinction between adnominal and actor-oriented intensifiers. Let us therefore briefly review a few tests that have been used in the relevant literature to show that the two types of SELF-intensifiers are not merely positional variants, but differ both syntactically and semantically (we will use English data, but similar tests can be applied to other languages too). (9) First of all, adnominal intensifiers and actor-oriented intensifiers impose different constraints on the referential properties of the associated NP. Adnominal intensifiers require NPs whose referent is "capable of being identified," or should at least have "specific reference" (cf. Edmondson and Plank 1979:380 f.). This restriction does not apply to actor-oriented intensifiers, which can also combine with indefinite NPs with a generic interpretation (cf. [12] and [13] from Moravcsik 1972: 274):

(12) An engineer should know this himself.

(13) ?An engineer himself should know this.

Secondly, adnominal intensifiers can be adjoined to animate and inanimate NPs. Actor-oriented intensifiers, as their name suggests, are restricted to animate NPs bearing the macro-role of an actor (this point is emphasized by Hole 2002):

(14) The gardens are quite ugly, but the castle itself is wonderful.

(15) My dog opened the fridge door himself.

(16) ??The wind opened the fridge door itself.

Thirdly, adnominal and actor-oriented intensifiers behave differently in negative sentences. Focusing on the sloppy reading of the possessive his, the sentence in (17) says that only Bill's lawn was mown whereas Max's lawn was not. In sentence (18), by contrast, both Max's and Bill's lawn were mown, but Max delegated the job to someone else.

(17) Max himself did not mow his lawn, but his brother Bill did.

(18) Max did not mow his lawn himself, but his brother Bill did.

A last point showing that adnominal and actor-oriented intensifiers should be kept apart in a typological study is that many languages make a lexical distinction between both types of expressions, that is, (17) and (18) are translated using different lexical material. We will return to this point in Section 2.3.

In addition to the two uses of intensifying self-forms exemplified in (10) and (11), namely, adnominal and actor-oriented intensifiers, there is a third use type which can approximately be paraphrased by inclusive focus particles like also or too. An example is given in (19).

(19) How can Jim complain about Jack's snoring if he snores himself? [approximately equal to] '... if he snores, too.'

Like the actor-oriented intensifier in (11), himself in (19) occurs in a nonjuxtaposed position to the NP it interacts with. Such instances of self-forms have been referred to as "adverbial-inclusive" or "additive" SELF-intensifiers because they are part of a verbal projection and have an additive semantics like also. These intensifiers, which can be differentiated from actor-oriented intensifiers on distributional grounds (cf. Siemund 2000; Gast 2002), will play no role in the subsequent discussion (note that they are also very rare crosslinguistically and basically restricted to Germanic and Slavonic languages).

2.3. Formal identity vs. differentiation of adnominal and actor-oriented intensifiers

As has been illustrated using data from English and German, languages may differ as to whether or not SELF-intensifiers and reflexives are formally identical. The additional syntactic distinction between two use types of SELF-intensifiers introduced in the preceding section allows us to go one step further. We may now ask whether there are languages that differentiate between adnominal and actor-oriented intensifiers at a lexical level, too. And in fact, such languages do exist and are not even rare. Examples of languages that have special actor-oriented intensifiers include Koyra Chiini (cf. Section 3), Tetelcingo Nahuatl, and Wardaman. (10,11) Since all those languages also have specialized reflexive markers, the English self-forms in their three uses correspond to three different expressions in those languages. We will here use data from Tetelcingo Nahuatl for illustration, which is an almost extinct Uto-Aztecan language spoken in a suburb of Cuautla (Morelos, Mexico). In Tetelcingo Nahuatl, reflexive relations are indicated by a derivational prefix mo- (cf. [20]).

(20) tiktlasohtlas mo-tl[??]ka-ikni kyenami taha you.will.love.him 2.POSS-man-brother like you to-mo-tlasohtla. 2.SG-REFL-love 'You shall love your neighbor as yourself.' (Mark 12:31 [Wycliffe translation])

Adnominal intensification in Tetelcingo Nahuatl is expressed using the combination [sie + pronoun] in juxtaposition to the NP (cf. [21]). So far, the division of labor established by the intensifying and reflexivizing expressions in (20) and (21) more or less corresponds to the situation in German as illustrated in (7) and (8) above.

(21) pos [sie yaha David] kihtoa ipa inu libro de salmos ... well SELF he David he.says in DET book of psalms 'and David himself saith in the book of Psalms: ...' (Luke 20:42 [Wycliffe translation])

However, once we turn to the encoding of actor-oriented intensification, we see that yet another expression (sa -siel) is necessary because sie yaha cannot adjoin to a verbal projection. The actor-oriented intensifier sa -siel of Tetelcingo Nahuatl is not directly related to the adnominal intensifier in etymological terms (12) and differs from it in both form and syntactic class: sa ('only, merely') is an adverbial element, and -siel is a relational noun which takes possessor prefixes. Sa -siel can be used as a translational (near) equivalent of Engl. himself and Germ. selbst as illustrated in (6) and (9) above: (13)

(22) Sowa [okichihchik i-kal sa i-siel]. John he.built.it 3.POSS-house only 3.POSS-SELF.AO 'John built his house himself.'

In the following discussion we will precisely be interested in those cross-linguistic patterns of identity and differentiation between reflexives and the two types of SELF-intensifiers as exemplified by the data from English, German, and Tetelcingo Nahuatl above. These patterns are summarized in Table 1. Our study will show that not all logically possible patterns of formal identity or relatedness are equally attested. Interestingly, the cross-linguistic patterns of formal overlap seem to indicate that reflexivity is more closely related to actor-oriented than to adnominal intensification. While it is relatively difficult to find languages that use the same expression for the encoding of reflexivity and adnominal intensification, and that oppose such an expression to a formally distinct actor-oriented intensifier, languages which lump together actor-oriented intensification and reflexivity as opposed to adnominal intensification seem to be more widespread. Evidently, such a bias in the actual distribution of logical possibilities is in need of explanation, the more so as it contradicts our expectations and those expressed in the previous research.

3. Crosslinguistic relatedness patterns in the domain of SELF-intensifiers and reflexives

There are five logically possible relatedness patterns between adnominal intensifiers, actor-oriented intensifiers, and reflexive markers. (14) First, all three types of expressions may be encoded identically. This pattern, which can be found in English, is also attested in Turkish (pattern I):

(23) Turkish

a. reflexive

ben kendi-m-i sev-er-im.

I REFL- 1SG-ACC like-AOR-1SG

'I like myself.'

(Muinnever Ozkurt pers. comm.)

b. adnominal intensifier

miudur-un kendi-si biz-im-le konus-acak.

director-GENSELF-3SG.POSS We-GEN-with talk-FUT

'The director himself will talk to us.'

(Munnever Ozkurt pers. comm.)

c. actor-oriented intensifier

Olga cocuk-lar-in-1 kendi-si okutur.

Olga child-PL-3SG.POSS-ACC SELF-3SG.POSS teaches

'Olga teaches her children herself.'

(Munnever Ozkurt pers. comm.)

In a second type of language, exemplified by German above, adnominal and adverbial intensifiers are formally identical, and are both differentiated from reflexives. Another language of this type is Tzotzil (pattern II): (15)

(24) Tzotzil

a. reflexive lah yil s-ba ta nen li h-Sun-e.

he.saw 3.POSS-REFL in mirror DET MASC-John-CL

'John saw himself in the mirror.'

b. adnominal intensifier

ta stal s-tuk li preserent-e.

will.come 3.POSS-SELF DET president-CL

'The president himself will come.'

c. actor-oriented intensifier

li h-Sun-e [s-tuk la sva?an s-na].

DET MASC-John-CL 3.POSS-SELF he.built 3.voss-house

'John has built his house himself.'

Third, reflexives and actor-oriented intensifiers may be formally related, while adnominal intensification is expressed by a lexical item of its own. This situation can be found in Tarascan. There is a verbal suffix -kua.Ie which indicates either reflexivity or actor-oriented intensification, but never adnominal intensification. For the latter function, only the loan word misimu (< Spanish mismo) is available, which is widely used in both spoken and written language (as in folk tales and the Wycliffe bible translation) (pattern III).

(25) Tarascan (Charapan)

a. reflexive

Juan ese-kuare-sti.

John see-REFL-PAST.3.IND

'John looked at himself.'

b. adnominal intensifier

uihtsindikua hanosti [huramuti misimu].

yesterday came president SELF

'Yesterday, the president himself came.'

c. actor-oriented intensifier

Juan u-kuaae-sti kumainchikua.

John make-AO.SELF-PAST.3.IND house

'John built his house himself.'

In a fourth type of language (pattern IV) there are different lexical expressions for the three functions under consideration. This was illustrated for Tetelcingo Nahuatl above. A three-way opposition of this type can also be found in Koyra Chiini (also known as "Western Songhay"). There is a reflexive anaphor (or "light reflexive") ngu and a heavy reflexive marker homo ('head'), which are sometimes combined (cf. [26a]). Adnominal intensification is expressed using the NP-adjunct jaati(r) (cf. Arabic Oaat 'self'), and actor-oriented intensification by the adverbial element huneyno (cf. huna 'life'), which takes a pronominal possessor:

(26) Koyra Chiini

a. reflexive i si naaney ngi-ye bomo.

they IMPF.NEG trust REFL-3PL REFL

'They don't trust themselves.' (Heath 1999: 330)

b. adnominal intensifier

ay ta jaatir di cindi ka goy dooti.

I TOP SELF.ADN DEF remain INF work there

'I myself continue to work there.' (Heath 1999: 217)

c. actor-oriented intensifier

yee fatta ay huneyno.

1SG.SUBJ.IMPF exit lSG.POSS SELF.AO

'I will go out by myself.'

(Heath 1999: 219)

Finally, there are languages in which adnominal intensifiers and reflexive markers are identical or formally related, and are opposed to a special actor-oriented intensifier (pattern V). Mitla Zapotec is a case in point: (16)

(27) Mitla Zapotec (17)

a. reflexive

bawia lagahk loa.

I.saw SELF me

'I saw myself.'

b. adnominal intensifier

lagahk Juan kayuhn-ni rolihdz-ni.

SELF Juan is.building-3sG house-3sG.POSS

'Juan himself is building his house.'

c. actor-oriented intensifier

Juan ensilaani kayuhn-ni rolihdz-ni.

Juan SELF.AO is.building-3sG house-3sG.voss

'Juan is building his house himself.'

The five patterns of relatedness exemplified above are summarized in Table 2. The letters A, B, and C stand for the expressions associated with the meanings given at the top of each column. If two cells within a row are filled with identical letters, this means that the expressions associated with the relevant columns are formally related. The different types of relatedness patterns have been assigned Roman numbers (I-V). For the discussion to follow, types III and V are the most prominent ones and, therefore, highlighted. The last but one column indicates the number of languages in our sample exhibiting the relevant pattern (cf. also the table in Appendix 2):

Our study into the relatedness patterns between the three expressions at issue has led to an interesting finding: as can be gathered from Table 2, type III seems to be considerably more common than type V. In our 72-language sample, there are only three languages that can more or less safely be grouped under type V (Mitla Zapotec, Amharic, Malagasy), while as many as ten languages can be categorized as displaying the pattern associated with type III. We have found this pattern in Cubeo, Japanese, Kinyarwanda, Karo Batak, Lavukaleve, Mezquital Otoml, Tagalog, Tarascan, Tukang Besi, and Totonac.

Given the importance of pattern III for the argument to be made in the following, we would like to provide more illustration, this time using data from Japanese. The adnominal intensifier of Japanese is jishin (cf. [28a]), whereas actor-oriented intensification and reflexivization both involve the lexical item jibun. When used as a reflexive pronoun, jibun takes the appropriate object case marker, namely, accusative wo or dative ni; when used as an intensifier, it combines with the instrumental postposition de (cf. [28b] and [28c]). (18)

(28) Japanese

a. adnominal intensifier

Taro-jishin/*jibun kyouju-wo sonkeishiteiru.

Taro-JISHIN/JIBUN professor-Acc honour

'Taro himself honours the professor.'

(Ogawa 1998: 165)

b. reflexive

Taro-wa jibun-wo/*fishin-wo semeta.

Taro-TOP JIBUN-ACC/JISHIN-ACC criticized

'Taro criticized himself.'

(Ogawa 1998: 165)

c. actor-oriented intensifier

Taro-wa jibun-de/*fishin-de (19) kuruma-wo arau.

Taro-TOP JIBUN-INSTR/JISHIN-INSTR car-ACC wash

'Taro washes his car himself.'

(Ogawa 1998: 173)

The fact that pattern III is more widespread than pattern V is remarkable, given that most if not all previous studies assume or simply take it for granted that reflexives are more closely related to adnominal intensifiers than to actor-oriented intensifiers (e.g. Faltz 1985; Levinson 1987, 1991; Kemmer 1993; Krnig and Siemund 2000a; Heine 2003). As has been mentioned in Section 1, the historical development of English shows how reflexive markers can arise from adding an intensifier to an otherwise referentially unspecific object pronoun. We would not like to challenge the general correctness of this analysis for English as well as many other languages. However, we believe that it has (incorrectly) been generalized that the process of grammaticalizing reflexive markers from SELF-intensifiers always involves the adnominal intensifier. What our data suggest is that the role of the actor-oriented intensifier has been underestimated in previous studies. In Section 6, we will show that in languages like Mezquital Otomi it is clearly the actor-oriented intensifier that has developed into a marker of reflexivity, not the adnominal intensifier. Moreover, we will argue that the historical development does not necessarily proceed from SELF-intensifiers to reflexives. The reverse process is also a possible development, and strategies of actor-oriented intensification may be based on reflexive markers (e.g. Japanese jibun-de).

Naturally, explanations for these unexpected diachronic relationships can only be given if we understand how the meaning of actor-oriented intensifiers relates to reflexivity. Moreover, we need to know how this relationship differs from the one between adnominal intensifiers and reflexives. Successfully pursuing these problems, thus, necessitates some knowledge of the meaning of actor-oriented intensifiers, to which we will turn in the next section.

4. The meaning of actor-oriented intensifiers

4.1. Reference to alternative propositions

In a few words, the contribution made by actor-oriented intensifiers to the meaning of a sentence can be characterized as follows: actor-oriented intensifiers are used to emphasize that the action described by a sentence is performed by the subject referent, and not by some other person. For example, (29) states that the action of repairing the bicycle was carried out by John, and not by someone else. (20)

(29) John repaired the bicycle himself.

The notion of 'emphasis' must of course be made more explicit. Using concepts of focus semantics (cf. Rooth 1985, 1992, 1996), we define "emphasis" as "the assertion or mention of a given value against the background and to the exclusion of a set of possible alternative values." The function of himself in (29) can accordingly also be described like this: the actor-oriented intensifier is used to relate the proposition John repaired the bicycle to alternative propositions in which someone other than John repaired the bicycle in question, while the other major coordinates of that event remain unchanged (referential interpretation of the NPs involved, tense, etc.). Examples of propositions potentially contrasting with (29) are given in (30).

(30) a. John had the bicycle repaired by a bicycle mechanic.

b. John's brother repaired the bicycle for John.

We can describe the semantics of actor-oriented intensifiers in more general terms if we make use of the macro-role "actor," used by Foley and Van Valin (1984) and in Van Valin (1993), among others. The role "actor" is a generalization over the semantic roles associated with the external arguments of transitive and unergative intransitive predicates. In (29), John is the "actor" (more specifically, "agent") in the action described by the verb repair (or the VP repair the bicycle). (21) This role is taken by some individual other than John in (30a) and (30b) (a bicycle mechanic and John's brother, respectively). (22)

So far, sentences with actor-oriented intensifiers do not seem to differ from sentences with adnominal intensifiers or free focus constructions. A sentence like JOHN/John himsELF repaired the bicycle likewise seems to put emphasis on the agentive involvement of John. A second important aspect of the function of himself in (29) is that in both (29) and the contrasting propositions (30a) and (30b), there is a presupposition to the effect that John, while not being the actor in the event described, still stands in some other thematic relation to it. We might say, he is an "event-external causer" in (30a) and a beneficiary in (30b). Such (presupposed) alternative thematic relations holding between the referent of the NP interacting with the intensifier and the events described by the alternative propositions can be observed in all instances of actor-oriented intensifiers (cf. Siemund 2000: Ch. 8; Gast 2002: Ch. 5). The presuppositional status of those semantic relations is witnessed by the fact that they are stable under negation. While negating that John has repaired the bicycle in question, (31) still implies that he was responsible for, or interested in, the bicycle being repaired:

(31) John has not repaired the bicycle himself.

According to the two empirical generalizations made above, the function of actor-oriented intensifiers can be described as in (32).

(32) Actor-oriented intensifiers are used to relate a proposition 7t to a set of alternative propositions R = {[[rho].sub.1], [[rho].sub.2] ... [[rho].sub.n]} in such a way that:

a. in the alternative propositions [[rho].sub.i], the actor-role is assigned to some individual y other than the referent x of the associated NP, and

b. x has a different thematic role in the alternative propositions, for example, that of an external causer or beneficiary (cf. [30a] and [30b], respectively).

Let us consider two attested examples in order to provide further illustration for our analysis. In (33) and (34) alternative propositions are explicitly mentioned in the context.

(33) He announced that he would pick his ministers himself, rather than submit to the dictate of party leaders used to bartering cabinet posts for parliamentary support. [BNC CR7 2043]

(34) If you are scoring the tests yourself, rather than getting the computer to do it for you, then always do so at the same time of the day--otherwise there would be variability due to time-of-day effects upon your scoring ability. [BNC A75 814]

In (33), the sentence He picked his ministers himself is opposed to a proposition in which someone other than the referent of he performs the action of picking the ministers, namely, party leaders used to bartering cabinet posts for parliamentary support (cf. [32a]). The referent of he is conceived as being negatively affected in this case (SUBMITS to the dictate ..., cf. [32b]). Similarly, in (34) the proposition you score the tests is related to you get the computer to score the tests, and the thematic role assigned by the verb score to its subject is filled by some individual other than the referent of the NP interacting with the intensifier (you vs. the computer; cf. [32a]). In the alternative proposition, the referent of you has the role of a beneficiary (cf. [32b]).

As has been pointed out, we regard the kind of reference to alternatives that is established by actor-oriented intensifiers as a means of emphasis. Given that actor-oriented intensifiers oppose propositions to alternative propositions in which the actor-role is assigned to some individual other than the referent of the associated NP, they can be said to emphasize the agent status of a referent in a given predication. This is more or less in accordance with the analyses presented by Edmondson and Plank (1978), Browning (1993), and Hole (2002). (23)

4.2. Actor-oriented intensifiers in self-affecting contexts

We can finally turn to the aspect of actor-oriented intensification that will be particularly central to a consideration of their relationship to reflexives: often, actor-oriented intensifiers are used to emphasize that an action which has a (negative) effect on some referent has been carried out by that very referent. (35) is a relevant example. The intensifier in (35) relates the proposition John has ruined his career to a set of alternative propositions in which someone other than John has ruined John's career, thereby emphasizing John's agentive involvement. This is illustrated by (36).

(35) John has ruined his career himself.

(36) John's career has not been ruined by Bill; John has ruined his career himself.

John's being negatively affected in (35) is presupposed, and the new information is that he is also the agent of that event. The SELF-intensifier is thus responsible for the fact that the main focus of (35) seems to be on John's "double involvement." A similar semantic effect can be observed when an actor-oriented intensifier occurs in a reflexive predication. Here, the intensifier likewise emphasizes that some referent is not only passively, but also actively involved in an action. Unfortunately, English self-forms are hard to identify as intensifiers in such contexts, since they cannot formally be distinguished from reflexive pronouns. Consider the examples in (37) and (38) (cf. Dirven 1973; Konig and Gast 2002 for such "hybrid" self-forms). (24)

(37) John KILLeD himself.

(38) John killed himSELF.

(37) presupposes that something happened to John, and it could be used as an answer to the question What happened to John? By contrast, (38) is ambiguous. It could either be used as an answer to the question Whom did John kill? or to the question Who killed John? (cf. [39] and [40]).

(39) - Whom did John kill?

- He killed himSELF (it wasn't Bill whom he killed). (adnominal)

(40) - Who killed John?

- He killed himSELF (it wasn't Bill who killed him). (actor-oriented)

In both (39) and (40), himself performs a two-fold or "hybrid" function: on the one hand, it is responsible for the reflexive interpretation of the predication; on the other hand, it functions as an intensifier. Since English does not formally differentiate between SELF-intensifiers and reflexives, and since sequences such as *himself himself are ungrammatical (where the first himself would be interpreted as a reflexive marker, and the second as an intensifier), the interaction between actor-oriented intensifiers and reflexive-marking is blurred in these contexts. However, the contrast becomes morphosyntactically visible in languages which do distinguish formally between SELF-intensifiers and reflexives. For example, Russian differentiates between an invariant reflexive pronoun sebja and an intensifier sam, which inflects for number, gender and case. (25) The sentence John ruined himself can thus be translated as shown in (41)-(43), depending on aspects of information structure such as those outlined above.

(41) John pogubil sebja. (plain reflexive pronoun)

John mined REFL.

'John RuiNed himself.'

(42) John [VP pogubil [NP sam-ogo sebja]]. (sam-ogo: adn. intensifier)

John ruined SELF-GEN REFL

'John ruined himSELF (he didn't ruin someone else).'

(43) John [VP sam-?? sebja pogubil]. (sam: actor-oriented intensifier)

John SELF-NOM REFL ruined

'John ruined himSELF (he wasn't ruined by someone else).'

(41) is a "neutral" statement about John--a common VP-focus structure. In (42), it is under discussion that John ruined someone, probably someone else. The intensifier, which is clearly adnominal (cf. the genitive case triggered by NP-internal agreement with sebja), (26) is used to emphasize that John ruined no one other than himself. Finally, in (43) it is under discussion that John was ruined, and the sentence provides the piece of information that John himself was responsible for his (own) misfortune. The intensifier sam is actor-oriented, and it is associated semantically with the subject NP (cf. the nominative case). The use of actor-oriented intensifiers exemplified in (38) and (43) will play a central role in the following sections, in which we will finally tackle the question of why actor-oriented intensifiers and reflexives are formally related in many languages.

5. Agentivity and involvement: a semantic link between actor-oriented intensifiers and reflexives?

The central issue addressed in this article is the unexpected strong empirical tie-up between actor-oriented intensifiers and reflexives. If we consider the function of actor-oriented intensifiers as outlined in Section 4, this relationship turns out not to be too surprising anymore. We have argued that actor-oriented intensifiers are sometimes used to highlight--though not to express by themselves--the fact that an object referent is also the actor of the relevant event (cf. [40] and [43] above). Reflexive markers, on the other hand, may be defined as linguistic devices that are used to indicate that two arguments of a predicate have identical referents. This seems to point in the following direction: while a reflexive marker GRAMMATICALLY INDICATES that two arguments have identical referents, the actor-oriented intensifier PRAGMATICALLY EMPHASIZES that fact. Since it is commonly assumed in grammaticalization research that (pragmatic) emphasis often gives rise to the development of grammatical formatives, we seem to have a first clue as to why actor-oriented intensifiers and reflexives are formally related in many languages. A similar point has been made by Edmondson and Plank (1978). Edmondson and Plank define the function of actor-oriented intensifiers in terms of a high degree of agency or "involvement" (cf. Note 23). They characterize the interrelation between actor-oriented intensifiers and reflexives as follows (cf. also Browning 1993):

Finally, we take up the question of the linkage between intensifiers and reflexive pronouns. One can see the common ground in these two phenomena most clearly in the value of [himself.sub.2] [actor-oriented intensifiers, VG & PS]. This intensifier relates actions or states with the participants in such actions or states in a particular way. The most direct involvement of an individual in an action that requires two or more participants is to fill both roles [actor and undergoer, VG & PS] simultaneously. (Edmondson and Plank 1978:407 f.)

The concept of "agency" or "directness of involvement" seems to provide a plausible link between actor-oriented intensifiers and reflexives. However, there is a conceptual problem that considerably complicates the matter (but which, as we will argue in Section 6, falls into place if we change our perspective a bit): given that actor-oriented intensifiers do not interact with object NPs directly, how can they specify the referential interpretation of those NPs? Let us illustrate this problem using an example from Old English. As was pointed out in Section 2.3, Old English had a series of object pronouns which could be used both in contexts of local coreference of the subject referent and the object referent, and in contexts of locally disjoint reference. Therefore, (44) is ambiguous (cf. Krnig and Siemund 2000a: 59). If we add an adnominal intensifier to hine, however, the object pronoun is interpreted as referring back to Judas.

(44) [Judas.sub.i] aheng [hine.sub.i/j].

'Judas hanged him(self).'

(45) [Judas.sub.i] aheng [[hine self-ne].sub.i/*j].

'Judas hanged him*(self).'

As has repeatedly been pointed out, the reanalysis of an adnominal intensifier as a reflexive marker is relatively well studied and we will not discuss it any further here. In this article, we are interested in the question of how and why an ACTOR-ORIENTED intensifier may trigger reflexivity. Let us therefore consider what happens if we adjoin an actor-oriented intensifier to the VP in such languages. This is illustrated in (the hypothetical) (46).

(46) [Judas.sub.i] [[aheng [hine.sub.i/j]] self-[??.sub.i]].

Judas [[hanged him] SELF-NOM]

'Judasi hanged [him.sub.i/j] himself.'

The SELF-intensifier in (46) emphasizes that the one who hanged 'the referent of hine' was Judas, and no one other than Judas. Moreover, it is presupposed that Judas stood in some (nonagentive) thematic relation to the action of hanging 'the referent of hine', in addition to the asserted actor-relation. But nothing is implied as to who that referent is. The referential interpretation of hine is simply not at issue, since the intensifier self interacts only with the NP Judas. Consequently, it is difficult to see how self in (46) should have had a reflexivizing effect in Old English. Do we have to discard our argument, then?

We don't think so. The crucial point is that actor-oriented intensifiers and reflexives interact only in a specific type of context, and they interact in a way that is crucially different from the interaction between reflexive markers and adnominal intensifiers. In Section 6, we will argue that there are two scenarios in which a relationship between actor-oriented intensifiers and reflexives can be established: on the one hand, reflexives may form part of a construction conveying the semantics of an actor-oriented intensifier (cf. Section 6.1). On the other hand, actor-oriented intensifiers can function as reflexive markers in contexts in which the main verb has undergone some operation of diathetic change--typically, they are used in combination with a middle marker as expressions of reflexivity (of. Section 6.2). In this case, the actor-oriented intensifier is not used to disambiguate between two different referential interpretations of an object pronoun ("referential disambiguation"). Rather, it emphasizes the agentive involvement of its associated NP, thus disambiguating between a transitive/active reading and a detransitivized/middle reading of a predicate ("role disambiguation").

6. Two scenarios relating actor-oriented intensifiers and reflexives

6.1. Oblique reflexives expressing actor-oriented intensification

As has been pointed out in Section 2.3, certain languages have special actor-oriented intensifiers which differ formally from the adnominal intensifiers (e.g. Koyra Chiini, Tetelcingo Nahuatl). Typically, the adnominal intensifiers of such languages cannot occur at a distance from their associated NPs. We may now ask what happens if: (a) a language has an adnominal intensifier which does not adjoin to verbal projections, and (b) it does not have a specialized actor-oriented intensifier either. Such languages often resort to a multi-morphemic strategy in order to express the semantics of actor-oriented intensification: a reflexive pronoun is embedded in some prepositional or postpositional phrase, and the whole PP is adjoined to the VP. This construction is akin to the English expression by himself/herself/itself as illustrated in (47).

(47) [John.sub.i] [VP did it [PP by [himself.sub.i]]].

Strategies of intensification as exemplified in (47) seem to be particularly widespread among Afro-Asiatic languages (e.g. Amharic, Hebrew, and Oromo), (27) but can also be found in Indo-European (Italian da se) and a number of other languages (e.g. Japanese jibun-de). Examples from Italian, Japanese, and Arabic are provided in (48)-(50).

(48) Italian

Olga fa lezione aisuoibambini da se

Olga does lecture to her children PREP REFL

'Olga teaches her children herself.'

(49) Japanese

Taro-wa jibun-de kuruma-wo arau.

Taro-TOP REFL-INSTR car-ACC wash

'Taroo washes his car himself.' (Ogawa 1998: 173)

(50) Arabic

Ar-ra?is-u katab-a al-kitab-a bi

DET-president-NOM wrote-3SG DET-book-ACC INSTR

nafs-i-hi.

REFL-GEN-3SG

'The president wrote the book himself.' (28)

The strategy illustrated in (48)-(50) achieves the semantic effect of actor-oriented intensification in a way which can be derived compositionally. Recall from Section 4 that actor-oriented intensifiers are typically used to oppose a proposition to a set of alternative propositions in which the relevant action is carried out by someone other than the referent of the NP interacting with the intensifier. To illustrate with another example, consider (51).

(51) ... "relevant function" means the performance by a person of the functions of one of the following--(i) manufacturer, (ii) convertor, (iii) packer/filler, (iv) importer, (v) wholesaler, or (vi) seller, either himself or through an agent acting on his behalf... [www (29)]

In (51), the actor-oriented intensifier himself contrasts with the instrumental PP through an agent. The contrast is thus between two opposing "instruments," and it could be illustrated as in (52).

(52) the performance by a person ... either through [[REFL].sub.F] or through [[an AGENT].sub.F].

The term "instrumental," of course, is to be understood in a metonymic sense here. The actor-oriented intensifier does not really correspond to an instrumental constituent, that is, the referent does not "make use of himself/herself." Rather, the instrumental adjunct is metonymically reanalyzed as standing for the agent role. This metonymic effect is exploited by the languages listed above to convey the semantics of an actor-oriented intensifier.

Since in Italian, Japanese, and Arabic actor-oriented intensification is expressed by an adjunct with a thematic role that is not subcategorized for by the predicate--in Japanese and Arabic an instrument, and in Italian an ablative PP ('from')--we will call these constructions "oblique reflexives." The existence of such constructions provides a first explanation for the strong empirical tie-up between reflexives and actor-oriented intensifiers: a number of languages use oblique reflexives to express actor-oriented intensification without there being any kind of formal relatedness of adnominal intensifiers and reflexives. For example, the Japanese SELF-intensifier jishin does not function as a reflexive marker, but a close empirical relation can be established between the actor-oriented intensifier and the reflexive anaphor of Japanese insofar as jibun forms part of the strategy of actor-oriented intensification [REFL + instrumental]. Consequently, Japanese is of type III in Table 2. This correspondence pattern is illustrated in (53) ([approximately equal to] is to be read as "formally related").

(53) [jishin.sub.ADN] [??] [[jibun.sub.REFL] [??] [jibun-de.sub.SELF.AO]]

Actor-oriented intensifiers that can be analyzed as "oblique reflexives" can also be found in many Bantu languages. For example, Kinyarwanda has an adnominal intensifier ub-, which takes possessive suffixes and adjoins to the fight of an NP (cf. [54a]). Actor-oriented intensification, by contrast, is expressed by a combination of the verbal reflexive marker ii(y)- with an applicative suffix -er-. The applicative marker gives the object agreement marker the interpretation of an oblique argument, thus assigning to the entire construction a meaning roughly equivalent to Japanese jibun-de (cf. [54b]).

(54) Kinyarwanda

a. nda-shaka ku-vuga-na na [direkteri ub-we].

1SG.PRES-want INF-speak-soc with director SELF-3POSS

'I want to talk to the director himself.'

(C. Emkow pers. comm.)

b. devoirs a-ra-z-ii-kor-er-a.

homework 3SG-PRES-OBJ-REFL-do-APPL-IMPF

'He does his homework himself.'

(C. Emkow pers. comm.)

6.2. Actor-oriented intensifiers as reflexive markers

The second type of context where actor-oriented intensifiers pattern with reflexives to the exclusion of adnominal intensifiers is instantiated by strategies of reflexivization that are semantically, syntactically, and/ or phonologically "weak." "Semantic weakness" implies that a reflexive marker--typically a verbal reflexive, in the terminology of Faltz (1985)--is ambiguous between a middle reading and a referential reading. This kind of ambiguity can be observed in most of the languages that have monomorphemic anaphors (though not necessarily SE-anaphors, in the terminology of Reinhart and Reuland 1993) or verbal reflexives. Russian is a case in point. The suffix -sja functions as a (referential) reflexive marker only in combination with a specific class of verbs, which we could call "typically self-directed predicates" (cf. [55]; see also Haiman 1983, 1995; Konig and Siemund 2000a; Smith 2004). Such predicates denote actions that are commonly carried out on oneself, for example, wash or shave. In combination with predicates that are not typically self-directed -sja often indicates diathetic operations that are not directly related to reflexivity. For example, it can be found as an impersonal passive marker (cf. [56]), as an "emotion middle" in Kemmer's (1993) terms (cf. [57]), and in sentences with unspecific object reference (cf. [58]). (59) illustrates that kusat'/ukusit' 'to bite' can also be used with a referential object.

(55) Ivan mo-et-sja dvazhdyv den'. (grooming/reflexive)

Ivan wash-3SG-REFL/MID twice in day

'Ivan washes twice a day.'

(56) Kak eto dela-et-sja?

how that dO-3SG-REFL/MID

'How is that done?'

(57) On bespoko-it-sja. (emotion middle)

He disquiet-3SG-REFL/MID

'He is worried.'

(58) Sobaka kusa-et-sja. (unspecific object reference)

dog bite.IMPF-3 SG-REFL/MID

'The dog bites.'

(59) Sobaka ukusi-l-a chelovek-a. (transitive/active)

dog bite.PF-PAST-FEM.SG man-GEN.

'The dog bit the/a man.'

The polyfunctionality of -sja may lead to ambiguities in certain cases. For example, myt'sja can be interpreted as both 'wash (oneself)' and 'be washed'. Such ambiguities can also be observed in languages whose reflexive markers we would call (monomorphemic) "reflexive pronouns" in the terminology of Faltz (1985). German is such a language. Consider the ambiguous example in (60).

(60) Hans hat sich verletzt.

Hans has REEL injured

'John got hurt.' or 'John injured himself.'

(60) is ambiguous because it is not clear whether Hans deliberately injured himself, or whether he inadvertently got hurt. In the first case, sich is interpreted as a referential reflexive pronoun, while in the second case, it functions as a middle marker. The difference between the two occurrences of sich can be illustrated syntactically: only referential sich can be topicalized and focused. Therefore, (61) only has an agentive reading.

(61) Sich hat Hans verletzt.

'It was himself that John injured.'

In order to avoid the type of ambiguity illustrated in (60), speakers may use the expression constituting the subject matter of this article: an actor-oriented intensifier can be added to the VP. The intensifier will invariably select for the referential interpretation of the reflexive pronoun. If we add selbst to the VP [sich verletzt] in (60), sich can only be interpreted referentially. This is shown in (62) (wieder mal 'once again' has been added in order to favor the reading of [62] in which selbst is in construction with verletzt, not with sich). (30)

(62) Hans hat [VP sich (wiedermal) selbst verletzt].

Hans has REFL (once again) SELF injured

'(Once again), John injured himself.' (not: 'John got hurt.')

(62) allows only for a referential interpretation of sich, while the middle reading is blocked. Why should this be so? Obviously, it is emphasis on agentivity, expressed by the actor-oriented intensifier that is responsible for the univocal meaning of (62). (60) above is not ambiguous with regard to the reference of the undergoer, who is in both cases Hans; it is ambiguous only with regard to the question of who is the (intentional) agent. Either no agent is explicitly mentioned and the action may not even have been carried out intentionally (the middle reading, cf. Engl. John got hurt); or John is the agent, which induces the referential reading John injured himself. By using an actor-oriented intensifier, John's agentive involvement is emphasized, via reference to alternative propositions in which someone other than John is the agent. As a consequence, the nonagentive reading is blocked, and the sentence is understood in such a way that John actively injured himself.

In German, actor-oriented intensifiers are used in the context of reflexive-marking only if emphasis is desired, or if the context does not clearly select for one or the other reading. It thus depends on the assessment of the speaker whether s/he will use an actor-oriented intensifier or not. In other languages, the semantic bleaching of reflexive markers has advanced further than in German, and certain contexts strongly disfavor if not categorically disallow a referential reading of reflexive markers. This situation can be found in Mezquital Otomi. Otomi has a middle and reflexive prefix n-, which in (63) adjoins to the verb hno 'to kill'. For independent (phonological) reasons, the resulting form is hno ([hno]).

(63) Ra Suwa bi hno.

Ra Suwa bi n- hyo

DET Suwa 3PAST- REFL/MID- kill

'Somebody killed Suwa.' (marginally also: 'Suwa killed himself.')

(Priego Montfort 1989: 120)

Under certain circumstances, the verb form hno in (63) can be interpreted as a (referentially) reflexive predicate. However, there is an increasing tendency in contemporary Otomi to use the actor-oriented intensifier sehe in order to make it clear that the relevant sentence is to be interpreted reflexively. (64) provides an example which unambiguously describes an action of Suwa's killing himself.

(64) Ra Suwa bi hno sehe.

Ra Suwa [[bi n- hyo] sehe]

DET Suwa 3.PAST- REFL/MID- kill SELF.AO

'Suwa killed himself.' (but not: 'Suwa was killed.')

(Priego Montfort 1989: 120)

In accordance with what we would expect from the perspective of grammaticalization theory (cf. Lehmann 1995), the adverbial element sehe shows a strong tendency to cliticize to the verb, thus losing its tones and becoming an affix -se. This is illustrated in (65).

(65) Ra Suwa bi hnose.

Ra Suwa bi- n- hyo -se

DET Juan 3PAST- REFL/MID- kill -SELF.AO

'John killed himself.'

In view of the fact that the nasalizing prefix n- is phonologically relatively weak (insofar as it is not segmental), it comes as no surprise that it is completely lost in certain contexts. For example, it can hardly be heard if the verbal root begins with/h/and contains a nasal consonant, since co-articulation will always trigger nasalization of the adjacent vowel. In such contexts, the actor-oriented intensifier sehe is often the only marker of reflexivity (cf. [66]).

(66) Bi hyeni sehe.

3PAST cut SELF.AO/REFL

'He cut himself.'

(Priego Montfort 1989: 120)

Again, the reason why an actor-oriented intensifier is used as a reflexive marker can be related to its emphasizing the agent role of the NP it interacts with, and the function of sehe in the Otomi examples provided above parallels that of German selbst in (62). The difference is that the Otomi prefix n- is phonologically much weaker than German sich, and that therefore, reinforcement (in the sense of Lehmann 1995) through an actor-oriented intensifier is more necessary from the perspective of "striving for clarity" (Deutlichkeitsstreben; cf. vonder Gabelentz 1901: 256).

Like Japanese, Otomi is one of the languages in which reflexives and actor-oriented intensifiers are formally related, while adnominal intensifiers are formally distinct; note that Otomi does not have a specialized adnominal intensifier, and it uses a focus construction to express the semantics of adnominal intensification. These facts are illustrated in (67).

(67) [[FOCUS CONSTRUCTION].sub.SELF.ADN] [??] [(n-) ... [sehe.sub.REFL] [??] [sehe.sub.SELF.AO]]

A similar strategy of reflexive marking can be observed in many languages that allow conversion of verbal roots, that is, diathetic alternations without any morphological indicator. For example, Guirardello (1999) reports for Trumai that reflexivization is expressed using an element falapetsi, which qualifies as an actor-oriented intensifier, in our terminology:

(68) Trumai

ha falapetsi ka_in.

I do.alone/SELF.AO FOC/TENSE

'I made (it) by myself.'

(Guirardello 1999: 325)

The verb make 'bite' can be used intransitively, with an unspecific (potentially reflexive) object. In order to unambiguously indicate a reflexive reading, falapetsi can be added (the first person pronoun in the dative is optional):

(69) Trumai

ha falapetsi ha make (hai-tl).

I do.alone/SELV.AO I bite (1-DAT)

'I bit myself' (lit. 'I did [it] alone, I bit (me).')

If we consider the strategy of reflexive marking employed by Otomi and Trumai, and if we compare the resulting patterns of formal relatedness to the one characteristic of Japanese and Kinyarwanda, a crucial difference between languages of the first and second type can be observed. In Japanese and Kinyarwanda, a reflexive marker (jibun) forms part of a strategy of actor-oriented intensification. The lexical element jibun combines with the postposition de to form the "complex actor-oriented intensifier" jibun-de. In Otomi, by contrast, the actor-oriented intensifier sehe forms part of a strategy of reflexivization, which is made up of the prefix n-plus that intensifier.

7. Conclusions: parameters in the diachronic development from SELF-intensifiers to reflexives (and vice versa)

In this article we have considered three different processes that relate SELF-intensifiers to markers of reflexivity. First, adnominal intensifiers can be used to mark an underspecified object pronoun as reflexive (OE hine + self-). Second, we have shown that (pro)nominal reflexives can be used in prepositional phrases to express the semantics of an actor-oriented intensifier ("oblique reflexives," Jap. jibun + de). Finally, we have demonstrated that actor-oriented intensifiers can also be used to reinforce semantically weak strategies of reflexivization, that is, they are often used in combination with middle markers to express (referential) reflexivity (Otomi (n-) + sehe). These processes are summarized in (70).

(70) a. unspecified object pronoun + adnominal intensifier [right arrow] reflexive marker

b. oblique preposition + (pro)nominal reflexive [right arrow] actor-oriented intensifier

c. middle marker + actor-oriented intensifier [right arrow] reflexive marker

The three developments represented in (70) are fundamentally different for several reasons. First, there is a crucial difference between (a) and (c) on the one hand, and (b) on the other: while (a) and (c) represent the derivation of a strategy of reflexivization from a SELF-intensifier (plus some other grammatical device), (b) is a process by which an actor-oriented intensifier is derived from a reflexive marker. Furthermore, there is a crucial difference between (a) and (c): in (a), the adnominal intensifier is used to resolve an ambiguity that concerns the referential interpretation of an object pronoun. The actor-oriented intensifier in (c), by contrast, is used to clarify the thematic role, the kind or degree of involvement of a given referent in the event denoted by the verbal predicate. We therefore consider it appropriate to refer to the process shown in (70a) as one of "referential disambiguation" or simply "reference disambiguation," while the process illustrated in (70c) should best be called "role disambiguation." Figure 1 summarizes the different relatedness patterns that have been described in this study.

[FIGURE 1 OMITTED]

The argument made in this article has a number of repercussions on our assumptions concerning the synchronic and diachronic relationship between SELF-intensifiers and reflexives. First, it is often assumed or at least implied that this process is unidirectional in the sense that SELF-intensifiers may develop into reflexives, but not vice versa (e.g. Faltz 1985; Konig and Siemund 2000a; Schladt 2000; Heine 2003). We have shown that the opposite direction is also possible, and that in languages of the Japanese type, a reflexive marker can give rise to an expression of actor-oriented intensification if it combines with an appropriate case marker.

The second corollary of our article concerns common assumptions about the type of SELF-intensifier that is involved in grammaticalization processes leading to the development of reflexives. It is often taken for granted that it is always the adnominal intensifier which is central to that process (cf. Faltz 1985; Konig and Siemund 2000a), but we have shown that there are languages where it is clearly the actor-oriented intensifier that gives rise to the formation of new reflexive markers.

Finally, we hope that our argument has made it clear that even a diachronic relationship as well studied as the one between SELF-intensifiers and reflexives requires cautious observation and fine-grained analyses if one is to avoid the pitfalls of rashly made generalizations and resulting oversimplifications. While it is certainly tempting to assume a uniform process of grammaticalization which can be summarized in the formula "SELF-intensifier > reflexive," a closer look reveals that adnominal intensifiers, actor-oriented intensifiers and reflexives form a "triangular" semantic field with various diachronic and synchronic connections between all of the three categories. This is illustrated in Figure 2.

[FIGURE 2 OMITTED]</p> <pre> Appendix 1. Abbreviations ACC accusative AOR aorist CL clitic DAT dative DET

determiner FOC focus (marker) FUT future tense GEN genitive IMPF imperfect(ive) IND indicative INF infinitive INSTR instrumental MASC masculine MID middle voice NEG negation NOM nominative PAST

past tense PF perfective aspect POSS possessive PREP preposition REFL reflexive SELF SELF-intensifier SELF.ADN adnominal intensifier SELF.AO actor-oriented intensifier SG singular SOC sociative SUBJ subject TENSE

tense TOP topic </pre> <pre> Appendix 2. Sample Table A. The sample ID Language Family Area

INT.ADN 01 Abkhaz N-Caucasian N-Eurasia -xata

Caucasus 02 Albanian

IE N-Eurasia vete Albanian

Balkans 03 Amharic Afro-Asiatic Africa

ras- Semitic North 04 Arabic,

Afro-Asiatic Africa nafs- Egyptian

Semitic North 05 Armenian IE S-Eurasia

ink'-/ir- Armenian Asia Minor 06

Bagvalal N-Caucasian N-Eurasia PRO-da

Caucasus 07 Bambara Niger-Congo

Africa yere Mande West 08 Basque isolate N-Eurasia -eu-/bera-

W-Europe 09 Bengali

IE S-Eurasia nije- Indo-Iranian

S-Asia 10 Bulgarian IE N-Eurasia sam

Slavonic Balkans 11 Cantonese

Sino-Tibetan S-Eurasia jihgei

SE-Asia 12 Comanche Uto-Aztecan America

p[beginstrike North

throughuend

strikethrough] 13 Cubeo Tucanoan America

-wahari South 14 Danish

IE N-Eurasia selv

Germanic N-Europe 15 Finnish Finno-Ugric N-Eurasia itse- N-Europe 16 Fulani Niger-Congo Africa bee hoore

Atlantic- Central Congo Atlantic 17 Gaelic, Irish IE

N-Eurasia fein Celtic

W-Europe 18 Georgian S-Caucasian N-Eurasia tviton

Caucasus 19 Greek,

IE N-Eurasia idhios Modern Greek

Balkans 20 Greenlandic, Eskimo-Aleut America

nammineq West North 21 Gujarati

IE S-Eurasia pote-

Indo-Iranian S-Asia 22 Hausa Afro-Asiatic Africa

kai- Chadic West 23 Hebrew

Afro-Asiatic S-Eurasia -atsm-

Semitic Middle East 24 Hungarian Finno-Ugric

N-Eurasia mag- E-Europe 25 Indonesian, Austronesian Austr./Oc. sendiri Riau Indonesia 26 Japanese isolate N-Eurasia jishin

E-Asia 27 Kannada Dravidian S-Eurasia

taan S-Asia 28 Karo Batak Austronesian Austr./Oc. jine

Indonesia 29 Kinyarwanda Niger-Congo

Africa ub- Atlantic-Congo Central

Volta-Congo 30 Koasati Muskogean

America -bi:no North 31 Korean isolate N-Eurasia casin

E-Asia 32 Koyra Chiini Nilo-Saharan Africa jaati(r)

West 33 Latin IE N-Eurasia

ipse Italic S-Europe 34 Lavukaleve

East Papuan Austr./Oc. --

Solomon Islands 35 Lezgian N-Caucasian N-Eurasia wic-

Caucasus 36 Lingala Niger-Congo

Africa mckc Atlantic-Congo Central

Volta-Congo 37 Lithuanian IE

N-Eurasia pat- Baltic NE-Europe 38 Malagasy Austronesian Africa tena-

South 39 Malayalam Dravidian

S-Eurasia tanne S-Asia 40 Mandarin Sino-Tibetan S-Eurasia ziji

SE-Asia 41 Maori

Austronesian Austr./Oc. tino

New Zealand 42 Maricopa Hokan America

maatm North 43 Mixtec,

Otomangue America maa- Chalcatongo

Mesoamerica 44 Nahuatl, Uto-Aztecan

America -no?ma? Classical Mesoamerica 45 Ndyuka Creole (E) America seefi

South 46 Ngiti Nilo-Saharan

Africa -tirc Central 47 Oromo, Afro-Asiatic Africa ofii Afaan

Cushitic East 48 Otomi, Otomangue

America [FOC CONSTR] Mezquital

Mesoamerica 49 Persian IE S-Eurasia

xod- Indo-Iranian Middle East 50 Pitjantjatjara Australian Austr./Oc. -nku

Pama-Nyungan Australia 51 Podoko Afro-Asiatic

Africa ba mudare Chadic

Central 52 Quechua Quechuan America kiki-

(Huallaga) South 53 Shona Niger-Congo Africa -mene Atlantic-Congo

South Volta-Congo 54 Somali Afro-Asiatic

Africa naft- Cushitic East 55 Soninke Niger-Congo Africa yinme

Mande West 56 Tachelhit Afro-Asiatic

Africa nit Berber North 57 Tagalog Austronesian Austr./Oc. mismo

Philippines 58 Tamil Dravidian S-Eurasia ANPH-ee

S-Asia 59 Tarascan isolate America

misimu Mesoamerica 60

Totonac Totonacan America mani?

Mesoamerica 61 Tukang Besi Austronesian

Austr./Oc. ala?a Indonesia 62 Turkish Altaic S-Eurasia kendi-

Asia Minor 63 Tuvaluan Austronesian Austr./Oc. loa Mayalo-

Polynesia Polynesian 64 Tzotzil

Maya America -tuk

Mesoamerica 65 Usan Trans-New Austr./Oc. -onou Guinea Papua New Guinea 66 Waiwai

Carib America rma

South 67 Wardaman Australian Austr./Oc.

namaj- Gunwingguam Australia 68 Wolof

Niger-Congo Africa ci bopp-

Atlantic-Congo West Atlantic 69 Yimas

Sepik-Ramu Austr./Oc. panawt-

Papua New

Guinea 70 Yoruba Niger-Congo Africa funra- Atlantic-Congo West

Volta-Congo 71 Zapotec, Otomangue America

lagahk Mitla Mesoamerica 72

Zoque, Mixe-Zoque America -ne?k[begin

Copainala Mesoamerica strikethrough

Itrike

throughend] ID Language

INTRO REF. Pattern 01 Abkhaz

[xa-t.sup.o]' -xe II 02 Albanian vete

(vet)vetja I 03 Amharic ba-gazza

ras- V e[??][??]- 04 Arabic,

bi nafs- nafs- I Egyptian 05 Armenian

ink'-/ir- ir- I 06 Bagvalal PRO-da

PRO-da I 07 Bambara yere i II 08 Basque -eu-/bera- buru-

II 09 Bengali nije- nije- I 10 Bulgarian sam sebe II 11 Cantonese

jihgei jihgei I 12 Comanche p[beginstrike

VERBAL II throughuend

trikethrough] 13 Cubeo baxu- baxu-

III 14 Danish selv sig

II 15 Finnish itse- itse- I 16 Fulani bee hoore hoore I 17 Gaelic, Irish

(PRO) fein (PRO) fein I 18 Georgian tviton

tav- I 19 Greek, idhios eafto II Modern 20 Greenlandic, nammineq

immi- II West 21 Gujarati potaan-

pote- I 22 Hausa da kai- kai-

I 23 Hebrew be-atsm- -atsm- I 24 Hungarian mag- mag- I 25 Indonesian,

sendiri (sen)diri I Riau 26 Japanese

jibun-de jibun III 27 Kannada taan

VERBAL + I

taan 28 Karo Batak bana bana III 29

Kinyarwanda -ii-...-er- -ii- III

[-REFL-... -APPL-] 30 Koasati

-ni:li VERBAL IV 31 Korean jikjob

caki IV 32 Koyra Chiini huneyno bomo IV 33 Latin ipse se

II 34 Lavukaleve -muan -muan III 35 Lezgian wic- wic- I 36 Lingala

mckc -mi- II 37 Lithuanian pat- save 11 38 Malagasy ihany no

tena- V 'only FOC' 39 Malayalam

tanne taan I 40 Mandarin ziji

ziji I 41 Maori tino

-anoo II 42 Maricopa maatm mat-

II 43 Mixtec, ma?[??][??] maa- I Chalcatongo 44 Nahuatl, -no?ma? mo-

II Classical 45 Ndyuka seefi seefi

I 46 Ngiti -tirc nd[begin II strikethrough

iendstrike

through] 47 Oromo, ofii of

II Afaan 48 Otomi, sehe [MD] +

III Mezquital sehe 49 Persian

xod- xod- I 50 Pitjantjatjara -nku

-nku I 51 Podoko ba mudare

be PRO II 52 Quechua kiki- kiki-

I (Huallaga) 53 Shona -zvi- ... -it- ..-zvi-

II [-REFL- ...

-APPL] 54 Somali naft- is II 55 Soninke yinme du II 56 Tachelhit

nit agayu II 57 Tagalog sarili

sarili III 58 Tamil ANPH-ee

ANPH-ee I 59 Tarascan -kuaLe/-kuLi -kuaLe/

III -kuLi 60 Totonac

-a?kstu MID + III

-a?kstu 61 Tukang Besi karama- karama-

III 62 Turkish kendi- kendi- I 63 Tuvaluan (ei)loa loa I 64 Tzotzil

Auk -ba II 65 Usan -onou -onou(mi) I 66 Waiwai rma

VERBAL II 67 Wardaman ngajbang VERBAL

IV 68 Wolof ci bopp- bopp- I 69 Yimas panawt- panawt- I 70 Yoruba

funra- ara II 71 Zapotec, ensilaani lagahk V Mitla 72 Zoque,

-ne?k[begin -win II Copainala strikethrough

Istrikethrough end] </pre> <p>Received 20 January 2004 Revised version received Hamburg University 31 January 2005

Free University, Berlin Hamburg, University

References

Browning, M. A. (1993). Adverbial reflexives. NELS 23(1), 83-94.

Cantrall, W. (1973). Why I should relate own, emphatic reflexives, and intensive pronouns, my own self. CLS 9, 57-67.

Dirven, R. (1973). Emphatic and reflexive in English and Dutch. Leuvense Bijdragen 63, 288-299.

Dowty, D. R. (1991). Thematic proto-roles and argument selection. Language 67(3), 547-619.

Eckardt, Regine (2001). Reanalysing selbst. Natural Language Semantics 9(4), 371-412.

Edmondson, J.; and Plank, F. (1978). Great expectations: an intensive self analysis. Linguistics and Philosophy 2, 373-413.

Faltz, L. (1985). Reflexivity--A Study in Universal Syntax. New York: Garland.

Farr, J. M. (1905). Intensives and Reflexives in Anglo-Saxon and Early Middle English. Baltimore, MD: Furst.

Ferro, L. (1993). On "self" as a focus marker. In Proceedings of the Ninth Eastern States Conference on Linguistics, ESCOL 92, Michael Bernstein (ed.), 68-79. Ithaca, NY: Cornell.

Foley, W. A.; and Van Valin, R. D. (1984). Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press.

Gast, V. (2002). The grammar of identity--intensifiers and reflexives as expressions of an identity function. Unpublished doctoral dissertation, Free University of Berlin.

--; Hole, D.; Siemund, P.; and Topper, S. (2003). The typological database of intensifiers and reflexives. [http://www.philologle.fu-bedin.de/-gast/tdir/].

Guirardello, Rachel (1999). A reference grammar of Trumai. Doctoral dissertation, Rice University.

Haiman, J. (1983). Iconic and economic motivation. Language 59, 781-819.

--(1995). Grammatical signs of the divided self. In Discourse Grammar and Typology, W. Abraham, T. Givon, and S. Thompson (eds.), 312-334. Amsterdam: John Benjamins.

Hall, B. (1965). Subject and object in Modern English. Unpublished doctoral dissertation, MIT.

Heath, J. (1999). A Grammar of Koyra Chiini. Berlin and New York: Mouton de Gruyter.

Heine, B. (2003). On reflexive forms in Creoles. Lingua 115(3), 201-257.

Hole, D. (2002). Agentive selbst in German. In Sinn und Bedeutung VI--Proceedings of the 6th Annual Meeting of the Gesellschaft fur Semantik, G. Katz, S. Reinhard, and P. Reuter (eds.), 133-150. Osnabruck: Institute of Cognitive Sciences.

Jayaseelan, K. (1988). Emphatic reflexive x-self. CIEFL Working Papers in Linguistics 5, 1-20.

--(1997). Anaphors as pronouns. Studia Linguistica 51(2), 186-234.

Keenan, E. (2002). Explaining the creation of reflexive pronouns in English. In Studies in the History of English: A Millenial Perspective, D. Minkova and R. Stockwell (eds.), 325-355. Berlin and New York: Mouton de Gruyter.

Kemmer, S. (1993). The Middle Voice. Amsterdam: John Benjamins.

--(1995). Emphatic and reflexive self. Expectations, viewpoint and subjectivity. In Subjectivity and Subjectivization in Language, D. Stein and S. Wright (eds.), 55-82. Cambridge: Cambridge University Press.

Konig, E. (1991). The Meaning of Focus Particles--A Comparative Perspective. London: Routledge.

--; and Gast, V. (2002). Reflexive pronouns and other uses of self-forms in English. Zeitschrift fur Anglistik und Amerikanistik 50(3) (Special issue: Reflexives and intensifiers--the use of self-forms in English), 225-238.

--; and Siemund, P. (2000a). The development of complex reflexives and intensifiers in English. Diachronica XVII, 39-84.

--; and Siemund, P. (2000b). Intensifiers and reflexives--a typological perspective. In Reflexives." Forms and Functions, Z. Frajzyngier and T. Curl (eds.), 41-74. Amsterdam: John Benjamins.

--; and Siemund, P. (2005). Intensifiers and reflexives. In The Worm Atlas of Language Structures, B. Comrie, M. Dryer, M. Haspelmath, and D. Gil (eds.), 194-197. Berlin and New York: Mouton de Gruyter.

Kratzer, A. (1996). Severing the external argument from its verb. In Phrase Structure and the Lexicon, J. Rooryck and L. Zaring (eds.), 109-137. Dordrecht: Kluwer.

Krifka, M. (1998). The origins of telicity. In Events in Grammar, S. Rothstein (ed.), 197-235. Dordrecht: Kluwer.

Kuno, S. (1987). Functional Syntax: Anaphora, Discourse and Empathy. Chicago: Chicago University Press.

Lehmann, C. (1995). Thoughts on Grammaticalization. Munich: LINCOM.

Leskosky, R. J. (1972). Intensive reflexives. Studies in Linguistic Sciences 2(1), 42-65.

Levinson, S. (1987). Pragmatics and the grammar of anaphora. Journal of Linguistics 23, 379-434.

--(1991). Pragmatic reduction of the binding conditions revisited. Journal of Linguistics 27, 107-161.

Merlan, F. (1994). A Grammar of Wardaman: A Language of the Northern Territory of Australia. Berlin and New York: Mouton de Gruyter.

Moravcsik, Edith (1972). Some cross-linguistic generalizations about intensifier constructions. Papers from the 8th Annual Meeting of the Chicago Linguistics Society, 272-277.

Nichols, J. (1992). Linguistic Diversity in Space and Time. Chicago: University of Chicago Press.

Ogawa, A. (1998). Emphatische Reflexiva im Sprachvergleich: Deutsch-Japanisch. Sprachtypologie und Universalienforschung 51 (2), 164-178.

Penning, G. E. (1875). A history of the reflective pronouns in the English language. Unpublished doctoral dissertation, University of Leipzig.

Poutsma, H. (1916). A Grammar of Late Modern English, Part II. Amsterdam: John Benjamills.

Priego Montfort, Maria E. (1989). Gramatica del Otomi (Hnahnu) del Mezquital, Mexico. Unpublished doctoral dissertation, University of Bielefeld.

Primus, B. (1992). Selbst--variants of a scalar adverb in German. In Informationsstruktur und Grammatik, J. Jacobs (ed.), 54-88. Opladen: Westdeutscher Verlag.

Quirk, R.; Greenbaum, S.; Leech, G.; and Svartvik, J. (1985). A Comprehensive Grammar of the English Language. London: Longman.

Reinhart, T.; and Reuland, E. (1993). Reflexivity. Linguistic Inquiry 24, 657-720.

Rijkhoff, J.; Bakker, D.; Hengeveld, K.; and Kahrel, P. (1993). A method of language sampling. Studies in Language 17(1), 169-203.

Rooth, M. (1985). Association with focus. Unpublished doctoral dissertation, University of Massachusetts at Amherst.

--(1992). A theory of focus interpretation. Natural Language Semantics 1, 75-116.

--(1996). Focus. In The Handbook of Contemporary Semantic Theory, S. Lappin (ed.), 271-297. London: Blackwell.

Sanchez, L. (1994). On the interpretation of intensified DPs and emphatic pronouns. In Issues and Theory in Romance Linguistics, M. Mazzola (ed.), 479-492. Washington DC: Georgetown University Press.

Schladt, M. (2000). The typology and grammaticalization of reflexives. In Reflexives: Forms and Functions, Z. Frajzyngier and T. Curl (eds.), 103-124. Amsterdam: John Benjamins.

Siemund, Peter (2000). Intensifiers--A Comparison of English and German. London: Routledge.

--(2002). Selber essen macht fett: Zur Syntax und Semantik von adverbialen Intensifikatoren. In Grammatische Kategorien aus sprachhistorischer und typologischer Perspektive: Akten des 29. Linguisten-Seminars (Kyoto 2001), Japanische Gesellschaft for Germanistik (ed.), 187-204. Hamburg: Iudicium.

Smith, M. (2004). Light and heavy reflexives. Linguistics 42(3), 573-615.

Topper, S. (2002). A cross-linguistic analysis of adverbial exclusive intensifiers. MA thesis, Free University of Berlin.

van Gelderen, E. (2001). A History of English Reflexive Pronouns. Amsterdam and Philadelphia: John Benjamins.

Van Valin, R. D. (ed.) (1993). Advances in Role and Reference Grammar. Amsterdam: John Benjamins.

Verheijen, C. R. (1986). Reflexives and intensifiers in modern British English. Unpublished doctoral dissertation, University of Leiden.

Visser, F. T. (1970). An Historical Syntax of the English Language. Leiden: Brill.

von der Gabelentz, G. (1901). Die Sprachwissenschaft. Ihre A ufgaben, Methoden, und bisherigen Ergebnisse. Tubingen: Narr.

Williams, E. (1981). Argument structure and morphology. The Linguistic Review 1, 81-114.

Notes

(1.) Some of the arguments made in this paper have been inspired by work done by Stephan Topper (cf. Topper 2002), whose support is gratefully acknowledged. Moreover, we would like to thank Florian Haas, Bernd Heine, Daniel Hole, Suzanne Kemmer, Shigehiro Kokutani, Elena Maslova, and two anonymous referees of Linguistics for critical comments and helpful suggestions. All remaining errors and inaccuracies are our own. Correspondence address: Volker Gast, Institut fur Englische Philologie, Freie Universitat Berlin, Gosslerstr. 2-4, 14195 Berlin, Germany. E-mail: gast@zedat.fuberlin.de.

(2.) The project was carried out under the direction of Ekkehard Konig, funded by the Deutsche Forschungsgemeinschaft (Ko 497/5-1/4).

(3.) In our areal classifications we follow Nichols (1992). However, Asia Minor, the Middle East, and South/east Asia are here subsumed under "Southern Eurasia." Unlike Nichols, we regard Armenian as belonging to Asia Minor (and, thus, to Southern Eurasia).

(4.) Eckardt (2001), Hole (2002), and Gast (2002) argue that this semantic effect can be derived from the assumption that intensifiers denote the identity function in focus. Although this point of view is certainly compatible with the arguments made in this article, it will not be pursued any further. It should be mentioned, however, that an analysis that tries to derive the formal overlap between intensifiers and reflexives compositionally from the assumption that both are expressions of identity is not trivial and requires that one takes into account specific interactions between verb meanings and the referential interpretation of arguments (cf. Gast 2002 for discussion). Moreover, the analysis of intensifiers as expressions of an identity function is more difficult to motivate for the nonjuxtaposed uses of intensifiers (cf. Hole 2002 for an attempt).

(5.) For a typological study such as ours it would be particularly infelicitous to apply the term "emphatic reflexive," which is clearly motivated by the analysis of English, to other languages. The use of this term would be as inappropriate as using the term "wh-pronoun" for interrogative words in general. Nevertheless, this is not to deny that what we call intensifiers bears a close semantic relationship to reflexives, which in some languages manifests itself in formal identity. Note further that in some languages, our intensifiers are related to expressions of identity like Engl. same, Ital. stesso, Span. mismo, etc. Cf. Note 4 on the relationship of intensifiers to the notion of identity.

(6.) Semantically, this boils down to functional application.

(7.) As one referee points out, "antecedent" would also be a possible terminological choice; but this would presuppose that the syntactic relation between intensifiers and the associated NPs is of the same type as the one between antecedents and anaphors; that is, this terminology would force us to assume that SELF-intensifiers are expressions of category NP that are juxtaposed to and bound by the preceding NP. However, this point of view is not generally accepted, so we prefer to use the more general notion "associated NP."

(8.) An anonymous referee of Linguistics correctly points out that the nonjuxtaposed self-intensifiers look like subject complements ("secondary predicates"), but note that this analysis is semantically difficult to motivate. According to such an analysis, a sentence like John drove the car himself would come to be analyzed in the same way as John drove the car tired, for which the following semantic representation can be given: [??]e[DRIVE(John) (the car) (e) & TIRED(John)(e)]. However, translating the first sentence in a similar way yields a semantic representation of the form: [??]e[DRIVE(John)(the car)(e) & HIMSELF(John)(e)]. Gast (2002), who explores such an analysis in interaction with focus semantic effects, points out a number of problems.

(9.) For a more complete discussion of such tests, cf. Edmondson and Plank (1978), Siemund (2000), and Gast (2002).

(10.) Wardaman has a verbal reflexive marker. For adnominal intensification, there is an element -namaj, while actor-oriented intensification is expressed using the word ngajbang (cf. Merlan 1994:110 ft.).

(11.) To provide an example of a less "exotic" language, Korean can arguably be regarded as making lexical differentiations between the three types of expressions. The adnominal intensifier of Korean is casin, the actor oriented intensifier jikjob, and the reflexive anaphor caki. The distributional boundaries are not always clear-cut, however, especially in view of the fact that casin can also be used as a marker of reflexivity. For a survey of the most central relevant facts, cf. Gast et al. (2003).

(12.) There is an indirect etymological relationship insofar as both expressions are related to the numeral 'one' (se in Classical Nahuatl).

(13.) All examples from Nahuatl, Zapotec, and Otomi without an indication of the source have been collected during field work by one of the authors.

(14.) By formal relatedness, we mean morphological similarity based on etymological relatedness or synchronic derivability. Complete identity can be regarded as a special case of formal relatedness.

(15.) Note that stuk in (24c) is clearly not contained in the NP li hSune; the clitic -e marks the right boundary of this NP.

(16.) In certain contexts, ensilaani can occur in a reflexivizing function too, for example, in ensilaani badzohnni 's/he negated him/herself'. This suggests that Mitla Zapotec has properties of pattern III languages as well. However, reflexivizing ensilaani does not seem productive. Note that classifying Mitla Zapotec as a pattern III language would yield even more support to our analysis; but it would be misleading since adnominal intensifiers and reflexives are clearly formally related, too.

(17.) Double consonants indicate tense articulation.

(18.) Below, we will say that jibun is always a reflexive, and that the semantics of an actor-oriented intensifier are a function of the combination of the reflexive marker with an instrumental case marker.

(19.) Note that the combination jibun-jishin-de is possible, but *jishin-de alone is not.

(20.) In the discussion to follow, we will focus on those aspects of the meaning of intensifiers that are relevant to their interaction with reflexive markers. For more complete discussions of these issues, cf. Siemund (2000: Ch. 8), Hole (2002), and Gast (2002: Ch. 5), as well as relevant references cited there.

(21.) The assumption that verbs or VPs assign thematic roles to subjects has repeatedly been challenged (e.g. Williams 1981; Kratzer 1996). We will stick to the more traditional viewpoint here. An analysis of actor-oriented intensifiers in Kratzer's (1996) framework has been provided by Hole (2002).

(22.) This characterization appears to be incompatible with those instances of intensifiers that have sometimes been referred to as "anti-assistive" (Hole 2002; Gast 2002). In such uses, an expression of the form x was the agent of an event e x-self (where 'e' is a variable ranging over events) contrasts with something like x was the agent of e, and was helped by some y. We believe that even in those cases, the agent of the alternative proposition is different from that of the proposition containing the intensifier, since such comitative expressions can be regarded as denoting predications with plural agents, from a semantic point of view. For example, in Fred performed e with the help of John there is a (collective) plural agent {Fred, John} or [Fred + John]. In logical terms, that plural referent is clearly different from the individual Fred. Note that this assumption does not conflict with the so-called "cumulativity universal" ([lambda][P.sub.(et)] [for all]x[for all]y[[P(x) & P(y)] [right arrow] P(x + y)]]; cf. Kritka 1998), since this is only an implication from distributive to collective predications, but not the other way around.

(23.) For example, Edmondson and Plank (1978), who distinguish categorically between intensifiers combining with agentive verbs and those forming a constituent with verbs of perception (which they classify as "stative"), describe the function of actor-oriented intensifiers as follows: "The agency/experiencer distinction is a function of the predicate type; non-stative predicates like roll when intensified determine their subjects as most agentive; intensified statives as in

(i) The president heard the news himself.

exclude anyone but the most directly involved experiencer" (Edmondson and Plank 1978: 406).

(24.) Some authors have argued that all instances of self-forms should be analyzed as intensifiers, and that the object pronoun is phonologically empty (cf. Jayaseelan 1988, 1997; cf. also Gast 2002: Sect. 7.9 for a similar approach). We will stick to the more traditional viewpoint here that himself has two lexical entries: one which projects to NP (the reflexive pronoun), and one which is an adjunct (the intensifier).

(25.) We are indebted to Elena Maslova for information on Russian. Any inaccuracies are our own.

(26.) In Russian, animate object NPs receive genitive case.

(27.) For more information and relevant examples, cf. Gast et al. (2003).

(28.) M. Nekroumi (pers. comm.).

(29.) Source: http://www.hmso.gov.uk/si/si1997/70648--k.htm (03/19/03).

(30.) Daniel Hole has pointed out to us that the insertion of wieder real is not a safe way of doing away with the possibility that sich could be in construction with selbst, since sich might have been moved leftwards across wieder real ("scrambling"). All we want to show is that there is a reading of (62) in which selbst does not form a constituent with sich, and wieder mal has an illustrative character here. The (assumed) constituent structure of (62) is indicated by the brackets.

Table 1. Relatedness patterns in English, German, and
Tetelcingo Nahuatl

 Adnominal Actor-oriented
 Reflexive intensifier intensifier

English self-forms self-forms self-forms
German sich selbst selbst
Tetelcingo Nahuatl mo- sie + pronoun sa -siel

Table 2. Relatedness patterns of SELF-intensifiers and reflexives

 n in
 SFLF.ADN SELF.AO REEL sample Examples

 I A A A 30 English, Mandarin

 II A A B 25 German, Tzotzil

[right III A B B 10 Tarascan, Japanese
arrow]

 IV A B C 4 Tetelcingo
 Nahuatl,
 Koyra Chiini

[right V A B A 3 Mitla Zapotec,
arrow] Amharic