Studies in African Linguistics Volume 23, Number 3, 1992-1994 DOGON PRONOMINAL SYSTEMS: THEIR NATURE AND EVOLUTION Christopher Culy The University of Iowa Koungarma Kodio Patrice Togo D.N.A.F.L.A., Bamako, Mali The Dogon language family has received little attention in the linguistics literature to date. In this paper we examine the binding properties of the pronominal systems of three Dogon languages, Donno S:>, T:>r:> S:>, and Togo Ka. We also posit the pronominal system of their common ancestor, and the changes from the common ancestor to the contemporary languages. In doing so, we find two ways in which languages can lose logophoricity: (1) the logophoric pronoun becomes a subject oriented reflexive, and (2) the logophoric pronoun is lost without any reflex. The Dogon languages thus give us insight into the nature of pronominal systems and how they evolve. 1. Introduction* The Dogon language family shows an immense amount of formal variation, in phonology, morphology, and syntax. In this paper we will examine the binding properties of the pronominal systems of three Dogon languages, Donno S:) (DS), T:)f;) S;) (TS), and Togo K§ (TK). We will see a variety of properties, many of which are interesting in their own right. However, perhaps the most interesting aspect is that TK seems to be in the process of changing from one type of system * We would like to thank Joan Bresnan, Douro Etienne Kassogue, Peter Sells, Tom Wasow and an anonymous reviewer for helpful comments. Of course, we remain responsible for any errors or omissions. 316 Studies in African Linguistics 23(3), 1992-1994 to another. This change provides evidence for the logophoric hierarchy of Culy [1991]. The last part of the paper will discuss the relationship among these pronominal systems. In particular, we will argue for certain properties that can be ascribed to their common ancestor. In doing so, we show two ways in which logophoric pronouns can change, a subject which has received little attention to date. 1 2. Background 2.1 Dogon. While there are many Dogon languages,2 with variation so extreme that some are not even mutually intelligible, the three languages that we will be looking at-DS, TS, and TK3-seem to be fairly closely related. DS and TS are geographically proximate, while TK is separated from the other two by the Bandiagara Escarpment. TK is the least conservative variety morphologically, while DS is the most conservative variety morphologically. However, DS and TS have been more influenced by Fula than has TK, as evidenced by a number of lexical borrowings. DS has two major sub-varieties, one spoken in and around the town of Bandiagara, and the other spoken in the smaller villages in the surrounding area. The town variety shows more influence, both lexically and phonologically, from Fula than does the outlying variety. Both of these varieties are almost completely mutually intelligible. 2.2. Binding properties. Dalrymple [1990] argues that the binding properties of any pronoun can be described in terms of combinations of five primitive pro perties, as shown in (1).4 (1) Primitive binding properties a. Subject Binding/Disjointness ([±sbD The pronoun must be bound to (respectively, disjoint from) a subject. 1 Though there has been some work on how logophoric pronouns arise, e.g. by Hyman [1979], [1981] Voorhoeve [1980], Frajzyngier [1985]. 2 See Bertho [1953] and Calame-Griaule [1956] for some discussion of the different varieties. 3 Data on DS are from fieldwork conducted by the first author in 1987-88, unless otherwise noted. Data on TS are from the second author, and data on TK are from the third author, both of whom are native speakers. Descriptions of other aspects of these languages can found in Kervran and Prost [1969, 1986], Calame-Griaule [1968], and Prost [1969], respectively. 4 Dalrymple does not use features to refer to the properties. We do so for typographical con venience. Dalrymple also provides a formalization of the binding properties, which we will not do here, for the sake of brevity. Finally, the Root S Binding property is not relevant for Dogon, so we will say nothing more about it. Dogan pronominal systems: their nature and evolution 317 b. Coargument Binding/Disjointness ([±co]) The pronoun must be bound to (respectively, disjoint from) an argument of the same predicate the pronoun is an argument of. c. Minimal Complete Nucleus Binding The pronoun must be bound by an argument in the smallest predicate-argu ment structure containing it and a subject of a tensed clause or possessor distinct from it (i.e., it must be bound in its minimal complete nucleus). i. Non-subjective ([+mcn]) The pronoun must not itself be a subject or possessor. ii. Subjective ([+mcnS]) The pronoun may itself be a subject or possessor. d. Minimal Finite Domain Binding ([ +mfd]) The pronoun must be bound in the smallest finite clause it is contained in. e. Root S Binding ([ +rt]) The pronoun must be bound somewhere in the sentence. English provides a simple illustration of these properties. English reflexive pronouns are generally [+mcn], since they must be bound in the smallest pre dicate-argument structure containing it and a subject or possessor distinct from it, and they cannot occur as subjects of tensed clauses or possessors.5 Some examples are given in (2). Notice in (2c) that even though 'herself' is an argument of 'picture', there is no possessor, so 'herself' is free to be bound by the subject. (2) English reflexives a. Pati talked to Leej about herselfi/j b. Pati said that Leej talked about herselfj/*i c. Pati liked the picture of herselfi d. Pati didn't like Leej's picture of herselfj/*i On the other hand, English personal pronouns are [-co], since they can't be bound by an argument of the same predicate. Some examples are given in (3). Note that when there is no subject or possessor of the predicate, then either the reflexive pronoun or the personal pronoun is allowed «2c) vs (3c». (3) English personal pronouns a. Pati talked to Leej about herk/*i/*j b. Pati said that Leej talked about heri/k/*j c. Pati liked the picture of heri/j d. Pati didn't like Leej's picture of heri/k/*j 5 But see Pollard and Sag [1994] for detailed discussion of some exceptional cases. 318 Studies in African Linguistics 23 (3), 1992 -1994 One final comment is in order before we move on to Dogon, and that is that Dalrymple deliberately did not discuss logophoric pronouns, which are an important part of Dogon. Logophoric pronouns are pronouns which must have as their antecedents the person whose thoughts, words, or state of mind are being reported. We will not give a detailed analysis of logophoricity,6 but will use [+log] to refer to logophoric pronouns and [-log] to refer to pronouns that cannot have a logophoric antecedent. 3. Donno (DS) S~ 3.1 Personal Pronouns. The personal pronouns are the easiest to describe. DS has first, second, and third person pronouns, distinct in singular and plural. The binding properties of the personal pronouns are also relatively easy to describe. Non-third person pronouns must be disjoint from coarguments (4a-b), but they can corefer with non-coarguments (4c). Note that in (4b) the m' is syntactically the possessor of sa 'word', but semantically the argument of 'talk' in just the same way that English myself is the object of the preposition about but semantically the argument of 'talk'. In both cases, they are coarguments of the subject. (4) Non-third person pronouns a. As direct object (disjoint from subject)7 *mi miii tcbcJaa bem IS IS-OBJ hit PST-IS 'I hit myself' b. As oblique (disjoint from indirect object) *Omar mi Je sa m m5 sayyaa be ross Omar IS with word IS talked PST 'Omar talked to me about myself c. As possessor (coreferential with subject) mi yaana m' m5 waa bem IS wife IS POSS saw PST -1 S 'I saw my wife' 6 See Sells [1987] for one such treatment. 7 Numbers used in the gloss of examples correspond to first, second or third person, hence, IS indicates first person singular, 3P third person plural. The following abbreviations are also used: AG Agentive OBJ Object PST Past AUX Auxiliary verb PI Plural REFL Reflexive DF Deftnite (detenniner) POSS Possessive SUB] Subject LOC Locative PRP Present participle LOG Logopboric pronoun PSP Past participle Dogon pronominal systems: their nature and evolution 319 Third person pronouns are similar to non-third person pronouns, as the examples in (5) show. (5) Non-third person pronouns a. As direct object (disjoint from subject) wo woil tebcJaa be 3S 3S-0BI hit PST 'Shei hit herj/*herselfi' b. As oblique (disjoint from indirect object) Anta wo Je so wo m5 soyyaa be Anta 3S with word 3S POSS talked PST 'Antai talked to himj about himijk/*himselfj' c. As possessor (coreferential with subject) Omar yaana wo m5 waa be 3S wife IS POSS saw PST 'Omari saw hisi/j wife' The one difference between third person and non-third person pronouns is their properties in logophoric contexts. Third person personal pronouns cannot have as their antecedents the person whose thoughts, words, or state of mind are being reported (6a); non-third person pronouns can (6b). (6) Personal pronouns in a logophoric environment a. Third person (disjoint from logophoric subject) Anta Omar woil we gi Anta Omar 3S-0BI saw said 'Antai said that Omar saw herj/*i' b. Non-third person mi Omar miil we gim Mi Omar IS-OBI saw said-IS 'I said that Omar saw me' We can summarize the properties of personal pronouns in DS as in (7). (7) Binding properties of personal pronouns in DS Non-third person [-co] Third person [-co, -log] 320 Studies in African Linguistics 23(3),1992-1994 3.2 Reflexives 3.2.1 The description. OS has two reflexive elements, an invariant reflexive pronoun sama, and a periphrastic reflexive, formed from the word for 'head' with a pronominal possessor which agrees in person and number with its antecedent. 8 Some examples are given in (8). (8) Reflexives in Oonno S:J a. Simple third person Anta sarna tcbeJaa be Anta REFL hit PST , Anta hit herself' b. Simple first person mi sarna tcbeJaa bem IS REFL hit PST-IS 'I hit myself' c. Periphrastic third person Anta ku wo ma tcbeJaa be Anta head 3S POSS hit PST , Anta hit herself' d. Periphrastic first person mi ku m' ma tcbeJaa bem IS head IS POSS hit PST-S 'I hit myself' Of these two strategies, it seems that the reflexive pronoun is the older one. There are three reasons for thinking this. The first reason is that the pronoun occurs only in the more conservative outlying variety of OS. The periphrastic strategy is found in both the town variety of OS, and in TS. The second reason for thinking that the pronoun is older is that Fula has the same compound strategy, as the example in (9) shows. (9) Fula reflexive Anta Ii'ii hoore mum Anta hit head 3S , Anta hit herself' 8 See Culy [1993] for detailed discussion of periphrastic reflexives. Dogon pronominal systems: their nature and evolution 321 Recall that TS and OS have had more contact with Fula than TK has, and fur thermore that the town variety of OS has had more contact with Fula than the outlying variety, especially with Bandiagara acting as the capital of part of the Fula-speaking Toucouleur empire in the 19th century. Finally, speakers of the town variety are much more likely to speak Fula as a second language than speakers of the outlying variety.9 The third reason for thinking that the reflexive strategy is older is that the pronoun, but not the periphrastic reflexive, has a cognate in TK, as we will see. Thus, it seems plausible that OS has borrowed the compound strategy from Fula, with the borrowing having completely replaced the indigenous reflexive pronoun in the town variety and coexisting with the pronoun in the outlying variety. 3.2.2 The binding properties. The binding properties of the periphrastic reflexive are a little easier to determine than those of the reflexive pronoun. In (10) we see that the periphrastic reflexive cannot be bound outside of its minimal complete nucleus, while (11) shows that it can have a non-subject antecedent. Recall from the discussion of (4b) that the periphrastic reflexive is the syntactic possessor of sa 'word' but the semantic argument of 'talk,' and hence in the same minimal complete nucleus as Anta, which is also an argument of the verb. (10) Periphrastic reflexive bound in its minimal complete nucleus Omar Anta ku wo m5 waa be 191 wa Omar Anta head 3S POSS saw PST know AUX 'Omari knows that Antaj saw herselfjl*himselfi' (11) Periphrastic reflexive with non-subject antecedent Anta Je ku wo m5 sa sayyaa bem Anta with head 3S POSS word talked PST-IS 'I talked with Anta about herself' To rule out [+mfd], we just need to note the ungrammaticality of (12), which contains only one finite verb. 9 Who in tum are more likely to speak Bambara as a second language than speakers of the town variety. 322 Studies in African Linguistics 23(3),1992-1994 (12) Periphrastic reflexive is not [+mfd] *Omar [ku be m5 le sa sayyezen' giaa] Omar head 3P POSS with word talk-PRP say-PSP ben bondaa be 3S-obj call-PSP PST ('Omar called them in order to talk with them') Although this evidence is all consistent with the periphrastic reflexive being [+co], the example in (13) shows that the [+co] account is untenable. The post position ne 'at' has its own predicate, yet the periphrastic reflexive can be bound by the subject, which is not a coargument. Thus, the periphrastic reflexive is [+mcn]: it must have an antecedent in the minimal complete nucleus containing it, and it cannot itself be a subject or possessor. (13) Periphrastic reflexive not bound by coargument [Kervran 1982:489] Golu sa-ga sayyew dyaa yeJaa Golou event-DF talking take come ku wo m5 ne wo daaa, yandalan kanaa yaga sayyi head 3S POSS at 3S arrived go over did other talked 'In relating the event, when he arrived at [the part about] himself, Golou passed over it and talked about something else' The binding properties of the simple reflexive are a little harder to determine. First of all, sarna must be bound in its clause, as seen in (14). (14) Simple reflexive bound in its clause a. Monoclausal Anta sama tebeJaa be Anta REFL hit PST , Anta hit herself' b. Biclausal Omar Anta sama waa be igi wa Omar Anta REFL saw PST know AUX 'Omarj knows that Antaj saw herselfj*himselfj' Dogan pronominal systems: their nature and evolution 323 That this restriction is not the [+mfd] restriction can be seen by the clause boundedness of the reflexive in (15), which contains only one finite verb, and in which 'Omar' is the understood subject of 'talk'. (15) Simple reflexive is not [+mfd] (cf. (12» Ornar sarna Ie sa sayyczen' giaa ben bondaa be Omar REFL with word talk-PRP say-PSP 3S-0BJ call-PSP PST 'Omarj called themj in order to talk with himJ*themj' What makes it difficult to determine the binding properties of sarna is that we have no examples of it occuring as the object of a simple postposition, 10 and it does not occur in the 'talk to X about Y' construction that we used earlier. The reasons sarna does not occur in this construction are slightly involved. First of all, in (16) we see that sarna cannot occur as a possessor. However, this does not seem to be a consequence of the binding properties of sarna. Rather, sarn a seems not to be able to occur with any overt case marking, which the possessor is, as shown by Embree [1993]. In particular, other pronouns must occur with Object case marking (e.g., ben in (15», even though it is optional for certain other NPs. However, sarna does not occur with Object case marking, as seen in (14b). Since sarna cannot occur with case marking, it cannot occur as a postnominal possessor of sa in (16). (16) Simple reflexive as possessor a. With semantically non-empty common noun *wo yaana sarna rna waa be 3S woman REFL POSS saw PST 'Hej saw hisj wife' b. With semantically empty sa *rni Anta Ie sa sarna rna sayyaa bern IS Anta with word REFL POSS talked PST-IS 'I talked with Anta about herself' Furthermore, pronouns cannot occur in the prenominal position with sa as seen in (17a) (cf. (11». This rules out sarna from occuring in this position, as in (I7b). 10 Most postpositions are complex, consisting of the simple postposition nc (cf. (13», combined with a common noun, the "object" being realized as the possessor of the common noun. 324 Studies in African Linguistics 23(3), 1992-1994 (17) Pronouns in the prenominal position with sa a. PersonaVlNon-reflexives *mi Omar Ie u sa sayyaa bem IS Omar with 2S word talked PST-IS 'I talked with Omar about you' b. Simple Reflexive *mi Omar Ie sarna sa sayyaa bem IS Omar with REFL word talked PST-IS 'I talked with Omar about himself' Given the the lack of the appropriate structures, we cannot tell if [+sb] is relevant for sama. Given that [+sb] does not playa role in any of the other pronouns or reflexives, we will assume that is not in fact relevant for sama. However, we still do not know whether sarna is [+co] or [+mcn]. Finally, third person reflexives are not used in logophoric environments to refer to the logophoric antecedent, as seen in (18a).11 Of course, they can refer to a non-Iogophoric antecedent in the same clause (18b), and non-third person periphrastic reflexives can have logophoric antecedents in the same clause (I8c). (18) Reflexives in logophoric environments a. Ungrammatical third person periphrastic reflexive *Omar inyemc ku wo ma samaa bem gi ross Omar LOG head 3S congratulated AUX-IS said 'Omarj said that hej congratulated himselfj' b. Grammatical third person periphrastic reflexive Omar Anta ku wo ma sarnaa be gi ross Omar Anta head 3S congratulated AUX said 'Omarj said that Antaj congratulated herselfj' c. Grammatical non-third person periphrastic reflexive mi ma ku m' ma sarnaa be gim ross IS IS-SUBJ head IS congratulated AUX said-IS 'I said that 1 congratulated myself' We can summarize the properties of the reflexives in DS as in (19). 11 This isn't quite true, as we'll see in the next section.