Yoruba: Anaphora Sketch

Yoruba Anaphora Sketch

Olúṣẹ̀yẹ Adéṣọlá and Ken Safir

1. Introduction

Yoruba is an interesting example of a notably unarticulated anaphora system, in which almost all local anaphoric readings are handled by a single form, and in which all non-local anaphora is achieved by otherwise independent pronouns. Since the anaphoric strategy includes a pronoun as a constituent part, we present the pronouns first.

2. Pronouns

Only the person and number features are marked on Yoruba pronouns (i.e., they are not marked for gender or animacy), but the forms of the pronouns vary depending on their case. Yoruba pronouns are divided into two classes, strong forms and weak ones. The strong forms are analyzed as nouns (see Awobuluyi 1978) while the weak forms are analyzed as clitics (see Akinlabi and Liberman, 2000). The following table shows the pronouns that are attested in the language:

	Strong Pronouns		Weak pronouns
	NOM/ACC	GEN	NOM	ACC	GEN
1^st Singular	Èmi	Èmi	mo	mí	Mi
2^nd Singular	Ìwọ	Ìrẹ	o	ọ́/ẹ́	Rẹ
3^rd Singular	Òun	ire`̣/oun	ó	un / V	Rẹ̀
1^st Plural	Àwa	Àwa	a	wá	Wa
2^nd Plural	Ẹ̀yin	Ẹ̀yin	ẹ	yín	Yín
3^rd Plural	Àwọn	Àwọn	wọ́n	wọ́n	Wọn

Note: V = vowel, NOM = nominative, ACC = accusative, GEN = genitive

3. Anaphoric Strategies

Apart from the use of otherwise independent pronouns to express anaphoric relations, there are only two other morphological strategies for achieving anaphoric readings. One, the null pronoun strategy, has a very limited distribution determined by the small class of verbs that are compatible with it. The second strategy, the ara-X strategy, is very widely used, lexically unrestricted, though subject to syntactic locality restrictions. We consider each strategy very briefly below.

3.1 Strategy A or the Ara-X strategy.

These forms are derived by combining ara ‘body’ with a genitive pronoun.[1] Thus, Yoruba has anaphors such as: ara à mi ‘myself’, ara à rẹ ‘yourself’, ara a rẹ̀ ‘himself’, ara a wa ‘ourselves’, ara a yín ‘yourselves’ and ara a wọn ‘themselves’. The last form, ara a wọn is the same morphologically with the reciprocal – ara a wọn ‘each other’. The form ara does not vary for person or number and always permits a literal reading, 'X's body', in the right context.

As noted above, this is by far the more productive of the strictly anaphoric strategies. In this strategy, the anaphor must be a co-argument with its antecedent, a condition that renders it consistent with accounts of local reflexivity and with the distribution that would be imposed by Principle A of the classical binding theory (Chomsky 1981).

(2) Adé ri ́ ara rẹ̀ (singular)

Ade see body his

‘Ade saw himself.’

(3) Adé àti Òjó ri ́ ara rẹ̀ (plural)

Ade and Ojo see body his

‘Ade and Ojo saw themselves’

Locality restrictions are familiar. While ara-X is most canonically a direct object, it can also be a prepositional object, depending on the preposition (see AQ 4.1.2.2-3). It is possible for the possessor of a direct object can be anteceded by a subject co-argument of the possessum (see AQ4.1.2.6, 4.2.3.4) although there tends to be an emphatic reading in these cases and it appears a simple pronoun would do without emphasis (much like English his own). It is also possible for a direct object to antecede an indirect object, particularly when the reading is reciprocal (see AQ (C29b)) though a direct object can only marginally antecede a prepositional object for a reflexive reading (see AQ 4.1.2.4) (a difference that may be a function of the plausibility of the examples). Prepositional objects appear to be able to antecede other coargument prepositional adjuncts (AQ 4.1.2.5). An anaphor cannot occur in the sentence initial position in Yoruba as illustrated in (4). This is consistent with the absence of reverse binding and with the rarity of backward anaphora in the language (but see AQ (C22e', f') for

some instances of backward anaphora).

(4) * Ara rẹ̀ rí Olú

body his see Olu

for ‘himself saw Olu’

One of the interesting properties of the ara-X strategy is that when the antecedent is plural, it is completely ambiguous between reciprocal and reflexive readings (in the absence of context). There is no independent form used exclusively for reciprocal readings.

(5) Àwọn obìnrin náà rí ara wọn

They woman the see body their

‘they saw each other’

‘they saw themselves'

‘they saw their bodies’

3.2 Strategy C or the Null Object Strategy

This is the less productive anaphoric strategy. In (6), a phonetically null object is used for anaphoric readings, although what is apparently the same reading is available using the ara-X strategy as in (7).

(6) Olú wẹ̀

Olu bathe

‘Olu washed himself’

(7) Olú wẹ ara rẹ̀

Olu bathe body his

'Olu washed himself’

This strategy works with only a very few verbs in the language, typically verbs of grooming.

3.3 An Exception to Local Obviation of Pronouns.

Pronouns normally cannot normally be used for coargument anaphora.

(8) a. * O tà̀n ọ́

You deceive you

for: ‘You deceived yourself’

b. *Wọ́n tàn wọ́n

they deceive them

for: ‘they deceived themselves’

There do exist certain special cases where a pronoun achieves a reflexive reading with a coargument antecedent in an apparent violation of Principle B of the classical binding theory. Such a case is illustrated in (9).

(9) O ò ri ́ ọ bi ́

You NEG see you QM

‘Can you see yourself?’

The sentence in (9) is an equivalent of the one in (10).

(10) O ò ri ara rẹ bí

You NEG see body you QM

‘Can you see yourself?’

In (9), the second person singular pronoun - ọ takes a co-argument antecedent similar to an anaphor. This is not expected under most theories of anaphora. It is important to note though that only the second person singular pronoun can be used in a sentence such as (9), and moreover, this is only possible in a yes-no question. All other pronouns are not acceptable (for similar readings) in the context.

4. Some Observations on the Yoruba Anaphoric System

4.1 Number Mismatches between Antecedent and Dependent Form

It is possible in a restricted context for a singular noun phrase to serve as the antecedent of a plural pronoun in Yoruba. The following is acceptable in a context in which all the participants coded in the plural pronoun are in front of a mirror. For example, if Ọlá and Àdìó are standing in front of a mirror and Àdìó looks at the mirror, the singular weak pronoun Ó ‘he’ can be used for Àdìó. In this case, Àdìó is still part of the referent of the plural pronoun wọn in a sentence such as (11a). However, it is impossible for Ó ‘he’ to be used for Àdìó in the same context if the anaphor ara wọn ‘themselves’ is used in place of the pronoun wọn ‘them’ as in (11b).

(11) a. Ó rí wọn nínú gíláàsì

he see them in glass

‘He saw them in the mirror'

b. *Ó rí ara wọn nínú gíláàsì

he see body them in glass

for ‘He saw themselves in the mirror'

On the other hand, there is a form of honorific marking where a plural pronoun can be used to track the referent of a singular antecedent in Yoruba. This is especially required when a younger person is talking about or to an elderly person. In such configuration, the pronoun is still syntactically plural, as illustrated in (12).

(12) Bàbá Olú fẹ́ràn ara wọn

father Olu like body their

‘Olu’s father likes themselves’

for: Olu’s father likes himself’

4.2 On the Logophoric Use of Third Person Pronouns

Yoruba differs from languages like English in that it's pronouns serve to morphologically distinguish reports of what is said by other people or what is said to be going on their minds, such as their thoughts, emotions or feelings. Yoruba requires that a third person strong pronoun be used to refer to the person(s) whose thoughts, hopes, beliefs or words are being reported (13). In contrast, a weak pronoun can be used in such constructions only if a strong pronoun is not available (14). If a strong pronoun is available in the embedded sentence, a c-commanding clause-mate weak pronoun is not allowed to have the same referent as the person whose thoughts, beliefs or words are being reported (i.e. the subject of the matrix clause) (15).

(13) Olú_i sọ pé òun_i rí Adé

Olu say that he see Ade

‘Olu said that he saw Ade’

(14) Olú_i ti kéde pé ó_i/j n´ bọ̀ lọ́la

Olu ASP announce that he PROG come tomorrow

‘Olu has announced that he is coming tomorrow’

(15) Olú_i sọ pé ó_j,*i rí baba òun_i

Olu say that he see father him

‘Olu said that he saw his father’

Note though, that an anaphor derived with a combination of ara ‘body’ and the third person (singular) strong pronoun cannot replace the strong pronoun in (13), as shown in (16), and thus a weak pronoun cannot antecede a strong pronoun that forms part of an ara reflexive, as in (17).

(16) *Olú_i sọ pé ara òun_i rí Adé

Olu say that body his see Ade

For: ‘Olu said that he saw Ade’

(17) * Olú_i sọ pé ó_j,*irí ara òun_i

Olu say that he see body his

‘Olu said that he saw himself’

An anaphor formed with a strong pronoun must not have a weak pronoun as its antecedent outside of a logophoric context either.

(18) * ó_irí ara òun_i

he see body his

for ‘he saw himself’

In other words, a further limiting factor on the distribution of ara+strong pronoun is that the strong pronoun must be licensed in a logophoric context.

5.0 Some Theoretical Questions

The coincidence of what appears to be an almost fully compositional anaphor for the ara strategy that permits a literal meaning as well as a reflexive one is perhaps not so uncommon, but the use of the same compositionally formed term to achieve a reciprocal reading is surprising, especially if the internal structure of anaphors is to be a guide to what sorts of uses the anaphor can be put to (see, for example, Safir, 1996). Also at issue is what the right internal structure for this form should be and whether or not its internal structure is different when it is interpreted differently. One possible structure of this complex anaphor is to assume that ara is in the specifier of D, where D is headed by the pronoun. Alternatively, perhaps ara heads an N that is a complement to D.

On the other hand, the absence of any dedicated reciprocal form suggests a functional extension of some sort to cover the semantic space of the reciprocal with the best form available. It is a theoretical issue to determine whether the best form available is chosen by a discourse functional, semantic or syntactic process, and whichever sort of process it is, why it is that the complex anaphor is selected as optimal for this purpose.

The existence of examples where a non-commanding antecedent can license a coargument anaphor, as illustrated in AQ4.1.2.5. This would appear to favor the reflexivity approach of Reinhart and Reuland (1993), for example over Principle A approaches or approaches that involve movement to a c-commanding position to establish anaphoric relations (e.g., Hornstein, 2001). Instances where locality is violated are very few in Yoruba, but there may be interesting questions to explore for the somewhat emphatic usage with ara-X in possessive position, but there appear to be locality restrictions to a clause for these cases too (see AQ (Dxa)).

Example AQ (B1a') suggests that connectivity effects might be profitably exported for Yoruba clefts.

Note that AQ (D4d) provides clear evidence for PRO (or for some theories, complement predication of an object) in Yoruba, since only object antecedency for the infinitival object ara-X is permitted in the complement infinitive for the verb meaning 'plead with'; Since direct objects can antecede reflexives (slightly less) well than subjects can, then the object of the infinitive ought to prefer the matrix subject as its antecedent or show signs of marginality if the direct object of the 'plead with' verb is its antecedent. Instead, it behaves as if the 'plead with' object is a subject antecedent, as the PRO analysis of infinitives would predict.

Also of interest is the logophoric system in Yoruba, but we have chosen not to present the distribution of logophoric licensing for strong pronouns as it is not deeply explored in the AQ. It would appear that the ara-X strategy is essentially unaffected, once one understands that a strong pronoun must be logopohorically licensed. With respect to the logophoric issues, however, we expect to develop a new questionnaire to explore such properties. In the meantime, consult Adesola (2001, 2005), Manfredi, (1987, 1995), and Pulleyblank, (1986) for further discussion.

[1] The extra ‘a’ in the anaphors is a genitive marker. I will ignore it in the rest of this work and in the other documents on Yoruba in this project.