This chapter deals with complementizers (like English that) used with verbs of speaking such as ‘say’, ‘tell’, ‘ask’, ‘shout’, ‘whisper’, in a reported speech construction such as She told me that she knew it. In general, the constructions that interest us are indirect-speech constructions (with person shift), but when a language does not have a special indirect-speech construction, we also consider direct-speech constructions (see Güldemann 2008 and 2012 for an in-depth typological study of quotatives).
Complementizers are defined here as elements that link the reported speech to the verb of speaking, not belonging either to the verb of speaking or to the text of the reported speech. Examples are say in African American English (see ex. 1), te in Lingala (see ex. 2), and postposed puris in Korlai (see ex. 3):
In quite a few languages, a form of the verb ‘say’ is used as a complementizer with verbs of speaking. If the ‘say’ form is a bare stem, the pattern is often considered as a kind of serial verb construction. Such serial verb constructions have given rise to a large amount of literature in the field of creole studies which we cannot possibly refer to here, but see e.g. Winford (1993: 291ff.) for an in-depth discussion of ‘say’ complementizers in English-based Caribbean creoles.
In this feature, we investigate asserted reported speech sentences like ‘She said that the boy did not feel well’, not sentences with directive modality like ‘She told the boy to stay in bed’, because these latter constructions often yield different syntactic structures.
Languages often show several different complementizer constructions.
In this feature we distinguish four values:
|Complementizer identical to bare ‘say’||5||24||29|
|Complementizer consists of ‘say’ plus some other marker||1||4||5|
|Complementizer not synchronically related to ‘say’||5||38||43|
There are 29 APiCS languages with a complementizer which is identical to the bare verb ‘say’ (value 1). It is not always easy to decide on the degree of grammaticalization of these verbs/complementizers (see Güldemann 2008, who criticizes the overgeneralization of the grammaticalization path ‘say verb’ > complementizer/ quotative marker for many African languages, and who instead proposes a whole range of other sources for such elements, e.g. similarity markers ‘be like, like this’, deictics, presentationals). But this matter is not of interest here. We only ask whether a verb which means ‘say’, ‘speak’ is used to introduce the content of the reported speech [Note 1: It could be that some of the se-forms in the English-based Atlantic creoles were directly borrowed from Akan quotatives/complementizers sɛ, se ('to be like', 'to say') (see e.g. Lord 1993: 181f). But since the Surinamese creoles have taki as their 'say'-verb (see ex. 5), and non-English-based creoles (e.g. Guinea-Bissau Kriyol with kuma) have similar constructions, the Akan forms had at best a reinforcing influence
on the English lexeme say, see §3 below.]. One example was seen in (1), and more examples are in (4)-(8).
In the Surinamese creoles the complementizer is taki or táa (< talk):
In San Andres Creole, the main verb se ‘say’ can be combined with the complementizer se.
In other languages this is impossible, for instance in Jamaican, where the complementizer se does not co-occur with se ‘say’ and piik ‘speak’, but with all other verbs of speaking Jamaican (Farquharson 2013), like taak ‘talk’ in (7):
Only five languages have a complementizer which consists of the verb ‘say’ in combination with some additional marker (value 2): Berbice Dutch (bifi dati ‘say comp’), Seychelles Creoles (poudir ‘for.say’), Sri Lanka Portuguese (falaa-tu ‘say-PFV’), Yimas-Arafundi Pidgin (maria-k ‘say-IRR’), and Angolar (fala ma ‘say comp’).
In addition to the verb fala ‘say’, Angolar uses the usual complementizer for object clauses ma. This construction (besides the construction with bare ‘say’) is only used when the recipient of the verb of speaking, here m ‘to me’, is expressed (Maurer 2013a).
Value 3 subsumes languages in which the complementizer is not synchronically related to ‘say’ even though there may be a diachronic link. Only five languages have this as their only complementizer strategy, whereas 38 languages share this option with another one. Here, we find complementizers going back to the various European lexifiers, e.g., ki, ke, que, dat, complementizers/quotatives like nde in Kikongo-Kituba (see ex. 10), but also new compound complementizers like Tok Pisin olsem (< all the same) as in (11):
(Smith & Siegel 2013) note that "in rapid speech the complementizer olsem may be reduced to a form which fortuitously is identical to se". But se is not the lexeme for ‘say, talk’ in Tok Pisin (which is toktok) and therefore should not be interpreted as a serial verb.
61 APiCS languages (14 of them exclusively) show no complementizer or linking element between the verb of speaking and the text of the reported speech (value 4). This strategy is not restricted to any area or lexifier, but, remarkably, 7 out of the 14 exclusively zero-marking languages are pidgins.
Besides the specific value distribution in the APiCS languages, it is interesting to note that only 25 languages show just one value, but all other languages have two, three or even all four values at their disposal. With an average of 1.82 value choices (close to two different options per language on average) this feature has the highest number of different values per language.
Bare ‘say’ constructions are almost exclusively concentrated in Africa and the Atlantic creoles. As has been extensively discussed (e.g. Lord 1993: 151ff, Boretzky 1983: 176ff, Holm 1988: 185ff, Parkvall 2000: 64ff), these ‘say’ constructions have clear counterparts in the African substrate languages. But we also find some languages in India and Southeast Asia, with Bislama as an outlier. As (Southeast) Asian languages also show bare ‘say’ constructions in complementizer function (e.g. Lord 1993: 207ff., Bisang 1992: 398f.) we can invoke substrate/adstrate influence here, too.
A final puzzling point: In this feature we see a striking difference between English- and French-based creoles. No single French-based creole shows a bare ‘say’ serial. We have no explanation for this, but there are other features where French-based and English-based creoles differ in a surprising way (e.g. Chapter 71 on noun phrase conjunction and comitative, and Chapter 112 on ‘hand’ and ‘arm’).