Chapter 98: Complements of 'think' and 'want'
1. Introduction ⇫ ¶
In the present chapter, we consider the similarities and differences between two complement clause types. On the one hand, we consider complement clauses which depend on the verb ‘think’ and where the subject of the main clause is different from the subject in the complement clause (e.g. English She thinks that her son is at home). On the other hand, we look at complement clauses of ‘want’ where the subject of the matrix clause is again different from the subject in the complement clause (e.g. English She wants her son to come home). For same-subject complements of ‘want’, see Chapter 97 (“Want complement subjects”).
In comparing these two complement clause types, two separate parameters are relevant:
(i) whether there is an overt complementizer, and
(ii) if so, whether both complement types show the same complementizer or different complementizers.
We distinguish five feature values:
 | Identical complementizer | 15 |
 | Different complementizer | 14 |
 | Only 'think' complement has a complementizer | 19 |
 | Only 'want' complement has a complementizer | 1 |
 | No complementizer | 24 |
| Representation: | 73 | |
If several different ‘think’ complement constructions or several different ‘want’ constructions are possible, the contributors were asked to choose the dominant construction.
2. The values ⇫ ¶
Fifteen APiCS languages show value 1, where the ‘think’ and ‘want’ complement clauses show the same complementizer.
The fourteen languages displaying value 2 show different complementizers.
The third value comprises languages in which the two complement types differ in that only ‘think’ complements have a complementizer (perhaps optionally), while ‘want’ complements lack one. This type is fairly widespread, too. One example comes from Pichi (cf. 4a-b), where we find the complementizer se in the ‘think’ construction, whereas in the ‘want’ construction there is no complementizer.
Likewise, the examples from Ambon Malay in (5a-b) show an optional complementizer (kata) introducing the ‘think’ complement clause, whereas the ‘want’-complement clause does not allow any complementizer.
In Berbice Dutch, the verb glofu ‘believe’ can introduce complement clauses marked by the complementizer dati ‘that’, by the serial complementizer bi(fi) ‘say’, or by a zero complementizer. However, in complement clauses depending on the verb suku ‘want’, no complementizer position is available.
The fourth value is the mirror image of value 3, i.e. ‘want’ complements have a complementizer, while ‘think’ complements lack one. But this type is found only in Louisiana Creole, where the verb ole ‘want’ optionally has the complementizer ke, whereas krwar ‘believe’ generally shows no complementizer.
In 24 APiCS languages both complement clause types have no overt complementizer (value 5). Examples come from Papiá Kristang and Kinubi.
3. Other differences between the two complement types ⇫ ¶
‘Think’ complements and ‘want’ complements often differ in ways that are unrelated to the presence, absence or form of the complementizer. In particular, the ‘want’ complements often carry a “subjunctive” or “infinitive” marker of some sort, which is lacking in ‘think’ complements. In this way, there can be a fairly striking difference between the two clause types even when the language has value 1 (same complementizer) or 5 (no complementizer in both cases).
Thus, in Nigerian Pidgin, different-subject ‘think’ and ‘want’ complement clauses are introduced by the same complementizer se, but the ‘want’ complement clause (cf. 8b) shows the subjunctive marker mek (from English make; cf. Ihemere 2006 for the use of mek in Nigerian Pidgin, and Yakpo 2009 for a similar situation in Pichi, see (4) above).
And in quite a few Atlantic English-based languages, the ‘want’ complement clause has a marker such as fi, fo or fu (deriving from for), corresponding to the infinitival marker to in English:
See also ex. (6)-(7) in Chapter 97. Such markers have often been called “infinitival” markers (cf. Mufwene & Dijkhoff 1989) or even “complementizers”, but we do not consider them complementizers here, as they do not occur in a clause-peripheral position. Their immediately preverbal position makes them more similar to modality markers. We did not single them out as a special type because one cannot readily distinguish such markers from other modality markers such as gò in (10).
Another way in which ‘want’ complements may be distinct from ‘think’ complements is that they may lack person marking, as in Seychelles Creole in (11), where the 3rd person marker i is missing (see also Michaelis 1994: 82-91).
Since most of the APiCS languages lack such agreement markers, this criterion cannot be used generally to classify the languages either.