In this section I first discuss current opinions in Latin linguistics concerning
All known languages, in spite of their considerable formal differences, have complex sentences, both those treated in this chapter and relative clauses. In view of this, it is not desirable to assume that earlier stages of Latin (e.g. Indo-European) did not allow the possibility of embedding predications in other predications. Yet, this is currently assumed, and in many cases complex sentences are explained as arising from two independent, simple sentences.  Such an assumption presumably arises from ideas on the development of the human brain, by analogy with the development of child language, regardless of the existence of empirical support (e.g. K.–St. II. 1). A second reason may be the fact that a number of subordinators can in all likelihood be shown to have developed from connectors  and adverbs, and it seems difficult to develop a hypothesis as to which subordinators existed in pre-historic periods. But the fact that we are unable to determine which specific subordinators existed does not mean that there were none. The Romance languages show no trace of highly common Latin subordinators such as ut and cum; if Latin had not been preserved, we would never be able to reconstruct these subordinators. General statements such as `In der ältesten indogermanischen Zeit gab es keine Nebensätze' (`In the oldest Indo-European period there were no subordinate clauses') are unfounded. 
In this section I treat two types of complex sentence and their current historical explanations. These are (i) clauses introduced by a subordinator and (ii) the AcI.
(i) Clauses introduced by a subordinator
(i) Clauses introduced by a subordinatorScherer (1975) clearly illustrates the ways in which he thinks specific constructions have come into being.  Owing to this clarity, he offers a
|Latin sentence||To be explained from||Scherer|
|(a) ubi sim nescio||`Where might I be? I do not know'||p. 238|
|(b) num me vituperas? :: multum abest ut te vituperem||`Are you criticizing me?' `Not at all; how could I criticize you?||p. 253|
|(c) dux milites hortatur ut fortiter pugnent||`… how they would fight courageously'||p. 269|
|(d) nemo tam stultus est quin id intellegat||`How could he not understand this? Nobody is that stupid'||p. 271|
convenient point of departure for the discussion. I first present, in table 7.7, his examples and the constructions which he uses as his starting point. The types of explanation are evident. In (a) a paraphrase with two independent sentences is chosen, and it is suggested that this structure is also possible in Latin. However, an independent (deliberative) question ubi sim does not seem very plausible, and in the paraphrase either nescio is one-place or the second argument is to be understood from the context (and this second argument would, of course, be ubi sim). (b) is slightly more complicated: in the Latin sentence ut does not mean `how' and the speaker is not wondering `how he could criticize'. Here, too, multum abest would lack an argument. The analysis also presupposes a kind of conversation. These remarks with regard to (b) in part also hold for (c). In (d) the order must be manipulated; the order given in the English sentence is not (or rarely?) found in Latin.
In giving these examples I do not intend to prove that subordination could never have arisen from originally independent sentences; on the contrary, it is evident that examples may be found. But the postulated development often presupposes forms and constructions which deviate rather essentially from Latin as it is known to us, and a too ready use is made of seemingly plausible paraphrases in a modern language.
(ii) AcI (a) arguo praetorem : pecunias cepisse → arguo: praetorem pecunias cepisse (b) nuntio eum: dictatorem esse factum → nuntio : eum dictatorem esse factum
(ii) AcIThe `true' AcI is often regarded as identical with or at any rate as having originated from the accusative + prolative infinitive. Above (p. 126) I have shown that synchronically the two constructions are certainly not identical. Here we address the alleged diachronic relationship. Many linguists  assume a so-called `reanalysis':
(a) arguo praetorem : pecunias cepisse → arguo: praetorem pecunias cepisse
(b) nuntio eum: dictatorem esse factum → nuntio : eum dictatorem esse factumAn objection against this analysis is the fact that it suggests that in Classical Latin (the situation to the right of the arrows) (a) and (b) had the same structure; this was not the case. Moreover, in (b) it is suggested (to the left of the arrow) that besides the `reporter' at one stage another person could occur as argument with verbs such as nuntio, something which is highly unlikely from a semantic point of view. Furthermore, it is unclear what pecunias cepisse might have meant.  For arguere we must assume that the three-place construction current in Classical Latin (arguere aliquem stupri) did not exist. Here, too, the facts are strained unnecessarily. 
From this section it must be concluded that the existing hypotheses concerning the origin of complex sentences are too simple. This is not to say, of course, that I have as yet offered an alternative explanation for the situation as we find it.
Pinkster, Harm (1942-) , Latin Syntax and Semantics [info], xii, 320 p.: ill.; 24 cm. [word count] [Pinkster].