How to investigate stigmatized constructions in French?

It is widely known that some constructions like left-dislocation (Les gateaux, je les ai mangés / The biscuits, I ate them.), though widespread in spoken corpora are marked in standard written French. Acceptability judgment experiments in written context show stigmatization compared to other pragmatically marked constructions (e.g. clefts), even in felicitous contexts. We replicated those experiments using spoken sentences, and a written speeded grammaticality judgment paradigm. 
In spoken French we did not find any differences to the previous experiments. Even with the right prosody, left-dislocations are judged significantly less acceptable than clefts in their respective felicitous contexts. The task demands seem to encourage metalinguistic judgments based on sociolinguistic parameters. 
Speeded grammaticality judgments (SGJ), however, show equally high acceptability rates for left-dislocations and clefts. Ungrammatical clefts with clitics, on the other hand, are judged strongly unacceptable while pragmatically odd left-dislocations without clitics are judged moderately acceptable. This is coherent with gradient acceptability accounts where clefts with clitics are ungrammatical syntax wise (hard constraint) and left-dislocations without clitics are possible grammatically, but need a salient alternative to be felicitous (soft constraint). 
This series of experiments shows that challenging tasks like SGJ provide results closer to the actual usage of stigmatized constructions.