(Gerardo, before) Then we agree that the "poverty of stimulus" is a false assumption. But the research of exactly what and how can a child learn from each stimular situation is still a relevant topic for a developmental account of language learning.
(Neil) And that is where we part company. Treating it in terms of "stimular situation" seems too limiting.
(Gerardo) "Too" limiting... for what purposes? Compared to which criterion? Don´t forget that the assessment of "too much" or "not enough" is relative to one or another purpose and criterion, so you cannot dismiss without further argumentation the purpose of having an analytical research program of children´s learning from stimular situations (a purpose that is valuable, at the very least, because it can give us effective tools for assessing and helping people with learning disabilities).
(Gerardo, before) It´s true that the child doesn´t need to acquire "rules of syntax". But he still needs to acquire generalizations (what´s "simmilar") and discriminations (what´s "not simmilar") of words and frames. And for such learning, the child needs experiential data which functions as positive or negative evidence for each generalization and discrimination.
(Neil) That's a standard view, but I do not agree with it. I cannot find any adequate account of "similar" in the literature.
(Gerardo) You don´t know any adequate account? What about the behavioral research about generalization and discrimination of stimulus classes? I´d quickly acknowledge that there´re many unanswered questions yet, but it would be both unfair and self-defeating to deny all the advances of such research area without further argumentation.
(Neil) It seems to me that "similar" is a perceptual judgement, and is prior to being able to discriminate. Roughly speaking, the ability to make judgements of similarity is not something learned through experiential data, but is instead something developed in order to make experiential data possible.
(Gerardo) Well, I´d acknowledge that a lot of similarities don't need to be learned through experiential data (they might be detected as similar thanks to our innate perceptual mechanisms), but you should also acknowledge that a lot of judgements of similarity must be learned through experiential data. For example a doll and a ball are very different objects in their perceptual features, and yet a child can learn that are "of the same kind" at least regarding certain features (i.e., they are "toys", they are "made of plastic", they are "properties of the same child", etc.). How would you account for such kind of judgments of similarity if you deny that they were learned through experiential data?