Can natural language semantics explain syllogistic reasoning?

It has long been known that conversion of the quantifiers can explain syllogistic reasoning errors (Newstead, 1989; Revlis, 1975; Wilkins, 1928). Geurts is more concerned with explaining how people solve valid syllogisms, and an integral part of his model is the (entirely legal) conversion of ‘no’ and ‘some’. This part of Geurts’ theory seems unexceptionable, but more problematic is his assumption that conversion of ‘no’ and ‘some’ can be carried out with negligible processing effort. While it may be true that these are relatively straightforward operations it is implausible that they put no discernible strain on working memory. Somewhat bizarrely, Geurts assumes that conversion does add difficulty when it is applied illicitly to ‘all’ and ‘some...not’. It is difficult to see how the same operation is both cost free and resource consuming depending on which quantifier it is applied to!