This work is part of a general research program, whose aim is to confront the predictions of linguistic formalisation to actual human processing mainly for lexical, syntactic and above all semantic issues, excluding as much as possible the interference with pragmatics. Examples of questions to be address are the following: Are there preferred readings? Is a sentence complex to interpret? Which linguistic ingredients are responsible for its complexity? We present here a semantic claim about quantifier scope taking , together with an experiment that we designed to test this claim, and which actually justifies our claim. Quantifiers are rather viewed as epsilon terms which better match the syntactic structure of the sentence.