During the last three decades, many recommendations for a gender fair Spanish language have been proposed, but, generally speaking, it has been the substitution of the so-called 'masculine generic' (e.g. los profesores 'the teachers [masc.]') by gender neutral (e.g. el pro¬fesorado 'the staff of teachers) or explicit references (e.g. los profesores y (las) profesoras 'the teachers [masc. and fem., resp.]') which has been favoured. Two important assumptions are implicit in these recommendations. First, the sentences containing the masculine forms would lead to associations primarily to men (thus leaving women 'invisible'), whereas sentences containing either the gender-neutral forms or the gender-explicit references would evoke a generic association. Second, the associations between form and mental representation are considered inalterable and unlikely to change over time. This paper intends to interrogate these assumptions by means of two questionnaire investigations that were carried out in Spain in 1995 and 2005 in which native speakers of Spanish were asked to complete specific filler sentences. The results of both investigations demonstrate that there is no clear-cut correspondence between certain linguistic forms and the mental (gender-) representations evoked in peoples' minds. For example, a masculine form is not automatically connected with a male image. The investigation also shows that some associations significantly change over time; for example, a clear male bias of the masculine form in the first study seems to have vanished within a time span of ten years. With respect to another aim of the gender fair recommendations - namely to make women more visible in public discourse – the investigations show that this, indeed, is best achieved by explicitly referring to women. Interestingly, this result appears to be stable over time. Altogether, it seems justifiable to conclude that the assumptions underlying the recommendations for a gender fair language can be challenged. At any rate, it is important to point out that these assumptions are subject to change over time - at least within politically influenced language matters – and perhaps faster than has been expected.