In mathematical activities and in the analysis of mathematics teaching-learning processes, "objects," "signs", and "representations" are often mentioned, where the meaning assigned to those words is sometimes very broad, sometimes limited, other times intuitive, allusive, or not completely clear. On the other hand, as international research in mathematics education has shown, the confusion between objects, signs, and representations is one of the main reasons of the difficulties in learning mathematics. But what kinds of objects are involved in teaching-learning mathematics? Why should we distinguish a knowledge object, and in particular a mathematical object, from one of its representations? What is meant by "sign"? Can we equate the term "sign" with the term "representation"? In this article we will try to provide an answer to these questions, taking into account the main contributions to mathematics education made by the semiotic theories that are considered the most relevant in the analysis of the cognitive processes involved in mathematical activities. In particular, we will refer to the semiotic representation registers theory, on which Duval's semio-cognitive approach is based. In general it will be shown that the choice of a semiotic approach to mathematics education assumes a fundamental theoretical choice closely tied to the fundamental distinction between classifying signs and classifying semiotic systems, which is often implicit or rather not emphasized enough. The example presented shows how the semio-cognitive analysis of the processes involved in the solution of a mathematical problem provides new and effective professional reading keys of students' difficulties in learning mathematics.