Currently, there is no single concept of the optimal spatial structure of mRNA during translation. It is known that many proteins, associated with the 5' end of mRNA, interact with proteins associated with the 3' end. Moreover, this interaction often affects the activity of these proteins. It is possible within the same mRNA molecule only when the mRNA forms a circular structure in which its ends are spatially close. Discovery of such proteins, in the 90s of the 20th century, made it possible to formulate the closed-loop mRNA structure hypothesis, in which it is assumed that the ends of translationally active mRNA are fixed next to each other due to the interaction of proteins and (or) RNA. However, later it was shown that a closed-loop structure is not always necessary for translation. Moreover, some authors have proposed a model according to which the translating mRNA, on the contrary, should be unfolded into a linear structure. Thus, the spatial structure of the translating mRNA does not have to be universal for all mRNA and can change dynamically, which affects its functional activity. In this review, we have summarized a variety of experimental data and concepts on the relationship between the spatial structure of mRNA and its translational activity.