Modular forms were actively studied by number theorists Hecke and Siegel in the 1930s, but it was not widely appreciated. Around the same time Hardy, in a series of lectures on Ramanujan's work delivered at Harvard in 1936, called modular forms -- as represented by Ramanujan's interest in the coefficients of the weight 12 form $\Delta(z)$ \Delta(q) = \sum_{n \geq 1} \tau(n)q^n$ -- "one of the backwaters of mathematics". The study of modular forms basically died off in the 1940s and 1950s. It was revitalized by Weil, Shimura et al. in the 1960s. See the introduction to Lang's book on modular forms for some relevant historical remarks.
[EDIT: As Emerton points out in his comment below, the full quote by Hardy is actually more complimentary, so let me include it here: "We may seem to be straying into one of the backwaters of mathematics, but the genesis of $\tau(n)$ as a coefficient in so fundamental a function compels us to treat it with respect." This is at the start of Chapter X of Hardy's "Ramanjuan: Twelve Lectures on Subjects Suggested by his Life and Work."]
