An important political and social phenomena discussed in several countries, like India and Brazil, is the use of WhatsApp to spread false or misleading content. However, little is known about the information dissemination process in WhatsApp groups. Attention affects the dissemination of information in WhatsApp groups, determining what topics or subjects are more attractive to participants of a group. In this paper, we characterize and analyze how attention propagates among the participants of a WhatsApp group. An attention cascade begins when a user asserts a topic in a message to the group, which could include written text, photos, or links to articles online. Others then propagate the information by responding to it. We analyzed attention cascades in more than 1.7 million messages posted in 120 groups over one year. Our analysis focused on the structural and temporal evolution of attention cascades as well as on the behavior of users that participate in them. We found specific characteristics in cascades associated with groups that discuss political subjects and false information. For instance, we observe that cascades with false information tend to be deeper, reach more users, and last longer in political groups than in non-political groups.