We consider the problem of computing the size of each $r$-neighbourhood for every vertex of a graph. Specifically, we ask whether the size of the closed second neighbourhood can be computed in subquadratic time. Adapting the SETH reductions by Abboud et al. (2016) that exclude subquadratic algorithms to compute the radius of a graph, we find that a subquadratic algorithm would violate the SETH. On the other hand, a linear fpt-time algorithm by Demaine et al. (2014) parameterized by a certain `sparseness parameter' of the graph is known, where the dependence on the parameter is exponential. We show here that a better dependence is unlikely: for any~$δ< 2$, no algorithm running in time $O(2^{o({\rm vc}(G))} \, n^δ)$, where~${\rm vc}(G)$ is the vertex cover number, is possible unless the SETH fails. We supplement these lower bounds with algorithms that solve the problem in time~$O(2^{{\rm vc}(G)/2} {\rm vc}(G)^2 \cdot n)$ and $O(2^w w \cdot n)$.