Diffusion coefficients, DN, for 2D vacancy nanopits are compared with those for 2D homoepitaxial adatom nanoislands on metal(100) surfaces, focusing on the variation of DN with size, N. Here, Nis measured in missing atoms for pits and adatoms for islands. Analysis of DN is based on kinetic Monte Carlo simulations of a tailored stochastic lattice-gas model, where pit and island diffusion are mediated by periphery diffusion, i.e., by edge atom hopping. Precise determination of DNversus N for typical parameters reveals a cyclical variation with an overall decrease in magnitude for increasing moderate O(102) ≤ N ≤ O(103). Monotonic decay, DN ∼ N-β, is found for N ≥ O(102) with effective exponents, β = βeff, for both pits and islands, both well below the macroscopic value of βmacro = 3/2. DN values for vacancy pits are significantly lower (higher) than for adatom islands for moderate N in the case of low (high) kink rounding barrier. However, DN values for pits and islands slowly merge, and βeff → 3/2 for sufficiently large N. The latter feature is expected from continuum Langevin formulations appropriate for large sizes. We compare predictions from our model incorporating appropriate energetic parameters for Ag(100) with different sets of experimental data for diffusivity at 300 K, including assessment of βeff, for experimentally observed sizes N from ∼100 to ∼1000.