An ecological community can be viewed as a network of interspecific interactions (community network), where a change in population density leads to cascading changes of other population densities. Ecological theory predicts that the community network structure (complexity, for example) influences community dynamics. However, there have been few empirical tests for the theory, partly due to the difficulty in quantifying, or even detecting the interspecific interactions in the field. Here, a “minimalist approach” is proposed to overcome the difficulty and enable testing the community network theory. In this approach any specific models of population dynamics are not assumed, but instead the interactions are inferred from time-series data obtained at the spatial and temporal scale of interest. Nonlinear dynamical theory suggests that a time series of a variable contains information about the whole system. We can use this to identify interspecific interactions, quantify their signs and strengths and evaluate its dynamical consequences from the empirical time-series data. Taking insect, fish and plankton communities as examples, I show how this approach enables us reconstructing the complex structure of community networks, revealing the fluctuating and condition-dependent nature of interspecific interactions and relating the community network structure to the dynamic properties at higher organization levels.