We use a set of evoked words to define the vertices of a network. The connections between vertices are established by individuals in a population that evoke these words. The resulting graph is called an Evoked Words Network, EWN. The data of evoked words comes from an epidemiological research in odontological public health. In this research we consider three concept themes or evocative words: mouth, disease, and health. We investigate these words in two populations: an upper middle class and a poor district of the city of Natal. We compare and analyze six EWNs. The distribution of connectivities of all of these EWNs indicates a scale-free structure, with the data fitting a power law. The analyzed quantities of the EWNs depend more on the concept theme than on the income of population. This conclusion is discussed in the context of language-based networks.