This paper has as main objective to demonstrate how the use of neural networks, self-organized maps type, is a potentially clarifying tool in the treatment, analysis and visualization of scientometric data, specifically, in the case of the analysis of the Spanish doctoral theses in teaching Social Sciences, indexed in TESEO database and defended between 1976 and 2014. A census of 301 doctoral theses has been recovered, analyzed according to autonomous communities (Andalusia and Catalonia), five-year term groups, thematic categories and educational stages. In Andalusia, research has concentrated its production in the stages of Primary and Secondary Education, and in the thematic of Didactics of Geography. The dissertations production is highest in the five-year period 1986-1990 and 2001-2005. In Catalonia, research deals mainly with the stages of Secondary and Higher Education, and the theme of Didactics of Social Sciences. The most productive five-year periods in Catalonia were 1991-1995, 1996-2000, 2001-2005 and 2006-2010.