1 Introduction

Our present knowledge of the properties of starburst galaxies, and of the physical conditions in nuclei and in extranuclear HII regions of spiral galaxies, rests on the analyses of a rather limited number of data sets. The selection criteria for the various samples necessarily introduce biases which influence the derived properties and which might explain why these studies sometimes reach conflicting results, for example on the role of the bar or of gravitational interactions in triggering starbursts. It is thus essential to multiply the observations of starburst galaxies, in order to make available the largest possible dataset for future analyses.

The goal of the present paper is to present a new sample of starburst galaxies which differs markedly in its selection from other samples. We wanted to avoid selecting small metal-poor galaxies (such as blue compact dwarfs) preferentially found in objective prism surveys, monsters (such as mergers or ultraluminous infrared galaxies) selected because of their unusual shape or luminosity, and active galaxies (such as Seyferts), which are probably related to starbursts, but should be considered separately. We thus devised selection criteria that would enable us to catch in our nets galaxies preferably resembling rather "normal" giant spiral galaxies. We believe that we have been fairly successful, and this new sample has been subjected to an extensive multifrequency analysis (Contini 1996), the results of which have been or will be reported separately (see Sect. 7).