===================================================================== ==> This proposal is an extension of PANDRE.PA_PROTO and LNORDH.SURVEY_1 ===================================================================== We propose to conduct CAM LW2 mapping observations of a broad sample of starless dense cores (with no associated IRAS sources) located within optically opaque spots of the sky. These cores are presumably pre-collapse and pre-stellar in nature and provide good laboratories for studying the initial conditions of star formation. The LW2/LW3 images of the rho Oph main cloud (Abergel et al. 1996) obtained as part of LNORDH.SURVEY_1 demonstrate that CAM is sensitive enough to detect dense clumps in absorption (against the diffuse mid-IR background produced by the interstellar radiation field or arising from cloud envelopes). The high angular resolution and high mapping speed of CAM thus offer a unique opportunity to search for small-scale structure within a large number of cloud cores. We have selected for study a sample of dark clouds/globules which all have a relatively strong (> 1 MJy/sr) estimated background at 6.75 um and span a wide range of properties (e.g. morphologies, environments). Our proposed absorption survey should bring invaluable information on the structure of pre-stellar cores and on the process of cloud fragmentation. It is crucial to study a large number of cores in order to build a complete theoretical picture. Based on the rho Oph example, we expect to discover many new pre-stellar and/or protostellar dense clumps that we intend to study in detail at a later stage by means of follow-up molecular-line and continuum observations with large ground-based (sub)millimeter radiotelescopes. By comparing the column densities derived from mid-IR extinction profiles with those estimated from mm continuum maps, an immediate scientific return of our survey will be the relative calibration of the uncertain dust absorptivities in the mm and the mid-IR for a whole range of core densities.