We present a 1200-micron image of the Great Observatories Origin Deep Survey North (GOODS-N) field, obtained with the Max Planck Millimeter Bolometer array (MAMBO) on the IRAM 30-m telescope. The survey covers a contiguous area of 287 square arcmin t
o a near-uniform noise level of ~0.7mJy/beam. After Bayesian flux deboosting, a total of 30 sources are recovered (>=3.5sigma). An optimal combination of our 1200-micron data and an existing 850-micron image from the Submillimetre Common-User Bolometer Array (SCUBA) yielded 33 sources (>=4sigma). We combine our GOODS-N sample with those obtained in the Lockman Hole and ELAIS-N2 fields (Scott et al. 2002; Greve et al. 2004) in order to explore the degree of overlap between 1200-micron- and 850-micron-selected galaxies (hereafter SMGs), finding no significant difference between their 850-micron to 1200-micron flux density distributions. However, a noise-weighted stacking analysis yields a significant detection of the 1200-micron-blank SCUBA sources, whereas no significant 850-micron signal is found for the 850-micron-blank MAMBO sources. The hypothesis that the 850/1200-micron flux density distribution of SCUBA sources is also representative of the MAMBO population is rejected at the ~4sigma level, via Monte Carlo simulations. Therefore, although the populations overlap, galaxies selected at 850 and 1200micron are different, and there is compelling evidence for a significant 1200-micron-detected population which is not recovered at 850micron. These are submm drop-outs (SDOs), with S_850/S_1200 = 0.7-1.7, requiring very cold dust or unusual spectral energy distributions (T_d ~ 10K; beta ~ 1), unless SDOs reside beyond the redshift range observed for radio-identified SMGs, i.e. at z > 4.
