A simple method for the detection of quasar-like or unusual objects by computer is described using the output produced by the COSMOS measuring machine. One method of finding quasars in quantity is to search objective prism plates by eye. However, while this method is successful, it does have some serious shortcomings. 1) The timescale for measuring a single plate from the United Kingdom Schmidt Telescope (UKST) survey by an experienced observer is of the order of one month. A second observer then goes through the plate again in order to reduce the number of missed objects. 2) Given these timescales, it is difficult to do many plates. 3) The plates have to be of the very highest quality. A small increase in the seeing and the numbers of objects detected drops sharply. 4) The sky distribution when measured in this way may reflect observer fatigue etc. in subtle ways. It occured to me at the Schmidt Conference in Cambridge held in December 1979 while listening to Cyril Hazard that it might be possible to classify objects on prism plates directly using COSMOS with simple parameters as had already been done by MacGillivray et all for star/galaxy seperation. Several parameters have been tried so far and it is clear that some parameters will detect some quasars and a different set of parameters other quasars. So far, no single parameter has been found to detect all quasars. One simple parameter that does seem to give a good chance of success is the ratio of the weighted semi-major axis to the semi-minor axis. These are merely two of the 18 parameters given out in the current version of COSMOS. (see the paper by Dr. R. Stobie this conference). Briefly, COSMOS produces output of a field in such a form that the ellipses around an image can be drawn using the second moments. All clean images have ellipses pointing in the Y direction. Large or contaminated images produce either circular images or orientations away from 90°. These are removed from the sample. The plot of COSMOS "magnitude" against axial ratio can be done. The points to the right of the graph and to the faint end are either artifacts caused by merged images or genuine images with a U-V excess for their magnitude. The selection of this subsample is done by computer program and an accurate overlay drawn. The skilled observer can then decide on which images merit further study. In order to give an impression of the scale of the problem for the visual observer, COSMOS detected over 75000 images in the unvignetted part of U.K.S.T. plate 2460. (RA 22h Dec-18.55). The final overlay represents a mere 1% of this total . Just how many of these objects are really QSO's (and I would not believe more than 20% were) is questionable until a full comparison is done with archive data and proper telescope observations made. At this stage I can only confirm that some of the selections are undoubted quasars . Given that 1500 quasars are already known, it is a fair and proper question to ask why we should be looking for a lot more. I would argue that there are many questions that cannot be answered by such a small sample. It only represents 2½% of the total UKST survey. One final detail, you can turn the algorithms around to look for circular objects on prism plates. These will be either very red objects, emission line only objects or very unusual things indeed (Kodak plate flaws are regretably included in this list). The difference is though, that these kinds of unusual objects go down to the limit of the direct Survey plates.