SCIENTIFIC ABSTRACT The primary objective of this proposal is to construct an ISO data-base for spectral classification of the brightest representive sources characterized by the IRAS Low Resolution Spectrometer (LRS) classification, the NASA Ames AI classification of the LRS and a sampling based on MK spectral type or IRAS color. In order to obtain the large and statistically significant data-base required to develop a comprehensive spectral classification scheme, we propose to complement and add to spectral measurements proposed by other central program and open time observers. We also need to include the spectral measurements made for calibration and "benchmarking" purposes. Consequently, except for the data for which the SWS and SOT consortia have the rights, we will have to wait until the data taken by others get into the public domain to proceed with the classification effort. The majority of our dedicated time will be devoted to observations with the SWS in the rapid scan mode, SWS01, and with PHOT in the PHT-S mode. We propose to concentrate this effort towards sources which are poorly segregated in the present classification schemes and are not likely to be observed for other reasons. Examples of such objects are MK A through S stars which were too faint for the LRS, and sources whose selection was based on their infrared colors and the emission line objects in the LRS classes 7, 8 and 9. OBSERVATION SUMMARY To fill our initial classification matrix, we required a minimum of two stars in each LRS, AI, MK or color catagory or 10% of the number of sources in a highly populated class with LRS spectra. These objects formed an initial list of 1316 sources, resulting in over 700 entries in each list for the two ISO launch dates. The mission data-base will be surveyed to identify which sources specified by other observers can be used for the classification objective of this proposal (using SWS01, SWS06, PHT40). The observing strategy is as follows: After the delivery of the central programme proposals we analysed the mission data-base containing the proposed targets and AOTs, to check the population of the various classes already entered. We then made a first attempt to complement these proposed observations for our purpose, to populate the various classes identified. If the source is brighter than 2Jy an SWS01 AOT is proposed, if fainter PHT40 is used. We do not wish to block any open time bids to observe the nominated targets. If an open time proposer wishes to make the same observation, we will drop the target and replace it with an alternative source from the original list of 700 for that launch date. This gives us a second attempt to fully populate the spectral classification scheme.