This paper reviews the outcomes of the 1988/9 UFC research selectivity exercise and uses these as a basis for recommendations to those undertaking the 1992/3 exercise. Attention is drawn to disparities between subjects which may result in inefficient allocations at institutional level. Of eight factors incorporated into a statistical model only size of cost centre is consistently associated with research selectively score. The only other variable that influences an individual cost centre's score in the majority of subject areas is that institution's score in other subjects - thus suggesting a ‘halo’ effect. Surprisingly, research expenditure per member of staff is not strongly associated with reseach selectively rating.