Fourteen sites occupied by Lagorchestes conspicillatus in the Northern Territory were examined from July to September 1986. The abundance of possible food items at each site was estimated by measuring the percentage cover of herbs, grasses and shrubs with foliage below one metre. Also recorded was the abundance of potential shelters, including grass tussocks, spinifex hummocks (Plectrachne and Triodia spp.) and low shrubs and trees. All sites possessed trees (more than 10 per ha) and shrubs (more than 5 per ha) together with more than 0.5% cover of green grass foliage. A model of suitable habitat was constructed using the 1986 results, and this was tested within the species' range in the Pilbara, Western Australia, and near Daly Waters in the Northern Territory in 1987. The abundance of L. conspicillatus was scored at control sites (sites with all features considered essential on the basis of 1986 data) and test sites (sites lacking 1-2 'essential' features). These data indicated that L, conspicillatus can occur at sites with as little as 0.2% cover of green grass provided there is more than 1.5% herb cover. Sites with no green grass (less than 0.2%) were usually not occupied. All sites occupied by L. conspicillatus had potentially suitable shelters, either shrubs, grass tussocks or spinifex hummocks within 50m of the feeding areas. Diet selection by L. conspicillatus during the 1986 and 1987 dry seasons was as follows: (a) herbs were eaten in preference to other food items except certain seeds; (b) grass formed at least 10% of faecal contents irrespective of herb cover; and (c) at sites with less than 0.5% herb cover the levels of grass in the faeces increased to compensate for the low availability of herbs.