Fauna conservation outside protected areas can make an important complementary contribution to conservation within reserves. This thesis aimed to contribute new information and analytical frameworks to the science of fauna conservation in human-modified landscapes. Two approaches were used: (1) empirical data collection and analysis, and (2) the discussion and development of conceptual landscape models. ¶ Empirical work focused on lizard distribution patterns in two production landscapes in southeastern Australia. Lizards were targeted because ectotherms are frequently neglected by conservation biologists. The “Nanangroe grazing landscape” was used for sheep and cattle grazing. In this landscape, approximately 85% of pre-European woodland cover had been cleared, and understorey vegetation was sparse. Lizards were surveyed at 16 landscape units, which were stratified by aspect, topographic position and amount of tree cover. Each landscape unit contained three sites, and each site contained three plots. Regression modelling showed that different species responded differently to their environment. For example, the four-fingered skink (Carlia tetradactyla) and Boulenger’s skink (Morethia boulengeri) were more likely to occur at woodland sites with northerly aspects, whereas the striped skink (Ctenotus robustus) and olive legless lizard (Delma inornata) were more likely to inhabit sites with a simple microhabitat structure. Statistical analysis further showed that the habitat attributes that lizards were related to varied continuously through space, and over different spatial scales. For example, invertebrate abundance (a proxy for food availability) varied most strongly over tens of metres, whereas the amount of grass cover varied most strongly over hundreds to thousands of metres. Thus, work at Nanangroe revealed spatially complex patterns of lizard occurrence and habitat variables. ¶ The “Tumut plantation landscape” was a spatial mosaic of native eucalypt (Eucalyptus) forest patches embedded within a plantation of the introduced radiata pine (Pinus radiata). In this landscape, thirty sites were surveyed for lizards. Sites were stratified by forest type and patch size, and included eucalypt patches, pine sites, and extensive areas of eucalypt forest adjacent to the plantation. Regression modelling showed that lizard species responded to various habitat attributes, including elevation, the amount of eucalypt forest within 1 km of a site, invertebrate abundance and ground cover. Variables related to habitat fragmentation often were significant predictors of lizard occurrence. However, work at Tumut suggested that important additional insights into lizard distribution patterns could be obtained by considering variables related to food and shelter resources, and climatic conditions. ¶ The Nanangroe and Tumut landscapes were in close proximity, but together spanned an altitudinal gradient of 900 m. An investigation of changes in lizard community composition with altitude showed that (1) only one species was common to Nanangroe and Tumut, (2) different species had different altitudinal preferences, and (3) ecologically similar species replaced one another with increasing altitude. These results highlighted that even in highly modified landscapes, natural gradients (such as climate) can play an important role in shaping animal assemblage composition and species distribution patterns. ¶ Empirical work suggested that, in some landscapes, the frequently used “fragmentation model” is a relatively weak conceptual basis for the study of animal distribution patterns. The fragmentation model implicitly assumes that “habitat patches” can be defined unequivocally across many species, and that patches are located within a relatively inhospitable matrix. Where these assumptions are breached, conservation guidelines arising from the fragmentation model may be too simplified. In spatially complex production landscapes, it may be more appropriate to maintain habitat heterogeneity at multiple spatial scales than to focus solely on the management of large, pre-defined patches. ¶ Given the potential limitations of the fragmentation model, a new, more holistic landscape model was developed. The “continuum model” was derived from continuum theory as developed for plant ecology. The continuum model recognises (1) spatial continua of environmental variables, and (2) species’ individualistic responses to these variables. For animals, key environmental variables may be related to the availability of food, shelter, sufficient space, and suitable climatic conditions. Unlike the fragmentation model, the continuum model is inherently process-based and thus may help to link the perceived gap between patterns and processes in landscape ecology. ¶ Three general conclusions arise from this thesis: 1. Some heterogeneous production landscapes support many native species, and therefore represent important conservation opportunities. 2. In some modified landscapes, the fragmentation model does not capture the complexity of animal distribution patterns. In those landscapes, conservation recommendations derived from the fragmentation model may be overly simplistic. 3. The continuum model may be a useful extension of the fragmentation model. It provides a process-based conceptual basis for empirical work on animal distribution patterns.