Nitrogen isotope signatures (δ(15)N) provide powerful measures of the trophic positions of individuals, populations and communities. Obtaining reliable consumer δ(15)N values depends upon controlling for spatial variation in plant δ(15)N values, which form the trophic 'baseline'. However, recent studies make differing assumptions about the scale over which plant δ(15)N values vary, and approaches to baseline control differ markedly. We examined spatial variation in the δ(15)N values of plants and ants sampled from eight 150-m transects in both unlogged and logged rainforests. We then investigated whether ant δ(15)N values were related to variation in plant δ(15)N values following baseline correction of ant values at two spatial scales: (1) using 'local' means of plants collected from the same transect and (2) using 'global' means of plants collected from all transects within each forest type. Plant δ(15)N baselines varied by the equivalent of one trophic level within each forest type. Correcting ant δ(15)N values using global plant means resulted in consumer values that were strongly positively related to the transect baseline, whereas local corrections yielded reliable estimates of consumer trophic positions that were largely independent of transect baselines. These results were consistent at the community level and when three trophically distinct ant subfamilies and eight abundant ant species were considered separately. Our results suggest that assuming baselines do not vary can produce misleading estimates of consumer trophic positions. We therefore emphasise the importance of clearly defining and applying baseline corrections at a scale that accounts for spatial variation in plant δ(15)N values.