We examined data on bird and reptile assemblages in a plantation landscape in southern New South Wales, south-eastern Australia, for evidence of threshold responses to the amount of native eucalypt vegetation in circular areas of 2000 and/or 1000 m around field survey sites. These circular areas contained varying proportions of native Eucalyptus and exotic radiata pine Pinus radiata forest thereby providing a basis for examining potential threshold effects in relation to the area of native vegetation cover. For bird species richness or the probability of detection of individual bird species we found no empirical evidence of a threshold response to the area of native vegetation cover, or any other potential explanatory variables. All relationships were characterised by considerable variability in the response data. "Broken-stick" relationships which involved sudden change points did not fit the response data better than smooth relationships obtained from generalised additive or linear models. As with birds, there was no evidence that a threshold model between lizard richness and the amount of native vegetation within 1000m described the relationship any better than a smooth, continuous or other type of relationship. Several related factors may explain our results. An important one is that species-specific responses to landscape conditions mean that marked thresholds will not be seen for an aggregate measure like species richness at a given value for a given landscape variable. Another is that factors other than the amount of native vegetation may significantly influence underlying patterns of species occurrence. This highlights a need to be aware of the potential effects of various ecological processes, even when a substantial amount of native vegetation cover remains. Our findings do not rule out the possibility of the existence of threshold relationships. However, irrespective of the choice of measure of predictor variable (e.g., the amount of native vegetation cover), it will often be difficult to detect and estimate threshold responses due to high inherent variability - a characteristic of the vast majority of ecological datasets. Furthermore, even if it is possible to estimate functional (threshold) forms and although they might be useful from an explanatory perspective, in most instances they are likely to be of limited value in a predictive sense. This calls into question the practical significance of the threshold concept. © 2005 Elsevier Ltd. All rights reserved.