Quantitative associations between animals and vegetation have long been used as a basis for conservation and management, as well as in formulating predictions about the influence of resource management and climate change on populations. A fundamental assumption embedded in the use of such correlations is that they remain relatively consistent over time. However, this assumption of stationarity has been rarely tested – even for forest birds, which are frequently considered to be 'indicator species' in management operations. We investigated the temporal dynamics of bird-vegetation relationships in young Douglas-fir (Pseudotsuga menziesii) forests over more than a decade following initial anthropogenic disturbance (commercial thinning). We modeled bird occurrence or abundance as a function of vegetation characteristics for eight common bird species for each of six breeding seasons following forest thinning. Generally, vegetation relationships were highly inconsistent in magnitude across years, but remained positive or negative within species. For 3 species, relationships that were initially strong dampened over time. For other species, strength of vegetation association was apparently stochastic. These findings indicate that caution should be used when interpreting weak bird-vegetation relationships found in short-term studies and parameterizing predictive models with data collected over the short term.