This study investigated the impact of lipid extraction, CaCO3 removal and of both treatments combined on fish tissue δ13C, δ15N and C:N ratio. Furthermore, the suitability of empirical δ13C lipid normalization and correction models was examined. δ15N was affected by lipid extraction (increase of up to 1.65‰) and by the combination of both treatments, while acidification alone showed no effect. The observed shift in δ15N represents a significant bias in trophic level estimates, i.e. lipid-extracted samples are not suitable for δ15N analysis. C:N and δ13C were significantly affected by lipid extraction, proportional to initial tissue lipid content. For both variables, rates of change with lipid content (ΔC:N and Δδ13C) were species specific. All tested lipid normalization and correction models produced biased estimates of fish tissue δ13C, probably due to a non-representative database and incorrect assumptions and generalizations the models were based on. Improved models need a priori more extensive and detailed studies of the relationships between lipid content, C:N and δ13C, as well as of the underlying biochemical processes. © 2008 The Authors.