Abstract Stable isotope methods are powerful, frequently used tools which allow diet and trophic position reconstruction of organisms and the tracking of energy sources through ecosystems. The majority of ecosystems have multiple food sources which have distinct carbon and nitrogen isotopic signatures despite occupying a single trophic level. This difference in the starting isotopic composition of primary producers sets up an isotopic baseline that needs to be accounted for when calculating diet or trophic position using stable isotopic methods. This is particularly important when comparing animals from different regions or different times. Failure to do so can cause erroneous estimations of diet or trophic level, especially for organisms with mixed diets. The isotopic baseline is known to vary seasonally and in concert with a host of physical and chemical variables such as mean annual rainfall, soil maturity, and soil pH in terrestrial settings and lake size, depth, and distance from shore in aquatic settings. In the fossil record, the presence of shallowing upward suites of rock, or parasequences, will have a considerable impact on the isotopic baseline as basin size, depth and distance from shore change simultaneously with stratigraphic depth. For this reason, each stratigraphic level is likely to need an independent estimation of baseline even within a single outcrop. Very little is known about the scope of millennial or decadal variation in isotopic baseline. Without multi-year data on the nature of isotopic baseline variation, the impacts of time averaging on our ability to resolve trophic relationships in the fossil record will remain unclear. The use of a time averaged baseline will increase the amount of error surrounding diet and trophic position reconstructions. Where signal to noise ratios are low, due to low end member disparity (e.g., aquatic systems), or where the observed isotopic shift is small (≤ 1‰) the error introduced by time averaging may severely inhibit the scope of one's interpretations and limit the types of questions one can reliably answer. In situations with strong signal strength, resulting from high amounts of end member disparity (e.g., terrestrial settings), this additional error maybe surmountable. Baseline variation that is adequately characterized can be dealt with by applying multiple end-member mixing models.