Abstract.-Patterns preserved in the fossil record are of the highest importance in addressing questions about long-term evolutionary processes, yet both the description of pattern and its translation into process can be difficult. With respect to gradual phyletic change, we know that randomly generated sequences may exhibit characteristics of a ''trend''; apparent patterns, therefore, must be interpreted with caution. Furthermore, even when the claim of a gradual trend can be statistically justified, interpretation of the underlying mechanisms may be challenging. Given that we can observe populations changing rapidly over tens or hundreds of years, it is now more difficult to explain instances of geologically gradual (as opposed to punctuated) change.Here we describe morphologic change in two bivalve lineages from the late Miocene Lake Pannon. We evaluate change according to the model-based methods of Hunt. Both lineages exhibit size increases and shape changes over an interval of nearly 4 million years. Size and two shape variables in the conjungens lineage are best fit by a model of directional evolution; remaining shape variables mostly conform to unbiased random walks. Body-size evolution in the diprosopum lineage is also significantly directional but all shape variables are best fit by the unbiased random walk model; the small number of sampling intervals available for this lineage (n 5 6) makes determination of the actual pattern more difficult. Model-fitting results indicate that the parallel trajectories of increasing log shell height over time in the two lineages can be accounted for by an underlying trend shared by both lineages, suggesting that the size increases may be a shared response to the same cause. The pace of phenotypic change, measured as Lynch's D, is slower than the neutral expectation for all size and shape traits.Our examples illustrate well the paradox of gradualism; the sequences exhibit significant directional morphological evolution, but rates of change as measured over the long-term are apparently too slow for directional selection or even drift to be the cause. Viewing long-term phenotypic evolution in terms of populations tracking peaks on adaptive landscapes is useful in this context. Such a view allows for intervals of directional selection (during times of peak movement-resulting in the overall trends we can detect) interspersed with intervals of stasis (during times of peak stability-resulting in overall changes that appear to proceed more slowly than the neutral expectation). The paradox of gradualism thus reduces to (1) peak movements and their drivers, which are not restricted in rate as are population-genetic drivers, and (2) the maintenance of stasis, on which no consensus exists.We can identify no environmental parameter in the central European Neogene that exhibits consistent change across the interval of gradual morphologic change. It may be that in Lake Pannon the long-term persistence of generally ameliorating conditions (plentiful resources and habitat space, few pr...