This paper presents a 500 year March-August precipitation reconstruction for the Bavarian Forest region of southeast Germany based on a composite well-replicated data set of 676 living and historical tree-ring width series from Norway spruce. Two versions of the chronology are developed. The cubic smoothing spline (SPL) chronology, standardized with a fixed 80 year spline function, retains decadal and higher frequency variation. The regional curve standardization (RCS) chronology uses regional curve standardization to retain additional lower frequency variation from the same data. Calibration V(1879-1978) of these chronologies with March-August precipitation indicates they explain 40% (SPL) and 34% (RCS) of the variability in the instrumental precipitation record. The SPL reconstruction models the high-frequency variation better, whereas the RCS reconstruction tracks the low-frequency trends more robustly. It suggests that spring-summer precipitation was above the long-term average for the periods 1730-1810 and 1870-2000, about average between 1560 and 1610 and that significantly drier periods occurred during 1510-60, 1610-35, 1660-1730 and 1830-70. The low-frequency trends of the RCS reconstruction during the 19th century were verified by comparison with a regional precipitation series derived from 14 long precipitation records from central Europe. Better verification results were obtained using the original (non-homogenized) records over this interval. These results suggest that some low-frequency variability may have been removed during correction of these early records. Periods of synchronous decadal variability were observed between the SPL series and independent dendroclimatic reconstructions from central Europe. The RCS reconstruction is the first dendroclimatic precipitation reconstruction in Europe to capture low-frequency information. These lone-term trends, however, are difficult to verify owing to the paucity of other proxy precipitation records in central Europe that portray low-frequency information. Further verification and testing of the RCS reconstruction will require the development of additional reconstructions from tree-rings (or other proxy data series) that similarly target loss-frequency variability.