We examined short-term (decadal) and long-term (millenial) processes o
f peat accumulation, and the links between them, in a Sphagnum bog in
continental Canada. A previously published model of bog growth was fit
ted to age profiles of the oxic acrotelm (surface, <60 cm thick) and t
he underlying, anoxic catotelm (210 cm thick). Approximately 5300 year
s of accumulation were represented in a radiocarbon-dated core that ex
tended to the base of the deepest part of the peat deposit. The model
estimated that the overall rate at which material entered long-term st
orage in the catotelm was 60-66 g . m(-2). a(-1). Although the decay r
ate coefficient was near zero, the bog stopped accumulating peat withi
n the past 1500 years, resulting from either a decrease in the rate of
transfer of material from the acrotelm or an increase in the rate of
decay of material at the top of the catotelm. The model of bog growth
estimated recent inputs to the acrotelm (90-930 g . m(-2). a(-1)) that
were twofold higher than published field measurements of aboveground
productivity, and decay rate coefficients (0.005-0.040 a(-1)) that wer
e 10-fold lower than published litter-bag measurements of mass loss. T
he pattern of mass loss over time, approximated from nitrogen concentr
ation data, deviated from the pattern predicted by exponential models
of decay. Calculations of the balance between additions to and losses
from the acrotelm suggest that the amount of material transferred to t
he underlying catotelm differs among microhabitats. Such spatial varia
bility in short-term processes is incompatible with long-term processe
s determining the position of the acrotelm-catotelm boundary. We discu
ss the applicability of the model to continental peatlands and suggest
ways to improve modelling of short-term autogenic processes.