• Community relaxation;
  • extinction debt;
  • extinction;
  • habitat fragmentation;
  • landscape;
  • moss microcosm;
  • species–area relation

Estimates of species loss due to habitat destruction are normally based on calculations employing the species–area relation, S = cAz. The validity of this approach is based on the assumption that the value of the exponent (z) defining the slope of the species–area relation in nonfragmented communities is at a steady state and that z is thus a constant. However, departure from such an assumption renders this approach unreliable. Here I report the results from a natural field experiment using “model” bryophyte-based microlandscapes designed to follow the species richness dynamics of microarthropod communities postfragmentation. Community isolation due to fragmentation initiated a delayed community relaxation process and resulted in substantial local extinction. Over the period of the experiment z declined in the control communities and yet remained fairly stable in the fragmented communities. I conclude that predictions of species loss due to habitat fragmentation that do not take into account the fact that often z may not be a constant may lead to error-prone predictions of future species loss.