Models of optimal parental investment predict that variation in certai
nty of paternity can affect the optimal level of paternal investment w
hen a male's expected paternity in different nesting attempts is not f
ixed throughout his lifetime. Several attempts to test this prediction
experimentally in monogamous birds have failed to induce a reduction
in care by males. This may be because the method used, detaining males
, is a poor model for what happens when a male's certainty of paternit
y is naturally reduced. We caught and detained female collared flycatc
hers Ficedula albicollis for 1 h immediately after laying on one or tw
o occasions in an attempt to induce variation in certainty of paternit
y for the males they were mated to. By capturing females immediately a
fter laying we hoped to exploit the existence of an ''insemination win
dow'' since males should be very sensitive to female absence during th
is period. The general effect of the experimental manipulation was con
sistent with reduced certainty of paternity: males responded by reduci
ng their level of paternal care to nestlings, and males mated to femal
es that had been caught on one morning fed nestlings significantly les
s often and made a smaller share of feeding visits than males mated to
control females. The effects of the experiment were generally weak, h
owever, and we argue that certainty of paternity may be fixed well bef
ore egg laying, in which case experimental manipulations are unlikely
to have large effects. It is difficult to predict the effects of natur
al variation in certainty of paternity on levels of male paternal care
because differential allocation by females mated to attractive males
may act in the opposite direction.