In cognitive skill acquisition it is generally assumed that the decrea
se of latencies found during training follows a power function. This i
mplies that the higher efficiency of performance found with increasing
practice is explained by assuming data-driven, unconscious learning m
echanisms. In contrast, some empirical findings within this field sugg
est discontinuities. Up to now, the geneses of such discontinuities re
mains rather unclear Therefore, the goal of the current experiment was
to examine whether discontinuous learning curves indicate the generat
ion of a consciously available and more efficient strategy. For this p
urpose 30 subjects were trained over 8 practice blocks, followed by tw
o transfer blocks. The results show that subjects with discontinuous l
earning curves compared to subjects without such discontinuities (a) w
ere faster at the end of training, (b) reached in the second transfer
block the level of latencies of the last practice block, and (c) after
the experiment, verbalized an easier strategy than that given by the
instruction. Taken together, these results suggest that indeed discont
inuities indicate the generation of an easier, consciously available s
trategy. The implications of these results are discussed with respect
to their meaning for conscious and unconscious processes within the fi
eld of cognitive skill acquisition.