The model theory of syllogistic reasoning assumes that reasoners can devise
models of the premises when the middle terms are contiguous. When the arra
ngement of the terms is not appropriate, they have to perform additional op
erations. Our first experiment shows that the time consumed by these operat
ions can be measured in the reading time of the second premise. This effect
is not predicted by other theories of syllogistic reasoning. The second ex
periment replicates the figural bias in a construction task. Both experimen
ts yielded a reliable effect of the number of models on reading time and ac
curacy. The number of models and the figure seem to be independent factors
that do not interact. An analysis of the invalid conclusions produced by th
e participants in Experiment 2 indicates that the invalid conclusions were
frequently consistent with any of the initial models.