The minimalist hypothesis of restricted inference processing (McKoon &
Ratcliff, 1992) states that the only automatic inferences to accompan
y comprehension are those that either are based on quickly available i
nformation or contribute to local text coherence. The present paper id
entifies serious shortcomings of the minimalist hypothesis. First, the
definition of minimalism is flawed with reference to the concepts of
automaticity and the two minimalist criteria. Second, minimalism is di
fficult to falsify. Third, the minimalist prediction that readers will
not inferentially link distant text ideas under local coherence is co
ntradicted by existing evidence. Fourth, the minimalist denial that re
aders routinely derive situation models from text is not a necessary i
mplication of the minimalist hypothesis. An alternate, global, analysi
s is that readers maintain coherence at multiple levels of test repres
entation, including the situational level; and that inference processe
s are guided by the collaboration of availability, convergence of acti
vation, and the search for explanations for text outcomes. (C) 1994 Ac
ademic Press, Inc.