Subjects in three experiments read texts describing the locations of l
andmarks in a fictitious town. Later they drew sketch maps and verifie
d sentences describing the relative locations of the landmarks. We pre
dicted that subjects would develop mental models of the town that were
organized around important landmarks (''anchors''), as are cognitive
maps constructed through real-world navigation. More specifically, we
expected that landmarks used in the text as reference points for descr
ibing the locations of some other landmarks would emerge as anchors in
the subjects' cognitive maps and would consequently be recalled more
accurately. Results showed that subjects represented such reference-po
int landmarks more accurately than they did the locations of other lan
dmarks. This effect was independent of (1) the perspective from which
the text was written (route or survey); (2) whether or not a map was p
resent at learning; (3) the order of information in the text (linear o
r anchors-first), and (4) the amount of information available to the s
ubjects while drawing sketch maps (the full text, the landmark names o
nly, or no information).