Table 2. Proportions of retrieved patterns in the three simulations.

simulation		control	schizophrenia	input-to-EC	EC-to-hip	SEM
lexical disambig-uation	strong vs. weak	0.88	0.81	0.76	0.85	0.04
	equal strength	0.77	0.67	0.55	0.74	0.03
	weak vs. strong	0.7	0.34	0.4	0.55	0.05
list learning	free recall	0.49	0.21	0.14	0.42	0.03
	intrusions	0.02	0.02	0.01	0.03	0.01
	recognition hits	0.88	0.74	0.81	0.80	0.01
	false alarms	0.09	0.16	0.21	0.10	0.01
source monitoring	source hits	0.58	0.42	0.23	0.53	0.02
	source errors	0.007	0.033	0.007	0.028	0.002
	omissions	0.41	0.55	0.77	0.44	0.02
	false alarms	0.024	0.086	0.016	0.061	0.01

For the lexical disambiguation simulation, the proportion of correct meaning retrieval is shown (i.e. the meaning consistent with the sentence context), in conditions in which it was the dominant of two homograph meanings (‘dominant’), the subordinate of the two (‘subordinate’), or one of two homograph meanings of equal frequency (‘equal frequency’). In the list learning simulation, free recall and recognition refer to the retrieval of list objects during a free recall or recognition test. Intrusions are recalled patterns that were not part of the studied list, while false alarms are foils falsely endorsed during the recognition test. In the source monitoring study, what is retrieved is either the correct source (source hit), the incorrect source (source error), or nothing (omission). False alarms here refer to the retrieval of a source for a foil object that had not been presented at study. Data are given for the intact model, the model with reduced input-to-entorhinal module connections (input-to-EC), reduced entorhinal module-to-hippocampus connections (EC-to-hip), or both sets of connections reduced (schizophrenia). To render an idea of the reliability of the model data, the standard error of the mean (SEM) is given for the scores pertaining to the normal model.