Three hypotheses on factors determining performance in a complex water maze were tested in rats prenatally exposed to phenytoin. The hypotheses were: 1) that increasing maze complexity would better differentiate experimental effects; in particular, that an expanded version of a maze originally described by Biel would better differentiate groups than Biel's original design; 2) that path order is an important factor determining performance; specifically, that path sequence AB would better differentiate experimentals from controls than the opposite order (sequence BA); and 3) that repeated trial failures interfere with learning, a problem putatively prevented by employing assisted (i.e., guided) escape. The specific prediction was that rats tested with assisted escape would learn faster and produce better group differentiation than rats tested with unassisted escape. Pregnant female Sprague-Dawley CD rats were gavaged on days 7–18 of gestation with propylene glycol alone (Control) or containing 100 or 200 mg/kg of phenytoin. Straight channel swimming trials followed by maze trials were begun on separate male/female offspring pairs from each litter on postnatal days 50, 70, or 90. The results confirmed hypothesis 1, i.e., the more complex maze better differentiated phenytoin-related group differences. This was true regardless of whether the phenytoin rats exhibiting circling were included in the analyses or not. The results disconfirmed hypothesis 2, i.e., that path order AB would better differentiate the groups than path order BA. Rather, the data supported the alternate hypothesis, that path order was not a significant determinant of prenatal drug-related maze deficits. This was unchanged regardless of whether phenytoin offspring exhibiting circling were or were not included in the analyses. The implication is that path B alone was sufficient to detect phenytoin's effects on maze performance. Finally, the overall results disconfirmed hypothesis 3, i.e., assisted escape failed to differentiate groups any better than unassisted escape regardless of whether circlers were or were not included in the analyses. However, when more detailed analyses were performed only on rats that failed to find the goal on the first or second trials of path B, an assistance effect was noted. The effect (fewer errors) only occurred in the controls and only on the first two trials of path B. It was concluded that assisted escape was, at best, a minor contribution to improved maze performance.