This paper presents a critical evaluation of 24 studies on the association between type of infant feeding and cognitive development published over the past 20 years. Validity and generalisability of study findings were assessed according to three methodological standards: clearly defined outcome, specification of partial vs. exclusive breast feeding and control of confounding. Only six of the 24 investigations met all three standards. The most frequent study flaw was failure to distinguish between partial and exclusive breast feeding. Studies which made this distinction found larger IQ advantages to breast-fed infants than studies that did not. Four of the six studies meeting all three standards found an advantage in cognitive development to breast-fed infants of the order of two to five IQ points for term infants and eight points for low birthweight infants. We conclude that the question of whether breast feeding and formula feeding have differential effects on cognitive development has not yet been comprehensively answered. Research to date provides only an indication of the effect of relatively brief durations of partial breast feeding and even briefer durations of exclusive breast feeding. Future studies should measure breast feeding as a continuous dose-type variable, examine longer durations of breast feeding and control for a full range of confounders using techniques that deal appropriately with multicollinearity.