The clear message running through the target article and commentaries is that the preference procedure can yield both familiarity and novelty effects depending on the experimental setting, and that this bidirectionality can make it difficult to interpret infants' behaviour in terms of the knowledge driving it. The commentators share our concerns about the interpretation of data obtained from modifications of the original habituation and preference procedures, and provide further examples of cases where uncertainty about the direction of an effect has led to differing conclusions about infants' cognitive abilities.
The commentators usefully add ideas on how to avoid the problem we describe. Our own suggestion was to follow the course of infants' looking over time. In the light of Hunter and Ames' (1988) model of infant attention, we argued that if infants are found to pass from a preference from one stimulus to a preference for the other, the former might more confidently be identified as a familiarity preference, and the latter as a novelty preference. Slater takes issue with this suggestion on two grounds: first, because evidence suggests that the pattern of preferences described by Hunter and Ames is not always found, and second, because the insertion of 'probe trials' testing infants' preferences might interfere with the habituation process. We address these points in turn.
We see the validity of Hunter and Ames' model as an empirical question. The point we wish to emphasise is the importance of assessing its validity in different experimental contexts, as we believe that regularities in the direction of infants' shifts in attentional preferences over time are not only potentially theoretically revealing but may also prove a useful experimental tool. Where a significant preference in a single direction does not allow a definitive conclusion to be drawn, as in Houston-Price (2002), obtaining preferences in both directions over time could solve the problem, if Hunter and Ames' model is correct.
However, as Slater notes, a familiarity effect is not always observed prior to a novelty effect. For instance, Fantz (1964) measured infants' looking times towards familiar and novel stimuli during 10 1-min intervals, and found no period of familiarity preference. Similarly, when Rose and Slater (1983) familiarized 3-and 5month-olds to a visual stimulus for 20 and 10 s respectively, infants showed a novelty preference on subsequent testing with no preceding preference for familiarity. While these studies may appear to fail to support the model of dynamic attentional preference presented by Hunter and Ames, they are also good examples of where the failure to find an effect tells us little about whether that effect