[Processing a hierarchy of stimulus pattern of 3 dimensions: a test of Navon's precedence hypothesis]

The purpose of the experiments was to test generalizations derived from Navon's (1977) perceptual precedence hypothesis (global features are extracted earlier than local ones in feature analysis). Experiment 1 compared the identification of a level in hierarchically organized stimulus patterns (small letters forming larger letters on higher levels) between two- and three-level structures. Results showed an effect on reaction time (RT) of the imbedment of a level within a structure of sub- and superordinated levels: The same level will be identified more quickly if it is the local level of a two-level letter than if it is the medium level of a three-level letter. This effect can be explained by the loss of the most easily noticed privileged structural position. This result and the observed longest RT for identifying the middle level in a three-level letter contradict expectations derived from Navon's (1977) hypothesis of perceptual precedence. Experiment 2 tested a variant of the precedence hypothesis by Kinchla and Wolfe (1979). On the basis of this hypothesis it can be expected that RT for identification of the middle level should converge toward those for the other levels if the middle level is presented at about 2 degrees. This hypothesis could not be confirmed. Results are discussed in the context of a proposed attentional explanation. This explanation suggests that a structural level is noticed more easily if it segregates perceptually from the other levels. In this respect, the global level usually but not necessarily has an advantage.