Semantic selectivity describes the phenomenon that selective processing is influenced by semantic relatedness. The load-insensitivity criterion of automaticity, which states that automatic processes are insensitive to variations in information load, and the intentionality criterion of automaticity, which states that automatic processes are not affected by the observer's intentions, were examined in five experiments in order to determine whether semantic selectivity reflects automatic processes;A priming-search task was developed in which participants reported the target (defined by initial letter) in a briefly presented backward masked search display consisting of a target and distractor. The search display was preceded by a prime that was associatively related to one of the words in the search display on half of the trials: When a related prime was presented, it was related to the target on half the trials;In Experiment 1, semantic selectivity was demonstrated using the priming search task: There was a semantic facilitation effect with higher report of semantically primed words that unprimed words. In Experiment 2, load was manipulated by presenting search displays comprised of one target and either one or three distractors: The semantic facilitation effect was obtained and it did not vary as a function of load. Thus, semantic selectivity met the load-insensitivity criterion of automaticity. In Experiment 3, participant's intentional shifting of spatial attention was manipulated by presenting a central arrow cue 150 msec before the search display. The central cue validly indicated the target on 75% of the trials. The semantic facilitation effect was found and it did not depend upon central cueing; thus, semantic selectivity satisfied the intentionality criterion of automaticity;Experiments 4 and 5 replicated Experiments 2 and 3 using response latency rather than report accuracy as the dependent variable. The priming-search task was modified such that search displays were presented until a response was made and were not backward masked. The response latency data mirrored the response accuracy data. Thus, semantic selectivity seems to reflect automatic processes but it does not seem to reflect the automatic capture of spatial attention. Recent neuropsychological data are discussed that support a multiple-attentions approach.