Perceptual units are involved in the reading process in a particular way. Reading starts with patterns of visual perception visual stimuli in combination with either previously learned information or a strategy for selectively attending to part of the stimuli. Whether the visual pattern selected for deeper information processing is a letter, a spelling pattern, or a word group depends not only on the stimulus display (perceptual unit) but also on the dominant contextual mechanism. Research indicates that these mechanisms are driven by "top-down" hierarchical operations on patterns, phonological events, and semantic/syntactic events; the more sophisticated learning strategies are used for more sophisticated tasks to keep processing time at a minimum. Six models of processing strategies have been used to explain the operations on perceptual units, and their interchangeability suggests that reading strategies specially for word recognition and decoding might be learned in sequence and then used for specific reading tasks. The way individuals learn to recognize the first perceptual units at the new processing levels (new contextual mechanisms) remains a difficult issue that has yet to be resolved. (Discussion following presentation of the paper is included.) (RL)