At what points in time do "bottom up" and "top down" information influence the word recognition process? In classical two-stage models of word recognition, processing of bottom-up information has to precede the use of top-down, "real world" knowledge, and the two forms of processing do not overlap. My findings, however, support a model in which multiple sources of information can have their effects very early in processing. Some examples from my research that support this view include the following:

• Lexical and sentential priming interact very early in processing (Hernandez & Bates, 1994).
• Very fine-grained timing manipulations affect the size of these two priming effects, and the presence or absence of an interaction (Hernandez, Fennema-Notestine, Udell and Bates, 2000).
• Age of Acquisition is a stronger predictor of object naming than word frequency (Fiebach, Friederici, Mueller, von Cramon, & Hernandez, 2003; Meshyan & Hernandez, 2002).
• Grammatical gender is a robust source of information for word recognition, for single words and for short phrases in Italian (Bates, Devescovi, Pizzamiglio, D'Amico & Hernandez, 1995; Bates, Devescovi, Hernandez & Pizzamiglio, 1996).