Abstract: A psycholinguistic approach to opacity: Report of a preliminary study in Hebrew Opacity is the strongest argument within theoretical phonology for intermediate levels of representation. Typically, this layered phenomenon does not fit well into optimality theory (OT; Prince & Smolensky 1993; McCarthy & Prince 1993a,b) because it relies on reference to intermediate forms that are not available within a pure OT framework. As a result, much work has been done to fit the theory to the data, as in Benua 1997 (output-output theory), McCarthy 1998 (Sympathy Theory), Sprouse 1998 (Enriched Input Theory), among others. In essence, the field has been split into two camps, those maintaining a parallel framework, and those instituting intermediate levels within OT. One venue in which opacity does not appear to be explored is psycholinguistics. In this paper, I question what speakers are actually doing with opaque forms, and present the results from preliminary experiments in psycholinguistics that are geared toward testing what the abstract phonological representations of opaque forms for speakers, in particular, Hebrew speakers, look like. The preliminary study consists of two priming experiments conducted by 10 adult Hebrew speaking participants (5 literate, 5 non-literate). Hebrew was chosen due to the wealth of opaque data in the language. Two groups of participants were used in order to test whether orthography has an influence on the abstract phonological representation. This is relevant in Hebrew because the writing system reflects all of the sounds in a word that were spoken at one time. The opaque forms, which I hypothesize to be historical remnants that are no longer productive in Hebrew, may be represented differently among literates and non-literates because of the characteristics of the writing system. In this paper, I concentrate on the results from Experiment 1. Experiment 1 considers the presence or absence of glottal stops in opaque forms, and in non-opaque forms in Hebrew. A typical example of opacity in Hebrew is given in 1. (1) UR /deS?/ Epenthesis deSe? Glottal Deletion deSe SF [deSe] Glottal stops in Hebrew never surface in coda position. Therefore, one possible alternative to positing intermediate levels within OT is to test whether there is any psycholinguistic evidence for positing a glottal stop in the underlying representation at all. Unfortunately, there are too few forms like that in 1 in Hebrew, which is another theoretical problem in itself that I will come back to in the conclusion. The low frequency of forms caused me to generalize to evidence of the presence of glottal stops in abstract representations, in general. An important note here is that while [deSe] is opaque, there are related forms, such as [deS?e] where the glottal stop surfaces regularly in the onset position. This pair of words is what forms the cornerstone of generative grammar, and what has traditionally served as an argument that forms like that in 1 have an underlying glottal stop. The experiment was set up as follows. Participants heard a prime, followed by a target, and the reaction time of word recognition was recorded. In the first experiment, participants had access to the following forms. (2) a. /CaCa?/ [CaCa] b. /CaCa/ [CaCa] c. /Ca?aC/ [Ca?aC] d. /CaCaC/ [CaCaC] e. /CaC?a/ [CaC?a] f. nonse words Above is one of two sets of forms used in Experiment 1. There are a number of predictions to be made from the interactions of the forms in 2. Based on established priming experiments, 2c and 2e should prime each other. If a glottal stop is present in the UR of 2a, this form should then prime 2c. If a glottal stop is not present in the UR of 2a, participants primed with 2b should recognize 2a faster than 2d because there would be exposure to one less segment, causing processing time to be shorter. If a glottal stop is not present in the UR of 2a, there should be no difference in reaction times for 2c or 2d when primed with 2a. The results from my experiment suggest that 2c and 2e are in fact in a priming relationship and that there is no evidence, based on reaction times, to support an analysis that posits underlying glottal stops for forms like 2a. If these preliminary results are supported by further experimentation, questions for theoretical phonology are numerous. To consider just a few, what is the abstract representation of 2a, then? Have the forms in question been lexicalized? Does opacity exist at all, or is it the result of historical changes in languages that should no longer be considered as part of a speaker's synchronic grammar? These questions, among others, are considered. Symbol Key ? glottal stop S voiceless alveopalatal fricative