A recent report of an association stimulus quality interaction in Spanish was construed as support for the use of the lexical route in word recognition in a shallow orthography (Bajo, Burton, Burton, & Can as, 1994). Because the interaction occurred against the backdrop of latencies in excess of 1 sec (and was not replicated in Bajo and co-workers' second experiment), the results must be considered equivocal. A naming experiment in Serbo-Croatian, a similarly shallow orthography which has not shown the interaction heretofore, focused on stimulus onset asynchrony (SOA) of prime and target. The interaction was absent at an SOA of 600 msec, replicating our previous results, but was reliable at an SOA of 40 msec. A second experiment in Serbo-Croatian followed Bajo and co-workers' experiments in that the task was lexical decision and the SOA was 800 msec. The experiment's primary feature was the use of phonologically ambiguous targets to enhance clean-up processes potentially responsible for the interaction. No interaction was observed in either the latency or error measures. The results are discussed in relation to dual-route theory and the phonological coherence hypothesis, and in relation to possible differences among the English, Spanish and Serbo-Croatian orthographies.