This study extended the work of Hoppe and Kess (1980) by manipulating the degree of possible readings of ambiguous sentences. Twenty four students attempted to resolve four levels of ambiguity either for biased or unbiased sentences. Findings showed that surface ambiguity was easiest to detect, underlying ambiguity less easy, and lexical-same and lexical different ambiguities the most difficult. This order of difficulty across the four levels was found for both biased and unbiased types. Bias made the detection of underlying ambiguity difficult, thus producing a different pattern of ordering of ambiguity detection between the biased and unbiased types. Unbiased surface ambiguity was more quickly detected than the biased one. The findings also showed that many subjects failed to resolve ambiguities, suggesting that linguistic intuition as revealed in ambiguity detection is not shared equally by all speakers of a language as has been assumed by many generative theorists.