The purpose of this study was to determine if speech language pathologists could reliably identify different perceptual variants of adductor spasmodic dysphonia (ADSD) using a classification system proposed by Bastian (2009). Audio samples of sustained vowel phonation, sentence production, passage reading, and conversation were pre-recorded from 16 patients with a primary diagnosis of spasmodic dysphonia. Four speech language pathologists specializing in voice disorders listened to the audio samples and identified the ADSD variant, the presence of co-occurring tremor, and overall severity of the voice disorder. Re-rating of half of the audio samples was conducted two weeks later. Inter-judge reliability for the diagnosis of ADSD variants was fair (©§ = 0.25). All four judges agreed on variant 19% (3/16) of the time, three judges agreed 62% (10/16) of the time, and two judges 19% (3/16) of the time. Inter-judge reliability agreement for identifying co-occurring tremor was fair (©§ = 0.35). All four judges agreed on the presence of co-occurring tremor 44% (7/16) of the time. Three judges agreed 50% (8/16) of the time. Inter-judge reliability for assessing overall severity was moderate (©§ = 0.41). Four judges agreed on overall severity 38% (6/16) of the time. Three judges agreed 19% (3/16) of the time. Two judges agreed 38% (6/16) of the time. Judges disagreed on overall severity 6% of the time (1/16). Further analysis revealed three sources of diagnostic discrepancies: 1) the co-occurrence of multiple perceptual characteristics that correspond to multiple variants, 2) the ambiguity in identifying co-occurring tremor, 3) misidentification of perceptually similar characteristics. Relevant spectrographic evidence was presented. Clustering and categorizing perceptual characteristics were proposed as a more reasonable approach to capture perceptual variation of ADSD to optimize clinical diagnosis and acoustic research.