The purpose of this study is to measure speech perception ability in presbycusis, using zero-error consonant-vowel (CV) syllables as a function of signal-to-noise ratio (SNR). Seventeen elder people with symmetric mild-to-moderate sensorineural hearing loss were tested with 14 syllables spoken by two different speakers at four different SNRs. Error rate and entropy calculation were used to analyze data. The results show that both error rate and entropy significantly increased as noise increased. In particular, /ba, va, ga, na/ were affected by noise level in both error rate and entropy, yet /za, Ʒa/ were only in entropy. We conclude that evaluating speech perception ability using zero-error CV syllables might scrutinize the unique characteristics of presbycusis. We also find that entropy analysis is very useful in even including response’s uncertainty which is not found in the typical error rate analysis.