Intelligibility of spoken letters of the alphabet

Abstract

A subjective experiment was carried out to assess the effect of spelling context on the ability of listeners to identify spoken letters of the alphabet. Individual letters, spoken by a male talker and a female talker, were concatenated to form strings of up to ten letters representing the spelling of a last name (surname) and one or two initials. Three spelling contexts were observed. First, real names were selected at random from the 20 000‐entry Bell Labs telephone directory. Second, strings were constructed using trigram statistics for letter sequences tabulated from the same directory. Third, letters were selected at random to form strings. In addition, three masking noise conditions were observed: no noise, low‐level, and high‐level broadband additive masking noise. The results show that the listeners made fewer errors identifying letters imbedded in real name or trigram strings than in random letter strings. The effect is especially pronounced in the presence of masking noise. Confusions among the spelled letters are also tabulated.