Igbo consonant confusion matrix: Issues on phonetic similarity

Abstract

This paper is on Igbo consonant confusion experiment aimed at producing an Igbo consonant confusion matrix. One hundred Igbo-speaking subjects were used for the experiment at the ratio of 1:1 for the control and experimental groups. Whereas the 28 consonants of standard Igbo were presented to the control group in frame in a noiseless environment, same material was presented to the experimental group under the condition of masking noise. The experiment and matrix reveal that the greatest confused consonant is [j]. All the consultants in the experimental group confused this consonant with another sound; hence it bears 0% in the confusion matrix. Conversely, the least confused of all the Igbo consonants is [kw]; followed by [f]; and [k m z] in that order. In the confusion matrix, they scored 94%, 92% and 88% respectively; only 3 people confused [kw] but 4 people confused [f] while 6 consultants confused each of [k m z]. The segments [j] and [Ɣ] have the closest perceptual similarity and therefore often confused perceptually, scoring as high as 56% across themselves in the confusion matrix. This result is surprising since they are not all that phonetically similar. However, we see the phonetic similarity between these confused sounds shown in the matrix - [ʧ] is confused with [t], as much as [l] is confused with [r]; both scoring 40% in the matrix. Also, [kp] is confused with [p] as much as [n] is confused with [l]; each of the pairs scored 28% in the matrix. On the other hand, many consonants were never confused for each other. For instance, none of the fricatives or approximants was ever confused for the voiceless bilabial plosive [p]. We discovered, however, that some consultants could have confused some segments not necessarily because the segments are phonetically similar but because they tended to consider meaningfulness of syllables rather than the segment they actually perceived. This calls for further investigation on segment confusion in tone languages.