The ability of 50 kindergarten children to discriminate and produce the phonemes typically used in early phonic-based reading instruction was investigated in matching-to-sample discrimination task and an echoic production task. The phonemes were presented to each child in isolation and in a word context in both tasks. The average time required to teach each subject was twenty minutes per day for five days. It was found that (1) more discrimination than production errors were made, a difference which was reliable only for the vowels; (2) vowels were easier to discriminate and produce than consonants; and (3) phoneme errors were fewer in words than in isolation, a difference which reached significance only in the production data. Most importantly, further data analysis revealed that frequency of a phoneme in the conversational speech of kindergartners and in the lexicon of the Southwest Regional Laboratory (SWRL) First Year Communication Skills program did not predict articulation difficulty of a phoneme either in words or isolation. The implications of the results for reading pedagogy are discussed. (TS)