The effects of certain linguistic dimensions on auditory blending performance and training were examined. Dimensions included type of phonological context, consonant-vowel or vowel-consonant (CV or VC); units to be blended, syllables or phonemes (S or P); and size of units, single or double. Six ordered 96-word training blends were administered to six groups of 20 preschool children each over a 4-day period. Two days after completion of training the children were given a 32-word random-ordered test consisting of four test words for each of the eight kinds of blends. Results of analysis of variance showed the following: (1) syllables were easier to blend than phonemes, (2) CV breaks were more difficult than VC breaks for phoneme blending, (3) length increased blending ease only in phonemes, (4) VC training was better than CV training, and (5) syllable blending training did not transfer to phoneme blending performance. It was concluded that phoneme and syllable blending involve separate concepts and that easy to hard training within one concept area should produce optimal results on both tasks. Tables and references are included. (Author/MS)