In this paper, speech recognition techniques are applied to automatically evaluate children's performance in a syllable blending task. Word verification is performed to filter out utterances pronounced incorrectly. For valid words, forced alignment is applied to generate syllable segmentations and produce the corresponding HMM log likelihood scores. Normalized spectral likelihoods and duration ratio scores are combined to assess the overall quality of children's productions. Speaker-specific information is further incorporated to optimize performance. Experimental results show that the automatic system correlates well with those of teachers, but requires no