In this study, the potential of Hidden Markov Model (HMM) for the development of Amharicrnspeech recognition system has been investigated and in the course of building the recognizer thernpopular toolkit Hidden Markov Model Toolkit (HTK) was used.In the process of building the recognizer, the speech data is recorded at a sampling rate ofrn16KHz and the recorded speech is then converted into Mel Frequency Cepstral Coefficientrn(MFCC) vectors for further analysis and processing.Since large vocabulary systems are envisaged, sub-word modeling is pursued. Sub-wordrnmodeling refers to technique whereby one HMM is constructed for each sub-word unitrn(phoneme, trip hone, syllable, etc .). Phonemes, tied-state trip hones and CV-syllables have beenrnconsidered as the basic sub-word units and have been used to build phoneme-based, tied-staterntrip hone based and CV -syllable based recognizer, respectively. In this study, an extensible l70-word vocabulary is constructed and both speaker-dependent andrnspeaker-independent models are built using 15 speakers (8 male and 7 female) in the age range ofrn20 to 30. Five untrained speakers who had no involvement in training the models are also used torntest the speaker-independent models.The results obtained are promising and have shown the potential of tied-state trip hones as goodrnsub-word units for Amharic. In fact, phonemes also have produced encouraging recognitionrnperformance. Even though CV -syllables appear to be more convenient for Amharic, this researchrnhas not proved that and is underscored for Fruther research.