Amharic is the official language of Ethiopia, which is characterized by very largernmorphological forms of words. This thesis is an investigation of the possibility ofrndeveloping an Automatic speech recognition system (ASR) for Amharic usingrnsynthesized Amharic speech generated through concatenation of prerecordedrnmorphemes, can be used to train a hidden markov model (HMM) based ASR system.rnThe development of HMM based ASR system requires identification of all possiblernwords and a construction of text and speech corpora containing multiple samples ofrnthe words to be recognized by the system. These data are then used as training sets inrnthe development of the models, the final objective being the construction of HMMrnmodels for each recognition unit. Since there are a large number of morphologicalrnforms for the words in Amharic, the effort of collecting the Amharic words forrnconstructing the text corpus and the recording and labeling of the same words for thernspeech corpus is extremely difficult. This thesis demonstrates that by developing anrnautomatic morphological expander, the effort of developing the text corpus is reducedrnto a manageable level. Additionally, a significant reduction in the speech corpusrndevelopment is achieved by using machine generated speech for training the HMMrnmodels of the ASR system. These reductions in the development efforts of the textrnand speech corpora greatly reduce the most prominent of the obstacles in developing arngeneral purpose Amharic speech recognizer.rnThe 62.37% word accuracy for naturally recorded speech indicates that usingrnsynthetic speech for training at least 62% of the words are correctly identified andrnsuggests that with synthetic speech some level of recognition is possible, giving thernimputes for more research in finding ways to increase this accuracy.