This paper presents original results on recognition, confusability and prosodic analyses of spoken digits over the local and long distance telephone lines. A database for the recognition experiments consisted of 300 speaker records of 10 digits from all across Slovenia. Confusion matrix analysis of the best test set recognition performance identified the most confusable word pairs Furthermore, the prosody of 10 Slovenian digits was studied in detail in the context of dialectal structure of Slovenian language.