Analysis and recognition experiments of nasals, liquids and fricatives for a 36 word English vocabulary, based on the alphabet and digits are reported. Speaker independent and mother tongue independent recognition tests have been conducted on 40 speakers, different from the other 40 speakers whose speech has been used for system training. Half of the speakers (for training and recognition) are non English speaking people, are not bilingual and come from different countries. Specific and simple analyses were designed for fricative and nasal/liquid recognition, yielding a 98.2 % recognition rate on fricatives and 90.5 % on nasals/liquids. It shows that it is possible to obtain good performance on difficult tasks with simple and specific analyses. These have been integrated in an automatic speech recognition system described in previous publications.