The regexes in my previous post may be borked due to email client stripping stuff out. Let's just assume that regex_consonant and regex_vowel are well formed patterns. The code below seems to be getting closer. It returns a plausible count for number of matches of regex 'expression' in string 'word': string get_m(string word) { regex c("^" + regex_consonant); regex v(regex_vowel + "$"); string re = "(" + regex_vowel + "+" + regex_consonant + "+)"; regex expression(re); string replacement = ""; word = regex_replace(word, c, replacement); word = regex_replace(word, v, replacement); boost::sregex_token_iterator itr(word.begin(), word.end(), expression); boost::sregex_token_iterator end; int size = 0; for(; itr != end; ++itr) { size++; } string s; stringstream out; out << size; s = out.str(); return s; } Is boost::sregex_token_iterator the appropriate function for subsequently counting the number of matches found in a given string? Also, don't hold back on how bad this code is, I'm here to learn! Mick