Pronunciation is an important subskill in second language learning, therefore worth evaluating. Its quality is commonly assessed in a global, impressionistic way by having learners read aloud. While this allows comparison of examinees' skills, ability to read aloud is a possible confounding variable. An alternative method is to have learners read texts in which only marked elements are judged as correct or incorrect. A study compared the reliability and validity of this and the traditional, holistic method. Pronunciation tests were developed for Dutch learners of French and German that incorporated words in which pronunciation errors occur frequently. The tests were administered to secondary school students (German=26, French=19). Recorded readings were rated in both the traditional way and with the marked-item method by teacher panels to measure reliability. Results suggest the new method can improve evaluation. To determine validity, another experiment simulated regular teacher evaluation of student pronunciation in French by comparing ratings of pronunciation in: (1) spontaneous speech; (2) a traditional read-aloud text; and (3) marked items in a read-aloud text. These results suggest that (3) marked items in a read-aloud text. These results suggest the new alternative is not preferable to traditional, holistic rating because it can not give a general impression of pronunciation quality. Further research is recommended. (MSE)