FIGS. 5B and 5C depict embodiments of the automatic statistical method for automatically generating the phonetic challenge list used by the phonetic difficulty metric. FIG. 5B relates to a method of identifying phonemes in a target language (i.e., the language which is being assessed) that may be challenging for non-native speakers of the target language to pronounce. At 530, a large corpus in the target language (e.g., English) is analyzed and at 535 the target language corpus' phoneme distribution (e.g., probability of phoneme n-grams appearing in the corpus) is extracted. Similarity, at 540 a large corpus in the native language (e.g., Korean) of the non-native speaker is analyzed and a corresponding phoneme distribution is extracted at 545. At 550, the phoneme distribution of the target language is compared to the phoneme distribution of the native language, using measures such as the Kullback-Leibler measure. Based on the comparison, at 555 the phonemes or phoneme n-grams that frequently appear in the target language (as indicated by the target language's phoneme distribution) but rare in the native language (as indicated by the native language's phoneme distribution) are included in a phonetic challenge list. Such a phonetic challenge list can be used by a phonetic difficulty metric to measure a text set's phonetic difficulty level for native speakers of that particular native language (e.g., Korean).