In English, prosodic parameters play a major role at two main levels. First, they indicate the intonation at the level of the utterance by marking the distinction between sentence types (statements vs questions) and they are related - although more or less directly- to the informational and grammatical structures of the utterance. Secondly, prosodic cues also contribute to marking the stress pattern at the level of the word (word stress or lexical stress). Even if it is useful to dissociate these two levels theoretically, when looking at their phonetic implementation in an utterance, it soon appears that the exact same prosodic cues are used (namely fundamental frequency, duration, and intensity). Contrary to what happens in tone languages, there is no pre-set prosodic configuration attached to each word in English. Yet, words in discourse retain a relative accentual independence even though the exact prosodic implementation of word stress depends on the specific intonational context expressed in a given utterance (Pierrehumbert, 1980). In French, stress pertains to the level of the group of words rather than to the individual word, which has no real accentual autonomy. Therefore, it is not surprising that French learners of English are faced with a major challenge: how to ensure the marking of lexical stress while, at the same time, using the same prosodic cues to indicate the intonational structure of the utterance. My hypothesis is that some intonational contexts impose a bigger constraint on French learners of English than others. These particularly challenging contexts are the final position at the boundary of non-final clause, or the boundary of a rising interrogative. Other contexts, like the quotation form or the final position of a statement, are less challenging for the intonational marking of lexical stress. To test my hypothesis, I collected passages of read speech by thirteen upper intermediate/advanced French learners of English along with the same passage read by ten native English speakers. Two trisyllabics carrying primary stress on the second syllable (com'puter, pro'tection) were placed in a series of intonational contexts under observation. The test-words were then extracted and submitted to native English listeners. The perceptual results show that the predicted ‘challenging’ contexts indeed caused substantial instability in the learners’ placement of lexical stress as perceived by native English listeners.