Some studies in experimental pragmatics have concluded that scalar inferences (e.g., ‘some X are Y’ implicates ‘not all X are Y’) are context-dependent prag- matic computations delayed relative to semantic computations. However, it remains unclear whether strong contextual support is necessary to trigger such inferences. Here we tested if the scalar inference ‘not all’ triggered by some can be evoked in a maximally neutral context. We investigated event-related potential (ERP) amplitude modulations elicited by Stroop-like conflicts in par- ticipants instructed to indicate whether strings of letters were printed with all their letters in upper case or otherwise. In a randomized stream of non-words and distractor words, the words all, some and case were either presented in capitals or they featured at least one lower case letter. As expected, we found a significant conflict-related N450 modulation when comparing e.g., ‘aLl’ with ‘ALL’. Surprisingly, despite the fact that most responses from the same par- ticipants in a sentence-picture verification task were literal, we also found a similar modulation when comparing ‘SOME’ with e.g., ‘SoMe’, even though SOME could only elicit such a Stroop conflict when construed pragmatically. No such modulation was found for e.g., ‘CasE’ vs. ‘CASE’ (neutral contrast). These results suggest that some can appear incongruent with the concept of ‘all’ even when contextual support is minimal. Furthermore, there was no signi- ficant correlation between N450 effect magnitude (‘SOME’ minus e.g., ‘sOMe’) and pragmatic response rates recorded in the sentence-picture verification task. Overall, this study shows for the first time that the pragmatic meaning of some can be accessed in a maximally neutral context, and thus, that the scalar infer- ence ‘not all’ triggered by some should be construed as context-sensitive rather than context-dependent, that is, more or less salient and relevant depending on the context rather than entirely contingent upon it.