Underlying much of the research on intervention effects, in which a quantificational or focusing element preceding a wh-phrase leads to degradedness, is the assumption that these effects are universal, and should therefore follow from basic properties of the grammar. This paper shows that unlike any other language documented until now, Amharic does not generally exhibit intervention effects. It is nevertheless empirically possible and theoretically preferable to retain the idea that these effects are derivative, rather than to consider their presence vs. absence a parameterized feature. Accordingly, two existing approaches to intervention effects are assessed vis-à-vis their ability to account for the exceptionality of Amharic: a hierarchical analysis, following Beck (2006), and an information structural-prosodic analysis based on Tomioka (2007a,b), whereby the effect is read off the linear string. The latter is claimed to better explain the data and correlate with independent aspects of Amharic, thus providing a general argument in favor of non-structural approaches to intervention effects. This analysis is also extended to alternative questions, in which an intervener preceding a disjunctive phrase removes the alternative question reading, allowing the sentence to be interpreted only as a yes/no question. In the process, many hitherto unknown properties of Amharic syntax, information structure, and prosody are brought to light.