Arylazide mediated photocrosslinking has been widely used to obtain structural constraints in biological systems, even though the reactive species generated upon photolysis in aqueous solution have not been well characterized. We establish a mechanistic framework for formation of adducts between photoactivated 3-hydroxyphenyl azide and RNA. Tethered to an internal site in an RNA duplex via a 2'-amido linkage, photolysis of the aryl azide yields a cross-strand cross-link. Analysis of the ability of reagents with diagnostic reactivities to intercept formation of this cross-strand cross-link supports the assignment that the photoactivated intermediate is the ketenimine or a ketenimine-derived ring expansion product. Neither the initially produced singlet nitrene nor the subsequently formed triplet nitrene contribute to cross-link formation. Argon matrix and time-resolved solution experiments show that photolysis of free 3-hydroxyphenyl azide releases (in <or=20 ns) either a ketenimine or azepinone intermediate that reacts with nucleophiles. Adenosine, uridine, and guanosine monophosphate nucleotides have approximately equivalent abilities to quench the cross-strand cross-link, indicating that the photoactivated intermediate reacts broadly with functional groups in RNA. The reactive intermediate forms an adduct with adenosine monophosphate when tethered to both an RNA duplex or unstructured single strand; thus, cross-link formation is independent of the local RNA environment. The lifetime of the reactive intermediate generated upon photolysis of free 3-hydroxyphenyl azide in 50 mM Hepes diamine buffer is found to be 60-160 micros or significantly shorter than large scale RNA folding events. RNA-tethered 3-hydroxyphenyl azide cross-linking in aqueous buffer can thus be used with confidence to map structural neighbors, including most dynamic interactions, in RNA.