This article considers a series of so-called image macro Internet memes and describes them as emerging multimodal constructions depending as much on image as on text, and apportioning roles to images much like constructional slots, for instance to fill in a subject role in a subjectless clause, or even to provide the main clause content to a textually given when-clause. In addition to extant or partially changed linguistic constructions, many examples also depend on specific top text/bottom text division of labor, and crucially rely on frame metonymy, with restricted formal means quickly cueing richly detailed frames (for instance by using iconic images). The popularity of memes, forming series and cycles of iterations and remixes, and their role in raising and maintaining discourse communities appears to be driven by a need to express and reconstrue viewpoints, often starting from ideas, affects or stereotypes presumed to be intersubjectively shared with viewers, whose responses they solicit. This article asserts that a proper description of Internet memes of the type considered demands a construction grammar approach, complemented by a comprehension of viewpoint dynamics in terms of a Discourse Viewpoint Space regulating the network of spaces and viewpoints.