This paper considers phenomena related to embedded interrogatives that do not fit the canonical profile of subordinate clauses. It focuses on restrictions on such noncanonical cases of subordination, here referred to as quasi-subordination, and makes the following claims. There are three points in the interrogative left periphery for building question meaning. The lowest point is CP, where interrogatives are differentiated semantically from declaratives. All embedding verbs that can take interrogative complements, can take CP+WH. The highest point is SAP. When its head is specified SAASK, the question denoted by the interrogative becomes a request for information by the speaker, directed towards the addressee. This is the structure we find in matrix questions (and quotations). In between these two levels is what I call PerspectiveP. Its head PerspCQ introduces PRO, an individual for whom the interrogative CP+WH is a potentially active question. That is, PRO is the perspectival center, the one from whose point of view the interrogative can be a request for information (signaled by the specification CQ for centered question). When PRO is bound by the speaker argument in the Speech Act Phrase, we get a matrix question; when PRO is bound by the subject of a matrix predicate we get quasi-subordination. Quasi-subordination is a hybrid between true subordination (with respect to pronominal interpretation, for example) and nonsubordination (with respect to intonation, for example). Restrictions on quasi-subordination are claimed to be regulated, in addition to standard selectional restrictions, by semantic compatibility between the implied ignorance of the individual who is the perspectival center of the question and the meaning of the embedding clause. Empirical support for this view of the interrogative left periphery comes from a range of phenomena from unrelated languages. While the idea of an articulated left periphery goes back to Rizzi (1997), the details of the present proposal are new. The paper discusses several implications of this view of the interrogative left periphery, connecting the specific claims to similar proposals about other clause types and to developments in our understanding of how complement selection works.