Much research in syntax has argued that certain domains, in current terms phases (Chomsky 2000, 2001, a.o.), are unique in only allowing constituents in their edge to be accessible by later operations. Using fieldwork data on cross-clausal scrambling in Balkar (Turkic), this paper argues for the following concepts about phases and their edges: #1: CPs are phases that allow multiple edge constituents, provided that tucking-in applies (Richards 1997, 1999, a.o.); #2: When a phase has multiple specifiers, their relative structural height determines the order in which they can be accessed by syntactic operations (Bošković 2016); #3: DPs are phases that uniquely lack a position for successive-cyclic A-bar movement from them (Bosque & Gallego 2014, Reeve 2018, van Urk 2019). This paper goes on to show that these considerations also make correct predictions about the syntactic behavior of the different varieties of embedded subject in Balkar, when combined with a particular proposal about the (anti-)locality of movement.