This paper investigates and formally compares the expressive power of dimensional (i.e., spatial, temporal, and spatio-temporal) query languages, where the dimensional extensions are supported in terms of ADTs (abstract data types). There are basically two approaches to the design of dimensional ADT extended query languages. One approach, by definition, adds semantics by interpreting an ADT attribute value associated with a database fact as an intrinsic (i.e., builtin) relationship with an underlying space. The other approach treats ADT attribute values as conventional attributes, where the dimension semantics (and space) associated with a database fact is an extrinsic property and controlled fully by the user. The comparison framework is based on the relational algebra (RA) and a single ADT extension to RA. Two comparison criteria of semantic equivalence also are defined. The one criterion of strict equivalent expressions imposes equal results, whereas the other (relaxed) criterion of snapshot equivalent expressions imposes equal snapshot results. For the strict criterion a certain class of intrinsic ADT extended languages is semantically richer than the set of corresponding expressions of a pure (i.e., extrinsic) ADT extended language. This is due to the properties of the built-in dimension support. For the relaxed criterion the same intrinsic language class is shown snapshot equivalent with corresponding expressions of the pure ADT extended language class. However, there is a class of expressions which relates database facts of non-intersecting subspaces, that is expressible only by the pure ADT language. In general, and despite differences, one language approach is not found strictly superior to the other. Rather, practically, the findings indicate multiapproach designs for user-level oriented query languages. Moreover, the findings also informally indicate that by extending the framework, e.g., allowing multiple orthogonal dimension ADTs, more involved problems arise, such as a kind of indeterminism of pure ADT extended languages, i.e., user-choices influence results of otherwise orthogonal dimensions.