Abstract. Trivariate C r macro-elements defined in terms of polynomials of degree 8r + 1 on tetrahedra are analyzed. For r = 1, 2, these spaces reduce to well-known macro-element spaces used in data fitting and in the finite-element method. We determine the dimension of these spaces, and describe stable local minimal determining sets and nodal minimal determining sets. We also show that the spaces approximate smooth functions to optimal order. §1. Introduction Let △ be a tetrahedral partition of a set Ω ∈ IR 3 . We denote the sets of vertices, edges, and faces of △, by V, E, and F , respectively. In this paper we study the superspline spaces ∈ C 2r (e), all e ∈ E},where in general we write P d for the d+3 3 dimensional space of trivariate polynomials of degree d. Here s ∈ C ρ (v) means that all polynomial pieces s| T associated with tetrahedra T sharing the vertex v have common derivatives up to order ρ at v. Similarly, s ∈ C µ (e) means that all polynomial pieces s| T associated with tetrahedra T sharing the edge e have common derivatives up to order µ at all points along the edge e.For r = 1 this space corresponds to a macro-element space first introduced in the finite-element literature in [23]. The analogous C 2 macro-element was developed in [16]. Both authors described their elements in terms of Hermite interpolation. It is well known, see Remark 1, that in order to construct similar macro-element spaces for higher values of r, we must work with splines of degree 8r + 1, and we must enforce C 4r supersmoothness at the vertices and C 2r supersmoothness around the edges of △. This suggest a natural set of Hermite data to associate with the element. But it is a nontrivial problem to describe what additional data 1)