IN THEIR INFLUENTIAL EMPIRICAL PAPER, Roll and Ross [11] state, We think that in many discussions of the CAPM, scholars were actually thinking of the APT and of process (1) [the factor model] with just a single factor. In light of recent developments in the asset pricing literature, I think it can now meaningfully be suggested that, in many discussions of the APT, scholars may have actually been thinking of a multi-beta interpretation of the CAPM. Indeed, much can be said in support of each of these perspectives, for a single paradigm is emerging which integrates fundamental aspects of both the CAPM and the APT. Broadly speaking, the equilibrium (CAPM) paradigm emphasizes the role of covariance between asset returns and an endogenous preference-based aggregate', while the arbitrage (APT) paradigm focuses on the covariance between asset returns and factors in the return generating process. The emerging equilibrium factor paradigm encompasses each of these relations as well as a third-covariance between the factors and the equilibrium aggregate. My thesis in Shanken [17] was that: (i) the arbitrage paradigm has not produced a refutable hypothesis and that (ii) recognition of the role of the factor/aggregate covariance relation appears to be essential if positive economic significance is to be attributed to asset pricing empirical work. Dybvig and Ross (D-R) attempt to resolve the empirical indeterminacy of the arbitrage paradigm by appealing to the condition that residual variance should not be too large. I explain, in Section II, that this condition is inherently ambiguous from an empirical perspective and does not permit the formulation of a precise testable implication. The equilibrium factor models which D-R discuss do yield testable restrictions on parameters in the return-generating process. However, consistent with my second point above, these restrictions involve joint hypotheses that cannot be verified without observing the entire universe of assets. Thus, contrary to their claim, considerations similar to those in Roll's [10] critique are indeed relevant to the testing of these models. This and other similarities between the equilibrium-APT and the CAPM are discussed in Section III. I begin by clarifying some basic terminology.