In order to construct a Krein-space theory (i.e., a *-algebra of (unbounded) operators which are defined on a common, dense, and invariant domain in a Krein space) the cones of a-positivity and generalized a-positivity are considered in tensor algebras. The relations between these cones, algebraic #-cones, and involutive cones are investigated in detail.Furthermore, an example of a P-functional 0 defined on (C 2 )® (tensor algebra over C 2 ) not being a-positive and yielding a non-trivial Krein-space theory is explicitely constructed. Thus, an affirmative answer to the question whether or not the method of P-functionals (introduced by Ota) is more general than the one of a-positivity (introduced by Jakobczyk) is provided in the case of tensor algebras. § 1. IntroductionThe present investigations are motivated by the algebraic approach to general (axiomatic) quantum field theory (QFT). The formalism introduced by Borchers [8] and Uhlmann [30] works for massive fields as well as for massless or gauge fields and is entirely equivalent to the Wightman axioms. Starting with a tensor algebra E® over some nuclear space E of test functions and a normalized continuous Hermitean functional W^E®' satisfying some further physically motivated conditions (linear program), a QFT is reconstructed by means of the GNS (Gelfand, Naimark, Segal) construction.It is now necessary to distinguish between i) massive fields, and ii) massless or gauge fields. In the case of i) W is taken to be a positive functional (Wightman functional), and thus the space of state vectors becomes a Hilbert space. In the case of ii) there are no-go theorems (see [22], [29]) stating that locality, covariance