We formulate a well-posedness and approximation theory for a class of generalised saddle point problems with a specific form of constraints. In this way we develop an approach to a class of fourth order elliptic partial differential equations with point constraints using the idea of splitting into coupled second order equations. An approach is formulated using a penalty method to impose the constraints. Our main motivation is to treat certain fourth order equations involving the biharmonic operator and point Dirichlet constraints for example arising in the modelling of biomembranes on curved and flat surfaces but the approach may be applied more generally. The theory for well-posedness and approximation is presented in an abstract setting. Several examples are described together with some numerical experiments.In particular we have in mind an example of the formwhere c 0 , c 1 are bounded but c(·, ·) is not coercive and a linear map T : W 1,q (Γ) → R N defined by (T η) j := η(X j ), j = 1, ..., N with X j ∈ Γ, j = 1, ..., N .1.1. Background. The study of saddle point problems is well documented, [4,16], with many applications, for example in fluid mechanics, [17], or in linear elasticity, [2]. Note that in many of the cases in which m = 0, the authors require some strong assumptions on c, at least positive semi definite, see [4,9,21]. The system (1.1) is an extension to that considered in [13]. The extended system is posed over an affine subspace of X × Y , rather than over the whole space. If in (1.1), we were to seach for a solution in X 0 × Y , the first equation were to be considered with test functions in X 0 and the third equation to be dropped, this recovers the abstract system studied in [13]. We will use the assumptions made in [13] together with an additional assumption to handle the constraint. The assumptions will be given in Section 2.In [7], the authors consider the approximation of Stokes flow by penalising the incompressibility condition. In particular, they show that the penalty terms approximate the pressure. We also consider an abstract problem with penalty and show that, in our setting, the penalty terms converge to the Lagrange multiplier associated to the constraints. Further to this, we show estimates between the solution to the problem with penalised constraint and the solution to the problem with enforced constraint. An abstract finite element theory with error bounds is presented. The results of this paper extend those of of [13] where for example, in [13, Section 6 and 7] it is shown that the well posedness theory in that paper may be applied to a problem with penalised point constraints without consideration of the convergence with respect to the penalty parameter.The motivation for the abstract setting is to handle second order splitting for a class of fourth order surface PDEs with point Dirichlet constraints arising in the modelling of biomembranes, [12]. The setting of [13] may be directly applied to the penalty approximation for fixed penalty parameter but does not handle the hard...