## 1. Precise statement and assumptions The claim is true under the following interpretation. Let \(N\) be a **finite** set of instruments, and let \[ \mathcal F\subseteq 2^N\setminus\{\varnothing\} \] be upward-closed: if \(T\in\mathcal F\) and \(T\subseteq U\subseteq N\), then \(U\in\mathcal F\). We seek a measurement model \[ y=A\theta+Bz, \] with: - an unrestricted scalar target \(\theta\in\mathbb R\); - a finite-dimensional, unrestricted nuisance vector \(z\in\mathbb R^d\); - known matrices \(A,B\); - finitely many scalar reading channels, partitioned among the instruments; - possibly **multiple channels per instrument**; - a nonzero coefficient of \(\theta\) in every scalar channel. A combination \(T\subseteq N\) receives all channels belonging to its instruments. It **identifies \(\theta\)** if any two parameter vectors producing the same available readings have the same target value: \[ A_T\theta+B_Tz=A_T\theta'+B_Tz' \quad\Longrightarrow\quad \theta=\theta'. \] The assertion is that such a system exists with \[ T\text{ identifies }\theta\quad\Longleftrightarrow\quad T\in\mathcal F. \] Important assumptions: - This is exact, noiseless identification—not estimation precision or statistical information. - Nuisance parameters have no bounds, priors, or additional constraints. - “Nonzero direct sensitivity” means a nonzero partial derivative with respect to \(\theta\), **holding the chosen nuisance coordinates fixed**. - There is no bound on the number of channels or nuisance parameters. - The empty combination cannot identify an unrestricted target. Thus excluding it from \(\mathcal F\) is necessary. ## 2. Checking the argument Let \(\mathcal M\) be the inclusion-minimal members of \(\mathcal F\). Because \(N\) is finite, \[ T\in\mathcal F \quad\Longleftrightarrow\quad T\text{ contains some }S\in\mathcal M. \] For each \(S=\{i_1,\ldots,i_k\}\in\mathcal M\), introduce separate nuisance coordinates \(z_{S,1},\ldots,z_{S,k-1}\) and readings \[ y_{S,j}=\frac{\theta}{k}+z_{S,j}\quad(j