This paper explores an approach to root and stem allomorphy that does not make use of context sensitive rules (i.e., secondary exponence) and relies on primary exponence only. In such a system, each feature is referenced by realisation rules only once and multiple exponence is eliminated. The study applies this system to the phenomenon of root and stem allomorphy in declension systems (McFadden in Glossa 3:8.1-36, 2018, Christopoulos & Zompì in Natural Language & Linguistic Theory 1–31, 2022). The paper argues that in this domain, the theory makes different empirical predictions than models based on context-sensitive rules. Specifically, the current model allows for an account of special nominative singular forms and the so-called pseudo-ABA patterns (Middleton in Morphology 31:329–354, 2021, Davis in Glossa 6, 2021). The proposal relies on the Nanosyntax model of grammar, using phrasal lexicalisation and lexicalisation-driven movements (Starke in Exploring nanosyntax, Oxford University Press, Oxford, pp. 239–249, 2018), though alternative implementations of this idea are conceivable.