The paper presents a feature-space theory of the production effect in recognition memory, extending attentional subsetting theory to explain why production differs from other manipulations like study time and spaced repetition. The authors propose that phonological features are drawn from a compact feature space, while deeper features are sparsely selected from a larger subspace. This model explains several findings, including the dependency of production effects on how other list items are encoded and the production advantage for homophones. The theory suggests that production enhances memory by increasing the number of encoded features and their distinctiveness, acting on early-accessed features. The model is applied to empirical phenomena, such as list-strength effects and mirror effects, and compared with other models. The authors argue that the production effect can be explained by examining the characteristics of the feature spaces influenced by experimental factors, rather than assuming different processes. The model is tested using data from a list-composition experiment, showing good fit and reproducing key qualitative features of the data. The paper also explores the sensitivity of the model to parameters and experimental factors, demonstrating how slight variations can change theoretical predictions.The paper presents a feature-space theory of the production effect in recognition memory, extending attentional subsetting theory to explain why production differs from other manipulations like study time and spaced repetition. The authors propose that phonological features are drawn from a compact feature space, while deeper features are sparsely selected from a larger subspace. This model explains several findings, including the dependency of production effects on how other list items are encoded and the production advantage for homophones. The theory suggests that production enhances memory by increasing the number of encoded features and their distinctiveness, acting on early-accessed features. The model is applied to empirical phenomena, such as list-strength effects and mirror effects, and compared with other models. The authors argue that the production effect can be explained by examining the characteristics of the feature spaces influenced by experimental factors, rather than assuming different processes. The model is tested using data from a list-composition experiment, showing good fit and reproducing key qualitative features of the data. The paper also explores the sensitivity of the model to parameters and experimental factors, demonstrating how slight variations can change theoretical predictions.