Reinstating details of previously seen stimuli given a recall cue is a common cognitive function. Studies suggest that such recollection involves hippocampus-mediated top-down reactivation of neural ensembles active during encoding, yet the neural mechanisms of such generative retrieval remain unknown. We formulate a biologically plausible hypothesis where transient cholinergic modulation, known to enhance backpropagated action potentials and trigger recall, initiates backward signals from sparse neurons discriminatively representing the cue to cortical areas encoding the trace. These signals flow through gap junctions, increasingly recognized for their prevalence and role in synchronization within memory-related brain regions, retracing the forward pass in reverse, reactivating the same neurons. We evaluate this mechanism's computational efficiency in two cue-based recall tasks: image-to-category classification and image reconstruction from sparse activations. For classification, we compare our method to an SVM classifier on many-example and one-shot learning using the Caltech face/motorbike dataset, ensuring robustness through multiple train/test splits and Gardner-Altman plots. For reconstruction, we use AlexNet to model feedforward visual processing and reconstruct input images from sparse activations at various layers, experimenting with normalization and sparsity. In image-to-category mapping, our mechanism achieves 96.8\% accuracy, comparable to SVM. In one-shot learning, it reaches 96.2\% accuracy for certain pairs, surpassing SVM. Our model successfully reconstructs images from sparse activations in higher AlexNet layers, even when using less than 1\% of their top activated neurons. This mechanism bridges sparse and distributed neural coding, with sparse neurons such as concept cells acting as pointers that initiate the reactivation of distributed cortical representations, offering a unified framework for understanding generative cognition. This could impact cued attention, mental imagery, and future episodic thinking. If validated through tissue experiments or cell-level simulations, this paradigm shift could redefine our understanding of a range of high-level cognitive functions.