Memory-related computations must occur across various brain states, each characterized by distinct oscillatory patterns. Theta oscillations in the hippocampal CA1 region are prominent during movement and are thought to organize network computations during active exploration, such as the virtual spatial paths formed by the sequential activity of CA1 spatial-tuning neurons, or 'place cells,' known as theta sequences. However, it remains unclear how theta supports memory during periods of immobility, when CA1 theta oscillations dissipate. To address this, we recorded laminar neural activity using Neuropixels probes in the dorsal hippocampus of 4 male and 4 female rats during a spatial working memory task. Surprisingly, while theta power in CA1 decreased below detection during pauses, prominent theta oscillations were observed in the dentate gyrus, which remained during both movement and immobility. Using a Bayesian decoding algorithm, we decoded the hippocampal spike information during immobility theta oscillations (“immobility theta”) and observed short, virtual spatial paths resembling theta sequences. Notably, these immobility theta sequences occurred at a theta phase distinct from those during movement. Importantly, while movement theta sequences encode the animal's current location and ongoing behavior, immobility theta sequences preferentially encode remote locations and predict future choices in working memory tasks. Although CA1 sharp wave-ripples (SWRs)---synchronous neural events during awake rest and sleep---can also predict future choices, immobility theta is more frequent during pauses at reward sites or choice points and has a longer duration. These findings expand the understanding of theta oscillations, revealing their critical role in memory coding beyond movement periods and highlighting their significance in learning and memory across diverse behavioral states.