The same cue can help or do little depending on when it lands
The clearest source of disagreement is timing. Work on closed-loop TMR, where cues are triggered automatically in response to a detected brain state during slow-wave sleep, reports effects on memory integration and emotional processing, but the results vary with the sleep stage being targeted and with the timing of delivery [2]. That pattern fits the broader picture: a cue delivered in one sleep state is not equivalent to the same cue delivered in another, so studies that differ in their targeting rules can reach different conclusions even when the underlying mechanism is the same.
This also helps explain why some studies find robust effects and others find weak ones without either being wrong. If the benefit depends on catching the brain in the right state, then a study's result reflects how often it succeeded in doing that, not simply whether TMR 'works.' The variability is a property of the method's sensitivity to timing rather than a contradiction between labs [2].
Breathing rhythm appears to gate when a cue can reactivate a memory
A newer line of evidence points to an even finer timing factor: the breathing cycle. During non-REM sleep, the phase of respiration influenced the timing of memory reactivation and consolidation, leading the researchers to describe breathing as an endogenous timing signal for reactivation [3]. If the respiratory phase opens and closes a window for reactivation, then studies that do not control for breathing phase would be sampling that window unevenly, which could produce inconsistent results across experiments.
This is a proposed gating mechanism rather than a settled explanation for the mixed literature, and the study focused on consistent effects of breathing phase rather than on reconciling divergent TMR outcomes [3]. Still, it offers a concrete reason why two otherwise similar cueing protocols might diverge: they may be delivering cues at different points in a rhythmic cycle that the sleeping brain uses to organize when memories can be reactivated.
Brain rhythms and who responds may separate strong results from weak ones
Another candidate factor is the fine structure of sleep oscillations. An exploratory analysis examined whether sleep spindles, brief bursts of brain activity during sleep, differ in frequency when they occur together with slow oscillations, and it explicitly acknowledged that TMR findings have been mixed [1]. This suggests the relevant variable may not be the cue alone but the particular oscillation pattern the cue coincides with, so studies capturing different spindle characteristics could be measuring different things.
Clinical work adds a complementary angle: when odour cueing during sleep was used to augment trauma-focused psychotherapy, outcomes improved, and physiological factors such as spindle coupling were identified as helping explain variability in how people responded [4]. That points to individual differences in brain physiology, not just protocol differences, as a source of divergent results. A cue that couples well with one person's sleep rhythms may couple poorly with another's, which would make group averages look inconsistent across samples.
Why the disagreement is informative rather than discouraging
Put together, these findings suggest the mixed literature reflects a method that is highly sensitive to state: sleep stage and cue timing [2], respiratory phase [3], and spindle coupling and individual physiology [1][4] each plausibly shift whether a cue strengthens a memory. Studies that differ on these dimensions are not necessarily in conflict; they may be probing different windows of the same timing-dependent process.
The practical implication is that the question 'does TMR work?' is probably the wrong one. The more useful question is under what conditions it works, and the evidence points toward delivering cues in the right sleep stage, at the right moment in the breathing cycle, and in brains whose rhythms the cue can engage. Much of the variability remains to be pinned down, since the oscillation analysis was exploratory [1] and the breathing result addresses consistent effects rather than the mixed literature directly [3], but the direction of the explanation is consistent across these studies.
Sources used in this answer
Modulating Sleep Oscillations Through Targeted Memory Reactivation: Linking Functional Dynamics to Brain Microstructure
An exploratory analysis of sleep spindles occurring with slow oscillations acknowledges mixed TMR findings and examines whether spindle frequency differs in that coupling, without clearly identifying what explains the variability.
The application of closed-loop targeted memory reactivation during slow wave sleep
Closed-loop TMR during slow-wave sleep affects memory integration and emotional processing, with outcomes varying by sleep stage and cue timing.
The breathing cycle gates memory reactivation during human NREM sleep
During non-REM sleep, the phase of the breathing cycle influences the timing of memory reactivation and consolidation, identifying respiration as an endogenous timing signal.
Odour cueing during sleep augments trauma-focused psychotherapy
Odour cueing during sleep augmented trauma-focused psychotherapy, with physiological factors such as spindle coupling helping explain variability in responses.