3.8 Breathing rhythm, sleep and memory consolidation
Breathing as a limbic “clock”. Zelano 2016: slow oscillations in the piriform correlate with breathing rhythm (r = 0.64–0.87), delta power is phase-locked to inhalation in the amygdala and hippocampus (7/7 patients), and mouth breathing abolishes the effect [76]. In healthy people, fearful faces are recognised ~28 ms faster and retrieval memory is better when the stimulus falls in the nasal inhalation phase [76]. In mice, the 4 Hz rhythm of the olfactory bulb drives the mPFC during freezing (Granger OB → dmPFC), olfactory bulbectomy reduces % freezing (effect size 0.77–0.97) and shortens freezing bouts (Markov, p = 0.008), and 13 Hz stimulation of the olfactory bulb reduces freezing (2.08) [78]. Folschweiller & Sauer synthesise: the respiratory rhythm is generated locally via the olfactory bulb, not by volume conduction [77]. Extrapolation to slow-breathing practice is mechanistically plausible but has not been tested directly.
Targeted memory reactivation (TMR). Rasch 2007 is the foundational experiment: rose odour present during learning and replayed during SWS improved location memory (97% vs ~86%; g = 0.69), with no effect when presented during REM, during wakefulness, or for procedural memory [80]. The Hu 2020 meta-analysis (N = 2,004) establishes the overall picture [83]:
| Subgroup | g [95% CI] | Notes |
|---|---|---|
| Overall | 0.29 [0.21; 0.38] | I² = 71%; Egger Z = 8.49 |
| After bias correction (3PSM) | 0.13 [0.06; 0.21] | Trim-and-fill 0.18; Vevea–Woods severe 0.05 (ns) |
| SWS | 0.27 [0.20; 0.35] | 70 experiments, N = 1,471 |
| N2 | 0.32 [0.04; 0.60] | |
| REM | −0.06 [−0.31; 0.18] | Minus sign inferred from CI (Appendix C) |
| Wake | 0.07 [−0.09; 0.23] | ns |
| Odour cueing | 0.32 [0.15; 0.50] | k = 32, N = 372; no different from sound (Q = 0.69) |
| Skill / declarative / conditioning | 0.44 / 0.23 / 0.03 | Conditioning ns |
Three observations. (1) The effect is real but small; after correction it is around g ≈ 0.13–0.18. (2) Odour is not superior to sound as a TMR cue, although it has the practical advantage of rarely causing awakening: Perl 2016 showed that odour during NREM increases delta (+9.8%) and slow spindles (+5.2%) without eliciting K-complexes (BF01 = 6.6), regardless of whether the odour is pleasant or unpleasant [109]. (3) The mechanism has neural evidence: odour during sleep activates the amygdala–hippocampus and increases connectivity with the vmPFC; reactivation of object category in the vmPFC predicts memory benefit (peak r = 0.70) [82]. Arzi 2012 demonstrated that it is possible to learn new tone–odour associations during sleep, but these transfer to the waking state only when learnt during NREM [81]. Daniels & Vermetten propose exploiting this application for fear extinction in PTSD (Hauner 2013: re-exposure to a contextual odour during a daytime nap enhanced extinction) [79].
3.9 Developmental window: attachment learning and the emergence of fear
In rat pups, pairing an odour with electric shock produces preference at PN8–9 but avoidance from PN10 (F(2,54) = 9.29), coinciding with the first activation of the amygdala by conditioning (2-DG) [92]. Moriceau & Sullivan 2006 showed that this window is controlled by corticosterone: at PN12–15, the same shock procedure produces avoidance when the mother is absent but preference when she is present (F(3,28) = 25.6); injecting corticosterone into the amygdala in the mother’s presence switches it to avoidance (F(2,20) = 35.4); amygdala muscimol blocks avoidance learning but not preference learning [93]. Neonatal attachment learning depends on noradrenaline from the locus coeruleus [25]. In humans, newborns whose mothers ate anise during pregnancy orient their heads towards anise odour (d = 1.29) and show fewer negative facial expressions (d = −1.72), although with only 24 infants and no randomisation [94]; a 2023 meta-analysis pooling 4 studies confirmed head orientation (d = 1.24; 40 infants) and mouthing (d = 0.72; 102 infants), but pooled negative expressions were not significant (d = −0.87; CI −2.39 to 0.66) [94]. The fetus discriminates the odour of amniotic fluid from around week 30 [2]. Implication: the early period has entirely different learning rules — odour–caregiver associations are formed regardless of the valence of the accompanying stimulus, a candidate mechanism for the origin of learnt valence in adults.
3.10 Ageing, neurodegeneration and depression
Epidemiology. Odour identification disorders rise from ~11–24% in middle age to 37–70% at age 70, and many people are unaware of it [97]; ~2% under 65 and ~75% over 80 according to another source [2].
Alzheimer’s disease. Olfactory regions (entorhinal, transentorhinal, anterior olfactory nucleus) are where AD pathology appears earliest (Braak I–II) [95]. In the WHICAP cohort (n = 1,037; 101 AD cases), each point decrease in UPSIT increased AD risk (HR 1.072 after adjustment); lowest versus highest quartile HR = 10.33 (4.55–23.49); in people without MCI, UPSIT predicted cognitive decline whereas the verbal memory test did not (RR 1.094 vs 1.010) [96]. AUC for predicting AD: UPSIT 0.65, verbal memory 0.73, combined 0.77 [96] — added value, not a replacement. Pooling 7,620 people followed for ~94 months, poor identification predicted conversion to AD within 2–10 years [97]. Note: the author of the UPSIT (Doty) is a co-author and has a financial interest [96].
Mortality. Loss of smell in people aged ≥57 is associated with a ~3–4-fold higher risk of 5-year mortality (Pinto 2014 via [97, 99]); each additional odour correctly identified ↔ ~6% lower risk (Wilson 2011 via [97]).
Depression — a bidirectional relationship. Kohli 2016 pooled data: patients with depression have lower threshold, discrimination and identification scores than controls, but the differences are small (e.g. identification 12.57 vs 12.98) [114]. The reverse direction is stronger: BDI rises with the degree of smell loss (normosmic 5.2; hyposmic 10.9; anosmic 14.2); BDI–TDI correlation r = −0.35 (n = 260); depression in 76% of patients with post-viral smell loss vs 6.7% of the general population [114]. In one study, only 28% (8/29) of patients with severe depression had olfactory impairment (Rossi 2015, via [114]). Athanassi 2021 showed that valence (hedonics) is the most sensitive index in depression — unpleasant odours are more unpleasant, pleasant odours less pleasant — whereas threshold is inconsistent; antidepressants improve both intensity and hedonics [115]. Soudry 2011 proposed a “common substrate” of amygdala–ACC–insula–OFC for olfaction and depression [19]. The olfactory bulbectomy model produces depression-like behaviour in rodents [77, 115].
PTSD. 45% of 100 refugees had odour-triggered panic attacks in the past month; diesel increased rCBF in the amygdala, insula and mPFC in veterans with PTSD [79]. Odour fear conditioning lowers the detection threshold for that odour, recovering after ~8 weeks (Åhs 2013 via [79]).
Scientific reference material, not medical advice. Figures are redrawn from published data, not copied from the original figures. GRADE and risk-of-bias ratings are the compilers’ own. Every number traces back to detailed reading notes on each paper.