The chemical induction of synaesthesia

Synaesthesia is a neurodevelopmental condition in which ordinary stimuli (inducers) evoke atypical, involuntary secondary experiences (concurrents), occurring in about 1–4% of the population. Previous work indicates a genetic basis with early environmental shaping of specific inducer-concurrent pairings, and reports exist of adult-onset synaesthesia following stroke, trauma or drug use. A contentious question addressed in the literature is whether synaesthesia-like experiences can be induced in non-synaesthetes; prior studies have shown that cognitive training, suggestion and several classic psychedelics can produce transient synaesthesia-like phenomena, with serotonin (5-HT2A) signalling implicated as a likely neurochemical mechanism. However, most pharmacological work has focused on serotonergic compounds, leaving open whether other neurochemical systems contribute, whether certain individuals are generally prone to chemically induced synaesthesia, and how recreational drugs modulate developmental synaesthesia. Luke and colleagues therefore conducted a broad survey of recreational drug users and self‑reported developmental synaesthetes to characterise the frequency, drug- and class-specificity, and phenomenology of drug-induced synaesthesia. The primary aims were to (1) estimate incidence rates of drug-induced synaesthesia across a larger set of commonly used psychoactive substances, (2) test whether induction clusters by drug class (suggesting neurochemical specificity), (3) describe the types of induced synaesthesia, and (4) examine how drugs affect existing developmental synaesthesias and whether some individuals are particularly prone to induction across drugs. The study used an online questionnaire to gather lifetime drug-use histories and self-reports of synaesthetic experiences under the influence.

The study recruited participants online over six months via English-language drug‑user and synaesthesia forums and social media; the survey was hosted on Qualtrics. Of 1,568 starters, 644 participants completed the survey (41% completion). The completed sample ranged from 18 to 74 years (mean 30.0, SD 12.2) and contained 457 controls (self-reported non-synaesthetes) and 187 self-identified developmental synaesthetes. Recruitment intentionally oversampled synaesthetes to ensure statistical power; all participants provided informed consent under local ethical approvals. Data were collected with a structured survey that recorded demographics, self-reported developmental synaesthesia (including childhood-only cases and family history), lifetime use and frequency of use for 28 psychoactive substances, and whether participants had ever experienced synaesthesia while under the influence of each substance. Synaesthesia was defined for respondents as a blending of senses (for example, shapes having a taste or sounds having a shape). For type specification, respondents could select from a 27×27 matrix of inducer–concurrent combinations. Developmental synaesthetes were additionally asked whether each drug had enhanced or suppressed their extant synaesthesia, and all participants were asked about “flashbacks” of drug-induced synaesthesia. Skip logic reduced irrelevant questions based on prior responses. Analyses were pre-specified to some extent: drugs were grouped into 12 classes, with ‘‘tryptamines’’ denoting classic psychedelics (LSD, DMT, psilocybin/psilocin, ayahuasca). The researchers computed incidence rates, correlations (Pearson) with bootstrap 95% confidence intervals (10,000 samples), and a drug-class incidence dissimilarity index to test within-class clustering using 10,000 random permutations. Four exploratory stepwise logistic regressions predicted binary occurrence of drug-induced synaesthesia from either binary drug-use indicators or frequency of use, performed separately for controls and synaesthetes; Nagelkerke R2 was reported and bootstrap resampling (1,000 samples) assessed internal replicability of retained predictors. Finally, a synaesthesia-proneness (SP) score was computed to quantify an individual's tendency to report induced synaesthesia across drugs, weighting reported frequency by inverse population incidence for each drug. The extracted text does not report some procedural details such as exact inclusion/exclusion criteria beyond self-report, nor does it provide dosing information for individual drug episodes.

Sample and usage patterns: Of the 644 completers, 457 were controls and 187 self-identified as developmental synaesthetes. Controls reported greater use of 19 of the 28 surveyed drugs relative to synaesthetes. Across drugs, use rates did not correlate with induction rates in controls (r = .01, p = .96) or in synaesthetes (r = .09, p = .64), indicating that higher usage did not explain higher induction incidence. Incidence of induced synaesthesia and correlations across groups: Induced synaesthesia was reported for many drugs. Among controls, the highest incidence rates were for LSD (57%), ayahuasca (49%), and psilocybin (45%), with a median incidence across all drugs of 15% (95% CI: 3, 25). Synaesthetes showed numerically higher incidence rates across almost all drugs (median 23% [95% CI: 9, 38]). Incidence rates for induced synaesthesia across the 28 drugs were highly correlated between controls and synaesthetes (r = .79, p < .001). In addition, incidence rates in controls correlated strongly with enhancement rates of developmental synaesthesia in synaesthetes (r = .93, p < .001), and within synaesthetes incidence and enhancement rates were strongly correlated (r = .78, p < .001). Flashbacks of drug-induced synaesthesia were reported by 24% of controls and 43% of synaesthetes (χ2(1) = 22.08, p < .001, phi = .19). Clustering by drug class: Visual inspection suggested that drugs within the same pharmacological classes tended to have similar induction rates. Quantitatively, a within-class incidence dissimilarity index (computed for classes represented by two or more drugs) yielded a mean value of 54% across seven drug classes in controls, indicating members tended to be within 54% of the median incidence for their class. Permutation testing (10,000 permutations) indicated that this clustering had a low probability of arising by chance. Significant within-class clustering was observed overall and specifically for tryptamines and phenethylamines in both groups; clustering for dissociatives was significant in controls only. The extracted text notes that some classes had too few members to allow robust clustering tests. Predictors of induction (logistic regressions): Four exploratory stepwise logistic regression models predicted whether participants had ever experienced drug-induced synaesthesia. In controls, the binary-use model was significant (χ2(6) = 70.48, p < .001) and accounted for roughly 20% of variance; six predictors were retained, with LSD, MDMA (Ecstasy) and methadone retained in ≥75% of bootstrap samples. A frequency-of-use model in controls was also significant (χ2(5) = 48, p < .001) explaining somewhat less variance; methadone and kava kava were retained in ≥70% of bootstrap samples. Across both control models, LSD, Amanita muscaria, methadone and kava kava were retained in the final models, though replication rates varied and methadone showed particularly high replicability. Among synaesthetes, the binary-use model was significant (χ2(5) = 85.7, p < .001) explaining a larger proportion of variance (reported R2 ≈ .60), with psilocybin and GBL retained in ≥70% of bootstrap samples; GBL was a replicable negative predictor. The frequency model was also significant (χ2(5) = 90.2, p < .001) with LSD, cannabis and dextromethorphan (DXM) retained in ≥70% of samples. Across the four analyses, LSD (use and/or frequency) was the only consistent predictor of drug-induced synaesthesia in both groups, though replication rates for LSD predictors were not uniformly high. Methadone (controls) and DXM (synaesthetes) emerged as other notable, relatively replicable predictors. The authors note that predictors other than LSD did not replicate across both groups. Phenomenology: Sound-colour synaesthesia was the most commonly reported type of drug-induced synaesthesia, followed by sound-space and sound-shape. Grapheme-colour synaesthesia—common in developmental cases—was infrequently reported. Chi-squared analyses across the four highest-inducing drug classes (tryptamines, phenethylamines, dissociatives, salvia) showed differences only for sound-colour incidence (χ2(3) = 18.84, p < .001, phi = .18). Subsidiary contrasts indicated higher sound-colour rates for tryptamines versus dissociatives (χ2(1) = 4.33, p = .038) and versus salvia (χ2(1) = 16.04, p < .001), and for phenethylamines versus salvia (χ2(1) = 8.09, p = .004). Overall, induced synaesthesia types did not appear strongly drug-class-specific. Proneness to induction: The synaesthesia-proneness (SP) score distribution was heavily skewed: many participants had never experienced induced synaesthesia (33% of controls, 17% of synaesthetes). Binary induction rates across all drugs were lower in controls (67%) than in synaesthetes (83%), Fisher's exact p < .001, phi = .17. A small subset of participants showed high SP scores, suggesting that some individuals are particularly prone to experiencing synaesthesia across multiple drugs, and such proneness was more common among self-identified developmental synaesthetes. The authors note that SP estimates are confounded by drug exposure (non‑users necessarily score low).

Luke and colleagues interpret their findings as corroborating a prominent role for serotonergic compounds—particularly tryptamine-class psychedelics such as LSD, DMT and psilocybin—in producing synaesthesia-like experiences, but stress that induction is not exclusively serotonergic. The highest induction rates were observed for tryptamines, and logistic models identified LSD use or frequency as the most consistent predictor across analyses, supporting the long-standing hypothesis implicating 5-HT2A receptor activity in drug-induced synaesthesia. At the same time, substantial induction associated with non-serotonergic compounds—most notably Salvia divinorum (a kappa opioid agonist), dissociatives (NMDA antagonists), and cannabis—argues for a more nuanced, multi‑pathway view in which different neurochemical mechanisms may converge on perceptual hyperexcitability or altered network dynamics. The authors further emphasise that drugs which induce synaesthesia in controls tend also to induce novel synaesthesias or enhance existing synaesthesia in developmental synaesthetes, with strong positive correlations between induction rates in controls and both induction and enhancement rates in synaesthetes. This overlap suggests shared neural mechanisms across induced and developmental phenomena, though the authors caution that drug-induced experiences typically lack the inducercurrent consistency and automaticity characteristic of developmental synaesthesia, consistent with a model in which long-term consolidation of inducer–concurrent associations is required for those hallmark features. Several limitations are acknowledged. The online, self-report survey design allowed wide reach but precluded experimental control over dose, context, and objective verification of developmental synaesthesia; developmental status was not validated with standard consistency or automaticity tests. The sample over-represents self-identified synaesthetes by design, so prevalence figures are not representative of the general population. The SP metric is confounded by participants' drug-use exposure. Despite these caveats, the large sample and the non-random clustering of induction rates by drug class are presented as evidence that the observed patterns reflect meaningful pharmacological effects rather than mere reporting noise. For future research, the authors recommend controlled pharmacological trials directly contrasting 5-HT2A agonists with drugs acting on other systems (for example kappa opioid or NMDA mechanisms) to test competing neurochemical models, and neurophysiological comparisons between drug-induced and developmental synaesthesia. They also suggest studying individuals who appear highly prone to chemically induced synaesthesia to clarify predispositional factors, and that contextual variables (such as concurrent music listening) be indexed to determine whether frequent sound-induced reports reflect genuine modality bias or situational confounds.