Participants in this study were 808 undergraduate students enrolled in a university psychology course on “Dreams and Sleep” within the general psychology curriculum. Of these, 616 students (241 males, 371 females; mean age = 19.44 years, SD = 1.295) provided complete responses to all questionnaires and were included in the analyses. This study was approved by the Ethics Committee of Toyo University (Approval number: P240021). All participants provided written informed consent in accordance with the Declaration of Helsinki.

The following three questionnaires were utilized in this study:

Typically, synesthesia is identified using behavioral diagnostic approaches that assess the consistency of inducer–concurrent mappings across time (e.g., consistency-based tests; Eagleman et al., 2007). However, this study used a self-report method to broadly assess both visual and non-visual synesthetic experiences, including spatial arrangement, personality, taste, and touch. Although self-report methods may yield higher estimates of synesthesia prevalence than methods based on temporal consistency (Rouw and Scholte, 2016; Ward and Simner, 2022) and may include individuals whose experiences are not temporally consistent, previous studies have demonstrated the validity of self-reports. For example, self-reported synesthetic profiles have been consistently correlated with objective behavioral indicators such as sensory hypersensitivity and imagery abilities (Rouw and Scholte, 2016). Furthermore, consistent cognitive profiles of synesthetes based on self-report data have also been established (Ward and Simner, 2022). Thus, while acknowledging its limitations, including the potential for inflated prevalence rates, the use of self-report methods is justified given the aims of this research.

Previous self-report studies have shown that synesthetic experiences cluster into multiple types, among which four types—Language – color, OLP, Spatial sequence, and Visualized sensation—occur with particularly high frequency. Accordingly, following prior work, the present study focused on these four major types (Ward and Simner, 2022). Other less prevalent types were also assessed in the questionnaire (see Supplementary Table S1), but were excluded from the main inferential analyses due to low endorsement rates and limited statistical power.

Using the questionnaire survey described above, this study aims to comprehensively and multidimensionally analyze the relationship between synesthesia and lucid dreaming, thereby offering new insights into the task of situating synesthesia within a continuum of general cognitive phenomena.

The survey was conducted during psychology classes focused on “Dreams and Sleep.” The questionnaires were distributed by the researchers responsible for the classes, and participants responded individually. The survey was conducted across three sessions: personality traits (TIPI-J) in the first, lucid dreaming experiences (LuCiD) in the second, and the synesthesia self-report questionnaire in the third. Each questionnaire session took approximately 10 min.

All analyses were conducted in R (version 4.3.2). Only participants with complete responses to all questionnaires were included in the analyses (n = 616; listwise deletion).

Responses to the Synesthesia Self-Report Questionnaire were summarized for the four primary synesthesia types targeted in the present study (Type 1: Language–color, Type 2: OLP, Type 3: Spatial sequence, Type 4: Visualized sensation).

Figure 1a shows the endorsement rate (self-report) of each synesthesia type and Figure 1b the distribution of the number of synesthesia types endorsed by participants. A participant was considered to endorse a given synesthesia type if they responded “yes” to at least one item belonging to that type; participants could endorse multiple types. A value of zero in panel (b) indicates participants who did not endorse any of the four primary synesthesia types. Accordingly, the overall endorsement rate (68.0%) reflects the endorsement of at least one synesthesia-like experience.

To examine the structure of item-level synesthesia responses, hierarchical cluster analysis (Euclidean distance and Ward's minimum-variance method, k = 4) was conducted. The dendrogram and cluster assignment are shown in Supplementary Figure S1. The four-cluster solution yielded the following clusters: (i) Language–color (e.g., letters–color, numbers–color), (ii) OLP (e.g., numbers–personality, days–personality), (iii) spatial sequence (e.g., numbers–spatial arrangement, months–spatial arrangement), and (iv) Visualized sensation (e.g., music–color, emotion–color). In this solution, the item “Posture–color” was assigned to the Language–color cluster (Supplementary Figure S1).

Internal consistency of the LuCiD was evaluated using Cronbach's α coefficients (Table 1). Cronbach's α was 0.79 for the LuCiD total score. Subscale α values were 0.87 (Insight), 0.82 (Control), 0.70 (Thought), 0.66 (Realism), 0.62 (Memory), 0.69 (Dissociation), 0.86 (Positive Emotion), and 0.87 (Negative Emotion).

Table 1 also reports descriptive statistics for the LuCiD total score and each subscale (mean item scores on the 0–5 scale). Because lucidity was operationalized as continuous LuCiD scores, participants were not dichotomized into “lucid” vs. “non-lucid” dreamers; accordingly, we do not report a single categorical count of lucid dreamers. Instead, the distribution of scores provides an overview of individual differences in lucid-dream features (e.g., Insight: M = 1.67, SD = 1.29, range = 0.00–5.00; LuCiD total: M = 1.63, SD = 0.84, range = 0.00–4.75).

Pearson correlation coefficients between LuCiD scores (total and subscales) and Big Five personality traits (TIPI-J) are shown in Table 2. The LuCiD total score was positively correlated with Extraversion (r = 0.15, p < 0.001) and Openness (r = 0.08, p < 0.05). Control and Thought were also positively correlated with Extraversion (Control: r = 0.13, p < 0.01; Thought: r = 0.11, p < 0.01) and Openness (Control: r = 0.09, p < 0.05; Thought: r = 0.11, p < 0.01).

Multiple linear regression analyses were conducted for the LuCiD total score and each LuCiD subscale (Tables 3, 4). Predictors retained in the final models were selected using an AIC-based stepwise procedure. Standardized coefficients are reported.

Table 3 summarizes the final model for the LuCiD total score (adjusted R² = 0.103). Positive coefficients were observed for Type 3 (Spatial sequence), Type 4 (Visualized sensation), and Extraversion, whereas negative coefficients were observed for the interaction terms Type 1 × Extraversion and Type 2 × Openness.

For the eight LuCiD subscales, Table 4 reports only statistically significant standardized coefficients (p < 0.05) from the AIC-selected final models (adjusted R² range: 0.024–0.076). Full coefficients for all predictors retained in the AIC-selected models (including non-significant terms) are provided in Supplementary Tables S2, S3. Across subscales, Type 4 was retained with positive coefficients in each model, with significant positive coefficients for most of the subscales, including Insight, Control, Thought, Realism, Memory, Dissociation, and Positive Emotion. Type 3 was retained in the final models for all outcomes except Negative Emotion and showed consistently positive coefficients, reaching conventional significance in the models for Control, Dissociation, and Positive Emotion. Type 1 was retained in multiple subscale models, showing negative coefficients for Insight and Control and a positive coefficient for Negative Emotion.

Personality traits contributed both as main effects and as moderators. Extraversion showed positive main effects on several outcomes, including Insight, Control, Memory, and Positive Emotion. Openness showed a positive main effect specifically for Thought (logical reasoning within dreams), whereas Agreeableness showed a negative association with Dissociation.

In addition, several interaction terms indicated moderation by personality traits: Type 1 × Extraversion was retained with negative coefficients for Insight and Thought; Type 2 × Extraversion was retained with a negative coefficient for Control; and Type 2 × Openness was retained with a negative coefficient for Dissociation. Type 4 also showed interactions with personality traits in specific affective outcomes (Type 4 × Extraversion for Negative Emotion; Type 4 × Agreeableness for Negative Emotion).

Figure 2 provides a schematic overview of the direction of standardized coefficients for the final models. Because final models were selected using an AIC-based stepwise procedure, some predictors were retained despite not reaching p < 0.05; these coefficients are shown as dashed lines in Figure 2 and reported in full in Supplementary Tables S2, S3.

The results of our regression models suggest that the direction of associations with lucid-dreaming indices may differ across synesthesia subtypes. In particular, Type 3 (Spatial sequence) and Type 4 (Visualized sensation) showed broadly positive associations, whereas Type 1 (Language–color) and Type 2 (OLP) showed weaker associations, or even negative associations in some models. This pattern points to heterogeneity that cannot be fully explained by treating synesthesia as a single, unitary phenomenon. Accordingly, as a working hypothesis for interpreting our findings, we introduce a framework that organizes synesthesia types along a conceptual-perceptual dimension.

Here, “conceptual” refers to processing in which stimuli are identified by category or concepts (e.g., phoneme of a grapheme “a”) rather than by lower-level physical features (e.g., font or size of a grapheme “a”). In contrast, “perceptual” refers to processing closely tied to a stimulus's sensory characteristics and that can change dynamically in response to context.

Type 1 (Language–color) and Type 2 (OLP) are often induced by symbolic stimuli such as letters, numbers, and days of the week, and previous studies have reported that the induced experiences are largely unaffected by perceptual features such as font or size. Therefore, these types can be regarded as “conceptual synesthesia,” driven primarily by conceptual processing (Simner, 2012; Dixon et al., 2004; Rich et al., 2005; Witthoft and Winawer, 2006; Ramachandran and Hubbard, 2003; Chiou and Rich, 2014). For example, the tendency to select a common color for the Arabic numeral “1” and the Japanese numeral “؀” suggests that the determining factor may be the concept they represent rather than their visual form per se (Asano and Yokosawa, 2012).

By contrast, Type 4 (Visualized sensation) implies that perceptual inputs such as music, emotion, pain, and odors are accompanied by visual experiences (e.g., color, brightness, and shape). For example, studies of sound–color synesthesia have reported that the magnitude or brightness of synesthetic experiences varies continuously in response to continuously changing stimulus frequencies (Chiou et al., 2013; Ward et al., 2006). This suggests the involvement of processing tightly coupled to the stimulus's sensory properties. In this sense, Type 4 can be regarded as a prototypical form of more “perceptual” synesthesia.

Type 3 (Spatial sequence) may appear conceptual insofar as it is triggered by symbolic elements such as numbers, time, height, and temperature. However, because the experiential content involves a sequence being evoked as a continuous spatial layout, it has been suggested that the inducer may be represented not as a discrete concept but rather as a continuous magnitude, and may thus function as a stimulus with perceptual qualities (Price, 2014; Gould et al., 2014).

Nevertheless, the conceptual/perceptual positioning is not fixed. For instance, for children before learning letters, letters may be processed more as mere shapes, and thus more perceptually. Furthermore, even in sound–color synesthesia, some cases involve discrete pitch-class concepts acting as synesthetic inducers (e.g., C as red, D as yellow, E as green), rather than continuous acoustic properties. In such instances, the synesthetic association is likely to have a stronger conceptual component (Itoh et al., 2019). The conceptual/perceptual organization used in this study is therefore a pragmatic framework that presupposes such continuity. In the next subsection, we explore how this distinction between “more conceptual” and “more perceptual” synesthesia might relate to dream experiences—particularly controllable lucid dreams—from the perspectives of SMCs and counterfactual richness.

Based on the conceptual–perceptual distinction introduced in the previous section, we interpret the associations between synesthesia types and lucid-dream outcomes (LuCiD) (Tables 3, 4, Figure 2).

Type 4 (Visualized sensation), which is characterized by perceptual features, showed a positive association with the LuCiD total score and yielded consistently positive coefficients across a wide range of subscales, including Insight, Control, Thought, Realism, Memory, Dissociation, and Positive Emotion. One possible interpretation is that Type 4 involves dynamically generated visual concurrents (e.g., color, shape, brightness) in response to continuous sensory inputs such as sounds, odors, pains–that is, the repertoire of possible SMCs is continuously and dynamically generated –thereby providing excessive counterfactual-richness (O'Regan and Noë, 2001; Ikegami and Zlatev, 2008; Seth, 2014).

Following the perceptual-presence account outlined in the Introduction, the excessive counterfactual richness of Type 4 thereby enhances perceptual presence in dreams—that is, the sense that the dream world is “really there.” A heightened perceptual presence may, in turn, provide a more coherent and action-relevant world model during dreaming, facilitating metacognitive monitoring of the ongoing experience (e.g., noticing inconsistencies and reflecting on one's state) and intentional intervention. This interpretation offers a plausible bridge from SMC/counterfactual-richness to LuCiD facets such as Insight, Dissociation, and Control, and may also be consistent with more elaborate in-dream cognition and recall (e.g., Thought, Memory) and a more realistic phenomenology (Realism). We emphasize that these links remain theoretical and should be tested more directly in future work using measures that capture SMC/counterfactual structure and perceptual presence during dreaming.

Similarly, Type 3 (Spatial sequence) also showed positive associations with the LuCiD total score and, in particular, with Control, Dissociation, and Positive Emotion. Because Type 3 is experienced as a sequential spatial layout that emerges as a continuous configuration, it may more readily increase potential SMCs related to spatial aspects of dreaming, such as scene construction and self-localization.

In contrast, for the more conceptually dominant types–Type 1 (Language–color) and Type 2 (OLP)–associations with LuCiD outcomes were generally limited. For Type 1, negative coefficients were observed for Insight and Control, while a positive coefficient was observed for Negative Emotion. Moreover, the retention of interaction terms such as Type 1 × Extraversion, Type 2 × Extraversion, and Type 2 × Openness in the final models suggests that the influence of conceptual synesthesia may not be uniform but may instead be conditional on personality traits and/or cognitive style. Conceptual synesthesia relies on conceptual processing that can become highly automatized through learning (Stroop, 1935; MacLeod, 1991), and thus does not necessarily involve the sequential, continuous updating of SMCs assumed for perceptually grounded synesthesia. Within the present framework (SMC/counterfactual-richness), the absence of a consistent “enhancement” pattern for Type 1/2 across LuCiD facets can therefore be regarded as compatible with their concept-dominant and automatic-processing characteristics.

In summary, organizing synesthesia types along a perceptual (Visualized sensation, Spatial sequence) vs. conceptual (Language–color, OLP) dimension is useful for understanding their contributions to lucid-dream outcomes. At least for the perceptually grounded types (Type 3/4), the observed pattern was consistent with our hypothesis that increased counterfactual-richness may facilitate dream controllability and metacognitive facets of lucidity. This conceptual reclassification may also offer implications for the cognitive and phenomenological understanding of synesthesia. Future studies should incorporate indices that more directly capture SMC/counterfactual-richness and examine their influence on dreaming and other everyday experiences.

The regression models suggested that personality traits can contribute to LuCiD outcomes both as direct effects (main effects) and as moderating effects via interactions with synesthesia types (Tables 3, 4, Figure 2). Below, we focus on Extraversion, Openness, and Agreeableness and interpret their main effects and interaction patterns for the LuCiD subscales. Bivariate correlations (Table 2) are referenced only as supplementary information.

The findings of the current study suggest the possibility of a continuous influence of synesthetic perceptual experiences extending into dream experiences, particularly lucid dreaming. This interpretation aligns with the Continuity Hypothesis (Schredl, 2003; Domhoff, 2017), which proposes that dream content is continuous with waking cognitive processing and everyday perceptual experiences. According to the Continuity Hypothesis, dreams reflect an extension of cognitive processes and sensory experiences that frequently occur in daily life.

Synesthetes engage in specific perceptual processes in response to particular stimuli (e.g., letters, numbers, sounds) during their waking lives. In particular, perceptual synesthetes (Visualized sensation, Spatial sequence) may inherently provide excessive sensorimotor contingencies (SMCs) to sensory stimuli. Such a perceptual processing style likely continues into dream experiences, potentially facilitating lucid dream experiences characterized by enhanced dream content control (due to excessive SMCs) and heightened perceptual presence. Consequently, these experiences may lead to increased insight (awareness of dreaming) and dissociation (the separation of self from dream content).

Conceptualizing synesthetic sensory-perceptual processing in daily life as potentially continuous with dream experiences may provide a more comprehensive and phenomenological account of the relationship between synesthesia and lucid dreaming. Future research should combine quantitative assessments of synesthesia (e.g., temporal-consistency measures indexing temporal stability) with phenomenological analyses of dream reports to clarify how everyday synesthetic experiences are reflected in dream content and dream awareness. Because the present study is based on self-report data, the mechanisms proposed here should be regarded as conceptual and interpreted with appropriate caution.

As anticipated from our a priori decision to use a broad self-report questionnaire (Rouw and Scholte, 2016), we observed a high endorsement rate of synesthetic experiences in this sample, which should not be interpreted as a prevalence estimate for the general population. Participants were recruited from a university course focused on “dreams and sleep,” a context that may have attracted students with a heightened interest in dreaming and related anomalous or introspective experiences. Moreover, because our synesthesia assessment relied on self-report rather than a behavioral consistency test, endorsement should be interpreted as indicating self-reported synesthesia-like experiences (or a broader synesthetic trait) rather than a definitive behavioral diagnosis of synesthesia. Accordingly, when compared with prevalence estimates for grapheme–color synesthesia reported in prior population-based studies (Simner and Hubbard, 2006; Ward, 2013), the elevated endorsement rate observed here likely reflects both sample-specific characteristics and the breadth of the self-report methodology (Eagleman et al., 2007; Rouw and Scholte, 2016; Ward and Simner, 2022). As noted in previous work, self-report approaches can capture a broader range of cross-modal experiences, including associations that may not meet strict temporal-consistency criteria (Rouw and Scholte, 2016; Ward and Simner, 2022).

Importantly, hierarchical clustering of item-level endorsements reproduced the targeted four-type structure in a manner broadly consistent with earlier self-report clustering studies (Ward and Simner, 2022), providing evidence that participants' responses were systematically structured rather than arbitrary. Nevertheless, this structural replication does not establish temporal consistency at the individual level; therefore, the present conclusions are restricted to associations between self-reported subtype endorsements and LuCiD outcomes. Notably, the relatively similar endorsement rates across the four targeted types (Figure 1a) are consistent with questionnaire-based clustering work: Ward and Simner (2022) reported broadly comparable cluster prevalences for Language–Color (0.684), Personification (0.440), Visualized sensations (0.576), and Sequence-space synesthesia (0.618).

Several additional limitations should be noted. First, both synesthesia and lucid dreaming were assessed retrospectively via self-report; individual differences in introspective tendency, response style, and dream recall could therefore inflate associations between subtype endorsement and LuCiD scores. Second, the cross-sectional design precludes causal inference: synesthesia-like experiences might facilitate lucid dreaming, lucid-dream propensity (or interest in dreaming) might increase synesthesia endorsement, or a third factor (e.g., imagery/absorption or metacognitive monitoring) may contribute to both.

In addition, the variance explained by the regression models was modest (adjusted R² = 0.103 for the LuCiD total score; 0.024–0.076 across subscales), indicating small effect sizes and substantial unexplained variability. Finally, because models (including interaction terms) were selected via an AIC-based stepwise procedure across multiple outcomes, some retained effects may be sample-specific. Accordingly, the present findings should be treated as hypothesis-generating and should be tested in independent samples using preregistered analyses, together with prospective dream measures (e.g., diaries) and behavioral diagnostics of synesthesia.

Finally, the internal consistency of the LuCiD subscales should be considered when interpreting the present findings. Although the LuCiD total score showed adequate reliability (Cronbach's α = 0.79), several subscales–particularly Realism (α = 0.66), Memory (α = 0.62), and Dissociation (α = 0.69)–exhibited relatively lower internal consistency. These subscales—Realism, Memory and Dissociation—may be particularly sensitive to state-dependent fluctuations and individual differences in metacognitive access to dream experiences, which could reduce item homogeneity. While these values fall within ranges commonly considered acceptable for psychological research, reduced reliability may increase measurement noise and attenuate effect estimates, especially in regression analyses. Accordingly, the results for these subscales should be interpreted with appropriate caution. Future research should re-examine item wording, assess test–retest reliability, and consider confirmatory factor analytic approaches to further refine the measurement properties of the LuCiD in relation to synesthesia research.

Finally, recent phenomenological work suggests that synesthetic experiences may involve dynamically generated sensory–affective qualities that are not fully captured by temporal consistency measures alone (Matsuda and Daikoku, 2025). Future research should therefore combine behavioral diagnostics (e.g., consistency testing) with self-report and phenomenological approaches to clarify which facets of synesthetic experience are most relevant to lucid-dreaming outcomes.