Voice-Hearing in the General Population: The Relationship Between Inner Speech, Absorption, and Voice-Like Hallucinations

Abstract

Background and Hypothesis

Inner speech models have been influential in explaining why some people hear voices in both general and clinical populations. To explore this, scholars developed a scale which linked inner speech to hallucination proneness in the general population. We predicted that we could replicate this finding across cultures, but that absorption would also be a significant predictor, with inner speech mediating absorption’s effect on voice hearing—providing an alternative conception of the inner speech model.

Design

We investigated the relationship between inner speech, absorption, and voice hearing in two studies. In Study 1, surveys were administered to undergraduates across five countries (total N = 825, N = 217 completing both the Varieties of Inner Speech Questionnaire and Tellegen Absorption scales). In Study 2, surveys were administered to online participants in the USA (N = 1000).

Results

We found that both inner speech and absorption predicted voice hearing in Studies 1 and 2. Study 2 suggested that absorption and inner speech have a complex relationship with one another, one best described by a model of moderated mediation.

Conclusions

We argue that absorption, which we characterize as an attitude toward the mind, may be an important, causally antecedent factor to inner speech in generating voice-like experiences in the general population. We believe this better captures the expansive range of phenomenological experiences represented in voice hearing. Perhaps a voice is less like a misattributed thought and more like a very vivid thought which, because it is felt as vivid, is experienced as sensory.

The experience of hearing “voices” is among the most striking symptoms of schizophrenia and to some, the most troubling. Voice content has been linked decisively to the prognosis of the illness, with more harsh and powerful voices associated with worse outcomes.¹ Because of this, it is crucial to understand these experiences more deeply—a relatively new effort in modern biomedical psychiatric science.^1–3

In this endeavor, psychiatric scientists have insisted that we understand the experience of voice-like hallucinations in the general population. The rediscovery of the remarkable frequency of these experiences has led many to point to a “continuum of psychosis”.^3,4 This has led to the hope that the study of voice-like experiences in the general population may lead to more effective treatments of psychosis—and a call for psychiatric science to shift from a focus on the biomarkers of disease toward a focus on apparently protective factors.⁵ While a consensus is emerging that there are multiple pathways to these experiences,^6,7 there is also an emerging consensus that this diversity may be reflected in the experience of those who come for clinical care.⁸ Thus, it may be that understanding the patterns of cause and phenomenology in voices and voice-like events in the general population potentially have a great deal to teach us about cause and phenomenology of voices in the clinical population and, ultimately, their management.⁹

Voices are fundamentally inner thought-like events that are experienced as not the voice-hearer’s own, and often not inside the voice-hearer’s head. Because we tend to think of voices as containing words, many scientists have proposed some version of a theory in which inner speech—for a variety of reasons and under a variety of conditions—is misattributed and experienced as if it is not one’s own.^1,2,10–14 Fernyhough,¹⁰ eg, has argued that inner speech—the act of talking to yourself in your head—is the product of a developmental process in which external speech is internalized in stages as fundamentally dialogic. That is, inner speech has the quality of a conversation, asking, and answering in turn; it is social. The dialogic quality of inner speech thus lays the groundwork for experiences in which thoughts are experienced as speaking to the speaker.

Different scholars have hypothesized different mechanisms to explain why some thought loses what William James called its “myness”.^15,16 Reality monitoring, eg, is the hypothesized process by which people decide whether information has an internal or external source through a set of judgments based on features of the event.^17,18 Some have argued that those for whom the reality monitoring process is disrupted, such as people with schizophrenia, are more likely to misattribute internal experiences as external ones.^19,20 Others have suggested executive and inhibitory control dysfunctions and distortions in top-down mechanisms.^21,22

In 2011, McCarthy-Jones and Fernyhough introduced the Varieties of Inner Speech Questionnaire (VISQ) scale to demonstrate a clear relationship between hearing voices and inner speech. The scale was later revised by Alderson-Day et al.²³ Both scales (the VISQ and VISQ-R) were found to predict auditory hallucinations in the general population as measured by the Revised Launay-Slade Hallucinations scale.²⁴ The original VISQ study²⁵ found that all but one of the VISQ subscales correlated with this scale, but after controlling for many factors (age, gender, anxiety, depression, and the visual items on the Launay-Slade scale) only the dialogic speech subscale significantly predicted voice hearing. The Revised Varieties of Inner Speech Questionnaire (VISQ-R) found that auditory hallucinations were positively associated with evaluative inner speech, dialogic inner speech, and the impression of other people in inner speech.²³

However, different work suggests that there may be other key predictors of voice-hearing experiences. In broadening the scope of studies of voice-hearing to include the full range of voice-like experiences in the general population, it is critical to note that those who have voice-like experiences often associate them with spiritual beings.^9,26 For example, charismatic evangelical Christians as well as practitioners of other religions (eg, traditional Akan religion in West Africa) often report learning to develop and control voice-hearing experiences as part of their religious practice.²⁷. Studies of religious voice hearing have highlighted a capacity for absorption as an important predictor of voice-hearing experiences. Absorption—typically measured by the Tellegen Absorption Scale²⁸—describes an individual’s tendency to be engrossed, or absorbed, in sensory or imagined events. Scores on the Absorption Scale are highly significant predictors of hallucination-like spiritual events (such as hearing the voice of a god; Luhrmann et al.²⁹). Indeed, a series of studies shows that absorption robustly predicts what participants call “voices” among evangelical Christians,^26,30 people who describe themselves as mediums or are otherwise involved in new age spirituality,³¹ people who are instructed to interact with invisible beings,³² and others.

Absorption also does not seem to be a factor that depends upon language. Inner speech models of voice hearing of course rely on thought having linguistic and specifically word-like qualities. “Speech” implies a sense that a speaker has spoken, yet “voice-hearing” experiences, whether audible or not, do not necessarily have words. For example, a Californian man with psychosis interviewed by Luhrmann reported a series of remarkable experiences driven by something he called energy. He started getting messages from a god. The God never spoke. The man never heard words. The energy would pull him and guide him. He’d feel something, and then he would find language to interpret it in his mind and the words would be there. In The Voices Within, Fernyhough² describes a man who “knew that one of the voices was there even though he wasn’t speaking” (p. 220). People sometimes report that a voice was present, even though it did not speak; they sometimes say that the voice had no words.³³ In fact, many of these experiences identified as “voices” are fleeting and ambiguous in psychosis and the general population alike, although in the general population they are far less common and intrusive. Questionnaires force participants to identify voices as more voice-like than they may be (see³¹) as do cultural expectations such as phrases like “hearing god’s voice.” In rich phenomenological accounts of inner experience,³⁴ in ethnographic and historical accounts of spiritual experience (eg, ^35–38), and in scientific reports by scholars with lived experience,^39,40 it seems clear that many experiences identified as voices are more complexly sensory and less verbal than we might assume.⁹ Of course, many voices are distinctly voice-like, both in the clinical and nonclinical populations.

Drawing on data from general population undergraduates in five countries (the USA, Ghana, Thailand, China, and Vanuatu), as well as nearly a thousand participants drawn from the general US population, we show that both absorption and inner speech are significant predictors of voice-hearing experiences and suggest that absorption may be a causally antecedent factor of the inner speech model of voice hearing. We argue that absorption—an experiential, immersive “attitude towards the mind”—may lay the foundation for other cognitive processes, such as inner speech, that in turn generate experiences of voice hearing. This expansion of the theory enables us to account for the variability reported in auditory hallucinations, including wordless voices, and raises the possibility that an attitude toward the mind is at least as important as words in inner speech.

Study 1

The data were collected as part of the project described in⁴¹ but the exploratory analyses reported here are new. The data collection was originally motivated by the observation that charismatic evangelical Christians routinely seek to interact with invisible beings. They speak of “hearing” God’s voice. Upon careful probing, we discover that to some degree, these experiences can be thoughts that feel as if they are not their own and even have sensory qualities. For more details concerning the data, previous project, its motivation, and structure, please consult.⁴¹

Methods

Participants, Procedures, and Measures

Undergraduate participants were recruited in two waves of data collection in five countries—from west to east: the USA, Ghana, Thailand, China, and Vanuatu.

N = 523 undergraduates completed a survey consisting of the Varieties of Inner Speech Questionnaire (hereafter VISQ). This survey included only the dialogic, evaluative, and other people subscales of the VISQ. This is because condensed inner speech did not correlate with proneness to auditory hallucinations in the original reporting on the scale.²⁵ We note that we used the VISQ rather than the VISQ-R because this study was administered before the VISQ-R was published in 2018. The survey also included the five auditory items from the Revised Launay-Slade Hallucinations scale.²⁴ Surveys were presented in the local language of instruction (English in the USA, Ghana, and Vanuatu; Thai in Thailand; Mandarin in China). The translations of these surveys were the results of an extensive process of translation and back-translation.

N = 519 undergraduates were recruited in another wave of data collection to complete a survey consisting of the Tellegen Absorption scale²⁸—a list of 34 statements about traits and preferences related to an immersive personal orientation to inner life and the world to which people respond “true” or “false”—as well as two questions about experiences of spiritual or religious voices (descriptions of hearing the voice of God or a spirit with one’s ears or in one’s mind) pulled from the Spiritual Events scale.⁴¹

Across both waves of data collection, a total of N = 825 participants filled out surveys. N = 217 undergraduate participants overlapped such that they were included in both waves of data collection. These 217 participants came from three of the five total countries, namely the USA, Ghana, and China and they completed all measures, including the VISQ, the Launay-Slade scale, the Tellegen Absorption scale, and the Spiritual Events scale. All questions were combined into “average scores” for each index. In calculating these scores, we included participants regardless of item-wise missing data. By default, scores for all variables were standardized before being entered into analyses. The work was approved by the Institutional Review Board at Stanford University (#36980).

Results

We replicated the findings in McCarthy-Jones and Fernyhough²⁵ that the dialogic subscale of the VISQ is a significant predictor of voice hearing in five countries. Pooling across countries, dialogic VISQ scores were a strong positive predictor of voice-hearing, as measured by Launay-Slade scores (β = .35, P < .001). Although we observed differences in overall rates of voice-hearing across countries (namely, lower rates in Thailand, β = −.34, P < .001; and China, β = −.23, P = .003 compared to the USA), we observed no significant differences in the strength of the relationship between inner speech and voice hearing (all interactions: |β| < .11, P > .215) (see Table 1 and Figure 1.) This pattern of results held for the other people and evaluative subscales of the VISQ, with one exception: the evaluative subscale VISQ scores were not significantly predictive of voice hearing in Vanuatu.

Table 1.

Regression Results From a Model in Which Voice Hearing (Measured by Standardized Launay-Slade Scores) Are Predicted by Inner Speech (Measured by Standardized VISQ Dialogic Subscores) and Country, Including Interactions. This Regression Includes Data From All N = 523 Participants Who Completed the VISQ.

Predictor	Estimate	Std. Error	t	P
Intercept	0.021	0.041	0.51	.608
Inner Speech (standardized)	0.345	0.042	8.16	<.001***
Ghana (vs. USA)	0.035	0.080	0.44	.661
Thailand (vs. USA)	−0.340	0.077	−4.40	<.001***
China (vs. USA)	−0.234	0.080	−2.94	.003**
Vanuatu (vs. USA)	0.082	0.090	0.91	.365
Inner Speech × Ghana	0.109	0.088	1.24	.216
Inner Speech × Thailand	−0.060	0.077	−0.77	.439
Inner Speech × China	−0.074	0.084	−0.88	.379
Inner Speech × Vanuatu	−0.106	0.101	−1.05	.294

Abbreviation: VISQ, Varieties of Inner Speech Questionnaire.

Figure 1.

Shows the relationship between inner speech (measured by the VISQ dialogic subscale) and voice hearing (measured by the five auditory items of the revised Launay-Slade scale) in N = 523 undergraduates in five countries. VISQ, Varieties of Inner Speech Questionnaire.

Given the relatively small sample sizes in each country, we caution against interpreting this as strong evidence against such differences; however, we do consider this some evidence that the relationship between inner speech and voice hearing is present across a diverse range of cultural settings.

Our analysis suggests that both inner speech and absorption may be predictors of voice-hearing experiences across diverse cultural settings: Across samples of undergraduates collected in the USA, Ghana, and China (N = 217), scores on the revised Launay-Slade Hallucinations scale were positively correlated both with dialogic VISQ scores (β = .45, P < .001, see Table 2) and absorption scores (β = .48, P < .001, see Table 3; these standardized beta coefficients and P-values come from separate regression models for each predictor, each including main effects and interactions with country) (see Figure 2).

Table 2.

Regression Results From a Model in Which Voice Hearing (Measured by Standardized Launay-Slade Scores) Are Predicted by Inner Speech (Measured by Standardized VISQ Dialogic Subscores) and Country, Including Interactions. This Regression Only Includes Data From the N = 217 Participants Who Completed the VISQ as Well as the Tellegen Absorption Scale.

Predictor	Estimate	Std. Error	t	P
Intercept	0.402	0.104	3.87	<.001***
Inner Speech (standardized)	0.448	0.088	5.08	<.001***
Ghana (vs. US)	−0.406	0.172	−2.36	.019*
China (vs.US)	−0.688	0.136	−5.07	<.001***
Inner Speech × Ghana	0.136	0.176	0.78	.439
Inner Speech × China	−0.183	0.132	−1.38	.168

Abbreviation: VISQ, Varieties of Inner Speech Questionnaire.

Table 3.

Regression Results From a Model in Which Voice Hearing (Measured by Standardized Launay-Slade Scores) Are Predicted by Absorption (Measured by Standardized Tellegen Absorption Scores) and Country, Including Interactions. This Regression Only Includes Data From the N = 217 Participants Who Completed the VISQ as Well as the Tellegen Absorption Scale.

Predictor	Estimate	Std. Error	t	P
Intercept	0.387	0.102	3.79	<.001***
Absorption (standardized)	0.479	0.098	4.90	<.001***
Ghana (vs. USA)	−0.194	0.171	−1.14	.257
China (vs. USA)	−0.721	0.135	−5.35	<.001***
Absorption × Ghana	0.017	0.154	0.11	.910
Absorption × China	−0.128	0.139	−0.92	.357

Abbreviation: VISQ, Varieties of Inner Speech Questionnaire.

Figure 2.

Shows N = 217 participants’ reports of voice-hearing experiences (Launay-Slade scores) plotted against rescaled scores of inner speech (dialogic subscore of VISQ) and absorption. Inner speech and absorption are predictive of voice hearing across diverse cultural settings. VISQ, Varieties of Inner Speech Questionnaire.

Study 1 was not large enough to evaluate the possibilities of mediation or interactions within or across these cultural settings, but at minimum we consider this to be a promising indication that the dynamic interaction between inner speech and absorption documented in Study 2 might be at play in a diverse range of social-cultural settings.

We also note that absorption is a predictor of spiritual experiences of all kinds.⁴¹ In Study 1, we found that absorption was a significant predictor of spiritual voice-hearing experiences—namely hearing the voice of God or a spirit with their ears or in their mind (β = .25, P < .001, see Table 4).

Table 4.

Regression Results From a Model in Which Spiritual Voice Experiences (Measured by Summed and Standardized Responses to Voice Questions in the Spiritual Events Survey) Are Predicted by Absorption (Measured by Standardized Tellegen Absorption Scores) and Country, Including Interactions. This Regression Includes Data From All N = 519 Participants Who Completed Tellegen Absorption Scale and the Spiritual Events Scale.

Predictor	Estimate	Std. Error	t	P
Intercept	0.032	0.038	0.83	.409
Absorption (standardized)	0.250	0.039	6.46	<.001***
Ghana (vs. US)	0.981	0.078	12.53	<.001***
Thailand (vs. US)	−0.373	0.072	−5.16	<.001***
China (vs. USA)	−0.641	0.075	−8.54	<.001***
Vanuatu (vs. USA)	0.220	0.083	2.67	.008**
Absorption × Ghana	0.083	0.070	1.18	.237
Absorption × Thailand	0.004	0.082	0.05	.957
Absorption × China	−0.099	0.082	−1.21	.227
Absorption × Vanuatu	0.011	0.081	0.13	.895

Abbreviation: VISQ, Varieties of Inner Speech Questionnaire.

Study 2

Here, we introduce a new study which explores the relationship of absorption and inner speech to voice hearing using data collected from online participants. This study allowed us to do several things: to replicate our findings, preregister our hypotheses, and increase the number of our participants, which enabled us to examine potential interactions between absorption and inner speech. It also allowed us to examine these relationships with the VISQ-R, which had not been published at the time of our initial study.

Methods

Participants

We recruited 1000 US adults to participate through the online survey platform Prolific. Participants were paid ~$13.50/h for their work. To be eligible for participation, participants had to be at least 18 years old, located in the USA (according to Prolific), fluent in English and could not have participated in any related pilot studies. After applying our preregistered exclusion criteria—including self-assessment questions about the quality of the survey as well as attention check questions (see preregistration at https://osf.io/nr9w4), our dataset contained responses from 975 participants. Participants ranged in age from 18 to 85 (median age was 35). By design, our sample contained roughly the same proportion of men and women (49% men, 48% women, 2% identified as some other gender, 3 participants had missing data). Most participants (55%) identified as White and no other race; 16% identified as Black or African American and no other race; the remaining 29% came from any other racial/ethnic category. Our sample was highly educated: 41% had some kind of college degree (AA, BA, BS), and a further 20% had some kind of graduate school or a graduate degree. All participants gave informed consent. The work was approved by the Institutional Review Board at Stanford University (#36980).

Procedure and Measures

As preregistered, our primary outcome of interest in this study was scores on an abbreviated version of the Revised Launay-Slade Hallucinations scale (In our preregistration, we followed Luhrmann, Weisman, et al.⁴¹ in using a 6-item abbreviated version of the revised Launay-Slade scale. To match the work done in McCarthy-Jones and Fernyhough,²⁵ and Alderson-Day et al.,²³ we report analyses done using only the five auditory items on the revised LS scale.),²⁴ a widely used measure of voice-hearing experiences and the scale used by McCarthy-Jones and Fernyhough²⁵ and Alderson-Day et al.²³ Participants were also asked about voice-hearing experiences beyond the Launay-Slade Hallucinations scale, including the 14-item Posey–Losch scale,⁴² another popular index of verbal hallucinations; as well as two questions about experiences of spiritual or religious voices (descriptions of hearing the voice of God or a spirit with one’s ears or in one’s mind) pulled from the Spiritual Events scale.⁴¹ Participants were also asked about absorption with the Tellegen Absorption Scale.²⁸

Our primary interest in this study was the degree to which such voice-hearing experiences were correlated with the VISQ-R²³ versus the Absorption scale.²⁸ In our results, we report analyses done using participant scores from the full VISQ-R. As noted, previous work has found that some subscales of the VISQ and VISQ-R were more predictive than others. As preregistered, we ran our analyses on the full VISQ-R as well as the dialogic, evaluative, and other people subscales. The pattern of results held when analyses were run using these subscales. We report only full scale results, both for simplicity and because we found that the full scale consistently outperformed the subscale scores in predicting voice hearing. Given that we are comparing the predictive power of inner speech and absorption, using scores from the full scale seemed to best represent the overall effect of inner speech.

In addition to these two preregistered predictors of voice-hearing experiences, participants completed the Porosity scale,⁴¹ which consists of 14 statements which assess a person’s beliefs that the boundary between the mind and the world is permeable (eg, Evil thoughts can go out into the world like Wi-Fi or a radio and cause bad things to happen to other people). Previous work has shown the affirmation of these statements to be a strong predictor of spiritual presence events, including reports of hearing the voice of God and other spiritual presences.⁴¹ We have argued that porosity and absorption reflect different attitudes toward the mind, which work together to facilitate voices and other unusual experiences.²⁹ Including the Porosity scale in the current study allowed us to further explore whether such beliefs might also play a role in other kinds of voice-hearing experiences.

Finally, following Luhrmann et al.,⁴¹ participants also completed two “control” measures which we hypothesized to be relatively less predictive of voice-hearing experiences than the Absorption scale or the VISQ-R: the Need for Cognition scale,⁴³ and the Sense of Control: Mastery subscale.⁴⁴

Participants received each index in a random order. Questions within each index were in the same order for each participant. All questions were combined into “average scores” for each index. In calculating these scores, we included participants regardless of item-wise missing data. By default, scores for all variables were standardized before being entered into analyses.

Hypotheses

We preregistered four hypotheses (see https://osf.io/nr9w4).

Hypothesis #1: Inner speech is a significant predictor of voice-hearing experiences. That is, people who score higher on the evaluative, dialogic, and other people subscales, as well as the full VISQ-R will also tend to report more experiences of voice hearing on the Launay-Slade Hallucination scale auditory subscale. These relationships were previously found in Refs.^23,25

Hypothesis #2: Absorption is a significant predictor of voice-hearing experiences. That is, people who score higher on the Tellegen Absorption scale²⁸ will also tend to report more experiences of voice hearing on the Launay-Slade Hallucination scale auditory subscale.

Hypothesis #3: Inner speech (as measured by VISQ-R) is a mediator of absorption’s effect on voice hearing. This hypothesis captures our addition to inner speech models of voice hearing. While inner speech affects voice hearing, we believe that absorption—a trait-like proclivity to pay attention to and be absorbed by one’s inner world—is also an important, and a causally prior variable which affects voice hearing. We hypothesized that absorption’s effect on voice hearing would be mediated by inner speech.

Hypothesis #4: Inner speech and absorption are stronger predictors of spiritual experience than other, less theoretically motivated, predictors. That is, when entered in a single model, the predictive power of scores on our predictor measures of interest (the VISQ-R and the Absorption scale) will be greater than the predictive power of scores on two “control” measures (the Need for Cognition scale and the Sense of Control: Mastery subscale). This was evaluated by coding the orthogonal contrast of predictors of interest (Inner speech, Absorption) versus control scales (Need for Cognition, Sense of Control). The parameter of interest is the interaction between predictor contrast and predictor score, which tests whether our predictors of interest were better predictors of spiritual experience than our control measures.

Results

This study yielded support for all of our preregistered hypotheses.

In separate regression analyses, both inner speech (β = .46, P < .001; Hypothesis #1) and absorption (β = .41, P < .001; Hypothesis #2) were significant predictors of voice-hearing experiences, as measured by the Launay-Slade scale. (We note that in our preregistration we intended to run mixed-effects models with random intercepts on participants. This proved untenable as we had the same number of participants as observations. Models were run as regressions using the function lm() in the R package stats.) In an additional regression analysis comparing our predictors of interest (VISQ-R, Absorption scale) to our two control predictors (Need for Cognition scale and Sense of Control: Mastery subscale), our predictors of interest were significantly better predictors of Launay-Slade scores than our control measures (β = 1.03, P < .001; Hypothesis #4). An exploratory analysis revealed that porosity—the extent to which a person believes that the boundary between the mind and the world is permeable—also correlated with Launay-Slade scores (β = .33, P < .001). See Figure 3 for all correlations between predictors and Launay-Slade scores.

Figure 3.

Shows participants’ reports of voice-hearing experiences (Launay-Slade auditory scores) plotted against rescaled scores of inner speech (VISQ-R), Absorption, Porosity, Sense of Control: Mastery subscale, and Need for Cognition. Correlations between each predictor and voice hearing are reported in each plot. VISQ-R, Revised Varieties of Inner Speech Questionnaire.

What Is the Nature of the Relationship Between Absorption and Inner Speech in Facilitating Voice Hearing?

Our hypothesis was that absorption—a proclivity to pay attention to and be absorbed by one’s inner world—amplifies or otherwise changes one’s relationship to one’s inner speech, which then facilitates voice-hearing experiences; in other words, that inner speech mediates the effect of absorption on voice hearing (Hypothesis #3). In line with this, a mediation analysis using the R package psych (All indirect effects reported are the mean bootstrapped indirect effects (ab path).),⁴⁵ showed that inner speech mediated the relationship between absorption and voice hearing (see Figure 4; β_indirect = .14, SE = 0.02, P < .001). We note that the reverse was also true: treating absorption as the mediator of a relationship between inner speech and voice hearing also yielded a significant mediation effect (β_indirect = .10, SE = 0.01, P < .001). Mediation analysis alone cannot determine whether absorption or inner speech is causally antecedent. Indeed, we find it plausible that this relationship could work both ways: stronger inner speech could lead to a more absorptive frame of mind. Inner speech and absorption may even be mutually reinforcing, causally speaking. However, we find it more theoretically plausible that being more absorbed in one’s inner sensory world affects the strength of one’s inner speech, which might then in turn facilitate the experience we call hearing a “voice.” We thus suggest that absorption might be causally prior to inner speech. After all, inner words are a relatively specific feature of mental life. Absorption seems to capture a broader range of phenomena and orientation in statements such as: “Sometimes I feel as if my mind could envelop the whole world”; “Sometimes thoughts and images come to me without the slightest effort on my part.”

Figure 4.

Shows the path diagram from the mediation analysis (Hypothesis #3). The key takeaway from this diagram is the difference between the c and c′ paths, which shows that inner speech mediates the effect of absorption on voice hearing.

In an exploratory analysis, we also examined the possibility that absorption and inner speech might be separate causal factors working together to facilitate voice hearing. In a regression analysis including absorption, inner speech, and the interaction between them as predictors of Launay-Slade scores, we found that both absorption (β = .27, P < .001) and inner speech (β = .34, P < .001) remained significant positive predictors, and showed a positive interaction effect (β = .10, P < .001). This suggests, first, that while both inner speech and absorption explain variance in participants’ Launay-Slade scores, they are not explaining the same variance; absorption and inner speech are not reducible to one another. Second, the interaction suggests that a participant who is high in absorption and has strong inner speech is even more likely to report a voice-hearing experience than we would predict on the basis of either predictor alone (Figure 5).

Figure 5.

Shows the interaction between inner speech (as measured by the participant’s average score on the VISQ-R) and absorption (as measured by the participant’s average score on the absorption scale) in predicting voice-hearing experiences (as measured by the participant’s average score on the Launay-Slade hallucination scale). Absorption is plotted in three groups: Participants whose scores were more than a standard deviation below the mean (lowest line, in red), participants whose scores were less than a standard deviation away from the mean (middle line, in blue), and participants whose scores were more than a standard deviation above the mean (highest line, in green). The different slopes of these lines show that a participant who is high in absorption and has strong inner speech is even more likely to report a voice-hearing experience. VISQ-R, Revised Varieties of Inner Speech Questionnaire.

One way to synthesize these findings is via a moderated mediation analysis (using the package lavaan⁴⁶). Results were consistent with the possibility that the relationship between absorption and voice hearing was mediated by inner speech (β_indirect = .17, SE = 0.02, P < .001), and that this mediation was moderated by absorption itself, as shown by the significant interaction effect between absorption and inner speech (β_b2 = .096, SE = 0.03, P < .001). The theoretical takeaway from this statistical model, then, is that absorption not only amplifies inner speech (with downstream consequences for voice-hearing experiences), but may also enhance the degree to which such inner speech drives voice-hearing experiences.

The current study suggests that absorption and inner speech have a complex relationship with one another, one that we believe is best described by a model of moderated mediation. According to this model, inner speech mediates the effect of absorption on voice hearing: that absorption is a background ability or orientation toward the mind which facilitates or otherwise enhances inner speech, and inner speech in turn facilitates voice-hearing experiences. In addition, this model suggests that absorption moderates the effect of inner speech on voice hearing, such that the generative effect of inner speech on voice hearing is stronger among people with a greater capacity for absorption. In other words, on this “moderated mediation” account, absorption not only facilitates inner speech but also enhances the degree to which inner speech drives voice-hearing experiences.

Prevalence of Voice-Hearing in a Large, General Population Sample

Our data suggest that voice-hearing experiences are, in fact, quite common among adults drawn from the general US population. Fully 79% of our participants (768 out of 975) in Study 2 responded yes to at least one of the five auditory questions on the Launay-Slade scale (ie, there was at least one voice-hearing experience included in the scale that they reported having experienced at least one time). Similarly, 90% of participants responded yes (more than “Never”) to at least one Posey–Losch question, and 43% responded yes to at least one spiritual voice hearing question. Some large epidemiological studies have estimated that only 3%-10% of general population adults report some kind of voice-hearing experience,^47,48 with some estimates as low as 1%.⁴⁹ Other research has found much higher rates, such as the original Posey–Losch study (1983–84), with 71%. Our current work, along with previous findings,⁵⁰ is consistent with these higher rates. While it is clear that hearing a voice is a real experience—people often report being startled, turning their head to see who has spoken; in interviews they provide convincing phenomenological details—it also seems clear that it is an ambiguous event whose recall is shaped by context.

General Discussion

Our findings provide further strong support for inner speech models of voice hearing. We replicated the original findings of the VISQ in five countries, and in a large online sample of US adults, we found that inner speech was a strong positive predictor of reports of voice-hearing experiences.

We also seek to extend this theory by offering absorption as another factor that also seems to play a generative role in producing such experiences and was also a strong predictor of voice experiences. We suggest that absorption and inner speech are not just two predictors of voice hearing, but in fact work together to generate the kinds of experiences we are calling “voices.” Moreover, we believe that there are reasons to suspect that absorption is in some sense causally prior to inner speech.

Inner speech models of voice hearing describe auditory hallucinations as inner speech, ie, for one reason or another, misattributed to external sources.^10,11,19,51 There are limitations to this approach. Hearing voices, as a phenomenon, seems to encompass a much broader swath of experiences than mere words, including feeling the presence of a social other with a message, or a general quality of “not me” thought (cf.^8,9). We also know that there are significant individual differences in the content of inner experience—more versus less spatial, more versus less vivid, more words versus more images, different sensory qualities, and even unsymbolized wordless thought.³⁴ People often say contradictory things about voice-hearing experiences: that they heard the voice audibly in their mind, or that they heard it outside their head but not with their ears. In fact, “voices” are not always voice-like—they may not have words, may not be external, may not be auditory only. They often seem more like some kind of mental experience felt to be not one’s own and so shaped by an experience of the mind.

We believe that by including absorption in the inner speech model of voice hearing, we can offer an account which can more gracefully include these less verbal dimensions of the phenomena we call voices. Furthermore, an emphasis on absorption shifts our attention to the general population. It invites us to consider whether hearing voices—thoughts felt not to be one’s own and often not within one’s mind—are a common human experience which is tweaked in psychosis, rather than focusing on the way psychotic deficit may be distributed in the population. It is worth pointing out that the predictive power of absorption on voice-hearing experiences is by no means obvious. The Tellegen absorption scale was initially developed to measure people’s susceptibility to hypnosis,²⁸ but the scale itself is not clinical in tone or scope. It includes “True/False” statements such as “I can be deeply moved by a sunset.” It is not particularly voice-focused, and lists no hallucinations. One item even seems to cut against the strength of a participants’ inner speech: “My thoughts often don’t occur as words but as visual images.”

We suggest, specifically, that absorption is a causally antecedent attitude toward the mind which makes a difference to how inner speech is processed and experienced. Absorption captures an experiential mindset which intensifies inner and outer sensory experience in ways that reflect both prior expectation and novel sensation.⁵² It makes experiences that require intense imaginative engagement, including spiritual experiences, feel more real.⁵³ Because absorption is a trait related to attention and sensory engagement, it suggests a potential mechanism by which reality monitoring is affected such that people experience voices or voice-like hallucinations. Two distinct aspects that the absorption scale may measure is the ability to suspend reality testing and a capacity for vivid mental imagery (⁵⁴; but see^41,55,56 for discussion). Richer mental imagery may change the top-down categorizations people make and suspending reality testing could alter the kinds of predictions being generated by the mind, perhaps at several levels. A highly absorptive person is one who experiences inner sensation—including inner speech—more intensely. Their inner experience then becomes more vivid. It feels more real, more alive, as if it is more in “living color.”⁵⁷ High absorption means that, at baseline, inner speech may be experienced more vividly, attended to more deeply. If voice hearing is misattributed inner speech, high absorption makes that misattribution more likely because the inner speech is experienced as stronger, more salient, more real. However, absorption is not uniquely attuned to inner speech, but all inner sensation. This helps to account for a broader range of phenomenological experiences of hearing voices.

While we find it plausible that absorption comes first causally, we do not claim that that absorption comes first temporally or developmentally. We mean to emphasize that absorption is trait like. This makes it possible to lay the foundation for how a person notices, attends to, and experiences their own cognitive, linguistic, and emotional processes like inner speech. We recognize that our single mediation analysis cannot definitively disambiguate whether inner speech mediates absorption’s effect on voice-like hallucinations or the other way around. However, regardless of the ultimate direction of the mediation, our results, including the interaction we find between absorption and inner speech, suggest that absorption and inner speech work together in facilitating voice-like hallucinations. Such a model might better match the experience of voices in the general population, which are often less frequently auditory, less mean, more responsive to cultivation, and perhaps more richly sensory.^8,58

Such a model might also better explain the complex relationship between psychosis and dissociation, a close cousin to absorption. Absorption has been described as making up a third of the items on the Dissociative Experiences Scale,⁵⁹ a common measure of dissociation.⁶⁰ Indeed, absorption, trance, hypnosis, dissociation could all be described as capturing different aspects of some kind of human capacity to experience agency as non-agentic: as not one’s own. Psychiatric scientists have proposed that the voices associated with psychosis are fundamentally a dissociative phenomenon, rather than a psychotic one.⁶¹ Those who are persuaded of this perspective point to the role of trauma in schizophrenia.⁶² They suggest that psychosis might best be understood as a kind of post-traumatic response and indeed argued that dissociation mediates the relationship between trauma and hallucinations.^63,64 Our work on absorption might invite us to take more seriously the possibility that dissociation is more interwoven with psychosis than mainstream psychiatric science now presumes.

The deepest question here is how we think about voice hearing. Do people have these vivid experiences—thoughts that do not feel like theirs, thoughts that feel as if they are perceived sensations—primarily because there is a deficit in the brain, or because of some more general feature of the way they relate to thought itself? This paper cannot answer that question. But an emphasis on inner speech can invite us to focus on deficit, on the problem which comes to be called “misattribution.” We seek to raise the possibility that the more basic phenomenon is that vividly experienced thoughts do, on occasion, feel as if they are not one’s own, and not internal. This shifts the main story to be about the complex, contradictory experience all humans have with their thoughts more generally.

Conclusion

We have shown, through data collected from a large sample of general population adults in the USA, Ghana, and China, that both absorption and inner speech predict voice-hearing experiences. We extend the inner speech model of voice hearing by suggesting that absorption is another factor which plays a generative role in producing voice-hearing experiences. In fact, we suggest that absorption and inner speech work together in producing voice-hearing experiences. Specifically, we argue that absorption as an attitude toward the mind, may be an important, causally antecedent factor to inner speech in hearing voices. In doing so, we believe we can better capture the expansive range of phenomenological experiences represented in voice hearing. Perhaps a voice is less like a misattributed thought due to a deficit and more like a very vivid thought which, because it is felt as vivid, is experienced as sensory.

Limitations

The primary limitation to this study is that it does not include clinical subjects. We also note that we did not preregister the interaction analysis between inner speech and absorption or the moderated mediation model we present in Study 2, and that we are relying on cross-sectional data to support our theory building: Drawing stronger causal conclusions would require preregistered longitudinal or experimental work.