Abstract
Inner speech is frequently assessed using self-report scales, but their validity is understudied. Uttl et al. (2011) found moderate correlations, perhaps because measures tap into different dimensions of inner speech. We expand on these preliminary results by investigating reliability and concurrent validity of seven inner speech questionnaires in a larger sample. Our results indicate that inner speech questionnaires are reliable but hold moderate concurrent validity, in line with Uttl and colleagues’ (2011) results. Specifically, our results suggest that some inner speech scales may capture a general conception of inner speech, while others may assess evaluative components of negative self-talk, self-regulation, and self-reflective processes, but not emotional valence. The results hold implications around further validity investigations of inner speech measures.
Introduction
Inner speech is often characterized as “the little voice in our head” (e.g., Perrone-Bertolotti et al., 2014), representing silent speech internally generated by oneself (Morin, 2012). Inner speech serves various cognitive functions, among which are self-regulation, problem solving, decision-making (Alderson-Day & Fernyhough, 2015), memory (Larsen et al., 2002), and self-reflection (Morin, 2018). Inner speech can be phenomenologically experienced as condensed, dialogical, spontaneous, effortful (Vygotsky, 1943/62), and containing the voices of other people (Toh et al., 2022). Its frequency varies widely but has been estimated as occurring on average 1/4 of our waking life (Heavey & Hurlburt, 2008).
Objectively assessing inner speech is challenging because it can be measured only indirectly (e.g., recording children’s private speech, adult think out loud, thought listing/sampling, articulatory suppression). By far the most popular method is the administration of self-report questionnaires, which has a long history in psychology (Gault, 1907). One main limitation of self-report measures is that they rely on retrospection and pre-existing views as to what inner speech functions theoretically are. The main question addressed here is: Are current inner speech questionnaires valid—that is, do they all measure the same construct? Uttl et al. (2011) showed that totals on four self-report measures of inner speech are reliable and somewhat valid. Our goal was to expand on these results by testing concurrent validity and reliability of seven inner speech questionnaires, including more recent ones, and by examining their subscales. We predict replicating and refining results previously obtained. We selected scales that specifically assess inner speech, as opposed to other constructs which are indirectly associated with it, such as imagined interactions (Honeycutt, 2015) and internal dialogues (Oles et al., 2020).
Methods
Participants, Procedure, Data Screening, and Analysis
Mount Royal University students (n = 547) completed an approved 30-minute questionnaire package via LimeSurvey for credit in the Psychology Program. We visually inspected the data, removed all reports containing large missing portions, or reports appearing to be random (e.g., those who took 5 minutes only to complete the whole study). We used SPSS v. 24 to perform normality, outlier, correlational, and reliability analyses. We did not perform Bonferroni corrections (neither did one of our sources for comparison—Uttl et al., 2011) because they are too conservative (increasing Type II error) for the number of comparisons we conducted. We performed 105 correlations, which would make the corrected alpha level below 0. Note that none of the scales described below contain the same questions, although many of the questions are conceptually similar.
Measures
Varieties of Inner Speech Questionnaire—Revised
The Varieties of Inner Speech Questionnaire—Revised (VISQ-R; Alderson-Day et al., 2018) uses 21 items assessing five factors on a 7-point Likert scale (from never to all the time); it is psychometrically acceptable in clinical and nonclinical contexts, and used in different cultures (e.g., Perona-Garcelán, 2017). The VISQ-R measures the phenomenological qualities of inner speech (dialogic, other people’s voice, condensed), as well as some of its functions (evaluative and motivational, and positive inner speech). An example of an item is “I talk to myself silently in an encouraging way”. We adopted all aforementioned subscales with the exception of the condensed inner speech subscale as face validity tells us that other measures do not capture this phenomenon.
Self-talk Scale
The 16-item STS (Brinthaupt et al., 2009) uses a 6-point Likert scale (from never to always) to measure self-reported frequency of self-talk (both inner speech and self-directed outer speech) in non-clinical populations. The STS is commonly used (e.g., Heavey et al., 2019), for example in performance contexts (e.g., Shi et al., 2017) because its factors focus on self-regulation via social assessment, self-criticism, self-management, and self-reinforcement (e.g., “I want to reinforce myself for doing well”).
General Inner Speech Questionnaire—Revised
The 75-item, 6-point (from never to all the time) is based on the exploratory General Inner Speech Questionnaire (GISQ; Racy et al., 2019). The GISQ-R asks what people think they talk to themselves “about” (e.g., career goals), use of inner speech “in order to” do things (e.g., plan in my head), and “when” they use inner speech (e.g., managing my eating habits), with an overall score representing perceived frequency of inner speech content and functions in general. Exploratory factor analysis suggested 15 factors requiring refinement (e.g., cognition, self-reflection, mental time travel, language, goals), and needs confirmatory analysis before more valid insights can be drawn, so we used a total score and took caution in interpretation.
Automatic Thoughts Questionnaire—Revised
The 5-point (not at all to all the time) Automatic Thoughts Questionnaire—Revised (ATQ-R; Kendall et al., 1989) measures frequency of cognitive self-statements associated with depressed mood, and comprises 30 negative self-statements such as “I hate myself”, and 10 reverse-scored self-verbalizations associated with positive affect, such as “I’m proud of myself”. The ATQ-R total score is obtained by inverting the 10 non-negative items and summing all items; a higher score indicates a higher frequency of depressive thinking.
Nevada Inner Experiences Questionnaire—Inner Speaking Subscale
The Nevada Inner Experiences Questionnaire (NIEQ; Heavey et al., 2019) contains five subscales estimating the frequency of inner experiences which have been identified using a thought sampling method— Descriptive Experience Sampling (DES; see Heavey & Hurlburt, 2008). These experiences are inner speaking, inner seeing, unsymbolized thinking, feelings, and sensory awareness. Here we were uniquely interested in the two inner speaking items (e.g., “How frequently do you talk to yourself in your inner voice?”). Estimates of frequency are obtained using a visual-analog scale from Never (0%) to Always (100%). For correlational purposes, we modified the coding so that Never corresponded to 1 and Always to 10. We also interpret these results cautiously, focusing on content validity rather than frequency, as the visual-analogue questionnaire was originally designed as an alternative to using Likert scales as frequency measures.
Self-verbalization Questionnaire
The Self-verbalization Questionnaire (SVQ; Duncan & Cheyne, 1999) is a 27-item questionnaire designed to measure the use of private speech (overtly vocalized speech directed at oneself) for self-regulatory purposes. It assesses three dimensions of self-regulation: cognitive, mnemonic, and attentional. Participants indicate their agreement or disagreement with each statement using a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). One sample item is “I sometimes plan my actions out loud when I’m getting organized”.
Inner Speech Scale
The Inner Speech Scale (ISS; Siegrist, 1995) is a 22-item questionnaire designed to measure the frequency of inner speech for self-reflection purposes, after Morin’s (2005, 2018) hypothesis of a link between these two constructs. Participants are invited to indicate the degree of agreement or disagreement with each statement using a 6-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). The scale was developed using German speakers and translated by the author into English. One example of an item is “Inner conversations help me to understand myself better”.
Results
Correlational matrix representing construct validity of several inner speech measures.
Note: All relationships are significant (p < .01) unless otherwise indicated; * = significant at the p < .05 level.
Discussion and Conclusion
Taken together, our results suggest that inner speech questionnaires sampled in the current study are generally reliable but hold moderate concurrent validity—a conclusion in accord with results obtained by Uttl and colleagues (2011). More specifically, the ISS, GISQ-R, and SVQ relate positively with most other measures, suggesting that these may capture a general conception of inner speech. Notably, evaluative inner speech, as measured by one subscale of the VISQ-R, was moderately related to most other scales and subscales, arguably because evaluation underlies many self-regulatory processes (e.g., STS subscales of self-criticism, management, social assessment) and self-reflective functions (e.g., ISS, GISQ-R), as well as negative thoughts (ATQ-R). Also, the ISS robustly correlated with all other questionnaires, suggesting that the self-reflective functions of inner speech (Morin, 2018) entail self-regulatory (STS, SVQ) and dialogic (VISQ-R dialogic) processes, and taps into inner speech in general (NIEQ, GISQ-R).
Interestingly, and mirroring our own results, the lowest correlations Uttl and colleagues (2011) observed were between the SVQ, STS, ISS, and the Self-Talk Inventory (STI; Calvete et al., 2005), the latter being comparable to the ATQ-R in that it assesses valence of inner speech as opposed to functions (e.g., STS) and content (e.g., GISQ-R). In other words, perhaps the ATQ-R weakly correlates with other inner speech questionnaires because it does not feature any factor targeted by these, such as self-reflection and self-regulation. Additionally, one could argue that “automatic thoughts” may be experienced in non-inner speech formats such as mental images. Thus, these results show how people think they experience inner speech in their every day lives, and points to the importance of considering valence, phenomenology, contents, and functions when choosing inner speech measures because each may have different implications for cognition and behaviour.
Likert scales used by our questionnaires may ask participants to determine if they talk to themselves about specific things or for particular reasons “rarely” or “often”—a rather vague frequency measure bound to be interpreted differently by volunteers. It is also important to note that our results are probably not generalizable beyond Canadian undergraduate students, who are characterized by unique gender, developmental, and cultural qualities. Another limitation in this study is the use of a large number of multiple comparisons, for which we did not correct because of the risk of Type II error (i.e., Bonferroni is too conservative), although not correcting also runs the risk of increased Type I error. Future studies will need to explore less conservative, more effective forms of correcting for multiple comparisons. For example, future researchers may want to use confidence intervals and bootstrapping in similar studies.
Although concurrent validity is not the only type of validity, one next research step in assessing the validity of these instruments could consist in comparing inner speech frequencies as measured by self-report questionnaires with thought sampling methods such as the DES (e.g., Hurlburt, 2017), where participants report what type of inner experience (including inner speech) they had right before hearing a beeper probe in their natural environment. Indeed, Hurlburt, Heavey, Lapping-Carr and colleagues (2021) recently offered such a comparison, suggesting that questionnaires substantially overestimate frequencies of inner speech occurring in everyday life. Such a comparison may be problematic however, as questionnaires probe inner speech use in specific situations (e.g., during self-regulatory efforts), whereas DES methods estimate it in general (Brinthaupt et al., 2022).
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The article processing fees were waived through Mount Royal University’s participation in the Canadian Research Knowledge Network (CRKN)-SAGE Publishing agreement found here: 
.