Someone pauses before answering a difficult question and reports that a whole sentence was somehow present, although no full sentence seemed to pass through the mind. Another person remembers a few words and a felt direction of thought. A third reports no words at that moment. A researcher cannot responsibly give these three reports the same code. Condensed inner speech concerns an experienced verbal thought that is abbreviated, fragmentary or less fully articulated than ordinary overt speech. Measuring it requires attention to what was experienced, how the report was elicited and what the method can miss.
This Methods & Measures entry concerns tools for studying that quality of inner speech. It is distinct from dialogic inner speech, which describes an exchange-like form, and from reporting little or no inner speech, which concerns frequency or presence. A brief inner phrase can be dialogic or monologic. A person who reports no speech at one sampled moment may have speech at others. A low questionnaire score on condensation cannot be read as proof that inner speech is absent or always fully formed.
Define the construct before choosing a scale
Ordinary spoken language unfolds as words, syntax and timing. Inner speech may retain some of those properties without retaining all of them. Researchers use “condensed” to name experiences in which a thought is felt as verbal but shortened, compressed or only partly specified. This is a phenomenological claim about how thinking seems to a person, not a measurement of the number of silent syllables produced in the brain. A participant may report the gist of an utterance with few experienced words, yet there is no direct way to reconstruct the unspoken full version.
McCarthy-Jones and Fernyhough’s Varieties of Inner Speech Questionnaire, or VISQ, treated condensed inner speech as a factor distinct from dialogic, evaluative or motivational qualities and the presence of others’ voices. That separation is central to construct validity. A report of an imagined back-and-forth does not tell us whether each turn is expanded or abbreviated. A fragment such as “later, maybe” does not tell us whether the person was addressing themselves, imagining someone else’s reply, or merely noting an intention.
The revised questionnaire, VISQ-R, further refined dimensions of reported inner-speech phenomenology. A questionnaire score summarizes endorsement of statements about typical experience. It does not count actual utterances in a given hour or directly verify their form. People differ in what they call a “word,” “voice” or “sentence” internally. A useful study states the items, response scale, scoring direction and time frame, rather than treating the score as a transparent quantity of condensation.
Fernyhough and Borghi’s review of inner speech as a language process and cognitive tool supports studying its varied forms rather than forcing one uniform model. It provides conceptual context, not a validation study of a particular condensation scale. For measurement claims, the original and revised questionnaire studies, momentary sampling work and phenomenological reports are more directly informative.
What a questionnaire can and cannot capture
A questionnaire can efficiently ask many people whether they recognize a recurring quality in their experience. It can test whether responses cluster into distinguishable dimensions and whether a proposed item behaves similarly across samples. It is especially useful for comparing broad tendencies if the wording is understandable and the factor structure holds. The VISQ and VISQ-R offer such operationalizations of condensation, but a factor label is a model-based summary of item responses, not proof that all respondents experience the same mental process.
Item wording matters. “I think in brief phrases” can invite a judgment about verbal length, while “I know what I mean without complete words” may also capture nonverbal thought. An item that asks about an “inner voice” may miss people whose inner speech is verbal but not heard as sound. If a scale combines these descriptions, a high score may represent several experiences. Researchers should inspect the actual items and avoid defining condensation solely from a subscale name.
Translation is not a mechanical step. Languages differ in typical pronoun use, ellipsis and the apparent completeness of fragments. Cultural norms can affect whether someone interprets inner speech as speaking, thinking, hearing or simply knowing. When using a scale across languages, a translated factor structure and comprehension testing are more informative than assuming that the same English item describes the same experience everywhere. Even a replicated factor structure does not establish identical meaning for every respondent.
Retrospective self-report also invites reconstruction. A person may know that their thinking often feels quick and therefore endorse condensed items, without accurately recalling particular moments. Someone who values verbal precision may classify the same fragment as “not really speech.” This does not make questionnaire data useless; it defines what they measure: considered reports about typical experience. To claim momentary prevalence or temporal change, collect data closer to the experience itself.
Momentary sampling changes the question
Experience sampling interrupts people during daily life and asks what was occurring just before an alert. Alderson-Day and Fernyhough compared general VISQ endorsements with reports collected through a smartphone app. Their study found that generalized endorsements and randomly sampled reports did not line up uniformly across all subscales, and momentary endorsements were lower. That result cautions against reading a typicality score as a direct count of episodes. It does not establish that questionnaires are false or that an app captures every fleeting experience.
A momentary prompt needs careful timing. “What were you thinking?” can elicit an explanation created after the alert; “what, if anything, was present just before the signal?” tries to reduce that reconstruction. Participants may still need seconds to notice, label and report a rapidly fading event. The prompt itself can turn attention inward and alter the next experience. Researchers should separate the target moment from the reporting moment and record response delay where possible.
Sampling frequency affects coverage and burden. A few alerts may miss rare forms of inner speech; many alerts may disrupt ordinary activity and encourage participants to monitor themselves. Alerts are not guaranteed to sample all contexts equally, because people may ignore them while driving, working or conversing. A report that condensed speech appeared in a certain proportion of answered prompts concerns those answered prompts, not automatically all waking thought.
Momentary coding should allow multiple qualities. A participant might experience a brief internal exchange with two imagined positions, each expressed only as a fragment. Coding it as either “dialogic” or “condensed” would impose an artificial choice. Similarly, a fragment can be evaluative and self-directed. Multi-label coding, together with clear definitions, preserves the dimensional structure that the questionnaire work identifies.
Descriptive interviews and the difficulty of articulation
Hurlburt, Heavey and Kelsey’s Descriptive Experience Sampling work uses randomly signaled moments followed by detailed interviews intended to clarify the experience at the instant of the signal. This approach can reveal distinctions that a fixed questionnaire might miss: inner speaking, inner hearing, unsymbolized thinking and other forms of experience need not be assumed identical. It is particularly valuable when a participant’s first phrase, “I said it to myself,” could refer to a fully worded utterance, a fragment or a post hoc paraphrase.
The interview is not a direct recording of thought. It depends on the participant’s memory of a just-past experience, the interviewer’s questions and the ability of both to distinguish description from explanation. Intensive training can improve the clarity of reports while also changing what participants notice. A method that produces rich descriptions from a small group cannot by itself estimate prevalence across a population. The relevant validity question is whether the procedure helps identify the target quality without leading the participant toward the investigator’s preferred categories.
A good interview asks for the exact form the person experienced, while allowing “I cannot tell” as an answer. Asking “What sentence was in your head?” presumes a complete sentence. Asking “Was it verbal at all, and if so which parts were present?” leaves room for a fragment. The distinction matters because a participant may convert a condensed experience into a grammatical sentence while explaining it. The polished report can be longer than the experience it describes.
Qualitative coding benefits from independent raters and examples that include borderline cases. A codebook can distinguish an experienced fragment from a complete sentence remembered only in gist, a nonverbal concept later put into words, and an absence of reportable content. Agreement between coders is useful but not sufficient; two raters can agree on a category that was poorly defined. Researchers should show how disagreements were resolved and preserve ambiguous cases rather than forcing every report into a sharp binary.
Condensation is not dialogue or absence
Dialogic inner speech refers to a conversational quality: a back-and-forth, a question and answer, or a represented other perspective. Condensation refers to how elaborated the verbal content feels. The dimensions can cross. A brief “go?” followed by “not yet” could be both dialogic and condensed. A long internal rehearsal of a speech could be monologic and expanded. A scale or interview that treats dialogue as evidence of condensation confuses two constructs that the original VISQ was designed to distinguish.
Absence raises a different issue. Nedergaard and Lupyan investigated people who reported very little inner voice and introduced the term anendophasia for that reported pattern. Their work can motivate sharper questions about individual differences, but it does not make every low momentary speech report evidence of a permanent absence. Nor is a highly condensed thought automatically close to no thought or no inner speech. It may carry a strong sense of words or communicative form despite little articulatory detail.
A person might also report unsymbolized thinking: a definite thought without experienced words or images. A researcher should not recode that as condensed inner speech merely because the person can later express the thought in words. Almost any thought can be described verbally after the fact. The measurement target is the form of the experience at the sampled moment. The boundary between condensed verbal experience and nonverbal thought may remain genuinely uncertain, which should be recorded as uncertainty rather than concealed.
These distinctions matter in studies of cognition. If participants with low inner-speech frequency perform differently on a memory task, the result does not identify condensation as the mechanism. If a condensation score correlates with another trait, it does not show that inner dialogue is responsible. Multidimensional measurement and planned comparisons are necessary before turning a broad association into a process claim.
Behavioral interference is indirect evidence
Researchers sometimes use articulatory suppression, such as repeating an irrelevant syllable, to disrupt covert verbal processing during another task. If performance changes, verbal resources may have contributed. Yet suppression also taxes attention, timing and motor control. It cannot tell an investigator exactly whether the person had condensed or expanded inner speech at the affected moment. A task effect is indirect evidence about a functional contribution, not a transcript of phenomenology.
A stronger design pairs behavioral manipulation with phenomenological measures while keeping their claims separate. A questionnaire can describe a typical tendency; sampling can document reported episodes; an interference task can test whether occupying speech-related processes changes performance. Convergence across them would be informative, but lack of convergence is also possible because they sample different times and aspects of thought. None should be treated as a gold standard against which the others are automatically judged.
Task instructions can themselves elicit inner speech. Asking people to plan verbally before a task, rehearse a sentence or report their thoughts aloud may produce more expanded language than the thought they would have had spontaneously. A study designed to measure natural condensation should distinguish elicited inner speech from everyday episodes. If it deliberately elicits speech, the inference should be limited to that instructed condition.
Validity, reliability and reporting
Reliability asks whether a measure yields sufficiently consistent information for its intended use. A trait questionnaire may show stable rankings over a short interval while momentary inner speech varies substantially with activity. Those are not contradictory results. The first concerns between-person endorsement; the second concerns within-person state. Researchers should specify which level they intend to measure and avoid treating a small set of beeps as a stable personal score.
Construct validity asks whether “condensed” items actually distinguish abbreviated verbal experience from speed of thought, poor memory for experience or discomfort with introspection. Convergent evidence might include similarly bounded descriptions in interviews; discriminant evidence would show that condensation is not simply identical to dialogic speech, frequency or general verbal ability. A correlation with a task outcome is not enough if the same questionnaire also captures attention, mood or self-consciousness.
Sampling plans should anticipate missingness. Alerts missed during social interaction or intense work may remove contexts where inner speech differs from quiet moments. State how many prompts were sent, answered on time and excluded, and why. If the analysis uses only people who completed many prompts, the sample may favor those comfortable with introspection. The number of observations per person matters for estimating within-person variation and should not be hidden inside a single overall percentage.
Questionnaires and interviews also need careful language around distress. Some people describe voices of others or critical self-talk, but those qualities are not synonymous with condensation and should not be turned into a diagnosis. The original VISQ considered several phenomenological dimensions and correlations with psychopathological variables; a methods page should not imply that one condensation score identifies a disorder. Clinical inference needs separate validated assessment.
What the evidence does not show
The cited studies do not provide a direct count of condensed inner words, a neural signature unique to condensation, or a universal boundary between fragmentary speech and nonverbal thought. A questionnaire factor reflects a pattern of responses under particular wording and samples. A momentary report reflects what a participant could notice and describe after an interruption. An interview can deepen description but cannot recover an exact hidden transcript.
Low endorsement of inner speech does not prove that a person never has it. High condensation endorsement does not prove a person always thinks in fragments. Dialogic form, self-evaluation, frequency and condensation are separable dimensions; evidence about one does not automatically transfer to another. The anendophasia literature raises questions about sparse or absent reported inner voice, but it should not be used to relabel every brief verbal experience as absence.
The evidence also does not show that condensed inner speech is inherently better, worse, more creative or more pathological than expanded inner speech. Functional claims require tasks that test a specified consequence and rule out alternative explanations. Reports about phenomenology should remain reports about experience unless an additional study connects them to behavior. A responsible result names the instrument, time frame, sample and uncertainty.
Reading a result with care
Return to the person who said that a whole sentence was present without full words. A careful investigator would ask whether any words were experienced, whether the person sensed a speaker or addressee, and whether the sentence was reconstructed only while reporting it. A questionnaire could place the person on a tendency scale; momentary sampling could show when such reports arise; an interview could clarify a particular episode. These are complementary but nonidentical observations. This is an Institute synthesis of the cited measurement literature, not a validated diagnostic procedure.
Condensed inner speech is difficult to measure precisely because it can be both meaningful and hard to spell out. The answer is not to force it into a complete sentence or to treat a sparse report as silence. Good methods preserve the distinctions the participant can make, record the distinctions they cannot, and tie any conclusion to the particular form of evidence collected.
Related Learning Hub entries
Research references
- McCarthy-Jones, S., & Fernyhough, C. (2011). The varieties of inner speech: Links between quality of inner speech and psychopathological variables in a sample of young adults. Consciousness and Cognition, 20(4), 1586–1593. View source
- Alderson-Day, B., Mitrenga, K., Wilkinson, S., McCarthy-Jones, S., & Fernyhough, C. (2018). The varieties of inner speech questionnaire – Revised (VISQ-R): Replicating and refining links between inner speech and psychopathology. Consciousness and Cognition, 65, 48–58. View source
- Alderson-Day, B., & Fernyhough, C. (2015). Relations among questionnaire and experience sampling measures of inner speech: A smartphone app study. Frontiers in Psychology, 6, 517. View source
- Hurlburt, R. T., Heavey, C. L., & Kelsey, J. M. (2013). Toward a phenomenology of inner speaking. Consciousness and Cognition, 22, 1477–1494. View source
- Fernyhough, C., & Borghi, A. M. (2023). Inner speech as language process and cognitive tool. Trends in Cognitive Sciences, 27(12), 1180–1193. View source
- Nedergaard, J. S. K., & Lupyan, G. (2024). Not everybody has an inner voice: Behavioral consequences of anendophasia. Psychological Science, 35(7), 780–797. View source