Should Filler Words and Pauses Be Preserved in Research Transcripts?

Should You Remove Every “Um,” “Uh,” and Pause From a Research Interview?

You have completed an interview and received an AI-generated transcript. The participant's answers look surprisingly polished. There are no filler words, repeated phrases, false starts, or indications of hesitation. Every response reads like a carefully written paragraph.

But the original conversation sounded different. The participant paused before answering certain questions, repeated some expressions, and occasionally changed direction halfway through a sentence.

Should these features be preserved, or are they unnecessary distractions? The answer depends on what your research is trying to understand and whether removing conversational details could change the evidence available for analysis.

Preserve What Matters to the Analysis, Not Every Sound Automatically

How Filler Words and Pauses Affect Qualitative Transcription

What Are Filler Words, Pauses, and Speech Disfluencies?

Spontaneous speech rarely resembles polished academic writing. Speakers pause, repeat words, restart sentences, and use expressions such as "um," "uh," or "you know" while speaking.

In speech research, disfluencies commonly include filled pauses, repetitions, repairs, and false starts. A silent pause is a period without speech, whereas a filled pause contains a vocal expression.

These features may reflect speech planning or interactional processes, but their significance depends on the conversational context. Researchers should not automatically interpret every hesitation as uncertainty, discomfort, or reluctance.

What Is the Difference Between Verbatim and Cleaned Transcription?

Research transcription practices differ in how much spoken detail they preserve. Terms such as full verbatim, clean verbatim, and intelligent verbatim are commonly used in transcription services, but their definitions are not universally standardized across research traditions.

The labels alone are insufficient. Researchers should specify exactly which features are retained or removed. A transcript described as "verbatim" may still omit pauses, overlapping speech, or nonverbal vocalizations unless its conventions are documented.

Why Might Filler Words Be Analytically Relevant?

Expressions such as "um" and "uh" can occur during speech planning and may influence how listeners process an utterance. Clark and Fox Tree (2002) examined the communicative functions of these expressions, arguing that they can serve as conventional signals associated with delays in speaking.

However, the presence of a filler word does not establish a participant's psychological state. A participant who says "um" before answering may simply be planning a sentence.

Filler words become especially relevant when the research examines conversational organization, response formulation, or interactional behavior rather than only the topics discussed.

When Do Pauses Matter?

Pauses may be important when researchers examine turn-taking, conversational timing, or how responses develop during interaction.

In conversation analysis, transcription conventions developed by Gail Jefferson provide detailed ways to represent pauses, overlapping talk, and other features of spoken interaction. Such notation allows researchers to examine conversational sequences more closely.

A study exploring how participants respond to sensitive questions may also require attention to pauses, although the researcher must avoid assuming that silence has one fixed interpretation.

By contrast, a thematic analysis focused on teachers' reported uses of educational technology may not require every brief pause to be represented in the working transcript.

The appropriate level of detail follows the analytical purpose, not a universal rule that more transcription symbols necessarily produce better research.

Can Removing Filler Words Change Participant Meaning?

Yes, particularly when automated cleanup removes expressions that function as qualifications rather than merely speech-planning sounds.

Consider the statement, "Well, I think it might help students." Removing "well," "I think," and "might" would produce "It helps students," which expresses a substantially stronger claim.

Not all these expressions are interchangeable with filler words. "Might," for example, is a modal verb expressing possibility. Treating every seemingly unnecessary expression as removable speech clutter can erase meaningful distinctions.

Researchers should distinguish between deleting a nonessential filled pause and changing the apparent meaning of a participant's statement.

What Does Qualitative Methodology Say About Transcription Decisions?

Poland (1995) identified transcription quality as an important aspect of qualitative rigor. His work emphasized the need to examine the relationship between recorded speech and its written representation.

Oliver, Serovich, and Mason (2005) discussed transcription as a process involving methodological choices rather than a neutral conversion of sound into text. Their analysis showed how transcription conventions can affect the representation of participants and subsequent interpretation.

Davidson (2009) similarly reviewed transcription as a topic in qualitative research, highlighting the importance of transcription practices and their relationship to research methodology.

These studies support a context-sensitive approach. They do not establish that every qualitative interview must use the most detailed possible transcription system.

How Does AI Transcription Handle Pauses and Repetitions?

AI transcription tools vary in how they represent disfluencies. Some systems omit filled pauses or repetitions as part of their default output. Others offer verbatim modes, word-level timestamps, or options for including disfluencies.

Researchers should distinguish between a system that fails to recognize a spoken feature and one that deliberately removes it during post-processing.

A transcript may also contain punctuation inserted automatically to improve readability. Such punctuation does not necessarily represent measured pauses or the exact timing of speech.

If the study requires precise pause duration, the researcher may need to examine the recording using appropriate audio-analysis tools rather than relying on punctuation or automatically generated timestamps.

Should Every Pause Be Assigned a Psychological Meaning?

No. Silence may occur for many reasons, including speech planning, turn-taking, distraction, technical delay, or consideration of a question.

A pause before discussing a difficult experience may be relevant to the analysis, but the pause alone does not establish embarrassment, dishonesty, anxiety, or emotional distress.

Interpretation should draw on the broader interaction, research design, and available evidence rather than treating isolated vocal features as diagnostic indicators.

How Should Researchers Establish a Transcription Convention?

Before preparing transcripts, researchers should determine which features are necessary for their intended analysis.

A practical convention may specify how to represent filled pauses, repetitions, incomplete sentences, laughter, overlapping speech, silence, and unintelligible passages.

For example, a researcher might retain meaningful repetitions and qualifications while omitting routine filled pauses that do not affect a content-focused analysis. Another study may require timed pauses and detailed overlap notation.

Consistency is important because different treatment across participants may create artificial differences in the transcripts.

The convention should also guide manual checking of AI-generated transcripts, particularly when automated cleanup has removed features that the study requires.

Deciding Whether to Preserve Hesitation in a Teacher Interview

The carefully cleaned version removes the filled pause and repetition while retaining the participant's uncertainty. The overedited version removes the qualification and transforms a tentative statement into an affirmative conclusion.

The example illustrates that a readable transcript can remain faithful without reproducing every speech-planning sound. The boundary is crossed when editing changes the participant's expressed position.

Common Misunderstandings About Filler Words and Pauses

Which Transcription Detail Is Appropriate for Your Study?

Start with the research question and analytical method. The purpose of transcription is to produce a representation of the recording suitable for the kind of interpretation you intend to undertake.

Researchers should apply the chosen convention through a consistent transcript verification workflow. The original recording should remain available under the study's approved data management arrangements so that important passages can be checked.

A useful transcript is not necessarily the longest or cleanest version. It is the version that preserves the evidence required for the analysis without introducing unsupported interpretations.

Before Removing Filler Words and Pauses

Frequently Asked Questions About Filler Words and Pauses

Preserve the Conversational Details Your Research Needs

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.