# Phenomenology of Reading - Research Overview
This note documents research on the phenomenology of silent reading, particularly the contested role of [[inner speech]]. It connects to [[Inner Speech and LLM Coupling - Research Discussion]] as foundational context for understanding how we experience LLM-generated text.
---
## The Empirical Conflict
Two studies produced dramatically different findings about inner speech during reading:
| Study | Method | Inner Speech Frequency |
|-------|--------|----------------------|
| [[Moore & Schwitzgebel (2018)]] - "The Experience of Reading" | Probe-based questionnaires | ~59% of samples |
| [[Brouwers, Heavey & Hurlburt (2018)]] - "Pristine Inner Experience While Silent Reading" | Descriptive Experience Sampling | ~3% of samples |
This is, as Hurlburt puts it, "a shockingly huge discrepancy."
### What Explains the Difference?
**Hurlburt's critique of questionnaire methods:** When you ask "do you experience inner speech while reading?" you prime participants to look for inner speech. [[Descriptive Experience Sampling]] tries to "bracket presuppositions"—training participants to report what's actually happening at the beep, not what they assume is happening.
**Moore & Schwitzgebel's reply:** DES has complementary shortcomings. The intensive method may disrupt the experience it's trying to capture. Apples-to-apples comparison shows smaller discrepancy than headline numbers suggest.
---
## What the DES Research Found
When Brouwers et al. sampled 16 college students reading Fitzgerald and Hemingway:
- **Visual imagery was frequent** (not inner speech)
- About **25% of samples** involved words in some form, but:
- Focus on *specific words* from text (not "reading aloud internally")
- Words *innerly spoken in response* to the text (commentary, not narration)
- Innerly spoken words *unrelated* to the text
- Only **~3%** involved actually "narrating the text" in inner speech
This suggests continuous inner vocalization may not be phenomenologically accurate for most reading.
---
## The Five Frequent Phenomena of Inner Experience
[[Russell Hurlburt]]'s broader research identifies five frequent phenomena:
1. **Inner speaking** (sometimes called "inner speech")
2. **Inner seeing** (visual imagery)
3. **[[Unsymbolized thinking]]** - thought directly present without words, images, or other symbols
4. **Feeling** (experience of emotion)
5. **Sensory awareness**
Each occurs in roughly a quarter of samples. Crucially: **individuals differ enormously**. Some rarely have inner speech at all; others have it constantly. There is no "normal" reading experience.
---
## Unsymbolized Thinking: A Deeper Look
This is a real term from Hurlburt's Descriptive Experience Sampling research—not a metaphor but a claimed phenomenological category.
### Definition
**Unsymbolized thinking**: Experiences where you're clearly thinking something specific—a definite thought with particular content—but without any words, images, or other symbolic vehicles present in awareness. You simply *have* the thought directly.
### Hurlburt's Example
A participant is thinking "I should call my sister" but:
- No words are present in experience
- No image of the sister
- No auditory "hearing" of the sentence
- The thought is simply present, contentful, but unsymbolized
### Frequency Claims
Hurlburt claims unsymbolized thinking appears in roughly **25% of random samples** of inner experience—making it as common as inner speech or visual imagery.
### Why It's Controversial
The claim challenges a widespread assumption: that all conscious thought must be *in* some representational format (language, imagery, etc.).
**Skeptical position ([[Eric Schwitzgebel]]):**
- Participants may be inaccurate reporters of their own experience
- What feels like unsymbolized thinking may be fleeting or peripheral symbolic content that's missed
- Introspection is unreliable, especially for subtle distinctions
**Hurlburt's defense:**
- DES methodology involves extensive training to "bracket presuppositions"
- Participants learn to distinguish what's actually in experience from what they assume must be there
- The phenomenon is robust across many studies and participants
### Theoretical Implications
If unsymbolized thinking is real and common:
1. **Thought doesn't require language** - at least not in a token sense. You can have a complete, determinate thought without any words in awareness.
2. **The architecture of conscious thought** may be more diverse than assumed. Not all cognitive content need be tokened in sensory-like formats.
3. **The "language of thought" hypothesis** (in its strong form) is challenged—or at least, the claim that conscious thought requires vehicles in the LOT.
4. **Reading may involve direct meaning-apprehension** without inner speech—grasping content without phonological mediation.
### Connection to Reading
If ~25% of general experience involves unsymbolized thinking, and only ~3% of reading involves narrating text in inner speech (Hurlburt's DES finding), then a significant portion of reading may involve:
- Visual imagery (seeing what's described)
- Unsymbolized thinking (grasping meaning directly)
- Sensory awareness (of the text, environment)
- Emotional response
This complicates models that assume reading *must* involve inner speech or phonological processing.
---
## Content-Dependent Reading Experience
[[Eric Schwitzgebel]] and Moore found reading experience varies with passage type:
- **Dialogue passages** → more inner speech reports
- **Vivid visual description** → more visual imagery reports
This suggests immersed reading may involve something other than inner speech—perhaps imagery, unsymbolized thinking, or direct meaning-apprehension.
---
## Pierini on Inner Speech and Poetry
[[Francesco Pierini]] (2025) - "Inner speech and the phenomenology of poetry" (*Philosophy and Phenomenological Research*) addresses the **form-content unity thesis** in poetry.
### The Dual-Aspect Structure
Pierini argues inner speech has two indissolubly connected aspects:
1. **Auditory aspect**: You experience the *sounds* of words (rhythm, stress, rhyme, tone)
2. **Semantic aspect**: You experience the *meanings* of words
Key claim: **these aspects cannot be experienced in isolation**. When you inner-speak a word, you don't first hear sound then get meaning—you get a unified experience where sound and meaning are fused.
### Why This Matters for Poetry
Poetry exploits this fusion. The sounds *are* the meaning-bearers; the meanings are *sonically textured*. This is why paraphrase loses something—changing sounds changes the meaning-experience.
---
## Subvocalization vs. Inner Speech Proper
Important distinction from the literature:
- **Subvocalization/Phonological processing**: Computing phonological information when reading—can be unconscious, involves minimal laryngeal movements
- **Inner speech proper**: Conscious experience of "hearing" a voice while reading
Evidence from Abramson & Goldinger (1997): silent readers take longer on words with long vowel sounds, showing phonological access even without conscious "hearing."
Much of reading happens *below* conscious awareness—the automaticity that skilled reading develops.
---
## Implications for LLM Coupling Research
This research complicates and enriches the [[Inner Speech and LLM Coupling - Research Discussion|hybrid inner speech]] thesis:
### Challenge to the Thesis
If only 3% of reading involves inner speech narration, then reading LLM output may not primarily engage inner speech machinery. It might involve:
- Visual imagery
- [[Unsymbolized thinking]]
- Direct meaning-apprehension without symbolic vehicle
The "hybrid inner speech" framing may need revision: perhaps LLM dialogue involves large amounts of **unsymbolized cognition** with only occasional symbolic punctuation.
### New Research Questions
1. **Phenomenal texture**: Does LLM text have different phenomenal texture than authored text? Is reading it more "opaque" (attending to how it says things) vs "transparent" (seeing through to meaning)?
2. **Authorless form-content unity**: Pierini's account links auditory-semantic unity to poetry. With LLM prose, is there form-content unity? If so, what does "authorial voice" mean when there's no author?
3. **Attention and decomposability**: Can you attend to LLM text's style independently of its content? If more attentionally decomposable than poetry or authored prose, that's phenomenologically significant.
4. **Immersion**: Does knowing text is LLM-generated prevent full immersion? Are you always somewhat "outside" the text?
5. **Unsymbolized understanding**: Is there such a thing as grasping LLM meaning without inner speech—directly apprehending content without auditory phenomenology?
---
## Sources
- [[Brouwers, Heavey & Hurlburt (2018)]] - "Pristine Inner Experience While Silent Reading" - *Journal of Consciousness Studies*
- [[Moore & Schwitzgebel (2018)]] - "The Experience of Reading" - *Consciousness and Cognition*
- [[Francesco Pierini]] (2025) - "Inner speech and the phenomenology of poetry" - *Philosophy and Phenomenological Research*
- [[Stanford Encyclopedia of Philosophy]] - Inner Speech entry
- [[Russell Hurlburt]] - Descriptive Experience Sampling research
- [[Peter Kivy]] - *The Performance of Reading* (2006)
---
## Connection to Related Notes
- [[Inner Speech and LLM Coupling - Research Discussion]] - Main theoretical framework
- [[Cognitive Phenomenology of Literature]] - Readwise highlights on Kivy's account
- [[Transparency and the Mindfulness Opacity Hypothesis]] - Transparency vs opacity in experience