Overview

This chapter explores the complex mental processes behind speech production, from idea generation to physical articulation. We examine how brain damage can cause Foreign Accent Syndrome, analyze different models of speech planning, and investigate various types of speech errors that reveal the hidden stages of language production.

Learning Goals

After studying this chapter, you should be able to:

• Explain the three core stages of speech production and their components
• Describe Levelt’s WEAVER++ model and its step-by-step process
• Identify different types of speech errors and what they reveal about production stages
• Understand how Foreign Accent Syndrome relates to speech planning breakdowns
• Compare WEAVER++ with Spreading Activation models
• Analyze evidence from tip-of-the-tongue experiences and picture-word interference tasks
• Explain the role of self-monitoring in speech production
• Describe how articulation transforms mental plans into physical speech

📖 Required Reading

• Traxler (2nd ed.), Chapter 2, pp. 39–50 — Speech Production & Comprehension (production focus for Week 2).
  Link to Chapter 2
  • speech production
  • speech errors
  • Tip-of-the-tongue experiences
  • Picture-naming and picture-word interference studies

Levelt_1999

Chapter 2 Lecture Notes: Speech Planning & Errors (Part I)

Key Focus: Speech production stages, models, errors, and Foreign Accent Syndrome

I. Opening Hook: Foreign Accent Syndrome (FAS)

What is FAS?

A rare neurological condition where brain damage causes a person to speak with a foreign-sounding accent (not actual “acquisition” of another language).

Key Examples

• Londoner → Scottish accent post-stroke (3 native Scots confirmed; Dankovicova et al., 2001).
• Belgian Dutch speaker → French/Moroccan accent; Norwegian → German accent (Moen, 2000, from shrapnel injury).

Link to Speech Production

FAS disrupts mental/motor steps of speech planning—this chapter explores these steps to explain accent changes.

II. Why Speech Production Matters

• Primary Communication Tool: Speech is how most humans share ideas.
• Myth vs. Reality:
  • ❌ Myth: “Idea → Pick words → Speak” (simple).
  • ✅ Reality: Complex hidden processes connect “idea” to physical speech (focus of this chapter).

III. 3 Core Stages of Speech Production (Griffin & Ferreira, 2006)

IV. Levelt’s WEAVER++ Model: Step-by-Step Production

Model Overview

• WEAVER++ = Word Encoding by Activation and VERification (Levelt et al., 1999; Roelofs et al., 2007).
• Key Insight: Ideas don’t directly activate speech sounds—intermediate steps are required.

High-Level Flow

Conceptualization → Lexical Selection (lemmas) → Formulation (grammar + phonology) → Articulation → Monitoring

(See figures: “Blueprint for the Speaker” and “Levelt Example: Production of COLUMN” for visual breakdowns)

Step 1: Conceptual Preparation (Lexical Concepts)

• Goal: Match nonlinguistic ideas to “lexical concepts” (ideas with a specific word in the language).
• Examples:
  • English has a lexical concept for “female horse” → mare (1 word).
  • No lexical concept for “female elephant” → combine female + elephant (2 concepts).

Step 2: Lexical Selection (Lemmas)

• Trigger: When a lexical concept has a matching word.
• Process: Select 1 word from semantically similar options (e.g., “cat” vs. “rat” vs. “dog”).
• What is a Lemma?
  Intermediate mental representation with:
  1. Semantic info: Word meaning (e.g., “cat” = small furry mammal).
  2. Syntactic info: How to use it grammatically (e.g., “cat” = noun; fits in “The ____ sat”).

Step 3: Morphological Encoding (Morphemes)

• Morpheme: Smallest unit of meaning (e.g., eat = root; -s = 3rd-person tense).
• Goal: Adjust word forms for grammar/meaning.
• Example:
  • Lemma eat → ate (past), eats (3rd-person present), eat (plural present).

Step 4: Phonological Encoding (Phonemes + Syllables)

1. Activate Phonemes: Retrieve speech sounds (e.g., cat → /k/, /æ/, /t/).
2. Syllabification: Organize phonemes into syllables (speech is produced in syllables, not isolated sounds):
   • Metrical structure: Create a frame with accents (e.g., banana → “o o’ o” [accent on 2nd syllable]).
   • Insert phonemes into the frame.

• Evidence: Escorting → “ess-core-ting” (not “ess-cort-ing”)—/t/ groups with -ing.

Step 5: Phonetic Encoding + Articulation

• Phonetic Encoding: Create a “gestural score” (motor plan for lips, tongue, etc.).
• Articulation: Motor system uses the plan to move articulators → sound waves for listeners.

V. Evidence for Production Models

1. Speech Errors (Slips of the Tongue)

Errors reveal stage breakdowns (not random):

2. Tip-of-the-Tongue (TOT) Experiences

• Definition: Know a word, but can’t recall its sound.
• Why It Happens: Lemma (meaning) is active, but phonological info is incomplete (Dell, 1986).
• Evidence:
  • Can guess syllable count or first phoneme (e.g., TOT for sampan → “2 syllables, starts with /s/”).
  • ~40% resolve within minutes (temporary phonological failure).
• TOT Examples:
  1. Scrooge’s first name → Ebenezer.
  2. Small Far East boat → sampan.
  3. Whale secretion for perfume → ambergris.
  4. Stamp collector → philatelist.

3. Picture Naming & Picture-Word Interference (PWI)

A. Picture Naming: Word Frequency Effects

• Experiment: Name objects vs. recognize them.
• Result: Unfamiliar objects take longer to name (Oldfield & Wingfield, 1965)—naming speed depends on how often you produce the word’s sound, not just the concept.

B. PWI Task

• Setup: Words printed over pictures; manipulate word-picture relationship.
• Conditions:
  • ✅ Identity: “lemon” over a lemon → faster (shared lemma).
  • ❌ Semantic Interference: “orange” over a lemon → slower (competing lemmas; Burki et al., 2020).

VI. Alternative Model: Spreading Activation (Dell, 1986)

How It Differs from WEAVER++

Key Phenomena Explained

1. Lexical Bias Effect: Sound errors mostly make real words (e.g., “blip” → “slip” [real] vs. “tlip” [nonword])—feedback from phonology to lemmas boosts real words.
2. Mixed Errors: Errors are semantically + phonetically similar (e.g., “oyster” → “lobster” [seafood + similar sound])—cascading activation boosts overlapping words.

VII. Critique of Lemma Theory (Caramazza, 1997)

• Lemma Theory Claim: Lemmas are universal (same for speech/writing) and store grammar.
• Counterevidence (Brain-Damaged Patients):
  • Modality Dissociation: Struggle with function words (e.g., the) in writing but not speech (or vice versa).
  • Inconsistent Errors: Say dish for a cook picture (speech) but write forks (writing)—no single lemma system.

VIII. Self-Monitoring & Repair

2 Types of Monitoring

• Pre-Output: Detect errors before speaking → no overt mistake (e.g., “instant repairs”).
• Post-Output: Correct errors after speaking.

Factors Influencing Monitoring

• Resource Load: Fast speech = more errors (planning uses resources needed for monitoring).
• Taboo Avoidance: Fewer errors that risk taboo words (e.g., toolkits → no “fooltiks”)—GSR higher for risky pairs (Hartsuiker & Kolk, 2001).
• Phrase Boundaries: Better detection at clause ends (lower planning load; Blackmer & Mitton, 1991).

IX. Articulation: Gestural Score to Sound Waves

1. Articulatory Apparatus

• Vocal Tract: Lips, tongue, velum, glottis → modify air flow to make sounds.
• Articulatory Phonology: Speech plan = gestural score (instructions for articulators; Browman & Goldstein, 1989–1991).

2. Gestural Score

• Definition: Blueprint for articulation—specifies:
  1. Which articulators to use (e.g., lips for /p/).
  2. Constriction location (e.g., alveolar ridge for /t/).
  3. Constriction degree (e.g., full blockage for /p/).
  4. Movement dynamics (e.g., fast lip release for /p/).

3. Coarticulation

• Definition: Gestures for one sound overlap with nearby sounds (efficiency).
• Example:
  • Pool: /p/ has rounded lips (anticipates “oo” vowel).
  • Pan: /p/ has flat lips (anticipates “a” vowel).

X. Revisiting FAS

FAS Causes (Linked to Production Stages)

1. Prosodic Disruption: Abnormal pitch, accent, or pausing.
2. Syllabification Disruption: Syllables spoken in isolation (e.g., “Ess-cort-us” instead of “Ess-core-tuss”).
3. Articulation Errors: Wrong phoneme location (back → front) or missing phonemes (e.g., spl → sl).

Conclusion

FAS = disrupted speech planning/articulation—not “learning” a foreign accent.

🧩 Self-Check Questions

Q1. Why do semantic distractors slow naming more than unrelated words?

Q2. What does a TOT reveal about word form vs meaning?

Q3. Give one morpheme-level error and explain it.

Q4. Why are proper names frequent TOTs?

Q5. How can speakers detect their own errors?

🧰 Key Terms

Lemma, Lexeme, Phonological encoding, Semantic interference, Phonological facilitation, Blocked-cyclic naming, Picture–word interference (PWI), Anticipation, Perseveration, Exchange, Blend, Morpheme error, Monitoring, TOT.