Inside My Speaking Score’s Learning Model
My Speaking Score (MSS) is an AI-powered platform for TOEFL Speaking practice. It uses ETS SpeechRater® and our own AI called VoX to give you fast, data-driven feedback. Thousands of test-takers and teachers use us. Our goal is simple: help you understand how you speak and how to get better—faster.
I recently analyzed 4,853 speaking attempts from 330 learners on our platform. Here’s what the data shows about how people really improve. But first, a bit of theory, because we've changed the game by changing the learning model.
1. How Learning Works in the AI Age
Old-school learning models, like Bloom’s Taxonomy, were built for teachers giving lessons step by step—from remembering facts to creating new ideas.
That model still helps for sure, but it doesn’t fully explain how people learn today, when AI feedback lets you test, reflect, and try again instantly.
That’s why we use L. Dee Fink’s “Significant Learning” model. Fink says real learning happens when you combine knowledge, practice, self-awareness, and motivation. It’s not a ladder—it’s a circle. Each part supports the others.
That’s exactly how My Speaking Score works.
2. The MSS Learning Cycle
Everything on MSS follows one simple loop:
Speak → Measure → Reflect → Adjust
- Speak: You record a TOEFL-style answer under real time limits.
- Measure: SpeechRater® and VoX analyze your speech and give a score.
- Reflect: You see exactly which skills are strong or weak.
- Adjust: VoX suggests what to practice next—then you try again.
This loop builds self-awareness, motivation, and steady progress. It’s not about perfection (don't let perfect be the enemy of good); it’s about small improvements that add up.
3. What’s Behind the Technology
- SpeechRater® analyzes spoken English automatically. It looks at pronunciation, rhythm, grammar, vocabulary, and organization.
- e-rater® analyzes writing. We use it in parts of the platform to understand grammar and vocabulary signals.
- VoX combines these results and shows you clear, actionable feedback.
Important note: The real TOEFL Speaking test is scored by AI and human raters together. MSS is a practice tool, not an official score provider.
4. What We Found
Between July 1–31, 2024:
- 330 learners submitted 4,853 speaking attempts.
- We measured every learner’s progress during their first 30 days.
5. Key Results You Can Understand
A. Overall TOEFL Speaking Score (0–30)
- Median gain: +1 point
- Average gain: +1.35 points
- 46% of learners reached a score of 26 or higher within 30 days
Meaning: Most learners made steady, measurable progress—and almost half hit their TOEFL goal.
B. Improvement by Task Type
Here’s how much learners improved on each TOEFL Speaking question type:
Takeaway: Learners improved on every task type, especially Q1 and Q3.
C. Speaking Speed (Words Per Minute)
- Median increase: +12 WPM
- Average increase: +15 WPM
This means learners spoke faster and more confidently—a good sign of fluency.
But remember: clarity matters more than speed.
D. Progress by Skill Area
We grouped SpeechRater features into the same three categories used by TOEFL raters.
1. Delivery
Average improvement: +0.35 SD
Learners spoke with smoother rhythm, fewer pauses, and clearer vowels.
2. Language Use
Average improvement: +0.25 SD
Vocabulary became more varied and grammar more accurate.
3. Topic Development
Average improvement: +0.19 SD
Ideas were more organized and connected.
Together, these show progress in how learners sound, the language they use, and how they build ideas.
6. The Signals You Can Train
Delivery
- Speaking rate (SR): how fast you speak
- Sustained speech (SS): how long you can talk smoothly
- Pause frequency (PF): fewer unnecessary pauses
- Distribution of pauses (DP): pausing between ideas, not mid-sentence
- Repetitions (Re): fewer repeated words
- Rhythm (Rh): natural stress and intonation
- Vowels (Vo): clear vowel pronunciation
Language Use
- Vocabulary depth (VDe): using precise words
- Vocabulary diversity (VDi): not repeating the same words
- Grammar accuracy (GA): correct forms
- Grammar complexity (GC): longer, more advanced sentences
Topic Development
- Discourse coherence (DC): logical flow and clear connections between ideas
7. Why This Learning Model Works
My Speaking Score fits Fink’s learning model perfectly because it makes you an active learner.
- You see your results instantly.
- You understand what they mean.
- You decide how to improve next.
- You practice again with purpose.
Every time you record, you’re learning how to learn — not just preparing for one test.
8. Responsible Use
- TOEFL Speaking uses AI and human raters. MSS is a practice tool, not an official test.
- Our scores are for training and improvement, not certification.
- We follow the NIST AI Risk Management Framework and the OECD AI Principles to make sure our system is fair and transparent.
9. Key Takeaway
AI won’t take the test for you — but it can accelerate your progress.
With real data, instant feedback, and smart guidance, you learn how to perform at your best.
- +1.35 point average gain on the TOEFL Speaking scale
- +12 WPM faster speech
- Clear progress in Delivery, Language Use, and Topic Development
That’s what data-powered learning looks like.
FAQ
Q1: Is My Speaking Score official TOEFL practice?
No. My Speaking Score is not endorsed by ETS. It uses the same kinds of analytics (SpeechRater® and e-rater®) to help you practice effectively.
Q2: Can my MSS score predict my real TOEFL score?
It’s a strong estimate — but not official. Use it to understand your performance trends and to target weak areas.
Q3: How often should I practice?
The best results come from consistent practice — short, focused sessions 3–5 times per week.
Q4: What’s the ideal speaking speed for TOEFL?
Most high-scoring speakers fall between 130–170 words per minute, depending on accent and clarity.
Q5: What’s the biggest improvement area for most learners?
Delivery — especially pausing naturally, reducing repetition, and improving rhythm.
Q6: How long does it take to reach a 26+ score?
In our July 2024 data, about half of learners reached 26 or higher within 30 days of consistent use.
Legal Notice
TOEFL®, TOEFL iBT®, SpeechRater®, and e-rater® are registered trademarks of Educational Testing Service (ETS).