The Flywheel Experiment: Can You Measure Your Way to Fluency?

Nearly twenty years ago, I stopped learning French.

I had studied it through middle school and high school, then continued for a couple of years in college. After that, life moved on. French did not disappear completely, but for all practical purposes, my serious study of the language went dormant.

Almost two decades later, I picked language learning back up.

This time, however, I approached it differently.

I am older. I have a career, a family, and considerably less free time than I did when I first sat in a French classroom. I also have years of experience working in business, finance, procurement, and operations. I spend a large part of my professional life thinking about systems, measurements, forecasts, processes, and the cumulative effects of thousands of small decisions.

Naturally, I started wondering whether I could apply some of that thinking to language learning.

That question eventually became The Flywheel Experiment.

The Problem With Measuring Language Learning

There is no shortage of advice about how to learn a language.

Read books. Listen to podcasts. Study vocabulary. Take lessons. Have conversations. Use an app. Watch television. Immerse yourself. Study every day.

Most of that advice is probably useful.

What I find much more difficult to answer is a simpler question:

How much?

How much reading does it take to noticeably improve?

How many hours of listening separate one stage of comprehension from another?

How much does speaking matter relative to reading?

What happens after 100 hours of serious exposure? What about 500? Or 1,000?

What happens when you stop using a language for a month, a year, or, in my case with French, the better part of two decades?

And can all of those different experiences be measured well enough to make useful predictions about where a learner is going?

Language proficiency is commonly described using levels such as those in the Common European Framework of Reference for Languages, better known as CEFR. A1 and A2 describe basic users, B1 and B2 independent users, and C1 and C2 increasingly proficient users.

CEFR gives us a useful way to describe what someone can do with a language.

What interests me is the road between those levels.

The Flywheel Experiment

I have started building a system to track that road.

At its center is an experience point system that converts different forms of language practice into a common measurement called XP. It tracks the work I put into a language over time and gives me a way to compare that accumulated effort with changes in actual proficiency.

The details of that system deserve their own article, and I will publish them separately.

For now, the important part is the hypothesis behind it:

Can accumulated language-learning activity help predict CEFR proficiency?

I do not mean perfectly. Language learning is far too complicated for me to believe that a single number can completely describe someone's ability.

Two people can spend the same amount of time learning a language and emerge with very different abilities. Even within one person, reading, writing, listening, and speaking can develop at very different rates.

That is part of what makes the question interesting.

If I track enough of the process, perhaps patterns will emerge. Perhaps certain kinds of practice will prove more valuable than I expected. Perhaps progress will accelerate at particular stages. Perhaps long periods away from a language will reveal how much knowledge is actually lost and how much merely becomes dormant.

Or perhaps the entire attempt to predict CEFR progress with a system like this will fail.

That would be useful information too.

The important thing is that I am making the predictions before I know the answers.

Why a Flywheel?

The name The Flywheel Linguist comes partly from another interest of mine: rowing.

Anyone who has spent time on a rowing machine understands the basic behavior of a flywheel. From a dead stop, getting the wheel moving takes effort. Each stroke adds energy. Once the wheel is spinning, the next stroke does not have to start from zero.

Language learning feels remarkably similar.

When you know very little of a language, everything is difficult.

A paragraph can take several minutes to understand. A native speaker sounds impossibly fast. A page might contain dozens of unfamiliar words. Even simple sentences can require conscious thought.

Then you learn a little more.

Vocabulary makes reading easier.

Easier reading allows you to read more.

Reading reinforces vocabulary and grammar.

Better vocabulary improves listening comprehension.

Better comprehension allows you to consume more interesting material.

More interesting material makes it easier to spend more time with the language.

Each turn of the wheel helps create the conditions for the next.

This is where I think language learning becomes particularly interesting. The value of learning another thousand words is not limited to knowing another thousand words. Those words make books easier to read, conversations easier to follow, and unfamiliar words easier to understand from context. Those improvements allow you to consume more of the language, which creates still more opportunities to learn.

Progress begins producing the conditions for more progress.

The wheel gets heavier with accumulated knowledge, but it also becomes harder to stop.

That is the flywheel.

What Happens When the Wheel Stops?

My history with French makes this question especially interesting to me.

I did not begin studying French recently from a completely blank slate. I spent years with the language when I was younger, then allowed it to sit largely unused for almost twenty years.

What remains after that kind of dormancy?

What disappears?

What comes back quickly?

Is relearning fundamentally different from learning the same material for the first time?

And if proficiency can decay, can that decay be meaningfully accounted for when trying to measure language ability over a period of years?

I do not have definitive answers to those questions yet.

I would rather document what happens than pretend that I do.

Fluency Is Not the Finish Line

There is also a larger reason I returned to languages.

I have a B.A. in English, and literature has always been one of my major interests. As I began taking language learning seriously again, I became increasingly interested in something obvious that I had rarely stopped to consider.

Every language is a library.

Learning Italian does not simply give me another way to order dinner or ask for directions. It provides access to Italian literature, history, journalism, film, humor, correspondence, and ideas in the language in which they were created.

French opens another library.

Spanish opens another.

Every language learned makes the accessible world a little larger.

Translation is an extraordinary achievement. Much of world literature would be inaccessible to most of us without translators, and reading a great translation is infinitely better than never encountering the work at all.

But translation is still an interpretation of an original.

I want to know what happens when the intermediary disappears.

I want to be able to take a book from a shelf that I once could not read, open it, and simply begin.

That is ultimately what I am working toward.

Fluency isn't actually the finish line. The books are.

Making Predictions Before I Know the Answer

One of the easiest ways to convince yourself that you understood something all along is to wait until after it happens.

I do not want to do that.

The Flywheel Experiment will make predictions.

I have built preliminary relationships between accumulated XP and the CEFR levels I expect that amount of work to produce. As I progress, I can compare those predictions with assessments of my actual abilities.

The numbers will almost certainly need to change.

That is not a flaw in the experiment. It is the point of the experiment.

If my predictions are too optimistic, I want a record showing that they were too optimistic.

If they are too conservative, I want to know that too.

If two people with identical XP could have dramatically different proficiency because one spent most of that time reading and another spent it speaking, then the system needs to account for that.

If a language left unused begins to deteriorate, the model should eventually reflect that as well.

Over time, I hope to replace assumptions with observations.

I will explain exactly how the XP system works, how I arrived at my initial CEFR estimates, and how I intend to test them in a separate post.

For this first entry, it is enough to put the hypothesis on the record:

I think the long process of learning a language can be measured well enough to make useful predictions about where sustained effort will take you.

Now I get to find out whether I am right.

Starting the Wheel

I am not beginning The Flywheel Linguist as someone who has already completed the journey.

That is important.

I am approaching this as a serious adult learner who wants to see how far deliberate, sustained effort can actually go.

I will document what I study, what I read, what works, what fails, what I change, and what happens when my theories encounter reality.

Some of this will be quantitative. Some of it cannot possibly be.

There will be books, language-learning experiments, progress reports, methods, failures, changes of direction, and probably more spreadsheets than any reasonable person needs to learn a language.

But underneath all of it is a much simpler objective.

Keep taking strokes.

Keep adding momentum.

Keep making the inaccessible a little more accessible.

And eventually, walk into another library.

The experiment starts here.

Master languages. Unlock literature.