LingQ: The Illusion of Learning

[davidalvarado]

I wanted to add to the discussion and talk about my personal experience with language learning, particularly English, which was my first foreign language.

TL;DR Immersion only actually works, but it’s an extremely lengthy and inefficient process.

When I was a kid, I liked browsing Wikipedia articles in Spanish (my native language) and I had very basic knowledge of English thanks to my school (think below A1). And I couldn’t help but notice that the Spanish article was a translation of the English one, and that the English vesion had more content! In my 8-year-old mind, I concluded that, therefore, everything in English is factual, more up to date and better than anything in Spanish, so I stopped trusting information given to me in Spanish until I could fact check it in English.
Next year, I switched schools to one that didn’t teach English, but my little fact checking habit persisted, and eventually led to me only watching Youtube videos in English as a teenager (which maybe you can surmise was a considerable portion of my day back then). I only watched videos in English, anime with English subs, video games in English, etc etc, the only Spanish in my life was my family and friends. I did almost no output, besides the occasional argument in forums or Youtube comments. That habit persists up until now, to the point that sometimes my English tends to be better than my Spanish in everything but pronunciation (which is not that bad, and I’m currently working on it).
But that’s just anecdotal, the hard proof is that I passed a C2 test (iTep, I thought it was way too easy, so maybe take it with a grain of salt) because I needed proof of second language skills to graduate from university.

Now, this whole process must have lasted about 8 years until I became truly fluent, so it’s not a particularly efficient way of learning, however, it is effective.

Now for the other side, I’m currently learning German using a mix of LingQ, Anki and the DW course, while generally following the Refold philosophy and constantly mining sentences, and I can say that progress must be at least 10 times faster, if not 100 times. By mining sentences your vocabulary goes up really, really fast and it creates a feedback loop because it’s easy to learn new words. I’ve recently switched to monolingual Anki cards, and it feels really nice to learn new words through their own German definition. And the last thing is the DW course, which constantly asks me reading comprehension questions (also monolingually) and has been a great mix of immersion and studying, because they give you grammar explanations in German. I’m currently solidly at a B1 level and I’ve been studying “seriously” (more than 1 hour daily, currently about 5 hours daily) for about one and a half years. Overall I’ve been studying German on and off for 5 years at this point. Mostly off, because if I had been studying this hard the whole 5 years I’m sure I’d be at least C1.

Also worth noting that I’m doing 0 output besides the DW course exercises.

[bbbblinq]

I aim for reading 3000 words each day, taking perhaps 30 minutes; and 1 million words a years. I notice a step change every 100,000 words; every month.
An hour or five a day would give me progress much more quickly.
I’m trying to compare methods based on the time spent. I’d also note that some methods would be pleasant to spend more time on and others not.

[AHumean]

I don’t know if I would call it a illusion as such, but what LingQ calls learned vocabulary isn’t really vocabulary that a user has actually learned. To be able to measure that, the people who design and maintain LingQ would first have to have at least a rough idea what constitutes ‘learning’. Then they would need an accurate, convenient, and usable gauge for what assessing what words a user has learned and the extent to which the user has actually mastered a given word. And I’m not sure to what extent LingQ’s management is actually interested in dealing with this problem. (For all we know, they might not consider this to be an issue of any importance). And to whatever extent it is interested, it is not clear to me to what extent they actually have a handle on the problem at all. At least not at the present time.

If we were really interested in accuracy in these matters, we might actually get it, but in addition to design and maintenance time, the users would have to be willing to pay for comprehensive assessments and evaluations. And the people who provide the services would probably want some prior assurance that it would be worth their while to do go to spend the time, effort, and money needed to get the job done.

The people who designed LingQ understand that a lot of the people who use LingQ want to be able to track their progress in acquiring their target language or languages. In other words, a lot of users - such as myself - want to be shown statistics to give at least a rough measure of how much progress we’re making in acquiring fluency in our target language or languages. One such stat they track is ‘learned words’.

Duolingo did not track ‘learned words’ for users when I first started using my free Duolingo account in 2018 and so far as I can tell, it still doesn’t.

I took my first Spanish course in high school in the late 80s and my last Spanish course was in Fall 1992, towards the beginning of my college career. I had also been using Duolingo to practice and improve my Dutch for over four years before I encountered LingQ, and I am still using it over two years later. Because of prior experience with language learning, I already knew quite well that mastery of vocabulary words usually lags a long ways behind exposure. So when I first ran into the ‘learned words’ metric in LingQ, I took it with several heaps of salt. At the same time, I’d rather have a flawed metric which can at least give me a rough idea where I stand than not have one at all. And this also allows me to give other people a rough sense of where I am in my journey along the road to native or near-native level fluency, although I keep reminding everybody of the limitations of these numbers.

The basic idea behind the ‘learned words’ metric, is that you can measure things if you substitute quantifiable things (such as hours of food deprivation) for things which are much harder to quantify (such as hunger). You might not be able to tell how hungry a given person is, but you can quantify when a given person ate, how much food that person ate, and so forth. In psychology the substitution for quantifiable metrics for things which are harder to quantify is called operational definition. The learned words metric is a prime example of it.

When you get down to it, what LingQ says is your ‘learned vocabulary’ is actually a metric LingQ uses to quantify user exposure to words. The LingQ website and app don’t actually know if you actually “know” a word or not. What LingQ does know is how much exposure you’ve had to a given word on LingQ. And it gives you points for things like looking at flash cards and selecting the correct option in multiple choice questions.

Context is a big part of reading and listening, regardless of whether you are working with a word, phrase, or sentence in your language or in some other language. The key to expanding your vocabulary in any language is to see, hear, and use words, phrases, and sentences in multiple contexts so that you can use the context to begin to determine what meanings a given word, phrase, or sentence might have. Then you use context (including background knowledge from other words, phrases, and sentences) to make note of what meanings are congruent with the context and which are not.

I do agree with you that some of the structural elements of language ought to be explicitly taught. That they ought to be broken down into more digestable pieces and then learners ought to be given multiple examples and provided lessons so that they can learn they rules/patterns and figure out how they apply. Things like conjugation charts definitely help. Duolingo isn’t terribly good at this, and LingQ is actually a good deal worse than Duolingo on this front. In my experience, grammar books are also of limited use to language learners who aren’t very advanced language learners because they tend not to explain things simply and in sufficient detail. They also tend not to provide nearly enough examples.

A series of lessons in which complex patterns and rules were broken down into relatively short, simple lessons and several examples were given would be very helpful. This isn’t difficult to do in LingQ, if you already know a language well enough, although it’s definitely going to take a while before most language learner feel comfortable trying to create such lessons on their own. For that reason, I think that LingQ ought to provide such lessons. And that would be a solid selling point for LingQ, if it did provide them.