Snarky Marketing vs. Real Data: Duolingo Really Works!

It’s become almost trendy to knock Duolingo in the last couple of years. In fact, positioning a new language learning product as the “anti-Duolingo” has become a fairly standard marketing tactic.

And while there are always reasons to debate a language learning method – no one approach is perfect in every way for every person – it’s good to know the difference between a social media pile-on and legitimate criticism.

A new paper has provided some more objective balance on that.

Published in Language Learning & Technology, a study by Smith, Jiang, and Peters (2024) took a rigorous look at how effective the app actually is for independent learners. Instead of just relying on anecdotal feelings, they measured outcomes across a comprehensive range of linguistic abilities. We’re talking both receptive skills (reading and listening) and productive skills (writing and speaking), alongside vocabulary, grammar, and pronunciation.

They tracked 48 independent learners tackling the Spanish course over a three-month period. The findings? After putting in an average of about 27 hours of study, participants showed significant improvement across all ability measures. A small sample, for sure, but 27 hours is remarkably low for meaningful gains – less than 15 minutes a day. For an app coming under flak for meaningless gamification, that is an excellent piece of defence.

When gamification earns its keep

Working in educational software development myself, I know how tricky it is to balance engaging user interfaces with genuine pedagogical value. It’s one thing to build a platform that people want to log into every day, but it’s another entirely to prove that it’s actually teaching them something substantive. This paper validates that the gamified approach, when engaged with consistently, really does translate into measurable linguistic gains. The researchers specifically noted that session completion, accuracy rate, and an overall positive user experience were directly linked to this observed growth.

This challenges one of the most common criticisms you hear: that apps like this only teach you how to pass their specific tests, or only build passive recognition. The fact that productive abilities like speaking and writing improved significantly is a huge win, as these are notoriously difficult to scaffold outside of a classroom or immersion environment.

Of course, this isn’t to say any app is a magic bullet. Twenty-seven hours of app time won’t instantly make you fluent enough to debate current affairs over a coffee. I’m a Duolingo user, and get frustration and satisfaction from it in, sometimes, equal amounts. It never hurts to spread your screen time over other language learning apps, either.

But as a tool for building a solid foundation and keeping learners consistently engaged with the target language, this fresh evidence is robust. It turns out that beneath the relentless notifications and leaderboard leagues, there is a some sound pedagogy at work.

The next time a social media ad pops up claiming they’ve “fixed everything the green owl gets wrong”, it might be worth taking it with a pinch of salt. Any tool that gets people to consistently show up, put the hours in, and tangibly improve is something worth celebrating!

The Perfection Trap: Why ‘Sounding Native’ Isn’t the Ultimate Accent Goal

A paper appeared this month in Language Teaching that reminds us all – in the classroom and learning at home – that accent isn’t everything.

Is English pronunciation teaching in China ready for the Intelligibility Principle? (Xue & Wu, 2026) looks specifically at English language teaching in China, and the pressure that aspirations of nativeness can put on students. For a polyglot community focused on becoming the languages we learn, this seems very familiar. The pursuit of a flawless, native-like accent can often feel like the ultimate prize – anything less is an imperfect near miss. Whether you are tackling the phonology of German, Spanish, or diving into a new challenge like Bulgarian, it is easy to obsess over perfectly mirroring native audio.

However, the authors gently remind us that this pursuit can be entirely unnecessary.

Historically, language instruction has been dominated by this Nativeness Principle, which dictates that learners should attempt to sound as close to native speakers as possible. It’s a hangover from a quite Victorian perfectionism in learning – learn by rote, never forget.  Yet, absolute native-speaker norms are an unrealistic objective for many learners – especially older ones (as much as it stings to hear that pushing 50!).

Instead, a more practical framework is the Intelligibility Principle, which argues that the primary pedagogical goal should simply be mutual understanding.

Does This Matter for the Polyglot Community?

Absolutely. The pressure to achieve a flawless accent can actively hinder our progress. An overemphasis on native-speaker models can add to language anxiety and can decrease a learner’s willingness to communicate. Not surprisingly, many of us have an aversion to feeling a fool – it’s great to push back against that as a language learner, but it’s not always easy.

The reassuring truth is that speech can carry a heavy foreign accent and still remain highly intelligible to listeners, and therefore functional. A lot has been made of Zelenskyy’s excellent communicative skills despite a strong accent, for example. Furthermore, because English and many other languages act as global lingua francas, learners frequently communicate with other non-native speakers. Therefore, a native accent is neither necessary nor particularly beneficial for mutual understanding.

The goal isn’t to erase our linguistic identity, but to ensure our message is clearly received.

Practical Steps: What Should Independent Learners Focus On?

The paper is quite rich on practical suggestions in this regard. When you are building your corpus of vocabulary or grinding through daily flashcards, it helps to know exactly which phonetic features actually impact comprehension. Instead of trying to master every subtle nuance, prioritise the elements that carry the highest functional load:

  • Vowel accuracy: Both the length (quantity) of a vowel and the quality of the vowel play significant roles in whether a word is understood.
  • Consonant clarity: Avoiding the simplification of diphthongs and preventing the conflation of voiced and voiceless plosives are critical steps for maintaining clarity.
  • The bigger picture: Suprasegmental features, which include stress, rhythm, and intonation, have a remarkably strong influence on how a listener perceives an individual’s oral proficiency.

Tools and Tactics for the Daily Grind

As independent learners and language hobbyists, we have to direct our own study routines. Here is how we can implement an intelligibility-first approach:

  • Embrace accommodation strategies: In real-world conversations, mutual understanding is often achieved through spontaneous adjustments like simplification and paraphrasing.
  • Leverage technology: Computer-assisted pronunciation training and artificial intelligence-powered tools offer instant, data-driven feedback and personalised suggestions for self-monitoring.
  • Recognise algorithmic bias: If you are using—or even developing—language learning apps, it is important to remember that commercial automatic speech recognition systems frequently exhibit bias, showing significantly higher error rates for non-native accents.

Ultimately, dropping the demand for native-like perfection takes a massive weight off the language learner’s shoulders. By focusing on practical intelligibility, clear suprasegmentals, and flexible communication strategies, we can speak our target languages with far more confidence and far less anxiety.

Have you evolved from learner perfectionism to more intelligibility-based goals yourself? Was it hard to let go, or did you feel a sense of relief at letting go? Let us know in the comments!

A mic stands on a stage, ready for #EdFringe

#EdFringe for Language Learners, 2026 Edition

#EdFringe is here again, and true to form, there’s something for everyone. Language learners and European culture vultures are no exception, of course, with some proper treats in this year’s rich programme.

As per last year (and the year before, and the year before that… I’ve been keeping tabs for a while now), there’s a good balance between shows in the target language as well as shows about target language counties. Here’s my watch list – but be sure to have a browse too, and let me know if I’ve missed any must-sees!

French

Music

Le Vent du Nord – Québecois progressive folk from a well-regarded five-piece.

Mary, Queen of Scots – Queen of 3 Kingdoms (Marianne Beate Kielland & Ben-San Lau) – French arias mingle with Scots and English in a programme celebrating the life of Scotland’s famous queen.

Afternoon Arias (Brian Bannatyne-Scott, Beth Taylor and friends) – French classical highlights from Berlioz to Debussy.

La Chatte Chanteuse (Kat Brooks) – It honestly wouldn’t be #EdFringe without a bit of Piaf. There’s that and more in this hour of chanson from Kat Brooks.

Theatre

Madame La Mort (Full Moon Theatre / Labyrinth Productions) – a radical reimagining of Rachilde’s French Symbolist play (in translation)

Comedy

Tori Morancay – Le French C’est Freak – An anglophone set that nonetheless plunges straight into the francophone world!

German

Music

Handel – Nine German Arias (Angela Hicks)  – Soprano Angela Hicks and ensemble present some of Handel’s few German-language works.

JS Bach (Aidan Jones) – Pianist Aidan Jones plays and presents, taking the audience on a winsome tour through the life of the great composer.

Scottish Lieder (Brian Bannatyne-Scott) – A lovely crossover presenting music of Schubert, Schumann, Loewe and Strauss, inspired by Scottish poets.

Comedy

Michael Brunström: William Tell vs the Algorithm – Swiss surrealist comedy (in English) that takes aim squarely at the Swiss background of this award-nominated return performer.

Jürgen Strack: Achtung! The Only Sauerkraut in Town – Riffing on his Germanness (with shades of Henning Wehn here), Strack has won fans for his sheer originality.

Spanish

Music & Dance

Sobremesa – Where Words and Music Meet at the Table (Nus Duo) – Billed as an interactive musical experience where audience and artists shape something wonderful from Spanish and Latin American texts.

Sounds of St Cecilia’s III: Spanish Flavours: Dance, Fire and Elegance (Cokus Duo) – A harpsichord-led exploration of the music of 18th-century Europe, with a nod to both France and Spain.

Flamenkids (TuFlamenco) – Flamenco is as permanent a fixture on Spanish #EdFringe as Piaf for the French. This family-friendly show introduces the rhythms of Spain in an hour-long show that sounds wonderfully interactive.

Alegria Flamenca (Alba Flamenca) – Appetite freshly whetted, if you’re now hooked on the flamenco check out this vibrant show, with a nearby bar on hand for tapas and drinks!

Theatre

Bull / Fight – Fresh Edinburgh ensemble Mythography presents this odyssey through Lorca’s Spain (in English).

Comedy

Escocia con Ñ (Jotace Loaiza) – Scottish life retold through Spanish eyes, in Spanish! An excellent (and rare) opportunity to attend a full Spanish-language set during the festival.

Mi Casa Es Su Casa (El Purnell) – Billed as a true ‘duo-lingo’ act (hope that’s been run past the owl!), laughs are promised for hispanists and non-hispanists alike.

And the rest…

With hundreds and hundreds of shows, I can’t possibly do the whole programme justice in one short blog post. While I’ve focused on French, German and Spanish, there’s plenty else there too, from the very tempting Four Courses of Italian Song (Anna Vanosi) and the intriguing I Can Make You Italian in 55 Minutes (Stefania Licari), to the Scottish and Norwegian sacred music presentation The Maid of Norway (Nordic Voices Norway) and online viral comedy hitster Thor Stenhaug. I’ll certainly be trying to tick some of those off my list this August, too.

Is there something you’ve bookmarked to see without fail? What other languages are represented in the listings? Let us know in the comments!

Forever the Optimist : Hugo’s In Three Months Language Courses

I loved Hugo’s In Three Months books as a language-obsessed teenager. Slim volumes, tiny, dense chapters… And of course, that promise that you’d reach some level of fluency in a quarter of a year.

I’ve collected a fair few of them over the years. My first was Italian in Three Months, of Hugo’s late-80s dark blue incarnation. I bought it ahead of a school wind band trip to Venice, and probably made my way through about a third of it before the trip was over, I packed my tuba away, and faddishly drifted onto the next project.

They’ve had a much longer history than that. Hugo’s Language Learning Books pops up in 1950s UK, and quickly starts churning out titles for speedy learners. Like other language series in the mid-20th Century, they expanded their Simplified System rapidly across languages, and became bookshop staples.

Expansion, incorporation and multiple reincarnations

By the end of the 80s, they’d already moved well beyond French, German and Spanish. The second-hand trail on eBay shows that spread, with the publication of courses from Arabic to Scottish Gaelic (still a solid reference for Gaelic grammar if you can get a copy). Curiously, the first foray into Japanese dropped the ‘in three months’, instead going for Japanese Simplified – the confidence wobble didn’t last long, as Japanese in Three Months is the title that made it into the 90s!

The In Three Months books have never really gone away. The series was picked up by Dorling Kindersley (DK), gaining a splash of the prototypical DK colour and gloss. While the range of languages is a little shrunken now, they’re still going strong, now as part of Penguin’s catalogue. Gone are the boxes of cassettes, replaced by online audio. But the familiar format remains: tight, reference-style chapters giving that “all the essentials for very busy people” vibe.

It’s in that spirit I picked up one of the new editions recently – appropriately, Italian in Three Months again, this time in its fancy new green sleeve. It’s a refresher ahead of a trip to Milan for a conference, and the perfect choice for that – not too heavy (for the suitcase or the reading).

But coming full circle like this takes me right back, and I can sense the excitement I felt in that titular promise all those years ago. Somehow, that promise still works. Long live the In Three Months series – may they continue to lure language nerds to their next obsession!

Gramophone Language Courses: The Original Multimedia Learning

If you’ve ever wondered about the origins of the multimedia language course, then some newly published archive material might surprise and intrigue you. The British Newspaper Archive recently added the early 20th-century title Sound Wave magazine to its growing catalogue. This record review title served phonographic fans from 1907 to 1933, and it’s surprisingly full of language learning history.

In those days, of course, it was the gramophone that reigned supreme. Recordings on the new flat disc format had been around since the late 1880s, but by the first decade of the 20th century, gramophones had become affordable enough for middle-class households. Sound Wave dates from that early tech spread, the publishers no doubt spotting a gap in the market for listening recommendations.

Only it wasn’t just music. What we’d now recognise as audiobooks was already in circulation – outfits like The Talking-Book Corporation were pumping out gramophone literature for adults and kids. There were elocution resources for improving one’s spoken English, too. One particularly enticing release was this special set of discs with the voice of Bernard Shaw himself (life imitating art – his own art!).

And language learning was there from the start too, in the form of regular ads from the Linguaphone Institute.

A 1907 advert from the Linguaphone Institute in the magazine Sound Wave

A 1907 advert from the Linguaphone Institute in the magazine Sound Wave

Linguaphone – a brand built on gramophone

Linguaphone is a real heritage brand for language learners, and pops up all over the newspaper archives. It started up in 1901, and is still going today – you may have come across their language training centres. Competition may have widened since then, but for over half a century they were the first word in audio course materials.

Testimonials in this 1927 edition vouch for their success. One C.B. of London reports that the Spanish course made travels “much easier and cheaper than they would otherwise have been”. A reviewer in 1929 praised the “French as it is really spoken” in a dialogue set in a hairdresser’s, on record no. 21 of that set. Yes, record 21 – these sets could run into dozens of discs, and usually shipped in a hefty, solid case.

No wonder they came with an equally solid price tag. In 1907, a box would set you back £3 and three shillings, easily several hundred pounds in today’s money. You can still pick them up second-hand today, and for much less – a lovely bit of language learning history.

Proto Language Lab

Beyond the well-heeled turntable owner, the gramophonic method wasn’t just for individuals; it was used in classrooms too. In 1914, a Leicester teacher, Mr. Cunfliffe, introduced records into his lessons at the Working Men’s College, to great success. One particularly modern-seeming innovation of Mr. Cunliffe’s was the provision of “24 pairs of hearing tubes” for the students! In this way, one element of language teaching that seems so late 20th-century, so proto language lab, had its roots decades before tape reels and CDs.

The BNA‘s inclusion of Sound Wave offers some lovely insights into the history of language learning and teaching. There’s doubtless much more to find in there. Let me know in the comments if you come across any other gems!

Screenscot of Cell to Singularity, an immersive casual clicker game available on Steam.

Cell to Singularity : Casual Play for TL Immersion

Osmosis isn’t just for cells – it’s for language learners too! Soaking up target language simply by placing it in your everyday line of sight is one of the most effective strategies for fluency. From your instagram feed to cosy telly-watching, consolidation can be about throwing more of the things you love in your way.

Gaming is another entz stream that is really easy to target language-ify, since many titles have multiple language options. The Steam platform is a particular goldmine here – a huge multi-platform marketplace, with loads of free-to-play offerings. The trick is to find quite text-heavy games with dialogue and interactions, exposing you to as much content as possible in-play. There’s honestly something for everyone here, from word games to fully-fledged RPG.

This week, I chanced across a casual clicker on Steam that has been working its quiet way into the hearts of users since its inception in 2018. It’s Cell to Singularity, a game that simulates the blossoming of life on Earth, from eukaryotes, to jellyfish, to humans (and beyond). It’s the kind of game you can have running inconspicuously in the background while you work, slowly developing and growing like a bonsai that needs occasional tending. Very Zen.

Screenscot of Cell to Singularity, an immersive casual clicker game available on Steam.

As you can see from the screenshot, it’s also a great way to revise the building blocks of life. That’s the root educational application the game has been feted for, covering evolutionary biology in a fun, laddered way. Switching my interface to German gives me a ton of fun natural world vocab.

Beyond word level

But the game is also full of conversational exchanges you have with the ‘supercomputer’ running your life simulation, as well as Wikipedia-style descriptions of all your finds. In short, it supports word, sentence and text-level language skills in a rich, engaging environment. What more could you ask for?

Screenshot from Cell to Singularity showing dinosaurs

The range of languages available right now is already impressive. Not only the ‘mainstream’ school ones, but also Korean, Japanese, Polish and Portuguese, amongst others.

Screenscot of the language options in Cell to Singularity, an immersive casual clicker game available on Steam.

Cell to Singularity currently has an 89% positive rating from thousands of Steam users. I wonder how many of them are playing to improve their target language? Hopefully I’ve enticed a few more of you to do just that!

Speaking dialect in the mountains

AI and Dialect: Can You Nudge a Language Model Off the Standard Path?

One of my quiet worries about the AI revolution in language is standardisation. When a model trained on billions of documents produces text, it tends to sand everything down to some kind of weighted average – grammatically impeccable, register-appropriate, but utterly devoid of local colour. And where’s the fun in that? You know my thoughts on dialect and language learning.

But here’s the thing: despite the odds, the standard grey slop isn’t a mathematical inevitability. Large Language Models are trained on billions of words. And that material will include plenty of non-standard language, too, simply by dint of the extraordinarily broad sweep of the training brush. Dialect literature, forum posts, subtitles, transcripts – there’s room for non-standard, more colloquial language in every context of the model’s background.

Which raises an interesting question: can we nudge AI output away from the standard, and towards something more authentically regional?

I’ve been playing around with exactly this, and the short answer is: yes – and it’s surprisingly decent at it.

Stepping off the well-trodden path

I’ve said from the start that AI is built for language learning content creation. There aren’t many fields where content accuracy doesn’t matter too much – as long as the language is correct and authentic-sounding. One of my go-to moves is pushing the LLM for a more colloquial register. Standard written language is fine for reading comprehension, but if you’re teaching people how to actually speak, you want contractions, ellipsis, informal vocabulary – the stuff that sounds like a real person. AI handles this pretty well once you prompt for it explicitly.

Create a model German dialogue on the topic “Reality TV”. The language should be contemporary, colloquial and natural-sounding German.

Claude did a pretty good job of this:

  • Lena: Ey, hast du letzte Woche „Love Island” geschaut? Ich konnte gar nicht aufhören – ich war bis halb zwei wach.
  • Kai: Natürlich hab ich das geschaut – wär ich ja bescheuert, das zu verpassen. Diese Tiara und Marco… ich meine, komm schon. Die spielen da nur rum.
  • Lena: Genau! Marco tut total so, als wäre er an ihr interessiert, aber das ist doch alles für die Kamera. Ich find das so zum Fremdschämen.
  • Kai: Ja, aber genau deswegen schaust du’s doch! Gib’s zu. Du liebst den Drama.
  • Lena: Okay, okay, ich geb’s zu. Aber ich find’s trotzdem fragwürdig, wie die Leute da so bloßgestellt werden. Manche wirken echt überfordert.
  • Kai: Absolut. Ich glaub, vielen ist nicht bewusst, wie sehr die Produktion das alles lenkt. Die schneiden das so zusammen, dass du denkst, die Person ist voll die Zicke – aber vielleicht war das gar nicht so gemeint.
  • Lena: Stimmt schon. Und dann wird die Person im Netz fertiggemacht. Ich hab mal gelesen, dass einige danach richtig psychische Probleme hatten.
  • Kai: Das ist so traurig. Und trotzdem melden sich jedes Jahr tausende Leute an. Ich versteh’s irgendwie – du willst berühmt werden, schnelles Geld, Follower… Der Lockruf ist halt stark.
  • Lena: Stimmt. Und seien wir ehrlich – wir beide gucken’s ja auch. Also sind wir Teil des Problems, oder?
  • Kai: Pff… ich würd sagen: wir sind Teil des Unterhaltsystems. Klingt besser. Jetzt mach die nächste Folge an.

Some great non-standard, everyday features in there. Colloquial contractions like -e dropping on verbs, decent use of particles like halt – not a bad job at all. Claude even threw in a bunch of usage notes without asking, too.

But colloquial is still, broadly, floating just beneath a standard umbrella. What if we go further?

Adding Local Flavour

When I’m probing the edges of what AI can do, I like to start with varieties I can actually judge. Austria is my German-speaking region of choice in this regard, being my year abroad destination of old – I still have a huge soft spot for Kärnterisch. So here’s the tweak:

Create a model German dialogue on the topic “Reality TV” between two young Austrian speakers from Kärnten. The language should be contemporary, colloquial and natural-sounding Austrian German.

Here’s what came back:

  • Magdi: Oida, hast du gestern „Love Island” g’schaut? I kann nimma! Die Lea hat den Stefan komplett ausg’spielt, echt a Wahnsinn.
  • Tobi: Na freilich, i schau des jo seit Wochn! Aber geh, die Lea, die tuat doch nur so. Die weiß ganz genau, wos sie macht — des is alles g’spield.
  • Magdi: Ja eh, ober trotzdem — der Stefan is so a Trottel! Wie kann man so blöd sein, gell? Der glaubt ihr jo alles.
  • Tobi: Najo, i mein, de san jo auch so a bissi animiert zum Streiten, oder? Die Produzenten woin das jo so. Ohne Zoff läuft des ned.
  • Magdi: Des stimmt scho, ober i find des trotzdem scheißn. Die tean so, ois ob’s echt wär, und dann geht’s nur ums Follower kriegen. Danoch san alle auf Instagram mit Werbung zuagekleistert.
  • Tobi: Haha, stimmt! Drei Wochen nach der Sendung verkaufen’s alle irgendwelche Protein-Shakes. Gott sei Dank ned mia. [lacht]
  • Magdi: Ober weißt wos? I schau’s trotzdem weiter. I bin schon voll drin, i muss wissen, ob der Kevin die Kathi kriegt oder ned.
  • Tobi: Na sag i dir: die kriegen sich ned. Die Kathi is viel zu schlau für ihn. Die geht da raus, macht a paar Storys, und dann war’s des.
  • Magdi: Oida, du bist so a Zyniker! Vielleicht is es jo echt?
  • Tobi: Magdi. Des is Reality TV. Des is so echt wie mei Oma beim Wrestling. [grinst]
  • Magdi: Haha, okay okay. Ober schauen tust’s trotzdem, gell?
  • Tobi: … Ja. Natürli. Um neun bin i dabei. Sag’s ned weiter.

If you’ve spent any time in Austria, this is probably just as joyful to you as it is to me. It’s added in so much southern colour, from the tag particle ‘gell’ to the reduced i from ich and past participles in g’. There’s even some fun wordplay in there (as real as my gran at wrestling!), so it’s not only dialect, but character. It’s laid it on thick!

There’s an element of pastiche to this, for sure. Nobody will claim this is indistinguishable from an authentic speaker of Klagenfurt. But it’s genuinely not bad, especially if you’re after a solid introduction to local speech features.

And then there’s Trøndersk…

That’s all very well for ‘mainstream’ foreign languages like German. To push the envelope a little more, it’s worth playing around with some lesser-learnt languages – Norwegian is always a good choice.

For my Norwegian test, I went with Trøndersk, the dialect cluster associated with the Trondheim region and much of central Norway. It’s one of the most immediately recognisable Norwegian dialects, with distinctive pitch accent patterns, characteristic -a verb endings, and a few phonological quirks that mark it out clearly from the Oslo Bokmål you’ll meet in textbooks.

How good a job would Claude do of this challenge?

  • Marte: Oi, har du sett at det e ny sesong av Farmen nå? Æ begynte å se på den i går kveld og klarte itj å slutt.
  • Jonas: Åja? Æ har itte sett mye reality-greier på ei stund, men Farmen e jo noe anna. Det e liksom ordentlig underholdning, da.
  • Marte: Ja, det e noko med den serien! Og den nye dama fra Steinkjer – ho e jo helt vill. Ho klaga på absolutt alt og alle og gjor ingenting sjæl.
  • Jonas: Haha, klassisk. Det e alltid nåkka en sånn type på realitygreier. Produsenten e nødt til å ha med dem, skjønner du? Ellers e det jo kjedelig.
  • Marte: Jaja, men det e jo litt ekkelt å innrømme at man liker det, liksom. Æ satt der og ropte på skjermen som et idiot.
  • Jonas: Haha! Ka sa du da?
  • Marte: Noe sånt som «koffor gjør’u det, din tulling!» Men mora mi satt rett ved sia og hun så bare på mæ som om æ var gal.
  • Jonas: Uff da, haha. Ka syns ho om programmet da?
  • Marte: Ho e heilt avhengig sjæl! Det e liksom heime hos oss at hele familien samles rundt skjermen på tirsdager. Det har vorre sånn siden æ va lita.
  • Jonas: Okei men ka med all den andre reality-søpla da? Eks på stranda og sånt? Det e jo bare tull, syns æ.
  • Marte: Ja nei, det gidder itj æ. Det e jo bare folk som vil bli kjendis og som itte har nåkka å by på. Farmen e annleis fordi det faktisk krev noe.
  • Jonas: Sant nok. Kameraten min i Ålesund e helt hekta på Paradise Hotel derimot. Han ser på det heile tida.
  • Marte: Seriøst? Koss klare han det? Æ holdt ut to episoder og ga opp. Alle folkan der oppe e jo helt… æ veit itte. Tomme, liksom.
  • Jonas: Haha, han sier det e «sosialt eksperiment». Æ tror bare han har lyst på drama heime fra sofaen sjæl.
  • Marte: Hahaha! Ja det e nok det. Uansett — skal du se Farmen med mæ på fredag? Vi kan bestille pizza og le av alle.
  • Jonas: Ja, det høres bra ut! Men du bestemme pizza-topping. Sist gang valde du masse squash og det va itte greit.
  • Marte: Oi, squash e jo godt da! Men okei, du får velge. Bare itte ananas.
  • Jonas: Deal. Vi ses fredag!

As far as I can tell – and I’m not so well placed here, I must admit – it’s done a solid job. The k- question words (rather than hv-), the reduced æ for jeg… Even the cultural references, selecting two actual Norwegian shows for the exchange. I’d trust this as a bit of pre-Trondheim trip prep!

That said, if you speak a Trøndersk dialect, what’s the verdict? Does this read as a reasonable attempt, or is it giving you second-hand embarrassment? I’d genuinely love to know. (Drop a comment below, or find me wherever I am on the internet this week.)

AI and Dialect for Language Learning

AI platforms clearly have some representation of dialect. Awareness is the wrong word for it – anthropomorphising LLMs and all that – but you know what I mean. The patterns are in the training data; the question is whether a prompt can reliably surface them, or whether the model’s default pull towards the standard is too strong.

The two languages I explored here seem to work well. The pull to parody is always something to watch, particularly when written material in a dialect tends to be humorous, rather than in-community. And rarer or more exclusively spoken varieties will no doubt be much more troublesome. But the more linguistic scaffolding you give the model – the more you treat prompting as a kind of sociolinguistic briefing – the better the output is likely to get.

That standard grey slop, it seems, has more texture than it first appears – you just have to know where to poke.

Correcting Mistakes: Should Teachers Do It Immediately or Later?

A small moment in a lesson recently got me thinking about the age-old question of correcting mistakes “live” in the classroom.

A student was responding to a question, and made a tiny vocab error – nothing dramatic, just the kind of slip that happens when you’re concentrating on your idea rather than the grammar. In other words, a big win for communicative flow, but at the expense of pronouncing a word completely correctly.

My instinct as a teacher was to jump in immediately and correct it.

But I stopped myself.

The student finished the thought. The rest of the class responded. And the conversation continued. I made a mental note of the slip, and later looped back briefly to the concept to model it correctly without calling anyone’s pronunciation out.

That said, I couldn’t help wondering afterwards: had that actually been the better choice?

Immediate versus Delayed Feedback

Language teachers have debated this question for decades. Should we correct errors immediately, as they happen? Or is it sometimes better to wait, allowing learners to finish speaking before stepping in?

Recent research has begun revisiting this question, and the answer appears to be – perhaps frustratingly – “it depends.” But the details are fascinating, and they tell us something important about how language learning actually works.

In second language acquisition research, this question is known as the timing of corrective feedback. It refers to when a teacher responds to a learner’s mistake – either instantly or after some delay.

Immediate feedback happens during the interaction itself:

Learner: Yesterday I go to the shop.

Teacher: Went — yesterday I went to the shop.

Delayed feedback happens after the task or conversation is complete:

“Earlier you said I go yesterday — remember we need went for past tense.”

It can be modelled without calling out the error, too:

Teacher: You said earlier that you went to the shop? Which shop?

The study by Li, Ou and Lee confirms earlier findings that both approaches have their place. Perhaps intuitively, results show that both immediate and delayed feedback improved motivation and learning outcomes compared to giving no feedback at all.

But where do their differences lie?

What each type seems to do best

Studies like this repeatedly suggest the two approaches may support different cognitive processes.

Immediate feedback

  • helps learners notice errors right away

  • supports rapid correction during conversation

  • often leads to faster improvements in accuracy

Delayed feedback

  • encourages reflection on language forms

  • allows learners to focus on meaning first

  • may trigger deeper discussion about grammar

Some studies even find that delayed feedback leads to more discussion about linguistic forms, while immediate feedback can produce greater improvements in accuracy over time.

In other words: one supports fluency and awareness, the other precision and correction. It’s all back to that communicative flow that I didn’t want to interrupt earlier.

Teacher Fading

This debate connects neatly to another concept that has recently attracted attention in language pedagogy: teacher fading.

Fading is the idea that teachers gradually withdraw support as learners become more capable, transferring responsibility for learning to the students themselves, a technique that is gaining traction as studies like this one explore how it affects the classroom dynamic.

In practice, teacher fading means:

  • correcting less frequently

  • letting learners negotiate meaning themselves

  • allowing conversations to run without interruption

In other words, teachers deliberately step back.

Seen through this lens, delayed feedback is not just a technique – it is part of a broader teaching philosophy. Instead of jumping in every time an error occurs, the teacher allows communication to unfold, intervening later only when it is useful.

The Balancing Act

The effect, I think, is a better balance between teacher guidance and student practice – taking down the guard rails just enough for them to try out their ‘talking hats’ without fear of being slapped down.

It’s this balancing act that’s always existed in language (and other) classrooms. If correcting language mistakes happens too frequently, it can impede communication and raise anxiety. If it doesn’t happen enough, learners may simply repeat the same errors.

Modern pedagogy increasingly suggests that the most effective classrooms combine:

  • moments of focused correction

  • periods of uninterrupted communication

  • gradual reduction of teacher intervention

In other words: a mixture of immediate feedback, delayed feedback, and teacher fading.

Putting it into practice

Mistakes do not need to be corrected instantly for us to learn from them. In fact, sometimes the best thing a teacher can do is let the conversation continue. Learning is not just about eliminating errors as quickly as possible. It is about building the ability to communicate – and that sometimes requires a little space for imperfect language to unfold.

And occasionally, the most effective teaching move is surprisingly simple. In the (doctored) words of Ronan Keating:

You say it best when you say nothing at all (at least for a little while!).

Why Music Might Help You Develop a Better Accent

Language and music have long been known to share deep cognitive roots. Both rely on rhythm, timing, pitch perception, and finely tuned listening. Research from the University of Cambridge has highlighted how speech rhythm and musical rhythm rely on closely related auditory timing mechanisms, suggesting that the systems that help us follow a beat may also help us decode the rhythm of speech.

Because of this overlap, researchers have long suspected that musical skills might transfer to – and even be implicated in – language learning. A very recent study published in the journal Language Teaching explored this connection in more detail, and came up with a particularly intriguing finding: certain types of music training may actually improve a learner’s ability to imitate the sounds of unfamiliar languages.

Now if you’re like many in the polyglot community – me included – this will be, ahem, quite literally music to your ears. Overseas pop is a popular way to practise a language and connect with your target language culture. This new study confirms our suspicions on the utility of music in language learning. But rather than focusing on listening skills, it explores the potential benefits for second-language phonology.

The Study

The researchers started off with a simple question. Might different approaches to music training influence learners’ ability to reproduce unfamiliar speech sounds? To this end, participants completed a set of speech imitation tasks involving languages they did not know.

There were two ‘modes’ of musicality in the experiment. Some participants had a traditional, performance-based music background. Others, though, had experience with embodied music training in particular – learning rhythm and musical patterns through physical movement, body percussion, and coordinated actions.

The results were clear; learners with embodied music training performed significantly better at imitating unfamiliar speech sounds than those without it. The researchers concluded that embodied musical activities can have measurable benefits for pronunciation skills and speech imitation. That’s quite striking – whole body activity as a route to strengthening speaking skills.

Why Might Music Help?

The connection between music and language perception has been explored for years. If rhythm and melody enlist the same mental structures as linguistic information, it makes sense that musical activities might activate and strengthen the perceptual systems used for speech. These overlapping processing pathways have been the focus of ongoing study, and researchers are still in the process of mapping out their myriad networks.

This musical crossover is not a new finding, either. Work at the University of Edinburgh has examined how singing and musical presentation can support memory and learning in a foreign language. And other research has found that people with stronger musical rhythm perception tend to process speech rhythm more consistently, as shown in research published in the Journal of Laboratory Phonology.

Several mechanisms, then, might explain the effect observed in the study:

  • Improved rhythm perception, which supports speech prosody
  • Greater sensitivity to pitch variation, important for both tone and intonation
  • Sensorimotor memory linking movement and sound
  • Enhanced auditory attention during listening tasks

Speech, after all, is not just a sequence of sounds. It is a rhythmic and melodic system – something much closer to music than we often realise.

What This Means for Language Learners

So the takeaway from all this for language learners? Chiefly, musical activities that involve rhythm and movement may be particularly good at supporting pronunciation development. The participants were introduced to a particular type of music training, but this needn’t necessarily mean that formal music training is required. The authors’ discussion reinforces the value of quite general practices that play on the musical qualities of language.

Examples you might work into your own learning could include:

  • shadowing spoken audio while gesticulating or signing the meaning of the words
  • chanting or rhythmic repetition
  • singing – vociferously – in the target language
  • dancing whilst singing along to your favourite foreign pop!

It all reminds me of the ‘body parts’ games I’d use as a language teacher – heads, shoulders, knees and toes and such like. Those words would stick with students. The same kind of embodied approach has been used for other vocabulary items, a memory technique known as pegging, suggesting that the kinetic aspect is just as important as the musical one.

And of course, any excuse to make language learning more fun is a good one.

Project Gutenberg Goodies : Free Reading for Language Learners

I’ve been spending a lot of time lately in the Project Gutenberg annals – wearing my research, rather than my language learning hat. Victorian novelists like Eliot and Hardy are a goldmine of dialect writing, which is what sent me back to this quietly heroic archive of public-domain texts.

And in doing so, I was reminded of something easy to forget: Gutenberg isn’t just a treasure trove for English literature. It holds an enormous amount of French, German, Spanish and Italian writing too, amongst other languages – much of it far more modern, linguistically speaking, than people assume.

When people hear “public domain”, they often imagine pre-modern archives full of dusty stuff only classicists would be interested in. But in most European contexts, public domain simply means “published roughly before the 1920s”.

That’s not ancient. That’s late 19th or early 20th century – and linguistically, that’s reassuringly modern.

Project Gutenberg is therefore an extraordinary (and free) resource for language learners who want authentic reading material that still feels recognisably contemporary.

Why the Language Isn’t “Too Old”

In French, German, Spanish and Italian, the core grammar and spelling conventions were largely standardised by the late 19th century. That means:

  • The verb systems are settled into the patterns we see today.
  • The spelling is (almost entirely) modern.
  • The syntax may feel more formal, but not archaic.
  • Most high-frequency vocabulary is still current.

You may encounter slightly more formal phrasing or the occasional dated word. But you are not learning an obsolete language – just a form of it a couple of generations removed. Think of it as reading early 20th-century English: recognisable, rich, and still practically useful as a linguistic template.

In fact, reading slightly older prose often strengthens your command of formal written style — which is still relevant in academic, journalistic and literary contexts.

So where to start, in an archive of thousands of resources? Here are a few highly readable gems, some of which you’ll almost certainly recognise as cultural touchstones. Download them to your reader of choice – I use the free send to Kindle service myself to get Gutenberg’s epub files onto my device. I’m also increasingly falling for the open source Calibre reader – not only free, but also not tied to any corporate behemoth.

🇫🇷 French

French orthography has been extremely stable for some centuries (as with English). A learner reading Verne or Leblanc is reading something visibly very close to modern written French.

🇩🇪 German

Kafka in particular feels strikingly modern. German spelling reforms have occurred since, but older orthography is easily recognisable (and Gutenberg editions are often standardised anyway).

🇪🇸 Spanish

Modern Spanish spelling was largely standardised in the 18th and 19th centuries. Early 20th-century prose feels very close to today’s standard language.

🇮🇹 Italian

Modern Italian largely crystallised in the 19th century. Many works from this period are linguistically very close to contemporary written Italian.

And more languages are available! I found Hamsun’s Norwegian classic Sult (Hunger) on there, for example.

Choosing the Right Level

The above selection cover a good range of books that should be accessible to lower intermediate learners upwards. But Project Gutenberg isn’t only for experienced readers. You can search for texts strategically:

  • Upper beginner: short stories, fairy tales, episodic narratives.
  • Intermediate: novellas, adventure fiction, children’s literature.
  • Advanced: literary realism, philosophical novels, modernist prose.

How you approach these works also makes a difference. Start with shorter chapters. Choose familiar stories. Use your Kindle’s dictionary function. Treat reading as graded exposure, not a heroic test of endurance. Little and often is often the best way to develop a foreign language reading habit.

Why Project Gutenberg Matters

There’s something quite powerful – not to mention digitally sovereign – about building fluency through public-domain literature. It costs nothing. It democratises cultural history. And it reminds us that “free” doesn’t have to mean “low quality”. In an era of subscriptions, paywalls and microtransactions, that feels quietly radical.

In fact, public domain literature doesn’t even have to mean fiction. There are plenty of non-fiction titles there, many on language itself. There’s a 19th-century Gaelic grammar, for instance, that teaches rules that are still relevant today. And if we suspend our “nearly contemporary” rule for a moment, there are historical treasures like this 16th-century French language primer, written for the English royal court. It’s surprisingly familiar to anyone who’s used traditional language learning textbooks.

Project Gutenberg isn’t a dusty archive. For language learners, it’s a modern treasure chest – hiding in plain sight.