Todd Hughes has a love-hate relationship with Danish. At 64, he can read a Danish novel with ease, but when placed among native speakers his confidence falters. His experience highlights the gap between sounding fluent and being truly proficient.
Why the distinction matters
Fluency often describes the smoothness of speech, while proficiency measures the ability to understand, respond and adapt across reading, writing, listening and speaking. Hughes, who trains language tutors at Rosetta Stone, stresses that real communication demands more than a polished delivery.
For me, it's a question of skill and a question of ability over a question of flow.
Native speakers constantly pause, stammer and change direction, making a flawless delivery an unreliable gauge of mastery. When someone claims fluency in multiple languages, Hughes asks, "What can you do?" He evaluates the four core skills, reading, writing, listening and speaking, because strength in one area does not guarantee competence in the others.
Technology's role
AI can generate practice exercises, correct writing and produce reading material at the appropriate level. Apple's Live Translation feature for AirPods, for example, promises real-time understanding of spoken conversation.
Hughes has been involved with language-learning technology since 1995, when his dissertation explored computer-assisted Spanish vocabulary acquisition. He likens AI to a "huge library" that never closes, capable of supplying beginner-level stories on demand.
I'm not there yet.
However, he remains skeptical about AI's ability to coach speaking. The nuance of tone, humour and cultural reference often escapes algorithmic translation.
Human interaction remains essential
Duolingo's chief product officer Cem Kansu agrees that real-time translation cannot replace the personal satisfaction of learning a language oneself. The company notes that many learners pursue English for career or academic advancement, goals that demand more than a single translated exchange.
Idioms illustrate the problem. The English phrase "don't pull my leg" translates literally into Spanish as "don't pull my hair". An app may swap the words, but it cannot gauge whether the remark is playful or rude, nor whether the speaker knows the listener well enough to use it.
In a live conversation, learners must interpret what another person says, choose an appropriate response, watch the reaction and adjust on the fly. Tools can suggest the next sentence, but they cannot perform the complex social dance of real communication.
When was the last time that you had a really, really, really interesting conversation with a computer?
Until AI can reliably read those subtle cues, language proficiency will continue to rely on human contact.