처음 1000단어가 일상 대화의 75%를 덮는다

How many words do you need? For comfortable conversation, roughly 2,000–3,000 word families. For reading novels without a dictionary, 8,000–9,000 (Nation 2006). But the most important number is smaller: your first 1,000 words, if they're the right ones, cover about 75% of everything said around you. Picking them by frequency instead of by textbook topic is the single highest-leverage decision a beginner can make.

How much does each layer of vocabulary cover?

Word frequency in every studied language follows Zipf's law: a tiny set of words does most of the work. Zipf gives the shape; corpus studies (Nation 2006; Schmitt 2008) give the percentages, and they repeat across English, Spanish, Mandarin, Japanese:

Vocabulary size	Coverage of everyday speech
Top 100 words	~50% (varies 45–52% by corpus)
Top 1,000 words	~75% (children's-book corpora land near 80%)
Top 2,000 words	~85%: most conversations followable with context
Top 5,000 words	~90–95% of casual TV and conversational text
~6,000 word families	Nation's strict 95% threshold for unsimplified text
~9,000 word families	The 98% needed to read novels comfortably

This table explains the beginner experience better than any motivational talk: the first 1,000 words are linear pain, and then everything accelerates, because with 75% coverage you can context-guess most of the rest.

How many words do you need to be "fluent"?

For conversational fluency, aim at 2,000–3,000 well-chosen word families plus the few hundred specific to your life: your job, your hobbies, your in-laws' favorite complaints. The 8,000–9,000 figure scares people, but it's the literary reading bar, not the conversation bar. Most learners drastically overestimate the vocabulary needed to start living in a language and drastically underestimate how much retrieval practice the words they "know" still need. (That second problem is the intermediate plateau.)

Why is frequency-first better than topic-first?

Because topics don't transfer; connectors do. Most apps and textbooks organize by theme ("food," "travel," "family"), which feels orderly and is statistically backwards. Spend a week on "fruits and vegetables" and you can discuss salad. Spend the same week on the connectors of speech (however, although, even though, in fact, basically, kind of) and you can suddenly discuss anything you already know nouns for. Textbooks front-load "colors" because they're concrete and teachable, but "would," "if," "even," and "actually" are vastly more common in real speech.

What's the right way to learn the first 1,000?

If you have 30 minutes a day, spend your first 90 days almost entirely here. Three rules:

1. Use a frequency-ordered source, not a chapter-ordered one. Free: Wiktionary's frequency lists; Mark Davies's COCA lists for English. Paid: the Routledge Frequency Dictionaries cover most major languages.
2. Learn every word inside a real sentence, never as a bare card with one translation. Words live in collocations.
3. Say each word out loud, 5–10 times, in different sentences. Speaking a word creates a measurably stronger memory trace than reading it silently: the production effect (MacLeod et al. 2010). This is also where spaced repetition alone falls short: cards show you the word; your mouth has to learn it too.

How Dia applies this

Every word you actively use in conversation with Dia is tracked in your personal word bank. Dia recycles the words you've already made yours into your next conversations, retesting them in real dialogue, where retention consolidates, and stretches you toward new words at the edge of your level. You don't study a list; the list assembles itself from how you speak. (Per-word frequency-rank coverage, "you've used 612 of the top 1,000," is on the roadmap; today the engine tracks what you use, not what you're missing.)

Frequency is free leverage. Use it.