---
title: "间隔复习 + 对话：为什么两者结合最有效"
description: "抽认卡训练识别，对话训练提取。两者结合，比单一方法记得更牢。"
canonical: https://talktodia.com/zh/blog/spaced-repetition-and-conversation
language: zh
published: 2026-05-16
updated: 2026-06-09
author: Bhada Yun (Founder, TalkToDia)
license: see https://talktodia.com/.well-known/ai-policy.txt
---

# 间隔复习 + 对话：为什么两者结合最有效

抽认卡训练识别，对话训练提取。两者结合，比单一方法记得更牢。

Spaced repetition is the most replicated finding in memory research, and it is still not enough to make you fluent. Flashcards train *recognition* — seeing a word and recalling what it means. Conversation trains *retrieval* — pulling the word out of your head in the half-second a real exchange gives you. You need both, and this post is the recipe for combining them without grinding 200 cards a day.

## What is spaced repetition, and why does it work?

Spaced repetition means reviewing material at expanding intervals (1 day, 3 days, 7 days, 21 days…) instead of cramming. It works because forgetting is fastest immediately after learning: Ebbinghaus's original 1885 self-experiment found most new material gone within a day, and Murre & Dros (2015) replicated the curve almost exactly with modern controls. Each well-timed review interrupts the decay and flattens the curve.

The effect is large. Cepeda et al.'s 2006 meta-analysis put spaced practice at roughly Cohen's d ≈ 0.6 over massed practice for long retention — in plain English, spacers remembered half again to twice as much a month later than crammers who studied the same total minutes.

## What is the optimal review schedule?

Roughly 10–20% of the time until you need the word — that's the recipe from Cepeda et al.'s 2008 follow-up. Need it next month? Review every 5–10 days. Need it in a year? Monthly is fine. There is no single magic interval; the schedule should stretch with your horizon. Most flashcard apps ship one default curve and never ask when you actually need the material.

## Why isn't Anki enough on its own?

Two reasons, one structural and one human:

- **Structural:** SRS solves recognition, not production. Being shown "однако" and recalling "however" is a different neural task from *reaching for* "однако" mid-sentence while someone waits. That gap is what SLA researchers call **automaticity** (Segalowitz 2010). Pimsleur's 1967 graduated-interval method trained both directions; most modern card apps quietly dropped the production half.
- **Human:** flashcards are lonely. Words divorced from any conversation you've ever had attach to nothing, and for most people the loneliness eventually beats the discipline. The graveyard of abandoned Anki decks is not a discipline problem — it's a design problem.

## What does conversation add that cards can't?

1. **Retrieval under pressure.** The half-second deadline of a real exchange is the exact skill fluency requires — and the one cards never simulate.
2. **Context.** You learn that "kind of" softens and "rather" sharpens — usage nuance that a definition side of a card cannot carry (lexical priming, Hoey 2005).
3. **Emotional tagging.** Words used in moments that mattered to you are recalled better, consistent with the affect-and-memory literature (Kensinger 2009).

The cleanest direct evidence for combining the two: Nakata (2008) and Nakata & Webb (2016) found spaced retrieval *plus active production* beats spaced retrieval alone for long-term retention.

## How do the common approaches compare?

| Approach | Trains | Strength | Weakness |
| --- | --- | --- | --- |
| Anki / classic SRS | Recognition | Free, precise scheduling | No production; high dropout |
| Gamified vocab apps | Recognition | Habit-forming streaks | Shallow context, topic-ordered |
| Conversation only | Production | Real retrieval, context, emotion | New words resurface by chance, not schedule |
| SRS + daily conversation | Both | Each covers the other's gap | Requires two habits — or one tool that does both |

## How TalkToDia closes the loop

Every word you actually use with Dia lands in your personal **word bank**, ranked by how actively you engage with it. The most active words get threaded back into your next conversation — so vocabulary keeps getting retested *in dialogue*, where production happens, instead of in a card queue. Smart flashcards are generated from your own conversations rather than a generic frequency deck, which means every card carries a memory of the moment you first used the word. (Calendar-style SRS scheduling à la Anki is on the roadmap; today the engine prioritizes by usage and recency.)

If you're past the beginner stage and wondering why your large passive vocabulary still doesn't come out of your mouth, that's the [intermediate plateau](/en/blog/breaking-the-intermediate-plateau) — and the fix is the same: keep the cards, add the conversation.

## FAQ

### How many new words should I learn per day?

For most adults, 5–10 genuinely new words a day is sustainable alongside reviews; 20+ is sprint pace that usually collapses within weeks. The constraint is not learning new cards — it is the compounding review load. Words you then actually use in conversation need far fewer scheduled reviews to stick.

### Is Anki enough to learn a language?

Anki is excellent for what it trains: recognition and recall on a schedule. It does not train producing words under time pressure, listening at native speed, or holding a conversation. Treat it as the memory layer of a system whose other layer is daily speaking practice.

### Why do I recognize words but freeze when speaking?

Recognition and production are different retrieval pathways, and flashcards only strengthen the first. The fix is retrieval practice under mild time pressure: daily conversation where you must reach for words, with a partner who recycles your recent vocabulary so retrieval gets retested before it decays.

### What is the forgetting curve?

The forgetting curve (Ebbinghaus, 1885; replicated by Murre & Dros, 2015) describes how memory for new material decays — steeply within the first 24 hours, then more slowly. Spaced reviews timed to interrupt the steep phase are the most efficient known counter to it.

## Sources

- [Ebbinghaus (1885) — Memory: A Contribution to Experimental Psychology](https://psychclassics.yorku.ca/Ebbinghaus/)
- [Cepeda et al. (2008) — Spacing effects on long-term retention](https://doi.org/10.1111/j.1467-9280.2008.02209.x)

---
Cite as: 间隔复习 + 对话：为什么两者结合最有效 — TalkToDia Blog, https://talktodia.com/zh/blog/spaced-repetition-and-conversation