How to Improve English Speaking | A Step-by-Step Plan and What to Do When You Plateau

Plenty of people who study English can read it well and follow most of what they hear, yet the words do not come when it is time to speak. This article is for learners who want a practical way to improve their English speaking, including those at an intermediate level who practise regularly and feel they have stopped moving forward.

It answers five questions in order: what to do step by step, what the 3/2/1 rule and the popular lists of speaking rules actually are, what the CEFR levels A1 to C2 mean for speaking, how to practise at home alone and for free, and why solo practice eventually stalls. The answer to the last question is the one discussed least: the sentences you could not say in a real conversation usually disappear before you get a chance to practise them.

Key takeaways

• Speaking improves in small steps: one situation, a few sentences, recall aloud.
• Nation’s 4/3/2 activity repeats one talk to new listeners in less time.
• The 3/2/1 rule is a shorter 4/3/2; one small study found only small gains.
• Popular lists of “7 rules” for speaking do not cite research behind them.
• CEFR levels run from A1 to C2 and describe speaking alone and with others.
• Self-talk, shadowing, a one-minute recording and recall cost nothing.
• Solo practice stalls when the sentences you failed to say are never collected.
• Karpicke and Roediger (2008) found that testing yourself beat studying again, a week later.

How to improve your English speaking, step by step

Most advice on improving English speaking arrives as a list of tips: watch films, learn vocabulary, talk to people, do not be afraid of mistakes. Each tip is reasonable, but a list does not say what to do first or what to do tomorrow. The steps below put the same ideas into an order that can be repeated every week, starting from a situation where English is actually needed.

The starting point is narrow on purpose. Speaking gets better in the situations that are practised, so it helps to pick one situation that will really come up in the next few weeks, such as introducing yourself at work, explaining a problem to a colleague, or describing your weekend at the start of a lesson.

• Step 1: choose one situation that will come up soon, and only one.
• Step 2: write three to five things you want to say there, in your own language.
• Step 3: find one natural English version of each, and keep only that version.
• Step 4: listen to each sentence read aloud, then say it out loud several times.
• Step 5: the next day, look only at your notes in your own language and say the English.
• Step 6: use at least one of the sentences in a real conversation that week.
• Step 7: right after that conversation, note what you wanted to say but could not.

Step 5 is the one most often skipped, because reading the English again feels like progress. A conversation offers no script to glance at. What speaking needs is the ability to turn what you mean into a sentence, and practising exactly that from the first day is closer to real speaking than rereading a phrase list.

Step 7 turns the week into a cycle. The notes from that conversation become the next week's step 2, so the material comes from the places where you actually got stuck instead of from a generic list. How to keep those notes in a form you can test yourself with is covered in the article on reviewing English lessons.

Grammar and vocabulary still matter, and they are easier to learn from sentences you are already using. When a sentence from step 3 contains a structure that is not yet clear, that structure is the one worth looking up, because there is already a reason to say it.

Techniques that work: timed retelling and the rules people quote

Two related questions come up again and again: what the 3/2/1 rule is, and what the seven rules for excellent English speaking are. They are very different things. One is a classroom activity with published studies behind it. The other is a list with no research attached.

The activity comes from language teaching research. Paul Nation described it in a 1989 paper titled “Improving speaking fluency”, and Arevart and Nation studied it in 1991. A learner gives the same talk or tells the same story three times to three different listeners, with less time for each telling: four minutes, then three, then two. Because the listener changes, the talk keeps a real audience. Because the time shrinks, the same content has to come out faster and with fewer stops.

In the 1991 study, learners spoke noticeably faster and hesitated less across the three tellings. That result describes what happened during the activity itself. It does not show how long the gain lasts afterwards, so it is better read as a way to practise being fluent than as proof of a permanent change.

The 3/2/1 rule is the same activity cut down to three minutes, two minutes and one minute, which fits more easily into a short class. Molina Barriga and Briesmaster tried this version in 2017 with 40 students at a technical university in Chile. They reported only a small improvement in words per minute and in pauses, and wrote that seven sessions were not enough. So 3/2/1 is a reasonable, shorter way to run the same idea, but on its own it rests on one small study.

The activity also works without a class. Alone, there is no new listener, but the timing and the repetition still apply, and a phone timer plus the phone's voice recorder is enough equipment.

• Pick a topic you can talk about for a few minutes, such as a recent trip.
• Set a timer for three minutes and tell it aloud, recording only your own voice.
• Tell it again in two minutes with the same content, then once more in one minute.
• Play back the last recording and note the words you searched for or avoided.
• Look those words up, and tell the same story again the next day.

One difference from the classroom version is worth knowing. Without a new listener each round, it is easy to slip from telling into reciting. Picturing a different person for each round (a colleague, then a friend, then someone who knows nothing about the topic) keeps the talk closer to speaking to someone.

The seven rules are a different matter. Lists with titles such as “seven rules for excellent English speaking” are widely shared online, but they do not cite research showing that the rules work. Some items overlap with sensible practice, such as learning whole phrases rather than isolated words and doing a lot of listening, and those parts are reasonable to keep. What such a list does not give is an order to work in or a way to check whether it is working, which is what the weekly steps and timed retelling provide.

What A1–C2 means for speaking

A1, A2, B1, B2, C1 and C2 are the six levels of the Common European Framework of Reference for Languages, usually shortened to CEFR, published by the Council of Europe. A1 is the lowest and C2 the highest. The Council of Europe groups them into three bands: basic user (A1 and A2), independent user (B1 and B2) and proficient user (C1 and C2). Its 2020 Companion Volume also describes a Pre-A1 stage below A1.

For speaking, the 2020 Companion Volume separates two things: speaking on your own, which it calls oral production, and talking with other people, which it calls oral interaction. The table below summarises the overall description of each level in plain words. It is a paraphrase, not the official wording.

Two details in these descriptions are easy to miss. First, much of the move from A2 to B1 is about independence: at A2 a conversation works when the other person helps, while at B1 a conversation on a familiar topic can be joined without preparation. Second, B2 is described in terms of ease for both sides (regular conversation without strain on either party), not in terms of sounding like a native speaker.

The descriptions are most useful for deciding what to practise next. If conversations on familiar topics already work but words still have to be searched for and some ideas get dropped because they feel too hard to say, the step from B1 to B2 is the one to aim for: saying more of what you mean, with fewer restrictions, at a steadier pace.

Daily practice at home, alone and for free

Practising alone at home costs nothing and needs no partner, and it is where most of the repetition happens. Each solo method trains a different part of speaking, so it helps to know what each one gives and what it leaves out.

Talking to yourself. Describing what you are doing while cooking, walking or tidying up forces you to build sentences on the spot, and gaps show up within a minute. Noting the missing words in your own language and looking them up later turns those gaps into material.

Shadowing. Repeating what you hear in a short audio clip, a moment after the speaker, trains rhythm, stress and the shape of common phrases. The sentences belong to someone else, though, so shadowing does not practise finding your own words.

Recording one minute of your own voice. Talking about one topic for a minute into a phone and then listening back reveals what goes unnoticed while speaking: long pauses, unclear words, or the same linking word used again and again. A recording of your own voice is all this method needs.

Recalling without the script. This means taking yesterday's sentences, looking only at the meaning in your own language, and saying the English. It takes a few minutes, and among the solo methods it is the closest to a real conversation, where nothing can be looked up.

A free AI chatbot can act as a partner that is available at any hour. The official plan comparison for ChatGPT shows that its free plan includes unlimited text chats, while voice chats and memory are limited. Typed role-plays are therefore free, but long daily voice sessions may run into the limit, and the chat will not reliably remember what gave you trouble last time.

An AI partner has one limit that matters for this article: it only knows what is typed or said to it. The sentence that would not come out in a meeting, a lesson or a chat with a friend never reaches it unless you bring it up yourself.

Why solo practice plateaus: the sentences you could not say

After a few months of regular practice, many learners reach a stage where they practise every day and nothing seems to change. What often happens at that stage is that practice keeps going over sentences you can already say, while the sentences that were needed in real conversations, and did not come, pass by without ever being practised.

The linguist Richard Schmidt proposed in a 1990 paper that second-language learners learn what they notice, and that input which passes by unnoticed tends not to be learned. It is a proposal about the role of attention rather than an experiment, and it does not claim that noticing a gap is enough to close it. It claims that noticing is where learning starts.

In speaking, the clearest moment of noticing is the moment a sentence will not come. In that moment it is obvious what is missing. The conversation keeps moving, though, so there is no time to find the answer, and by the time it ends, the feeling of being stuck is still there while the sentence itself has usually faded.

How that sentence is practised once it has been found also matters. In an experiment published in Science in 2008, Karpicke and Roediger had students learn Swahili–English word pairs and compared studying the pairs again with being tested on them again. On a test one week later, repeated retrieval produced a large benefit, while further study after the pairs had been learned added nothing. The experiment measured memory for word pairs, not speaking, but it suggests that pulling an item out of memory is what makes it stick.

Put together, the two ideas suggest a short loop you can run after any real conversation and needs only paper or a notes app.

• Right after the conversation, note one to three things you wanted to say but could not.
• Find one natural English version of each, and keep just that version.
• The next day, look only at your note and say the English sentence aloud.
• If it does not come, check it, say it again, and try once more a few days later.
• In the next conversation, use the sentence at least once on purpose.

The sentences that did not come out show exactly where fluency runs out, so practising them prepares you for the next conversation better than picking new phrases from a list. When the plateau shows up despite regular lessons, the article on whether online English lessons work looks at the hours involved and at what changes when progress stalls.

Summary

• Start from one real situation and a few sentences, and recall them aloud.
• Timed retelling (4/3/2 or 3/2/1) trains saying the same content faster.
• The 3/2/1 version rests on one small study; the 7-rule lists on none.
• CEFR A1–C2 describes speaking alone and with others; aim at the next step.
• At home, self-talk, shadowing, one-minute recordings and recall are free.
• Collecting the sentences you could not say gives solo practice a target.