Why rehearsing out loud works

Thinking is not speaking

The psycholinguist Willem Levelt described speech production as a three-stage process: conceptualisation (knowing what you want to say), formulation (choosing the words and constructing the sentence) and articulation (actually saying it). When you prepare for a conversation “in your head”, you mainly work on the first stage. You know what you want to get across. But formulation and articulation, the stages that unfold in real time when you are facing someone, remain largely unpractised.

This is why mental preparation gives a misleading sense of mastery. Researchers Leonid Rozenblit and Frank Keil demonstrated a related phenomenon, which they called the illusion of explanatory depth: we think we understand a subject in detail until we are asked to explain it out loud, step by step. The gap between “I know what I want to say” and “I can say it” only becomes apparent when we have to produce the words.

Rehearsing out loud forces that gap to show itself before the real conversation, when it costs nothing.

The production effect: you remember what you say better

In 2010, Colin MacLeod and his colleagues at the University of Waterloo gave a name to a phenomenon observed since the 1970s: the production effect. The principle is simple. When you read a list of words, you remember the ones you say out loud considerably better than the ones you read silently.

In a review published in 2017, MacLeod and Glen Bodner summarise what dozens of experiments have established since:

• Speaking out loud produces the strongest effect of all the modes of production studied. Writing, typing and spelling words out also help, but less so.
• Even mouthing the words silently, without making a sound, improves memory.
• Hearing someone else say the words helps a little, but less than producing them yourself.
• The advantage lasts: words spoken out loud are still recognised better a week later, and textbook paragraphs read out loud are recalled better the following day.

The leading explanation is distinctiveness. The act of saying something adds information to the memory trace: the movement of your mouth, the sound of your own voice, the sensation of having produced the sentence. When you try to remember it, these cues make the item easier to retrieve from everything else.

The role of your own voice

In 2018, Noah Forrin and Colin MacLeod went further. They compared four ways of learning a list of words: reading silently, listening to someone else read, listening to a recording of your own voice, and reading out loud yourself in real time. The ranking was clear: reading out loud came first, followed by listening to your own voice, then listening to someone else, and finally silent reading. Reading out loud gave a recognition advantage of around 12% over silent reading; listening to someone else, only 4%.

The authors draw a useful conclusion for anyone preparing for a conversation: a significant part of the benefit comes from the personal nature of the production. It is not just the sound that matters, but the fact that you are the one saying it, in your own words. Watching a video of someone expertly conducting a conversation to address poor performance or behaviour teaches you something. Conducting it yourself, however awkwardly, teaches you more.

Test yourself rather than reread

A second body of research reinforces the same message: the testing effect, or retrieval practice. In 2006, Henry Roediger and Jeffrey Karpicke asked students to learn short texts. Some reread them; others had to recall the content without the text in front of them.

Five minutes later, those who had reread the texts did better (81% versus 75%). Two days later, the pattern had reversed (54% versus 68%). After a week, the gap had widened: 42% for rereading, 56% for recall. In a second experiment, students who had reread the texts four times predicted that they would remember more than the others. They were wrong: a week later, they had the poorest results.

This last point is essential. Rereading, or mentally replaying a conversation, feels like learning because everything seems fluent and familiar. Having to produce the answer yourself is less comfortable, and it is precisely this effort that consolidates learning. Psychologists Robert and Elizabeth Bjork call these “desirable difficulties”.

Rehearsing a conversation out loud is a form of retrieval practice: you have to find the sentence, construct it and say it, without a crib sheet.

Mental rehearsal helps, but it has its limits

Visualisation has a good reputation, and rightly so. A meta-analysis by James Driskell, Carolyn Copper and Aidan Moran, published in 1994 in the Journal of Applied Psychology, concludes that mental practice genuinely improves performance. But the effect depends on the type of task and weakens as the time between practice and performance increases.

For conversations, mental rehearsal has an additional weakness: in your head, the other person cooperates. They react as you imagine they will, do not interrupt, and do not say “this is not the time” or “everyone does the same”. The real conversation does not follow your script. Rehearsing out loud with someone who responds, objects and asks further questions forces you to work on the part that actually causes difficult conversations to fail: responding to what you had not anticipated.

What research on simulation and role play shows

Beyond memory laboratories, research into professional training points in the same direction.

In 2020, Olga Chernikova and her colleagues at the University of Munich published a meta-analysis of 145 studies on simulation-based learning in higher education, covering more than 10,000 participants. The overall effect on complex skills was large (g = 0.85). For communication skills specifically, the effect was more modest but robust (g = 0.44), and role plays with simulated patients or pupils produced a medium effect (g = 0.63). The authors suggest an explanation: a simulation makes the gap between what you do and what you are aiming for visible, which no lecture can do.

An earlier study, now a standard reference, illustrates the other half of the picture. In 2004, William Miller and his team trained 140 professionals in motivational interviewing, a demanding conversational technique. All the trained groups improved more than the control group, but those who also received feedback on recordings of their interviews, or individual coaching, reached a higher level. One striking detail: participants’ self-assessments did not match their actual skill levels, as observed in the recordings. We are not very good at judging how we speak until we have heard ourselves.

What the research does not say

To give an honest account of these findings, three qualifications are worth making.

First, much of the research on the production effect concerns word lists, not conversations. The mechanism (production makes an item distinctive) is well established; the exact size of its effect in a real conversation is less clear.

Second, the production effect is stronger when comparing, within the same person, what they have said and what they have read silently. Between two separate groups, the recognition advantage exists but is smaller. Speaking out loud is not a magic formula; it is one tool among others.

Finally, remembering your sentences better is not enough to conduct a conversation well. Listening, judgement and managing emotions matter just as much. This is why the most useful practice combines three elements: speaking out loud, someone who responds, and precise feedback on what happened.

Put it into practice

A ten-minute exercise to try before your next important conversation:

1. 1Choose the conversation. Just one, a real conversation you need to have in the next few days.
2. 2Say your opening sentence out loud, three times. Not in your head. Out loud, as if the person were in front of you. Note what changes between the first and third versions.
3. 3Record yourself and listen back. Your phone is enough. You will hear what you do not notice while speaking: the justifications that come too early, the “just” and “a bit” that weaken the message.
4. 4Add resistance. Imagine the most likely objection (“it is not my fault”, “nobody told me”) and respond to it out loud. This is where most conversations go off course.
5. 5Keep just one adjustment. Not ten. One thing to do differently next time. (We explain why in the next article, One change at a time.)