One change at a time

Working memory has limited capacity

Working memory is the mental space where you temporarily hold the information you need to act. It is small. In an influential review published in 2001, psychologist Nelson Cowan estimates its capacity in adults at around four items at a time, far fewer than the often-quoted “seven plus or minus two”.

A difficult conversation already takes up almost all that space. You are listening to what the other person says, interpreting their tone, managing your own emotions, finding your words and anticipating what comes next. Adding a mental list of six instructions (“slow down, paraphrase, do not justify myself, give the figure, draw things to a close, avoid ‘actually’”) means asking an already full space to hold twice as much. In practice, most of the instructions fall away, often including those that mattered most.

One instruction, by contrast, fits. It can stay in mind throughout the conversation without displacing your ability to listen.

More feedback does not mean better feedback

The most widely cited meta-analysis on the subject is by Avraham Kluger and Angelo DeNisi, published in 1996 in the Psychological Bulletin. It covers 607 measured effects from more than 23,000 observations. On average, feedback improves performance (d = 0.41). But in more than a third of cases, it makes performance worse.

To explain this finding, the authors propose feedback intervention theory. The central idea is that feedback helps when it keeps attention on the task and how to carry it out. It becomes harmful when it shifts attention towards the self (“am I any good?”, “what do people think of me?”). At that point, mental energy goes into self-evaluation, defensiveness and emotion rather than improvement.

A long list of things to correct has precisely this effect. Beyond a certain number, it no longer reads as “here is what to adjust” but as “here is everything that is wrong with you”.

In a review of the literature on formative feedback published in 2008, Valerie Shute makes some very practical recommendations. These include presenting feedback “in manageable units”, because too much information at once leads to superficial learning and cognitive overload; keeping feedback “as simple as possible, but no simpler”; and focusing it on the task, not the person. She also notes that feedback that is too long or too complicated is often simply ignored.

What makes feedback genuinely useful

The aim is not to give less useful information, but to choose it more carefully.

John Hattie and Helen Timperley distinguish four levels of feedback: task (what was done), process (how to do it better), self-regulation (how to monitor yourself) and person (“well done”, “you are talented”). The last is the least effective because it says nothing about what to do next.

A more recent meta-analysis by Benedikt Wisniewski, Klaus Zierer and John Hattie, published in 2020 and covering 435 studies and around 61,000 participants, confirms the average effect of feedback (d = 0.48). Crucially, it shows that the most effective feedback contains the most useful information about the task and the method (d = 0.99).

This does not contradict the idea of making one change. Information-rich feedback is not the same as lengthy feedback. It is feedback that says precisely what to change, how and why. “In your opening, state the subject in the first sentence, before giving the context, so the person does not have to guess where you are heading” is rich in information. A list of ten vague observations is not.

Deliberate practice: one target at a time

Anders Ericsson’s work on expertise, beginning with his 1993 study of violinists, popularised the concept of deliberate practice. In his later writing, Ericsson describes it as training designed to improve a specific aspect of performance, with a clearly defined goal, focused attention, immediate feedback and the opportunity to repeat and refine.

The key word is “specific”. A musician does not work on “playing this piece better”. They work on three bars, at a given tempo, until they can play the passage cleanly. The same logic applies to conversations: you do not work on “handling an unhappy client better”, but on “letting the person finish their sentence before replying”.

There is an important qualification. More recent research, notably the 2014 meta-analysis by Brooke Macnamara, David Hambrick and Frederick Oswald, shows that deliberate practice explains a substantial share of differences in performance in highly structured fields such as music or games, and a much smaller share in professional work. Practice does not explain everything. But the principle of working on a specific target with feedback remains one of the best-documented ways to improve.

Too much guidance creates dependence

Research on motor learning provides a further argument. According to the guidance hypothesis, proposed by Alan Salmoni, Richard Schmidt and Charles Walter in 1984, feedback that is too frequent and too detailed helps during training but harms lasting learning: the person relies on external correction rather than developing their own ability to spot mistakes. In 1990, Carolee Winstein and Richard Schmidt showed that reducing the frequency of feedback improved skill retention.

For conversation, the implication is simple. The aim is not for a coach or a tool to correct every sentence, but for you eventually to hear for yourself when you start justifying your position. One clearly understood thing to watch for builds that ability. Ten simultaneous corrections short-circuit it.

Turn the change into an “if… then” plan

Choosing a change is not enough; it also needs to happen at the right moment. Peter Gollwitzer and Paschal Sheeran analysed 94 tests of implementation intentions: plans framed as “if this situation arises, then I will do this”. Their average effect on goal attainment is large (d = 0.65). The authors show that these plans make the triggering situation easier to recognise and the response more automatic, reducing the demands on working memory.

Applied to a conversation:

• Vague goal: “I need to justify myself less.”
• Implementation intention: “If the person says ‘it is not my fault’, then I will respond with a question before giving my point of view.”

The second version is a single change tied to a specific moment. This is the format most likely to hold up under the pressure of a real conversation.

How to choose the right change

Whether you are a manager, a trainer or practising on your own, here are three criteria for choosing the change to work on:

1. Impact. Which adjustment would make the greatest difference to the outcome of the conversation? Often, it is the opening or the response to the first objection, not verbal habits.
2. Observability. The change must be visible or audible: you can check whether someone “states the subject in the first sentence”, but not whether they “are more empathetic”.
3. Immediate feasibility. It must be possible to apply it on the very next attempt, not after three months of work.

The other observations do not disappear. They wait their turn. Once the first change is established, it frees up space for the next.

What this does not mean

“One change at a time” does not mean ignoring a serious problem. If a conversation has gone wrong on a matter of substance, such as false information or an unacceptable remark, that issue becomes the priority. The principle concerns how to improve, not what you allow yourself to notice.

Nor does it mean the debrief should lack substance. You can note ten observations. Only one, however, becomes the goal for the next attempt.

Put it into practice

An exercise for your next debrief, whether you are receiving or giving the feedback:

1. 1List everything. Write down every observation without filtering.
2. 2Circle one. Apply the three criteria: impact, observability and feasibility.
3. 3Write it as an “if… then” plan. “If [specific moment], then I will [specific behaviour].”
4. 4Rehearse the conversation again straight away. Not next week. Go through the same passage again, keeping just that one instruction in mind.
5. 5Check, then move on. When the change happens without effort, choose the next one on the list.