How the four weeks are built

This chapter turns the models of the book into four weeks of practice that any couple can do at home, without equipment, money or a therapist. Each week works on one term of the drive $\mu = J – L – c$ and adds a measurement, so that by the end the couple has a month of their own data.

The weeks build on each other: the protected hour of week 1 continues through all four weeks, and each new week adds one practice.

Week 1: measure and close the leak

Practice. Choose one hour each evening, the same hour if possible, when both are at home. At the start of the hour, both phones go to another room, set to allow calls from a short list of people who might need you urgently. Televisions and laptops stay off unless both choose to watch the same thing together. What you do in the hour is free: cook, talk, walk, sit together.

Measurement. At the end of the hour, each partner writes down how many minutes inside the hour went to screens anyway, and how many times they picked up a device. Be honest; the point is to see the leak, not to win.

Why. Chapter 5’s simulation found that moving the phone out of reach removed the leak completely, while willpower in the presence of the phone did not. Week 1 tests that on your own evenings. Expect the first evenings to feel restless: the learning machinery is still expecting the variable-ratio reward. If the model is right, the restlessness should fall within the week; the log will show whether it does.

Week 2: add micro-novelty

Practice. Each day, one partner (alternate days) introduces one small, new, positive shared experience: a dish neither has cooked, a different route for the walk, a song from the other’s teenage years, a question never asked before. It should take minutes, not hours, and should be learnable: something that tells you about the world or each other, not random chaos.

Rules. Keep the core predictable while the content changes (Chapter 4). Novelty is not testing the partner, provoking jealousy or springing unwelcome surprises. If the partner does not enjoy something, that is information, not failure.

Measurement. Each evening, each partner rates two things from 1 to 7: wanting (“Today I looked forward to time with my partner”) and liking (“Today I felt warmly towards my partner”). Chapter 2 predicts that novelty should move wanting more than liking.

Week 3: joint attention and zero-point listening

Practice. Fifteen minutes a day, inside the protected hour, in three parts.

1. Five minutes: one speaks, one listens. The speaker talks about anything that mattered today. The listener does not plan a reply, give advice or judge; they attend to the voice and the person, returning attention whenever it drifts. This is the listening practice of Turning the Hearing Inward, the second book of this library, applied to a partner: attention rests on the sound before it moves to the reply.
2. Five minutes: switch roles.
3. Five minutes: one shared object of attention, in silence or almost: a walk, a sunset, a piece of music, a meal eaten slowly.

Measurement. Record the minutes of fully shared attention in the evening, and write one sentence together about something the day meant to both of you. Over the week, these sentences become a small record of shared meaning (Chapter 6).

Week 4: repair speed and early warning

Practice. Agree, at a calm moment, on a repair signal: a word, a gesture or a phrase that either partner can use to say “I want to stop this escalating and come back to us”. Agree that the signal is honoured even when the disagreement is not resolved; the issue can be returned to later.

Measurement. When a conflict happens, note the time it starts and the time both partners feel “back to normal”. The difference is the recovery time. Keep a list.

Early warning. Chapter 6 showed that recovery time lengthens as a relationship approaches its tipping point. If recovery times grow for two weeks in a row after the programme, treat it as a warning: protect the evening hour more strictly (reduce $L$), add novelty and shared meaning (raise $J$), and consider outside support.

The daily log

All four weeks use one simple log, kept by each partner or together. A spreadsheet or a sheet of paper works. The columns are:

import csv, io
from math import log2
from statistics import mean, median

TEMPLATE = "date,week,protected_min,screen_in_protected_min,pickups,shared_min,novelty,wanting,liking,conflict,recovery_min\n"

week1 = """2026-11-02,1,60,22,9,35,0,4,6,0,
2026-11-03,1,60,15,6,40,0,4,6,1,180
2026-11-04,1,60,9,4,50,0,5,6,0,
2026-11-05,1,60,6,3,55,0,5,6,0,
2026-11-06,1,60,4,2,58,0,5,7,0,
2026-11-07,1,60,3,1,60,1,6,7,1,45
2026-11-08,1,60,2,1,60,0,5,7,0,
"""
rows = list(csv.DictReader(io.StringIO(TEMPLATE + week1)))
leak = [int(r["screen_in_protected_min"]) for r in rows]
rec = [int(r["recovery_min"]) for r in rows if r["conflict"] == "1"]
print(f"leak L in the protected hour: day 1 {leak[0]} min -> day 7 {leak[-1]} min (week total {sum(leak)} min)")
print(f"pickups per hour: {mean(int(r['pickups']) for r in rows):.1f}   shared minutes per evening: {mean(int(r['shared_min']) for r in rows):.0f}")
print(f"wanting {mean(int(r['wanting']) for r in rows):.1f} / 7   liking {mean(int(r['liking']) for r in rows):.1f} / 7")
print(f"conflicts: {len(rec)}   median recovery time: {median(rec):.0f} min")

def H(minutes):
    total = sum(minutes)
    return 0.0 - sum(m / total * log2(m / total) for m in minutes if m)

print(f"attention entropy of the protected hour, day 1: {H([60 - leak[0], leak[0]]):.2f} bits, day 7: {H([60 - leak[-1], leak[-1]]):.2f} bits")
# leak L in the protected hour: day 1 22 min -> day 7 2 min (week total 61 min)
# pickups per hour: 3.7   shared minutes per evening: 51
# wanting 4.9 / 7   liking 6.4 / 7
# conflicts: 2   median recovery time: 112 min
# attention entropy of the protected hour, day 1: 0.95 bits, day 7: 0.21 bits

In the example, the leak inside the protected hour falls from 22 minutes on the first evening to 2 on the seventh, and the attention entropy of the hour falls from almost one bit to about a fifth of one. These are the quantities Chapters 1 and 5 asked about, measured on an ordinary evening.

After the four weeks

At the end, sit down together with the log. Three questions are worth asking, and the numbers help answer them.

1. Where was the leak, and did it close? Compare the first and last weeks’ screen minutes and pickups.
2. What moved: wanting or liking? If wanting rose while liking stayed high and steady, the pattern matches Chapter 2’s prediction.
3. How long did repairs take? A median recovery time measured in minutes is a good sign; one measured in days is the early warning of Chapter 6.