A Wandering Mind Is an Unhappy Mind — The 250,000-Moment Happiness Study Run on iPhones

Opening — The Two-Page Paper That Changed How Happiness Is Studied

The classic method of happiness research is the survey. "Over the past month, how happy have you been overall?" The problem is that human memory is a terrible accountant. We remember last month not as it was actually experienced but as a summary of a few impressive moments.

In 2010, Harvard's Matthew Killingsworth and Daniel Gilbert chose a different method. They built and distributed an iPhone app (trackyourhappiness.org) that sent notifications at random moments of the day and asked exactly three things: How do you feel right now? What are you doing right now? Are you thinking about something other than what you are currently doing?

The data gathered this way amounts to roughly 250,000 momentary samples collected from 2,250 adults in 83 countries. The paper published in Science has a body of barely two pages, but I chose it as the fifth installment of Psychology Through Papers because both the results and the method are beautiful.

Three Findings — In Numbers

First, the mind wanders for nearly half of waking life. In 46.9% of all samples, people were thinking about something other than what they were doing. What is more, this rate was almost independent of the type of activity. Whatever people were doing (with a single exception — in the paper's phrasing, making love), the rate of mind wandering never fell below 30%.

Second, whether the mind was wandering predicted happiness better than what people were doing. In the regression analysis, "am I mind-wandering right now" explained momentary happiness more strongly than "what kind of activity am I doing right now". Even during commuting (the byword for unpleasant activities), the moments of being absorbed in it were not as bad as expected, and even during pleasant activities, happiness dropped once the mind had drifted away.

Third — the most counterintuitive result — even pleasant mind wandering was no better than focus. Splitting the content of wandering thoughts into pleasant, neutral, and unpleasant, both unpleasant and neutral wandering were clearly less happy than the focused state, and even pleasant daydreams were no happier than being absorbed in the task at hand. On top of this comes the time-lag analysis: mind wandering now predicted unhappiness a moment later, while unhappiness now predicted later mind wandering more weakly. The result suggests that the causal arrow may point in the direction of "mind wandering → unhappiness" (though, since this is not an experiment, it is not definitive).

The last sentence of the paper is one of the most frequently quoted closing lines in the history of psychology papers: "A human mind is a wandering mind, and a wandering mind is an unhappy mind."

Understanding It in Code — The Structure of Experience Sampling Data

The real lesson of this study lies in its method. Let us build toy data to see what experience sampling data looks like and why "within-person analysis" matters.

import random
import statistics

random.seed(7)

# toy model: each person has a happiness baseline (trait),
# and each sampled moment is focused or wandering (state)
people = []
for pid in range(200):
    baseline = random.gauss(60, 12)          # stable individual differences
    moments = []
    for _ in range(30):                       # ~30 pings per person
        wandering = random.random() < 0.47    # 46.9% in the real study
        state_penalty = 8 if wandering else 0 # wandering costs ~8 points
        mood = random.gauss(baseline - state_penalty, 10)
        moments.append((wandering, mood))
    people.append(moments)

# WRONG-ish: compare across ALL moments, ignoring who they came from
all_w = [m for p in people for (w, m) in p if w]
all_f = [m for p in people for (w, m) in p if not w]
print(f"between-sample gap : {statistics.mean(all_f) - statistics.mean(all_w):.2f}")

# BETTER: within-person gap, then average over people
gaps = []
for p in people:
    w = [m for (is_w, m) in p if is_w]
    f = [m for (is_w, m) in p if not is_w]
    if w and f:
        gaps.append(statistics.mean(f) - statistics.mean(w))
print(f"within-person gap  : {statistics.mean(gaps):.2f}")

In this toy example the two values come out similar. But if there were an individual difference such that "people who are happier to begin with wander less", a comparison that lumps the whole sample together would mix the effect of the state with the effect of the person. This is why the actual paper used a multilevel model (a regression with person as a level): the 250,000 "moments" are nested inside 2,250 "people". When reading experience sampling studies, the habit of checking "is this a within-person effect or a between-person effect" is a filter that screens out half the papers in this field.

The Limitations, Stated Fairly

As the rules of this series demand, we look at the limitations too. First, the sample consists of volunteers who owned iPhones (early adopters, by 2010 standards). Representativeness is limited. Second, it is a correlational study. The time-lag analysis hints at a direction, but there is no experimental manipulation. Third, later studies have refined the picture: there are reports that, among kinds of mind wandering, deliberate wandering on interesting topics (planning, creative daydreaming) can be neutral or even positive. The summary is being polished from "all mind wandering is poison" into "the problem is the wandering you get dragged into unintentionally".

Even so, the core finding — that attention strongly shapes the quality of experience — has never had its direction reversed in follow-up research.

What We Take Away — The Unit of Happiness Is the Moment

1. See happiness as a matter of attention, not only of conditions. We tend to think of happiness as a function of conditions (salary, house, vacation), but this data says that even within the same conditions, where your attention sits makes a large difference to the happiness of the moment. Changing conditions takes years; bringing your attention back takes seconds.

2. The rediscovery of boring chores. Times when thought "comes loose" — commuting, dishwashing, walking — are hotbeds of mind wandering. Bringing attention back to the senses during that time (the soles of your feet, the sound of water, the breath) is the whole of mindfulness training, and this study shows the expected payoff of that training in data. It shares roots with the handling of rumination covered in A Prescription for Overthinking Nights.

3. Flow is a happiness skill too. The flow installment treated flow as a performance skill, but this study shows that flow is also a happiness skill — flow is the state in which mind wandering is structurally impossible. Designing task difficulty and blocking interruptions is not productivity tuning; it is quality-of-life tuning.

Guide to Reading the Originals

• Original paper: Killingsworth, M. A., & Gilbert, D. T. (2010). A wandering mind is an unhappy mind. Science, 330(6006), 932.
• Follow-up refinement: Smallwood, J., & Schooler, J. W. (2015). The science of mind wandering. Annual Review of Psychology, 66, 487-518.

Reading tips: the original paper is just two pages of text and a single figure, making it the lowest barrier to entry in the series. Spend three minutes on the happiness-versus-mind-wandering scatter by activity in Figure 1. The methodological details are in the supporting material. The next installment features the same author — The Money and Happiness Debate.