Joseph Murphy’s ideas vs the research: what holds up, what doesn’t

The scorecard

Each row is one of Murphy’s ideas, in our words, with the strongest study we could verify and our verdict. Effect sizes are given where the study reports them in its abstract or full text; where it does not, we say so rather than borrow a number from a blog.

How we graded

We took Murphy’s ideas from the 1963 text and matched each to the closest published research. We read each study’s abstract or full text on 20 October 2026; numbers we could not find in a primary source (several circulate online) are left out. A meta-analysis, which pools many studies, counts for more than a single experiment, and an experiment counts for more than an observational study that follows people over time.

Psychology has its own reliability problem, and it matters here. When the Open Science Collaboration (2015) repeated 100 published studies, 97% of the originals had reported significant results but only 36% of the repeats did, with effects about half the size. So “one study found” is weaker than it sounds, which is why the single-study rows above say so.

• Supported: a meta-analysis or a replicated finding points the same way as Murphy.
• Supported, small effect: the direction is right, the size is modest.
• Supported, one study: a single experiment points Murphy’s way; treat it as provisional.
• Partly supported: research confirms part of the idea, not Murphy’s explanation.
• Mixed: it helps some people and harms others.
• Contradicted: the best studies point the other way.
• Not supported: no study we know of supports it.

What holds up

Sleeping on a problem. Murphy’s chapter on scientists claims that problems handed over before sleep come back solved. Wagner and colleagues (2004) trained people on a number task with a hidden shortcut, then let them sleep or stay awake for 8 hours: more than twice as many found the shortcut after sleep. But sleep did nothing for people who had not practised the task first. The research supports “work on it, then sleep on it”, not “hand it over”.

Habits. “Habit is the function of your subconscious mind,” Murphy writes in chapter 18. Habit research would phrase it differently: habits form when the same action is repeated in the same context until the context triggers it (Wood & Neal, 2007). In Phillippa Lally’s 2010 study, 96 volunteers repeated one simple behaviour daily in the same context for 12 weeks. Among the 39 whose data fitted the model well, it took a median of 66 days to near its peak of self-reported automaticity, from 18 to 254 days (the longer times are extrapolated beyond the 12 weeks), and missing a single day did not derail it.

Gratitude and forgiveness. Both are rituals Murphy recommends, and both have a research base, with effects smaller than he suggests. Gratitude exercises raise well-being a little compared with another activity (d = 0.17), and do no better than other psychologically active exercises (Davis et al., 2016). Forgiveness programmes led by a professional show moderate effects, larger when they last longer (Wade et al., 2014).

Rehearsal and process. Mental practice of a skill improves performance, though a 2020 update found the effect small. And picturing the steps works better than picturing the prize: in Pham and Taylor’s study, process simulation improved studying and grades.

What holds up only in part

The drowsy state. Murphy puts the threshold of sleep at the centre of his method. Recent sleep research gives that moment real weight: in a 2021 study of 103 people by Lacaux and colleagues, those who spent at least 15 seconds in N1, the first and lightest stage of sleep, found a hidden rule in a number task nearly three times as often as those who stayed awake (83% vs 30%), and the advantage vanished in those who reached deeper sleep; sleep stage was observed, not assigned at random. And in a study of 49 people, Haar Horowitz’s team (2020) found that a spoken cue played at sleep onset turned up in 67% of dream reports, though it also turned up in 52% of reports from people who heard it awake, a gap that was not statistically significant. What none of this shows is that an idea “impressed” at that moment changes events.

Not forcing. Murphy writes “Easy does it” in chapter 8: effort works against you. Wegner’s white-bear experiments (1987) support part of this: people asked not to think of a white bear kept thinking of it, and when later asked to think about it, did so more often than people asked to think of it from the start. That is evidence about controlling thoughts, not about getting results.

What research contradicts

Picturing the outcome. In chapter 6 Murphy calls visualizing “the easiest and most obvious way to formulate an idea”. Oettingen’s research points the other way for one kind of visualizing. In four studies, among them graduates looking for a job and students with a crush, positive expectations (judging the outcome likely) predicted effort and success weeks to two years later, while positive fantasies about the outcome predicted less (Oettingen & Mayer, 2002). In four experiments, induced positive fantasies even lowered energy, measured by physiological and behavioural indicators, compared with other kinds of thoughts (Kappes & Oettingen, 2011).

Affirmations. Murphy’s Baudouin technique condenses a wish into a short phrase repeated like a lullaby. Wood’s 2009 experiments asked people to repeat “I’m a lovable person”: those with high self-esteem felt slightly better, those with low self-esteem felt worse. This is not the same as “self-affirmation” in research, which means writing about your values, and whose effects did not reproduce in a large, well-powered school replication (Hanselman et al., 2017).

What has no support

An obedient subconscious. Murphy’s model is a mind that accepts whatever the conscious mind decrees. Psychology does describe a large automatic, nonconscious side of the mind (Bargh & Chartrand, 1999). But that research is about habits, quick evaluations and goals running in the background, not a servant that carries out instructions. Some of its most famous findings have also failed to replicate: when Doyen’s team repeated a study claiming that words about old age made people walk slower, 120 participants timed by infrared sensors showed no difference.

Money and the body. Murphy’s chapters claiming that belief brings money and resolves illness have no research behind them, and critics of positive thinking add that such claims blame people for their misfortune. For health, ask a professional.

The evidence-based version of Murphy’s method

Keep Murphy’s rituals, swap his explanations for what the studies support. The table pairs each aim with his instruction and the version research favours.

Where the night ritual fits

The “if X, then I do Y” line in the table comes from research on implementation intentions: across 94 independent tests, Peter Gollwitzer and Paschal Sheeran (2006) found a medium-to-large effect on reaching goals (d = 0.65). A 2025 meta-analysis of 642 tests by Sheeran, Listrom and Gollwitzer reports effects from small (d = 0.27) to medium (d = 0.66) depending on the outcome. It is the strongest evidence in this whole area, and it does not appear in Murphy’s 1963 book.

Dream Architect’s Full flow borrows the idea lightly: after the moment, the emotion and a sensation, you pick “Tomorrow’s sign”, an ordinary cue you will notice the next day, and “One small step”, a gesture of under a minute that closes the evening. To be clear, the app does ask you to picture the moment your dream is real, so it does not avoid outcome imagery: it keeps that scene ordinary and in the present tense, adds a cue and a step, and the next morning asks whether anything reminded you of it. The cue to notice tomorrow is the part closest to what this scorecard supports, though it is not a full if-then plan.

Limits of this scorecard

Murphy did not run experiments, so we matched his ideas to the nearest research, and the match is ours to defend. Most of these studies used students in short lab sessions. Some effect sizes we wanted were not in any source we could read, so we left them out. And a verdict of “not supported” means no evidence, not proof of the opposite. We will update the table when stronger studies appear.

Questions people ask next