I did Zone 2 only for a month: what actually changed

It’s 6:50am on the Isar and I am, once again, going deliberately slowly. Week three of this. A man with a dog overtakes me. He is walking.

I’d decided to spend a month doing nothing but Zone 2. No intervals, no hard efforts, no accidental race with someone on the towpath. Just the slow stuff, four or five times a week, for thirty days, to find out what actually changes — because I’ve written about why Zone 2 is the training that works and about how it stacks up against intervals, and both of those posts are essentially me relaying other people’s data. I wanted to know what a month of it feels like from the inside.

Here’s the honest headline, so you can leave early if you like: a month of Zone-2-only training changed how the training felt quite a lot, changed my resting heart rate barely at all, and the gap between those two things turned out to be the most useful thing I learned.

It also turns out that gap says much less about my physiology than about my instruments, which I found mildly humbling.

What I actually did

The rules were deliberately boring:

• Four to five sessions a week, 45-75 minutes each. Mostly walking-to-jogging along the Isar, one longer session at the weekend through the Englischer Garten, and two indoor bike sessions on the mornings Munich decided to be Munich about the weather.
• A hard heart-rate ceiling. I picked a number at the top of my conversational range and treated it as a wall rather than a target. If I went over, I slowed down. This is much more irritating than it sounds — every gentle incline becomes a negotiation.
• No hard sessions at all. No intervals, no hill sprints, no “well, I’m nearly home so I might as well.” Strength training stayed in (three times a week, unchanged) because removing it would have confounded everything and also because I like it.
• One measurement I cared about: my time for the same 6km Isar loop, at the same heart-rate ceiling. Plus resting heart rate, tracked overnight, which I looked at far too often.

The one thing I did not do, which I’ll come back to, is take a proper baseline first. I just started. Bit of a schoolboy error, that.

What happened

Week one was the worst. Not physically — physically it was nothing, that’s the point — but psychologically. Holding a heart-rate ceiling when you’re used to running by feel means spending a lot of the first week walking up hills you’d normally jog. I felt like I was wasting the mornings. My weekly training felt so easy that I genuinely wondered whether I was doing anything at all.

Week two, the loop got quicker. Same ceiling, same route, roughly twenty seconds a kilometre faster than where I started. That’s the first thing that surprised me: the change didn’t show up as “I feel fitter”, it showed up as the same effort covering more ground.

Weeks three and four, that continued and then flattened. By the end I was around forty seconds a kilometre quicker than the first week at the same heart rate, most of which had arrived by day fifteen or so. The sessions also stopped feeling like a chore, which I’d put down to my legs no longer arguing with the pace.

Resting heart rate did almost nothing. My overnight rolling average drifted from somewhere around 58 to somewhere around 55-56 — call it two to three beats, which sounds like a result until you look at the day-to-day spread, which was easily that wide on its own. One glass of wine, one short night, one mildly stressful Tuesday, and the number moved as much as a month of training did.

I want to be careful here, so: this is one person, no control, no lab, thirty days, and the resting heart rate figure comes from a phone’s overnight estimate rather than an ECG. It’s a diary with a heart rate monitor attached. Treat it accordingly.

The bit that made this worth writing

I went looking for what should have happened in thirty days, and found a study that’s essentially my experiment done properly, in 1996, with needles.

Phillips and colleagues took seven trained volunteers, put them through two hours of cycling at 59% of their pre-training VO2 peak, five to six times a week, and biopsied their vastus lateralis at baseline, after 5 days, and after 31 days. (According to PubMed: Phillips et al., American Journal of Physiology, 1996, DOI.)

The timeline is worth having in full:

• By day 5, muscle lactate accumulation, phosphocreatine breakdown and glycogen depletion during a standard exercise challenge were all already significantly reduced. Within five days. The body had got noticeably better at the job before anything structural had changed.
• Those early gains happened before mitochondrial capacity moved at all. Succinate dehydrogenase activity — their marker of muscle oxidative potential — only rose by day 31, and then by 41%.
• VO2 peak rose 10%, and only at day 31. Nothing at day 5.

So the shape of the response is: fast improvements in how you handle a given workload first, and the engine itself getting bigger later. But note where later landed — inside the month. By day 31 the slow adaptations weren’t pending. They were measurable, and substantial.

Two caveats before I lean on any of that. Phillips’s team sampled at day 5 and day 31 and nowhere in between, so the study tells you what had happened by each of those points, not when anything started or stopped. “The fast adaptations finish and then the slow ones begin” is a story the data is compatible with, not one it demonstrates. And their volunteers cycled two hours a day, five to six days a week — roughly triple my volume — so their thirty days are not my thirty days.

What survives both caveats is the bit that genuinely reframed my month, and it isn’t about timing at all. The mitochondrial adaptations turned up in a muscle biopsy and a laboratory VO2 test. I own neither. Something similar was very likely happening in my legs; it just has no route to my phone. What I could see was pace at a fixed heart rate — the fast half of the response — which duly flattened around day fifteen and then looked, from the outside, like nothing else was happening.

So the month didn’t fail to produce the slow adaptations. It failed to show them to me, because the only instruments I had were the ones that read the fast ones. That’s a measurement problem wearing a physiology costume, and it’s the trap in every thirty-day experiment I’ve ever seen someone run.

The four things I got wrong

1. No baseline period. I started training on day one and started measuring on day one, so I have no idea what my resting heart rate was doing before. Two weeks of just measuring first would have cost me nothing and made the whole thing interpretable.

2. I picked a metric I had no way to actually read. Resting heart rate genuinely predicts mortality — the numbers on that are strong — but it’s also so twitchy that population-scale wearable data can spot flu outbreaks from it. Radin and colleagues used Fitbit resting heart rate and sleep data from 47,000 consistent wearers to improve state-level influenza-like-illness prediction across five US states, with model correlations to CDC rates of 0.84 to 0.97. (According to PubMed: Radin et al., The Lancet Digital Health, 2020, DOI.) A signal sensitive enough to see flu season arriving is a signal that moves for reasons that have nothing to do with your training.

I want to be careful with the arithmetic here, because the obvious version of this argument is wrong and I believed it for a while. “My daily spread is ±3 bpm and the effect is 2-3 bpm, therefore I can’t see it” doesn’t hold: thirty readings with a 3-bpm standard deviation average down to a standard error of roughly 3/√30 ≈ 0.6 bpm. On paper, a 2-3 bpm shift is comfortably detectable. Averaging is genuinely powerful, and I don’t want to talk anyone out of it.

What breaks it is everything the averaging quietly assumes:

• The readings aren’t independent. A bad week of sleep drags five consecutive nights up together. Whatever my effective sample size was, it was closer to the number of weeks than the number of days — and four is a much less impressive number than thirty.
• There was nothing to average against. A precise mean is useless without a prior mean to compare it to, and I never took one.
• The confounders drifted with the intervention rather than around it. More daylight, more time outside, better mood, a gradual shift in when I went to bed. Those don’t cancel out; they ride along.

So the honest diagnosis isn’t “the signal was too small.” It’s “one number, no before, and readings that move in clumps” — which is a far more fixable problem, and the reason the first rule below is the one that actually matters.

3. I was probably the wrong person for a big effect. The largest analysis of this question — 191 studies, 215 samples — found that endurance training reliably lowers resting heart rate, but that the size of the drop was positively related to the starting resting heart rate and negatively related to age. (According to PubMed: Reimers, Knapp & Reimers, Journal of Clinical Medicine, 2018, DOI.) The people who get the dramatic 8-10 bpm drops are the people who started high. I started at 58. There wasn’t much room.

4. Thirty days was a calendar decision, not a physiological one. I picked a month because a month is a nice unit for a blog post, which is not a reason. Twelve weeks would have been the honest window — long enough for the changes to outgrow the things I couldn’t control.

How to run this better than I did

If you want to actually learn something from a self-experiment rather than generate a month of pleasant vibes, four rules:

1. Baseline first, for two weeks. This is the one I’d fix before anything else. Measure without changing a thing, and write down the weekly averages as well as the daily ones — a fortnight of before turns a meaningless number into a comparison, and the week-to-week wobble tells you honestly how big an effect you’d be able to spot.
2. Pick the metric your instruments can actually resolve, not the most impressive one. For aerobic training that’s pace or power at a fixed heart rate on a fixed route. It moves in weeks, it’s directly observable, and it doesn’t need a lab. Resting heart rate moves slowly and noisily. Mitochondrial density doesn’t move on your phone at all.
3. Change one thing. I kept strength training constant, which was right. I did not control sleep, alcohol, work stress or the fact that Munich in that stretch was gorgeous and I was outside more generally — all of which push resting heart rate around, and all of which drifted in the same direction as the training.
4. Give it twelve weeks. Not because nothing happens in thirty days — plenty does — but because twelve weeks is where the changes get large enough to clear the confounders you couldn’t control. It’s roughly what our 12-week VO2 max plan allows for, and that isn’t a coincidence.

Would I keep it?

The Zone-2-only part, no. Permanently skipping hard efforts means neglecting VO2 max, which is the fitness metric most tightly bound to how long you live, and Phillips’s volunteers earned their 10% VO2 peak bump with two hours of cycling a day, five to six days a week — a volume I am not going to reproduce alongside a job. I’ve gone back to roughly four easy sessions and one hard one, which is where the evidence pointed all along.

The Zone-2-heavy part, absolutely. The ceiling rule turned out to be the useful bit: it stopped me drifting into the vaguely-uncomfortable middle intensity that feels like training and mostly just makes you tired. Four weeks of enforced slowness reset what “easy” was allowed to mean, and that’s stuck.

And the actual finding, the one I’d hand to anyone about to start a thirty-day anything: your body will change faster than your instruments can prove it did. That’s not a reason to skip measuring. It’s a reason to measure the thing you can actually see moving, take a baseline before you start, and be patient with everything else.

The man with the dog is still quicker than me on the hills. Some things a month won’t fix.