Claimed breakthroughs can sometimes be artifacts of oversimplified statistics
For a quarter century, the standard model of cosmic expansion has held together remarkably well — until one stubborn number refused to cooperate. The Hubble tension, a five-to-seven standard deviation disagreement between how fast the early and late universe appear to be expanding, has resisted every proposed remedy, including a provocative new model imagining dark energy once ran in reverse. A rigorous statistical analysis published in Physical Review D now confirms that this sign-flipping model edges predictions in the right direction but does not close the gap — and warns that some earlier claims of progress were likely mirages born of oversimplified mathematics. The universe, it seems, is not yet ready to give up its deepest secret.
- A five-to-seven sigma disagreement between two independent ways of measuring cosmic expansion is too large to be noise — it signals something genuinely missing from our best theory of the universe.
- The LsCDM model, which imagines dark energy flipping from a pulling force to a pushing one early in cosmic history, raised hopes by nudging predictions closer to locally observed expansion rates.
- Researchers at Plaksha University and the Federal University of Latin American Integration applied non-Gaussian statistical tests — more honest to the messiness of real data — and found the apparent improvement largely evaporates under scrutiny.
- Early-universe data from Planck, the Atacama Cosmology Telescope, and galaxy surveys hold together well, but the moment nearby supernova measurements enter the picture, both LCDM and LsCDM show unresolved strain.
- The study lands as a methodological warning: statistical shortcuts that assume tidy bell curves can manufacture the appearance of breakthroughs where none exist, raising the bar for what counts as a genuine solution.
For nearly a century we have known the universe is expanding, and since the late 1990s we have known that expansion is accelerating — a discovery so consequential it earned Saul Perlmutter, Brian Schmidt, and Adam Riess the 2011 Nobel Prize in Physics. The explanation, dark energy represented by a cosmological constant Lambda, anchored the LCDM model that has guided cosmology ever since. LCDM explains the ancient light of the cosmic microwave background, the clustering of galaxies, and the brightness of distant supernovae with remarkable precision. But one problem has refused to yield: the Hubble tension.
The Hubble constant measures how fast the universe is expanding right now. Two independent methods of calculating it — one working backward from early-universe physics, the other measuring nearby supernovae calibrated against pulsating Cepheid stars — disagree by five to seven standard deviations. That gap is far too wide to blame on measurement error. Something in our understanding appears to be fundamentally incomplete.
Among the many proposed fixes, the LsCDM model attracted serious attention. It keeps LCDM largely intact but adds a striking twist: dark energy once had the opposite sign, pulling matter together in the early universe before flipping to its familiar repulsive behavior when the cosmos was less than a third of its current age. Earlier studies suggested this single change could ease both the Hubble tension and a related disagreement called the S8 tension.
A new analysis in Physical Review D, led by doctoral student Sehjal Khandelwal alongside professors Abraão Jessé Capistrano de Souza and Suresh Kumar, subjected that claim to a harder test. Standard tension measurements treat all data as if they follow a simple bell curve — a shortcut that can badly misrepresent disagreements when datasets differ wildly in precision and shape. The team instead applied non-Gaussian statistical methods and a posterior predictive check, drawing on data from Planck, the Atacama Cosmology Telescope, the South Pole Telescope, the Dark Energy Spectroscopic Instrument, and the Pantheon Plus supernova catalog.
The results split cleanly along a fault line. Early-universe data and galaxy clustering agreed well under both models — the cosmological foundation held firm. But adding nearby supernova measurements exposed persistent, unresolved tension in both LCDM and LsCDM alike. The sign-flipping model did shift predictions toward locally measured expansion rates, but the observed Hubble constant still landed in an improbably unlikely corner of its predictions. Progress, not resolution.
The broader lesson is methodological as much as physical. When rigorous statistics replace convenient shortcuts, some celebrated breakthroughs dissolve into artifacts of oversimplified analysis. The Hubble tension endures, and closing it entirely will demand better data, sharper theory, or both.
For nearly a century, we have known the universe is expanding. Then, in the late 1990s, two independent teams of astronomers made a discovery that upended everything: the expansion itself is accelerating. Saul Perlmutter led one team, the Supernova Cosmology Project. Brian Schmidt and Adam Riess led the other, the High-Z Supernova Search Team. Their work was so consequential that it earned them the 2011 Nobel Prize in Physics. The explanation they settled on was dark energy—a mysterious force, usually represented by a constant called Lambda, that pushes space itself apart. Paired with cold dark matter, this framework became known as LCDM, the standard model of the cosmos for the past quarter century.
LCDM is remarkably good at what it does. It explains the cosmic microwave background, the ancient light left over from the Big Bang. It accounts for the way galaxies cluster across space. It predicts the brightness of Type Ia supernovae, those exploding stars we use as cosmic measuring sticks. But there is one problem that refuses to go away: the Hubble tension.
The Hubble constant, written as H0, is simply a number that tells us how fast the universe is expanding right now. Cosmologists have two main ways to measure it. The first method uses the cosmic microwave background—essentially working backward from the physics of the early universe to predict what the expansion rate should be today. The second method is more direct: measure nearby supernovae and calibrate them against pulsating stars called Cepheids. These two approaches should agree. They do not. They disagree by five to seven standard deviations, a gap so large that it cannot be dismissed as measurement error or coincidence. Something fundamental in our understanding of the universe appears to be missing.
This persistent mismatch has sparked a decade of proposed solutions. Some physicists have imagined new particles. Others have suggested that gravity itself might work differently than Einstein described. One popular family of ideas proposes that dark energy is not constant at all—that it changes as the universe ages. A model called LsCDM takes this idea seriously. It keeps most of LCDM intact but adds a radical twist: dark energy once had the opposite sign. Early in cosmic history, it pulled matter together, almost like ordinary gravity. Then, when the universe was less than one-third of its current age, it flipped. After that moment, it began pushing space apart as we observe it doing today. Earlier studies claimed this single change could ease both the Hubble tension and a related problem called the S8 tension, without damaging the model's ability to explain the early universe.
But a new analysis, published in Physical Review D and conducted by researchers at Plaksha University and the Federal University of Latin American Integration, asked a harder question: Does easing tension actually mean the disagreement goes away? The team, led by doctoral student Sehjal Khandelwal and professors Abraão Jessé Capistrano de Souza and Suresh Kumar, realized that cosmologists typically measure tension using a statistical shortcut—treating all measurements as if they follow a simple bell curve. But real cosmological data are messier than that. When one dataset is extremely precise and another is loose and lopsided, this simplified approach can badly overstate or understate how serious a disagreement truly is.
To test this properly, the researchers gathered the latest data from multiple sources: the Planck satellite's observations of the cosmic microwave background, measurements from the Atacama Cosmology Telescope and the South Pole Telescope, galaxy-clustering data from the Dark Energy Spectroscopic Instrument, and the Pantheon Plus supernova catalog calibrated with the SH0ES project's local Cepheid measurements. They then ran both LCDM and LsCDM through rigorous statistical tests—not just the standard rule of thumb, but also more sophisticated techniques that do not assume bell-curve distributions, plus a check called posterior predictive testing that asks whether the model's predictions match what was actually observed.
The results told two very different stories. When the researchers looked only at the cosmic microwave background and galaxy-clustering data, the news was reassuring: both models performed well, and the two datasets agreed with each other remarkably well. The foundation of early-universe cosmology appeared solid. But when they added measurements from nearby supernovae, the picture changed. Both LCDM and LsCDM showed real, unresolved tensions. The sign-flipping model did shift predictions in the right direction, moving them closer to the locally measured expansion rate. Yet the observed Hubble constant still fell in a surprisingly unlikely region of the model's predictions. It was progress, but not resolution.
The finding carries a broader lesson. Claimed breakthroughs in resolving cosmological tensions can sometimes be nothing more than artifacts of oversimplified statistics, not genuine physical insight. When rigorous, non-Gaussian methods are applied, LCDM earns a real but partial victory. The Hubble tension, however, remains stubbornly unresolved. Cracking it completely will require better data, sharper theory, or both.
Notable Quotes
Claimed breakthroughs in resolving cosmological tensions can sometimes be just an artifact of oversimplified statistics, not real physical insight.— Research team (Khandelwal, Capistrano de Souza, Kumar)
The Hearth Conversation Another angle on the story
Why does it matter whether we use a bell curve or a more rigorous statistical method? Isn't the physics the same either way?
The physics is the same, but how we measure disagreement between datasets is not. A simplified statistical approach can make a real problem look smaller than it is, or make a minor discrepancy look catastrophic. We need to see the actual shape of the data, not force it into a shape that's convenient for calculation.
So LsCDM—the model where dark energy flips sign—doesn't actually solve the Hubble tension?
It moves in the right direction. The predictions shift closer to what we measure locally. But it doesn't eliminate the fundamental disagreement. The observed expansion rate still falls in an unlikely region of what the model predicts.
What does that tell us about dark energy itself? Is it really changing over time?
We don't know yet. This analysis doesn't rule out sign-flipping dark energy, but it also doesn't confirm it. What it does confirm is that we can't use sloppy statistics to claim victory. The tension is real, and it's still there.
If both LCDM and LsCDM have problems, what's the next step?
Better observations and better theory. We need more precise measurements of nearby supernovae and their calibration stars. We may also need to rethink fundamental assumptions about how the universe works. The tension is telling us something is missing from our picture.