Top 100 Unsolved Mysteries of the Universe, Episode 36: The CMB Lensing Anomaly and the A Posteriori Tension Problem. Picture yourself looking through an old pane of window glass at a poster covered with fine print. The glass is not broken, just slightly wavy. Sharp corners get rounded, straight lines bend a little, and the fine ridges on the poster lose some of their crispness. If you only stare at the final image, you may think the poster itself was printed badly. But the real culprit may be the sheet of glass in the middle. That is the basic image behind CMB lensing. The microwave background is like a plate from the early universe. Later clusters, filaments, and large-scale structure are like the uneven glass along the line of sight. They gently stretch the plate, remap it, smooth the peaks and troughs, and even twist part of the cleaner polarization pattern into a little extra swirl.
The awkward part is what happens when cosmologists try to read that distortion. Inside the CMB temperature and polarization power spectra, the fit has often shown a mild preference for a bit more lensing smoothing than the baseline model expects, as if the cosmic glass were slightly better at blurring the image than it should be. But when that preference is placed beside CMB lensing reconstruction, weak lensing, redshift-space distortions, and structure-growth probes, the story does not always close cleanly. So the real puzzle is not whether CMB lensing exists. It does. The puzzle is why the same cosmic structures seem to report slightly different lensing strength depending on which reading chain we use.
This is where mainstream cosmology gets trapped in a gray zone. The signal does not look solid enough to count as clean new physics, but it also does not sit comfortably as mere noise. In practice, the internal CMB fit often compresses several effects into one convenient knob: how much the acoustic peaks are rounded, how much E-mode polarization leaks into B-like swirl, how much the damping tail is smoothed, how much the whole image looks softly blurred. But once everything is squeezed into one knob, beam effects, foreground cleaning, absolute calibration, covariance choices, parameter degeneracies, and even the habit of noticing a tension first and adding freedom later can all get folded into the same number. It becomes like using one dial to control sharpness, contrast, focus, and glass distortion at once. The picture may look tidier after you turn it, but you no longer know whether the universe changed, the pipeline changed, or the knob merely absorbed several mismatches into one cosmetic fix.
EFT rewrites the question before it rewrites the answer. Lensing should not first be treated as an inventory photograph of invisible matter. It should first be read as a path problem: how does the foreground base map rewrite the background plate? In that picture, the CMB is not a warehouse photo from which we directly count cosmic stock. It is an early-universe plate. What we read today is that plate after it crossed a later network of structure that pressed, stretched, and re-engraved it on the way here. Once you read the problem that way, a stronger-than-expected lensing preference no longer has to mean that the universe is hiding one more unseen lump. It can instead mean that a single-window fit has over-compressed line-of-sight environment, path embossing, late-time image shaping, and statistical uplift into one overly convenient lensing dial.
That is why EFT demands a harder audit, not an easier one. The real task is not to keep arguing over where one posterior parameter should land. The real task is to put CMB lensing, weak lensing, strong-lensing time delays, distance residuals, and structure formation back onto one common foreground base map and ask a tougher question: can the same map explain the dynamics ledger, the imaging ledger, and the growth ledger at the same time? If it can, much of the anomaly looks more like a compression-language problem than a new cosmic substance. If it cannot, the failure points back to the base map itself, telling us that our reconstruction of the line-of-sight environment is still wrong.
That is also where EFT places its guardrails. First, EFT is not denying real lensing. It fully accepts that later structure genuinely remaps the background plate. What it rejects is the habit of promoting one single-window posterior preference directly into the universe’s final account book. Second, EFT is not saying every tension is just systematics in disguise. Its demand is stricter than that. If a feature survives cross-window bookkeeping on one common base map, keep it. If it changes the moment you swap windows, change the compression scheme, or alter the fitting language, then it looks less like a new law written by the universe and more like a mixed-up number hiding inside an effective smoothing knob. In EFT, the CMB lensing anomaly and the a posteriori tension problem are best treated not as a premature coronation for a new particle, and not as a free pardon for every systematic, but as a test strip for whether we have been mixing up the foreground base map and the background plate all along. Tap the playlist for more. Next episode: The Precision-Physics Problem of Recombination History. Follow and share—our new-physics explainer series will help you see the whole universe more clearly.
EFT
Language