Holiday Message | Hits and Misses
What this covers
Sean Carroll walks through a selection of his own physics papers—some highly cited, others nearly ignored—to examine what makes a paper a scientific "hit" or "miss." The episode pairs technical depth with reflection on the process of doing research: how ideas form, why some land broadly while others languish, and what that tells us about how science actually advances. Carroll draws on concrete examples, from early work on Lorentz invariance violations and closed timelike curves to later investigations of dark energy, space from Hilbert space, and black hole information.
The argument hinges on a distinction that may surprise working physicists: effort and impact are decoupled. A paper written over a weekend can become foundational; years of labor can produce work nobody cites. What matters instead is opening a new playground—giving the field a new toy to work with, whether a quintessence model, a modification of gravity, or a fresh angle on black hole thermodynamics. Misses, Carroll contends, usually reflect community taste, snobbery, or timing rather than merit. Even strong papers on de Sitter space compactification or firewalls failed to catch on partly because they fell outside the prevailing bandwagon. Along the way he discusses collaborators like Alan Guth and Eddie Farhi, explains why Citation Count as Quality Proxy is genuinely useful despite its flaws, and reflects on lessons learned about balancing quirky personal interests with work that moves the discipline forward.
Carroll argues that the scientific impact of a physics paper (its citations) is loosely but genuinely correlated with quality, yet largely decoupled from effort, and that the most reliable route to a 'hit' is opening a new playground of solvable problems for others, while 'misses' often reflect taste and bandwagon dynamics rather than the merit of the work.
- Papers that give the field a new 'toy' to play with (Lorentz violation, modified gravity) get heavily cited because they generate follow-on work.
- Effort and impact are uncorrelated: highly-cited papers were sometimes written over a weekend on a plane.
- Misses (time machines, de Sitter compactification, firewalls) often failed to land because of community taste, snobbery, or being orthogonal to the prevailing bandwagon, not because they were wrong.
Opening new theoretical possibilities for exploration drives high-impact physics papers.
- The most reliable way to produce a high-impact ('hit') physics paper is to open up a new set of possibilities — to give theoretical physicists a new toy or playground to play in — because the resulting flood of follow-on work generating constraints, variations, and refinements drives citations, as demonstrated by Carroll's Lorentz-violation and modified-gravity papers.
“So give people a new playground to play in. That is the lesson for getting songs on your hit record there.”
- When you produce a paper you are proud of but the community does not pay it the attention you hoped, especially when your idea is orthogonal to what colleagues expect a solution to look like, the correct response is to work harder — write follow-up papers, develop the ideas further, demonstrate they solve the problems you care about, and advertise them widely — because suggesting an unconventional angle requires extra effort to convince people you are on the right track.
“Your job is to work harder now, write a follow-up paper, write other papers, develop the ideas further, show that they really do solve the problems that you care about, figure out what the implications are... and advertise it widely.”
- While it is important to do work that has impact and is considered interesting by the field, it is not important to chase research bandwagons simply for the sake of being on the bandwagon; Carroll deliberately pursued the emergence of spacetime from quantum information in the de Sitter / real-world context rather than the dominant AdS/CFT setting everyone else worked in.
“it's important to do things that have an impact and are considered to be interesting by the rest of the field, but it's not important to chase bandwagons simply for the sake of being on the bandwagon.”
- Carroll was randomly assigned George Field as a PhD advisor in the Harvard astronomy department, which proved enormously consequential because Field — though a theoretical astrophysicist specializing in the interstellar medium — had recently taken up particle-physics cosmology and was the closest available match to Carroll's theoretical interests.
“I got lucky because it was a completely random assignment of advisor and student.”
- A scientist should spend some time writing quirky papers they personally find cute, but it cannot be the only thing they do; to sustain a career and have legitimate scientific impact one must also write papers that other people care about and that actually affect the field, which Carroll learned the hard way after nearly failing on the job market by working only on idiosyncratic problems.
“just writing my own little papers that I thought were cute, but no one else cared about was not really the only thing I should be doing. You should spend some of your time doing that, but it can't be the only thing.”
Citation counts, while imperfect, reliably correlate with genuine scientific quality and impact.
- The number of citations a scientific paper receives is positively but imperfectly correlated with its quality and importance; there are highly cited papers that are nonsense and uncited papers that are great, but having a positive impact on the field is genuinely correlated with being a good paper, which makes citation counts a useful first-pass check on whether someone is a serious scientist.
“Having a positive impact on the rest of the field is correlated with being a good paper. And it's easy to check by anyone by going online.”
- There is no correlation between how much work goes into a paper and how good or important it turns out to be; a paper that required years of solid effort may end up uninteresting, while a single flash of insight written up over a weekend can become highly cited.
“As far as I know there's no correlation between how much work you do on a paper and how good it is or how important it is.”
- Who counts as an author on a physics paper is fluid and depends on how conversations evolve: Ken Olum's contributions to the time machine work progressed from meriting a simple acknowledgement, to a strong acknowledgement, to full co-authorship, illustrating that the threshold for authorship is determined by the substance of a collaborator's input rather than fixed at the project's outset.
“Ken's contributions went from meriting an acknowledgement in the paper to meriting a very strong acknowledgement in the paper to being a co-author on the paper.”
Physicists habitually reason classically despite learning quantum mechanics, failing to take wave-function reality seriously.
- Despite learning quantum mechanics, physicists habitually think classically — treating the world as classical physics plus occasional jumps — and routinely turn quantum wave functions into realizations of stochastic processes when reasoning about black hole information or eternal inflation, talking about other parts of the multiverse as if their wave functions have already collapsed; taking the quantumness of unobserved parts of the universe seriously is something the field has not yet properly done.
“we physicists, despite the fact that we learn quantum mechanics, we always think of the world classically at the end of the day. We think of the world as classical physics plus some occasional jumps”
- Conventional quantum gravity quantizes a pre-existing classical theory (general relativity or strings traveling through spacetime), but taking many-worlds seriously implies you should instead start with the vector in Hilbert space — the wave function — as fundamental and derive the classical image of spacetime from it, rather than building space and time in by hand.
“You should be starting with the vector in Hilbert space, the wave function, the quantum thing, and you should be deriving a classical image, whatever that image is, from that quantum thing.”
- In quantum field theory the vacuum state has high entanglement between nearby degrees of freedom and low entanglement between distant ones, establishing a relationship between entanglement and spatial geometry; reversing this, Carroll, Cao and Michalakis treated the amount of entanglement between abstract quantum degrees of freedom as a measure of distance and showed that, in the weak-gravity linear regime, the resulting emergent geometry obeys Einstein's equation — without ever starting from a metric.
“rather than saying, here's the distance, I can calculate the entanglement. Let's say, here's the entanglement. Let me assert that I can treat that entanglement as a measure of distance.”
Dark matter remains the simplest, most robust framework despite observational discrepancies traceable to ordinary matter complications.
- Although Carroll was sympathetic in 2004 to replacing dark matter by modifying gravity, twenty years later he is much more skeptical because the idea has essentially been ruled out by the data; as a methodological stance, once ideas are ruled out by data he loses interest in them even if others continue holding on.
“Today, 20 years later, I'm much more down on that idea because it's basically been ruled out by the data. I know that people are hanging on to it, but once ideas are ruled out by the data, I lose interest in them.”
- Both the regime where dark matter becomes necessary in spiral galaxies and the regime where cosmic acceleration sets in share the common feature that gravity is weak — the relevant acceleration scale in galaxies is numerically similar to the Hubble constant and the cosmological constant — which motivated trying to modify general relativity so that it behaves normally when spacetime is curved but deviates when spacetime becomes nearly flat.
“what is common between these two features... And the answer is that gravity is weak, right?”
- It is always good to base your starting point on the simplest models that are robust and work — such as the cold dark matter model in which particles are created cold and barely interact — because the simplest robust framework basically works even if it does not fit the data exactly, and discrepancies often trace to the complications of ordinary matter rather than to exotic dark matter.
“it is always good to base your starting point on the simplest models that are robust and work, right?”
Closed timelike curves are mathematically tractable solutions unfairly stigmatized by 'time machine' framing in physics culture.
- The spacetime around a single infinitely long, perfectly straight cosmic string is perfectly flat everywhere except exactly at the string itself, where there is a conical singularity; the string's energy produces no local gravitational pull but creates a deficit angle, so a gyroscope carried around the string fails to return to its original orientation because it senses the total curvature enclosed by its loop.
“the space time outside the cosmic string is perfectly flat. There's no gravitational field.”
- Lie groups of continuous transformations are themselves manifolds with a natural metric, and the geometry of the 2+1 dimensional Lorentz group SO(2,1) happens to be exactly that of three-dimensional anti-de Sitter space, an insight that allowed the time machine non-creation result to be proved cleanly using only diagrams of trajectories in anti-de Sitter space rather than heavy equations.
“The Lorentz group in three dimensions, the group of Lorentz boosts and rotations and so forth, has a geometry that happens to be exactly that of three dimensional Anti-de Sitter space.”
- Work on closed timelike curves was looked down upon by many physicists purely because of its 'time machine' framing — Carroll was personally asked at a postdoc interview what Feynman would think, and Kip Thorne noted his NSF grant could not fund time-machine work — even though identical content labeled as 'classifying violations of global hyperbolicity in 3D Einstein solutions' would have drawn no objection.
“if we had called it classifying the space of violations of global hyperbolicity in three dimensional solutions to Einstein's field equations, no one would've objected, even though it was exactly the same content to the paper.”
- Because perfectly straight parallel cosmic strings have no dynamics along their length, the theory of such strings in 3+1 dimensions is mathematically equivalent to the theory of point particles in 2+1 dimensional gravity, allowing the Gott time machine problem to be analyzed as moving point particles with conical gravitational fields in a plane.
“the theory of perfectly straight parallel cosmic strings in three plus one dimensions is the same as the theory of point particles in two plus one dimensional gravity.”
- If you start in a 2+1 dimensional spacetime with point particles that do not already contain a Gott time machine — moving too slowly relative to each other — then you cannot accelerate them fast enough to create one, because doing so while conserving energy-momentum is impossible given a timelike total momentum; thus closed timelike curves of the Gott type cannot be created from initial conditions that lack them.
“if the particles aren't already moving fast enough, you can't get them to move fast enough to make a Gott time machine.”
- Huw Price's double standard principle holds that whatever criteria of naturalness or simplicity you apply to the universe's initial conditions you must equally apply to its final conditions; cosmologists routinely cheat by imposing low-entropy conditions on the early universe while declining to impose them on the late universe, which is only justified if time has an intrinsic direction — something the time-symmetric laws of physics and most cosmologists themselves deny.
“Price's double standard principle is, if you think something is natural for the early universe, you should think it's natural for the late universe, okay? For the beginning and for the end.”
- Instead of a constant vector field (which is ill-defined in curved spacetime), a slowly-changing scalar field whose gradient is a vector can violate Lorentz invariance and serve as dark energy; if it is a pseudo-scalar with a symmetry that keeps its mass low and forbids couplings to other fields, it can naturally couple only to electromagnetism, producing observable cosmic birefringence of about one degree — below existing limits but reachable by improved data such as the cosmic microwave background.
“if you wanted to have a scalar field that could be the dark energy, and that scalar field has a natural symmetry that both keeps its mass low and prevents it from coupling to other fields, then there's a natural way to do it.”
- Any modification of general relativity that does weird things with the curvature scalar (such as adding a 1/R term to the Lagrangian) generically brings a new dynamical degree of freedom to life; specifically, R + 1/R gravity is equivalent to a scalar-tensor theory with the metric plus a scalar field, and in such theories the ordinary Schwarzschild solution still holds, so no new effects appear far from gravitating objects.
“Whenever you have a theory of gravity that is some modification of general relativity just by doing weird things with the curvature scalar, you basically bring a new degree of freedom to life.”
The universe's low initial entropy is explained by de Sitter space spawning inflating baby universes via quantum fluctuations.
- Rather than asking why the Big Bang had its special low-entropy properties, the right question is why we do not live in empty de Sitter space, because that empty accelerating universe is the genuinely high-entropy configuration; Carroll and Chen proposed that empty de Sitter space can spawn baby universes via quantum fluctuations that look like inflating big bangs, a process that runs in both time directions and thereby explains low initial entropy without fine-tuning or cheating.
“Jenny and I started asking, why don't we live in De Sitter space? Why don't we live in an empty universe? Why is there a Big Bang at all?”
- Time's arrow exists because entropy increases, and the explanation has two parts: Boltzmann's definition that entropy is the logarithm of the number of microstates within a macrostate, and the past hypothesis — the crucial initial condition that the universe began in a low-entropy microstate — which together account for the second law of thermodynamics.
“the entropy is the logarithm of the number of ways you can arrange the system so that it looks macroscopically the same... The other half the answer, why is entropy increasing, is because it started in a low entropy microstate.”
- Contrary to the usual assumption that extra dimensions start small and three become large, Carroll, Johnson and Randall showed that starting from a large higher-dimensional de Sitter space with only an electromagnetic field, quantum fluctuations can dynamically produce localized tube-like regions that are macroscopic in N-2 dimensions while curling the remaining two into a small sphere — compactifying ordinary dimensions rather than uncompactifying small ones.
“Rather than starting with them all small and asking, can you make some big, if you start with all of them big, can you make some small?”
Black hole information paradox resolves through many-worlds entanglement accounting across different wave-function branches.
- Whether two subsystems are entangled can have different answers depending on whether you ask about the global wave function or about a single branch; the early-late photon entanglement is a global-wave-function (information-conservation) question while horizon smoothness is a branch-level (measurement-outcome) question, so the two required entanglements are not in conflict, and there may be enough branches of the black hole wave function that no individual branch — and hence no infalling observer — encounters a firewall.
“whether or not two subsystems of the universe are entangled with each other can have a different answer depending on whether you're saying, do you mean in the global wave function of the universe or do you just mean on a branch?”
- The AMPS firewall paradox arises because an outgoing late Hawking photon must be entangled with an early outgoing photon (to conserve information) yet also maximally entangled with an infalling photon (for the event horizon to look like empty space), but the monogamy of entanglement forbids a maximally-entangled system from also being entangled with a third — forcing the conclusion that the horizon is not empty space but a high-energy firewall.
“If you have an outgoing photon that is maximally entangled with an ingoing photon, which is what you need for everything to look like empty space... then that outgoing photon cannot also be entangled with anyone else.”
Dynamical dark energy faces theoretical difficulties that scalar-field models cannot overcome.
- A quintessence scalar field proposed as dynamical dark energy must have a mass around 10^-33 electron volts — vastly smaller than the electron (~half a million eV) or the lightest neutrinos (~10^-3 eV) — and in quantum field theory such tiny masses normally require a protecting symmetry, so simply writing one down to fit the data, without explaining the smallness or why the field has not already been detected through its couplings, is theoretically unsatisfying.
“The mass of the Scalar field that you need to be quintessence is something like 10 to the minus 33 electron volts.”
- The cosmological constant is the simplest, most robust candidate for dark energy, but even if dark energy turns out to be something dynamical instead, the puzzle of why the cosmological constant is so small remains unsolved either way — you cannot escape that problem by adopting a dynamical model, and early hopes that scalar-field dynamics could explain the smallness essentially never work.
“even if you have a non cosmological constant candidate for what the dark energy is, it is still a puzzle why the cosmological constant is small. That's just a puzzle one way or the other.”
- A massless copy of electromagnetism coupling only to dark matter (with a dark photon and dark charges) need not have an extremely weak coupling, because the dynamical relevance of electromagnetism depends not only on coupling strength but on particle mass — heavy particles accelerate and radiate less — so if the dark matter particle is heavy (around a trillion electron volts) the dark electromagnetic interaction can be of ordinary strength and still escape all observational constraints.
“if you did have dark electromagnetism, it wouldn't have to be that weak. Because there's a trade off.”
- A constant background vector field filling spacetime and coupled to electromagnetism (a Chern-Simons-like Lorentz-violating term) would rotate the polarization plane of photons traveling cosmological distances — cosmic birefringence — which can be tested observationally by comparing the known polarization angle of distant radio galaxy jets against their observed polarization, yielding tight limits on the violation.
“it would cause what we now call cosmic birefringence... a photon interacting with this vector field, stretching throughout all the universe would rotate its polarization.”