Here's another list of false beliefs. These are held by science popularizers. Whether they actually believe these beliefs, or just utter them for the purpose of getting more viewers, is an unanswerable question:
The curved space near massive object can be pictured as a deformed rubber sheet
This one is due to Einstein, unfortunately. You put balls on a rubber sheet, and you see that they roll towards each other. The reason this is a terrible explanation is because you have the Earth's gravity doing the pulling, not the curved space. The actual geodesics on a curved space like the rubber sheet are repelled by the central mass. The reason things attract in relativity is because of the time-dilation factor, and this is the dominant effect. It is just as easy to explain things correctly, in terms of time slowing down near a massive object, and world-lines trying to maximize their proper time with given fixed endpoints, but popularizers never do this.
A variable speed of light can replace inflation.
This appeared in a recent popular show, and it is based on the following bogus idea: if light moved faster at early times, then all the universe could have been in communication! The reason this is false is because no matter how the speed of light is imagined to vary, one can recoordinatize space-time in terms of the intersections of light cones, and unless these lightcones split instead of merge, you get the same communication paradox--- new regions coming into causal contact are coming into causal contact for the first time.
Mesons and Baryons are made of quarks like atoms are made of protons, neutrons and electrons.
This is insidious, because its true for heavy mesons. But it's much more false than true for pions and protons and all the excitations at lower than 1GeV, because of the vacuum condensates. There is no reasonable model of light pions which does not take into account their Goldstone nature. This type of explanation also leaves out Nambu and Skyrme, both of whom were unjustly ignored for too long.
String theory is a theory of strings
This picture is not good for someone who doesn't already have a sense of string theory, because if you start out making home-made models of relativistic strings, you will never get anything like the correct string theory. The strings you naively picture would not have the special light-cone interactions that strings do in the Mandelstam picture, and they would not obey Dolen Horn Schmidt duality. They would just be conglomerations of point particles held together by rubber bands. They would have the wrong spectrum, and they would be full of ghosts.
The only proper way to say what strings are is to say right off the bat that they are S-matrix states, and that they are designed to be an S-matrix theory with linear Regge trajectories. They have a string picture, but the constraint that exchanging strings in the S-channel is dual to exchanging them in the T-channel is all-important, just as it was all-important historically. Without this, even with the Nambu action, you are at a loss for how to incorporate interactions. It is not obvious that the interactions are by topology unless you know Dolan Horn Schmidt.
It is also important for realizing that string interactions are somewhat holistic (that they become local on the light cone is the surprise, not the other way around). You add them order by order in perturbation theory by demanding unitarity, not by asking what happens when two strings collide in the usual sense. These "strings" are strange new things born of 1960s Chew-isms, and their closest cousins are flux lines in gauge theory, or fishnet Feynman diagrams, not a collection of point masses held together by spring-like forces.
This is also insidious, because Chew, Mandelstam, Dolen, Scherk, and all that generation developed the greatest physical theory the world will ever see, and their reward was: "You're fired". (in Scherk's case, "You're crazy"). Then they were heckled for thirty years, while their work was appropriated by a new generation, who described them as the deluded misguided Chew-ites who discovered something great by accident.
There is more than a snowball's chance in hell for large extra dimensions
The idea that there are large extra dimensions was very popular in 2000, but it's completely preposterous. Large extra dimensions bring the planck mass down to about a TeV, giving neutrinos generic majorana masses which are in the KeV-MeV range, so you need to fine tune. They lead to essentially instantaneous proton decay, and huge CP violations in strong interactions, so you need to fine tune some more. To avoid proton decay, there is a clever mechanism due to Arkani-Hamed and Schmalz which puts the quarks and leptons in different places in the extra dimensions. This idea is appealing only at a superficial first glance, because it requires that the SU(2) and U(1) of the standard model be extended in the extra dimensions, which affects their running immediately. The theory predicts unambiguously and model-independently that proton decay suppression requires huge electroweak running at around a TeV. That's a signal you haven't seen a hint of at 100GeV collisions. Come on. In addition, how do you stabilize large dimensions? It's the same fine-tuning as before, so the number of problems has gone up.
A low Planck scale would completely demolish the predictivity of string theory. You can squeeze a lot of stuff into large dimensions. In my opinion, it is this brand of string theory that the critics correctly criticize as fundamentally non-predictive.