[drpossum]

No, because an untestable theory is just a useless theory (i.e. it is not meaningfully predictive). There are more difficult and confrontable issues such as predictions models will make like "supersymmetry", "the proton decays", and "early universe signatures in CMB" which could be, but have not been, observed.

You can sometimes make models which can put these under various parts of the proverbial rug and, sure, claim we just need to build the next generation of experiment, but at that point what are you really doing? I suspect this is what "worse than you think" is trying to get at.

[cainxinth]

Many testable theories started as untestable theories, including some of the most famous theories of all like evolution and atomic theory.

[esperent]

This is a valid point. Some other theories that were originally thought of as radical and untestable were plate tectonics and germ theory.

On the other hand, I think the reason why this gets brought up so often for string theory is the degree to which it appears untestable. Multiple curled up dimensions and something like 10^500 possible versions of the theory, and with everything operating below the Planck length, which according to other widely accepted theories is the smallest possible measurement. How do you measure something smaller than the smallest measurable thing?

I think the only comparably untestable theory was atomic theory when first suggested by the Greeks 2000 years ago. Germ theory and evolution were relatively tame by comparison.