I see. Believe me, I understand. I've been waaaay down that road.
When I studied physics in college, they made sure to take us down it. The Newton/Laplace clockwork universe was made very clear to us, as were the laws of thermodynamics, and Maxwell's electromagnetism. We got very good at manipulating them, too. It all made sense--common sense--and it all worked perfectly. Or so they made us believe.
Then, in our junior year, they started to introduce us to certain uncomfortable facts. Experimental facts. Facts that couldn't be made to fit into the commonsense framework that we'd just spent two years and thousands of dollars building in our own heads. Clocks slow down when they move. Rulers get shorter when they move. Light waves arrive in tiny lumps of energy. A small bit of matter takes multiple paths to get from one place to another. Outcomes of certain events are apparently changed retroactively.
The demonstrable behavior of the universe is different from what very well-trained common sense would expect.
I've never witnessed it myself, but many professors who teach junior physics have stories of students who have become emotionally distraught at this. But it's important that every professional physicist, at least once, undergo the experience of having a perfectly coherent physical worldview smashed irretrievably by experimental fact.
The cold reality is that nature, at its core, does not conform to man's common sense. Mathematics contradicts it, experimental fact contradicts it, and the two of them agree with each other. It takes humility to yield to this, but the universe is the way that it is, and not how we would wish it to be.
From here: http://math.ucr.edu/home/baez/RelWWW/tests.html
"The Classical Tests
Anyone who has taken a first course in general relativity will be aware of the so-called "classical solar system tests":
light bending: according to gtr, the images of stars near a foreground isolated massive object such as the Sun should be displaced outwards from their usual positions by an amount first computed by Einstein in 1915. This effect could be consider the simplest kind of gravitational lensing effect. As is well known, a team of astronomers, including Arthur Stanley Eddington, confirmed Einstein's prediction that a stellar image which according to Newtonian gravitation would just graze the limb of the Sun would be displaced outwards by 1.75 seconds of arc during the solar eclipse of 1919. Many further observations of other exclipses have since been made, but these observations were too uncertain to provide very accurate tests. However, the effect should be the same for any frequency of EM radiation, including radio waves, and in recent years the Very Long Baseline Array Interferometer (VBLI) array has been used to confirm the prediction of gtr to within 0.1%, Most recently, the Hipparchos star mapping mission confirmed the effect for occultations by the major planets such as Jupiter, as well as our Sun.
Shapiro time delay: according to gtr, the travel time of a radar signal should be longer if the radar beam travels near an isolated massive object such as the Sun, by an amount first computed by Shapiro. Radar ranging experiments using the Viking and Voyager spacecraft (whose position is known with great accuracy) have confirmed this effect to within about 1% of the value predicted by gtr, and observations of the Pulsar PSR 1937+21 have provided an independent verification that this effect works as advertised (to within about 5% of the value predicted by gtr) elsewhere in the Universe, not just in our own solar system!
Time delay and light bending observations are actually measuring the same PPN parameter, gamma. The percentage accuracies quoted above are the percentage to which gamma conforms to the value it has according to gtr; this is far more physically relevant than reporting the percentage accuracy of individual predictions, because it shows that if gtr is not the correct classical field theory limit of the presumed quantum theory of gravitation, then this limit must be a good mimic of gtr, at least in so far as the amount of spacetime curvature caused by a given amount of mass-energy is concerned. Here is a graph, taken from this paper by Clifford Will, of various measurements of gamma.
gravitational red shift: according to gtr, static clocks held close to an isolated massive object run more slowly than static clocks which are further away, in the sense that radio signals and other EM radiation "climbing away" from a massive object are red-shifted, as recieved by a distant static observer, by an amount first computed by Einstein as early as 1913. Today, this effect would be considered a test of principle of local position invariance, and thus a test of a large class of "metric gravitation theories", not just a test of gtr. The effect has been verified by progressively more accurate experiments since the classic experiment of Pound and Rebka in 1961, and the best current experimental results confirm the effect to within about 0.02% of the value predicted by assuming LPI.
In addition, careful observations of a very rapidly rotating pulsar (neutron star) have confirmed to within a few percent. that this effect also works as advertised elsewhere in the universe. Futhermore, the ASCA X-ray satellite has observed very strongly gravitationally redshifted light emitted from the inner edge of the accretion disk a supermassive black hole in the galaxy MCG-6-30-15. This light appears to come from r = 10m to r = 3m; note that the inner edge of the accretion disk is located at r = 6m, so these observations appear to confirm that very hot material is leaving the disk and plunging into the event horizon, effectively disappearing at about r= 3m, just as gtr predicts!
Last but not least, the clocks aboard the Global Positioning System satellites in effect test every day both the gravitational red shift effect and the kinematical red shift predicted by str to within 50 nanoseconds per day (the str effects and gtr effects account for a comparable amount of the rate by which the satellite clocks differ from the rate of comparable clocks on the Earth's surface), and if either of those predictions were not correct, the clocks would drift by 40 microseconds per day, which would immediately render the GPS system completely useless!
Recall from the above that gravitational redshift is actually a measure of the post-PPN parameter alpha; here again is the graph from the paper by Will summarizing various measurements of alpha.
precession of periastrion of objects such as a planet orbiting a more massive object such as a star: according to gtr, non-circular orbits of planets around a star (or of satellites around the Earth) are not exact Keplerian ellipses, but are more like ellipses which slowly "rotate" so that their point of closest approach (periastrion) slowly "precesses" in the direction of motion of the orbiting object. The precession predicted by gtr was first computed by Einstein in 1915, who found that it precisely accounted for an unexplained "extra-Newtonian" precession of the planet Mercury in its orbit around the sun, in the amount of about 43 seconds of arc per century. (The total precession is much larger, about 5600 seconds per century, but almost all of this is explained by perturbations of the orbit of Mercury due to Jupiter and other Newtonian astrodynamical effects. Because in the limit of weak fields and slow motion (the motion of Mercury around the Sun falls into this category), gtr behaves pretty much like Newtonian gravitation, the previously known explanations for all but the ``extra-Newtonian'' 43 seconds also apply in gtr, but gtr's first great success was that it also explained the "extra" precession which could not be explained by Newtonian astrodynamics. Since 1915, progressively more accurate observations have tested the effect using the orbits of all the inner planets (specifically: the ``extra-Newtonian'' precession of Venus is observed to be 8 seconds of arc per century, and the ``extra-Newtonian'' precession of Earth is observed to be 5 seconds of arc per century, both also in perfect agreement with the prediction of gtr), the orbits of various asteroids, and the orbits of various spacecraft.
The PPN parameter which controls this precession is betabar; the above mentioned observations show that betabar agrees to 0.1% with the gtr value. The main source of uncertainty in these measurements is the value of the quadrupole moment of the Sun; helioseismology suggests that this moment is less than 10^(-7), but if it were larger, the estimated value of betabar might be affected.
In addition to the above mentioned classical solar-system tests, since 1969, when the Apollo astronauts placed a laser reflector on the surface of the Moon, the Earth-Moon distance has been tracked continuously with an accuracy of 15 cm (!). This LURE data shows that the PPN parameter alpha1 is within 0.1% of the gtr value, and that the PPN parameter zeta3 is within 10^(-8) of the gtr value. According to gtr, the spin axis of a spinning gyroscope in orbit about a massive object should precess; this effect was first noticed by de Sitter, and is called geodetic precession (if the massive object is itself spinning, there is an additional, much smaller, gravitomagnetic precession of the spin axis). The LURE data has confirmed this prediction to within 0.7 % of the predicted value for the Earth-Moon system, considered as a ``gryoscope'' in orbit about the Sun. In addition, measurements of the Sun's alignment with the ecliptic shows that the PPN parameter alpha2 is within 10^(-6) of the gtr value, and measurments of Earth tides show that the PPN parameter xi is within 0.1% of the gtr value.
As we'll see below, the remaining PPN parameters are constrained by strong-field measurements coming from observations of binary pulsars; at present, all the PPN parameters are known to agree with the gtr values to between 10^(-3) and 10^(-20) (!). Actually, these figures do not by themselves do justice to the actual known contraints upon the PPN parameters. In fact, when we combine deductions from several different types of measurements, we find not only that the resulting estimates are consistent with one another, but that they ``trap'' the space of metric theories which are in agreement with all known data to within a very small neighborhood of the gtr PPN values. For example, here is a graph, taken from this paper by Esposito-Farese, showing the constraints on the PPN parameters gamma = 1 + gammabar and beta = 1 + betabar, given by combining
- the observed ``extra-Newtonian'' precession of the perihelia of Mercury,
- tracking of the Earth-Mars distance, using a radar reflector placed on Mars as part of the Viking mission,
- Very Long Baseline Array (VLBI) interferometry measurements "