Light emitted from a source bends around intermediate mass usually called the deflector according to Einstein's Theory of General Relativity. Not all the light emitted from the source reached the observer, only that light which bends through the correct angle. The amount of deflection caused by huge deflectors, like clusters of galaxies, is at most 10's of arcseconds. So lensing candidates require a great deal of coincidental alignment to start with. But with the huge number of galaxies and clusters of galaxies out there, there are dozens of confirmed examples.

The most extreme bending of light, strong lensing, is when the lens is very massive and the source is close enough to it: in this case light can take different paths to the observer and more than one image of the source will appear. A multiple image, shown below, is the first example of a double image was found in 1979, of a quasar. The number of lenses discovered has been used to estimate the volume of space back to the sources. This volume depends strongly on cosmological parameters, in particular the cosmological constant. If the source varies with time, the multiple images will vary with time as well. However, the light doesn't travel the same distance to each image, due to the bending of space. So there will be time delays for the changes in the images. These time delays can be used to calculate the Hubble constant. A few systems with these time delays have been found and are under study. Much of the subtlety in this work lies with constructing the model of the mass distribution forming the lens.

For weak lensing, in many cases the lens is not strong enough to form multiple images or arcs. However, the source can still be distorted: both stretched (shear) and magnified (convergence). If all sources were well known in size and shape, one could just use the shear and convergence to deduce the properties of the lens. However, usually one does not know the intrinsic properties of the sources, but has information about the average properties. The statistics of the sources can then be used to get information about the lens. For instance, galaxies in general aren't perfectly spherical, but if one has a collection of galaxies one doesn't expect them all to be lined up. Thus, if this set of galaxies is lensed, on average, or statistically, there will be some overall shear and/or convergence imposed on the distribution, which will give information about the intervening lens(es). There is a distribution of galaxies far enough away that can be treated as sources, and thus clusters nearby can be "weighed" (i.e. have their mass measured) using their lensing. Superclusters have been considered as well. In addition, theories of cosmology predict the distribution of large scale structure, the distribution of matter in the universe. The statistical properties of the large scale structure (e.g. the probability of finding a galaxy at one place when there is another a certain distance away) can also be measured by weak lensing, because the matter will produce shear and convergence in distant sources (which can be galaxies, or the cosmic microwave background, for example). Weak lensing is a useful complement to measures of the distribution of luminous mass such as galaxy surveys. Lensing measures all the mass, in particular the dark matter as well as the luminous matter.

In some cases the lensing is of an image that is so small or faint that one doesn't see the multiple images-- the additional light bent towards the observer just means that the source appears brighter. (The surface brightness remains unchanged but as more images of the object appear the object appears bigger and hence brighter.) This lensing can have effects in many measurements, as sources which would have otherwise been too dim become visible. This can be helpful, as when one wants to view objects that would otherwise be too far away. It can also be a problem, for example when one is trying to measure all objects brighter than a certain amount in a certain region and lensing introduces objects by magnifying objects enough to bring them into the sample. There are ongoing searches to use lensing to find a type of dark matter called MACHOs (massive compact halo objects). Although MACHOs, as dark matter, cannot be seen themselves, if they pass in front of a source (e.g. a star nearby), they can cause the star to become brighter for a while, e.g. days or weeks. This effect has been observed (here is a Aug 1998 news article) but determinations of the dark matter are not yet conclusive. Observations are underway by many groups.