Lytton ⁵⁾ provides an in-depth summary of the historical developments of NDT, backcalculation and theoretical considerations as well as associated technology in his state-of-the-art presentation in 1988.
He also illustrates some of the concerns regarding the differences between backcalculated results using different backcalculation programs on the same deflection data.

These are typically technical problems but they are exacerbated by the continuing development of similar backcalculation programs.
In many cases, new programs have little to differentiate them from existing software other than a name.
Rada et al ⁶⁾, in their description of the SHRP backcalculation procedure software selection originally included a table listing the most common backcalculation procedures in use at that time.

This table, somewhat modified, is included here as Table 1 to illustrate similarities and differences between programs. In reviewing Table 1 it should be kept in mind that the CHEVRON and ELSYM5 numerical integration routines are identical, and until recently, produced erroneous results under certain circumstances due to an error in the integration procedure.
This error was corrected in 1992 by Irwin ⁷⁾ and verified by comparison with the BISAR program.

The programs listed in Table 1 are by no means a comprehensive listing of backcalculation routines.

Other programs in use today include COMDEF, DBCONPAS, PROBE, ILLIBACK, LMBS, DEFMET, RPEDD1, PHONIX, PEACH, FALMAN, CLEVERCALC, EPLOPT, OAF, SEARCH, EFROMD and more.

Most of the programs rely on linear elastic layered theory, or a variation thereof, for the basic structural model.
In comparing results from these programs, the primary criterion used for evaluation of accuracy is based on the goodness of fit of computed deflections to measured deflections.
As computing power has increased, so has the ability to improve the goodness of fit.

An important fact to keep in mind is that, in many cases, improving the goodness of fit does not necessarily mean that the theoretical model better represents actual pavement response.

If an existing pavement structure is in such a condition that it clearly violates some of the fundamental assumptions of elastic theory, then a good fit between measured and calculated deflections should not be expected, and goodness-of-fit should not be the determining factor for deciding if a solution is realistic or not.

This point is also made by Lytton ³⁾ who discusses the need for experience in analysis, with materials and with deflections to ensure that the backcalculation process yields the most acceptable set of moduli for a given deflection basin.
It should be noted that essentially all pavements violate the fundamental assumptions of linear elastic theory, albeit to differing degrees.

Also important, and related to the issue discussed above, is the fact that backcalculation provides a modulus value that is a layer parameter and not necessarily the layer material modulus, which can be measured using laboratory tests on a sample of the layer material.

This is due to the geometry of typical deflection basin measurements which is typically on the order of a 1.8 meter (6 ft.) length so that the effect of horizontal layer and material variability over that dimension is included in the backcalculated moduli.

This variability includes damage such as cracking, both on the macro- and micro-structural level.
Simply stated, the problem lies with the fact that the in-situ modulus is not known, so that backcalculated values cannot be validated directly.