This study integrates on-site studies of two marble-clad buildings with laboratory tests in an attempt to elucidate the causes of deterioration of the panels on these buildings. The buildings each contain about 45,000 panels and are 83 (ACB) and 72 (TC) storeys tall respectively. The taller building was 17 years old at the time of study and the latter is presently 27 years old. There are minor differences in the panel sizes, anchoring and arrangement on the buildings but they have not been found to be significant. Similar marble types from the vicinity of Carrara, Italy were used on both buildings. The primary manifestation of panel deterioration was the progression from flat panels upon installation to bowed or more accurately dish-shaped geometry. On ACB 31% of the building developed bows over 1 cm in 17 years and 26% on TC in 20 years. All panels on both buildings were measured periodically, demonstrating a demographic distribution of the bowing. On both buildings the bowing is most extensive in those regions exposed to the most thermal heating, either from direct sunlight or from reflected sunlight, regions of shading showed negligible bowing. Therefore, the south- and west-facing panels and particularly those on the upper floors generally showed the most bowing. Strength measurements from 122 panels removed from ACB show an average strength loss of 26% when compared to Virgin panels that had been kept in temperature controlled storage for the same period. Comparable tests on 158 panels from TC show average strength losses of about 35% over the longer period of time. The strength distribution in panels containing significant bow is not uniform. At the top and lower edges of panels from TC the strength loss is only about 19%, however, where the amplitude of the bow is largest the loss averages 50% for the 158 panels. Additionally, test specimens from areas of bow curvature show the extended portion of the bow to be 25 to 30% weaker than the inside of the bow. Petrographic observations document that microfractures are concentrated in the extended portions contributing to the significant strength loss in those areas. Thermocouples to record air temperatures were placed outside and behind the panels, as well as on outside and inside panel faces and in the center of the panel. These clearly demonstrate that thermal heating, not air temperature correlate with panel bowing. Recognizing that differential thermal expansion of calcite can result in weakening of marble led to controlled laboratory experiments to attempt to define the strength-loss curve. Portions of thirty-one panels were subjected to a cycle of twelve hours in an oven at a given temperature followed by twelve hours at room temperature. Three temperatures of 12, 66 and 107 °C above 24 °C room temperature were used. Initial data contained 200 cycles, but has now been extended to 450 cycles. Initially, panels show a sharp decrease in strength during the first 25 cycles, with the rate of decrease becoming significantly less with increasing cycles. The strength decrease is interrupted by periodic, temporary increases, which appear to be related to increases in the residual elastic strain of the marble. Upon release of this strain, the strength drops to low levels again. By about 120 cycles the panels appear to reach a basic strength beyond which further loss is minimal. Porosity measurements made on panels removed from TC show a sub-population of panels that have values significantly higher than the general panels. Average porosities for the general population are about 1.5% with a standard deviation of 0.3%. The smaller group of panels has values ranging from 2.1 to 5.8%. The porosity measurements are all made close to the upper or lower edges of the panels to minimize the influence of pore space associated with fractures created during bowing. As would be predicted, this group has average strengths 72% lower than the general population. Some of the high-porosity panels were in shaded areas on the building and had not developed significant bows, nor do they presently show any ordered distribution on the building. It is hypothesized that this group of panels had intrinsically high porosities when quarried. These panels do not have any macroscopically recognizable features to distinguish them, so with their low strength they pose a significant threat to the cladding integrity. An acoustic instrument, which measures compressional-wave travel time between two transducers, was developed and is used to detect the panels of higher porosity. A correlation between longer travel times and higher porosity offers a non-destructive method of evaluating panel strength on panels attached to the building and the potential for correlating strength changes with time.