Synchrotron radiation circular dichroism (SRCD) spectra were recorded for a family of 12 DNA duplexes that all contain nine adenines (A) and nine thymines (T) in each strand but in different combinations. The total number of AT Watson-Crick (WC) base pairs is constant (18), but the number of cross-strand (CS) hydrogen bonds between A and T varies between 0 and 16. the maximum possible. Eleven of the duplexes have one or more A tracts, and one duplex has T tracts. The signals due to hybridization were found from Subtraction of spectra of single strands from spectra of the duplexes, The residual spectrum of the T-tract duplex T(9)A(9):A(9)T(9) (5'-3':3'-5') significantly differs from that of the A-tract duplex A(9)T(9):T(9)A(9), but only below 210 urn, which suggests that the signal in this region depends oil the superhelicity of the duplex. A principal component analysis of all residual spectra reveals that spectra of A-tract duplexes can be obtained to a good approximation as a linear combination Of just two basis spectra. The first component is assigned to the spectrum of 18 WC and 8 CS pairs, whereas the second component is that of 8 CS pairs. This interpretation is supported by separate experiments on duplexes of varying lengths but with similar arrangements of the A and T's and by experiments on two other duplex families of 14 and 30 base pairs. The best correlation is obtained by the assumption that cross-strand interactions occur as long as there are two adenine neighbors in a strand. Our data indicate that a circular dichroism spectrum of a duplex containing only A and T can simply be inferred from the number of WC base pairs and the number of CS interactions, and we provide reference spectra for these two interactions. Finally, time dependent density functional theory calculations of the circular dichroism spectra for an isolated WC base pair and two different CS base pairs (between adenine N-6 amine and thymine O-4 or between adenine C-2-H and thymine O-2) were performed to provide some additional support for the interpretation of the experimental spectra. We find large differences between the two calculated CS spectra. However. there is a reasonable qualitative agreement between the calculated WC and the C-2-H center dot center dot center dot O-2 CS spectra and those deduced from the experimental data.