It is well known from optical and resonance Raman studies that the degeneracy of the excited electronic states Q and B of heme chromophores are generally split by electronic and vibronic perturbations induced by the protein matrix. A direct observation of this band splitting at room temperature is generally considered impossible because of the large overlap between the band originating from transitions into the x- and y-components of the excited states. Recently, however, Dragomir et al. showed, for cytochrome c, that the fine structure of the circular dichroism (CD) associated with the Q- and B-bands can be used to estimate the respective splitting of these bands. In the present study we measured and analyzed the CD spectra of the B-bands of ferrous deoxymyoglobin (deoxyMb), ferric (high spin) aquo-met myoglobin (metMb) and ferric (low spin) myoglobin cyanide (MbCN). Thus, we discovered a significant non-coincidence between the positive Cotton band in the CD spectrum and optical absorption spectrum of deoxyMb with respect to both, wavenumber position and band profile. A similar though less pronounced effect is discernable for MbCN, whereas a close correspondence between CD and absorption spectrum was observed for met-Mb. Our observations suggest that the ferrous deoxy state of the protein is significantly more perturbed than the investigated hexacoordinated ferric states. We performed a consistent modeling of the absorption and CD profile of deoxyMb and of depolarization ratios of resonance Raman lines measured with 442 nm excitation. We found that the observed non-coincidence between absorption and dichroism results from a large electronic perturbation which causes a substantial splitting of the B-band. Only the y-component of the B-band exhibits a significant positive Cotton effect. Our results question the validity of some earlier attempts to rationalize the B-band in terms of a Gaussian distribution of Fe²⁺-displacements.