We have performed ab initio CASSCF, CASPT2, and EOM-CCSD calculations on doubly deprotonated p-coumaric acid (pCA(2-)), the chromophore precursor of the photoactive yellow protein. The results of the calculations demonstrate that pCA(2-) can undergo only photoisomerization of the double bond. In contrast, the chromophore derivative with the acid replaced by a ketone (p-hydroxybenzylidene acetone, pCK(-)) undergoes both single- and double-bond photoisomerization, with the single-bond relaxation channel more favorable than the double-bond channel. The substitution alters the nature of the first excited states and the associated potential energy landscape. The calculations show that the electronic nature of the first two (pi,pi*) excited states are interchanged in vacuo due to the substitution. In pCK(-), the first excited state is a charge-transfer (CT pi,pi*) state, in which the negative charge has migrated from the phenolate ring onto the alkene tail of the chromophore, whereas the locally excited (LE pi,pi*) state, in which the excitation involves the orbitals on the phenol ring, lies higher in energy and is the fourth excited state. In pCA(2-), the CT state is higher in energy due the presence of a negative charge on the tail of the chromophore, and the first excited state is the LE state. In isolated pCA(2-), there is a 68 kJ/mol barrier for double-bond photoisomerization on the potential energy surface of this LE state. In water, however, hydrogen bonding with water molecules reduces this barrier to 9 kJ/mol. The barrier separates the local trans minimum near the Franck-Condon region from the global minimum on the excited-state potential energy surface. The lowest energy conical intersection was located near this minimum. In contrast to pCK(-), single-bond isomerization is highly unfavorable both in the LE and CT states of pCA(2-). These results demonstrate that pCA(2-) can only decay efficiently in water and exclusively by double-bond photoisomerization. These findings provide a rationale for the experimental observations that pCA(2-) has both a longer excited-state lifetime and a higher isomerization quantum yield than pCK(-).