A series of linearly arranged donor-spacer-acceptor (D-S-A) systems 1-3, has been prepared and characterized. These dyads combine an Os(II)bis(terpyridine) unit as the photoactivable electron donor (D), a biphenylene (2) or phenylene-xylylene (3) fragment as the spacer (S), and a N-aryl-2,6-diphenylpyridinium electrophore (with aryl = 4-pyridyl or 4-pyridylium in 1 or 2/3, respectively) as the acceptor (A). Their absorption spectra, redox behavior, and luminescence properties (both at 77 K in rigid matrix and at 298 K in fluid solution) have been studied. The electronic structure and spectroscopic properties of a representative compound of the series (i.e., 2) have also been investigated at the theoretical level, performing Density Functional Theory (DFT)-based calculations. Time-dependent transient absorption spectra of 1-3 have also been recorded at room temperature. The results indicate that efficient photoinduced oxidative electron transfer takes place in the D-S-A systems at room temperature in fluid solution, for which rate constants (in the range 4 × 10(8)-2 × 10(10) s(-1)) depend on the driving force of the process and the spacer nature. In all the D-S-A systems, charge recombination is faster than photoinduced charge separation, in spite of the relatively large energy of the D(+)-S-A(-) charge-separated states (between 1.47 and 1.78 eV for the various species), which would suggest that the charge recombination occurs in the Marcus inverted region. Considerations based on superexchange mechanism suggest that the reason for the fast charge recombination is the presence of a virtual D-S(+)-A(-) state at low energy--because of the involvement of the easily oxidizable biphenylene spacer--which is beneficial for charge recombination via superexchange but unsuitable for photoinduced charge separation. To further support the above statement, we prepared a fourth D-S-A species, 4, analogous to 2 but with a (hardly oxidizable) single phenylene fragment serving as the spacer. For such a species, charge recombination (about 3 × 10(10) s(-1)) is slower than photoinduced charge separation (about 1 × 10(11) s(-1)), thereby confirming our suggestions.