We describe spontaneous parametric down-conversion experiments in which induced coherence across two coupled interferometers results in high-visibility single-photon interference. Opening additional photon channels allows "which-path" information and reduces the visibility of the singlephoton interference, but results in nearly perfect visibility when photons are counted in coincidence. A simplified theoretical model accounts for these complementary observations and attributes them directly to the relations among the vacuum fields at the different crystals.Complementarity, often discussed in quantum optics as wave-particle dualism, is a fundamental principle of undiminished interest [1][2][3][4]. "Which-path" information about a single photon is possible only at the cost of decreasing interference fringe visibility V . This can be expressed quantitatively by the formula K 2 + V 2 ≤ 1, where the "contrast" K is a measure of which-path information [5]. Recent work [6,7] demonstrated among other things that single-photon interference is determined by the mode function associated with the photon, as quantum field theory predicts [6]. In this paper we describe experiments on spontaneous photon down-conversion (SPDC) demonstrating that which-path information in single-photon interference reduces fringe visibility but results in very high visibility in second-order interference measurements. These experiments may serve to clarify the roles of mode functions and vacuum fields in SPDC, which has for many years been a primary platform for fundamental studies in quantum optics and conceptual foundations of quantum theory.Although the signal and idler photons in SPDC have no fixed phase relation, the combined entity, or biphoton, carries observable phase information about the pump field [8,9]. In particular, by varying the phase delay between the pump fields for two crystals, first-order singlephoton interference has been observed [10]. In these experiments the idler photon modes i1 and i2 of crystals BBO1 and BBO2 (Figure 1) were aligned and indistinguishable, and interference of the signal photon channels s1 and s2 was observed as a consequence of induced coherence between BBO1 and BBO2.Suppose now that a third pumped crystal (BBO3) is added. If the s1 and s3 modes are aligned and indistinguishable, the possibility of generating a signal photon at BBO1 or at BBO3 might suggest that single-photon interference between the channels s1 and s3 should be observable at detector A behind beam splitter BS1. But we now have "which-path" information: detection of photons i3 with a detector D behind the beam splitter BS2, for instance, implies that a photon pair (i3,s3) must have been emitted from crystal BBO3. In spite of this which-path information, however, we observe highvisibility signal-photon interference fringes at A when these signal photons are counted in coincidence with idler photons at D.The experimental setup employed three BBO crystals for biphoton generation (Figure 1). All three crystals had a length of 4 mm and were...