We report the first measurement of the quantum phase-difference noise of an ultrastable nondegenerate optical parametric oscillator that emits twin beams classically phase-locked at exact frequency degeneracy. The measurement illustrates the property of a lossless balanced beam-splitter to convert number-difference squeezing into phase-difference squeezing and, thus, provides indirect evidence for Heisenberg-limited interferometry using twin beams. This experiment is a generalization of the Hong-Ou-Mandel interference effect for continuous variables and constitutes a milestone towards continuous-variable entanglement of bright, ultrastable nondegenerate beams. and progress has been made in this direction [9]. Recently, Silberhorn et al. made a beautiful demonstration of continuous-variable entanglement of picosecondpulsed optical beams, by simultaneous squeezing of the number sum and of the phase difference [10]. For highprecision measurements, however, stable CW beams are preferable. One interesting system for this purpose is the ultrastable nondegenerate optical parametric oscillator (OPO), which emits intense twin beams. In a type II OPO, these twin beams are orthogonally polarized. It is thus easy to separate them spatially and to subsequently make their polarizations parallel. Then, the twin beams can be made indistinguishable by locking their frequency difference to zero, which has been an experimental challenge. This Letter reports the first experimental demonstration of nonclassical interference of such macroscopic boson fields.In general, a quantum interference experiment consists in "splitting" a quantum field into two subfields, each experiencing its own phase evolution, and then "recombining" these subfields and performing a measurement. The corresponding probability distribution presents fringes which give information about the phase difference of the two subfields. Examples include the Mach-Zehnder interferometer for light and the Ramsey interferometer for matter, which are isomorphic to each other. "Nonclassical interference" may either mean that waves of a nonclassical nature are involved (e.g. matter waves), or that their behavior itself has no classical optical equivalent. The latter situation is what interests us, and is determined by the role of the vacuum modes of the quantum field, i.e. the physics of the "splitting." The physics of the beam splitter ( Fig. 1) takes center stage here and also determines the phase measurement noise. Take the example of an initial N -photon Fock state |k a ,ǫ, ω; N a , wherek andǫ are the unit wave and polarization vectors and ω the angular frequency. The beam splitter output is given by the interference of this state with the corresponding polarization-and frequency-degenerate vacuum state |k b ,ǫ, ω; 0 b . As was first demonstrated by Caves in 1980 [11], this yields a (classically intuitive) binomial probability distribution of the photon number between modes c and d, with standard deviation ∆N out − = ∆(N c − N d ) ∝ N 1/2 . Using the Heisenberg inequality...