We report the development of a fluorescencebased immunoassay technique relying on the physical phenomena of random number fluctuations and diffusion, which we review. By determining the autocorrelation of the fluctuations in the fluorescent intensity, this method is able to measure the amount of labeled antigen or antibody that is bound to micrometer-sized carrier particles in solution. The principal advantage of this technique is its insensitivity to small, fast-diffusing sources. It also discriminates against weakly fluorescent contaminants of size comparable to the carrier particles. We demonstrate these attributes by using two model systems: a human IgG assay and an idealized system consisting of polystyrene fluorescent spheres and rhodamine dye. and diffusion of particles, in which the diffusion coefficient is inversely related to particle size. To distinguish between large, slowly diffusing sources of fluorescence (due to Ag*-Ab or Ab-Ab* attached to carrier particles) and small, rapidly diffusing ones, we simply keep track of the fluctuations in fluorescence intensity that occur for a given small volume element in the solution. Clearly, any fluctuation in intensity that occurs due to a fluctuation in the number of particles present in the volume will, on average, persist for a time which depends inversely on the rate of diffusion of the particles into or out of the sampling volume. For the large tagged carrier particles these fluctuations will be very long lived. For any small, freely diffusing fluorescing molecules there will effectively be no "memory" of a fluctuation having occurred at a finite time earlier, provided that the sampling volume is made appropriately small. the root mean square magnitude of the fluctuations is N1'2. A convenient way to monitor these fluctuations is to evaluate the familiar intensity autocorrelation function, [1] in which I(t') is the fluorescent intensity originating from OV at time t' and the symbol ( ... )tt indicates an average of the intensity product over a large number of sampling times t'.We now periodically sample the fluorescent intensity at hundreds of locations within the solution, using in each case a volume of the same size (bV 10-6 cm3). The large number of samples increases the signal/noise ratio of function C(t). When time t equals an integral multiple of the repetition period T, the two intensities that form the product in Eq. 1 refer to the same volume element, for a given t'. The usefulness of the autocorrelation function in our application depends upon the fact that a number fluctuation in WV has a finite lifetime, r. Fig. la is an idealized trace of the intensity I(t') showing two fluctuations, each of which repeats after the period T (» >> T). In reality, the intensity fluctuations do not resemble the regular, symmetric pattern of Fig. la; rather, they appear as "random" noise (as illustrated in Fig. 2). This idealization was chosen to permit a simple algebraic computation of C(t) as discussed below. We assume a mean number of particles N per...