The collection efficiency eta(M) of single fibers, both isolated and within a parallel array, was determined as a function of accumulated particle mass for Stokes numbers St=0.3-3 and interception parameters In=0.04-0.32. This regime is dominated by inertia, interception and particle bounce. Measurements were made with an optical in-situ particle counting technique. Monodisperse polystyrene aerosols (1.3, 2.6, 3.6 and 5.2 mu m) were deposited onto steel wires of 8 and 30 mu m, respectively. Particle accumulation reached values on the order of M=0.5-1 mu g/mm fiber, corresponding to 10(4)-10(6) particles per mm, depending on size. eta(M) was found to increase for all operating conditions between about 2 and nearly 50 fold, most strongly however for those conditions were the bare fiber is ineffective as a particle collector, i.e. for very low or very high values of St number, corresponding to low collision or low adhesion probabilities, respectively. The absolute efficiency increased to values approaching or exceeding unity, i.e. the effective fiber collision cross-section exceeded the diameter of the bare fiber. For fibers within an array of parallel fibers (equivalent packing density similar to 0.004), the initial efficiency eta(0) was higher than for an isolated fiber by almost an order of magnitude. However the increase with loading was substantially smaller, typically by an order of magnitude. The efficiency increase can be described by a power law of the type eta(M)/eta(0) = 1 + b . M(c) where b and c are empirical fit coefficients. Within an array, the exponent c is on the order of 0.7 +/- 0.05, but lower than reported in earlier work on fiber arrays which suggest a value near unity. For isolated fibers, c -> 1 as interception becomes dominant (at very low flow velocities) while at high impact velocities and significant particle bounce c approximate to 0.5. The coefficient c correlates with fiber Reynolds number, but not with other parameters. The coefficient b is inversely proportional to eta(0) (consistent with earlier work, however with significantly lower values than previously published) and a function of (St/R)(2). The experimental section of the paper is preceded by a literature review on single-fiber efficiency data and models for the inertia-interception regime, including both information on bare fibers and dust loaded fibers. An improved, general fit function with physically meaningful limits for St -> 0 and St -> infinity is proposed for the efficiency of bare fibers. (C) 2009 Elsevier Ltd. All rights reserved.