The morphology and rheology of carbon nanofiber (CNF) suspensions were studied. The CNFs, produced by decomposing organic vapors at elevated temperature in the presence of metal catalysts, have characteristic diameter and length of 100 nm and 20-100 mu m, respectively. The CNFs, as delivered, have a strong tendency to clump into mm-sized agglomerates. The efficacy of CNF/glycerol-water suspensions was studied vs. two processing parameters: mechanical sonication and chemical treatment. Experimental measurements revealed that sonication alone reduces the size of CNF clumps from millimeter to micrometer scale, but cannot achieve uniform dispersion. The chemically untreated sonicated suspensions contain clumps of nanofibers with a characteristic size of 20x50 mu m, together with smaller aggregations of partially dispersed nanofibers. In response to this unsatisfactory dispersion, the effect of acid treatment before dispersion was investigated. This acid treatment, which makes the surface of the CNFs more hydrophilic, greatly improves dispersion in the aqueous solution: treatment followed by sonication results in a uniform dispersion of individual nanofibers. At the same time, however, we observed that surface treatment and subsequent sonication greatly shorten the nanofibers. The rheology of CNF/glycerol-water suspensions is highly non-Newtonian both in shear and extensional flows, with strong dependence on the dispersion, particle length, and concentration of the CNFs. As the solvent is Newtonian, all of the elastic and strain-rate dependent behavior in the CNF/aqueous suspensions derives from the addition of nanofibers. The steady shear viscosity of the untreated-sonicated (poorly dispersed, with longer fibers) suspensions is highly shear thinning with a viscosity that increases three orders of magnitude as concentration varies from 0.5 wt% to 5 wt%. Beyond 5 wt% the suspensions are too viscous to be effectively mixed by sonication. When the CNFs are chemically treated and then sonicated (resulting in much better dispersion but shorter fibers), the viscosity exhibits little shear thinning, and only varies by a factor of two from pure solvent to 5 wt%. In small amplitude oscillatory shear measurements, we found strong indications of elastic behavior in both the treated and untreated suspensions, with elastic modulus G' greater than loss modulus G''. In particular, for both systems G' exhibits a low-frequency plateau when nanofiber concentration is 3 wt% or above, a characteristic of elastic solidlike response. Again, there is a strong dependence on CNF dispersion and fiber length: At low frequencies, the elastic modulus of the 5 wt% untreated suspension (with agglomerates and longer fibers) is four orders of magnitude larger than that of the 5 wt% treated suspension (with uniformly dispersed, shorter fibers). In addition, G' of untreated suspensions is a much stronger function of concentration than that of treated suspensions, indicative of network formation. The rheology and morphology of nanofiber suspensions were related by identifying morphology of the suspensions with the assumptions of the kinetic theory-based elastic and rigid dumbbell constitutive models; the approach is to specify the parameters in the kinetic theory models in terms of microscale morphological features measured in the SEM. Of those investigated, the elastic dumbbell model with anisotropic hydrodynamic drag is the most successful, effectively modeling the small amplitude oscillatory shear and steady shear behavior of the treated sonicated suspensions. As for the treated unsonicated and untreated sonicated suspensions, which contain mesoscale agglomerates not present in the underlying assumptions of the dumbbell models, it is discovered that the elastic dumbbell with parameters assigned from morphological measurements predicts the correct trends in the steady shear experiments, but fails to accurately predict the small amplitude oscillatory shear experiments.