Detection of single analyte molecules without the use of any label would improve the sensitivity of current biosensors by orders of magnitude to the ultimate graininess of biological matter. Over two decades, scientists have succeeded in pushing the limits of optical detection to single molecules using fluorescence. However, restrictions in photophysics and labelling protocols make this technique less attractive for biosensing. Recently, mechanisms based on vibrational spectroscopy, photothermal detection, plasmonics and microcavities have been explored for fluorescence-free detection of single biomolecules. Here, we show that interferometric detection of scattering (iSCAT) can achieve this goal in a direct and label-free fashion. In particular, we demonstrate detection of cancer marker proteins in buffer solution and in the presence of other abundant proteins. Furthermore, we present super-resolution imaging of protein binding with nanometer localization precision. The ease of iSCAT instrumentation promises a breakthrough for label-free studies of interactions involving proteins and other small biomolecules. T here are more than 2,000 proteins secreted by the human body at concentrations that are below the detection limit of the current commercial biosensors 1 . To achieve a sufficient sensitivity for early and robust diagnosis of health hazards such as cancer and lethal infectious diseases, scientists have explored a fury of strategies based on mechanical 2 , electrochemical 3 and optical 4,5 interactions. The ultimate performance would be to count the analyte molecules one at a time. Furthermore, a practical biosensor would ideally detect the molecule of interest directly and without the need for labelling, amplification or sophisticated architectures.A well-established approach for label-free detection is via the vibrational signatures of molecules, for example via Raman spectroscopy. This method has an exquisite selectivity, but singlemolecule sensitivity has only been reached by enhancing the Raman cross-section in the near field of surfaces 6 or tips 7 . A common alternative strategy relies on the detection of the refractive index change due to analyte binding. The oldest and commercially available implementation of this technique records the shift of the plasmon resonance of a gold-coated prism 8 , whereby the specificity and selectivity are provided by surface chemistry 9 . While submonolayer detection is readily achievable in such a device, a single molecule does not manifest a significant change in the refractive index of the sensor medium. To increase the sensitivity, confined modes of plasmonic nanostructures [10][11][12] and dielectric microresonators 13,14 have been investigated, but these techniques are intrinsically limited. First, there is a compromise between the size of the sensor active area and its sensitivity. Higher sensitivity comes through confined intensity distributions and, thus, at the cost of more restricted active area, fewer binding receptors and thinner functionalization 15 . Se...