Nanomaterials offer opportunities to construct novel compounds for many different fields. Applications include devices for energy, including solar cells, batteries, and fuel cells, and for health, including contrast agents and mediators for photodynamic therapy and hyperthermia. Despite these promising applications, any new class of materials also bears a potential risk for human health and the environment. The advantages and innovations of these materials must be thoroughly compared against risks to evaluate each new nanomaterial. Although nanomaterials are often used intentionally, they can also be released unintentionally either inside the human body, through wearing of a prosthesis or the inhalation of fumes, or into the environment, through mechanical wear or chemical powder waste. This possibility adds to the importance of understanding potential risks from these materials. Because of fundamental differences in nanomaterials, sound risk assessment currently requires that researchers perform toxicology studies on each new nanomaterial. However, if toxicity could be correlated to the basic physicochemical properties of nanomaterials, those relationships could allow researchers to predict potential risks and design nanomaterials with minimum toxicity. In this Account we describe the physicochemical properties of nanoparticles (NPs) and how they can be determined and discuss their general importance for cytotoxicity. For simplicity, we focus primarily on in vitro toxicology that examines the interaction of living cells with engineered colloidal NPs with an inorganic core. Serious risk assessment of NPs will require additional in vivo studies. Basic physicochemical properties of nanoparticulate materials include colloidal stability, purity, inertness, size, shape, charge, and their ability to adsorb environmental compounds such as proteins. Unfortunately, the correlation of these properties with toxicity is not straightforward. First, for NPs released either unintentionally or intentionally, it can be difficult to pinpoint these properties in the materials. Therefore, researchers typically use NP models with better defined properties, which don't include the full complexity of most industrially relevant materials. In addition, many of these properties are strongly mutually connected. Therefore, it can be difficult to vary individual properties in NP models while keeping the others constant.