In the history of science and technology, researchers have always undertaken endeavors to enhance the degree of “directness of measurement”. With “directness of measurement”, we conceptualize that the essence of the object under investigation, the structure, dynamics, and function of biological samples in the scope of this Review, is entirely and straightforwardly assessed by the measurement, bypassing hypotheses and intricate data analysis. When the degree of measurement directness is low, conclusions derived from the gleaned data often differ depending on the formulation of hypotheses, analysis models, and the data interpretation. That is why conceptual consensus about a specific issue of the studied object is rarely reached based on indirect data. This is often the case for studies of structure–function relationship of proteins. Previously, scientists had to understand distinct qualities of proteins by measuring a sample containing a huge number of molecules. The results are ensemble-averaged quantities (often at equilibrium) that provide only limited information on the proteins of study. Proteins are dynamic in nature and work at the single-molecule level. Protein molecules fluctuate, undergo structural changes, bind to and dissociate from interaction partners, and traverse a range of energy and chemical states during molecular action. Most, if not all, of these dynamics and their statistical distributions are convoluted and hidden in ensemble averaging measurements. To overcome the limitations of ensemble measurements, single-molecule biophysics was developed more than two decades ago, with the use of fluorescence microscopy,1,2 optical spectroscopy,3,4 and optical and magnetic tweezers,5,6 whose performances were further improved by the advancements of improved optical microscopes, lasers, electronics, computers, and high-sensitivity video cameras and sensors. Using these techniques, our understanding of the functional mechanism of proteins has made significant steps forward. Moreover, super-resolution optical microscopy techniques bypassing the diffraction limit for fluorophore localization have recently been added to fluorescence microscopy.7−9 However, the degree of directness of measurement is still limited, because the protein molecules themselves are invisible in these single-molecule measurements. Protein structure is typically studied by X-ray crystallography, electron microscopy (EM), and nuclear magnetic resonance (NMR) spectroscopy. To date, these techniques have revealed detailed three-dimensional structures of over 94 000 proteins (Protein Data Bank (PDB), http://www.rcsb.org/pdb/home/home.do), with a growth rate of about 8000 novel structures per year (2010–2013). Yet, these techniques make use of ensemble averaging, and, more seriously, the obtained structures are merely limited to static snapshots of fixed conformations. Thus, the simultaneous and direct observation of structure, dynamics, and function of single protein molecules has long been infeasible, and hence the materialization of a technique allowing such an observation has long been awaited in biological sciences. An ideal microscopy technique that allows simultaneous observation of structure, dynamics, and function of single protein molecules has to meet all of the following conditions (see Table 1): (i) in-liquid specimen imaging, (ii) high spatial resolution, (iii) high temporal resolution, (iv) low invasiveness to the specimen, and (v) direct imaging of the specimen without the use of markers (in other words, resolving the structure of the specimen itself). Although efforts have been made to develop environmental electron microscopy techniques capable of observing unstained biological specimens in solutions,10 the strong electron dose that is required to achieve high contrast and spatial resolution instantaneously denatures the sample. Achieving the above-described goals by EM is a highly difficult, if not impossible, task. Conventional atomic force microscopy11 (AFM) meets most of the above-mentioned conditions, except for the third condition, that is, high temporal resolution, and the fourth condition, that is, low invasiveness, is only moderately satisfied. Table 1 Feasibility Comparison of Three Types of Microscopy