In this work, we use AFM measurements in conjunction with dialysis experiments to study the synthesis mechanism and physical state of dendrimer-stabilized platinum nanoparticles. For characterizing particle size distributions by high resolution transmission electron microscopy and AFM, sample preparation by drop evaporation presumably minimizes the risk of sample bias that might be found in spin coating or dip-and-rinse methods. However, residual synthesis by-products (mainly salts) must be removed from solutions of dendrimer-stabilized metal nanoparticles prior to AFM imaging. Purification by dialysis is effective for this purpose. We discovered, by UV-visible spectrophotometry and atomic absorption (AA) spectroscopy, that dialysis using 'regular' cellulose dialysis tubing (12 000 Da cut-off) used in all previous work leads to substantial losses of poly(amidoamine) (PAMAM) dendrimer (G4OH), PAMAM-Pt(+2) complex, and PAMAM-stabilized Pt nanoparticles. Use of benzoylated dialysis tubing (1200 Da cut-off) shows no losses of G4OH or G4OH-Pt mixtures. We use AFM to see whether selective filtration during dialysis introduces sampling bias in the measurement of particle size distributions. We compare results (UV-visible spectra, AA results, and AFM-based particle size distributions) for a sample of G4OH-Pt(40) divided into two parts, one part dialysed with regular dialysis tubing and the other with benzoylated tubing. Exhaustive dialysis using benzoylated tubing may lead to the loss of colloidal Pt nanoparticles stabilized by adsorbed dendrimer, but not Pt nanoparticles encapsulated by the dendrimer. The comparisons also lead to new insights concerning the underlying synthesis mechanisms for PAMAM-stabilized Pt nanoparticles.