Turbulent flow within the cylinder of a reciprocating internal combustion (IC) engine plays a fundamental role that affects engine performance, exhaust emissions, and in particular, their cycle-to-cycle variation. In efforts to predictively simulate engine performance, large-eddy simulation (LES) continues to find increasing interest and with enhanced computational capabilities, LES begins to become a design tool. Validation of LES as well as further enhancements can be guided by experimental data, such as results from particle image velocimetry (PIV) measurements. In this context, this paper discusses the analysis of subgrid-scale (SGS) contributions to LES and assesses appropriate cut-off scales at which the SGS model can derive quantitative scaling information from resolved scale input. Residual stress is one of the most frequently unclosed parameters in SGS models. A one scale-similarity approach had originally assumed that the Leonard stress As such, the relationship between and the becomes of particular interest. Here, LES and PIV data from the transparent combustion chamber (TCC) engine are analyzed via a spatial filtering technique. The analysis reveals spatial limitations of scale-similarity for and . The analysis also dissects the spatial uniformity of scale-similarity for and . The results provide guidance for the validation of models and the choice of filter length scales in simulations.