<p>Determining the representative elementary volume (REV) is critical for ensuring the statistical representativeness of porous media characterization. This study evaluated the influence of particle morphology and size distribution on REV estimates in monodisperse glass beads and polydisperse sand grains. We tested two hypotheses: (i) porous systems composed of nearly regular particles but exhibiting different pore-size distributions have distinct REVs, and (ii) the REV associated with structural complexity varies according to specific multifractal parameters, namely the capacity (<i>D</i>_<i>0</i>), information (<i>D</i>_<i>1</i>), and correlation (<i>D</i>_<i>2</i>) dimensions. Three-dimensional X-ray microtomography (XCT) images were analyzed using a dynamic stability criterion to determine REV values for morphological properties (porosity, anisotropy, connectivity, pore number, and tortuosity) and multifractal dimensions (<i>D</i>_<i>0</i>, <i>D</i>_<i>1</i> and <i>D</i>_<i>2</i>). Glass beads exhibited smaller and more consistent REVs, with window lengths ranging from 150 to 300 pixels, reflecting their high geometric uniformity. In contrast, sand samples required larger window lengths (125–400 pixels), and REV was strongly dependent on the analyzed property. Porosity and tortuosity REVs increased with grain size for sands, with anisotropy and the multifractal dimensions <i>D</i>_<i>0</i> and <i>D</i>_<i>2</i> showing the opposite trend, requiring larger representative scales for finer, structurally more complex grains. Our findings demonstrate that REV is not a fixed property but varies according to the structural attribute under investigation. Consequently, selecting REV based solely on properties with limited sensitivity to pore-network complexity, such as porosity, may result in over- or underestimation of representative scales, potentially affecting the characterization and modeling of transport processes in porous media.</p>