Tree-structured probabilistic models offer a balance between interpretability and tractability, making them a cornerstone in density estimation and graphical model learning. In this work, we present an empirical evaluation of three recent approaches for learning tree distributions, benchmarking them against the classic Chow-Liu algorithm. Although Chow-Liu is decades old, our results show that it continues to outperform or match the newer methods in most scenarios, including both synthetic and real-world datasets. A key contribution of this study is the implementation and practical evaluation of two online learning algorithms, which had not previously been tested in practice or compared directly with the other methods we use. We analyze performance across multiple dimensions: realizability of the data distribution, alphabet size, and dataset scale. Our findings reaffirm the competitiveness of the Chow-Liu algorithm while also offering a comparison among the three methods, shedding light on their respective strengths and limitations across different regimes.