We present the results of a study comparing three directed-edge representations in node-link diagrams. Node-link diagrams are probably the most popular type of graph representations; nodes are depicted as dots and links as straight or curved lines connecting the nodes. An arrowhead placed at the end point of a link is traditionally used to indicate edge direction. However, Holten and Van Wijk showed that this arrow representation was not optimal. Their user study compared the performance (in terms of reading time and correctness) of various directed-edge representations. A tapered representation - wide at the start and narrow at the end - showed the best performance. This paper presents the results of a follow-up study comparing the performance of the tapered representation with two other representations: one using biased curvature and the other using animation to indicate edge direction. We tested the three representations using a more realistic setting than the original article. We used random small-world graphs generated with the Barabàsi-Albert model, we used the Fruchterman-Reingold algorithm to lay out the graphs, and we varied graph density and link length to measure their influences. Overall, our study shows that the tapered and animated representations perform significantly better than the biased-curvature representation, with no significant differences between tapered and animated except for medium-length representations where tapered was faster. The article presents detailed results and provides practical recommendations on the use of directed-edge representations.