The use of pheromone trails in ant colony organization is an important model for understanding collective decision-making and complex adaptive systems. The ant Lasius niger L. (Hymenoptera: Fomicidae) is one of the main model organisms used for such studies. Key to understanding pheromone trail use by ants is knowing how well trails are followed. The results of a previous study suggest that L. niger trail following is poor, with between 60% and 70% accuracy at a T bifurcation. It is hypothesized that the true trail following accuracy is higher, and that the low accuracy reported previously is the result of a methodological error. Specifically, it is hypothesized that 'task state' (i.e. what the ants 'thought they were doing') affected pheromone following accuracy. In the present study, the task state of the ants is set experimentally to one of three states: scouting (completely naive), recruited (having information that food has been found, but not where it is) and shuttling (having a strong memory of the location of a food source). Trail following accuracy is tested for each group. Trail following is found to be more accurate than previously reported: 83%, 82% and 74% correct decisions for scouts, recruits and shuttlers, respectively. However, the difference between the three groups is not significant. Importantly, very high inter-trial variation is reported both in the present study and in experiments from other research groups. This variation is unexplainable by trail strengths or colony-level differences, and is highlighted as an important factor when experimentally measuring trail following.