In face-to-face software-training settings, the behavior-modeling method has been shown to be more effective than either instruction-based or exploration-based methods. Yet, in online asynchronous software-training settings the behavior-modeling method may not be superior. This study provides evidence as to the relative effectiveness of these three software-training methods for the online asynchronous environment. A field experiment was conducted adapting the three methods to an online asynchronous environment, and including a face-to-face behavior-modeling control condition. Results indicate that when conducting software training, it may be almost as effective to use an online asynchronous behavior-modeling method as it is to use a more costly face-to-face behavior-modeling method. Although somewhat better results sometimes are evident for face-to-face behavior-modeling, observed differences are not significant, nor are their patterns consistent. Within the limits of instrument translation concerns, there seems to be no significant interaction of learning style and training method on training performance or trainee reactions to their training. The study provides evidence as to the relative effectiveness of various methods, particularly those of online asynchronous nature, for software training. Within its limits, this research takes a first step in developing a strategy for online asynchronous software training.