Bar charts may be appropriate when the data reflect an accumulation. Thus, a graph of the distribution of a set of values should be a bar chart. Before using a bar chart ask yourself what the meaning of a zero would be. If it would mean the absence of any representatives in a particular set, then bars are fine. For example, no blue feathers would generate a null bar in a graph of number of individuals vs feather color.
    On the other hand, many graphs are submitted to EER as bar charts when they should be box plots instead. For example, if the value you are displaying is an average, then it does not represent an accumulation. Most of the time you will also want to depict a measure of dispersion (e.g. standard deviation) around an average. Then, not only is a bar chart wrong, but it obscures the effort to show dispersion. Although one can take certain measures to make the deviation clear on a bar chart, the bars are still wrong. Use a box plot.
    Excellent examples of box plots include the following from volume 6 (available free):

    Sébastien Barot, Mikko Heino, Loretta O’Brien & Ulf Dieckmann, 2004. Estimating reaction norms for age and size at maturation when age at first reproduction is unknown. Evolutionary Ecology Research, 2004, 6: 659–678 (Fig 4)
    Ingolf Kühn, Roland Brandl & Stefan Klotz, 2004. The flora of German cities is naturally species rich. Evolutionary Ecology Research, 6: 749–764

    The latter paper also shows examples of proper use of bar charts.