Range is not a measure of center, so this choice is eliminated.
Of the three measures of central tendency, choose the mean, unless you have a good reason not to.
For instance, if there is an outlier, a number so large or so small that it distorts the mean, the median might be a better choice. In this case, there is no such outlier. I have no reason to ever choose the mode, except that it is the easiest to find (there could be a bimodal set of data, but that's a different, more involved, story).
In this case, the median is 4, the mode is 4, and the mean is 3.89. The mean shows that the "average" is slightly less than 4, a good description.