experimentally. As mentioned earlier, the issue becomes how to model the maxilla. One
possible reason the maxilla may experience small loads is the presence of the hard palate.
Unlike the mandible, the maxilla has the hard palate which may serve to eliminate or
greatly reduce twisting and bending (Daegling and Hylander 1997).
Measurements of different size/shape variables were also taken from each
specimen in order to determine whether or not a relationship exists between these
particular measurements and the fractal dimensions of the mid-palatal sutures. Only
three regressions showed significance, but the correlation values were very weak (Table
6). When these measurements were regressed against the ruler fractal dimension, palate
height and palate depth in Procolobus badius showed significance. Interestingly there
were no significant regressions in Colobuspolykomos for the ruler fractal dimension.
However, the opposite is true for the information fractal dimension. No significant
results were found for Procolobus badius, but the regression of information fractal
dimension versus facial width in Colobuspolykomos showed significance. The fact that
the so-called equivalent fractal dimensions yield different significance is further
evidence that these are not equivalent measures. More than likely the significant P-
values for these three regressions reflect a type I error instead of real significance,
although there is no way to truly know if a type I error was committed. The results of the
regressions suggest that there is no predictable pattern between either of the fractal
dimensions and any of the size/shape variables.
One problem limiting interpretation was small sample sizes. When dealing with
biological samples, obtaining sufficiently large sample sizes can be a problem. An
attempt to deal with this problem was made by bootstrapping the data. However, as