“Using the complete paragraph [see the above image], I drew in the lines with a ruler reading normally (top to bottom, left to right), running approximately from the center of the spaces without intersecting any other letters or punctuation. Longer lines superseded shorter lines where there was a choice, and no space was allowed to be part of two lines. Spaces at the end of lines were not included. In two cases that were difficult to determine, I counted the shorter line. This still remains slightly arbitrary, and another person repeating this process could come up with somewhat different counts. My resulting counts follow: (48 spaces were not part of any line, and are not counted)
Y, count, expected Poisson proportion
0, 34, 0.57
1, 9, 0.32
2, 3, 0.09
3, 3, 0.018
4, 1, 0.002
Which results in a mean of 0.56, which corresponds to an average of 2.56 lines spanned. Fitting a Chi-Square test of homogeneity with the null hypothesis that the observed cell counts will be approximately that expected under a Poisson distribution, I get a Chi-Square statistic of 17.624, df=4, with a p-value of 0.0015. This seems to indicate that my initial idea does not fit the data well. Longer lines are more common in the paragraph than might be expected if this length followed a Poison distribution.
I note that the number of lines spanning 2 rows is much greater in the complete paragraph compared to the partial paragraph where I did my initial counts. Also, my counting was much more careful this time. The two difficult cases where I chose the shorter lengths will tend to favor the null hypothesis.”
