408 JOURNAL OF LAW AND POLICY
Table 6: State v. Cunningham: counts of shared and nonshared tokens by
dialog-type, speaker, and n-bar
(^) Other Sharing Self Sharing
(^) DialogType Actual Randomized Actual Randomized
Speaker Nbar OS NOS OS NOS SS NSS SS NSS
CARNES 1 159 933 1627 9293 110 254 558 3082
(^) 2+ 30 1827 285 18285 65 554 119 6071
COURT 1 175 1238 2029 12101 110 361 761 3949
(^) 2+ 32 2623 315 26235 42 843 295 8555
MURPHY 1 261 2679 4023 25377 324 656 2462 7338
(^) 2+ 23 5575 451 55529 80 1786 316 18344
WOLFE 1 11 100 163 947 4 33 46 324
(^) 2+ 1 167 14 1666 1 55 10 550
- Discussion
Applying the method proposed to the data examined here
supports the conclusion that the prosecutor and potential juror
persisted in their respective communication strategies. The fact
that allo-repetition effects that distinguish the actual dialogue
from its ten randomized counterparts do not appear strengthens
the argument that mutual understanding did not emerge during
the interaction. This analysis coincides with the determination on
appeal that the jury member did not unambiguously demonstrate
understanding of the concept of presumed innocence.
IV. CONCLUSIONS
While the current work is in the spirit of the traditions in
statistical methods for authorship attribution,^48 it is focused on
(^48) See generally GEORGE U. YULE, THE STATISTICAL STUDY OF
LITERARY VOCABULARY (1944) (examining word-distribution from different
portions of author’s works throughout history); Harald Baayen et al., Outside
the Cave of Shadows: Using Syntactic Annotation to Enhance the Authorship
Attribution, 11 LITERARY & LINGUISTIC COMPUTING 121, 121–32 (1996)
(reporting an experiment in which statistical measures and methods previously
applied to words and their frequencies of use are applied to rewrite rules);