The true tree/doppelganger pairs, as determined by the public’s vote:
1D, 2A, 3B, 4C, 5H, 6E, 7I, 8G, 9F


Total number of votes: 43
Total number of valid votes: 41
Total number of unique visitors: 4186
Percent of visitors who voted: 1.03%
Note: 2 people’s votes were invalidated because they paired up the same tree or doppelganger more than once by accident. Repeated attempts to contact these voters about the error were unanswered.

Number of votes received for each tree/doppelganger pair:

Percent of votes received for each tree/doppelganger pair:i

Number of voters who are known to have come from southern Illinois: 13 (30.23%)
Number of voters who are known to have come from Chicago: 2 (4.65%)
Number of voters who are known to have come from outside Chicago or southern Illinois: 16 (37.21%)
Number of voters of unknown origin: 12 (27.91%)
Note: The above statistics include all voters, regardless of the validity of their votes. Voters’ locations were determined by their email addresses and the content of their messages. Also note that while Chicago has only 2 known voters, it is certainly possible that all 12 voters of unknown origin were actually from Chicago. This, however, seems unlikely.


True tree/doppelganger pairs, in descending order of their winning percentages:

Analysis: Note that pair 1D was the only one to receive a true majority of the vote. The percentage of the vote each true pair received descends at a fairly regular slope: about 2-5% points between each pair and the next. This tends to indicate that while the public was more certain about some pairs than others, there was no major break-off point where the pairings suddenly became untenable; instead, they very slowly became less certain. However, the difference between the voting percentages of pairs 1D and 3B, at approximately 50%, and pair 9F, at less than half of that, is certainly noteworthy.


The percentages by which the true pairs beat their nearest rivals, in descending order,
by doppelganger:

Analysis: There is a major split between the three pairs with very high winning pecentages, 1D, 3B, and 7I, and the other six with much lower percentages. It seems that the public was very certain that D was the Doppelganger for Tree 1, B for 3, and I for 7. The extremely low cushion between the rest of the true pairs and their runners-up, however, as seen in the bloc of 8G, 2A, 6E, 4C, 9F, and 5H, when looked at in comparison with the generally strong winning percentages listed at the top, shows that while the public seemed to narrow in on certain general possible sets of trees and doppelgangers (that Tree n definitely went with either Doppelganger x , y or z, but certainly not any of the others), the actual specific pairs (whether it was actually Doppelganger x or Doppelganger y or Doppelganger z) were often chosen by a fairly small margin. (Pair 5H’s 0% winning percentage will be explained below.) It is also worthwhile to note that pairs 1D, 3B, and 7I received some of the highest absolute winning percentages and all of the highest buffering percentages above their nearest rivals. This indicates that these three pairs were truly in a bloc of their own. 2A, on the other hand, received a very high absolute percentage, but still only beat its rival by a comparitively small amount, indicating that there was fierce competition between two--and only two--rivals for Tree 2.

by tree:

Analysis: The statistical blocs here seem to be pairs 1D, 3B, 2A; pair 7I; pairs 6E and 5H; pairs 4C and 8G; and pair 9F. Once again, the top three pairs scored significantly higher than the six below them. There are a few interesting things of note here. The first is that, confusing as it may seem, the public seemed less certain about which doppelganger went with which tree than it was about which tree went with which doppelganger!

The second, as seen in the zero or negative percentages in pairs 4C, 8G, and 9F (as well as 5H in the preceeding set of statistics), is that some true pairs could not be determined solely through a simple predominance of votes, but had to be looked at in relation to other pairs that were more clearly true pairs. Therefore, while pair 8G had more votes than 8C, it had the same number of votes as 6G. However, pair 6E had more votes than 6G, so Tree 6’s Doppelganger was clearly E. Thus the 8G/6G tie was broken in favor of 8G. The cases of pairs 4C and 5H are similar. In the case of 9F, while it had fairly significantly less votes than 4F, it had slightly more votes than 9E. 4F, meanwhile, had slightly less than 4C, so in the end the somewhat unsatisfactory--but still logical--choice was 9F. This also made sense because it happened to leave each doppelganger with at least one tree, and each tree with only one doppelganger. However, this last point, it must be noted, was solely a byproduct of the just-outlined series of choices, and not a reason for, or influence on, those choices.

The last thing of note about this set of statistics is that it helps us understand, to a certain extent, the way in which voters made up their minds about which tree went with which doppelganger. It is clear, from the above 3 sets of statistics, that the trees with lower numbers and the doppelgangers nearer to the beginning of the alphabet were chosen with very high certainty--1D, 2A, 3B, 4C. It is thus possible to believe that voters’ creation of these pairs was influenced by their numeric and alphabetic order. This would seem to indicate that the trees and doppelgangers were not paired together because they were true tree/doppelganger pairs, but solely for convenience. The fact that the last tree has a negative percentage associated with it could be seen to bolster this opinion. However, the strong votes for 7I, which did exceedingly well in all of the above statistics, seems to counter this argument, and to point to the fact that the first four trees and doppelgangers simply happened to go with each other. For more on this, see the analysis of the trees’ and doppelgangers’ respective physical characteristics, below. The negative percentage associated with 9F also turns out to be a red herring, because its negative percentage is actually related to Doppelganger F, not Tree 9--Tree 9 is clearly linked with Doppelganger F, it is just the nature of Doppelganger F’s relationship with Trees 4 and 9 which is uncertain. Therefore, the statistics above seem to indicate that while some tree/doppelganger relationships were somewhat ambiguous than others, each pairing is, in the end, both the logically inevitable outcome of the vioting process, and also due to credible physical and metaphysical analysis, not accident or convenience. To repeat: each pairing seems to have been chosen, not randomly, but because the public truly did believe in and understand the concept of tree doppelgangers, and was in fact able to determine which tree in particular went with which doppelganger.


Number of votes received, by pair:

Analysis: From this it can be seen that votes were distributed fairly evenly within the following groups: 12 to 20 votes; 4 to 10 votes; 3 votes; 2 votes; and 0 to 1 votes. The clearest statistical aberrations are the 26 pairs which received 2 votes and the 11 pairs which received 3 votes. Together these make up 45.68% of the total number of possible pairs, but only 23.04% of the 369 total votes cast in this process. The top statistical voting bloc, from 12 to 20 votes, consists of only 8 of the 81 possible tree/doppelganger pairings, or 9.88%, but those 8 pairs have a total of 127 votes, or 34.42% of total votes cast. This may be seen as indicating that approximately half of the possible pairings were generally understood to be highly unlikely, thus garnering very low votes (0-3), while the 8 most popular pairings were seen as very strong pairs, as seen in their garnering of approximately a third of the total vote. Again, this is a strong indication that the public understood what a tree doppelganger is, and was able to find strong relationships between the trees and the doppelgangers.


Tree/doppelganger pairs that had 0 votes:
1C, 2D, 2F, 5B, 5C, 9C

Analysis: Doppelganger C had the most trees with which it was not paired at all (1, 5, and 9). This may partly be attributed to the fact that it had 2 very heavily-voted-for Trees (2 and 4), which together received over 65% of its votes. Thus it seems that its lack of links with trees 1,5, and 9 comes from the certainty that it was linked with either Tree 2 or 4.

Trees 2 and 5 both had two Doppelgangers with which they were not linked (D and F, and B and C, respectively). Tree 2 had overwhelming votes for Doppelgangers A and C, especially A, which together received over 75% of its votes. Therefore, its situation seems similar to that of Doppelganger C. However, Tree 5 had a different statistical spread. Its votes were much more evenly spread out, evidencing a possible lack of conviction of the part of the voters, so it seems unusual that two of its possible mates were still not considered to be viable options.


Analysis of Each Voter’s Percentage of Votes for the True Tree/Doppelganger Pairs:

Number Correct	Number of Voters	Percent of Voters	Percent Correct
0	2	4.88%	0%
1	6	14.63%	11.11%
2	10	24.39%	22.22%
3	9	21.95%	33.33%
4	6	14.63%	44.44%
5	3	7.32%	55.56%
6	2	4.88%	66.67%
7	1	2.44%	77.78%
8	0	0%	88.89%
9	2	4.88%	100%

Analysis: There is an almost perfect curve that quickly moves up from 0 correct through the maximum of 2 and 3 correct, then back down to 7 correct at almost the exact same rate, with the slight exception of the 2 people who were very impressively able to get all 9 correct. Note that it is impossible to get 8 out of 9 correct in a matching exercise such as this. 19.51% of the voters were able to correctly perceive at least half of the true tree/doppelganger pairs, which is exactly the same percentage of voters that got only 0 or 1 pair correct. While this statistic initially seems the most damning for the idea that people were able to correctly perceive the nature of the tree/doppelganger pairs, and that responses were made at random, certain aspects of these results bear a closer look. First, if responses were made at random, the number of true pairs each voter correctly chose would be expected to be spread evenly across the above chart--i.e., there would be approximately 4.5 voters each who got 0 pairs correct, 1 pair correct, 2 pairs correct, 3 pairs correct, etc. As it stands, it seems much more likely that the curve is as geometrically perfect as it is because of the nature of the task. In other words, this was an admittedly difficult task, and it should be expected that each voter would only be able to correctly perecive the true nature of a small number of the pairs; this is why the project was realized in a statistical manner in the first place, instead of having only a few experts try to solve the problem. Therefore, it seems that the difficulty was such that each voter could only be expected to get 3 or 4 correct, but that, when looked at across the general population, that chance went up or down according to each voter’s individual abilities. When looked at in this way, it can be seen that while 46.34%, almost half, were able to get 2 or 3 correct, the number who got less than this average, 19.51%, is much less than the number who got more than it, 34.15%. There seems to be no other way to understand the unusual curve found in this chart other than this: that it must come from the nature of the task, and thus either the nature of the voters or of the trees and doppelgangers, or of a combination of the two.


A subjectively-chosen set of general categories of physical tree/doppelganger characteristics (this is not an exhaustive list of either categories or their member elements; also, these characteristics may not all be found in this particular set of trees and doppelgangers):
Height: Small, Medium, Large
Width: Thin, Average, Wide
Shape: Triangle, Round, Oval, Square, Rectangle, Mushroom, Uncategorizeable
Color: Blue, Blue-Green, Green, Light Green, Dark Green, Purple, Red-Purple
Age: Young, Adult, Old
Taxonomy: Angiosperms (deciduous), Gymnosperms (conifers)

Subjective analysis of the relationship between each true tree/doppelganger pair in respect to the above categories:
1D: Tree 1 and Doppelganger D are both medium height, average width, triangular, green, adult angiosperms. They thus share 6 out of 6 possible physical characteristics.
2A: Tree 1 and Doppelganger A are both small, wide, semi-triangular angiosperms. Tree 1 is young and light green while Doppelganger A is young and purple. They thus share 4 out of 6 characteristics.
3B: Tree 3 and Doppelganger B are both medium height, average width, triangular, adult angiosperms. Tree 3 is light green while Doppelganger B is blue-green. They share 5 out of 6 characteristics.
4C: Tree 4 and Doppelganger C are both small, generally triangular youth. Tree 4 is a thin, dark green gymnosperm, while Doppelganger C is a wide, green angiosperm. They share 3 out of 6 characteristics.
5H: Tree 5 and Doppelganger H are both off-triangular trees. Tree 5 is a tall, wide, green, old gymnosperm, while Doppelganger H is a small, average-width, purple, young angiosperm. They share 1 out of 6 characteristics.
6E: Tree 6 and Doppelganger E are both thin, off-rectangular, angiosperms. Tree 6 is small, dark green, and young, while Doppelganger E is large, green, and old. They share 3 out of 6 characteristics.
7I: Tree 7 and Doppelganger I are both wide, off-round, adult angiosperms. Tree 7 is tall and light green while Doppelganger I is medium height and dark green. They share 4 out of 6 characteristics.
8G: Tree 8 and Doppelganger G are both small, thin, green, young angiosperms. Tree 8 is of uncategorizeable shape, while Doppelganger G is oval. They share 5 out of 6 characteristics.
9F: Tree 9 and Doppelganger F are both medium-width. Tree 9 is a large, uncategorizeably-shaped, light green, old angiosperm, while Doppelganger F is a small, triangular, dark green, adult gymnosperm. They share 1 out of 6 characteristics.

Analysis of the analysis: When compared to the percentages by which these pairs were chosen, the above characteristics have some relevance (as can be seen in the relationship that exists between the scores pairs 1D and 9F received from the above analysis and in the percentages they received from the public’s vote), but also seem to indicate that physical characteristics were not the only characteristics by which pairs were chosen. For instance, 8G seems to have a very high correspondence between physical characteristics, but had a fairly low share of the voters’ confidence. Therefore, this data seems to support the idea that voters chose their pairs not simply because of physical characteristics, but also because of qualities that can only be described as contextual or metaphysical.

However, the following proposed hierarchy of physical characteristics seems to hold true:
Shape: Very important. Besides 8G, all other pairs are at least generally the same shape.
Width: Important. Besides 4C, all other pairs are at least generally the same width.
Age: Somewhat important. Several pairs have different ages, but they do not differ by an extreme amount.
Height: Not so important. Several pairs have different heights, the difference between which is often fairly extreme--for instance, 5H, 6E or 9F.
Color: Not so important. Note that despite the extreme color difference found in 2A, it had one of the highest voting percentages of all the true pairs.
Taxonomy: Not important. Note that neither of the gymnosperm trees is paired with the only gymnosperm doppelganger.

The following statistics are from awstats:

Number of times visitors looked at large-scale pictures of trees, doppelgangers, and billboard: 94,894
Number of times visitors looked at video of billboard: 59
Number of Visitors who added the site to their Favorites list: 61


Number of visits by day:

Analysis: Note the huge increases in traffic on October 5 and 17. These increases largely seem to be due to the fact that on those dates, links to this site were posted on several other sites (I did not participate in these postings, they were done by other people acting autonomously). These other sites are well-known sources for web users to find lists of popular and unusual web destinations. Therefore, it can be seen that a very large percentage of this site’s traffic was due to inter-internet traffic (people who saw descriptions of this site listed on another site), as opposed to traffic from people who logged in directly to this site due to outside stimuli. This is in turn an interesting comment on the price of media, since there was no cost involved in having this site listed on the other sites. However, it can also be seen in the statistics below that most people did not stay long enough to vote, so perhaps the “inter-internet” traffic was not really “keeper” traffic--traffic that would guarantee actual involvement rather than just quick browsing. It may be hazarded that those people who took the time to remember the unusually-long address for this website after seeing it on a billboard or poster, and then logged in when they got they chance later in the day, would have more personally invested in the situation, and would thus be more likely to vote.


Average number of pages looked at, by day of the week:

Average number of pages looked at, by hour of the day, military time:

Amount of time each visitor stayed at the site:

Analysis: It is worth noting that once people decided to stay, there is a fairly gradual slide from the number of people who stayed from 30 seconds to 30 minutes, which is fairly impressive. This seems to indicate that if the site was attractive to people’s taste at all, it would interest them enough for a fairly involved investigation. The fact that 14 people stayed for more than half an hour is a stunning fact. It is also interesting to note that the number of the people who stayed over 5 minutes--61--is fairly close to the number of total voters--70.49% of them.


Links from an external page (other web sites except search engines):

Keyphrases used on search engines that led to the site:

Keywords used on search engines that led to the site:

The following statistics are from webalizer:

Visitors’ origins:

(“Other” includes, in descending order: United States, Netherlands, Japan, Denmark, New Zealand (Aotearoa), Germany, France, Finland, Singapore, Belgium, Sweden, Switzerland, Norway, Israel, Mexico, Brazil, Italy, Hungary, Iceland, and Estonia)

The following statistic is from EMC (Electronic Marketing Company):

Daily vehicular advertising impressions for the billboard: 29,200

Analysis: It is interesting to consider which method of dissemination received the most “impressions,” as understood in the advertising world. The billboard had 29,200 impressions a day, which makes for 905,200 impressions over the entire month. The number of impressions for the smaller posters in southern Illinois cannot be precisely measured, but with 100 signs, one can guess that at least 20 people saw each sign each day, for a total of 2000 impressions a day and 62,000 in the month. However, it is likely that many of these signs were taken down before the end of the month, so that estimate is more likely to be around 40,000. The number of people exposed to this website through the various postings that were set up on other sites is also impossible to estimate, but they could be tens of thousands a day. However, since most of these postings would only be active or obviously accessible for a short period, most likely a day or two, the total web impressions are probably only in the range of 150,000.

If you have any question about this project or any other, please contact Chris Wildrick, the project manager, at chris.wildrick at gmail.com. Thank you for your interest!