Spotlight on stylometric text analysis - Who wrote The Childhood of the Phantom
| Country: | |
|---|---|
It has long been speculated that "The Childhood of the Phantom", the Sunday story from 1944, may have been written by Alfred Bester who is mentioned as a possible ghost writer during Lee Falk's military service.
This article takes a deeper dive into the issue by seeing whether a stylometric analysis can provide some answers.
The childhood story, like other stories written by Lee Falk, follows the Three-Act Structure.
- The first act portrays the young Phantom's joyful upbringing in the Deep Woods. The first major turning point occurs when his parents inform him that he must go to America to receive education and learn about civilization.
- The second act delves into the young Phantom's life in America, with the midpoint marked by his successful adaptation to his new surroundings.
- The third act commences with the second significant turning point, when his father requests his return to the Deep Woods. The climax of the third act reaches its peak when his father tragically passes away.
Who wrote "The Childhood of the Phantom" ?
The Corpus
The Phantom Sundays
PS_002: "The Precious Cargo of Colonel Winn", PS_003: "The Fire Goddess", PS_004: "The Beachcomber", PS_005: "The Saboteurs", PS_006: "The Return of the Sky Band", PS_007: "The Impostor", PS_008: The Marshall Sisters Pt.1: "Castle in the Clouds", PS_009: The Marshall Sisters Pt.2: "The Ismani Cannibals", PS_010: The Marshall Sisters Pt.3: "Hamid the Terrible", PS_011: "The Childhood of the Phantom", PS_012: "The Golden Princess", PS_013: "The Strange Fisherman", PS_014: "Queen Pera the Perfect"
Mandrake the Magician Sundays
MS_012: "The Ghost Bear of Glass Mountain", MS_013: "The Theatre Mysteries", MS_023: "Mystery of the Girls with Red Hair", MS_024: "Cloud City", MS_025: "Gloria Golden", MS_026: "The Garden of Wuzzu", MS_027: "The Circus Adventure", MS_028: "The Santa Claus Pirates"
Comics written by Alfred Bester
Starman: AC-Bester_067: "The Menace of the Invisible Raiders!", AC-Bester_068: "The Blaze of Doom!", AC-Bester_078: "The Little Man Who Wasn't There!", Green Lantern: GL-Bester_10: "The Man Who Wanted the World!"
Comics written by Gardner Fox
Starman: AC-Fox_065: "The Mystery of the Undersea Terror!", AC-Fox_066: "The Case of the Camera Curse!", AC-Fox_069: "Adventure of the Singapore Stranglers"
Analysis
Voyant Tools
An analysis using the Voyant Tools shows that the corpus with 27 stories consist of 87,280 total words and 7,557 unique word forms.
In this corpus, the stories of Bester have a higher vocabulary density, a measurement of vocabulary usage in comparison to the length of a document. (Think of how many words will be read on average before a new word is encountered.)
| 1 | 2 | 3 | 4 | 5 | |
|---|---|---|---|---|---|
| Highest: | AC-Fox_069 (0.406) | GL-Bester_10 (0.405) | PS_009 (0.387) | AC-Fox_065 (0.382) | AC-Bester_078 (0.380) |
| Lowest: | PS_006 (0.198) | PS_010 (0.206) | PS_011 (0.223) | PS_007 (0.236) | PS_004 (0.236) |
Comparing the readability index the childhood story has the lowest score. The readability index is an estimation of how difficult a text is to read. The estimation is made by measuring a text's complexity. Measurable attributes of texts such as word lengths, sentence lengths, syllable counts, and so on give us ways to measure the complexity of a text. The Voyant Tools uses the Coleman–Liau index, and the output is approximates the U.S. grade level.
| 1 | 2 | 3 | 4 | 5 | |
|---|---|---|---|---|---|
| Highest: | MS_028 (7.863) | MS_026 (7.730) | MS_024 (7.378) | MS_027 (7.360) | MS_025 (7.222) |
| Lowest: | PS_011 (3.547) | PS_004 (4.493) | MS_012 (4.668) | MS_013 (4.767) | PS_014 (4.836) |
JGAAP
Using the Java Graphical Authorship Attribution Program we can compare an unknown text against known texts.
We adding the texts (exept the childhood story) from the corpus under the three authors: Falk, Bester and Fox. An analysis where the childhood story is our unknown text we choose the event driver: Character NGrams (n: 10), the event culling: Most Common Events (numevents: 50) and the analysis methode: Nearest Neighbor Driver with metric Alt Intersection Distance.
In an organized list from one to twenty-seven, this analysis suggests that this story was most likely written by the same author as:
- score 01 to 15: PS_003, PS_012, PS_002, PS_004, PS_007, PS_013, PS_008, PS_010, PS_014, PS_006, PS_005, PS_009, MS_027, MS_024, MS_026, MS_028,
- score 15: AC-Bester_067,
- Score 18 to 21: MS_013, MS_023, MS_012, MS_025,
- score 21 to 27: AC-Bester_078, AC-Fox_065, AC-Bester_068, AC-Fox_066, GL-Bester_10, AC-Fox_069
'stylo' package in RStudio
A stylometric text analysis of the Avon Novels using the 'stylo' package in RStudio shows that it is possible to use stylometry to identify the author of a book. The text in the comic strips are a bit different than a novels. In the novels the dialogues looking about this: "Hello," said the Phantom. But in the Sunday stories the speech bubble are more like this: Hello!
Two analysis of the Phantom Sundays were done: first 0-902 MFW 2-gram (fig. 1) and the second 0-902 MFC 3-grams (fig. 2). Both using the Boostrap Consensus Tree. The result for the MFW 2-grams grouping all stories close in writing style. The MFC 3-grams shows a slightly larger variation, but this is most likely due to the ammount of dialogues from different characters in the stories. The analysis shows no clear indication that anyone other than Lee Falk was the author of these stories.
A third analysis was done adding some Mandrake Sundays to the corpus. This analysis is a cluster analysis using the 100 MFW (fig. 3). Here one see a greater distance between the stories with the Phantom and Mandrake. But otherwise no indication of a ghost-writer.
A fourth analysis was done using the Phantom and Mandrake sundays and the "Starman" stories by Bester and Fox. This Boostrap Consensus Tree using 0-902 MFW with classic delta distance. This analysis (fig. 4) shows that the Starman stories are distinct from the Mandrake and Phantom stories. This suggests that the Mandrake and the Phantom texts are more closely related stylistically and represent the style of one author. The childhood story is placed closely together with the other Phantom stories, but not in isolation. This may indicate a general similarity in style with the other texts in the Phantom Sunday group.
The closest node to the childhood story is "Hamid the Terrible" and "The Golden Princess", strengthening the argument that the childhood story was also written by the same author. It is worth noting that the childhood story is clearly separated from the Fox and Bester groups, which weakens the possibility that the text was written by Bester.
A fifth analysis (fig. 5) is a Boostrap Consensus Tree using 100-500 MFW with classic delta distance. The childhood story clusters closely with the group of Lee Falk's other stories with the Phantom.
In the 1–1000 MFW figure, the childhood story is centrally grouped with the other Phantom Sundays. This suggests that across a wide range of most-frequent words, the stylometric features align well with Falk's typical patterns. In the 100–500 MFW figure, the childhood story remains in a similar position, reinforcing that its stylistic features are stable within this narrower range.
In both figures (fig 4 and 5), the distance (branch length) between the childhood story and the nearest Phantom Sundays is very short, indicating a high level of similarity. If the childhood story were written by another author, we would expect it to cluster farther away from the Phantom Sunday group or show closer affinities to other groups (e.g., AC-Bester or AC-Fox).
Based on figures 4 and 5, it is highly probable that "The Childhood of the Phantom" was authored by Falk. Its consistent grouping with the Phantom Sundays and its distance from works by Bester and Fox support this attribution. Further analysis with additional parameters (e.g., different stylometric measures or textual features) could provide additional confirmation, but the current evidence strongly favors Lee Falk as the author.
-
fig. 1
-
fig. 2
-
fig. 3
-
fig. 4
-
fig. 5