Spotlight on stylometric text analysis - The WW2 Years

word cloud

It has long been speculated that some stories with "Mandrake" and "The Phantom" may have been written by a ghost writer during Lee Falk's military service. And that the ghost writer could be Alfred Bester.

Cambridge Summer Theatre

A somewhat strange coincidence is that in the Starman story "The Time-Machine Crime" (Adventure Comics 080, November 1942), where Bester is listed as an author on GCD, the villains use a time machine to kidnap William Shakespeare. They trick him into helping with the plan to rob the wealthy audience during a performance at the Cambridge Summer Theatre. A theatre co-sponsored by John Huntington and Lee Falk at that time. The theater staged Shakespeare's "Othello" the week of August 10, 1942, and it is reasonable to believe that the story of "The Time-Machine Crime" was inspired by this production.

There are anecdotes that mention something about Alfred Bester being a ghostwriter for some Mandrake and the Phantom stories, while Lee Falk was in the army. And there are several who have tried to identify which stories these could be. This article takes a deeper dive into the issue by seeing whether stylometric analysis can provide some answers.

Analyses can be done (and have been done) with various software such as JGAAP, Voyant Tools and the 'stylo' package in RStudio. But in this article, similar analyses are tested using ChatGPT 4o.

Mandrake and the Phantom

In the years 1942 and 1943, Lee Falk was Chief of Radio of the Foreign-Language Division OFF/OWI. And in the years 1944 and 1945 he was a private in the army, according to him, in the signal corps. To visualize these two time periods, two timelines can be set up:

Timeline 1942-1943: Mandrake the Magician and The Phantom Stories

	1942																								1943
	Jan		Feb		Mar		Apr		May		Jun		Jul		Aug		Sep		Oct		Nov		Dec		Jan		Feb		Mar		Apr		May		Jun		Jul		Aug		Sep		Oct		Nov		Dec
MD	Lothar the Champ													The Rumor Factory				Baron Kord														The Witch of Kaloon												The Earthshaker
MS	The Hungry Isles									Pacifica													Mystery of the Girls with Red Hair								Cloud City																Gloria Golden
PD			The Inexorables																									Bent Beak Broder								The Phantom's Engagement				High Seas Highjacker
PS					The Impostor															Castle in the Clouds												The Ismani Cannibals					Hamid the Terrible
Lee Falk	Chief of Radio of the Foreign-Language Division OFF/OWI																																								writing

Timeline 1944-1945: Mandrake the Magician and The Phantom Stories

	1944																								1945
	Jan		Feb		Mar		Apr		May		Jun		Jul		Aug		Sep		Oct		Nov		Dec		Jan		Feb		Mar		Apr		May		Jun		Jul		Aug		Sep		Oct		Nov		Dec
MD			The Dome									Doctor Congo								The Mirror People				The Ice Lady							The Old Ones						The Mysterious Prince							The Sleeping Beauty
MS			The Garden of Wuzzu							The Circus Adventure							The Santa Claus Pirates						Fountain of Youth							Kingdom of the Wind							The Atalan Deep									The Twins of Karana
PD					Diana					The Crooner							The Maharajah's Daughter														The Blue Gang					Lago the Lake God					The Wild Girl					The Mermaids of Melo Straits
PS	Hamid the Terrible												The Childhood of the Phantom													The Golden Princess											The Strange Fisherman
Lee Falk	writing				Fort Devens and other Army Training Camps																														??

An anecdote by Bester tells of a meeting with Lee Falk. In this, Bester mentions that he ghost-wrote for Falk. He also mentions which house he lived in and that he had original pages with Green Lantern on the wall. The last two pieces of information allow the anecdote to be placed somewhere between the fall of 1943 and the spring of 1944.

It must also be remembered that in the timelines above, the stories are inserted as they were published in the newspapers. However, the scripts themselves must have been written at least two to three months before this time, due to deadlines for the artist and follow-up work at KFS.

The Corpus

Mandrake Dailies: MD_030 "The Earthshaker", MD_031 "The Dome", MD_032 "Doctor Congo", MD_033 "The Mirror People", MD_034 "The Ice Lady", MD_035 "The Old Ones", MD_036 "The Mysterious Prince", MD_037 "The Sleeping Beauty", MD_038 "Pleasant Valley".

Mandrake Sundays: MS_012 "The Ghost Bear of Glass Mountain", MS_013 "The Theatre Mysteries", MS_022 "Pacifica", 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", MS_029 "Fountain of Youth", MS_030 "Kingdom of the Wind", MS_031 "The Atalan Deep", MS_032 "The Twins of Karana".

Phantom Dailies: PD_016 "The Inexorables", PD_017 "Bent Beak Broder", PD_018 "The Phantom's Engagement", PD_019 "High Seas Highjacker", PD_020 "Diana", PD_021 "The Crooner", PD_022 "The Maharajah's Daughter", PD_023 "The Blue Gang", PD_024 "Lago the Lake God", PD_025 "The Wild Girl", PD_026 "The Mermaids of Melo Straits", PD_027 "Princess Valerie".

Phantom Sundays: PS_001 "The League of Lost Men", 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", PS_015 "King of Beasts".

Analysis

The groups MD, MS, PD and PS are written slightly differently and to be adapted to two different characters and two different publishing media (daily newspapers and in Sunday newspapers as a comic attachment), and analyses are performed for each of the four groups.

The goal of the analyses of the stories is to investigate whether there are features that indicate the likelihood of a ghost-writer for some of them.

1. Overall analysis

Several stylistic and narrative analyses have been done to uncover Lee Falk's writing style, including analyses related to three-act structure, average sentence length, and frequency of modal expressions, action verbs, and exclamations. This is to try to find out if parts stand out, which could indicate a ghost-writer.

Result
Mandrake dailies
The stories are evenly balanced in both their three-act structure and language. The use of modal expressions and exclamations is low, and the act distribution is quite even across the stories. This indicates a uniform authorial signature and high narrative stability.
Mandrake Sundays
The stories are distinguished by a high average number of words per line and a high proportion of modal expressions and exclamations. Several stories also have a deviating act distribution. This may indicate script changes or possible ghost-writing in parts of the corpus.
The Phantom dailies
The stories in this category are the most structurally and linguistically stable. The distribution of acts is even, and the use of language varies little. This suggests an author and a fixed writing model over time.
The Phantom Sundays
In these stories there are some episodes with a high use of modal expressions and exclamations, as well as an uneven act structure. Some stories stand out as possibly written by a different hand, especially PS_009 and PS_010. Overall, this gives reason for further investigation.
Overall assessment
Across the analyses, we see that PD and MD appear consistent in both language and structure. MS and PS have several stories that break the pattern - with a higher degree of emotional language and structural deviations. This applies especially to PS_009 "The Ismani Cannibals", PS_010 "Hamid the Terrible" and MS_029 "Fountain of Youth". The combination of three-act breaks and stylistic peaks points to possible ghost-writing or other authorial influence in these stories.

• 
  MS
• 
  PS

2. PCA - stylometric features for the four subcorpora

A promt asking for a PCA analysis on stylometric features (Sentence length, Word length, Word variation, Use of exclamations) for the four groups. The main purpose here is to find out if any of the stories within the groups stand out significantly, and then to conduct further analyses focusing on these.

Result

• MD: Clearly defined clusters with MD_031 and MD_038 as outliers. The remaining MD stories are closely clustered.
• MS: More dispersed than MD. MS_022 and MS_012 stand out with distinct style.
• PD: High variation, with PD_016 and PD_021 far removed from the others.
• PS: Large distance for PS_010, and to a lesser extent PS_005 – both distinctly different.

Examining the results

• MD: MD_031 "The Dome" (January 31, 1944 - June 24, 1944) might be interesting to look into further. MD_038 is outside the WW2 period.
• MS: MS_022 "Pacifica" (April 19, 1942 - November 1, 1942) might be interesting to look into further. MS_012 is outside the WW2 period.
• PD: PD_016 "The Inexorables" (February 2, 1942 - January 9, 1943) might be interesting to look into further, as well as PD_021 "The Crooner" (May 22, 1944 - August 26, 1944).
• PS: PS_010 The Marshall Sisters Pt.3 "Hamid the Terrible" (July 11, 1943 - June 25, 1944) might be interesting to look into further. PS_005 is outside the WW2 period.

An analysis of a thematic overview for each group, based on the text's word usage and expressions (including keywords, focus of action and any characteristic expressions) does not fully confirm the findings in the PCA analysis.

A PCA analysis of stylistic features grouped by thematic cluster for the entire corpus also does not provide any clear indication that there are individual stories that stand out significantly.

ChatGPT - 4o: PCA - rolling.classify()

In the 'stylo' package in RStudio there is a rolling.classify() funktion. This function can also be promt for in ChatGPT.

The texts in the MD, MS, PD and PS groups are joined together to create four new text. One PD text, one PS and on.

The rolling.classify() function divided each texts into sliding sections (windows) of 1000 words, with steps of 500 words. The analysis was done for three stylometric measures:

• TTR: The Type-Token Ratio is a simple measure of lexical diversity in a text, calculated by dividing the number of unique words (types) by the total number of words (tokens). It indicates the richness or variety of vocabulary used in a text.
• MSTTR: Mean Segmental Type-Token Ratio is a lexical diversity measure that calculates the average TTR across multiple segments of a text. It helps to evaluate the richness of vocabulary used in different parts of a text.
• HDD: Hypergeometric Distribution D is the vocabulary variation based on word selection in text windows.

The goal is to see if any of the windows stand out from the rest of the text, which may indicate a ghostwriter in parts of the text.

Result
Mandrake dailies
The TTR variability in the text is generally moderate and relatively stable. A few peaks indicate increased lexical variation, but these tend to occur in longer stories. There is no clear indication of author change or marked deviations.
The text appears with high MSTTR and HDD throughout. A clear low point in HDD partially coincides with lower MSTTR, which may indicate a simpler, more repetitive sequence. Overall, both measures support a consistent authorial signature.
Mandrake Sundays
The TTR variability in the text shows significantly greater variation than for the dailies. Individual stories such as MS_012 and MS_022 are associated with windows with high TTR, which may be due to dialogue-heavy or varied language use. MS_018 shows clearly low TTR, which may reflect a more action-driven narrative style. Some stylistic breaks may indicate editorial or author-driven differences.
The text shows more dynamics in both MSTTR and HDD measures. Several local lows in MSTTR and HDD point to text passages with less variation and possible shifts in writing style. This may indicate either thematic repetition or the involvement of multiple writers.
The Phantom dailies
The Phantom daily strips generally have low TTR, with several long windows with an almost flat profile. This points to a more repetitive, firmly anchored language - something common in serial structure. PD_016 and PD_020 give small peaks, but otherwise there is little to suggest stylistic break or variation between writers.
The text is the most consistent, with small variations in both MSTTR and HDD. This confirms a homogeneous, formulaic style and provides little basis for suspecting ghostwriting in this material.
The Phantom Sundays
The Phantom Sunday Pages stand out with the most obvious stylistic shifts. PS_008, PS_009 and PS_010 have a complex structure and high TTR, which may suggest both thematic depth and a different style. This may point to an alternative author or a different writing model precisely during this period.
The text has clear lows in both HDD and MSTTR, sometimes in agreement. This may indicate episodic sequences with less linguistic variation, possibly written with a different writing technique or authorial signature at times.

Same character - two different authors

The Starman Corpus

These stories were published in the "Adventure Comics" 1938 series by DC Publication. In the issues 65-80 (Aug 1941 to Nov 1942)

Starman-AC_065: "The Mystery of the Undersea Terror", Starman-AC_066: "The Case of the Camera Curse", Starman-AC_067: "The Menace of the Invisible Raiders", Starman-AC_068: "The Blaze of Doom", Starman-AC_069: "Adventure of the Singapore Stranglers", Starman-AC_070: "Adventure of the Ring of Hijackers". Starman-AC_071: "The Invaders from the Future", Starman-AC_072: "The Case of the Magic Bloodstone", Starman-AC_073: "The Case of the Murders in Outer Space", Starman-AC_074: "The Case of the Monstrous Animal-Men", Starman-AC_075: "The Strange Case of the Luckless Liars", Starman-AC_076: "(The Case of the Sinister Sun)", Starman-AC_077: "Finder's Keepers", Starman-AC_078: "The Little Man who wasn't There", Starman-AC_079: "The Tune of Terrific Toby", Starman-AC_080: "The Time-Machine Crime".

AT GCD issues 67, 68, 78, 79 and 80 are credited Alfred Bester while the others are credited Gardner Fox.

Analysis

The goal of the analyses of the Starman stories is to investigate whether there are features that indicate the likelihood of who wrote which stories.

ChatGPT - 4o

Overall Overview: Authorship Analysis

The stories were uploaded to chatgtp. It was written that Gardner Fox was the author of most of the stories but that there was also an unknown author on some of them. Could the theme and language suggest something about which stories were written by whom?

issue	GCD	chatgtp - author?	chatgtp - comment
65	Fox	Fox	Technological threat from the sea, classic Fox.
66	Fox	outside Fox	Occult and voodoo theme, little science.
67	Bester	Fox	Inviso solution, FBI collaboration – familiar Fox element.
68	Bester	Fox	Patriotic and technical, typical Fox.
69	Fox	Fox	War, sabotage, Asian setting – Fox theme.
70	Fox	Fox	Good mix of railway drama and sabotage.
71	Fox	Fox	Time travel and technology, familiar Fox plot.
72	Fox	outside Fox	Supernatural/mythological story, weaker logic.
73	Fox	Fox	Very technical (rocket trajectories + law!), strong Fox style.
74	Fox	doubtful/Fox	Animalization machine is borderline Fox – parodic.
75	Fox	Fox	Hypnosis, crime and rational explanation of supernatural.
76	Fox	Fox	Solar fantasy with explained planet enemies, classic Fox.
77	Fox	Fox	Ghost character (The Mist), familiar from Fox’s Starman writing.
78	Bester	outside Fox	More like parody, Scooby Doo-like structure, Children + comedy.
79	Bester	outside Fox	Humor and parody. Very uncharacteristic for Fox.
80	Bester	Fox	Shakespeare and time machine, rational logic and heavy Fox feel.

Gardner Fox's stylistic features

Based on analysis of Fox-proof stories (#65, 67–71, 73, 75–77, 80):
Linguistic tone: Combines serious, technical prose with dry one-liners and patriotic comments.
Science and explanation: Fox always attempts to provide a technical explanation (even if it is pseudoscientific) – as in the time machine, rocket launchers (#73), invisibility solution (#67, 77) and rays (#76).
Starman role: Ted Knight appears as a heroic but weak social figure, who is activated through "summons". The internal conflict (playboy vs. hero) is Fox's work.
Narrative moves:
Stories often start with news reports or a public threat.
Starman is informed through Woodley Allen or radioactive signals.
Stories often end with a moral lesson or logical explanation.

Stories with a possible different author

The following stories break with the Fox style in several areas:
issue 66 – "Camera Curse": Focuses on magic cameras, voodoo, mystical powers – no technology.
issue 72 – "Magic Bloodstone": Uses Arabic mysticism and magic stones as a propulsion – very pulp-like.
issue 78 – "Little Man who wasn’t there": Scooby Doo-like comedy with a child sidekick (“Mike Muggins”), very little of Fox’s technical or logical approach.
issue 79 – "Tune of Terrific Toby": A sort of morality fable with heavy use of slapstick and self-deception – this is more of a social comedy than a superhero story.

Statistical Stylometry

A more specific query against stylometric features in Fox stories versus non-Fox stories confirms the findings in the first analysis.
Average Sentence Length
The sentences are on average shorter in stories attributed to Gardner Fox than with those by other suspected authors.
Average word length
Fox stories tend to use somewhat longer words.
This indicates a somewhat higher level of abstraction and explanatory language, typical of Fox's technical and explanatory style.
Type-token ratio (linguistic variation)
Fox stories clearly show higher lexical variation.
This suggests a more varied and precise use of words – consistent with Fox's habit of scientific terminology and differentiated narrative voice.
Total number of words per story
Fox stories tend to be somewhat longer, but not extremely so.
This may indicate that he emphasized developing plot and explanation rather than just action.
Technology frequency: Fox vs. Others
Fox stories have a significantly higher incidence of technological terms.
This is in line with Fox’s well-known interest in explaining Starman’s powers and enemies’ devices via pseudoscientific logic.
Non-Fox stories use technology less or only as a backdrop, not as a motor in the plot.
Modal verbs (must, should, might, etc.)
Fox stories have a higher use of modal expressions.
This suggests that the characters in Fox’s stories more often express logic, duty and possibility – e.g. "I must act fast!" or "He could be dangerous!"
This fits with Fox's analytical and explanatory writing style.
Action verbs (hit, run, fly, etc.)
Here the difference is more even.
Both Fox and the other authors use high-frequency action verbs, but there is a slight tendency for other authors to use even more physical drama.
This may indicate that non-Fox stories place a greater emphasis on visual comedy and exaggerated physical conflict.
Exclamations (wow, yow, ugh, etc.)
The non-Fox stories use far more exclamations.
This points towards a more cartoonish-parodic or "pulp comic" tone - as in the stories with "Terrific Toby" and "Little Man Who Wasn't There".
Fox's style is generally more controlled and matter-of-fact, even in exaggerated situations.

• 
  Average sentence length
• 
  Average word lenght
• 
  Type-Token ratio
• 
  Word count per story
• 
  Technical terms usage
• 
  Modal verb usage
• 
  Action verb usage
• 
  Exclamations usage

PCA

Asking chatgtp to make a PCA analysis of the Starman corpus to visualize a 2D view of stylistic differences between the stories.

The PDA figure shows that the assumed Fox stories form a clear cluster on the left – with a relatively consistent stylistic signature. The non-Fox stories, assumed by Bester, (#66, #72, #78, #79, and partly #74) lie further out, especially #78 and #79, which stand out as linguistically and thematically different (distinctly humorous/parodic). Story #74 (“Animal-Men”) is close to the Fox cluster, which is consistent with the assessment of borderline case.

• 
  PCA

Summary

Total number of stories analyzed by ChatGPT: 16 (AC #65–#80). Strongly indicated Gardner Fox: 11 stories. Probably written by another author, Alfred Bester, issues 66, 72, 74*, 78, 79. Note that #74 is a borderline case, and may be a collaborative project or a careful experimentation by Fox himself.

ChatGPT's analysis of the presumed author for the stories differs from who is credited on GCD for six of the stories, something that should be investigated further.

issue	title	GCD - author	chatgtp - author	chatgtp - comment
66	"The Case of the Camera Curse"	Fox	outside Fox	Occult and voodoo theme, little science.
67	"The Menace of the Invisible Raiders"	Bester	Fox	Inviso solution, FBI collaboration – familiar Fox element.
68	"The Blaze of Doom"	Bester	Fox	Patriotic and technical, typical Fox.
72	"The Case of the Magic Bloodstone"	Fox	outside Fox	Supernatural/mythological story, weaker logic.
74	"The Case of the Monstrous Animal-Men"	Fox	doubtful/Fox	Animalization machine is borderline Fox – parodic.
80	"The Time-Machine Crime"	Bester	Fox	Shakespeare and time machine, rational logic and heavy Fox feel.

Note: Most of these analyses and several others have been pre-analyzed using the 'stylo' package in RStudio and Voyant Tools. Those that were not analyzed with these two tools are post-tested in analyses with these two tools. Analyses using the 'stylo' package in RStudio and Voyant Tools are consistent with the findings in the analyses using ChatGPT.

This article, or section of an article, is very short. You can help MandrakeWiki by expanding it.

External links

• GCD: Alfred Bester
• GCD: Gardner Fox
• GCD: Issues with covers and longest comic story for series Adventure Comics (198 series)
• WikiPedia: Starman (Ted Knight)
• Don Markstein's Toonopedia: Starman. Archived from the original on October 23, 2017.