Tracing the Letters of J.R.R. Tolkien is a digital humanities project that aims to provide new methods of reading and understanding the main corpus of primary sources related to the titular author. One of the preeminent English writers of the past century, Tolkien’s letters illustrate a vast collection of the author’s insight into a myriad of matters – social, cultural, political, and, of course, literary. Through a text-mining approach with Voyant online software, this project organizes the letters into categorized data in order to sort through them for specific themes and issues related to his creative and writing processes as well as the ways in which Tolkien discussed them with different audiences. Specifically, the letters are divided between relevant letters to his family, his editors and publishers at Allen & Unwin, and his innumerable fans later in his career. Ultimately, this project seeks to aid Tolkien scholars with a relatively new way of approaching these important sources.

The dataset created for this project comes from The Letters of J.R.R. Tolkien: Revised and Expanded Edition, published in 2023 by HarperCollins and originally compiled and annotated by Humphrey Carpenter, Tolkien’s original biographer, and Christopher Tolkien. This text includes 354 letters written by Tolkien throughout his life, from a letter to his future wife Edith in 1914 to his final letter to his daughter Priscilla in 1973. Among these published letters, approximately half of them were chosen to be involved in the text-mining dataset. In narrowing the focus of this project, it was decided that each letter involved should connect in some way to Tolkien’s writing or creative processes. This scope encompasses any discussions involving the act of writing and topics pertaining to his writing, such as criticism, inspiration, or commentary. Hence, many of the letters that merely exchange formalities or discuss routine interpersonal matters are omitted from the dataset, as well as those discussing more technical issues with his publishers, such as pricing and the matter of overseas pirated editions. However, this selection process was done on a careful case-by-case method, as the ways Tolkien approached his creative decisions are varied across the letters; for example, the letters wherein he discussed his visions for cover art and how the narrative was to be divided into multiple books (with respective titles) are evidently tied to this project’s goal and are thus included. All in-text revisions and footnotes by the editors were deleted from the corpus, leaving only Tolkien’s original words, and each letter’s salutations were also excluded in order to avoid greeting words from skewing the results. Finally, the published collection includes some letter drafts that were written but never ultimately sent – these are also included so long as they pertain to the project goal, such as Tolkien’s extensive draft letter to an alternate publishing company which features an immense explanation of his writing and fictional histories. In summary, all letters in this dataset were chosen because they involve subjects related to Tolkien’s creative and writing processes. All others from the book were excluded because their subjects are not relevant to the project goal and provide no insight into the central question.

The methods of this text-mining project involve a cross-analysis between the selected corpus of data, divided into three categories: Family Letters, Publisher Letters, and Other Letters. The first category naturally includes letters from the dataset sent to members of Tolkien’s immediate family, usually correspondence with his sons. The second category includes letters sent to his publishers and those whom Tolkien chose to be involved in the editing process of his works, such as Rayner Unwin, as well as a few letters intended to be sent to alternate publishing companies. The third category includes letters to figures outside of the other two categories, such as fans, friends (including C.S. Lewis), and other persons interested in his work for specific reasons (such as inquiries from directors about dramatic adaptations). The third category features the most diverse collection from the dataset, involving so many different recipients, whereas the first category is largely centered around his correspondence with his son Christopher while writing The Lord of the Rings, and the second category is almost entirely fixed on his contacts in the Allen & Unwin publishing company. Each category reveals a rich variety of material regarding Tolkien’s views, methods, and beliefs, discussed in detail below.

All Letters

With the project methods transparently discussed, this project begins with a baseline analysis of the entire corpus of selected letters, pictured in full below:

From this overarching view, the Cirrus and Summary tools provide a cursory glance at the corpus. The Cirrus word cloud features some unsurprising names for this particular author, such as hobbit and elves, as well as his talks on God and men, and the word ‘English’ is fairly common given his role as a philologist, showing up in discussions on linguistic history and creative writing. These words were expected given the selection process for the dataset and the project’s focus on Tolkien’s writing, but the other prominent words showcase more sides to the subject. For example, the word ‘Children,’ appearing 106 times across the corpus, is seldom used in reference to his family, but often in relation to how The Hobbit is written with children in mind whereas The Lord of the Rings is not, as well as discussions both religious (‘children of God’) and literary (children as written in his books). This tool also points out how the term ‘story’ is a common way in which he refers to his writing. ‘History’ is also a very important word for Tolkien, as evidenced by its use in both scholarly talks in his field and the histories that he writes of in his books. The Summary tool points out that interestingly the most common word shared across the letters is “world:”

By using the Contexts tool below, readers can see that Tolkien’s usage of this word is often split between his commentary on the physical world of human civilization, often employed in his talks with his son during the second World War, and his way of referring to his fiction, the world of Middle-earth; Tolkien himself defines this dichotomy as the ‘Primary World’ and the ‘Secondary World,’ the latter of which encompasses his fiction.

In viewing the corpus altogether, this text-mining approach does not bring to the forefront much in the way of surprises. The most common words are rather expected, involving names of characters and words important to Tolkien’s interests, like ‘history’ and ‘english.’

Family Letters

The first specific category involves letters sent to his family members. The full text-mining results are pictured below:

In comparison with the entire corpus, this category’s Cirrus tool highlights some different emphases unique to letters with his family. Words like ‘love,’ ‘life,’ and ‘young’ are more prominent here than the other two categories, which is not unexpected with personal correspondence. Specific terms from his writing, such as ‘hobbit,’ are not nearly as common here, but verbs such as ‘wrote’ and ‘work’ as well as nouns like ‘book’ and ‘chapter’ make up the difference. Using the Contexts tool, Tolkien often refers to his creative writing as his ‘work,’ both as a literary endeavor and in the active sense of writing it.

The Cirrus tool shows a more personable vocabulary in this category, which is to be expected, although the acronym ‘c.s.l.’ for his friend C.S. Lewis appearing with relative frequency was a surprise, which should illustrate a level of importance to that connection to have been brought up in some frequency. As a profoundly religious person, the name ‘God’ shows up frequently, and although it is used more in the larger Other letters category, in Family letters it is far more often in reference to his beliefs specifically rather than in connection with his writing.

In total, this category’s results are not wholly surprising, but they do show which parts of his writing he sought to discuss with his son Christopher in the writing process, as much of this category consists of letters he sent during World War II describing chapters he was working on while his son was at war.

Overall, text-mining the Family letters category reveals how central Christopher Tolkien was in the writing process compared to the rest of his family – none of Tolkien’s other children communicated in this way over his creative writing nearly as much as his third son. Viewing these letters in the Context tool, it is clear that this subgroup of letters reveals the most about Tolkien’s social commentary during the second World War.

Publisher Letters

This corpus involves letters sent to his publishers at Allen & Unwin, largely consisting of Stanley and Rayner Unwin, as well as some unsent drafts to alternate publishing houses. The full text-mining results for this corpus are presented below:

The Cirrus tool demonstrates some expected terms, such as ‘book,’ ‘world,’ ‘read,’ and ‘work,’ as well as terms from his writing such as ‘hobbit,’ ‘rings,’ and ‘elves.’ Notably, the term ‘sequel’ shows up more prominently here than the other categories, where it exclusively refers to The Lord of the Rings in relation to The Hobbit, and there are examples from the Context tool of Tolkien noting how wider in scope this ‘sequel’ was getting.

On a related note, temporal terms like ‘long’ and ‘time’ are especially prevalent in this corpus, revealing through the Contexts tool that discussions on the timeframe of Tolkien’s writing were a common occurrence in these letters. Using this software, specific patterns in his publisher letters can be observed that divulge prominent repeated issues in the publication process, such as the usage of ‘time.’ For instance, the term ‘world’ is repeatedly used to explain to his editors his abovesaid notion of ‘primary’ and ‘secondary worlds,’ and the term ‘written’ is most often used when giving progress updates. Moreover, the term ‘west’ appears in this corpus almost exclusively in regards to his fictional mythology; these patterns show that Tolkien was in dialogue about the details of his fiction as well as strict editorial matters, and the language used when discussing this subject appears more formal than in their counterparts in the Family Letters category. Additionally, the word ‘read’ appears frequently alongside references to Rayner Unwin, highlighting his role as Tolkien’s early proofreader and test audience for his book.

Overall, the text-mining approach for the Publisher Letters category did not bring about much that was surprising, as the subject of time was a common complaint of Tolkien’s, and the limited number of correspondents naturally emphasized important figures such as Rayner Unwin. Still, this approach demonstrates the specific emphases central to the publication process that differ from his discussions on the same subjects in the other categories.

Other Letters

The final category encompasses letters Tolkien sent to a variety of figures outside the other two categories, mostly consisting of fans inquiring about his stories. The full text-mining results for this category are presented below:

From the Cirrus tool, it is clear how much more varied this category’s vocabulary is, befitting the more varied audience, although similar emphases on words like ‘world,’ ‘course,’ and ‘story’ are shared from the others. More specific terms from his writing are present here, such as ‘Frodo,’ ‘Valar,’ ‘age,’ and ‘Sauron,’ illustrating how much dialogue he had with his fans regarding literary inquiries. The Contexts tool is especially important for this category, as many words shared from the other two are used in relation to his narrative, such as the term ‘children’ to refer to fictional characters, such as the ‘children of Hurin.’

The Terms tool also provides an effective way at previewing common topics of discussion across these letters, including words like ‘men,’ ‘power,’ ‘names,’ ‘history,’ and other references to both life and writing.

The prevalence of the term ‘men’ in this category compared to the other two is of some interest: using the Context tool, it is apparent that Tolkien uses the word most often as a metonym for the human race as he writes it in his fiction.

Overall, this is clearly the most diverse category of the three, which is not surprising given the audience, but the variety of topics and discussions makes for an important use of Voyant in locating specific phrases and themes within them.

Additional Analysis

Each of the three categories offer different ways of viewing similar issues and topics, based on the context of the intended audience, and a text-mining approach allows for an easy way of pinpointing examples. For instance, different usages of the same word across categories can illustrate how common certain discussions are for certain letters, such as the simple word ‘good.’ Using the Summary and Context tools across all three categories, the word is used the least in Publisher letters, where it is used far more in an affirmative sense regarding decision-making. In contrast, the word frequently appears across the other two categories in discussions about morality and ‘good’ in opposition to Tolkien’s view of ‘evil,’ another word that shows up more in Other letters:

This demonstrates that among family and fan letters, the notion of ‘good’ was a common object of inquiry, especially regarding the Other letters, as questions of morality in The Lord of the Rings were prevalent. Another word of interest is ‘war,’ which shows up far more frequently in Family letters than the other two categories. This is not wholly unexpected, as a large part of the Family letters involve correspondence with Tolkien’s third son Christopher between 1943 and 1945, who served in the Royal Air Force during the war. Using the Contexts tool, it is apparent how often the word was understandably on Tolkien’s mind, and many of these letters reveal his thoughts and discussions on the nature of human conflict. In contrast, most of the word’s usage in the other two refer to his fictional wars in his writing, specifically in explanations of their narrative role, although an interesting example of its usage can be found in the Publisher letters, where he discusses his potential titles for divisions of The Lord of the Rings, and many of his ideas use the word ‘war’ before he decided on the final three (none of which employ the word). In the Other letters, he answers multiple queries about potential inspirations from the World Wars, of which he repeatedly denies in all except ideas of war-torn landscapes present in his stories.

Using the Links tool for the word in the Family letters corpus, Tolkien’s views become apparent through the proximity of words like ‘stupid’ and ‘waste,’ although similar Links for the other categories more often links ‘war’ to those in his books, showing how Tolkien reserved his most critical opinions for his personal letters.

The topic of inspiration leads into another interesting word across this corpus: ‘allegory.’ In discussing his creative approach to writing, Tolkien frequently denounces any presence of ‘allegory’ in his work, nor does he desire readers to be searching for one. The word most frequently appears in the Publisher letters where, for instance, he warns his proofreaders against approaching his work allegorically. But in the Other letters, he is more open to discussions on applicability over allegory, explaining potential inspirations and ways that his stories could possibly be compared to the ‘Primary world;’ many of these examples are tied into larger discussions on his creative decisions which he discusses with both family members and fans, such as the Ring’s symbolic role of evil in the story. This is perhaps one of the more important words to search for with this software, as it is part of numerous highly relevant passages wherein Tolkien describes these decisions and writing processes.

The Collocates tool provides additional insight by pointing to other words in connection, which gives some idea about which items were discussed regarding allegory, including words like ‘moral’ and ‘truth,’ characters like ‘wizards’ and ‘orcs,’ negative words like ‘ridiculous,’ and even the instance of Tolkien refuting any inspiration from the ‘atomic’ bomb.

In summary, Voyant’s text-mining tools demonstrate connections and contrasts between these divisions of Tolkien’s letters that would be hard to track or appreciate otherwise. In the broadest sense, Voyant’s Cirrus tool illustrates differences in vocabulary used for these three different audiences: a more focused vocabulary for Publisher letters, a personal vocabulary for Family letters, and a varied array of words for Other letters. In review, this does not appear surprising given their respective audiences, but the way the vocabulary is employed can be compared for important patterns. In the Publisher letters, matters of story were not discussed nearly as much as other issues like time and editing, except when Tolkien discusses reader feedback from Rayner Unwin, for example. Likewise, much of the references to specific narrative terms from his writing are concentrated in the Other letters, specifically fan letters, where he would write substantial answers to fan questions about his stories. These kind of response letters to fans progressively got shorter and more concise over the years, which is evident by looking at the letters themselves, but Voyant helps in showing what points he did decide to respond to in those later years. All in all, despite some of the results being unsurprising to one already familiar with the author and these letters, using Voyant created a clear overarching view of a corpus that has already been extensively researched without it. Patterns of vocabulary amongst different audiences are made visible, and specific topics and their prominence are tracked within their distinct contexts. Importantly, the various Voyant tools employed in this project allow for unique ways of visualizing this text-based data that was not previously applied to these letters.

Conclusion

Voyant’s text-mining software provides an important, novel way of approaching a text-based dataset for specific questions and purposes. Naturally, software such as Voyant does not replace traditional close reading, but this approach allows for easier analysis of textual patterns and connections. The different ways and contexts in which specific words are used provide valuable insight into the details of the content that may not be apparent elsewise. For such an expansive dataset of text, Voyant’s tools allow for quick pinpointing of passages containing important recurring words and phrases, such as the above examples of ‘war’ and ‘allegory,’ as well as demonstrating specific trends over time, which was not as relevant to this project’s aims. With Voyant, this project provides useful visualizations for each dataset that can be used by anyone interested in scouring the letters with questions on Tolkien’s creative and writing decisions, processes, and explanations.