Wilder Women Writing
An Investigation of Authorship Based on Selected Works of Laura Ingalls Wilder and Rose Wilder Lane
Kristina Runyeon-Odeberg
Degree Project
Main Field of Study: English Credits: 30
Semester/Year: Spring 2020 Supervisor: Rachel Allan Examiner: Terry Walker
Course code/Registration Number: EN005A Degree Programme: English MA
Contents
1. Introduction...1
2. Background...2
2.1 Laura Ingalls Wilder...2
2.2 Rose Wilder Lane...3
2.3 Corpus-Based Approaches to Determining Authorship...4
2.3.1 Register Features...4
2.3.2 Corpus Stylistics...5
2.3.3 Idiolect...6
2.4 Previous Research...7
2.4.1 Ingalls Wilder or Wilder Lane?...7
2.4.2 Stylistic Exploration of Texts...8
2.4.3 N-grams in Authorship Studies...9
3. Aim and Hypotheses...10
4. Material and Method...11
4.1 Material...11
4.1.1 Limitations...13
4.2 Method...13
4.2.1 Register Features: Verbs...13
4.2.2 Register Features: Nouns...14
4.2.3 Register Features: Pronouns...15
4.2.4 Stylistic Features: Common Descriptive Adjectives...15
4.2.5 Stylistic Features: Adjective-Adjective Constructions...16
4.2.6 Stylistic Features: Verb-Verb Constructions...16
5. Data Analysis...17
5.1 Register Features...18
5.1.1 Verb Tense...18
5.1.2 Nouns...20
5.1.3 Pronouns...21
5.2 Stylistic Features...23
5.2.1 Common Descriptive Adjectives...23
5.2.2 Adjective-Adjective Constructions...28
5.2.3 Verb-Verb Constructions ...30
i
6. Summary and Discussion...33
6.1 Register Features...33
6.2 Stylistic Features...34
7. Conclusion...36
References...38
Primary Sources...38
Secondary Sources...38
ii
1. Introduction
In May of 1930, Laura Ingalls Wilder brought the handwritten manuscript for her memoirs to her daughter, Rose Wilder Lane. At the time, Ingalls Wilder had only published short columns for magazines; Wilder Lane, however, was an award-winning writer (Ingalls Wilder 2014, xix; Holtz 1995 [1993], 113). From these memoirs, eight books were published. These books would sell more than 60 million copies, be translated into 45 languages, and undergo a revival when they were televised in the 1970s and 1980s. The television show, called Little House on the Prairie, has been one of the most popular and long-running shows to date, and it was in syndication at least until 2016 (Fraser 2016, 5). The year 2020 marks the 85th anniversary of the publication of Little House on the Prairie (Smith Hill 2007, 161). Ingalls Wilder, who was the sole credited author of these books, would become one of the most popular authors of children’s books in the twentieth century (Miller 2016, 1). Today, Wilder Lane is primarily remembered for being one of the founding mothers of the Libertarian movement rather than award-winning writing. The Little House books, however, still receive attention and recognition from readers worldwide. Titles from the series are appearing on lists such as 100 Best American Authors (Goodreads 2020). The story of Laura Ingalls Wilder is a phenomenon that continues to draw interest from both fans and critics.
Apart from fans and critics, scholars have taken interest in the Little House books, spanning from such subjects as the importance of horses (Blackford and Lockhart 2018) to gender studies (Romines 1997). For a long time, the story was simply that Ingalls Wilder had written her own memoirs. Soon after the death of Wilder Lane, scholars began to question the authorship of Ingalls Wilder’s books. One of the enigmas to capture the interest of scholars was how Ingalls Wilder seemed to be a full-fledged writer from the first of the Little House books, which was published when she was in her mid-sixties (Miller 2008, 23). Pioneer Girl, the original manuscript that formed the basis for the Little House books, has been described as much less endearing than the finished books, thus raising the question of co-editing (Miller 2008, 25). Others have raised the question of lackluster writing in The First Four Years, which was published in its manuscript version (Ingalls Wilder 1971, xiv; Moore 1975, 110).
Given that the manuscripts received for publishing from Ingalls Wilder evidently required no editing (Moore 1975, 118), the difference in quality has puzzled the minds of many.
The extent of Wilder Lane’s involvement with co-editing or even authoring has not yet been fully determined, so the question of authorship still remains. This thesis is concerned with finding evidence of the involvement of Wilder Lane in the Little House books. If any such evidence is found, corpus stylistics methods will be used to evaluate it.
2. Background
Anyone who investigates authorship of the Little House books also needs to be familiar with who Ingalls Wilder and Wilder Lane were, and how they became writers. This biographical background information is presented first. Section 2.1 provides background on Ingalls Wilder and how she started writing; Wilder Lane’s introduction to the trade is covered in Section 2.2.
The following two sections, 2.3 and 2.4, deal with approaches to authorship and previous research.
2.1 Laura Ingalls Wilder
Laura Ingalls Wilder was born in 1867 in Pepin, Wisconsin, as the second child of Caroline and Charles Ingalls (Fraser 2016, 31); she died in Mansfield, Missouri in 1957 (Fraser 2016, 476). She was a pioneer, a schoolteacher, an expert hen farmer, a writer, and a businesswoman. While Ingalls Wilder did not graduate from high school (Smith Hill 2007, 58), she was nevertheless a good scholar, because she earned her first teacher’s certificate before her sixteenth birthday. She then taught in three schools between 1882 and 1885 before marrying Almanzo Wilder (Smith Hill 2007, 59). The oldest child and the only one to survive past infancy was Rose. The writing career of Ingalls Wilder started later in life, when she shared her expertise on raising hens in Mansfield, Missouri: her insights were published in the Missouri Ruralist in 1911 (Smith Hill 2007, 96). During the 1920s, articles by Ingalls Wilder on cooking and homemaking negotiated (and edited) by Wilder Lane were published in periodicals (Smith Hill 2007, 122). In 1917, she took on the position as Secretary-Treasurer for the National Farm Loan Association, a position with more fiscal than writing responsibility; her knowledge of farming must have been seen as beneficial, because she stayed on for ten years. She also formed two all-female clubs, which were intellectually oriented (Smith Hill 2007, 123). In 1930, she wrote her memoirs under the title of Pioneer Girl, using a pencil and notebooks (Smith Hill 2007, 130–131). Wilder Lane typed out and edited this manuscript and tried to promote it (Smith Hill 2007, 133). Eventually, at the request of a publisher, Pa’s stories from Pioneer Girl were strung together by Wilder Lane with descriptions of everyday pioneer life (Smith Hill 2007, 135–138) and became Little House in the Big Woods, the first book to be published in the series. The rest of the Little House books were published at regular intervals until 1943, when the final Little House book to be published during Ingalls Wilder’s lifetime, These Happy Golden Years, first went into print (Smith Hill 2007, 180).
2.2 Rose Wilder Lane
Born in 1886 in De Smet, South Dakota, Wilder Lane traveled with her parents to Mansfield, Missouri at a young age, but her grown-up travels would take her to Europe and the Middle East. Wilder Lane was a talented student and graduated from high school, but the Wilders could not afford to send their only child to college (Holtz 1995 [1993], 44). Therefore, when her high school studies had concluded, Wilder Lane sought and found an income in Kansas City working as a telegraph operator. She was a self-taught touch typist and kept her employment until there was a telegraph strike in 1907. She returned to her parental home in Mansfield, Missouri, although her stay would be brief, because she received a post as manager of a Western Union telegraph station in Mount Vernon, Indiana (Holtz, 1995 [1993], 46–49). During this time, Wilder met and married Gillette Lane, a journalist. Married life did not suit Wilder Lane; she lived a rather destructive lifestyle, taking drugs and even trying to commit suicide (Holtz 1995 [1993], 51–52). Wilder Lane was unemployed after layoffs at Western Union but soon found work through a journalist friend and became an editorial assistant at The San Francisco Bulletin. She eventually moved on from editorial tasks to serials, one of which was about Henry Ford. This serial was made into a book and published in 1915, allowing Wilder Lane to send money for Ingalls Wilder to visit her daughter in California (Woodside 2017, 24). In 1918, there were layoffs at The San Francisco Bulletin, but Wilder Lane was offered a position at the Red Cross publicity bureau in London (Holtz 1995 [1993], 82). In 1929, Wilder Lane was living in her parental home again and needed work. Ingalls Wilder brought her daughter the draft version of Pioneer Girl, Ingalls Wilder’s memoirs intended for adults. According to Holtz, Wilder Lane took Pa’s stories (e.g. Ingalls Wilder 2014, 36–38; Ingalls Wilder 1971 [1932], 87–96) from this manuscript and made them into what would become Little House in the Big Woods, a book intended for children (Holtz 1995 [1993], 224–225). Harper Brothers accepted and published the book in 1932, credited to Ingalls Wilder. The very next year, according to Holtz (1995 [1993], 239), Wilder Lane used material from Pioneer Girl to write and publish Let the Hurricane Roar, a project that angered Ingalls Wilder. However, according to Woodside (2017, 202), mother and daughter overcame their difficulties and continued to collaborate on the entire Little House series. Wilder Lane gained recognition for a political work called Discovery of Freedom, which is considered one of the cornerstones of the Libertarian movement (Woodside 2017, 140). The extent of Wilder Lane’s contributions has been investigated (e.g. Moore 1975, 1980; Holtz 1995 [1993]; Miller 2016; Woodside 2017) but neither fully proven nor fully disproved, and to my knowledge, not been investigated from a stylistic viewpoint.
2.3 Corpus-Based Approaches to Determining Authorship
In this section, some of the register and stylistic features that can be used to gain insights into the authorship of literary works are presented. The register features are addressed before the stylistic features.
2.3.1 Register Features
Biber and Conrad define register, genre, and style as approaches to spoken or written text (2009, 15). Generally speaking, the definition of a register is coupled to a situation or a certain use of spoken or written text: it could be said to be one of several approaches to text (Biber and Conrad 2009, 6, 15). Registers are both identified by characteristics such as their setting and their audience (Biber 2006a, 477) and their linguistic features, such as verb tense and aspect (Biber 2006a, 479). Genre deals with entire texts and is concerned with the type of expression/rhetoric, expected (conventional) features, and format. Register deals with parts of texts and examines the features of expression, the linguistic characteristics, and interpretation.
Style is closely related to register, but it differs from register in being concerned with the features of the text that are valued æsthetically, meaning its grammatical and lexical features (Biber and Conrad 2009, 16, 71). Others speak of registral labels, such as lexico-grammatical features (Askehave and Swales 2000, 200), which is close to Biber and Conrad’s definition of style. Yet others view registers to be of importance when determining the purpose of communication (Tsiplakou and Floros 2013, 124). When it comes to studying register features, corpus methods are often used (Biber and Conrad 2009, 58). The use of computer tools for searches helps identify complex language patterns and analyze them; the amount of text examined through corpora can be much larger than those analyzed by hand (Biber and Conrad 2009, 74). Register analysis examines the features of expression, the linguistic characteristics of the texts, and their interpretation (Biber and Conrad 2009, 15). In a comparative study of registers including newspapers and fiction, the frequencies of verb tenses, nouns and personal pronouns are among the lexical features that have been addressed (Biber et al. 1999). Verb tenses are observed to be distributed as follows: the frequency of simple past tense is greater in fiction than in newspapers (Biber et al. 1999, 460). The present perfect is more frequent in newspapers than in fiction, and the frequency of the past perfect is more frequent in fiction than in newspapers (Biber et al. 1999, 461). Nouns are more common in newspapers than in fiction (Biber et al. 1999, 65), and personal pronouns are more common in fiction than in newspapers (Biber et al. 1999, 235). For this investigation, the register of fiction is relevant, since the Little House books were all marketed as fiction. However, as
Wilder Lane learned the writer’s trade in the newspaper world, the register of newspaper writing also needs to be considered.
2.3.2 Corpus Stylistics
In addition to register features, authorship has also been investigated from a stylistic viewpoint using corpus stylistics, which could be described as a discipline that investigates language patterns, especially, but not exclusively, in literature, using corpus methods (Shepherd and Berber Sardinha 2013, 70). Corpus stylistics could be said to occupy a space in between linguistics and literature. Stubbs (2005, 5) claims that both linguists and literary critics have been reluctant to accept it. Linguists, according to Stubbs (2005, 5), seem more interested in general theories than limited texts, and literary scholars seem to be skeptical of using statistics to determine writing style. However, Stubbs (2005, 5) argues that it is necessary to gain an understanding of the background of common features in order to understand what sets a text apart from others. According to Baker (2010, 101), corpus stylistics is used to be systematic and remove subjective thinking often connected to literary analysis. Corpus stylistics may involve examining features at a word or phrase level; in this thesis, the word level is in focus. Examples of features to examine include the use of adjectives or keywords, namely words that are more commonly used than in a reference corpus. Keywords are also a content-driven phenomenon: a book about farming in the Dakota territory set in the latter half of the 19th century would be very unlikely to contain the word car in any other sense than a train carriage. However, the word plow is likely to be used more frequently than in a reference corpus. Other examples of investigations may be related to morphology, lexis, syntax, semantics, or discourse (Burke 2015, 432). The presence of certain features, lexical or otherwise, may provide insights on authorship (Toolan 2013 [1998], 169).
Another way of phrasing this is to say that fictional style consists of deliberate choices made by authors in order to control how their story is conveyed (Biber and Conrad 2009, 144).
Corpus stylistics methods may be used to locate evidence of and analyze these choices.
There are different viewpoints on how corpus stylistics should be carried out. On the one hand, Wynne (2006, 224) prefers to analyze a corpus in its entirety to be more empirically sound than investigating individual examples. On the other hand, McEnery and Hardie (2012, 161) find the intuitive approach to be beneficial when studying corpora, because its characteristics can be located, allowing for rapid and effective interaction with a corpus. This thesis will attempt to combine both approaches while investigating authorship.
2.3.3 Idiolect
Another way of investigating authorship is to determine the concept of idiolect, which is formed from the Greek prefix idio- and a back-formation of dialect (Hazen 2006, 512). Some may define idiolect as the language pattern used by an individual during a certain period (Merriam-Webster 2020); however, according to Hazen, there is more to the word. Idiolect could refer to all utterances of one person (Hazen 2006, 512) or what is spoken or written, not thought (Hazen 2006, 512). These features will distinguish an individual from other speakers or writers, provided they have the same dialect (Hazen 2006, 512). Perhaps an idiolect is best described as “linguistic fingerprinting” (Coulthard 2004, 432). There is no standard formula or procedure for this fingerprinting process (Solan 2010, 406). The Unabomber was found and caught based on linguistic analysis, but a good memory was involved too. The perpetrator’s brother recognized some token expressions in a manuscript published by The Washington Post and claimed his brother had used them in a manifesto. A linguistic expert then chose twelve expressions from the manifesto that could be expected to appear in a text arguing a case, and it was ultimately found that the only hits containing all of these expressions led back to the manifesto (Coulthard 2004, 433). While this thesis does not aspire to catch criminals, it is concerned with finding evidence of ghost writing in the Little House books and thus will look for characteristic traits of writing.
Rosenthal and Yoon (2011) investigated authorship of Supreme Court decisions, because there was a debate regarding how much the actual Supreme Court Justices wrote themselves, and what, if any, writing they delegated to their clerks (Rosenthal and Yoon 2011, 284). The assumption made was that with increasing degrees of delegation, the writing style would appear more heterogeneous. If justices wrote only by themselves, their writing style would be more distinct writing than if they had delegated the writing to one or many clerks (Rosenthal and Yoon 2011, 284). There had been speculation regarding which justices wrote on their own and which justices delegated their work, but this was speculation, and the researchers endeavored to find out if there were any scientific grounds on which to base this speculation (Rosenthal and Yoon 2011, 285). Their method was to use 63 function words, including these examples: a, at, by, even, for, if, in, no, on, one, some, than, the, to, up, what, and when. These words were evaluated using statistical methods to determine where the greatest variation occurred (Rosenthal and Yoon 2011, 287). Rosenthal and Yoon (2011, 293) were not able to discern if the greater variation among some authors was coincidence or not, and their hypothesis was not possible to prove statistically. They combined different texts of known authorship (Rosenthal and Yoon 2011, 299), but similar results could be observed
when combining Tolstoy’s War and Peace in a rearranged order (Rosenthal and Yoon 2011, 300). The observations did show that there was greater variation among the texts supposedly written by one justice compared to another, which, according to Rosenthal and Yoon (2011, 306), indicated some evidence of delegated writing. Varied style, or a varied choice of words, may indicate ghost writing. However, using similar words to describe a similar event may also indicate ghost writing, which will be addressed in Section 5.1.2.
2.4 Previous Research
In this section, previous research regarding the authorship of the Little House books is presented along with other authorship studies using corpus-based methods.
2.4.1 Ingalls Wilder or Wilder Lane?
During the lifetime of both Ingalls Wilder and Wilder Lane, no one questioned the authorship of the Little House books. However, after the death of Wilder Lane, the question of authorship of the Little House books credited to Ingalls Wilder sparked the interest of scholars. One of these scholars is Rosa Ann Moore. Moore cited part of the prologue of The First Four Years (Ingalls Wilder 1971) as an example of lackluster writing (Moore 1975, 110). In comparison, a selection from By the Shores of Silver Lake (Ingalls Wilder 1971 [1939]) was praised for its easily-flowing language (Moore 1975, 109). Moore went on to examine notes and edits of By the Shores of Silver Lake; its plot and theme was the subject of dispute between Ingalls Wilder and Wilder Lane (Moore 1980, 104–105), but they kept on collaborating. The dispute of authorship, however, has not been fully resolved. One biography of Wilder Lane claimed that daughter, not mother, was responsible for what readers admire and claim that the skillful writing of Wilder Lane transformed the works of Ingalls Wilder (Holtz 1995 [1993], 380).
Others accuse Wilder Lane of theft and claim that Wilder Lane used Ingalls Wilder’s material because of jealousy and/or rivalry (Fraser 2016, 345). In comparison to the writing of Ingalls Wilder, Wilder Lane’s characters have been called flat; her language has been compared to commercialized magazine writing and been labeled overwrought (Fraser 2016, 345). There is still controversy about how great the involvement of Wilder Lane was in the writing process that formed the Little House books. To complicate matters, according to Miller (2016, 2, 145) neither Ingalls Wilder nor Wilder Lane ever admitted to any involvement from Wilder Lane.
2.4.2 Stylistic Exploration of Texts
In this section, studies that have influenced this one are addressed, namely stylistic explorations of works by Fowles, Conrad, and Meyer.
In Ho’s (2011) exploration of The Magus by John Fowles, the original and revised versions were compared using four corpus tools. Ho had two hypotheses. One of them was that a literary shift occurred in the revised version from story events to character-focused information (Ho 2011, 14), and the other was that more figurative expressions, such as similes and metaphors, were used in the revised version (Ho 2011, 31). To test these two hypotheses, Ho used four different corpus analysis tools. The first tool measured text similarities (Ho 2011, 68) and was concerned with the quantitative side of the analysis, listing matching words and chapters that had undergone the most changes, to name a few examples (Ho 2011, 75–
79). The second tool was targeted at locating plagiarism and confirmed the results of the first tool (Ho 2011, 83); however, neither of these initial analyses were concerned with how the second edition had been changed, but that it differed from the original version. The problem of measuring content was not addressed by using either of these two tools (Ho 2011, 90). The third tool was an interface to a grammatical part-of-speech tagger (Ho 2011, 118) and a semantic, or “word sense” tagger (Ho 2011, 118). By using this tool, it was possible to measure the increase or decrease of word tokens and differences on the word level (Ho 2011, 121, 140). The fourth and final tool was used for investigating figurative speech such as metaphors and similes. This required reading through concordance results to locate the correct instances of the investigated features (Ho 2011, 157). According to Ho, the advantage of the word-tagging tool is that it can be used to identify significant literary textual features (2011, 201). Both hypotheses regarding the revised version of The Magus were confirmed (Ho 2011, 188). The advantages of using corpus methods to study a literary text according to Ho are that it is possible to identify the number of changes and what has not been changed, and to be able to present statistics on a chapter level to establish a pattern pertaining to the revision process (Ho 2011, 199).
One way of applying stylistics to investigate either style or authorship is to study the use of adjectives in narrative descriptions and how they are used (Toolan 2013 [1998], 56), an approach used by the following two studies.
Stubbs (2005) explored Heart of Darkness using computer-based corpus methods with different approaches. One of them was to search for expressions pertaining to vague vision, such as mist, haze, and blurred (Stubbs 2005, 10–11), another was to find the most common lexical words, with focus on the ten most common nouns (including proper nouns) and verbs
(Stubbs 2005, 11). Yet another approach was to find collocations and analyze what certain adjectives denoted (Stubbs 2005, 14–15), and another was to investigate nominal groups and phrases. The conclusions were that using a computer to investigate corpora may reveal patterns that would most likely be missed by a close reading of the same text, and to add systematic data to literary analysis (Stubbs 2005, 21–22). There are both elements of close reading and computer investigations in the analysis carried out for this thesis.
Cesiri and Coccetta (2019) carried out a stylistic exploration of the Twilight saga by investigating collocates of the most frequent adjectives as well as keywords. The object of the study was to find stylistic evidence of what literature critics have said: that the plot and characters are underdeveloped (Cesiri and Coccetta 2019, 52). Their findings were that the narrative descriptions focus on external qualities, for example eyes, face and expression. All of these, including the eye color of characters, seem to be tied to emotions (Cesiri and Coccetta 2019, 52), except if the emotions are already expressed; in this case, the eye color is not mentioned. Their conclusions were that the critics were right; the popularity of the books has no support in well-developed literary merits (Cesiri and Coccetta 2019, 57). This thesis investigates frequent pronouns and descriptive adjectives. However, there is another way of investigating authorship: the use of n-grams, or clusters of words, which will be discussed next.
2.4.3 N-grams in Authorship Studies
Antonia et al. (2014) used n-grams of n=1 to 5 to investigate authorship of 174 Renaissance plays and 254 articles from Victorian periodicals (2014, 151). The overall findings were that 1–4-grams performed well in authorship, but that using 5-grams did not increase the accuracy (Antonia et al. 2014, 147). Their findings were that for frequent hits, individual words worked best for authorship attribution. For rare occurrences, 3-grams brought the best performance (Antonia et al. 2014, 154), and finally, for a defined list of function words, 2-grams worked best. Their recommendation was to use up to 2-grams for the investigated Victorian corpus (Antonia et al. 2014, 153) and 3-grams for the corpus consisting of Renaissance plays (Antonia et al. 2014, 154). Something to consider is that the larger the size of n, meaning the number of words included in the n-gram, the more n-grams there will be; however, the likelihood of them occurring repeatedly is smaller. From a linguistic point of view, though, 2- 3-grams will be more commonly repeated (Antonia et al. 2014, 154).
Wright (2017) has also investigated authorship using n-grams. The corpus used consisted of e-mails from Enron (Wright 2017, 221), and the samples from different authors
were extracted at random in different sizes (Wright 2017, 222). It appears to be a common practice to combine several texts from one author rather than use texts separately (Wright 2017, 223). In essence, the results were that the bigger the original dataset size had been, or the higher number of/longer e-mails that the sample had been taken from, the greater the accuracy of the authorship attribution (Wright 2017, 228). Four-grams were found to perform best, but the accuracy was better for some authors than others (Wright 2017, 228).
Not all n-gram studies have been as successful, though. Grieve (2007) found the performance of word n-grams to be poor when using collocates to investigate the authorship of newspaper columns (Grieve 2007, 263). There was only a 75% success rate in identifying the author using two-word collocations and 53% for three-word collocations when there were only two possible authors (Grieve 2007, 263), and the results became poorer for each added possible author. However, when the use of punctuation was included as a variable, Grieve was able to identify the author with an accuracy from 93 to 95% (2007, 262). This thesis contains a brief investigation of 3–4-grams of the most frequent adjective across the corpora in search of similarities or differences, which is discussed in Section 5.2.1.
3. Aim and Hypotheses
Since the 1970s, scholars seem to have agreed to disagree on the authorship of the Little House books. The authorship of these books has been investigated from the standpoint of style and comparison between drafts and finished material, as per Section 2.4.1. The Little House books have also been investigated in terms of authorship from a biographical (e.g.
Miller 2016) and political standpoint (Woodside 2017). I have not been able to locate any investigations based on register features or corpus linguistics, though. Therefore, the aim of this study is to investigate the authorship of the autobiographical fiction of Ingalls Wilder using exactly those two approaches.
The investigation of authorship will be done by comparing three corpora (see 4.1).
Two of them contain works of known authorship, and they are works by Ingalls Wilder (henceforth Laura) and Wilder Lane (henceforth Rose), respectively. The third corpus (henceforth Little House) contains material attributed to Ingalls Wilder, but there may be evidence of ghost writing or co-editing. The investigation will examine various lexical features in each corpus, aspiring to identify similarities and differences. From these investigations, there may be conclusions to draw regarding the authorship of the Little House books, traditionally credited to Ingalls Wilder, but with a disputed amount of involvement from Wilder Lane.
The register features investigated are drawn from features of newspaper writing and fiction (Biber and Conrad 2009, 77). In Sections 2.1 and 2.2, the different writing backgrounds of Ingalls Wilder and Wilder Lane were discussed. Considering their differences, the hypothesis concerning register features is that there should be more register features of newspaper writing in Rose, and less in Laura. In Rose, a lower frequency of simple past tense is expected than in Laura (Biber and Conrad 2009, 77). In the Laura corpus, the expected results are a lower frequency of nouns and a higher frequency of personal pronouns than in the Rose corpus (Biber and Conrad 2009, 77). Overall, Little House should show a mixture of the two register features (Biber and Conrad 2009, 77). The stylistic features of common adjectives, adjective-adjective constructions and verb-verb constructions are exploratory by nature.
Specifically, the research questions are:
What is the normalized frequency of simple past tense, present perfect and past perfect across the three corpora, respectively?
How does the normalized frequency of nouns vary between the corpora?
How does the normalized frequency of pronouns vary between the corpora?
In what contexts do descriptive adjectives found across the corpora appear?
What adjective-adjective constructions are present in the corpora?
What verb-verb constructions may be observed in the corpora?
4. Material and Method 4.1 Material
As mentioned in Section 3, the material for this thesis consists of three corpora, called Laura, Rose, and Little House. The works that were used to build these corpora and the reasons for choosing them will be discussed in the following paragraphs.
The first corpus, Laura, was built from three works by Ingalls Wilder: The First Four Years, On the Way Home, and West from Home. The First Four Years is an account of the early married life of the young Wilders, best labeled as fictionalized autobiography. The second work in the corpus, On the Way Home, consists of Ingalls Wilder’s diary that she kept when the young Wilder family relocated from the Dakota territory to the Ozarks in Missouri.
The third work, West from Home, consists of letters and postcards to Almanzo Wilder, written while Ingalls Wilder went to San Francisco to visit her daughter in 1915.
For the second corpus, Rose, representing Wilder Lane’s authorship, a deciding factor was the availability of works from the same register on electronic format, and it was possible to locate
two novels: Diverging Roads, the story of Wilder Lane’s early adulthood as a telegraphist and beginning reporter, along with Young Pioneers (first published as Let the Hurricane Roar), an account of a pioneer couple and their struggle. As Wilder Lane wrote the setting for On the Way Home, and a small part of West from Home is known to have been penned by her, these contributions were also included in the Wilder Lane corpus.
The third corpus, Little House, was built from three novels credited to Ingalls Wilder but with some level of involvement from Wilder Lane, Little House in the Big Woods, On the Banks of Plum Creek, and These Happy Golden Years. Little House in the Big Woods and These Happy Golden Years were the first and final Little House books to be published during Ingalls Wilder’s lifetime. These two were chosen to have samples from the beginning and the end of the Little House series. Table 1 shows the word counts of the three corpora for the sake of comparison.
Table 1: Details of the three corpora used (word counts according to Sketch Engine)
Corpus Works included Word count
Laura (Ingalls Wilder) On the Way Home (diary part) The First Four Years
West from Home
10,356 24,359 20,629
Total 55,344
Rose (Wilder Lane) On the Way Home setting and contributions to West from Home Young Pioneers
Diverging Roads
10,074 28,014 82,727
Total 121,415
Little House
(Ingalls Wilder and Wilder Lane)
Little House in the Big Woods On the Banks of Plum Creek These Happy Golden Years
32,199 55,648 64,616
Total 152,464
There are similarities between some works that have justified the addition of them to two of the corpora. In both Young Pioneers and On the Banks of Plum Creek, there are grasshopper infestations (Wilder Lane 1976 [1933], 33–44; Ingalls Wilder 1971 [1937], 192–204). If there is any evidence of similarity, it is most likely to appear in the two novels that tell of similar events. Since the aim of this thesis is to determine authorship, the most suitable material for the task needs investigation, and the similar storylines of the aforementioned works certainly warrant a stylistic comparison. However, to make this stylistic comparison, there are limitations to be considered.
4.1.1 Limitations
The three corpora differ in size; therefore, any observations must be normalized. The chosen normalization frequency is 10,000 words in this thesis. The Laura corpus is not a homogeneous corpus; the three works that were chosen for it are a novel (The First Four Years), a travel account in the form of a diary (On the Way Home), and a collection of letters (West from Home). Therefore, its register or lexical features may not be as coherent to the genre of fiction as those of the other two corpora.
4.2 Method
First, the electronic versions of works to be included in the corpora were investigated for scan errors (Eder 2013b, 607). When a work is scanned, there are a number of pairs that may be mistaken: the letter m may end up as the word in, the letter l may end up as the number 1, the letter h may end up reading li, and the letter c may end up as an e, or vice versa (Eder 2013b, 607). As she is a frequent pronoun in all three corpora, it should not read slie, because that would affect the number of hits. An example encountered in the raw material for the three corpora featured in this thesis was ta jee instead of to see (Wilder Lane 1919, 35).
Wordsmith Tools (Scott 2008) was used to extract some of the basic statistics: the number of sentences, the average sentence length, the average word length, and the standardized type/token ratio. This provided a bird’s-eye view of the corpora involved, a capability Sketch Engine (Kilgarriff et al. 2020 [2004]) does not have in as much detail.
While it is possible to obtain the number of sentences and words, and thus the average number of both, the rest of these statistics are not readily available. Tagging may be useful when looking for different phenomena (Shepherd and Berber Sardinha 2013, 72), and Sketch Engine (Kilgarriff et al. 2020 [2004]) has the ability to tag word classes, which was useful for identifying the register features investigated for the corpora as identified in the first three research questions, i.e. tenses, nouns and pronouns. This particular feature was also useful when investigating the final three research questions, i.e. the distribution of personal pronouns, consecutive adjectives, and consecutive verbs.
4.2.1 Register Features: Verbs
First, the verb tenses to find different register features were investigated, namely the simple past, present perfect and past perfect tenses. These were then compared for each corpus by using the search tags shown in Table 2.
Table 2: Sketch Engine tags used for verb tenses
Tense Search tag Example
(simple) past VVD swam, did ... swim
be, past VBD was, were
have, past VHD had
have, present, third-person VHZ has
have, present, non-third-person VHP have
The VVD search tag not only catches the simple past tense using a single verb; it also catches the verb do, which is used as an auxiliary in this tense (Eastwood 2002 [1994], 78). There is another tag, VVN, which produces past participle hits; however, these occurrences appeared as hits while using the tags VHZ, VHP, and VHD. The hits from these searches were sorted into the correct category: simple past tense, present perfect, or past perfect. As the tags VHZ and VBD exclude other verbs than have and be, and VVD excludes was, were and had, all of them were needed. The verb have in past tense may signify simple past tense, as in Mary had a little lamb, past perfect, as in Peter had taken the bus, or conditional, as in if I had a million dollars, I would... The occurrences of had were searched, and the hits that did not meet the requirements for inclusion were removed from the actual count before normalization. The VVD tag may also cause errors: the verb got, for example, is identical in both the simple past and perfect tenses. However, it can be deceptive if it is preceded by have in the present or past tense, making it present or past perfect. To make it more complicated, have is sometimes omitted in American English. Therefore, some extra attention is required if or when the verb form got is encountered. The concordance search for all three corpora was scrutinized, and the hits including have got, including where have had been omitted before got, were removed from the frequency count before it was normalized.
4.2.2 Register Features: Nouns
Nouns also have tags in Sketch Engine. For the search criteria of this thesis, there were four tags of interest, which are shown in Table 3.
Table 3: Sketch Engine tags used for nouns
Noun form Search tag Example
Singular or mass NN table (singular), hair (mass)
Plural NNS horses
Proper, singular NP Laura
Proper, plural NPS Wilders
With nouns, there are fewer potential error sources than verbs, but the search results still need to be checked for occurrences of single letters of the alphabet when the Sketch Engine tag NN (noun, singular) is used. The plural tag does not seem to create ambiguous hits according to a close scan of Laura and a bird’s-eye scan of Rose and Little House. In a register study by Biber and Conrad in 1999 (Biber and Conrad 2009, 77) there is no mention of the frequency of proper nouns, but these were included as a matter of curiosity. Among the nouns, some words may create potential error, of which one is call. This was found in the concordance, and removed from the count, as “Some call it Stony Creek” (Ingalls Wilder 1976 [1962], 48).
Single letters of the alphabet, such as o in one o’clock may also end up here, so an analyst has to be watchful of them in the hit list. The tag NP lists some occurrences such as dates, which were removed, and abbreviations, but received hits from the latter category were kept, because the US signifies the United States, two proper nouns, to mention one example.
4.2.3 Register Features: Pronouns
The normalized frequency of personal pronouns was compared between corpora using the designated tags for that purpose, namely:
PP, which lists the personal pronouns (I, you, he, ...)
CDZ, which lists the possessive pronouns using the ’s construction (one’s, Laura’s...)
An error that may be encountered using these tags is that a possessive pronoun with (or without) the ’s construction may be a determiner, so the concordance hit list was scrutinized for such instances, and the irrelevant hits were removed from the actual count before they were normalized. Furthermore, some distinguishing features were investigated. All works, although the Laura corpus is diverse, are centered on women. Therefore, the frequency of the pronoun she across the corpora compared to other pronouns was considered. Simple searches were made for I, you, he, she, we and they for all corpora, including the added specification of the hits being pronouns. A possible error here is that the personal pronoun you may refer both to one person or several people. An actual error that was encountered during the simple search is that Sketch Engine does not show any hits whatsoever for the personal pronoun I; the pronoun tag has to be removed to find the occurrences of this particular form.
4.2.4 Stylistic Features: Common Descriptive Adjectives
The most common descriptive adjectives were searched for using the simple word class search function in Sketch Engine. No tagging was needed here, as the tool provides a ready-
made solution. Four descriptive adjectives out of the top ten were found to be common to the corpora, namely little, good, long, and new. A possible source of error for the adjective little is that it may be part of an adverbial, as in only a little; it may function as a determiner, as in little progress has been made, or it may be a pronoun, as in little of the building remained.
Therefore, those hits were removed before normalization. As an initial investigation, the top 3–4-grams were determined for little across the corpora by using the N-grams function and the Advanced search tab chosen. When this search had provided its results, the Advanced search tab on the Word Sketch was used with the adjective word class chosen from the drop- down menu. This feature was used to determine what nouns the five adjectives common to the corpora modified, to investigate any further differences across the corpora. Some differences in meaning were found; these are discussed in Section 5.2.1.
4.2.5 Stylistic Features: Adjective-Adjective Constructions
The use of adjectives may be a distinguishing feature of narration (Cesiri and Coccetta 2019).
During the reading of the works that form the corpora of this thesis, sometimes the use of adjective-adjective constructions caught my attention. Sketch Engine has a corpus query language, henceforth CQL (Kilgarriff et al. 2020 [2004]), which offers the possibility of searching for adjective-adjective constructions. The following sequence is entered into the Sketch Engine concordance search: [tag="J.*"]{1}[tag="J.*"]. This will produce adjective- adjective constructions using no commas. In order to find the hits with adjectives using commas, it is possible to use the CQL search [tag="J.*"]","[tag="J.*"] to find two consecutive adjectives separated by a comma and then keep adding ","[tag="J.*"] to find three or more consecutive adjectives separated by a comma. These searches were all performed, and the results were normalized (Biber 2006b, 256). In order to investigate how the most common adjectives were used, 3–4-grams were extracted for the top adjective.
4.2.6 Stylistic Features: Verb-Verb Constructions
Another possible distinguishing style feature may be the use of verbs in sequence, a feature which was noticeable during the reading of the material for the Rose corpus, as in (1) below:
-(1) “Suddenly she thought, hoped, asked, Had I taken it myself, to play with?” (Wilder Lane, in Ingalls Wilder 1976 [1962], 81)
Four verbs appear in sequence here: the first three verbs are part of one verb phrase, and the fourth, which is the auxiliary of the past perfect tense, is part of the next verb phrase. In order to find verbs in sequence, the query [tag="V.*"]","[tag="V.*"] was entered, and the additional string ","[tag="V.*"] was added as needed to catch three or more verbs in sequence. This eliminated duplicate hits from the previous queries concerning the present and past perfect tenses, which would have shown up without the comma in the query. This first query contained results of the type <verb, simple past tense> + <verb, present participle>;
sometimes, more than one of the latter occurred. These hits were separated from the verb-verb constructions. Another search then used the query [tag="VVG"]","[tag="VVG"], adding the sequence of ","[tag="VVG"] until there were no more hits received. The hits were recorded, and all the concordances were checked for the number of consecutive present participles to ensure that the correct number of occurrences had been detected. A possible source of error is that the hit “denying, appealing, swearing” (Wilder Lane 1919, 73) may show up as two hits for two consecutive present participles, one hit for the first two and another for the final two, respectively. This kind of faulty hit was found and removed. Another possible source of error is that an adjective, as in “the driving team” (Ingalls Wilder 1971, 23), may be tagged as a present participle, although no such hits were found. When the faulty hits had been removed, the remaining ones were normalized (Biber 2006b, 256).
5. Data Analysis
Table 4 shows some basic statistics of the three corpora as per Wordsmith Tools (Scott 2008).
Table 4: Basic statistics of the corpora
Parameter Laura Rose Little House
tokens (words used for word list) 55,417 121,415 152,527
types (distinct words) 4,844 9,444 6,756
standardized type/token ratio (STTR) 39.20 39.47 39.39
mean word length (characters) 4.03 4.26 4.10
number of sentences 3,311 9,342 12,700
mean length of sentences (in words) 16.74 13.00 12.01
The mean word length does not reveal any notable differences between the corpora, but the average sentence length does. With an average sentence length of 16.74 words, Laura has the longest sentences. Rose is notably shorter at 13.00 words and Little House has the shortest sentence length of them all at 12.01 words.
As can be seen in the table, the standardized type-token ratios of the three corpora, respectively, are close to each other. The type/token ratio of Laura is 39.20, while the corresponding numbers for Rose and Little House are 39.47 and 39.39, respectively. From this point, all results are from Sketch Engine, as explained in Section 4.
5.1 Register Features
This section will present the results of the investigated register features: verb tense, nouns, and pronouns.
5.1.1 Verb Tense
First, the verb tense was investigated as described in Section 4.2.1. Table 5 shows the normalized number of hits per 10,000 words for each of the investigated verb tenses (Biber 2006b, 256).
Table 5: Frequencies of verb tenses per 10,000 words across the corpora
Corpus Simple Past Present Perfect Past Perfect
Laura 602.40 27.83 42.28
Rose 769.10 17.55 87.24
Little House 973.93 9.25 60.01
The lowest number of simple past tense is found in Laura, which has a frequency of 602.40 hits per 10,000 words, while the highest frequency is found in Little House, which has 973.93, and Rose has 769.10 constructions with simple past tense. The present perfect is most frequent in Laura with 27.83 hits and least frequent in Little House with 9.25 of them; Rose has 17.55 hits. As for the past perfect tense, it is most prominent in Little House, with 87.24 hits, and least prominent in Laura, with 42.28 hits, while Rose is in between the two with 60.01 hits.
The use of different verb tenses reflect the types of text in each corpus. The simple past tense is used extensively in Little House, as is shown in (2):
-(2) “A water-pail stood on a bench by the door. A boughten [sic] broom stood in one corner. [...] Some kind of short, white sticks lay in [a] trough [below the
blackboard]” (Ingalls Wilder 1971 [1937], 150).
The simple past tense is used in preference to past progressive, which could have been used to set the scene; the water pail could have been standing on a bench, the broom could have been standing in the corner, and the short white sticks (chalks) could have been lying in the trough.
The present perfect was most prominent in the Laura corpus, possibly due to the diary writing approach (Haegeman 2013, 89), which (3) shows:
-(3) “Well, we have come to the bluffs. [...] Grass has not grown on the face. [...] And the first oak trees we have seen. We have been going over the bluffs...” (Ingalls Wilder 1971 [1962], 29).
Here, the diary writing style of Ingalls Wilder is evident; this was probably one of the first works she wrote. Another defining characteristic of a diary is the lack of personal pronouns, which will be discussed in Section 5.1.3.
The Little House corpus has the highest normalized frequency of past perfect tense; (4) contains a description of the Ingalls family bulldog, Jack:
-(4) “All day long for many, many days, Jack had been trotting under the wagon. He had trotted all the way from [...] Indian Territory [...]. He had learned to take his rest whenever the wagon stopped” (Ingalls Wilder 1971 [1937], 1).
This text is clearly directed at a younger audience. The simplicity and repetition helps get the message across. Day, trotting, and wagon are repeated; the omitted passages cover U.S.
geography to show the way the Ingalls family wagon had taken from Kansas to Minnesota.
The passage is also at the beginning of a book, which motivates a short history of the travels leading the Ingalls family to their new home. The Little House books are each part of the series, but they may be read independently. How the past perfect is used in Rose is worthy of attention, as Wilder Lane sometimes takes the opportunity to recap when other alternatives are available. This is shown in (5):
-(5) “...she had had moments of criticising Louise and momma. But she had quickly hidden the criticism in the depths of her mind” (Wilder Lane 1919, 127).
Here, Wilder Lane is using the past perfect tense to let the main character, Helen, rethink her feelings with the arrival of her sweetheart. An alternative way of telling this part of the story
would be to put the actual thoughts when they happen, for example to write: she sometimes criticized Louise [...] in her thoughts, but quickly hid the criticism. However, the differences (and likeness) in use of verb tense between the corpora was not the only investigated register feature. The distribution of nouns, which was another, will be discussed next.
5.1.2 Nouns
The distribution of nouns and proper nouns was investigated and normalized. Table 6 shows the differences across the three corpora.
Table 6: Frequencies of nouns and proper nouns per 10,000 words across the corpora
Corpus Nouns Proper Nouns Combined
Laura 1,849.85 453.70 2,303.55
Rose 1,849.49 273.80 2,123.29
Little House 1,660.65 654.06 2314.71
The count of nouns per 10,000 words shows that Laura and Rose are very close at 1,849.85 and 1849.49, but Little House has the least amount of nouns; there are 1,660.65 of them. As for proper nouns, Little House has the highest normalized frequency at 654.06, followed by Laura at 453.70, while Rose has the lowest normalized frequency at 273.80. Combined, Little House has the highest normalized frequency at 2,314.71; Laura has 2,303.55, and Rose has 2,123.29 of them. To show how nouns may be used in Little House, (6) was extracted:
-(6) “The cloud was hailing grasshoppers. The cloud was grasshoppers. Their bodies hid the sun and made darkness” (Ingalls Wilder 1971 [1937], 195).
In the first sentence, Ingalls Wilder (or possibly Wilder Lane, as this example comes from Little House) could have written hailed instead of was hailing, but the following sentence uses a noun as a complement to the verb be. The third sentence could have been worded as ...and it became dark; again, there is a noun used as a complement; this time, though, the verb is make.
In (7), a similar event is described, where Wilder Lane is the credited author:
-(7) “Grasshoppers were coming out of the sky, out of that cloud. The air twinkled with their shining wings, coming down. The cloud was grasshoppers” (Wilder Lane 1919, 22).
The second sentence from Little House and the third sentence from Rose are identical, and there are other similarities, but the first and second sentence above shows more verbalization than nominalization, in contrast to the example from Little House. The repeated coming and twinkling of the air give a more verbal impression. For the sake of comparison, the original passage from Pioneer Girl is shown in (8):
-(8) “...there was a light colored, fleecy cloud over [the sun...]. And then we saw that the cloud was grasshoppers, their wings a shiny white” (Ingalls Wilder 2014, 79).
Wilder Lane had access to the manuscript for Pioneer Girl when she wrote Young Pioneers, so Ingalls Wilder’s unfavorable reaction of her daughter publishing her material without asking does motivate using the word of theft (Fraser 2016, 345). Shining wings in 7 bears some resemblance to shiny white in 8. Anyone who performs a search for “the cloud was grasshoppers” on Google Scholar (or regular Google) will see results (e.g. Behmer and Joern 2012, 3; Nugent 2020, 68; Skinner et al. 2002, 102) pointing to this exact scene from On the Banks of Plum Creek, which is an indication that this is a rare way of constructing a sentence.
The examples so far show some of the register features of the corpora, but to get more information, the third register feature, namely that of pronouns, needs to be discussed.
5.1.3 Pronouns
The pronouns in the corpora per 10,000 words are shown in Table 7.
Table 7: Frequencies of pronouns per 10,000 words across the corpora
Corpus Personal Possessive Combined
Laura 772.97 125.58 898.55
Rose 1,099.27 280.26 1,379.52
Little House 875.29 204.77 1080.12
The highest frequency of personal pronouns is found in Rose, which produced 1,099.27 hits, followed by 875.29 in Little House, and Laura at 772.97. Rose is also the corpus with the highest normalized frequency of possessive pronouns at 280.26, though it is considerably smaller than the frequency of personal pronouns. Little House has the second-highest normalized frequency of possessive pronouns with 204.77 of them. Again, Laura has the lowest normalized frequency of possessive pronouns, with 125.58 hits. The distribution of
pronouns across the corpora was also investigated. Table 8 shows the differences between the corpora for frequencies per 10,000 words.
Table 8: Pronouns by type across the corpora
Corpus I you he she we they
Laura 120.52 40.47 71.73 71.37 95.76 86.73
Rose 114.27 95.41 122.22 313.52 27.06 59.83
Little House 103.76 75.76 102.58 168.37 24.27 115.24
The highest frequency of any pronoun overall is she, which has a normalized frequency of 313.52 words in Rose, followed by 168.37 in Little House. Another notable differences is that the highest frequency of he also occurs in Rose, at 122.22 hits, followed by 102.58 in Little House. The third and final difference to note is that the highest normalized frequency of they is found in Little House at 115.24 hits, followed by Laura at 86.73 hits and Rose at 59.83. (9) shows the use of personal pronouns and was extracted from Rose:
-(9) “She did not want to stand there [...]. She did not want to follow the old stale road [...], which had not changed since she could remember. She felt that she should be doing something, she did not know what” (Wilder Lane 1919, 25–26).
There is a sense of pronouns being stacked in the passage, and a wordlist search in Sketch Engine confirms that for Rose, the word she appears in second place after the, even more commonly used than and. This is shown in (10):
-(10) She was a graduate of Weeks’ School of Telegraphy, she told him breathlessly. She could send perfectly, she wasn’t so sure of her receiving, but she would be awfully careful not to make mistakes. She had to have a job, she just had to have a job; it didn’t matter how much it paid, anything. She felt that she could not walk out of that office. She clung to the edge of the counter as if she were drowning and it were a life-line (Wilder Lane 1919, 47).
In Little House, the top pronoun is also she, but it is in fifth place overall after the, and, to, and was. Fiction has its distinguishing register features as discussed in Section 2.3.1;
however, one of the defining register features of diary writing is to omit the subject of the sentence (Haegeman 2013, 89), hence avoiding I, for example, or we, as (11) shows:
-(11) “Started at 8:30. Soon crossed Little Elm Creek and drove through beautiful woods of elm, oak, ash, hickory and walnut” (Ingalls Wilder 1976 [1962], 44).
Since the passage is retelling the journey of the small Wilder family, it is assumed that we is implied. Ingalls Wilder could have written at least two we: one at the beginning of each sentence, and if there had been a comma before and drove, a third we could have been inserted. However, the subject has been omitted in all possible places. If the identity of the travelers had been unknown, any personal pronoun could have been inserted.
In the case of the pronoun I, however, Ingalls Wilder makes a point of using the subject, which may explain the highest normalized frequency of that particular personal pronoun. In another diary entry, as shown in (12), Ingalls Wilder wrote:
-(12) “Started at 9. We are following down the valley of the creek on a nice level road.
Reached Schuyler [...]. Here we had to get the tires set, so we did not leave town till 3. I met an interesting woman...” (Ingalls Wilder 1976 [1962], 36).
According to style features of diary writing (Haegeman 2013, 89), perhaps a few of the occurrences of we or I in (10) could be removed. However, Ingalls Wilder seems to use the different pronouns in order to distinguish between what happened to the small Wilder family, we, and what she herself, I, experienced. With the register features accounted for, I will now present the results of the investigation of the stylistic features.
5.2 Stylistic Features
As the searches for register features showed some areas worthy of further investigation, three stylistic features were researched: common descriptive adjectives, adjective-adjective constructions and verb-verb constructions.
5.2.1 Common Descriptive Adjectives
As mentioned in 4.2.4, a search for the top ten descriptive adjectives showed that there were four adjectives that were common to all three corpora: little, good, long, and new. Therefore, the main focus will be on these adjectives and what nouns these words may modify; in other
words, how they are used. The top ten descriptive adjectives from all three corpora are shown in Figure 1, where the horizontal axis represents the number of hits per 10,000 words, rounded to the nearest integer.
Figure 1: Most common descriptive adjectives in the corpora
The top descriptive adjective across all the corpora is little, with 33 hits per 10,000 words in Little House, followed by 29 hits in Laura, and 23 in Rose. Good is in second place across all the corpora, too, with 25 hits in Laura, followed by 22 hits in Little House, and 20 in Rose.
The third most common descriptive adjective in Rose is long, with 13 hits, which has 14 hits in Little House, and 9 in Laura. The fourth most common descriptive adjective in Little House is new, with 11 hits in Little House, followed by 9 in Laura and 8 in Rose.
While frequencies reveal the most commonly used adjectives, the use of adjectives may contain more information. One way of investigating their use is to look at clusters of words, or n-grams. Going back to the discussion in Section 2.4.3, according to Antonia et al.
(2014, 154), 2–3-grams have been determined to be more frequent, and according to Wright (2017, 228), 4-grams perform best. Therefore, 3–4-grams for the most frequent adjective little were researched; the hits received across the corpora were for 3-grams. Table 9 shows 3- grams per 10,000 words for the three corpora which received more than 5 hits before normalization.
Table 9: 3-grams connected to little per 10,000 words
Corpus/Rank Laura Rose Little House
1 a little way 1.99 a little while 0.66 was a little 1.57 2 was a little 1.45 the little house 0.58 of the little 0.72 3 only a little 1.26 of the little 0.58 and a little 0.72
4 in a little 0.58 a little while 0.72
5 and the little 0.58 the little house 0.59
6 was a little 0.50 in the little 0.59
7 little David John 0.50 a little way 0.59
8 to the little 0.41 and the little 0.52
9 only a little 0.41 in a little 0.46
10 just a little 0.41 only a little 0.39
As the table shows, two 3-grams are present in all three corpora. The first is only a little, which has 1.26 hits in Laura, followed by 0.41 hits in Rose and 0.39 hits in Little House. The second is was a little, which has 1.45 hits in Laura, followed by 1.57 hits in Little House and 0.50 hits in Rose.
A concordance search showed that only a little is used to indicate slightly different meanings across the three corpora, In all three corpora, the 3-gram only a little is mainly
related to short distances or a young age, actual or perceived. Some differences in the use of this 3-gram will be discussed. In Rose, one instance referring to a character has a belittling connotation, as shown in (13):
-(13) “Her mother seemed suddenly to see in her a stranger. ‘But—good gracious, Helen!
You’re only a little girl!’ ” (Wilder Lane 1919, 13–14)
In Little House, only a little may show a considerably different meaning. In (14), Pa is looking at the four-year-old Laura admiringly after she has dragged her older sister Mary clear across the woodbox:
-(14) “ ‘You’re only a little half-pint of cider half drunk up, but by Jinks! you’re as strong as a little French horse!’ ” (Ingalls Wilder 1971 [1932], 36)
While (13) and (14) are not two accounts of the very same subject matter, they do touch on the subject of would-be frail girls. However, the implication of calling an eighteen-year-old young woman little, as Wilder Lane does, and a four-year-old girl little, as Ingalls Wilder does, are two different matters. Wilder Lane lets her main character be belittled, while Ingalls Wilder conveys a very different message. A passage matching (14) can be found in Pioneer Girl, without a capital J in Jinks and the exclamation mark afterwards, but otherwise identical (Ingalls Wilder 2014, 32). This means that Wilder Lane’s only contribution is light editing, if she has contributed to the passage at all. It is a fair assumption to make that this 3-gram has been used by two different authors to convey different meanings.
Adjectives are mostly used as (pre-)modifiers of nouns or complements of the verb be (Eastwood 2002 [1994], 251), and the former use will be investigated here. As a modifier, little may be used differently, depending on its meaning. Apart from meaning small in size, the word may also be used as an intensive, as a diminutive, referring to something pleasing (Merriam-Webster 2020), or even as a term of endearment. In Laura, little most frequently modifies house, followed by way, and tree. In Rose, little modifies house, followed by girl, and boy. However, little girl not only refers to children; the expression tends to belittle young women (e.g. Wilder Lane 1919; 62, 116, 130). In Little House, little modifies girl, followed by house, and way; its modification of girl is shown in (15):
-(15) “The boys played boys’ games on one side of the schoolhouse; the little girls played on the other side. Mary sat with the other big girls...” (Ingalls Wilder 1971 [1937], 157).
Here, little girls evidently means that a difference in age implies engaging in different activities during recess, although it is notable that children played separately.
The second most frequent adjective across all the corpora, good, modifies deal in Laura; all the occurrences there are used in the meaning favorable. Another two nouns modified by good are crop, in the sense of depending on just that to repay debts, and job. In Rose, good modifies time, job and prospect. In Little House, the adjective good most commonly modifies girl, night, as in the pleasantry, and time. Just as for little, 3–4-grams were researched for good, but the amount of information that was possible to extract was did not provide much insight.
In Laura, the adjective long modifies teeth in a bull rake, measurements of clothes, and braids of hair. Long is most commonly used in Little House to modify time, drive, and way. In Rose, long modifies time, letter, and line; the word line is even modified twice by it, as in (16) below:
-(16) “As far as I could see, covered wagons stood one beyond another in a long, long line” (Wilder Lane; in Ingalls Wilder 1971 [1962], 27).
A descriptive adjective that had slightly different meanings in the corpora was new. The adjective new is used in Laura to modify home, followed by house, and land. In Rose, new modifies clothes, house, and hat. It is notable that in Laura, the modified nouns have to do with necessary things for a farmer, which is the case in Little House as well, only to replace worn out items, while in Rose, two out of three modified nouns seem to be connected to luxury, as in (17):
-(17) “She fortified herself with a new hat and a veil with large velvet spots, yet at the very door she had a moment of panic...” (Wilder Lane 1919, 108).
It is also worth noting that large seems to be the preferred descriptive adjective in Laura, big in Rose, and that both big and large appear among the top ten descriptive adjectives in Little
House. Having addressed the common descriptive adjectives, I will now discuss the use of consecutive adjective constructions.
5.2.2 Adjective-Adjective Constructions
Adjectives were discovered to occur in collocation with each other during the generation of word lists, with and without commas in between. In (18) and (19), consecutive adjectives from Little House may be found (Ingalls Wilder 1971 [1937]):
-(18) “Mary was a good little girl.” (32)
-(19) “The sun was shining and all around the wagon was clean, wide space to be explored.” (9)
The objective here is to investigate differences between the corpora regarding the frequency of this occurrence. The number of normalized adjectives occurring consecutively are shown in Table 10.
Table 10: Adjective-adjective constructions in the corpora
Corpus Comma? 2 consecutive 3 consecutive 4 consecutive
Laura Yes 32.34 0.90 0.00
No 8.67 0.18 0.00
Total 41.02 1.08 0.00
Rose Yes 26.16 0.50 0.00
No 20.61 1.57 0.17
Total 46.76 2.07 0.17
Little House Yes 30.11 0.79 0.07
No 19.81 1.57 0.00
Total 49.91 2.36 0.07
The Laura corpus has the highest frequency per 10,000 words of two consecutive adjectives using no commas, at 32.34 hits. Little House is the next highest at 30.11 hits, and Rose is the lowest at 26.16. Constructions with two consecutive adjectives using commas are most frequently found in Rose, where there are 20.61 hits; Little House has 19.81 hits, and Laura has 8.67. The highest total normalized frequency of two consecutive adjectives is found in
