The Texas Historical Markers Program: Racial and Ethnic Narratives

: In this article we explore the text of the over 16,000 historical markers erected in the state since 1936, using GIS and corpus linguistics to determine the where , how , what , and when of how Texas memorializes its racial and ethnic groups. Unsurprisingly, our results indicate that the story of Texas is implicitly a narrative of white people. More interestingly, the term “African (Americans)” begins to be commemorated especially after the 1990s, but only in stories of community, religion, school, and children, as Texas historical markers do not to dwell on narratives of slavery, the civil rights movement, and lynchings. “Indians” and “Mexicans” in the 1930s and 1960s exemplify the most egregious case of derogatory semantics we found in the markers. As concerns racial and ethnic groups, in general they tend to be memorialized where they were historically present, whether or not such groups are still there. The analysis also reveals the increasing concentration of the markers in urban areas.


Introduction
Commemoration practices in the United States are often geared towards the establishment and reinforcement of identities, especially when it comes to race and ethnicity.The various groups that constitute American society have largely assimilated into a multicultural and multiethnic population, but they have also often engaged in brutal conflicts with each other, including the Native American genocide, slavery, and the civil rights struggle, to name a few [1][2][3][4].The shared memory of these conflicts has often helped the victims of violence unite and confront past and present injustices perpetrated by the dominant groups, with a reverberation in commemoration practices [5][6][7], as demonstrated by the removal, relocation, or renaming as of February 2021 of more than 160 Confederate statues in the aftermath of the murder of George Floyd [8].Geographers have tracked and interpreted the contested memories of the races and ethnicities of the United States through the lens of political and economic perspectives that reveal social injustice and that view the landscape as a text written and read by social agents [4,9].In doing so, the "new" cultural geographers who emerged in the 1980s contend that the common sense shared in a society is also artificially constructed by social agreements [10][11][12].The text and language metaphors are central to this theoretical framework that sees landscape as "communicative devices that encode and transmit information", like written and spoken words [13] (p. 4).This is the literary concept of intertextuality, the idea that all texts constantly write and rewrite each other.When defining a text as an object of interpretation, reading situates the text in a context, defined as an ideological structure that social members believe to be true [14].However, the endless revision of meaning relegates the context to an arbitrary and ephemeral status, one in which what is believed to be true today can be refuted and rejected tomorrow.This upheaval of the accepted common sense is especially evident during political revolutions, in which a new ideology and a new common sense replace the old [15,16].Still, this process does not always work as planned.To counteract the desire of the dominant social groups to establish an everlasting narrative, counternarratives such as the George Floyd protests may emerge.
Inheriting the metaphors of text, context, and intertextuality, geographers who study commemoration in critical perspective developed a concept of "textual politics", where language and narrative in the commemorative inscriptions forge one-sided interpretation of history [17][18][19].The power of text can stand up for social minorities when they accurately record their historical plight as a means of protest, but it is more often prominent as those in power decide what to include in or exclude from commemoration to promote an "official" narrative.In the United States, race and ethnicity are crucial to understanding textual politics because, as Bright et al. [20] suggest, Anglo whites and males wield their privilege to author inscriptions etched on memorials, silencing African Americans, Native Americans, women, and other minorities.Hanna and Hodder [21] demonstrate how geographers can study textual politics not only by counting the number of historical markers by topic but also by examining the inscription's accessibility and legibility on the landscape.
In studying the narratives of commemoration, geographers have borrowed theories and methods from linguistics [22,23], including corpus linguistics (CL), a technique which uses digital methods to analyze and interpret "big data" of text [24][25][26].Narratology, the art of temporally sequencing events, has also had a crucial impact on commemorative storytelling and its geographic implication.Thus, memorial facilities appropriate historical spaces as a narrative medium at various scales of analysis [27].Narratologists employ a poststructuralist conceptualization of text, one in which narratives are constructed by social agreements and understood differently depending on context.Since writing and rewriting allow multiple interpretations of a certain narrative, geographers' role is to anchor those readings in space, for example, by explaining regional uniqueness or by mapping a character's travel route.This anchoring process creates spatial patterns that work as narrative sequences and that help readers make sense of the complicated interactions that occur in a text [28][29][30].
Commemorative studies often adopt mixed methods analytical frameworks in order to take advantage of quantitative and qualitative tools and techniques and overcome the limitations of both [31].By doing so, researchers are able to question both where the memory is located and how it is narrated.As concerns the where, in part as a result of the "spatial turn" in the digital humanities [30,32,33], GIS is increasingly being used in collaborative studies on public memory [20,21,34].This interdisciplinary trend stems from humanities scholars' attempt to combine historical GIS, narratology, and textual analysis [35].As for the how, some geographers have tackled this question by conducting content analysis and discourse analysis.In the former, the researcher counts and analyzes the number of commemorative inscriptions, with emphasis on text and narrative topics; in the latter, the emphasis is on interpreting the author's intent and context in which the text was created, including power dynamics and issues of social justice [18,19,36].In this article, we study the how predominantly from a content analysis perspective.Geographers are especially interested in the underrepresentation and/or stigmatization of racial and ethnic minorities, under the assumption that uncomfortable realities are revealed only after debunking the ostensible innocence of everyday language.A characteristic of these studies is that they follow a deductive approach in which the researcher decides which terms are selected for analysis before reading the text [18,20,37].For example, Hanna and Hodder [21] group monument inscriptions by the predetermined categories "Native American", "Segregation or civil rights", and "Slavery or emancipation".
In our study we employ instead an inductive approach using CL, a technique that has been used by geographers [24][25][26], although not to study commemoration.CL computes and indexes large bodies of digitized text (a corpus) in search of grammatical, thoughts, and sentiments patterns-this distant reading of text mirrors what GIS does in its search for spatial patterns in geographic data [38].For example, word frequency counts provide a window into which themes are commemorated.In our study, "African American" emerged as a key element of commemoration because it is one of the most frequent terms mentioned in the corpus we examined.In doing so, CL shifts the focus of the analysis from the researcher to the commemorative inscription itself [39]-in other words, CL is especially suited to content analysis.Another characteristic of CL is that it can perform semantic analysis-this process is called tagging in corpus linguistics-and therefore may reveal broader patterns beyond the lexical meaning of individual terms; by doing so, words and phrases are placed in context to disambiguate their grammatical usage and implication, more fully satisfying the how question posed by discourse analysis [40,41].Combined with word frequencies, semantical analysis can reveal general or specific trends in corpora of millions of words [42].The result is that both the exceptional and the unexceptional emerge, as it should be, considering that the exceptional can open a window into what is not immediately visible.CL is especially useful when it comes to comprehensive reading of thousands of historical markers as opposed to the previous methods, which examine textual politics and conduct discourse analysis by selecting only a few sample inscriptions [17,19].In this article, we look at the Texas Official Historical Markers program through a combination of CL, narratology, and GIS to determine the where, how, and when of Texas racial and ethnic narratives.

Materials and Methods
A historical marker is a small, durable object etched with inscriptions commemorating historical sites, individuals, societies, events, and other significant topics.These inscriptions tend to be articulated narratives difficult to analyze at scale, as is the case in Texas, where, as of June 2022, there were more than sixteen thousand markers, for a total of over two million words.Texas' is by far the largest historical marker program in the United States (Virginia's, the second largest program, includes slightly below 2600 markers) [43].The Texas official commemoration program has been reviewed in some detail elsewhere, but never studied in its entirety as we do here [44][45][46].
The history of Texas is characterized by a myriad of conflicts and reconciliations among different groups (e.g., Native Americans, European Americans, Hispanics, African Americans, and Asian Americans) and by the central role played by its victorious war of independence from Mexico in 1836 (later to be known as the Texas Revolution).Many authors have highlighted Texas' uniqueness [47][48][49], but no researchers have used its vast commemorative program to answer the question of how the state tells its history.In this article, we look at how Texas markers record the history of the state's various racial and ethnic groups, in which narrative context (positive, negative, or neutral), and where and when these stories are told.
The Texas Historical Commission (THC) has administered the Texas Official Historical Marker program since 1962.The first historical markers, erected for the centennial anniversary of the Texas Revolution in 1936, were typically made of granite and came in different shapes, colors, sizes, and engraving styles; the text engraved was shorter than in today's aluminum plates [50].While members of the public can propose that a marker be erected, it is the THC that has the final say on the creation of a marker: as the guidelines clearly state, "the wording of the state marker inscriptions is the sole responsibility of the THC" [51].Typically, the first draft of a potential new marker's inscription is proposed by a county's historical commission, to be sent to the THC for final approval at the state level.After the manufacturing and dedication of a new marker, the THC updates its markers dataset every first day of the month on its official website [https://atlas.thc.texas.gov/(accessed on 1 June 2022)] and allows the free download of the dataset's latest version.Monthly updates include the addition of newly built markers, the correction of errors, and the filling up of missing information.Users can view the markers, as well as other state landmarks and the location of the state's historic cemeteries, on the THC's interactive map.
The marker dataset is provided in .csvand .shpformats to allow users to open, view, and modify the data in a GIS environment.The THC dataset we used is updated to 1 June 2022.Once downloaded, the data were preprocessed, which involved filling up missing information (year of erection, latitude/longitude, and inscription), correcting diacritic marks, and counting the number of racial/ethnic words per marker.The following userprovided online databases helped supplement the missing element of the official dataset: Our study covers the period from 1885, when the first marker was erected, to 2019.We excluded markers erected after 31 December 2019, to facilitate our analysis, which is organized by decades.Another reason for excluding the years 2020-2022 is that inscriptions on many markers erected during that period were missing from the THC dataset, replaced by the note "marker pending".The final dataset contained 2,141,918 words inscribed in 16,235 markers.
We performed corpus linguistics analysis on the Texas historical markers dataset using Wmatrix [41].The software performs three main functions.First, it generates two frequency lists.One tabulates all the words, while the other classifies words by part of speech (POS) based on the Constituent Likelihood Automatic Word tagging System (CLAWS4) [40].The word and POS frequency lists highlight the most prominent lexical and grammatical features of the text analyzed.Second, Wmatrix analyzes collocation-the occurrence of two or more words within a short distance of each other [41] (p.16)-and identifies statistically significant word combinations within a span of two or more words.There is no agreement on the best size of a word span, but for texts in English corpus linguists usually employ a span of four words to the left and right of the word analyzed [52,53].Collocation is often used to infer the narrator's underlying intention, ideology, or assumption when they use a word [24].Third, Wmatrix highlights semantic collocates to identify each word's role and usage within a sentence.For instance, users can organize the terms "happy", "sad", and "angry" into a single category of "emotion" to examine the sentimental discourse running through a corpus.To do so Wmatrix uses an automatic tagging system called the UCREL Semantic Analysis System (USAS), developed by the University Centre for Computer Corpus Research on Language (UCREL) at the University of Lancaster in the United Kingdom.The system uses a customized version of the Longman Lexicon of Contemporary English in which words are subdivided into twenty-one semantic categories identified by alphabet letters, which are then further subdivided into 232 subsets.For instance, Tag S relates to terms relative to social actions, states, and processes; S2 refers to people in general, and S2.1 to human females.Comparing word and semantic collocates is a fundamental step in discourse analysis because the former examines words within the text while the latter allows the researcher to start inferring the intentions and ideology of the narrator-the THC in this case.
Although Wmatrix can help answer questions of what and how, inquiries about the where and when also matter.The issue of when is worth examining because commemoration trends tend to vary over space and time.To look for spatiotemporal patterns in the text of historical markers, we used SaTScan to answer the question of where and when a certain word was most used.(For geographers' use of SaTScan, see [24,54,55].)We selected the Poisson model option to measure the probability of word occurrences in space and time; Monte Carlo replication in SaTScan enhanced the robustness of results by comparing 1000 independent trials-the original data plus 999 randomized permutations-to increase the p-value to 0.001 [56].The Poisson model of space-time clusters requires three input files: geographic coordinates, case, and population.The Texas dataset assigns a pair of latitude and longitude coordinates to each marker, which is therefore recorded as a point, and SaTScan determines the odds of a marker mentioning a certain word by chance, starting with a null hypothesis that all words in the text of marker are randomly chosen.A cluster is found once this hypothesis is rejected, indicating that spatial and temporal patterns in the usage of words are present and are statistically significant.
Output clusters are displayed as circles on the ground and are mappable in GIS.The circle becomes a spatiotemporal cylinder on the y-axis.The smallest possible cluster with time contains only one marker and has a radius of zero; the cluster can be so large to include 50% of all words in the dataset.There is no consensus on proper cluster size.As Kulldorff [57] did, in this study we set a cluster's maximum size at the 50% level.

Quantitative Summaries
Table 1 shows the twenty words most frequently mentioned in Texas markers, subdivided by categories.The first column lists the most frequent terms and the next four list the most frequent adjectives, common and proper nouns, and verbs.Most words are typically found in any English language corpus ("a", "the", "in", "of", and the like) but others are more specific ("Texas", "church", and "cemetery").Wmatrix detected word pairs like "United_States" and counted them as a single term, and did the same for "american", which includes "American", "AMERICAN", and "american".Wmatrix disambiguated each word's POS depending on its context and linguistic patterns, which allow differentiating adjectives from nouns for the word "native"."American", "mexican", "african", "indian", and "german" were also recorded both as nouns and adjectives, although they do not appear in the table as they are not in the list of top twenty terms by category.The automatic disambiguation process usually requires manual corrections, including in this study [42,58].For example, Wmatrix initially categorized as adjectives the word "civil_war" (mentioned 2216 times), "baptist" (1507), and "methodist_church" (1399), so we had to manually reclassify them as nouns.In the proper noun column, we also merged "U.S".(889) with "U._S".(537) and changed the ranking of the term accordingly."Texas_1936" (799) and "Texas_Sesquicentennial" (436) were removed from the list of the top twenty proper nouns because they appear at the end of many inscriptions to mark the occasion for the erection of the marker, as in "Erected by the state of Texas 1936" and "Texas Sesquicentennial 1836-1986" [50].Overall, the POS table supports the idea that Texas' uniqueness derives from its geography as a multicultural borderland [47][48][49].Given the marker's nature as a historical text, it is not surprising that most adjectives relate to time ("historic", "new", "early", and "old").The distinctive characteristic of the Texas narrative is more forcefully witnessed by race and ethnicity ("mexican", "african", "indian", and "german"), with other adjectives primarily referring to historical or geographical significance ("original", "nearby", "prominent", "oldest", and "present").Counter to the stereotype that "everything is bigger in Texas", "small" is found more often than "large", possibly to remark on the state's progress from a humble start to the current prosperity.Examples include a church which started as a "small building" ("Harmony Baptist Church" marker), a "small community" of ethnic settlers ("Gruenau Turn and Schuetzen Verein"), and a "small group" of people gathering in association ("The Woman's Study Club of Holland").Such examples of historical contrast are a popular literary technique used to add dramatic flavors to the storytelling.Geographic themes dominate the common nouns list, with a majority related to types of buildings ("church", "cemetery", "school", "building", "house", and "home") and others more general ("community", "area", "site", "land", and "property").As expected in a state program, there are references to jurisdictions ("state", "county").This wide array of geographic reference is due to the marker's versatile spatiality: markers can tell stories that have occurred not only at one location but also along a route or in a region [27].The geographic specificity of Texas stands out more conspicuously in the proper nouns list."Texas" is of course the most frequently used term but note that "Mexico" is more frequent than "United_States", due to Mexico being closely intertwined with the history and geography of the state, especially in the 1800s.The term "civil_war" is also prominent, due to the erection of hundreds of markers on the centennial anniversary (1965) of the end of the Civil War.
The preponderance of the past tense form of verbs attests to the historical and commemorative nature of the dataset.The terms "built" and "served" are recorded twice as a past tense and as a past participle.Following the three forms of the verb "be" at the top of the list, "became" registers both the passage of time and the change of landscape."Recorded" is mainly used as a signature, as in "Recorded Texas Historical Landmark"."Served", "died", "buried", and "erected" are characteristic of the 1936 markers celebrating the heroes of the Texas Revolution, which include military rank and affiliation, battles fought, and when they died.In 1936, markers were also erected along highways to introduce travelers to local history, typically with information about when a county was established and where its name came from.
In addition to their commemorative nature of places and events, the markers also tell the unique history of the peoples of Texas (Table 2).The five most frequent racial/ethnic words are "indian" (2055 times), "mexican" (1281), "german" (1256), "african" (1213), and "spanish" (884).Note that the totals vary from the table because we excluded markers erected in the 1940s and 1950s, a period during which only twenty new markers were installed.None of the racial/ethnic words in Table 2 were mentioned more than five times between 1940 and 1959.Included in the word counts are all forms of a term: singular and plural, upper and lower cases.Depending on context, these terms may refer to people, languages, or architectural styles.To explore the racial and ethnic theme, Table 2 includes terms that do not appear in the top twenty list but are variations of the five terms listed above, including "native", "black", and "negro" to testify to historical changes in American linguistic practice [59][60][61].The words' polysemy demanded a close reading to remove usages of no interest to this study, such as when "black" refers to a color or to a last name.Manual checking dramatically reduced the count of "native" from 1776 to 135, and of "black" from 1208 to 478 ("negro" had no use other than racial).Table 2 clearly illustrates that the official Texas historical markers program memorializes some groups more than others and that this changes over time.The word "indian" is the most frequent (2055) overall and also the most frequent until the 1970s."African" remains rare until the 1990s, when the terms started to be used together with "American" to replace "black" or "negro".The term "negro" came under scrutiny by activists-who favored "black"-in the 1960s [59], but the Texas markers program kept using it until the 2010s, although only in conjunction with the names of social organizations or buildings.The use of "black" almost disappeared in the 1990s but gradually regained popularity in the next three decades."African" suddenly appears after Reverend Jesse Jackson proposed the term "African American" in 1988.The most prominent feature of "african" is its increasing frequency of use, which stands in contrast with other racial and ethnic terms, whose popularity tend to come and go.Overall, the 1990s are a turning point for cultural diversity as the new entries "native" and "african" became more and more used.
The word counts in Tables 1 and 2 are absolute and therefore must be taken with caution when comparing across decades, as there is a risk of over or underrepresentation.Relative frequencies (Figure 1)-obtained by dividing absolute frequencies by total word count-are more appropriate indications of relevance.Note how the absolute frequency value (left) for the combination "indian+native" peaked in the 1960s, but its relative frequency-and therefore its prominent role as a topic for commemoration-was actually much higher in the 1930s.In the case of "african+black+negro", the relative frequencies confirm a steep increase in the 1990s and in the two decades that followed, but such an increase is not as strong as Table 2 would suggest.All other groups remained below the 0.1% value, except for "mexican" in the 1930s, a result of the 1936 commemoration of Texas independence (98% of the 1930s markers, or 1078 out of 1095, were erected in 1936).The term "spanish", in reference to the rulers of Texas before Mexico, mirrors the pattern of Mexico in most decades, in spite of Spain's defeat in 1821.Poyo and Hinojosa [62] (p.395) note that early Texas historians downplayed the Spanish colonial system as "pervasively backward, irrational, inferior" and emphasized the enlightening role of Anglo Americans against "ignorance and despotism".Overall, the analysis of relative frequencies flattens temporal differences.With the exception of "indian+native" until the 1980s, "mexican" in 1936, and "african+black+negro" starting in the 1990s, Texas is quite consistent when it comes to which groups are commemorated.Digging deeper, SaTScan reveals local differences that are not evident at the scale of the state (Figure 2), while also confirming that sites of commemoration tend to concentrate around the largest cities for public attention and support [5].This also tends to occur in commemoration practices outside of Texas, influenced by cultural traditions and the heritage of specific places [20,63,64].Figure 2 summarizes the results of SaTScan analysis, and Table 3 includes information on the statistically significant clusters identified in Figure 2. Note that each cluster s statistical significance is defined by the p-value and the log likelihood ratio (LLR): high LLR values indicate a low probability that a cluster may occur by chance.As concerns the p-values, a cluster is generally statistically significant when its pvalue is less than 0.1 (confidence level of 90%) or less than 0.01 (99%).Thus, the fifteen clusters in Figure 2 are all statistically significant.
The cluster for the combined terms "indian+native" is by far the largest in size and also the earliest in time (1930s-1970s).It is centered in the western part of the state, historically a frontier land into which Anglo settlers moved and where they encountered Native American tribes [47,49,65].Note that "indian+native" markers are also numerous in Fort Worth, Austin, and San Antonio, cities that have played a prominent role in the history of the "Old West".The large size of the western cluster tells us that the pair "indian" and "native" is dispersed enough that smallest, more localized clusters, do not emerge.As for the topics of the markers, they memorialize violent events for the most part, including whites fighting Native Americans at Forts Belknap and Clark (in Newcastle and Brackettville), a ranch established after the removal of nomadic buffalo hunters in northwest Texas (Lubbock), the victims of multiple Indian attacks (Junction), or a compassionate Indian agent murdered by a white man (Newcastle).Others memorialize the collaboration between Indians and the U.S. army, as is the case with Seminole scouts (Brackettville) and with Tonkawas serving the Confederacy (Newcastle).
The word "mexican" forms two clusters, one in South Texas that extends as far north as San Antonio and is the result of the settlement and migration of Hispanics to the area [48], and the other centered in El Paso in the western part of the state.While El Paso s cluster is large, small clusters, temporally and geographically concentrated, are found around Nacogdoches and Abilene in different parts of Texas.In Nacogdoches, four Digging deeper, SaTScan reveals local differences that are not evident at the scale of the state (Figure 2), while also confirming that sites of commemoration tend to concentrate around the largest cities for public attention and support [5].This also tends to occur in commemoration practices outside of Texas, influenced by cultural traditions and the heritage of specific places [20,63,64].Figure 2 summarizes the results of SaTScan analysis, and Table 3 includes information on the statistically significant clusters identified in Figure 2. Note that each cluster's statistical significance is defined by the p-value and the log likelihood ratio (LLR): high LLR values indicate a low probability that a cluster may occur by chance.As concerns the p-values, a cluster is generally statistically significant when its p-value is less than 0.1 (confidence level of 90%) or less than 0.01 (99%).Thus, the fifteen clusters in Figure 2 are all statistically significant.To conclude this part of the analysis, we compared the location of markers with population distribution at the county level, like others have done [20,64].To do so, we mapped 2020 census population data and superimposed the clusters just described for comparison; for population of German ancestry, we used the ethnic table from the American Community Survey in 2015 [66].The population was normalized by county total as in Figure 1 to allow for meaningful comparison.We also adopted the collective category "Hispanic" from the census to overlay the word clusters of "mexican" and "spanish", distributed in the similar regions.Moreover, the American Community Survey has a county-level table for Mexicandescent but not for Spanish-descent.Overall, the distribution of the Hispanic, German, and Black population tends to match markers clusters (Figure 3).In the case of Hispanics, they are and have always been especially numerous in South Texas and along the border, seeking cultural homogeneity and physical proximity to Mexico.Early German colonizers The cluster for the combined terms "indian+native" is by far the largest in size and also the earliest in time (1930s-1970s).It is centered in the western part of the state, historically a frontier land into which Anglo settlers moved and where they encountered Native American tribes [47,49,65].Note that "indian+native" markers are also numerous in Fort Worth, Austin, and San Antonio, cities that have played a prominent role in the history of the "Old West".The large size of the western cluster tells us that the pair "indian" and "native" is dispersed enough that smallest, more localized clusters, do not emerge.As for the topics of the markers, they memorialize violent events for the most part, including whites fighting Native Americans at Forts Belknap and Clark (in Newcastle and Brackettville), a ranch established after the removal of nomadic buffalo hunters in northwest Texas (Lubbock), the victims of multiple Indian attacks (Junction), or a compassionate Indian agent murdered by a white man (Newcastle).Others memorialize the collaboration between Indians and the U.S. army, as is the case with Seminole scouts (Brackettville) and with Tonkawas serving the Confederacy (Newcastle).
The word "mexican" forms two clusters, one in South Texas that extends as far north as San Antonio and is the result of the settlement and migration of Hispanics to the area [48], and the other centered in El Paso in the western part of the state.While El Paso's cluster is large, small clusters, temporally and geographically concentrated, are found around Nacogdoches and Abilene in different parts of Texas.In Nacogdoches, four markers surrounding the city hall refer to "mexican" fifteen times in total (1979, 2008, 2009, and 2019).In the case of Abilene, there is only a single marker, "Mexican-American/Americanization School", but in it "mexican" is used ten times (1997).This is not the case for the El Paso cluster: in only three markers the term occurs more than twice, with a maximum of five in the "Trinidad Concha" marker.
For those who know the history of the state, the size and location of the "german" cluster in central Texas is no surprise (the LLR value is second only to "native+indian", indicating high statistical significance).German migrants predominantly settled in Fredericksburg, New Braunfels, and Industry in the central part of the state, but two small clusters are found near each other (4.7 miles) in Dallas.The term "german" in the two Dallas clusters is used eight times in the marker "Sons of Hermann in Dallas" (2011) and eight times in two separate markers-four times each in "St.Paul's Evangelical and Reformed Church" in 1989 and "Zion Lutheran Church" in 2006.
As concerns "african+black+negro", two clusters are statistically significant, the first and largest is in east Texas for the decades 2000s and 2010s, the second includes Shamrock and Wichita Falls in the 2010s.The two cities share similar commemorative narratives centered around African American churches and schools.Additionally, a marker in Shamrock tells the story of African American soldiers helping move Native Americans to reservations ("Buffalo Soldiers at Fort Elliott", 2012) and one in Wichita Falls commemorates the influx of Black immigrants attracted by the oil boom of 1918 ("Dr.Annie Davis Roark", 2016).
Finally, the spatiotemporal clusters for "spanish" are not very well defined: most are small and far from each other, and the highest LLR score here is the lowest recorded (Table 3).The two clusters with relatively high LLR score include three cities which came to prominence during the Spanish colonial era from 1690 to 1821: San Antonio, Goliad, and Nacogdoches.The third around El Paso is spatially and especially temporally similar to the cluster for "mexican".In Amarillo, identical markers were placed around the city in 1965 to explain that the name of the city comes from Arroyo Amarillo, the Spanish name given to a nearby creek.The cluster, however, is short lived and in fact the term was not used in any other city marker from 1974 to 2011 ("American Legion Hanson Post No. 54").The fifth cluster is very small and only contains one marker in Wills Point: in it, the term "spanish" recurs six times ("Philip Nolan Expeditions into Spanish Texas", 2014).
To conclude this part of the analysis, we compared the location of markers with population distribution at the county level, like others have done [20,64].To do so, we mapped 2020 census population data and superimposed the clusters just described for comparison; for population of German ancestry, we used the ethnic table from the American Community Survey in 2015 [66].The population was normalized by county total as in Figure 1 to allow for meaningful comparison.We also adopted the collective category "Hispanic" from the census to overlay the word clusters of "mexican" and "spanish", distributed in the similar regions.Moreover, the American Community Survey has a countylevel table for Mexican-descent but not for Spanish-descent.Overall, the distribution of the Hispanic, German, and Black population tends to match markers clusters (Figure 3).In the case of Hispanics, they are and have always been especially numerous in South Texas and along the border, seeking cultural homogeneity and physical proximity to Mexico.Early German colonizers settled in the central part of the state, and they are still there.Blacks were typically taken to Texas against their will from the eastern U.S. cities and ports by slave traders and owners.After emancipation, freed Blacks remained in the eastern part of the state, often moving to its cities for job opportunities and a chance at creating strong communities, but the oil boom attracted African Americans to the northwest part of the state as well [67].Also notable is the lack of overlap, except in a few areas, between Hispanics and Blacks, with the two groups historically divided along a line that runs from Texarkana to San Antonio, a pattern that continues today [48].
ing to its cities for job opportunities and a chance at creating strong communities, but the oil boom attracted African Americans to the northwest part of the state as well [67].Also notable is the lack of overlap, except in a few areas, between Hispanics and Blacks, with the two groups historically divided along a line that runs from Texarkana to San Antonio, a pattern that continues today [48].
Finally, the clusters for "indian" are the only ones that do not overlap with current population distribution, the tragic result of the expulsion of Native Americans from much of the state and the scattering of those who stayed across Texas, including in some large cities.It is striking that while other ethnic groups have remained in the same places where they were originally-a history of survival-for Indians the markers tell a story of defeat.This is the dark side of the myth of the frontier that has captured so much of the state s imagination.

Qualitative Semantics
In this section we address the how and why of commemoration in Texas by looking at collocation.Collocated pairs of words can follow each other (e.g., "African American") or be separated by one or more words ("band" of "Indians", "Mexican" general Santa "Anna").Wmatrix generates collocation lists by single words or by semantic tags.The collocate lower-case and upper-case initials are listed separately ("School" and "school" as collocates of "African") as are plural and singular forms of a noun ("German immigrant" vs. "German immigrants").Close reading after processing in Wmatrix is a necessary step to remove nonrelevant collocations ("native stone").Wmatrix sorts results by the log likelihood (LL) value, which measures the probability of a meaningful association.Finally, the clusters for "indian" are the only ones that do not overlap with current population distribution, the tragic result of the expulsion of Native Americans from much of the state and the scattering of those who stayed across Texas, including in some large cities.It is striking that while other ethnic groups have remained in the same places where they were originally-a history of survival-for Indians the markers tell a story of defeat.This is the dark side of the myth of the frontier that has captured so much of the state's imagination.

Qualitative Semantics
In this section we address the how and why of commemoration in Texas by looking at collocation.Collocated pairs of words can follow each other (e.g., "African American") or be separated by one or more words ("band" of "Indians", "Mexican" general Santa "Anna").Wmatrix generates collocation lists by single words or by semantic tags.The collocate lower-case and upper-case initials are listed separately ("School" and "school" as collocates of "African") as are plural and singular forms of a noun ("German immigrant" vs. "German immigrants").Close reading after processing in Wmatrix is a necessary step to remove nonrelevant collocations ("native stone").Wmatrix sorts results by the log likelihood (LL) value, which measures the probability of a meaningful association.A collocation with high LL means that words pairing in the text is intentional rather than occurring by chance.All collocates in Table 4 are statistically significant: LL values above 15.13 are equivalent to a p-value of less than 0.0001.Table 4a-c allows us to inquire about how different groups are characterized in the historical markers.Overall, all five groups are associated with positive, negative, and/or neutral narratives.For "indian" (Table 4a), markers overwhelmingly describe violent encounters between white colonists and Native inhabitants, including "Indian raid", "Indian fighter", "hostile Indians", "Indian attacks", and "savage Indians".All other pairs in the table are the names of Indian tribes (e.g., "Comanche Indians"), with no judgment.The collocates of "mexican" (Table 4b) are a mix of military and cadastral terms-a result of Mexico's land grant policy and the conflicts that ensued with Anglo settlers.Later, mark-ers memorialize the role Mexican "descents" played in Texas history, from civil rights to everyday culture, in a vigorous affirmation of identity.The themes of commemoration for "german" stand in strong contrast to those for "indian", and emphasize cultural origin through immigration, language, family, and heritage.The only negative term, "German prisoners", refer to soldiers interned in camps in Texas during the two world wars.For the most part, "African American(s)" are commemorated in markers related to education, community, and religion, with only one pairing-"enslaved African"-testifying to slavery (Table 4c).Finally, for "spanish" the collocates refer to early exploratory expeditions as well as architectural terms ("Colonial", "style", and "Revival") (Table 4c)."Spanish" is also paired with "mission(s)", a center of religious conversion and practice as well as the social, administrative, and economic keystone of colonial Spain.As already mentioned, it is worth noting that while both Spain and Mexico ruled what is today Texas, negative connotations are associated more often with Mexico than with Spain, in spite of the arguably bloodier and more genocidal conduct of the latter.Considering that the two anniversaries of the Texas Revolution-the centennial in 1936 and the sesquicentennial in 1986-account for 10% of the total number of markers (1609 out of 16,235), we can confidently claim that hostility towards Mexico and Mexicans is in considerable part a result of the outsized role the Texas Revolution has on the collective memory of Texans [68].
In the last part of the analysis, we revisit collocation, shifting from lexical to semantical analysis-from content to discourse analysis.Automatic tagging also often required manual correction, as in "Indian reservation", which Wmatrix misclassified as an expression of doubt, and thus tagged as A7-in Table 5.We kept this and similar mistakes in the table as they are statistically significant but ignored the negative connotation.The semantic tags in Table 5 confirm the findings from the collocate analysis at the lexical level.Both "indian" and "mexican" are marked by negative or at least violent narratives, most evident in the prominence of tags G3, E3-, and their subsets: see, for example, G3c (infantry, cavalry, garrison) as a subset of G3 (raid, war, army) with positive signs occurring only in the sense of "belonging to a group" (tag S5+).Several neutral tags are collocates of these two groups, as in "native" Z2/S2mf (american), "indian" I2.1/S2mf (agent), and "mexican" I1 (grant).W3/M4 registers a perceived deep relationship of Native Americans with the natural environment.Finally, some hydrographic features in Texas are still named after their native name (e.g., Caddo Lake, Bowles Creek, and Navasota River).M tags mark the relation between "german" and migration, and in this narrative Germans also strive to improve their socioeconomic status in the new continent (T2+).Tales of "german" heritage (A9+/S1.1.1),language (Q3), and people and religion (S9/S2mf) also occur.The term "African" gained popularity in the 1990s, primarily in association with "american" (Z2) and "americans" (Z2/S2mf).The term "black" has a strong association with education (P1/S2mf) and children (S2mf/T3-)."Spanish" collocates with tags M7/S7.1 (colonial), which refers to both a political system and an artistic style.Immigration (M1) and American (Z2) are also significantly paired with this term, but Texas history adds a more distinctive flavor, with expedition (M1) and explorers (M1/S2mf) added to the mix."Spanish" is associated with language because many features of the natural (e.g., rivers) and built (churches) environment have Spanish names.Texas itself derives its name from the Spanish transcription of the Caddoan Indian term Teychas, meaning allies or friends [49].
Table 6 looks at the semantic tagging of the five racial/ethnic words by decade to examine how their characterization changed over time.For simplicity, the table only lists the most likely collocate per decade rather than listing the top ten as in Table 5.We also separated "native", "black", and "negro" from "indian" and "african" in order to trace when the transition in their use occurred.The most striking feature of this part of the analysis is that the topics of commemoration change from narratives of war and violent colonization to narratives of peace, development, and community.Each racial/ethnic group presents a similar trajectory, with some differences.For example, while "indian" has come to be associated with neutral collocates that refer to areas of settlement, movement, and villages (M7), the term "mexican" maintained for a long time its linguistic association with war, and even its association with politics (G1.2) is mainly due to its collocation with the Texas "revolution"."German" also has a strong relationship with immigration topics (M1/M7/S2mf), beginning in 1936 and continuing to this day and without interruption.In the 1970s, "African" and "black" started being collocated with religion (S9/S2mf) and community (S5+/O4.3c).The term "black" follows a pattern we already encountered in Table 2: decrease in use in the 1990s and recovery in the last three decades."Negro", on the other hand, formed linguistic pair with "servant" and "slave" (S7.1-/S2mf) in the 1960s and then disappeared, to be revived in the 2000s in conjunction with the names of African American organizations and buildings, such as the Houston Negro Chamber of Commerce (I2.1/S5+c) and the Cora Anderson Negro Hospital (B3/H1c).Finally, the term "spanish" had no particular connotation throughout the study period, being associated with exploration (M7/S7.1)and colonial architecture (M7/S7.1).Mission also tops the 1970s list, although with the already mentioned misclassification of tag Wanted (X7+).

Discussion
All commemoration practices are the expression of social forces and vary over time and space [5,18].Commemoration serves the present by celebrating selected events and people from the past and places and spaces are themselves a narrative medium [27,69].This dynamic is clearly present in the Texas Official Historical Markers program.In this article, we have chosen to focus on five racial/ethnic terms with high frequency of commemoration, but other groups are also remembered in the Texas markers, usually only locally and for only one or a few decades: "french" (mentioned 312 times), "english" (269), "czech" (267), "korean" (144)", "irish" (139), "swedish" (138), "italian" (101), and "polish+pole" (99).Interestingly, identity groups with few markers are primarily associated with neutral themes like immigration and culture, similarly to "german", and as is the case for German Americans, some of these groups are themselves members of the hegemonic group, which is the likely reason for the neutral feelings.These markers tell a story of migration and settlement, civic engagement, and religion.Moments of self-assertion, such as riots, strikes, mutinies, or civil rights events are rare in the markers' narratives for these groups.
The cases of "indian" and "mexican" in the 1930s and 1960s exemplify the most unfortunate case of derogatory semantics.As white colonists waged wars aimed at expelling native tribes from Texas, 1936 markers in particular offer a one-sided narrative of the story, typically recalling the tragic histories of white women or children murdered during raids and often exaggerating the brutality of Indian warriors [70].The collocates "hostile Indians" and "savage Indians" build a strong narrative framework that emphasizes emotional hatred and oversimplify the social, economic, and racial factors behind this confrontation.When atrocities are committed against natives they are most often ignored, if not celebrated as heroic acts of defense, according to a mythological narrative of the frontier that justifies violence as a product of the harsh environment the Anglo Texans encountered, together with their purported superiority over other races and a lack of self-doubt that defines their individualism and lawlessness [49].Another myth that looms large in the collective memory of Texans-the Texas Revolution-has been re-examined in recent years as an attempt by the Anglo Texans of securing chattel slavery and lucrative cotton businesses; this is a narrative that runs counter to the traditional view, which the markers reflect, that disguises the economic motivations of the event by focusing exclusively on a narrative of independence as an act of self-affirmation and heroism [68].A narrative of independence not only justifies violence, but it sanctifies it by aligning the Texas Revolution to the American Revolutionary War (1775-1783) [5].

Scholars have remarked on the relation between war memorials and nationalism in the
United States [3][4][5]68] and in Texas, too, markers memorialize fallen soldiers and veterans of the Texas Revolution, the Civil War, the Spanish American War, the two world wars, and the Korean and Vietnam wars.This commemoration serves to focus the public's attention on patriotic acts, events, and people, while discouraging and stigmatizing dissenters, "others", "noncitizens", and "enemies".
A term that is conspicuously missing from the list of ethnic and racial terms commemorated is "white", except in the form of "german", "spanish", "french", "english", "czech", "irish", "swedish", and "italian".For comparison, African Americans are grouped into the more general category of "african+black+negro" as a result of the loss of ethnic specificity caused by the characteristics of the slave trade.Asian Americans are, like "whites", memorialized as "korean", "chinese", "japanese", and "vietnamese", but of course these groups came to Texas in significant numbers only much later.Interestingly, Native Americans are uniquely commemorated both as a single undifferentiated group ("indian+native") and by nation ("comanche", "cherokee", "apache", "lipan", "wichita", and the like), in the latter case to distinguish the "good" Indian from the "bad" one [71].Texas markers tell in large part a story of colonization and of often violent and bloody encounters between different racial and ethnic groups and in this sense to say that the colonizers are "white" is redundant.For instance, the "Site of the McLaurin Massacre" marker reads (emphases added by the authors): On April 19, 1881, Catherine "Kate" Ringer McLaurin (sometimes McLauren) was with her three small children and 14-year-old Allen Lease in the garden when a band of Lipan Apaches started to plunder her home.Lease, thinking there were pigs in the house, went to investigate the noise and was shot and killed.Catherine was also shot, dying hours later, but her children were unharmed.Maud, age 6, went for help because her father, John McLaurin, was away.Neighbors gave chase for 70 miles before soldiers from Fort Clark took command.Soldiers trailed the party into Mexico, reportedly killing all but two.Note that only "Lipan Apaches" are identified by their ethnicity and that the word "white" is not needed because the marker assumes (correctly!) the audience already knows that all named individuals are white.
We have already remarked that the collocates of "african" consist in great part of positive terms related to community, religion, school, and children.Although "enslaved African" ranked twentieth in the list, what Hanna and Hodder [21] have noted for Virginia is true also for Texas: its historical markers prefer not to dwell on narratives of slavery, emancipation, the civil rights movement, raids, massacres, and lynchings.To give the THC credit, the agency in 2006 launched the "Undertold Marker" program to assess which topics and stories had not been memorialized, in recognition of the fact that the centennial markers of 1936 had placed some groups-African Americans, Mexican Americans, Native Americans-in an "unflattering or unfortunate context" [46] (p.60).As a result, the THC has erected more than one hundred Undertold Markers in the ensuing years, often challenging and offering a counterpoint to earlier narratives.However, this is often a recognition of an outcome rather than a reflection on process: recent inscriptions celebrate what African Americans have "achieved" as a result of the civil rights movement, but they omit the protests, crackdown, backlash, and violence that got them there.

Conclusions
Geographers have adopted the intertextuality idea that all texts write and rewrite each other.This applies to commemoration as well: as the social context changes, so changes who and what are commemorated.The Texas Revolution is an interesting example of these dynamics.In 1936, for the centennial anniversary, markers reflect the state-sanctioned viewpoint that sees "indian" and "mexican" as the counterforce to the founders of the Republic, in a narrative largely fruit of the writings of influential historians-George Pierce Garrison, Eugene C. Barker, Walter Prescott Webb, and T.R. Fehrenbach-that justified the revolt against the Mexican government and promoted the myth of the frontier and a narrative of individualism steeped in the American tradition.More recently, and galvanized by the civil rights movement, Hispanics, African Americans, Native Americans, and other minorities have started to counter these myths, either by highlighting their groups' contributions to Texas history [72][73][74][75] or by accusing Anglo colonists of committing racist crimes [75][76][77].Other scholars have emphasized the role of slavery and the cotton trade as motivations for the Texas Revolution [78,79] or have highlighted the atrocities committed by the Anglo forces under the guise of self-defense [80].This trajectory is partially reflected in the Texas historical markers narrative, with a more positive or at least neutral characterization of minority groups in recent decades, a thematic transition from war to peace, and the opening of the program to contributions from the public.These recent trends have also been observed at the national scale and counterbalance the one-sided narratives of the past, although some researchers have called for more proactive policies and coordinated efforts [7,18].Changes in the narrative of Texas commemoration are an example of poststructuralist intertextuality, in which a new text challenges outdated modes of interpretation.One remarkable aspect of this new sensibility is to be found in the THC's decision not to change the text of the centennial markers, even when they are known to be inaccurate or problematic.Instead, new markers are placed to counterbalance the narrative of old ones, thereby entering the two narratives in a conversation-an example of intertextuality by the state.This is unusual when it comes to commemoration, because as a rule new perspectives remove the legacy of old ones to promote new values [6,15,16,63].Interestingly, the THC occasionally edits the text of some markers, but those erected in 1936 are treated as special, not to be touched.As stated in the Texas Centennial Marker Policies [81]: . . .The inscriptions for some 1936 markers may be inaccurate, incomplete or confusing.However, because these inscriptions are part of the state's 1936 historic preservation effort and have acquired historical significance in their own right, the THC will not revise or alter 1936 inscriptions. . . .This echoes T.R. Fehrenbach's sentiment, as expressed in the second edition of his opus magnum, Lone Star [49]: . . .It has been said that each generation must rewrite history in order to understand it.The opposite is true.Moderns revise history to make it palatable, not to understand it.Those who edit "history" to popular taste each decade will never understand the past-neither the horrors nor glories of which the human race is equally capable-and for that reason, they will fail to understand themselves.The 1968 Lone Star was in some ways highly original. . . .I have seen no reason to change this, which makes the current edition an update, not a revision, from the ephemeral perspectives of the nineties.

Figure 1 .
Figure 1.Word frequency by decade.The charts show the absolute (left) and relative (right) frequencies.The 1930s and 1960s are connected by dashed lines to indicate the hiatus during the 1940s and 1950s.

Figure 1 .
Figure 1.Word frequency by decade.The charts show the absolute (left) and relative (right) frequencies.The 1930s and 1960s are connected by dashed lines to indicate the hiatus during the 1940s and 1950s.

Figure 3 .
Figure 3. Percent of population by county.Source: United States Decennial Census 2020.

Figure 3 .
Figure 3. Percent of population by county.Source: United States Decennial Census 2020.

Table 1 .
Word and POS frequency lists.

Table 2 .
Word frequency list by decade.

Table 6 .
Semantic collocate list by decade.