Unsettling the Slave Master: Resistance and Transgressive Behavior in a Caribbean Slave Colony

Abstract

Previous research has characterized those resisting slavery as quite atypical of the enslaved population: most of them being young, male, and engaged in particular occupations. In this article, we study transgressive behavior among an enslaved population quantitatively. We employ a unique census from the Caribbean island of St. Croix in 1846, which allows us to study not only the characteristics of those that transgressed the masters’ order in some way, but also to compare them with those of the entire enslaved population on the island. We find that the individuals in our dataset who transgressed the oppressive institution were, in many respects, quite typical of the entire enslaved population under study. Opposition to the oppressive system could be found among all groups of enslaved persons in the studied society. Nonetheless, we find that specific characteristics, such as marital status and gender, were more likely to be associated with transgression on St. Croix.

Introduction

It is by now well-established in scholarly research that many enslaved individuals tended to resist their status (for some recent literature reviews, see Egerton 2016; Sidbury 2010). Such research spans the study of slave revolts and rebellions (some key examples from recent years include Brown 2020; Genovese 1992; Gonzalez 2019; Harpham 2015; Hoffer 2010; Kars 2022; Smith 2005; Rediker 2013; Zoellner 2020), the history of runaways or marronage (e.g., Corneiro 2019; Foner 2016; Franklin and Schweninger 1999; Pargas 2021; Read and Zimmerman 2014), and maroon communities (e.g., Alston 2023; Kent 1965; Maris-Wolf 2013; Price 1973; Sivapragasam 2020). Less dramatic than full-scale revolts or marronage, yet no less important, was the everyday resistance and small-scale confrontations that many enslaved put up (Camp 2002; 2004; Craton 1982, 31–60; Gaspar 1992; Herskovits 1941, 99–105; Kafka 1997; Kolchin 1978; Stampp 1956). Such resistance is key to adequately understand what daily life entailed for the enslaved. However, analyzing everyday resistance empirically can be challenging due to sources’ constraints, albeit narratives by enslaved persons often testify to various forms of everyday resistance (see for example Blassingame 1977; Escott 1979; Fisch 2007; White and Burnard 2022).

Yet, just how common everyday resistance was, what forms it took, and who undertook it, is still not fully understood. A particular limitation of previous research concerns how typical or atypical those resisting slavery were. This issue is going to be in focus in this article. Scholars have argued that those individuals who resisted slavery were atypical of the enslaved population at large, hinting at selection bias (see for example Bailey 1980; Kolchin 1978). Rather than focusing on the representativeness of the characteristics of those that resisted slavery, previous research has mostly focused on the processes of slaver resistance (e.g., analyzing the enfolding of rebellions) or on narratives of particular individuals (e.g., analytical narratives of individual runaways). Some scholars have, however, attempted to provide a characterization of who a typical rebel would be. In common for all of these studies is that they all have focused upon resistance in the form of runaways. The typical runaway has been characterized as young, male, and doing itinerant work (Franklin and Schweninger 1999; Geggus 1985; Hodges and Brown 2019; Johnson 1981; Meaders 1975; Read and Zimmerman 2014; Wallace 2017; White 1980; 1991). This research is limited in two important ways. Firstly, there is a lack of research on other forms of resistance against slavery beyond runaways and their profiles. Secondly, the existing research is limited in that there generally is no corresponding detailed individual-level data on the characteristics of the entire enslaved population in these societies to compare the runaways with. Claims concerning the representativity – e.g., in terms of the gender, ethnicity, occupation, age – of the individuals in previous research have therefore remained quite vague.

The aim of this article is to contribute to the field of research on forms of resistance against slavery by quantitatively studying who, within a slave society, exhibited transgressive behavior. By transgressive behavior, we here mean behavior that challenged the social order in slave societies, including challenging the masters as individuals. This includes conscious resistance against the slavery regime, but it also includes behaviors where the enslaved individuals did not necessarily intend to challenge the oppressive system as an institution, but which the masters interpreted as challenging, either of the system or to the masters as individuals. Examples of the former could be running away or outright rebelling against the slavery regime, whereas examples of the latter could be absconding temporarily to meet friends or family, shirking at work, or talking back to a master. Our research question is: Who, and how typical of the whole enslaved population, were the enslaved persons who exhibited transgressive behavior?

To answer our research question, we draw empirical evidence from a uniquely complete slave census undertaken in a Caribbean slave society – namely, St. Croix in what was the Danish West Indies (current-day US Virgin Islands) in the mid-nineteenth century. The strength of these records is that they encompass the entire enslaved population on the island, enabling us to study whether the types of behavior that these sources provide evidence of were more common among certain strata of the enslaved population.

St. Croix resembled most islands in the Caribbean: a plantation economy almost entirely based on cash crop production, primarily sugar, for export (Dookhan 1994, chap. 5; Hall 1994; Olsen 2017; Sveistrup 1942; Sveistrup and Willerslev 1945). As for much of the region, production had, for a long time, relied nearly solely on enslaved labor. At its peak in the late eighteenth century, around 90 percent of the population on the island was enslaved, but the share had by the time under study in this article decreased somewhat (Theodoridis et al. 2024, tbl. 1). The decrease was due to demographic factors in combination with the international abolition of trade in enslaved people, prohibiting further imports (Gøbel 2016). Manumissions of enslaved people already living on the island further reinforced the trend towards a decreasing share of the population being enslaved (Hall 1994, chap. 8). The Danish West Indies had also experienced substantial resistance from the enslaved, including two large-scale revolts in the eighteenth century (albeit never on St. Croix), as well as substantial marronage (Hall 1985; Roopnarine 2010; Sebro 2013; Simonsen and Christensen 2023; Westergaard 1926).

Factors influencing resistance against slavery

The present study is most closely related to much previous research on direct forms of resistance against slavery, and especially quantitative studies of runaways. As is clear from this research, runaways could include everything from people absconding temporarily – hiding in a local forest, or going away to visit family members – to individuals trying to run away to freedom permanently (see for example Morgan 1985, tbl. 7; Mullin 1974, tbl. 3; White 1991, tbl. 15). According to the quantitative research trying to understand who these runaways were, one key factor is the gendered nature of this act, with every study finding that those running away were primarily men (see for example Franklin and Schweninger 1999, tbls. 3–4; Geggus 1985, tbl. 1; Hodges and Brown 2019, appendix table 1; Johnson 1981, 418; Meaders 1975, 292; Read and Zimmerman 2014, tbl. 1; Wallace 2017, tbl. 2.6; White 1980, tbl. 1; 1991, tbl. 19). Even in the absence of comparable data on the entire enslaved population, a limitation in all of these respective studies, the gender ratios of the runaways are many times so skewed that it seems reasonable to assume that the runaways indeed were atypical of the enslaved population in terms of their gender. One possible explanation suggested by several previous scholars could be that gendered norms – especially concerning child-rearing – among the enslaved put more pressure on women to stay put, rather than to run away (Franklin and Schweninger 1999, 210–11; Johnson 1981, 418; Read and Zimmerman 2014, 409; White 1991, 137). Other scholars have, thus, argued that resistance by enslaved women has been underestimated or ignored in much previous research (see for example Araujo 2015; Bush 1984; Camp 2002; 2004; Ellison 1983; Kafka 1997).

The age of the enslaved might also have been important: running away was more likely among the younger strata of the population, generally below the age of 40 (see for example Franklin and Schweninger 1999, 210; Geggus 1985, 124; Hodges and Brown 2019, appendix table 2; Johnson 1981, 418; Meaders 1975, 292; Read and Zimmerman 2014, 409; Wallace 2017, tbl. 2.8; White 1991, tbl. 20). This could be due to several factors, both in terms of the ability to escape (elderly enslaved being less able to undertake the physical challenge of this type of resistance), or the expected future gains from doing so.

There is no consensus in the previous research on whether or not individuals born in Africa – and hence likely born into freedom – were more prone to resist slavery than those who had been born enslaved in the Americas, as this might have differed by context. Persons born in Africa were thus quite common among the runaways in studies on colonial South Carolina or Saint Domingue (Geggus 1985, tbl. 1; Johnson 1981, tbl. V; Morgan 1985, tbl. 1; White 1980, tbl. 5), but less common in studies from other parts of the United States (Franklin and Schweninger 1999, 232–33; Wallace 2017, tbl. 2.13; White 1991, tbl. 14). Just how typical these individuals were of the entire enslaved population in their respective historical contexts is, however, not possible to tell from these studies as there is no corresponding data for the entire enslaved populations these study refer to.

The occupation of the enslaved might also have mattered in undertaking acts of resistance. Some scholars have found that more highly skilled enslaved laborers were quite common among runaways (Morgan 1985, tbl. 4; Mullin 1974, tbl. 2; Wallace 2017, tbl. 2.18; White 1980, tbl. 3; 1991, 124). Others have argued that individuals who were put to work in itinerant and specialized jobs had potentially more opportunities to run away than those holding more generic jobs or at a specific location only (Geggus 1985, 125; Johnson 1981, 424–25; Meaders 1975, 308–9; Read and Zimmerman 2014, 409–11). But the occupation might also have mattered in another way, by reinforcing incentives or disincentives to undertake resistance. One classic theory from the research on political rebellions is the relative deprivation theory. This theory suggests that people partake in rebellions or revolts if they belong in groups suffering from relative deprivation vis-à-vis other groups in society, i.e., if they feel they are treated wrongly relative to other groups (Goodwin and Jasper 2015, 53–54; Gurr 1971; for a review of empirical studies on the topic, see Østby 2013). This could have been important for slave rebellions, so that enslaved individuals who were assigned to positions of lower status would be more likely to resist the system of slavery than those who were assigned to positions associated with somewhat higher status.

Finally, research on modern social movements has emphasized a number of additional factors influencing why individuals may resist oppression. We are here able to study whether certain structural conditions influenced the probability to resist the oppressive institution of slavery (Goodwin and Jasper 2015, 54). One potential structural condition is the size and structure of the establishments where the enslaved individuals worked. This has, in other historical contexts, been shown to be important, for example, in the case of trade union activities (e.g., Kaufman 1983). A large establishment size might be a challenge for creating trust and well-functioning social networks within a group. Trust among the enslaved was undoubtedly crucial for their resistance (e.g., Kyles 2008; Lussana 2013). Michael Johnson’s research also suggests that the size of the group mattered when running away – most of the runaways escaped in small groups – for the very same reason (Johnson 1981, tbl. I). In addition, the size of an establishment might be associated with different management practices lending themselves to varying degrees of resistance.

Primary sources

The current study is based on the 1846 census of the island of St. Croix in the Danish West Indies. The census has been made available online by the Danish National Archive. The contents of the census was digitized as part of the construction of a large panel dataset on the economic and demographic history of the island (Rönnbäck et al. 2024). This census provides a unique opportunity to study not only the individuals’ transgressive behavior, but also those who seemingly were not involved in such acts or behavior. This can help enlighten us specifically on the individual and/or social characteristics that made such behavior more likely, and thereby how typical those who exhibited it were of the entire enslaved population.

The census contains information on the age, gender, religion, birthplace, and marital status of the population, including the enslaved members of the population. As the census was collected by plantation or geographical address in the towns, we also know the place of residence of all the individuals, as well as the number of persons living on and working at different locations. Most importantly for our study, the census recorded two additional pieces of information about the enslaved: how the masters evaluated their “moral character,” and whether they had ever been “punished” for some crime. We have not been able to determine the authorities’ intention when including these two latter questions in the census. It does, however, seem plausible that this information was recorded as part of the process of amelioration of slavery underway in the Danish colony at this time (Hall 1994, chap. 11). The census therefore provides us with a snapshot of a slave society on the verge of emancipation (something that happened quite suddenly two years after the census was taken).

A first key piece of information for our study is the information on “punishments” meted out against the enslaved individuals who ostensibly had committed some crime. The full query for this column in the census is “If ever as criminals punished by Judgement or by the Governor General’s Resolution and when and how punished.” Everyone convicted of, and hence punished for, a crime would thus presumably be reported as such in the census. The source does not report how or why the master would have kept this information, so while some masters might have kept records containing information on this, others might have just put down in the census what criminal punishments they remembered. It is thus possible some underreporting of some crimes, more likely perhaps on large-scale plantations. We would, however, not expect this underreporting to be very large. Slave laws in the Danish West Indies were for a long time draconian, with very brutal physical punishments meted out – including branding, mutilations, amputations, and the ubiquitous whippings. For repeated or more severe crimes, the punishment was death, and then often a very painful death at that. The acts that were criminalized in the slave laws were, as previous scholars have noted, preoccupied particularly with acts of resistance against the slavery regime (Olsen 2009, 6; for research on other Caribbean colonies, see, for example, Paton 2001). The acts criminalized included anything from congregating in public or running away, to disobedience, thefts, sabotage or violence against members of the masterclass (Boyer 2010, 26–30; Dookhan 1994, 154–56; Hall 1977, 174–75; 1994, chap. 3; Olsen 2009).

The majority of the masters only responded to the first part of the query – if the enslaved individuals ever had been punished – and generally failed to answer both how and when enslaved had been punished. Albeit the census did not explicitly inquire what crime the enslaved were punished for, this was, nonetheless, reported in some cases. John, a 38-year-old field laborer on the Cane Garden estate, had been punished for having run away and having been absent for two weeks. Furthermore, only in a handful of cases is the punishment spelled out: the 59-year-old carpenter Petrus was, for example, punished with 150 lashes and “wore irons for 6 months” albeit the reason for his punishment is not reported. As it appears, our source informs us on whether a person had been punished, rather than on the nature of the crime or the punishment meted out. This information is, however, relevant to us as masters also reported when the enslaved had never been subjected to any such punishments. The source thus provides direct evidence on whether or not an individual had been punished for a crime. The source thereby enables us to estimate the likelihood of being punished for a crime.

Many of the acts criminalized and punished have in previous research been considered as one or another form of resistance to slavery. In cases where the source explicitly reveals the crime committed, these are predominantly of three types: insubordination (e.g., threatening somebody), marronage, or thefts (see Table 3). All of these types of acts have been characterized as resistance strategies against slavery in previous research (e.g., Herskovits 1941, 99–105; Craton 1982; Gaspar 1992, 134), or as small-scale confrontations (Kolchin 1978), but we have no way of knowing the intent behind the individual acts in our sample. There are, furthermore, cases in the census that could have little or nothing to do with direct resistance against slavery. Kitty Roberts, a 56-year-old cook working in Frederiksted, was punished for “fighting in the streets.” The source does not report who this street-fighting woman had fought with. If she had been fighting a master, or some other person in authority, then this could certainly be classified as an act of resistance against slavery. In many slave communities, there was, however, substantial violence between enslaved persons (e.g., Forret 2015, 2008); the chance that Roberts was fighting another enslaved person is not negligible.

The transgressive behavior that the punishment information in the census potentially can reveal are just the tip of the iceberg. Masters (as well as the colonial authorities) had an interest in detecting and punishing criminal acts of resistance, no matter who the culprit was (Hall 1977, 184). But a number of crimes were undoubtedly never solved. Individuals committing crimes who were never identified – for example, people committing thefts or acts of sabotage who were never caught – would obviously not have been reported. We cannot, from our source, know how common the undetected and/or unreported transgressive acts were on St. Croix – we can only measure acts that were both detected and reported to the colonial authorities. Another limitation due to the very nature of our source – a census – is that it requires that the individuals were alive at the time the census was take and still living on St. Croix. The most brutal form of punishment allowed by the slave laws was death (Boyer 2010, 26–30; Hall 1994, chap. 3; Olsen 2009). Anyone subjected to this punishment would thus not appear in the source. Furthermore, anyone sentenced to transportation, and enslaved persons who successfully ran away from the island (Corneiro 2019; Hall 1985), would not feature in our source as they were no longer living on the island. The number of death or transportation sentences was, however, by the nineteenth century quite low in absolute terms (Simonsen 2017, fig. F). Yet another limitation would be if some masters were more inclined to report crimes to the colonial authorities than others. Slave masters had the prerogative to punish those that they enslaved for misdemeanors (Dookhan 1994, 154–56). It is possible that some masters – for example, rural planters – used this prerogative to punish enslaved persons themselves rather than taking the time to report crimes to the colonial authorities, due to the geographical distance to where the legal courts were located, even in cases where the colonial authorities in theory should have been notified. Urban slave masters may, on the other hand, have had easier access to the colonial legal system due to their geographical proximity. If that indeed was the case, we would expect an underreporting of crimes committed on rural plantations. In the analysis, we control for geography by including a rural dummy as a control variable.

A second piece of information that we will use as a complementary indicator of transgressive behavior is the information on the “moral character” of the enslaved. The column where the masters filled in information about the “moral character” of the enslaved persons was most often used to enter value judgments about the enslaved individuals – e.g., “bad” or “good.” In a smaller number of cases, the information was somewhat more substantial, referring to more specific character traits. In this study, we will use this information on the “moral character” of the enslaved as a second indicator of transgressive behavior. In the vast majority of cases, we do not know why the master characterized these individuals in negative terms. It is important to remember that this characterization was provided by the master, and does not probably reflect the real character or intentions of the enslaved very well. “Good” character traits, according to a slave master, entailed a number of characteristics – e.g., hardworking, obedient, and submissive (Simonsen 2017, 51). The terror and violence underlying the system of slavery undoubtedly led many enslaved persons to hide their true intentions from their masters (Blassingame 1972, 132–53; Harpham 2015, 261). Anecdotal evidence from the source, suggesting that this indeed could be the case, is the example of the leaders of the 1848 slave revolt on St. Croix: John Gottliff (a.k.a. Budhoe) from Estate La Grange, Peter Benjamin Rankin and Frederik from Mount Pleasant, Martin William from Ham’s Bay, Cancer from Mount Washington, Isaac from Estate Prosperity, and Moses Robert from Butlers Bay (Highfield 2018, 86; Holsoe 2009, 194–96). Several of these individuals can be identified in the census taken just two years prior to the revolt: Gottliff was there described by the master of the La Grange plantation as having an “indifferent” moral character, two of the others (Isaac from Estate Prosperity, and Frederik from Estate Mount Pleasant) are possibly classified in negative terms, but the remaining four were all described in positive terms as to their “moral character.” ¹ It does not seem far-fetched to assume that several of them had successfully put up a façade before their masters, while in reality they might have been scheming a rebellion.

Both these indicators – punishments for crimes, and the depiction of the individuals’ “moral character” – will exhibit ample proof of various forms of transgressive behavior. Although the intent behind the recorded actions cannot be definitively determined, there is little reason to doubt that much of the behavior indicated in the sources was likely deliberate and aimed at resisting either a specific slaveholder or the condition of enslavement itself. We therefore believe that the empirical evidence can give us some measure – albeit crude – of resistance against slavery. The two indicators might then be complementary: while the “punishments” information captures specific, and presumably quite serious, offenses (even if the source in many cases does not report what the exact offense was), the “moral characterization” variable would reflect a more vague and subjective impression by the master. To the extent that the information in the census reflects transgressive behavior with no intention of resisting the institution – i.e., punishments for criminal acts or negative depictions as to the “moral character” of the enslaved that have nothing to do with any resistance against the slavery regime – this will then act as noise in the statistical analysis undertaken in this article. What we can use this information for is thus to study whether punishments and/or negative depictions were more common for certain members of the enslaved population than for others.

Analyzing factors associated with transgressive behavior

In contrast to previous related studies, which have all lacked detailed individual information about the characteristics of the whole enslaved population, the source used in this study provides us with such valuable information. We are thereby able to robustly test the representativeness of those who exhibited transgressive behavior. To do so, we undertake a multivariate logistic analysis aimed at identifying the factors associated with the probability of having been punished for a crime, and an ordered logistic analysis of the characterization of the individuals’ “moral character.” This will help us shed light on how typical or atypical those exhibiting these types of behaviors were of the entire enslaved population. Our analysis is for both outcome variables expressed in odds ratios. An odds ratio above 1 means a higher probability of a negative depiction as to the “moral character” than the benchmark category, or a higher probability of having been punished for a crime than the benchmark category.

The information on punishments is standardized into a dummy variable: whether or not an individual ever had been punished for at least one crime. We classify the information on the “moral character” of the individuals into a categorical variable depending on how the master evaluated the individual – very negatively, negatively, neutrally, positively, or very positively – based on the wording in the source. Again, it is important to remember that this is an evaluation of how well the enslaved conformed to what the masters wanted, and not an evaluation of the individuals as human beings. Keywords in the source classified as negative include “bad,” “lazy,” “idle,” and many others. If the master had added the reinforcement word very (e.g., “very bad,” “very lazy”), we consequently classified this as very negative. We classify a characterization as neutral if keywords such as “common,” “ordinary,” or “tolerable” were employed. Positive keywords were “good,” “fair,” “diligent,” and several others. If the reinforcement word very was employed for positive words (e.g., “very good”), we correspondingly classified the characterization as very positive. As a result, our variable can take either of five values: from very negative to very positive.

In order to test whether there are gender differences as to transgressive behavior, we employ the information on gender from the source in the form of a dummy variable. In order to test whether persons holding certain occupations were more prone to transgressive behavior, we group occupational titles in the census by broad definition, dividing the sample into five categories: field laborers, craftsmen, domestic workers, others, and unknown. This allows us to distinguish those having a higher chance of being put to some itinerant or skilled work – such as craftsmen – from those more probable to have remained at a specific location – such as field laborers or domestics.

Place of birth is included in the analysis as a categorical variable. The three islands constituting the Danish West Indies (St. Croix, St. Thomas, and St. John) are grouped together into one category. Another category contains individuals born in Africa, as per the source, albeit no further specification as to where in Africa the person was born is provided. There are also a small number of other places of birth in the sample, the vast majority being other islands in the Caribbean. We group all of these other locations together into one category labelled “Other.”

In order to test whether the size of an establishment carried any importance for transgressive behavior, we use the number of enslaved persons recorded on a particular geographical location as an explanatory variable. As the census was undertaken by geographic location (by plantation in the countryside, or by address in the two towns on the island), we can easily calculate the number of enslaved persons per establishment. As we would not expect the relationship to be necessarily linear, we divide the size of establishments into a categorical variable taking three values: small establishments (1st–25th percentile of the enslaved population in the sub-sample, which in this sub-sample means 8 or less enslaved individuals); medium-sized establishments (25th–50th percentiles, i.e., 9–90 enslaved individuals); and large establishments (51st–100th percentiles of the enslaved population, i.e., establishments with more than 90 enslaved individuals).

Some previous research has, as we noted above, suggested that age was a crucial variable for certain acts of resistance, such as marronage. From the census, we know the individual’s age at the time of the census. Only in a handful of cases do we also know the year (and hence at which age) a crime had been committed. We can therefore, unfortunately, not test whether transgressive behavior was associated with age at the time of the act. We do, however, include age as a control variable in our analysis, as the probability of ever having been punished for a crime undoubtedly would increase with age. Age is then used as a continuous variable.

Finally, we also make use of several other variables available in the census – religious affiliation, marital status, and place of residence – as categorical variables in order to test whether individuals exhibiting transgressive behavior were typical or atypical of the entire population in any of these regards. All standard errors are clustered geographically, by the location (plantation or urban location) where the enslaved persons were reported to live. Robustness tests, including variations to the model specification such as introducing location fixed effects, are reported in the article’s online Supplementary material.

Overview of the data

Table 1 reports an overview of our sample of observations. In total, there were 16,480 enslaved persons who lived on St. Croix at the time according to the census (see Table 1, column A). Most of the enslaved persons were born in the Danish West Indies and were living on rural plantations at the time of the census, the majority belonging to the Protestant faith. The dominant type of occupation was the field laborer on plantations. We also find a large number of domestic or house servants (either in cities or on plantations), as well as different craftsmen (e.g., blacksmiths, carpenters, coopers, and masons).

Table 1. Descriptive statistics of dataset

Source: Danish West Indies Panel (henceforth DWI panel), (Rönnbäck et al. 2024).

Information on “moral character” was diligently filled in by virtually all slave masters on the island; information is missing for less than one percent of the enslaved population. Data are more frequently missing for the variable “punishments”: information for this variable is, unfortunately, only available for a sample of 4,176 enslaved persons, ca. 25 percent of the enslaved population (see Table 1, column B; a more formal analysis of missing data is provided in the online Supplementary material Table A1). The sample containing this information (column B of Table 1) is quite representative of the full population when it comes to most characteristics, including age and gender of the enslaved, their “moral character,” and their marital status (cf. columns A and B). There is, however, a certain difference between the full population and the sample in terms of their place of residence: while most enslaved lived in the countryside when looking at the full population (92 percent), only 69 percent of the sample lived in the rural areas. We also have a corresponding underrepresentation of large establishments (but since all of them are situated in the countryside, this variable is not statistically significant on its own as an explanation of missing data, see online Supplementary material Table A1). This is something that must be taken into consideration when analyzing our sample. It is also important to remember that our study is based on one census – i.e., a cross-section of data. We are, for that reason, unable to determine the direction of causality of any association that we may identify.

Who was depicted in negative or positive terms?

We begin our analysis by studying factors associated with being depicted in negative terms by the slave master as to the individual’s “moral character.” As we have nearly complete information on this variable in the census, we can study almost the full enslaved population on the island, including more than 16,000 individuals. Column A of Table 2 shows the results from an ordered logistic regression with moral character as the outcome variable.

Table 2. Factors associated with characterization of enslaved person or enslaved person having been punished for crimes (odds ratios, clustered standard errors in parenthesis)

Source: Rönnbäck et al. 2024

Note: * = statistically significant at the 10 percent confidence level; ** = statistically significant at the 5 percent confidence level; *** = statistically significant at the 1 percent confidence level. Standard errors clustered by geocode (plantation or urban location).

What the estimates show is that unskilled and unmarried men are the most likely to be depicted in negative terms by the masters. Men were, for example, according to these estimates, 29 percent more likely to be characterized in negative terms than women were. Married individuals were, 34 percent less likely to be characterized in negative terms. That marital status mattered might be an indication that family obligations would have had an impact on how the individuals behaved, at least overtly. We will return to the gender-patterns below, discussing in greater detail the gendered nature of the more specific information on the “moral character.” The older a person gets, furthermore, the higher chance of being depicted in negative terms. The latter is hardly surprising, as many of the positive characteristics were associated with being able to work hard, and ageing was consequently negative from the masters’ point of view.

Individuals holding more skilled occupations (craftsmen) or domestics are substantially less likely to be depicted in negative terms than the field laborers in our sample (29 percent less for craftsmen, and 23 percent less for domestics, according to the estimates). It is here important to remember that our data do not allow us to discuss the direction of causality, so we cannot unambiguously interpret this evidence as less transgressive behavior among these groups. It is, in this case, rather probable that some degree of reverse causality might have been in operation, so that individuals that a master believes had certain positive characteristics – e.g., diligent or intelligent – were the ones selected to receive a training for skilled occupations or elevated to more privileged positions. Vice versa, should someone holding these more privileged occupations exhibit signs of transgressive behavior – e.g., being insubordinate – it is possible that they would have been demoted from that position as punishment.

The other characteristics that we can analyze – including place of birth, place of current residence, the size of establishment, or the individual’s religious affiliation – do not seem to have been associated with how the individuals were depicted in terms of their “moral character.”

Who was punished for crimes?

We next turn to factors associated with our other outcome-indicator, punishments for crime. Column B of Table 2 shows the results from the econometric analysis employing this as the outcome variable.

The estimates suggest that enslaved men were considerably more likely to have been punished for some crime than enslaved women (with an estimated risk 351 percent higher than the women). This estimate would suggest a gendered pattern to criminal acts. We return to this issue below, discussing what we can deduce about the gendered pattern of criminal acts.

Another characteristic found in some previous research is that many rebels had been born in Africa, rather than in the Americas. The origin of the runaways may very well have differed in different historical contexts. The estimates in Table 2 do not suggest that Africans were more prone to have been punished for crimes in the case under study here: on the contrary, persons born in Africa were, if anything, seemingly less likely to have been punished for some crime than those who had been born on the Danish West Indies (47 percent less likely, according to the estimates). This result is, however, not robust to changes in the model, as shown in the article’s online Supplementary material (Table B1, models 3–6). There might also be survival bias in our estimates: the slave trade had in the Danish case been criminalized in 1803, so few persons born in Africa would have entered the island after that date. Those reported as born in Africa in the census were consequently significantly older than the average of the entire enslaved population (56 years of age, compared to the average enslaved person’s 29 years). While we control for the age of the individuals in our regressions, it is possible there is a particular effect for this group that the general age variable fails to capture fully. We can therefore only tentatively conclude that persons born in Africa at least were not more likely to have been punished for some crime than persons born in the Danish West Indies, but there might be a recollection bias, with an underreporting of crimes committed far back.

Yet another variable of interest in our analysis is the occupational status of the enslaved. Our results show no statistically significant association between the occupational categories and the probability of having been punished for criminal acts: the differences, in odds ratios compared to the benchmark category (field laborers), is not statistically significant at conventional levels for any of the occupational groups. This evidence would thus not lend support to the occupation of the enslaved being an important explanatory factor for transgressive behavior in our case. The results are furthermore robust to all changes in the specifications we carry out in our robustness checks in the online Supplementary material (see Table B1).

In addition, age was associated with the probability of having been punished for a crime if we include all age groups (increasing by 3 percentage points per year). This is hardly surprising as the census asked whether a person had ever been punished for a crime. If we exclude children below the age of 15 from the sample (see Table B1, models 3 and 6), the estimated odds ratio is reduced substantially, and the estimate is furthermore on the margin of being statistically significant.

We also analyze whether the size of an establishment mattered. Our expectation was that this would be associated with the probability of resisting slavery. We can, however, find no such relationship in our sample. The place of residence (urban vs rural) does likewise not seem to have been associated with having been punished for a crime. As the odds ratio for having been punished for a crime was not lower for the urban population, it does not seem as if the urban reporting bias in our sample affects these estimates much. Religious affiliation and marital status were, finally, not associated with the probability of having been punished for some crime.

What types of transgressive behavior did the enslaved exhibit?

Our source can also shed some light on the types of transgressive behavior that the enslaved exhibited. For a small number of cases, we know what acts individuals were punished for, reported in Table 3.

Table 3. Types of crimes that the enslaved were punished for, by gender

Source: Rönnbäck et al. 2024.

Prior to the analysis, it is noteworthy that the type of crime committed is available only for a small fraction (17 percent) of those reported to have ever been punished for some crime, requiring caution in the inference drawn upon these data. What we nonetheless can conclude is that the majority of all criminal offenses that do appear in our sample – i.e., insubordination, marronage, and thefts – all have been classified as examples of everyday resistance against slavery in previous research (e.g., Craton 1982; Gaspar 1992, 134; Herskovits 1941, 99–105). A few women were, in addition, punished for other acts: one for having neglected her own child, another for having caused the death of an unnamed child, and a third for having pretended pregnancy. It is certainly possible that some of these were also acts intended as resistance against the oppressive system, but this is potentially less clear-cut than, say, marronage (see Araujo 2015 and Ellison 1983 for examples of similar acts interpreted as acts of resistance). As can be seen in Table 3, there are furthermore some gender differences: many of the men were punished for thefts or robberies, whereas women were punished for running away to a greater extent, as well as for crimes related to child caring/rearing.

The information on the “moral character” in the census can provide additional insights into behaviors that the slave masters deemed troublesome. In several cases, the masters used keywords concerning the enslaved individuals that give some indication as to why a master evaluated them negatively. These are shown in Table 4.

Table 4. Negative characterizations of the enslaved, by gender

Source: Rönnbäck et al. 2024.

A few persons were described as either physically or mentally ill – whether these illnesses were real, or the individuals in question feigned illness as a resistance strategy is not possible to tell from the source (Gaspar 1992, 134). There were also a number of people described as drunkards, which, according to some scholars, may itself have been a form of resistance against slavery (Golden 2023). More importantly, our sources also shed light on a number of forms of transgressive behavior that seem to reflect a more direct and unequivocal resistance against the exploitation that the enslaved were suffering from. A fair share of the enslaved (13 percent of the sample for which this is known) were described as being insubordinate in various ways: the 35-year-old seamstress Ann Mary was, for example, described as “self-willed,” whereas the 24-year-old field laborer Daniel was labelled as “insolent.” Other terms commonly employed about the enslaved persons were “impudent,” “saucy,” or “quarrelsome.” Most of these were reported as never having been punished for a crime, so their insubordination had never reached a level where the masters found it necessary to report them to the authorities for some crime committed. Whether the master had meted out some punishment of their own is, as noted at the outset of this article, not possible to tell. Another fair share (16 percent of the sample) were described as lazy or idle. A few individuals were described as runaways (4 percent of the sample) or as thieves (1 percent of the sample): the 7-year-old field laborer Manuel was, for example, only characterized with the words “runs away,” as was the 54-year-old woman Cecilia. These character traits so far reflect quite well some of the key crimes committed by those who were punished (see Table 3). The most common negative characterization employed by the masters – for 52 percent of the sample for which the nature of the transgressive behavior is specified – were, however, that the enslaved persons were “meddling.” Exactly what was meant by this is unfortunately hard to determine, but a possible interpretation is that these persons might have tried to interfere with how the masters managed the establishment (including the oppression of the enslaved) in various ways, but potentially not in a manner serious enough for the master to consider it criminal.

Discussion

It is by now well-established that violence was an integral part of the system of slavery (e.g.. Baptist 2014; for references to much of the older literature, see Farnsworth 2000, 145). The physical violence was furthermore but one part of the repressive nature of the slavery regime. Just as important was the additional fear of punishment (Gutman and Sutch 1976, 58; Sutch 1975, 342). The very point of employing violence was as a rule not just to punish a particular person into submission, but also to terrorize the other enslaved – on a plantation, or even more broadly, throughout a whole community or society. The point of a brutal whipping was then to set an example for everyone else in the enslaved community. Punishments against the enslaved were for that particular reason often meted out in public (Altink 2002; Gutman and Sutch 1976, 59).

In this article, we examined how typical or atypical those exhibiting transgressive behavior were of the entire enslaved population. For that purpose, we analyzed what individual and/or social characteristics made such behavior more probable, using punishments for crimes and the masters’ depiction of the “moral character” of the enslaved as indicators of such behavior. While these indicators are far from perfect, we argue that they provide some indication on patterns among the enslaved population.

One key conclusion from our analysis is how typical, in so many respects, those exhibiting transgressive behavior were of the entire enslaved population. Persons committed for crimes could be found in all sub-categories of the population: among men and women; among persons born on the islands or in Africa; among all types of occupations and religious affiliations, etc. The same goes for the probability of being depicted in negative terms. More formally, we find no statistical association between the probability that they had been punished for some crime or depicted in negative terms, and a number of the explanatory variables that we explore in our analysis, including the occupation of the enslaved, their place of residence or the size of the establishment they lived and worked on, their religion or their marital status. One interpretation of this overall pattern is that slavery was such an oppressive institution that various types of resistance against it were not exclusive to any particular group of enslaved individuals.

Previous research on runaways has suggested that the type of occupation the enslaved were forced to do might have mattered for their propensity or chances to run away (Geggus 1985; Johnson 1981; Meaders 1975; Read and Zimmerman 2014; Wallace 2017; White 1980). While some have argued that this might have been due to the location of the work enabling some to run away more easily, others have argued that enslaved persons with higher learning (e.g., literate individuals) might have been more prone to run away. In our study, we find no association between our indicators of transgressive behavior and any of the occupational groups of the enslaved. While having certain occupations or skills thus might have made it easier for some individuals to take such a premeditated act as running away, the transgressive behaviors that we are able to measure here (with acts such as insubordination, idleness or “meddling” thus being very important) might not have been any easier for anyone possessing a certain skill set or occupation than for others who did not. It is important to remember, though, that we cannot determine the direction of causality in our sample.

Another important finding is that persons born in Africa were no more probable to be depicted in negative terms than persons born elsewhere, and they were as likely (or potentially even less likely) to have been punished for some crime than those who had been born into slavery in the Danish West Indies. If we interpret the latter as an indicator of acts of resistance against the slavery regime, the persons born in Africa were in our case not more prone to resist the institution than those born into slavery on the island. This case would then seemingly be in contrast to what David Geggus found in his study of runaways on Saint Domingue, where persons born in Africa were common among the runaways (Geggus 1985), but possibly in line with what scholars found in their study of runaways in the United States (Franklin and Schweninger 1999; Wallace 2017). One key factor here might be the different socio-historical contexts under study. Geggus’ study was concerned with runaways on Saint Domingue just prior to the Haitian revolution. Many of the runaways in that study were probably stolen from Africa quite recently. The slave trade had in the Danish case under study here been criminalized in 1803, more than forty years before the census employed for the study was taken. Few persons born in Africa would have entered the island after that date, and those reported as born in Africa were all significantly older than the average of the entire enslaved population. Persons who had been born in Africa, and also survived long enough to be included in the census underlying our study, might thus have become assimilated into the oppressive system, and therefore exhibit no higher propensity for transgressive behavior than other enslaved persons on the island.

That said, some groups were under- or over-represented among those punished for some crime or depicted in negative terms. One such group were married individuals: they were no less likely to have been punished for some crime, but they were significantly less likely to be depicted in negative terms as to their “moral character.” While the census does not allow us to study directly whether parenthood mattered, we believe the finding that marital status mattered for transgressive behavior is well in line with findings in previous research from other historical contexts. Family obligations – perhaps most importantly for women – might have been one variable of importance concerning resistance against slavery (Franklin and Schweninger 1999, 210–11; Johnson 1981, 418; Read and Zimmerman 2014, 409).

One of the most clearcut findings from our logistic estimates is that enslaved men faced a higher probability of having been punished for some crime compared to enslaved women, and also were more likely to be depicted in negative terms by the masters. This gendered pattern of transgressive behavior would be in line with most previous research that has characterized runaway slaves, at least, as predominantly male (Franklin and Schweninger 1999; Geggus 1985; Johnson 1981; Meaders 1975; Read and Zimmerman 2014; Wallace 2017; White 1980). We do, however, not believe that this necessarily ought to be interpreted as showing that men resisted their enslavement to any greater degree than women did. The evidence we present on the nature of the transgressive behavior, instead, leads us to a somewhat different interpretation: that enslaved men and women resisted slavery in different ways. This would be in line with research by, for example, Stephanie Camp on the gendered nature of resistance against slavery (Camp 2002, 2004). Some of this resistance might have been covert, such as go-slows (labelled “laziness” by the masters). Other forms of resistance could be more confrontational, such as talking back to a master (perhaps labelled as “meddling”), or explicitly refusing to work. From the limited evidence we have, reported in Tables 3 and 4, it seems as if the enslaved men and women exhibited somewhat different transgressive behaviors: men were more often punished for thefts and robberies, whereas women to a larger extent were punished for crimes related to children or childcaring, as well as for running away. Men whose “moral character” was characterized in negative terms were to a much greater extent characterized in ways we might classify as confrontational (“meddling”) than women, who to a greater extent employed potentially covert tactics (such as go-slows, labelled as “laziness”). This could then explain the gendered nature of punishments: the risk of apprehension would be much greater for confrontational methods of resistance, than for covert methods of resistance. This would also impact the estimated probability of having been punished for a crime. Our conclusion is therefore that we do find that there was a gendered difference in the transgressive behavior that we study. This was, however, possibly not a difference in terms of the probability for resistance against slavery per se, but in the nature of action employed doing so.

Conclusion

In this study, we studied the transgressive behavior exhibited by enslaved persons and the degree of representativeness of the individuals exhibiting such behavior compared to the whole enslaved population. We employ a unique census including detailed individual-level information on the whole enslaved population on the Caribbean island of St. Croix, a few years before the enslaved were finally emancipated. In particular, the census records two pieces of information that we use as indicators of transgressive behavior – namely, punishments for crimes and depictions of the “moral character” of the enslaved individuals. The indicators are certainly far from perfect, as many criminal acts, for example, likely went unrecorded in our source. We do, however, believe that they provide useful indicators of transgressive behavior broadly defined, but more specifically also of resistance against the slavery regime or the slave masters.

Previous research has in many cases emphasized how atypical those resisting slavery were of the entire enslaved population; they were, according to this research, predominantly young males, and engaged in certain, specialized occupations. Our results show that in the case under study, the island of St. Croix, transgressive behavior could be found among all groups of enslaved individuals on the island. In many respects – including their religious affiliation, their place of residence, their place of birth, or their occupation – those exhibiting such behavior were quite typical of the entire enslaved population.

However, we find that not all groups exhibited such behavior to the same extent. Importantly, married individuals were seemingly less likely to exhibit transgressive behavior. This is well in line with previous research on the topic, suggesting that family obligations, most importantly parenthood, played a crucial role. Gender also played a role for transgressive behavior. One the one hand, men seem to have been punished for crimes significantly more often than women and were also more likely to be depicted in negative terms. On the other hand, as far as we can tell, the types of behavior seem to have differed between enslaved men and women. This also had consequences for the likelihood of being detected and punished for particular acts. Our conclusion in this regard is therefore that women in the case under study might not necessarily have resisted the oppressive system any less than the men did, but that they resisted it in different ways than the men.