Faces of violence: homicide trends and cultural meanings: Amsterdam, 1431-1816.

It would be hard to imagine a historical setting without violence. Throughout the ages, for example, the rulers of states have waged war against their enemies or employed police forces against lawbreakers. The focus of my research is what their subjects did to each other in ordinary social intercourse. In the historiography of interpersonal violence we can distinguish two largely independent approaches. The first, influenced by anthropological notions, may be termed the cultural approach. In it, violence is analyzed primarily in qualitative terms. The emphasis lies on concepts of honor, ritual behavior and the meaning contemporaries attributed to violent acts. Natalie Davis pioneered in this type of study with an article, published two decades ago, on religious riots in sixteenth-century France.(1) She interpreted them as symbolic acts aimed at a purification of the community, borrowing concepts from the anthropologist Mary Douglas. Davis took up the subject anew in a later work concerned with French royal pardons, mostly in homicide cases.(2) Robert Muchembled based his study of social life in Artois on the same type of source; in his case, pardon letters addressed to the Burgundian and Habsburg overlords.(3) Both authors stress the ritual quality of violent behavior in the fifteenth through seventeenth centuries. Parallels for a more recent period are to be found in Frevert's book on duelling.(4) She extends the scope of the cultural approach by analyzing the role of violence as a constitutive element in male self-perception. Next to these historians, the anthropologist Anton Blok has been studying violence in relation to concepts of honor and infamy both in past and contemporary societies.(5) In a recent programmatic article he concludes that a scholarly interpretation of a violent incident is bound to be defective if it does not take sufficient account of the situational context.(6)

The second approach is primarily quantitative and mainly concerned with homicide. Its representatives include Roger Lane, one of the first American historians to embark on this kind of research, and J. S. Cockburn, whose recent study of an English county covers four centuries.(7) Most scholars working within this tradition consider the incidence of homicide as an indication for the level of violence in past societies. They follow criminological usage in taking as their measure the homicide rate, defined as the annual average of killings, over a specified period, per 100,000 population in a given area.(8) A rise or a decline in this rate would mean that the society in question became respectively more or less violent. However, the available data so far do not cover all societies evenly. We are relatively well-informed about national trends in the statistical period.(9) For the preceding centuries, lacking national statistics, our evidence is largely limited to England. In that country, the homicide rate underwent a long-term decline from the thirteenth century until its reversal in recent decades. The trend is unmistakable, but the corpus of studies on which it is based leaves some questions of methodology undecided.

This article has three aims. First, I want to indicate ways in which the cultural and statistical approaches may be integrated. This can only be done with the help of a solid theory of long-term change. A major conclusion will be that it is crucial to gather contextual evidence on homicide cases. Second, I present quantitative and qualitative data on homicide in Amsterdam from the fifteenth century to the beginning of the nineteenth. They are based on my own archival work, supplemented with information from a few Dutch publications. As far as contextual evidence is concerned, I am concentrating on the killer-victim relationship. Woven into these discussions, the third aim, is a critique of some of the methodological premises implicit in earlier historical investigations of homicide. Together, the three efforts constitute a formidable task and I have to admit in advance that the Amsterdam data, gathered so far, provide just a partial answer to the set of problems raised. I hope, however, that my reflections may be useful as an agenda for future research.

The longitudinal set of data collected in statistical investigations provides an unambiguous picture. Even skeptical historians, who distrust grand theoretical schemes, admit that the bare figures indicate a trend. The statistical approach has led to two major findings: a long-term decline in homicide rates from the late Middle Ages until the middle of the twentieth century (the recent rise in the rates, intriguing as it is, is beyond the scope of my present reflections) and an increasing proportion of family homicides. The second finding will be commented upon below. Absolute figures have been calculated for enough English towns and regions in order to construct an average national trend. In England the aggregate homicide rates declined from about 20 (per 100,000) in the thirteenth century; to about 15 toward the end of the Middle Ages; approximately 7 around 1600; between 4 and 5 in 1700; around 2 in 1800; approximately 1 at the beginning of the twentieth century.(10) Although these figures are national averages and the local or regional studies on which they are based diverge in methods and types of sources used, the overall trend is so marked that it must be acknowledged as real. The more haphazard figures available for other European countries in the pre-statistical period generally support the hypothesis of a decline over the long term. An incomplete set of Swedish data suggests that the incidence of homicide steadily decreased in that country, especially between the 1630s and 1760s.(11) Homicide rates reported for a few towns and regions in Continental Europe in the fourteenth, fifteenth and sixteenth centuries are as high or higher than their English counterparts.(12)

Moving to possible explanations, several authors argue that the evidence about a long-term decline in homicide serves to confirm Norbert Elias's theory of the "civilizing process." According to Elias, the personalities of individual people changed as societies became more complex. Notably, the gradual monopolization of violence by state organizations was accompanied by a socio-psychological process leading to an inner pacification of aggressive urges.(13) Thus, the long-term decline in homicide resulted from a taming of aggressive impulses in daily social intercourse, which in its turn was a function of the rise and growth in power of states in Europe. Elias's own work mainly covers the twelfth through eighteenth centuries. Carol and Peter Stearns, however, found comparable changes in the personality structure of Americans in the nineteenth and twentieth centuries. According to these scholars, the overt expression of anger met with increasing social restrictions during this period.(14) Although anger is not quite the same as violence or aggressiveness, their findings may be taken as support for Elias's theory. To be sure, what they claim to observe is a transformation in emotional standards (the "emotionology") with respect to anger, which may indicate a parallel change in actual behavior.

One factor complicates the seemingly straightforward exercise of linking the long-term decline in homicide to the "civilizing process." Elias's theory does not simply or merely posit a decline in aggressiveness. Rather, the economy of emotions as a whole was increasingly subject to constraints and regulation. The behavior of individual people became more evenly controlled and directed less by the impulse of the moment. Although Elias himself never explicitly said so, the possession of a psychic equipment of emotional control in theory allows for the use of controlled violence. A carefully planned act of killing, in which the victim is "depersonalized," may be compatible with a "civilized" personality on the part of the perpetrator, or, in the Stearns' terms, an emotionology which constrains the expression of anger.(15) Consequently, we cannot automatically assume that homicides of this type will be rare in complex societies. On the other hand, it is plain that most forms of violence, also when an act is carefully planned, are valued negatively in the contemporary Western world. These qualifications definitely suggest that different types of violence can have different meanings and that this may be important for a theoretical understanding of the long-term development of homicide. Different meanings is precisely what it is all about in the cultural approach to the history of violence.

The most valuable contribution of the cultural historical approach lies in the "discovery" of ritual violence and its important role in preindustrial Europe. Notably in village and neighborhood communities an intimate relationship existed between violence and popular ritual. Fights among rival youth groups or between the male inhabitants of two adjacent villages were a common occurrence. The fact that these clashes were customary rituals, in no way prevented them from being tough and serious at times. Festivals often were the occasion, with a relative absence of violent behavior on normal days. Almost always violence was bound to the symbolic hierarchy of the village or neighborhood community.

Homicidal incidents in local communities, too, seldom implied just straightforward aggression. Attackers were led by cultural codes when deciding which part of their adversary's body to hit. To injure the other's head with a knife, for example, particularly humiliated him, even though, in order to kill him, it was less efficient than stabbing in the belly.(16) The jealous husband in sixteenth-century France, who murdered his unfaithful wife on Magdalen day, provides an example of a homicidal incident linked to the festive calendar.(17) The studies from which these examples are taken suggest that ritual violence was especially prevalent in the late medieval and early modern periods, which implies that it declined afterwards. Indeed, Muchembled says that the traditional fights among youth groups became criminalized and were suppressed from the seventeenth century onward.(18) The custom of duelling enjoyed a greater longevity, but it became obsolete in the first half of the twentieth century.(19) In the elitist culture of the duel as well as with clashes in village or neighborhood communities, the triad violence-honor-ritual was a crucial element. The message inherent in a violent encounter, whether or not leading to death, was at least as important as the bare physical events.

Although ritual aspects of violence are a central theme in what I termed the cultural approach, they were not the only contextual aspects identified by historians. Several authors, notably those working on the Middle Ages, were struck by the hot-tempered quality pervading the sociability of the age. Hanawalt, speaking of fourteenth-century England, concludes that "men were quick to give insult and to retaliate with physical attack." Three to four out of every five homicides were motivated by some sort of argument.(20) Tavern brawls, too, might easily end in the death of one of the contestants, whether or not passions had been inflamed by alcohol consumption. Such incidents were fairly frequent in fourteenth-century Venice and they formed the context of the majority of homicides committed in fifteenth-century Utrecht.(21) Impulsive violence, then, seems to have been quite common in the Middle Ages and possibly even later. Needless to say, Elias's theory could very well account for this. So it is not surprising that a group of Dutch sociologists and criminologists, several of whom are inspired by the work of Elias, also employ the concept of impulsive violence. In their studies of aggression and crime in the Netherlands today they especially search for its opposite, which they usually define as instrumental violence.(22) However, from the point of view of a strictly formal analysis, "impulsive" and "instrumental" are not opposites. Moreover, this dichotomy leaves no room for the ritual aspects which figured so prominently in the cultural approach.

Hence, I propose a slightly more complex model. The simple dichotomy is replaced by a system of two related but distinct axes. The one has as its opposites impulsive vs. planned or "rational" violence; the other has ritual or expressive vs. instrumental violence. The axes are distinct, because their determining factors are different. In the first case it is the actor's personality. A carefully planned murder out of jealousy or revenge, for example, may be called rational, even if the perpetrator is caught. This axis is the one associated most closely with Elias's theory and the observations on which it is based. The social meaning of the act is the determining factor in the second axis. While ritual violence is guided by the implicit cultural codes of the community, its instrumental counterpart is primarily a means to an end: usually to exploit the victim's property or body. The word "axis" was chosen deliberately, since we have to do with a continuum. No attack, for example, is absolutely instrumental or absolutely ritual. In principle, every violent incident can be characterized by its position on this axis, either closer to the one pole or its opposite. The same can be done for the other axis.

This, finally, enables me to link the cultural and statistical approaches. If one violent incident can be assigned a position on the impulsive-rational and ritual-instrumental axes, the average of a hundred incidents can be plotted there as well. Admittedly, the interpretive exercise needed for that is far from easy and it would be useful if some standard criteria were available. When the qualitative character of violence has been quantified in this way, the results facilitate a diachronic analysis. For the study of homicide it means that, apart from establishing the absolute figures, we have to collect contextual evidence on as many cases as we can. In view of the work already done, the hypothesis seems justified that long-term trends moved from a dominance of impulsive violence to a greater share of rational violence and in the direction of a marginalization of ritual aspects and a growing prominence of instrumental aspects. Support for that hypothesis is to be found in the studies cited above: ritual as well as impulsive violence appear to have been especially characteristic of past societies. Less evidence can be cited for contemporary societies, but a collection of recent studies by Dutch criminologists is suggestive. They found that a significant number of killings were planned, some of them taking place in the sphere of professional crime. Moreover, the use of controlled violence constituted a behavioral ideal among street robbers.(23) If there really has been a qualitative shift in the character of homicide, its exact timing would be a matter for future research.

The quantification of contextual aspects has not been entirely neglected in historical studies of homicidal violence. Most notably, we find it in the identification of family homicide as an important theme of research. This leads to a consideration of the killer-victim relationship as a variable to be studied, since a proportion of family victims necessarily implies a proportion of non-family victims. The killer-victim relationship is a traditional subject in criminology, but it has received less attention from historians of preindustrial Europe.(24) In criminological literature various classificatory schemes in this respect are to be found.(25) So we can take our pick. At the very least, we should distinguish an intermediate category of victims who are neither family nor total strangers. Besides, I think that the concept of family homicide is too limited. The intimate quality of the mutual relationships of spouses, parents and siblings seems to be the crucial factor here, which is suggested by the fact that historians explicitly exclude from the family category apprentices and servants, even though in preindustrial Europe they were considered as belonging to the household. Therefore, I propose to distinguish instead an extended category of intimate victims. They include unofficial partners, mistresses and lovers.

Supposedly, the proportion of homicides on intimates will be found to be increasing over time, as was the case for the proportion of homicides on family members in England. In that country the percentage in question rose from about 8 in the fourteenth century, to about 20 in the seventeenth and nearly 50 by the middle of the twentieth. It is difficult to interpret these figures from the perspective of family history, with its huge and undecided controversies.(26) Alternatively, we may look for explanatory suggestions in the Stearns' work. They demonstrate that efforts to regulate anger in America from the 1830s onward, had the family as an important target.(27) It may be hypothesized that, as the dominant emotionology increasingly banished anger in marital, parent-child and other intimate relationships, the customary rules for its expression were gradually forgotten. This would increase the chances for tensions to remain submerged and, in extreme cases, to explode suddenly in homicide. Of course not every killing of an intimate person was the culmination of a previous history of tensions. Thus, a situational analysis along the lines outlined above still is imperative. It is difficult, however, to combine the killer-victim aspect with the other contextual aspects as conceptualized in my axes. For example, slaying one's lover after years of intense but submerged conflict, seems to be neither impulsive nor rational. To a large extent, the axes of violence on the one hand and the killer-victim dimension on the other are independent variables; with one major exception though. When someone kills a total stranger, there is no previous history per definition. Such an incident, therefore, suggests impulsive violence, associated with a relatively unrestrained expression of anger.

The classification in terms of intimates rather than just family implies a methodological innovation. It follows from my concern to focus on the qualitative character of violence rather than on formal categories. Another methodological critique has to do with the type of source used. It is understandable that those who first tentatively construed an overall trend, simply laid all available sets of data next to each other. For England, this means that the national average is based on studies making use of indictments (hence, cases that came to court) as well as studies making use of coroners' inquests (i.e. inspections of all suspect bodies found). The time has come to be more sophisticated. Criminal prosecutions simply cannot be trusted as a source for the incidence of killing. The drop in homicide indictments in Surrey and Sussex around 1700, for example, partly resulted from an unwillingness of the courts to initiate cases against unidentified offenders.(28) Even more dramatic evidence comes from sixteenth-century Amsterdam. Jean Jungen compared the criminal records and the judicial body inspections over the years 1524-65 and he found that only one out of every nine homicides committed in Amsterdam in that period led to a criminal prosecution. The killers who escaped a murder or manslaughter trial either had fled or were reconciled to the victim's relatives.(29) This definitely suggests that criminal prosecutions constitute an inferior source for the construction of homicide rates. Reports of body inspections are to be preferred at all times.

If those reports are not available, an exception can be made for the study of early periods; that is, if the court records yield relatively high homicide rates. The latter qualification is necessary. No conclusion can be drawn from underreported figures if they are low in a period when we had expected them to be high. If, on the other hand, the rates in early periods, even though underreported, are extremely high, they definitely indicate a trend. This is the case with the figures for fifteenth-century Amsterdam, presented by Boomgaard.(30) They refer to cases of homicide with identified killers, most of whom had fled the city. Records are extant for eleven fiscal years during the period 1431-62. The names of fifty-four killers were known and together they made forty-four victims. Consequently, the homicide rate was c. 47 or c. 59, depending on whether the victims or their aggressors are taken as the measure.(31) For reasons of comparison the first should be done, since all figures for later periods are based on body inspections. The actual homicide rate possibly was higher still, as there might have been cases with unidentified killers which remained unrecorded.

As noted above, the figures for sixteenth-century Amsterdam, presented by Jungen, are based on reports of body inspections. Representatives of the court carried them out and they concluded whether or not the person in question had died violently. With the help of Jungen's data we can calculate the homicide rates for the years 1524-65 and 1560-90. The rate was about 28 in the first period and it lay between 21 and 24 in the second.(32) The sixteenth-century body inspections were not registered separately but inserted in the registers of criminal sentences. After 1600 this was no longer done. Presumably, inspections still were performed in the first half of the seventeenth century, but the reports have not been preserved. The extant series of separate registers starts in the late 1660s and with a few gaps it continues until August 1817.(33)

The professionalization which had progressed in the meantime, causes a technical problem for today's investigator. It was no longer the schout and the schepenen who did the inspections. The Amsterdam lector of anatomy, assisted by two surgeons, drew up the reports.(34) In their conclusions they seldom went beyond the boundaries of their professional competence: they were content with listing the medical causes of death. Presumably, the court had further information on the circumstances, but the reports are silent about it. Only in some cases a clerk noted this in the margin. Especially during the first decades of the eighteenth century it was often added that a killer had fled or had been caught. In other cases, fortunately, the sort of wounds inflicted often leave no room for doubt, while additional medical information is sometimes helpful too. Still, in a minority of cases it proved impossible to decide whether an inspected body had become the victim of homicide or an accident or whether the person in question had been killed by his own hands or those of another. Those uncertain cases figure in table 1 as possible homicides.

One more set of inspections, those performed on dead babies, has been omitted from my calculations. To refrain from a discussion of infanticide follows from the theoretical perspective which I adopted. Homicide was studied as an indication of the level of violence prevailing in a particular society. That makes it imperative to keep infanticide figures entirely apart.(35) They tell us little about people's propensity for aggression. The story behind infanticide is usually that of desperation, shame and conflict with the community's sexual norms. The perception of the act by the killer and those around her could be diverging too. While many unmarried mothers may have thought themselves excused because they saw no way out, other people considered their offense a major assault on Christian morality. The tremendous differences in the social contexts of infanticide and homicide oblige modern investigators always to discuss the two crimes separately. Here I am concentrating on the violent death of non-infants, including children and adolescents.(36)

At first sight it looks as if the gaps in the documentation represent the crucial years, with the homicide rate rising sharply during the first period for which the data are missing and declining again during the second. The figure calculated from the reports over the years 1667-79 is especially surprising. Not only would it imply a very steep decline from the 1590s, it is also decidedly lower than the level prevailing a few decades later. It is the only figure contradicting the hypothesis of a steady decrease over the centuries. However, there is good reason to assume that precisely in this period cases were underreported. An urban ordinance dated 3 June 1692 explicitly said that unofficial inspections unknown to the court had been common and it prohibited any further continuation of this practice.(37) If the ordinance was rigidly enforced, this would explain the subsequent rise in registered body inspections, although the number of the alleged unofficial ones in the 1660s and 1670s would have to have been enormously high in order to make the total number of inspections approach the figure prevailing from 1693 onward. Whatever was the case, it is improbable that the lower rates prevailing from 1752 onward were biased by underreporting. During the eighteenth century bureaucratization definitely progressed in the city, also and especially in the judicial realm, which makes it unlikely that the practice of unofficial body inspections was resumed in this period. The fact that after 1750 the number of default cases against fugitive homicide suspects decreased, is in line with the declining trend in table 1.(38)

Table 1

Homicide (on non-infants) in Amsterdam body inspection reports, 1667-1816

                     Certain             Possible
                     homicide            homicide
                   (ann. aver.)         (ann. aver.)

                        per                  per         Total per
Period      abs.   100,000 pop.   abs.   100,000 pop.   100,000 pop.

1667-1679    5.5        2.9       1.2        0.6             3.5
1693-1709   18.0        9.0       2.4        1.2            10.2
1710-1726   17.1        8.4       0.7        0.3             8.7
1752-1767    3.3        1.6       1.9        0.9             2.5
1768-1783    3.7        1.8       4.4        2.1             3.8
1784-1799    3.1        1.5       2.4        1.1             2.6
1800-1816    2.3        1.0       0.9        0.4             1.5

The actual homicide rate necessarily stood somewhere between the levels indicated in the first and third column. The ratio of possible to certain cases is very modest up to 1726, which means that the margin of doubt is modest as well. The actual level in the 1660s and 1670s can be put at c. 3.25; it represents a minimum rate, on the premise that there was underreporting. The average level of actual homicide during the years 1693-1726 was c. 9. From 1752 onward the ratio of possible to certain homicide lay between 0.4 and 0.7, except in the fifth period when it was over one. That period may have witnessed a relatively high number of suicides and accidents. If this supposition is correct, the temporary peak during the years 1768-83 was not as marked as it seems. Thus, the homicide rate for the second half of the eighteenth century may be estimated as standing between 2 and 3. In the early nineteenth century it was somewhere between 1 and 1.5. That was the overall trend. The reports also inform us about the victims' sex and approximate age. Table 2 presents the most striking results.

That women and children were less likely to be stabbed to death than adult men is perhaps not surprising. The figures become especially noteworthy when compared with the frequencies of table 1. The correlations, negative and positive, are nearly perfect: when the homicide rate is high, the proportion of stabbings goes up and the proportion of female and young victims goes down. Information on the perpetrators is lacking in most cases, but women who killed men, especially who stabbed them, must have been a rarity. Thus, the relatively high levels of homicide prevailing in the 1690s and the first quarter of the eighteenth century were due mainly to frequent knife-fights among men.

Table 2

Characteristics of homicide victims (non-infants, certain and possible) in
Amsterdam, 1667-1816

                        %child or
Period       % female   adolescent   % stabbed

1667-1679      21.8        2.3        49.4
1693-1709      13.0        2.0        74.6
1710-1726      13.2        1.3        82.8
1752-1767      27.2        9.6        28.9
1768-1783      35.2        7.8        20.9
1784-1799      38.6       10.2        17.0
1800-1816      33.3       20.0        29.1

Total          20.6        4.7        58.1

(source: body inspection reports)

Can we learn more about the social context of homicidal violence in Amsterdam from the late seventeenth century to the early nineteenth? The body inspection reports sometimes contain information on the circumstances in which the incident took place. In three cases, for example, it was noted that an attacker, after a previous quarrel, had secretly approached his adversary and stabbed him while he was urinating. Those are obvious cases of ritual violence, meant to humiliate the victim. Such information, however, was not given very often. The only way to gather contextual evidence on homicide in Amsterdam more systematically is to turn to the criminal records. My series of trials ending in a public punishment, collected in earlier research, forms the basis. It covers the period 1651-1750 and includes all cases of prosecuted homicide, except those in which the suspect had successfully pleaded self-defense. Because this series ended just before the great decrease in the incidence of killing, it was necessary to gather additional information for subsequent years. References provided by Sjoerd Faber enabled me to trace homicide trials up to 1810. For a diachronic comparison I am distinguishing three periods: 1651-1700 (34 killers making 36 victims); 1701-1750 (62 killers, 66 victims); 1751-1810 (29 killers, 34 victims).(39)

Because we are now looking for trends in the character of homicide, its actual incidence having been established, it is no problem that our set of data is much smaller than that extracted from the inspection reports. What matters, is its representativeness. When we consider the collected cases as a sample of homicide in Amsterdam, the only systematic bias is that the killers were arrested. It is difficult to assess different perpetrators' chances of getting caught in various kinds of situations. At least we have one quantitative measure to compare the two sets of data. The victim's sex is the only variable present in the trial series as well as the inspection reports. In the former series the percentage of female victims is 36.8, which is considerably higher than in the inspection reports. Clearly, people who killed women were more likely to get caught (and less likely to claim self-defense) than those who killed men. However, the difference between the two series is largely caused by an exceptionally high number of female victims in the third period of the trial records. They made up 25% and 22.7% in the first and second period, respectively, a proportion comparable to that found in the inspection reports. The first and second periods, then, can be considered fairly representative. The period 1751-1810, in which no less than twenty-six out of thirty-four victims (76.5%) were women, has to be handled with greater caution.

The number of female killers in the trial series, ten altogether, was too small to allow a meaningful comparison over time.(40) Together they made twelve victims, eleven of whom either were the woman's own (noninfant) child or another adult female. The twelfth victim was the woman's husband. It means that of all spouse and lover killers tried only one was female. The killer-victim relationship could be ascertained in most cases. In view of the smallness of the sample, I grouped the types of relationship in just four categories.(41) The "unclear" category is mostly made up of cases in which the records do not reveal whether or not killer and victim already knew each other before the homicidal incident. As a rule, the victims in question either were acquaintances or strangers, not intimates.

Table 3

Intimates and non-intimates as victims in Amsterdam homicide trials in three
periods

                        1651-1700       1701-1750     1751-1810

Group of victims       abs.     %     abs.      %    abs.     %
Intimates                4    11.1      9     13.6    16    47.1
Acquaintances           13    36.1     32     48.5    10    29.4
Strangers               17    47.2     20     30.3     5    14.7
Unclear                  2     5.6      5      7.6     3     8.8

Total                   36   100       66    100      34   100

With the proportion of acquaintances more or less fluctuating and that of the unclear cases slightly on the rise but remaining well under 10%, our attention is drawn to the two other categories. The approximate ratio of strangers to intimates changed from 4:1 in the first period, 2:1 in the second and 1:3 in the third. In the latter period, if all unclear cases were in fact strangers, this ratio still would be 1:2. It may be objected that I just cast doubt on the representativeness of the trials in the third period. However, I am not prepared to consider the observed trend entirely an artefact. For one thing, the percentage of every subcategory of intimates increased. A trend can be observed also with respect to the killer's age: 27.1, 30.0 and 33.4 years, respectively, in the three periods. This trend is compatible with a shift from fights among strangers to conflicts among intimates. Finally, since the number of homicide victims in the inspection reports was much lower after the middle of the eighteenth century, the trials of the third period represent the most intensive sample. By the early nineteenth century the homicide rate had dropped considerably in comparison with a hundred years earlier. Significantly, this drop was accompanied by a rising proportion of homicides on intimates, although the dimensions of the change may have been more modest than the figure for the period 1751-1810 indicates.

Some other variables are informative. That of the killer's birthplace, for example, shows a fluctuating pattern. Those born in Amsterdam made up 29%, 58% and 38% in the three periods, respectively.(42) While for the first this is about 15% less than in the general population, for the second it is about 10% higher. A number of homicide convicts either had a previous arrest or one or more additional offenses listed in their sentence. Those killers can be considered as the group with a criminal background. They made up 53%, 31% and 24% in the three periods, respectively. If the data on place of birth and criminal background are combined, the contrast between the second half of the seventeenth century and the first half of the eighteenth is especially marked: among the perpetrators of homicide a shift took place from immigrants with a criminal background to residents without such a background. In the second half of the eighteenth century, on the other hand, immigrants again were a little overrepresented among the killers, but very few killers had a criminal background. The latter observation is not surprising, since most victims were intimates.

A characterization of the court cases with the help of the two axes I distinguished is a hazardous enterprise. No homicidal incident simply can be labelled according to one archetype or the other. Let me briefly review the cases, restricting myself to what is relevant for the discussion of trends. Throughout the period 1651-1750 impulsive violence was overwhelmingly dominant. About three fifths of the killings in that period resulted from conflicts arising in a tavern or street without any discernable previous history. There was no planning involved; just a little in some cases, when a quarrel was interrupted and one of the protagonists used this opportunity to go home and get a knife. When two persons fought, a third sometimes intervened, which might make him either the killer or the victim. In other cases the killer had been provoked, because the victim was teasing him. The non-settled population was involved in this kind of violence too. One pickpocket stabbed another to death as they fought over the division of the spoils. Apart from these brawls, about half of the robbery-related homicides of this period had impulsive features. A man had been drinking with another, for example, and suddenly decided to try to take his possessions. Or a woman killed another out of desperation, because she owed her money which she could not pay. During the period 1751-1810, on the other hand, only between a third and a half of the homicides can be called impulsive. That does not mean that the remaining cases involved rational violence. Careful planning remained rare; it was only a feature of some robberies.(43)

Less can be said about the ritual-instrumental axis. From the available documentation it is difficult to determine to what extent the tavern brawls and comparable quarrels also had expressive features. One isolated incident of ritual murder certainly was exceptional for its extremity.(44) Most cases of instrumental violence were homicides related to a property crime. They occurred throughout the period 1651-1810.

As noted above, the contextual evidence related to the two axes of violence on the one hand and the killer-victim relationship on the other constitute separate variables. We may ask, however, to what extent the killing of an intimate person was an eruption of pent-up tensions. Homicides of that type, resulting from deep-seated conflicts, can hardly be called impulsive. Neither can they be termed rational, however, since a sudden eruption seems incompatible with advance planning. On the ritual-instrumental axis they would be closer to the former pole, even though they appear to be more common in modern times. I am calling this type of homicide on an intimate person tension-related. The opposite type, described in several Amsterdam trial records, may be termed anger-related. We encounter it notably among men who killed their wife or concubine. Characteristically, the perpetrator was said to be drunk and annoyed over a specific thing. Sometimes the incident was a beating that got out of hand. Every story of a partner killing is like that up to the 1720s. From then on tension-related cases surface in the court records and they become more frequent in the period 1751-1810. All but a few of the cases in that period belong to one of four groups, which are about equally frequent: tension-related homicides on intimates, anger-related homicides on intimates, impulsive violence toward acquaintances or strangers, killings related to robbery.

It can be concluded that the contextual evidence assembled so far indicates a trend. Straightforwardly impulsive violence predominated until the middle of the eighteenth century, becoming less prominent since then. Notably the incidence of tavern brawls decreased. A shift took place around the same time from the killing of strangers to the killing of intimates. From the 1720s onward more homicides on intimates were tension-related. The observed trend may have continued after 1810. In his study of royal pardons in the Netherlands Van Ruller discusses 122 homicide cases from the period 1814-1870. Most cases, he says, either were robbery-related or involved marital/love problems. He identified 48 female and 43 male victims.(45)

These findings, although they allow no firm conclusions yet, definitely indicate that it is worthwhile to investigate changes in the character of homicidal violence. The evidence on the development of homicide rates, on the other hand, is conclusive. These rates definitely declined in Amsterdam from the middle of the fifteenth century until the beginning of the nineteenth. Up to the 1720s they were far above the English national average. The high level at which Amsterdam began was not exceptional within the Netherlands. This is indicated by a converging figure for Utrecht-City in the first half of the fifteenth century. The homicide rate in that town was about 50 then.(46) It may very well be that the steep decline observed in the Netherlands is more characteristic for the European development than the English curve. It is my claim that the Amsterdam figures are better, because they are based on body inspections. The English evidence, I think, should be evaluated anew and all figures based on criminal prosecutions left from consideration. It would be interesting to conduct a longitudinal study of body inspections in London and see if it would produce a curve similar to the one reconstructed for Amsterdam. Finally, we may ask whether towns witnessed a higher incidence of violence than rural areas in the period 1400-1750 or possibly beyond it. Thus, from a European perspective both the incidence and character of homicide still require further historical elucidation.

Department of History Rotterdam, The Netherlands

ENDNOTES

I am indebted to Sjoerd Faber for helping me locate homicide trials in the period 1751-1810. Desiree Herber and Stefanie Reesink assisted with the collection of data in the Amsterdam archive. Helpful criticism of my ideas came from Herman Franke, Eric Johnson, the participants in the session on long-term trends in crime during the SSHA conference in New Orleans (Nov. 1991) and my colleagues of the research group on the history of mentalities at Erasmus University (March 1992).

1. Natalie Zemon Davis, "The Rites of Violence," Society and Culture in Early Modern France (Stanford, 1975).

2. Natalie Zemon Davis, Fiction in the Archives. Pardon Tales and Their Tellers in 16th-Century France (Stanford, 1987).

3. Robert Muchembled, La violence au village. Sociabilite et comportements populaires en Artois du 15e au 17e siecle (Turnhout, 1989).

4. Ute Frevert, Ehrenmanner. Das Duell in der burgerlichen Gesellschaft (Munich, 1991).

5. Anton Blok, Infamy (Oxford, 1989); see also Anton Blok, The Mafia of a Sicilian Village, 1860-1960. A Study of Violent Peasant Entrepreneurs (New York, 1974).

6. Anton Blok, "Zinloos en zinvol geweld," Amsterdams Sociologisch Tijdschrift 18, 3 (1991): 189-207.

7. Roger Lane, Violent Death in the City. Suicide, Accident and Murder in 19th-Century Philadelphia (Cambridge, 1979); see also Roger Lane, "Urban Homicide in the 19th Century. Some lessons for the 20th" in James A. Inciardi and Charles E. Faupel, eds., History and Crime, Implications for Criminal Justice Policy (London, 1980), pp. 91-109. J. S. Cockburn, "Patterns of Violence in English Society. Homicide in Kent, 1560-1985," Past and Present 130 (1991): 70-106.

8. This definition allows me, when referring to the homicide rate throughout this article, to omit the use of "per 100,000 inhabitants."

9. Summarized in Jean-Cluade Chesnais, Histoire de la violence en Occident de 1800 a nos jours (Paris, 1981); more recent figures in Herman Franke, "Gerveldscriminaliteit in Nederland. Ein historisch-sociologische analyse," Amsterdams Sociologisch Tijdschrift 18, 3 (1991): 13-45.

10. I take as my point of departure the synthesis in Lawrence Stone, "Interpersonal Violence in English Society, 1300-1980," Past and Present 101 (1983): 25-6 (largely based on Ted Robert Gurr, "Historical Trends in Violent Crime. A Critical Review of the Evidence," Crime and Justice, An Annual Review of Research 3 [1981]). Two major studies presenting homicide figures were published since then: John M. Beattie, Crime and the Courts in England, 1660-1800 (Oxford, 1986), esp. 108 on Surrey and Sussex and Cockburn; "Patterns of Violence," on Kent. Whereas the Kent rates around 1600 and the Sussex ones around 1700 were below the national average, the rates for urban Surrey in the late seventeenth century were higher than the national trend even in the Elizabethan period.

11. Eva Osterberg, "Criminality, Social Control and the Early Modern State. Evidence and Interpretations in Scandinavian Historiography." Paper presented at the Fourth IAHCCJ Conference (Stockholm, 1990); see also Eva Osterberg, "Social Arena or Theatre of Power? The Courts, Crime and the Early Modern State in Sweden," in Heikki Pihlajamaki, ed., Theatres of Power, Social Control and Criminality in Historical Perspective (Publications of Matthias Calonius Society 1) s.1. 1991: 8-24.

12. Homicide rates in 14th-century Florence were as high as 152 and 68 in two three-year periods. Marvin Becker, "Changing Patterns of Violence and Justice in 14th- and 15th-century Florence," Comparative Studies in Society and History 18, 3 (1976): 287. The rates calculated for Artois and Cologne in the 15th and 16th centuries are minimum figures, based on an incomplete set of data. In both cases, the minimum homicide rate was 10. Muchembled, "La Violence au Village," pp. 19-21; Gerd Schwerhoff, Koln im Kreuzverhor Kriminalitat, Herrschaft und Gesellschaft in einer fruhneuzeitlichen Stadt (Berlin, 1991), pp. 282-84.

13. Norbert Elias, Uber den Prozess der Zivilisation. Soziogenetische und psychogenetische Untersuchungen, 2 vols. 2nd ed. (Munich, 1969).

14. Carol Zisowitz Stearns and Peter N. Stearns, Anger: The Struggle for Emotional Control in America's History (Chicago, 1986).

15. See also the remarks in Randall Collins, "Three Faces of Cruelty: Towards a Comparative Sociology of Violence," Theory and Society 1 (1974): 415-40.

16. Muchembled, La Violence au Village, pp. 167-83.

17. Davis, Fiction in the Archives, p. 29.

18. Robert Muchembled, "Anthropologie de la violence dans la France moderne, 15e-18e siecle," Revue de Synthese IVe serie, nr. 1 (1987): 31-55.

19. Cf. Frevert, Ehrenmanner.

20. Barbara Hanawalt, "Violent Death in 14th and Early 15th-Century England," Comparative Studies in Society and History 18 (1976): 311.

21. Guido Ruggiero, Violence in Early Renaissance Venice (New Brunswick, NJ, 1980), pp. 171-80; see also Dirk Arend Berents Misdaad in de middeleeuwen Een onderzoek naar de criminaliteit in het laat-middeleeuwese Utrecht. s.1. (1976), pp. 65-6.

22. Cf. the special issue of the Amsterdams Sociologisch Tijdschrift 18, 3 (1991). Most contributors adhere to a dichotomy in one way or the other. Franke, Gerveldscriminalikeik in Nederland distinguishes instrumental from impulsive violence; Nico Wilterdink, "Inleiding" p. 9, adheres to the same distinction, but he equates impulsive with expressive; Blok, "Zinloos en zinvol geweld," p. 194, contrasts violence called instrumental or technical from violence called expressive, ritual, symbolic or communicative. Blok is the only one who emphasizes that these are different aspects (rather than mutually exclusive types) of violence, which can be combined in one and the same act.

23. Cf. P. van den Eshof and E. C. J. Weimar, "Moord en doodslag in Nederland. Nederlandse gegevens in international perspectief," Justitiele Verkeninngen 18, 1 (1991) and the contributions to the special issue of the Amsterdams Sociologisch Tijdschrift, 1991.

24. One exception is Hanawalt, "Violent Death in 14th and Early 15th-Century England," pp. 309-10. Based on a sample of 63 cases, she considers the geographic origin of homicide victims and concludes that most of them were neighbors or acquaintances. Barbara Hanawalt, Crime and Conflict in English Communities, 1300-1348 (Cambridge, 1979) has a whole chapter on killer-victim relationships, but the subject is discussed only with reference to all categories of offenses combined.

25. The one used by Eshof and Weimar, "Moord en doodslag in Nederland," comes closest to the classification proposed below.

26. Compare the discussion in Stone, "Interpersonal Violence in English Society, 1300-1980," p. 27, from which the cited figures were taken. Cockburn, "Patterns of Violence in English Society," pp. 93-6, contests these figures, mainly because he proposes to include infanticide in the category of family homicide. As explained below, I favor keeping infanticide figures entirely apart.

27. Stearns & Stearns, Anger.

28. Beattie, "Crime and the Courts in England, 1660-1880," pp. 108-09.

29. Jean A. G. Jungen, "Doodslagers en hun pakkans in het 16e eeuwse Amsterdam," in Cyrille Fijnaut and Pieter Spierenburg, eds., Scherp Toezicht. Van 'Boeventucht' tot 'Samenleving en Criminaliteit,' (Arnhem, 1990), pp. 84-7.

30. Jan Boomgaard, Misdaad en straf in Amsterdam. Een onderzoek naar de strafrechtspleging van de Amsterdamse schepenbank, 1490-1552 (Amsterdam, 1991), pp. 92-3 (table 4.1; I discarded the data for the years 1426-30, because they did not refer to fiscal years and most killers were unidentified). I am grateful to Jan Boomgaard for making the data available to me before he published his book.

31. Jan de Vries, European urbanization, 1500-1800 (London, 1984), p. 271 counts 14,000 inhabitants in 1500. Paul Bairoch et al., "La Population des villes Europeennes. Banque de donnees et analyse sommaire des resultats, 800-1850 (Geneve, 1988), p. 53 count 3,000 in 1400 and 15,000 in 1500. Thus, Amsterdam s population by the middle of the fifteenth century may be estimated at eight or nine thousand.

32. The annual average of body inspections in which the conclusion was murder/manslaughter was 8.2 in the period 1524-65, Jungen, "Doodslagers en hun pakkans in het 16e eeuwse Amsterdam," pp. 84-5. At that time Amsterdam had nearly 30,000 inhabitants. In an unpublished thesis, Jean A. G. Jungen, God Betert. De Amsterdamse lijkschouwingsrapporten in de jaren 1560, 1570, 1580 en 1590 (Free University Amsterdam, 1982: 19) the body inspections of the years 1560, 1570, 1580 and 1590 were taken as a sample. The result for that period is an annual average of 10 homicides and 1.5 uncertain cases. This period overlaps with the one Jungen dealt with in his later article, in which he incorporated the 1560 data. The population estimates for 1600 given by De Vries (65,000) and Bairoch et al. (54,000) diverge widely. Since Amsterdam had over 100,000 inhabitants according to the census taken in 1622, De Vries's figure seems closer to reality. His estimate for 1550 is 30,000, therefore, I take the population in 1575 to be 47,000.

33. GA Amsterdam, archive nr. 5061, inv. nrs. 640c-640g.

34. They alternately styled themselves de gezworen chirurgijns van den gerechte (the sworn surgeons of the court) or de gequalificeerden tot het schouwen der lijken der nedergeslagenen (the committee authorized to inspect the bodies of the slain), which clearly indicates that their task was to look for signs of a violent death.

35. This is also done in Beattie, "Crime and the Courts in England, 1660-1880." For a discussion of infanticide in Amsterdam, see Sjoerd Faber, "Kindermoord, in het bijzonder in de 18e eeuw te Amsterdam," Bijdragen en Mededelingen betreffende de Geschiedenis der Nederlanden 93 (1978): 224-40.

36. The resulting file of noninfants who certainly or possibly had been killed by another person consists of 1,091 victims, out of a total of 1,451 noninfant corpses inspected. Reliable population estimates for Amsterdam are available. The number of inhabitants in the seven periods, distinguished respectively, can be put at: 190,000; 200,000; 205,000; 210,000; 210,000; 215,000; 220,000. In the tables every average was calculated directly from the total number and rounded off at one decimal, which means that average parts sometimes do not add up to average sums.

37. Ordinance inserted in GA Amsterdam archive 5061: nr. 640G.

38. Information on default cases in Sjoerd Faber, Strafrechtspleging em criminaliteit te Amsterdam, 1680-1811, De nieuwe menslievendheid (Arnhem, 1983), pp. 94-5, 98; Marijke Gijwijt-Hofstra, Wijkplaatsen voor vervolgden. Asielverlening in Culemborg, Vianen, Buren, Leerdam en IJsselstein van de 16e tot eind 18e eeuw (Dieren, 1984), p. 156.

39. Data extracted from GA Amsterdam, archive nr. 5061: nrs. 308-515, 581-598, 605-614 (the records for the years 1706-09 are missing). Just once it was clearly a matter of two people together killing one person. For the rest, differences in numbers between killers and victims are due to cases in which one person killed two or more others. Trials for mere complicity or attempted murder were not included; neither were trials for infanticide. One decade,

1711-20, with 27 prosecutions for murder or manslaughter largely accounts for the greater frequency in the second period.

40. There were 2, 3 and 5 female killers in the three periods, respectively.

41. The intimate victims include spouses, lovers and first-grade family; the acquaintances include kin from the second grade (one case in each period), work mates, fellows in crime and other acquainted persons.

42. Unadjusted for missing cases. Percentage of cases with information on the killer's birthplace missing: 0, 8, 7, respectively.

43. One case of rational violence took place already in 1664, when a hired killer shot a Jew at the request of another Jew. GA Amsterdam, archive 5061: nr. 316, fo. 64 et seq.

44. In 1742 an old man managed to overpower a younger man whom he hated. He tied his victim to the wall, chopped off his head with a chisel and skinned him. GA Amsterdam, archive 5061: nr. 402, fo. 106 et seq.

45. Sibo van Ruller, Genade voor recht. Gratieverlening aan ter dood veroordeelden in Nederland, 1806-1870 (Amsterdam, 1987), pp. 126-29. Many pardons were rejected, which means that this sample of homicide is not biased in the direction of cases which were thought most likely to be pardoned.

46. Berents, "Misdaad in de middeleeuwen," p. 66; Dirk Arend Berents, Her werk van de vos. Samenleving en criminaliteit in de late middeleeuwen (Zutphen, 1985), pp. 136, 210 (note 103). The annual average of homicide cases was 6.4. Presumably, this figure includes default cases. In his calculation of the homicide rate, Berents takes the town's population to be 10,000. Based on recent demographic literature, I corrected this to 12,000. This results in a homicide rate of 53.3 Berents also tells us that his figures concern the number of killers and that the number of victims was slightly lower. Therefore, I lowered the rate to 50