The Last Visitation of the Plague in Sweden: the Case of Bräkne-Hoby, Blekinge in 1710-11

The last visitation of the plague in Sweden: the case of Bräkne-Hoby, Blekinge in 1710-11

By ROGER SCHOFIELD

SUMMARY: In an earlier study of the plague in Colyton, Devon the household distribution of deaths was studied to see whether this provided a method of the identifying the causative disease. In this article, a known epidemic of plague in the Swedish parish of Bräkne-Hoby was studied as a means of testing out the generality of the household distribution of deaths. It was discovered that, in this case, the very heavy mortality was due to two radically different means of spreading the disease, initially the classic bubonic one through the rat flea, and latterly, and somewhat surprisingly, the pneumonic one, through the infection of the inhabitants by their own friends and neighbours.

One of the difficulties in studying epidemics in the past is that the disease behind the epidemic is often not identified as such in the burial register.1 Although there are certain pointers to the nature of disease, such as its seasonality, these are in general far from conclusive. Accordingly in a study that I made of the epidemic in Colyton, Devon in 1645-6, I decided to investigate the patterns of deaths within the household, just in case this gave additional information on the nature of the infection.2 One method of achieving this is through a ‘family reconstitution’ of the parish register, but for a parish of the size of Bräkne-Hoby, about 2900 inhabitants in 1749; this is a difficult and extremely laborious undertaking.3 In later epidemics, in this part of Sweden, it is possible to avoid this, by also using two alternative sources, a ‘husförhörslängd’, which is a household examination book, and the tables known as the ‘statistiska tabeller’, which contain an aggregate analysis of the parish registers. I therefore decided to take an epidemic in Sweden, in 1710-1, that was spread over several parishes, and to the church officials, at least, seemed to be clearly one of plague. The parish of Bräkne-Hoby is located in Blekinge, a coastal county in the extreme southeast of Sweden. The parish is about 8 kilometers wide, and runs up from the coast in the south, to the border with Kronobergs Län in Småland, some 33 kilometers inland.4 The parish sits astride the main coastal road, which runs from a garrison town, Karlshamn, 22 kilometers to the west, to the major naval port of Karlskrona some 33 kilometers to the east. The population of Bräkne-Hoby was 2837 in 1754, and it was scattered over the whole area in settlements ranging from isolated farmsteads to fair sized villages.5 The economy of the parish was based on agriculture and fishing, and its settlements were located in shallow fertile valleys and clearings, separated by extensive tracts of rock-strewn and wooded terrain. Most of the larger villages were in the south of the parish where the arable land was to be found. Conversely, the far north of the parish was hilly and covered by woods, and the settlements were made up of hamlets and isolated farmsteads, with quite small

1 I am very grateful to Professor Dr Anna Christina Ulfsparre, formerly Landsarkivarie in Lund, who both encouraged me to analyse the parish register of Bräkne-Hoby and who secured copies of the mantalslängder. I am especially grateful to Dr Bodil Persson for her knowledge and unrivalled experience of plague mortality in eighteenth century Sweden. She has been responsible for many of the changes to the original text. 2 See, Schofield, R., ‘An anatomy of an epidemic: Colyton, November 1645 to November 1646’, in P. Slack (ed.), The plague reconsidered, pp. 95-126, 1977, p.121. 3 Leslie Bradley carried out the reconstitution of the parish register, and most of the tabulations, on which this article is based, between 1978 and 1980. Unfortunately he died at the end of 2003, but had kept in touch, and commented extensively on the later versions of this article. Historically the parish, therefore, included the area of the modern, and separate, parish of Öljehult. 4 Bräkne-Hoby, Landsarkivet i Lund, Statistiska Tabeller, GIII:1. The demography of the province to the south and west of Blekinge, the called Skåne, has recently been studied; see. Persson, B. E.B, Pestens gåta Farsoter i det tidiga 1700-talets Skåne. 5 Bräkne-Hoby, Landsarkivet i Lund, Statistiska Tabeller, GIII: 1. 2

parcels of land. As the burial register gave the hamlet, or farmstead, of the residence of the population, it was relatively easy to study the geographical distribution of the epidemic mortality.

Figure 1.

The essential background to the plague epidemic was war. For it was through war that the military activities across the Baltic, and more particularly in Estonia, that brought the plague to Sweden. The particular form of permanent military billeting, that was current in Sweden at the time, ensured that the effects of movements of troops and sailors would reach deeply into a coastal parish

like Bräkne-Hoby. The parish had a considerable number of sailors living permanently in cottages scattered all over its area. When they were not campaigning, the sailors eked out a living either by smallholding, or by labouring. These activities were supplemented by payments in cash, or in kind, raised by way of taxation from other members of the community. In times of war the sailors naturally served in the fleet, whose main southern base was at Karlskrona, returning to the parish at the end of the campaigning season. In 1709 the plague had spread from Turkey northwards through Eastern Europe, reaching the southern coast of the Baltic, where the Swedes were engaged in defending their overseas provinces.6 Refugees from Estonia brought the plague to Stockholm in August 1710, where it inflicted its heaviest mortality at the end of October.7 From Stockholm it was spread by ship to several places, including Kalmar, a naval port some 100 kilometers to the Northeast of Bräkne- Hoby. It reached Kalmar, in September, and then spread into the surrounding countryside.8 Meanwhile, on 20 September a hospital ship from the Nyenska fleet, which had seen active service in Estonia, arrived back at its base in Karlskrona. The ship’s arrival sparked off a severe epidemic in the crowded barracks, and in the surrounding town.9 The naval Chaplain, stationed in Karlskrona, wrote in the burial register that the pestilence began there in the month of November in 1710, and he provided the most secure evidence on the timing of this outbreak.10 In a later entry in the burial register he wrote that the plague continued for the whole of 1711 and only slackened at the end of that year when more than 6,000 had been buried.11 At Karlshamn, a town some 22 kilometers to the west of Bräkne-Hoby, the sickness broke out in December 1710 and in January 1711 six soldiers of the garrison died of ‘plague’.12 Hult describes the ensuing panic in the town with desperate measures being taken by the inhabitants to bury their dead in the churchyard, and not in the official plague burial ground, which was located outside the town. On 23 May the Lord Lieutenant of the county complained that an epidemic disease was reported from the deaneries of Medelstad, Bräkne and Lister, which surrounded Bräkne-Hoby. On 15 June three-quarters of the town of Karlshamn had been infected, and over 400 had died.13 The situation in the countryside was becoming desperate; the disease was reported from every deanery and every parish. Many homesteads were deserted, and the quartering of large numbers of military personnel was causing serious difficulties.14 As the highroad between Karlshamn and Karlskrona crossed the parish of Bräkne-Hoby, it was inevitable that the infection would reach the parish. But was it really plague that caused the massive increase in mortality? As the burial records do not state the cause of death, we have to rely, first on the contemporary accounts we have cited, and second on checking whether the characteristics of the epidemic agree with the accepted characteristics of plague. We shall begin with a general discussion of mortality by plague.

6 Hult, O.T., ‘Pesten i Sverige 1710’, Hygienisk Tidskrift, VIII, Stockholm, 1916, p.95. 7 Hult, ‘Pesten i Sverige’, p.101. 8 Hult, ‘Pesten i Sverige’, pp.143-4. 9 Hult, ‘Pesten i Sverige’, p.154. 10 Hult, ‘Pesten i Sverige’, p152. 11 Hult, ‘Pesten i Sverige’, p.155. 12 Ibid. Wieselgren, in a typical migratory legend, writes that the plague came to Karlshamn in a ship that was sailing adrift with no surviving crew. A man, who went aboard and seized a coat, became the plague’s first victim. Wieselgren, P. Ny Smålands beskrifning, vol. 2, 1844-5, p. 771. 13 Hult, ‘Pesten i Sverige’, p.156. 14 Hult, ‘Pesten i Sverige’, p.156.

The plague bacillus has a marked toxicity in man. It produces cell death and provokes generalised inflammatory reactions, especially in the nervous tissues.15 The accumulations of the bacillus obstruct the dilated capillaries, leading to bleeding. And in the capsules of the enlarged lymph nodes they infiltrate the nerve fibres with a blood-tinged oedema. It seems that it is this distension and infiltration by the bacilli, which are responsible for the terrible pain of plague. Clinically, two forms of plague are distinguished; differing essentially in the way the bacterium enters the patient. This can be either through the skin (bubonic plague), or through the lungs (pneumonic plague). In the bubonic form of plague the incubation period after a penetration of the skin is from one to six days. Normally, the onset is very acute, with the patient’s temperature rising to between 39 and 40 oC. The point of infection is almost always through a fleabite on a limb. The survival rate varies from epidemic to epidemic but normally it lies between 20 and 40 per cent of those infected.16 From the second, or third, day there is an enlargement of the lymph nodes, usually in the groin but sometimes in the armpit or the neck, draining the site of infection. The lymph nodes become hard, and very painful, and tend to suppurate. These are the buboes. At the end of eight or ten days, the disease may suddenly terminate and convalescence begins. Otherwise, a stage of acute septicaemia sets in and the patient’s temperature rises to between 40 and 42o C, and at this point most deaths occur. Pneumonic plague passes through the lungs, and the incubation period is from only one to three days. The onset begins with a body temperature of only 38oC, but otherwise it is very severe, with the pulse rate reaching 120 per minute. In every case death occurs within two or three days. Pneumonic plague is transmitted directly from person to person by droplets of saliva, which are emitted while speaking or coughing. For plague to be spread in this way, a plague abscess in the lung of an infected person has to rupture, infecting the droplets. In cold and damp environments these droplets remain infective for a long time, and so can be inhaled so long as they are in suspension in the atmosphere. The plague bacillus then enters the body through the mucous membranes in the nose, mouth or lungs. The same process can occur when objects soiled by spittle are handled, and the fingers later touch the mucous membranes of the mouth, eyes and nose, for dried spittle remains infective for a long time in a cold country. Thus, although pneumonic plague is a very unusual form of plague, its high degree of inter-human infectiveness and 100 per cent death rate, make it an exceptionally dangerous one.17 In 1710-11 the disease behind the mortality was accepted by the Swedish vicars as being caused by plague; but were they mistaken in their attribution? This has now become an issue, for while the discussion of bubonic plague is no longer concerned with the question of which flea it was that caused the infection, since the supporters of the human flea appear to have given up the battle.18 Moreover, wholly new diseases have been pressed into service, such as whatever disease

15 See Pollitzer, Plague, p.440-53. The discussion, presented here, is mainly taken from Biraben, ‘Current medical views’. 16 Biraben, ‘Current medical views’, p. 27. 17 Biraben, ‘Current medical views’, p. 30. Pollitzer, Plague, pp. 451-3. The leading modern text on pneumonic plague is still Wu, A treatise on pneumonic plague, Geneva, 1936. 18 See Audoin-Rouzeau, F, and J-D Vigne, Revue de Paléobiologie, 13, 1, pp. 125-145, and Audoin-Rouzeau, F; Bull. Soc. Path. Exo, 92; 5, Pt. 2, pp, 422-26.

may have lain behind ‘hemorrhagic plague’.19 But in the circumstances, we can at least emphasise that at the very high scale of mortality found in Brkne-Hoby, the deaths, however caused, must have been so extensive as to infect almost everyone.

The Bräkne-Hoby Parish Register gives the impression of having been well kept. In the case of baptism entries, for example, the register gives the full name of the father, together with the settlement of residence, the first name of the child and a list of ‘sponsors’. Military occupations are always specified; indeed sailors (båtsmän) were resident as a permanent reminder of the danger of Danish raids on the coast. The burial entries are similarly detailed. Those for children specified the first name of the child, the full name of the father, or stepfather, and occupation, as above. From January 1697 to late January 1711 the date and time of day of death are added in all cases and, for children, the full name of the mother too. These details are omitted from 29 January 1711, when the epidemic was in full flood. The lack of a date of death, after the first two months of the epidemic, is unfortunate because the interval between death and the burial recorded in the registers could be quite long. Over the period 1697-1710, when birth dates are given, the median interval was 6 days, and the central half of the distribution of intervals fell between 5 and 9 days after death. These rather long intervals reflect the scattered patterns of settlement in the parish with many hamlets, especially in the north, lying up to 25 kilometres from the church and burial ground, which were located in the main settlement of Bräkne-Hoby, near the southern end of the parish. Indeed, some of the hamlets and farmsteads in the northern part of the parish were much nearer to the neighbouring parish church of Backaryd, and the Backaryd register records the burials of a number of persons resident in these northern settlements, sometimes in addition to registration in the Bräkne-Hoby register. Unfortunately the Backaryd register is much less informative than that of Bräkne-Hoby. The settlement is always recorded, but personal information varies between full names and such brief notations such as ‘4 sons of Nils’. Clear identification was therefore sometimes difficult or even impossible, but of the 173 burials recorded in the Backaryd register as from Bräkne-Hoby settlements 38 per cent could definitely be identified as duplicates of burials recorded in the Bräkne-Hoby register. A curious feature is that even amongst this last group the dates recorded in the two registers rarely agreed and the discrepancy was sometimes considerable. The Bräkne-Hoby register is, therefore, used as the main source, supplemented where appropriate by entries from the Backaryd register. Finally, on the 14 April 1711 the parish opened up a new burial ground, in which were buried 139 sailors and their dependants, up to the 13th July 1711. The old burial ground, however, continued to be used by the non-military inhabitants.20 The poll tax register (‘mantalslängd’) was normally drawn up annually in February or in March.21 It recorded for each taxation unit, that is a settlement or part of a settlement, the full name of each head of a household, the presence of a wife, the first names of co-resident sons and

19 Susan Scott and Christopher Duncan, describe them ‘as any of a diverse group of viral diseases that characterised by sudden onset, fever, aching, bleeding from internal organs, petecchiae (spots resulting from the effusion of blood under the skin) and shock’, Biology of plagues: evidence from historical populations, p. 385. 20 Bräkne-Hoby, Parish Register, pp. 155, 173. 21The Mantalslaengd is in Riksarkivet, Länsräkenskaper, Blekinge Län (1710, 1711, 1714), pp.204-27.

daughters over the age of 15, and of servants. Sailors were personally exempt, but their dependants were listed, hence a common form of entry is ‘sailor X’s wife’. The record then tabulates the numbers in the unit falling into a number of categories, such as the number of adult farmers and craftsmen and their wives, the number of adult sons and daughters, servants, and old people over the age of 63. Children under age 16 were not liable to taxation and so they were neither listed, nor included in the tabulation.22 Their existence, therefore, had to be discovered by linking burial entries in the parish register to the ‘mantalslängd’. In view of the very limited range of names in common use; the possibility of confusion between individuals is obvious, especially in the larger settlements where there could well be two or three individuals with the same name. Fortunately, however, the frequent addition of the wife’s name to a record considerably reduced the difficulties of identification except when whole families moved between settlements. A further difficulty with the poll tax listing lay in the fact that the taxation unit was often not necessarily coterminous with a household or contained instead several residential units. For example a farmstead might comprise a divided farmhouse, or several cottages, either grouped around a yard or more widely scattered. The poll tax listing, therefore, provides only a rather weak indication of co-residence, and the following assumptions were made in order to arrive at a finer ‘household’ division. Firstly, each name was taken to indicate a separate household, except where a family relationship was indicated, as in the case of a son or daughter, and servants were taken to belong to the household immediately preceding their names. Secondly, as sailors were not subject to tax, they were not recorded in the register. Consequently there was no way of knowing whether they were present in the parish and this uncertainty had to be taken into account in the household analysis. Finally, several soldiers and their wives and children were recorded in the parish register. In view of this uncertainty about their presence in the community, they have been excluded from consideration in the family analysis, though they have been included in aggregate totals of deaths before, and during, the epidemic. Table 1 shows that during the period 1680-1708 there were steady, though small, excesses of baptisms over burials. The excesses of burials over baptisms in 1691, 1709 and 1710, though substantial, are no greater than are likely with any parish of this size when attacked by the occasional epidemic. Almost always, in the absence of migration, the retained excess would have led to an increase in the population, but there is little sign of any substantial increase in the annual number of baptisms until 1700, or in the number of burials. The massive scale of the mortality in the epidemic of 1710-11 is clear from the table, which also shows how the earlier pattern of a surplus of baptisms over burials soon re-established itself afterwards. This was probably brought about by the immigration of young adults with full reproductive potential after an epidemic.

III

In order to set the mortality crisis of 1710-11 in its demographic context, the frequencies of events recorded in the parish register between 1677 and 1730 were inspected.

22 Old and destitute persons, even when included in the nominal list, are not included in the tabulation, though there is usually a note to indicate their omission.

Table 1. Bräkne-Hoby, annual totals of baptisms and burials and natural increases. Baptism Burial Natural Baptism Burial Natural increase increase

1677 60 80 -20 1704 116 59 57 1678 63 149 -86 1705 105 88 17 1679 61 77 -16 1706 100 88 12 1680 81 37 44 1707 130 49 81 1681 78 32 46 1708 102 45 57 1682 74 51 23 1709 79 100 -21 1683 89 39 50 1710 88 100 -12 1684 84 40 44 1711 83 1346 -1263 1685 72 36 36 1712 130 39 91 1686 63 36 27 1713 95 46 49 1687 99 39 60 1714 115 45 70 1688 69 35 34 1715 89 38 51 1689 105 50 55 1716 92 36 56 1690 85 61 24 1717 106 83 23 1691 61 125 -64 1718 99 39 60 1692 80 63 17 1719 85 48 37 1693 74 57 17 1720 97 39 58 1694 86 50 36 1721 60 32 28 1695 83 68 15 1722 95 45 50 1696 103 55 48 1723 82 59 23 1697 69 53 16 1724 83 90 -7 1698 77 70 7 1725 83 28 55 1699 99 40 59 1726 81 49 32 1700 68 61 7 1727 58 64 -6 1701 91 92 -1 1728 66 46 20 1702 125 42 83 1729 61 43 18 1703 93 62 31 1730 80 39 41

Source: Fly leaf of the Bräkne-Hoby Parish Register.

The registers also show that in the six years preceding the epidemic Bräkne-Hoby shared several common demographic characteristics with English parishes at the same date. For example, infant mortality in Bräkne-Hoby was rather high by English standards: it stood at 241 per thousand, as against 195 per thousand live births.23 Furthermore, the proportion of marriages with one partner coming from outside the parish, and it was overwhelmingly the groom, was at the comparatively low figure of 26 per cent. This is not to say that the population was immobile, for there were many ‘life cycle’ servants who moved from farm to farm, as in England. In fact the extensive area covered by the parish and the separation of its population into settlements, often divided by woodland, resulted in a considerable mobility at marriage, but it was mobility from within, and not across, parish boundaries.

23 For England the infant mortality figure is taken for the 25-year period 1700-24, illegitimate births being scored at double the mortality rate for legitimate births. Wrigley, Davies, Oeppen and Schofield, English population history from family reconstitution (1997), p.219, table 6.2.

If we now assume that the infant mortality rate can be used as an indicator of the general level of mortality in the period before the crisis, then we can use the observed distribution of burials by age, to calculate not only the prevailing level of fertility, but also the rate of population growth and the probable population size. All these calculations assume the population to have been subject to constant rates of fertility and mortality. This may well not have been the case, but even a very rough indication of population size and of the level of the vital rates, would be useful information. If we now compare the percentage distribution of burials by age in Bräkne-Hoby in the 15 years immediately before the epidemic (1695-1710), with that of a stable population experiencing the Princeton ‘North’ pattern of mortality at an almost similar level, i.e. with an infant mortality rate of 242 per thousand. We find that the best fit between the figures actually observed and the model table is for a population with an expectation of life at birth of 28.6 years (for males and females combined), and which was growing at about 1 per cent per annum.24 Thus although the fit is far from perfect, it is better than can be obtained with other levels of mortality, or with other rates of population growth. It therefore seems to be the case that at least part of the natural increase was being retained in the parish, which in consequence was growing fairly rapidly in size. Moreover, in such a population the crude annual death rate is about 35 per thousand and, because of its stable rate of growth, the crude birth rate is rather higher at 46 per thousand. If Bräkne-Hoby had also experienced these rates in the 15 years before the crisis, the average annual totals of 93 live births and 70 burials would have been produced by an estimated population of about 2000 people.

We can also get a quick appreciation of the scale of the epidemic mortality, if we plot the distribution of burials for every half-month between 1 December 1710 and 31 August 1711. This is done in figure 2, where it can be seen that the number of burials falls off sharply in August 1711. Accordingly, the crisis period has been defined as being the eight months between 1 December 1710 and 31 July 1711.

24 The Princeton’ North’ pattern of mortality has relatively high child rates; it was taken from various Scandinavian populations in the late nineteenth century, and the high child rates also seem true of Bräkne-Hoby around 1700.

During this period, 1404 burials were recorded in the Bräkne-Hoby register. To this must be added 106 burials from the Backaryd register of Bräkne-Hoby residents that were not recorded in the Bräkne-Hoby register, making a total of 1510 burials in the crisis period.25 Assuming an initial population of 2000, as was estimated above, it would appear that massive three-quarters of the original population died during the crisis. Even if some allowance is made for immigration, the death rate is still very high, probably even higher than the upper ranges of village mortality in England during the Black Death and very much higher than the later plague death rates in the sixteenth and seventeenth centuries. For example, in Colyton in 1645-6, the death rate was 20 per cent, and in Eyam in 1665-6 it reached 50 per cent.26 Indeed of all the diseases it is only plague that has a severe enough case-fatality rate, somewhere between 60 and 85 per cent that can inflict mortality on this scale. If we were to take a generous figure, that is if we were to assume that the case-fatality rate was 80 per cent in this case, then no less than 94 per cent of the population of Bräkne-Hoby was probably infected.27 If we now consider the evidence of the registers kept by the other parishes that were neighbouring to Bräkne-Hoby; we shall be able to see whether the epidemic had a similar monthly profile, along the Blekinge coastline.

25 In these figures soldiers and their families have been included, but the settlement at Skörje has been excluded because it was split between Bräkne-Hoby and Backaryd parishes. 26 Schofield, ‘An anatomy of an epidemic’, p.98; Bradley, ‘The most famous plague’, p.77. 27 The ‘case-fatality’ figures for the gross plague rates in Provence in 1720-22, (termed ‘lethality’) work out at an average of 78 per cent. Benedictow, ‘Morbidity’, p.410. In general, see Morris, ‘Plague in Britain’, p.38.

Table 2. Monthly profile of the epidemic December 1710-August 1711. Hallaryd Åryd Backaryd Bräkne-Hoby No. % No. % No. % No. %

December 2 0.4 0 0 24 3.4 19 1.3 January 13 2.6 70 15.4 66 9.3 121 8 February 29 5.8 81 17.8 93 13.1 199 13.1 March 56 11.2 95 20.9 55 7.8 245 16.1 April 93 18.6 103 22.6 79 11.1 268 17.6 May 208 41.7 52 11.4 164 23.1 399 26.3 June 83 16.6 39 8.6 120 16.9 189 12.4 July 6 1.2 9 2 98 13.8 71 4.7 August 9 1.8 6 1.3 11 1.6 9 0.6

Total 499 100 455 100 710 100 1520 100

Source: Parish Registers

The monthly profile of deaths in the epidemic is taken as running from 1 December 1710 to 31 August 1711. Table 3 gives the monthly figures of deaths, and the percentage of all deaths that they comprise for each of the parishes.

Figure 3, shows the shape of the monthly profile proportional picture for Bräkne-Hoby which is based on the largest number of burials, which are given on the right of the graph. They build up to reach a maximum in late May, and then peter quickly out by the end of August 1711. Backaryd is actually located to the north and east of Bräkne-Hoby, but it is shown on the left of that parish, so as not to hide any of Bräkne-Hoby’s monthly columns. It would do so, because in this parish the epidemic lingered longer in June and July, just as it did in the north of Bräkne-Hoby close to the

borders with Backaryd. All the surrounding parishes with records, therefore, had an epidemic at the same time as in Bräkne-Hoby.28 We have not so far considered the inland side of Bräkne-Hoby, lying toward the north of the parish. This takes us into the county of Kronoberg in Småland, where infantry regiments were stationed backing up the garrisons in Karlshamn and Karlskrona.29 The county seems to have been clear of the epidemic until the end of February 1711, when the disease came from Kalmar and Blekinge to Kronoberg, rather than vice-versa.30 At any rate, Hult includes in his account of the epidemic a table for the deanery of ‘Konga och Uppvidinge’, which gives for each parish the number of infected houses, and the number of deaths, as at 25 August 1711. This Table shows that all the parishes, which lay along the border with Blekinge to the south, had extremely high death rates.31 Tingsås, the parish which bordered on Bräkne-Hoby, had its first deaths of the epidemic in January 1711 and by the end of August had registered 296 of them, a probable epidemic death rate of 45 per cent.32

The parish seems to have been open to the epidemic largely because of the traffic along the coastal road, but also because of the return of war personnel in the form of the off-duty sailors, going home to their cottages. We can now consider the internal geography of the epidemic, by looking at how the burials were distributed between the various settlements in the parish. It appears that the epidemic arose independently in two foci: a smaller one around Grårör, in the extreme north west of the parish and a major one at Torp, which is located towards the southern end of the parish. Torp lay some 1.5 kilometres from the main road linking the ports of Karlshamn and Karlskrona (see Figure 1). Two settlements, Rödby and Stora Silpinge, both near the road linking Torp and Backaryd, were soon infected. By mid-January only six of the 65 settlements in the parish produced more than two burials, but by the end of January this was true of thirteen of the settlements, and all but two of them were on, or near to, main roads. By the end of February the epidemic was widespread, though there were still 23 settlements unaffected, and 24 more in which there had been no more than two burials. The peak came rapidly; the epidemic lingering on most severely in the northern part in the second half of May, by which time only 16 of the 65 settlements had not been seriously infected. The decline was then quite rapid in all the settlements of the parish.33

28 These registers are in Landsarkivet i Lund: Backaryd C:1; Hällaryd C:2; Åryd C1:1. 29 Hult, ‘Pesten i Sverige’, pp. 142-3. 30 This was the conclusion of Hult, ‘Pesten i Sverige’, pp.142. However, in Ysane, a coastal parish in Blekinge some 30 kilometers to the west of Brảkne-Hoby, the earliest plague burials were considerably earlier. On 23rd and 24th October of a båtsman and four soldiers from the Kronoberg’s Infantry Regiment, were buried while on their way home from Karlskrona after a period of service on board the naval fleet. Persson, Pestens gåta, p.290. 31 The parishes were (from west to east) Tingsås, Södra Sandsjö, Älmeboda and Långasjö. 32 Hult, ‘Pesten i Sverige’, 143. The calculation is based on the average of the pre-crisis burial rate, and an assumed death rate of 35 per 1000. The fact that the surviving parish register has many fewer burials recorded is very puzzling. The table XIII given in Hult’s article is stated to be a copy of a report from a local ‘befallningsman’, or a local judicial administrator, in the Konga and Uppvidinge district. Thus this table seems to be more complete than the local parish registers, such as the one for Tingsås. 33. For modern India, where the plague had a similar geographic pattern to that in Brkne-Hoby , see the Plague Commission, ‘On the spread of epidemic plague’, J. of Hygiene, (1910), pp. 349-415.

Table 3. Age distribution of burials (12-1710 to 22-1-1711) Age group Number %

0-1 month 3 3.6 1 mth-1 year 3 3.6 1-4 years 7 8.4 5-9 years 10 12 10-15 years 8 9.6 16-19 years 7 8.4 20-29 years 12 14.5 30-39 years 7 8.4 40-49 years 6 7.3 50-59 years 10 12 60-69 years 6 7.3 70-79 years 3 3.6 80+ years 1 1.2

Total 83 99.9

Source: Bräkne-Hoby Parish Register.

From 1 December 1710 to 22 January 1711 the parish register gives the ages of everyone that was buried, and this information is analysed in column 1 of Table 3. Thereafter, the ages of the deceased are known only when the corresponding baptism can be traced. Since the registers prior to 1695 were not examined, this means that the ages of persons of over 15 completed years cannot be traced from a baptism, and our analysis, therefore, cannot distinguish between the age groups of over 15 years. In the circumstances, it can reasonably be assumed that the following groups of people had reached the age of 16: married persons, persons described as servants, and all persons named in the poll tax register, at least if they are not described in the register as ‘son of’ or ‘daughter of’.

Table 4. Age structure of burials in Bräkne-Hoby1 during pre-crisis and crisis periods, compared with Colyton2 and Eyam3 Age group Pre-crisis period Crisis period (years) (%) Bräkne-Hoby Colyton Eyam Bräkne-Hoby Colyton Eyam 0 32.3 15.5 22.6 3.6 7.9 8.8 1-4 17.4 16.6 14.5 10.6 11.1 10.7 5-9 4.9 5.9 3.1 14.4 11.7 12 10-15 4 4.14 3.04 13.7 10.14 10.24 16+ 41.4 58.04 56.84 57.7 59.24 58.44 1. Bräkne-Hoby: Pre-crisis: 1695-1710 (except 1701, 1705-6 and 1709-10); Crisis: 23 Jan 1711-July 1711. 2. Colyton: Pre-crisis: 1633-42; Crisis: Nov 1645-Nov 1646. 3. Eyam: Pre-crisis: 1651-64; Crisis: Sep 1665-Oct 1666. 4. Approximated from the 10-19 age group. Sources: Bräkne-Hoby parish register; Colyton: Schofield, ‘An anatomy of an epidemic’, pp.98-100; Eyam: Bradley, ‘Mostfamous of all English plagues’, p. 70.

With these assumptions it is possible to work out the age distribution of burials during the epidemic from the 23 January until the end of July 1711.34 Again, Table 4 compares the age distributions of burials in Bräkne-Hoby with those for the English parishes of Colyton and Eyam in 1645-6 and 1665-6, both for the pre-crisis and the epidemic periods. The calculation of the pre-crisis distribution for Bräkne-Hoby is based on the aggregation of burials for the eight years in the period 1695-1710, but omitting the years 1701, 1705-6, and 1709-10 in which there were minor burial crises.A further two points may be noted. Firstly, that in the pre-crisis period the proportion of all burials that were infant burials in Bräkne-Hoby is very much higher than those in the two English parishes. Secondly, that in all three parishes the epidemic produced a considerable shift in the percentage of burials, with a decline for the infants and for all the younger age groups, especially in the age range of between 5 and 15 years. The burial distributions just noted ignore two considerations. Firstly they take no account of the relative sizes of the different age groups and secondly, as already pointed out; the deaths are not all due to the epidemic disease. A more refined calculation will, therefore, determine the excess of burials in the crisis period over the normal pre-crisis number for each age group, and will relate this to the size of the group.35 For each age group the annual pre-crisis age grouping of deaths were subtracted from the annual numbers of deaths during the crisis period, in order to estimate the number of excess deaths due to the epidemic. In order to bring out the unevenness in the age-incidence of the excess mortality, the proportional age structure of the population was estimated. The number of expected excess deaths in each age group was then calculated, on the assumption that the epidemic mortality had struck with a fixed risk of dying at all ages.36 The age structure that was used was that

34 All soldiers and their families have been excluded from this analysis. 35 Schofield, ‘An anatomy of an epidemic’, pp. 109-112. 36 This is because model life tables, under theoretically ‘stable’ conditions of fertility and mortality, provide information on the proportional age structure of the population. Moseng in his study of plague mortality of Allerum, in Skåne, considers this ‘an unfruitful’ use of the equal-incidence hypothesis. Instead he re-calculates the number of expected deaths due to the epidemic by making them proportional not to the proportion in the age structure, but to the proportion of deaths produced by the specific mortality schedule. So instead of testing the number of deaths due to the epidemic against the expected number based on an ‘equal-incidence’ hypothesis, he tests them against the normal course of dying by age that underlies the mortality schedule. By allocating the expected excess epidemic

for mortality of the Princeton ‘North’ pattern with an expectation of life of 30 years at birth (27.1 years for males), and a 1 per cent per annum rate of growth as calculated for Bräkne-Hoby, as outlined above.37 The ratio between the observed number of excess deaths in each age group, and the number expected, on an ‘equal-incidence’ hypothesis, then served as an index of inequality in the age- incidence of mortality during the epidemic.

Table 5. Excess mortality by age: Bräkne-Hoby, 23 January-31 July 1711 Age group Annual Annual Recorded Model Expected Excess crisis pre-crisis excess age excess ratio burials burials (1)-(2) structure1 burials (3) ÷ (5) (1) (2) (3) (4) (5) (6)

0 70.5 19.9 50.6 4 78.5 0.64 1-4 201 6.4 194.6 12 235.5 0.83 5-9 273 1.5 271.5 12 235.5 1.15 10-15 264 1.7 262.3 11 215.8 1.22 16+ 1207.5 24.4 1183.1 61 1196.9 0.99

All 2016 53.9 1962.1 100 1962.2 1 1 . Coale and Demeny, Regional model life tables, model N, level 5, e0=28.5 (male and female), population growth 1% per annum Source: Bräkne-Hoby, Parish register.

The resulting ‘excess rates’ are exhibited in Table 5. In this Table column 1 contains the ‘annual’ crisis burials, i.e. the totals for the eight-month crisis period, multiplied by 1.5 to bring them up to an ‘annual’ basis. In Bräkne-Hoby the excess mortality due to the epidemic was milder than average for the infants, and for the young children aged 1-4 years. It fell more severely on the age group 5-15, and as severely as the ‘equal incidence’ hypothesis would predict from age 16 upwards. The epidemic, therefore, was swamped by the normal mortality under age 5, and did most damage to the mortality rates of those aged between 5 and 15 years. 38

deaths proportionally to the m(x) values, ages with high mortality in the table get given massively inflated expected excess death ratios. Therefore, the ‘excess ratios’ of the children and young adults are inflated, and those of the under-5 years and the adults of over 50 years are heavily depressed. It seems to make little sense to expect that the epidemic deaths will also be subject to the normal age-specific pattern of mortality, which is what this alternative method entails. Moseng, ‘Nordens siste pestepidemie’, pp.185-190. 37 Coale and Demeny, Regional model life tables, pp.252, 398. 38 For the spread of an epidemic of bubonic plague in modern India see Plague Research Commission, ‘Reports on plague,’ J of Hygiene, vol. 7, 1907, pp. 724-98. The age and sex-specific incidence rates are given in table XIV, on pp. 763-4. For the spread of epidemic plague see Swellengrebel und Hösen, ‘Von rattenpest ohne menschenpest, Zeitschr. Hygienen u. inf. krankheiten, vol.79, p. 436.

Table 6. ‘Excess ratios’ for Bräkne-Hoby 1710-11, Colyton (1645-6) and Eyam (1665-6) Age groups Bräkne-Hoby Colyton Eyam (Br-H) 0 0.64 2.49 2.36 1-4 0.83 1.16 0.99 5-9 1.15 1.28 1.15 10-9 (10-5) 1.22 1.23 1.1 20+ (16+) 0.99 0.8 0.86 Source: Bräkne-Hoby: Table 5; Eyam, Colyton: Schofield, ‘Anatomy of an epidemic’, pp.113-4. The Colyton figures were recalculated using the age structure for a population with e0=35 years and a growth rate of 0.03%.

For comparative purposes Table 6 shows the experience of two plague epidemics, in Colyton in 1645 and in Eyam in 1665-6. The age groups used in the calculation of the ‘excess ratio’ for Colyton and Eyam are not precisely the same as those used for Bräkne-Hoby, but some comparisons are possible. The major difference between the experience in Bräkne-Hoby, and in Colyton and Eyam, is that in the former the excess ratio was less than unity (0.64) for infant mortality, while in the English parishes, infants were by far the worst affected with excess ratios of 2.49 and 2.36 recorded. In all these epidemics there was also a similar excess ratio for the age groups 5-19, and the ‘equal incidence’ hypothesis explained the mortality of those aged 20 and above. This is explained partly, but by no means wholly, by the fact that infant mortality was higher in Bräkne-Hoby in the pre-crisis period. If, by way of experiment, Bräkne-Hoby is given the same age level of pre-crisis mortality as prevailed in Colyton, the excess ratio rises from 0.64 to 1.25, but it is still far short of the ratio of 2.49 in Colyton. The fact that the infant ‘excess’ rates were so low in Bräkne-Hoby therefore, must still be accounted for.

Table 7. ‘Excess ratios’ for Bräkne-Hoby; December 1710-February 1711 and March-July 1711. Age group Bräkne-Hoby December-February March-July 0 -0.31 0.58 1-4 0.57 0.88 5-9 0.94 1.24 10-5 1.02 1.31 16+ 1.18 0.95

Source: Bräkne-Hoby, Table 5.

A preliminary examination of the age data is attempted in Table 7, by dividing the epidemic into two periods. Here there would appear to be a change in the pattern between the early part of the crisis period (December-February) and the later months (March-July). If we calculate the ‘excess ratio’ separately for these two periods, it appears that crisis mortality in the December-February period fell more heavily on adults (i.e. the 16+ age group) and relatively lightly on infants and the 1-4 age group. In the later period from March to July, the table looks much more like that of the seventeenth century English parishes of Colyton and Eyam, though it is still very light for the mortality at all ages under five, and especially so in the case of infants. By comparison with the English examples, there seems to have been some factor reducing the relative impact of crisis

mortality on infants, throughout the crisis, and an additional factor, operating in the first three months of the crisis, increasing the mortality of adults.39

In order to pursue the analysis further, a household analysis of the impact of crisis mortality was attempted. As has already been pointed out, only the units listed in the poll tax register can be used, subdivided into households according to criteria, which were specified above. In almost all the settlements there are other households about which some information can be derived solely from the 1695-1710 register, but they still have to be rejected because of insufficient evidence as to which members of the household, over the age of 15, were resident in the settlement at the beginning of the crisis period. Moreover, there is no information as to any servants who may have been co-resident. These rejected households comprised almost half of the total number of households on which we have some record. Of course, the critical question is whether the households that are used in the analysis were, indeed, representative of the total. The most that can be said on this point is that there are no obvious grounds for suggesting that they are unrepresentative. Of the 475 households suitable for this analysis, 189 (40 per cent) present no difficulty in that the size of the household at the beginning of the crisis, and the number who died in the crisis, are clearly determined. These are known as Group A. The remaining 60 per cent, which constitutes Group B, present one or more of the following difficulties. Serving sailors appear in the form of a relative, who is listed, e.g. the ‘wife of Båtsman Varg’, leaving considerable doubt as to whether the ‘båtsman’ himself was absent on duty or whether he was resident, and omitted because he was exempt from taxation. Alternatively, a son or a daughter, appears in the burial register in the crisis period, of whom there is no previous record. They have been assumed to be over the age of 15, since they do not appear in the 1695-1710 register. As they also do not appear in the 1710 poll tax listing for their settlement, they may have been away in service. What is not clear is whether they rejoined the household before death, or died away from home. There is a similar doubt about the death of servants appearing in the 1710 poll tax listing and not recorded as buried in the same settlement. For Group B, therefore, a ‘maximum’ and ‘minimum’ size of the household was calculated. The ‘maximum’ assessment took the greatest possible household size and the appropriate burials, while the ‘minimum’ took the smallest possible size. All the tables were worked out three times: once using ‘Group A’ only, once using ‘Group A’ plus the ‘minimum’ assessments for ‘Group B’, and once using ‘Group A’ plus the ‘maximum’ assessments for Group B. Statistically, the calculations using ‘Group A’ only may appear to be more reliable, but we do not know whether these households are sufficiently representative, especially as they exclude serving sailors and servants. The other two sets of calculations are based on all the available households, but they involved assumptions, which could not be proved. For example, the ‘maximum’ calculation is subject to the criticism that individuals are taken into account only because they died. Furthermore, it cannot take into account similar individuals who joined the household after the 1710 tax assessment, but did not die. This last point introduces an undesirable bias.

39 The normal seasonal rhythm is for infant deaths to peak in this period from December to February), and for adult deaths from March to April, so it is not the normal pattern of seasonality that is producing this effect. Schofield and Wrigley, ‘Infant and child mortality’, p.89.

In an article on the plague epidemic in Colyton in 1645-6 I argued that to catch a disease outside the household a person must both meet someone carrying the disease and catch it from him or her.40 But everyone else in the household can do the same, and if we make the simplifying assumption that each member of the household is equally likely to catch the disease outside, then the number of members of the household catching the disease is directly related to the size of the household.41 If the chance of catching the disease from other members of the household living in close physical proximity were high, we should expect that the numbers of infected people around, and hence the size of the household, would considerably influence an individual’s chance of infection. Thus the size of household might be expected to be an important factor in catching diseases spread by contagion or by human parasites such as lice or fleas. For example typhus would be a case in point, or bubonic plague when the human (and not the rat) flea is the significant vector, or indeed where the spread of plague is directly through the pneumonic form. But if the members of the household were mobile, and if their chance of meeting an infected person outside the household and catching the disease from him were high, as might be the case during an epidemic of influenza, there would be little additional risk of infection from other members of the household, and we would expect only a weak association between the chance of infection and household size. Finally, if the chance of catching the disease from another member of the household is zero then the number of infected people in the household is quite irrelevant, and an individual’s chance of infection will have nothing to do with household size. Thus, if bubonic plague is only introduced into households by rats and is only transmitted to man by the rat flea, and never from man to man by the human flea, and never pneumonically, we would expect to find no association between the chance of infection and household size. Once we have made the logical distinction between the clustering of a disease in certain households, and the association of its incidence with household size, we have a useful means of identifying the relative importance of different factors in the transmission of the disease. Thus we would expect the incidence of airborne infections, such as influenza, to be weakly clustered by household and strongly associated with household size. Diseases spread by contagion, or by human parasites, for example typhus or bubonic plague through the human flea, might be expected to show a moderate clustering by household and a very strong association with household size. On the other hand, diseases spread by animal parasites, such as bubonic plague through the rat flea, would show marked differences between households, depending on the movement of the animal hosts, and this would be reflected in a strong clustering by household, quite independently of household size. If, however, an individual infected with plague became infective pneumonically, one would expect all the other members of the household to be at a severe risk of infection; infection, that is, would be strongly clustered by household, and very strongly related to household size. Thus, so far as bubonic plague is concerned, the distinction between household clustering and association with household size would also appear to provide a means of discriminating between the

40 The argument is set out in detail in Schofield, ‘An anatomy of an epidemic’, p.101-115. 41 We can use the binomial formula again to express our expectations under random conditions in more formal terms. If p is the chance of catching the disease and there are n people in a household, the average number of people infected in a series of ‘trials’ involving households of size n is np. If p is constant across households of all sizes then the number of people infected depends on the size of the household.

human flea and the rat flea as the significant vector in a given plague epidemic. That is, strong clustering by household would suggest the rat flea, while a strong association with household size would indicate the human flea. Pneumonic plague would combine the two into strong clustering by household, and a strong association with household size.

Table 8. Household size and death rate by household size, Bräkne-Hoby Household 95% No. No. dying in + Deaths per household No. deaths confidence households persons size interval 0 1 2 3 4 5 6 7 8+

Size of household and numbers dying, both known (Group A)

1 16 6 22 22 6 27.3 + 19 2 21 7 3 31 62 13 21.0 + 10 3 9 11 12 4 36 108 47 43.5 + 9 4 6 7 15 10 2 40 160 75 46.9 + 8 5 2 3 6 7 2 3 23 115 59 51.3 + 9 6 2 4 3 1 7 3 1 21 126 62 49.2 + 9 7 0 0 0 1 1 3 1 0 6 42 28 66.7 + 14 8+ 1 0 0 2 1 1 2 1 2 10 83 50 60.2 + 14

Total 57 38 39 25 13 10 4 1 2 189 718 340 47.4

Group A plus maximum of Group B

1 16 23 39 39 23 59.0 + 15 2 39 22 17 78 156 56 35.9 + 8 3 27 28 25 9 89 267 105 39.3 + 6 4 12 18 34 19 6 89 356 167 46.9 + 5 5 6 8 21 18 15 7 75 375 199 53.1 + 5 6 2 5 6 6 14 5 2 40 240 128 53.3 + 6 7 1 0 2 3 6 7 7 1 27 189 121 64.0 + 7 8+ 1 0 2 5 2 6 8 9 3 38 334 210 62.9 + 5

Total 104 104 107 60 43 25 17 10 3 475 1956 1009 51.6 cont… 20

Table 8. cont… Group A plus minimum of Group B

1 72 25 97 97 25 25.8 + 9 2 48 34 19 101 202 72 35.6 + 7 3 23 27 24 13 87 261 114 43.7 + 6 4 9 15 27 13 3 67 268 120 44.8 + 6 5 3 5 14 19 7 4 52 260 138 53.1 + 6 6 3 5 3 4 9 5 3 32 192 102 53.1 + 7 7 0 0 3 3 3 3 4 0 16 112 66 58.9 + 9 8+ 1 0 1 3 3 3 7 2 3 23 192 118 61.5 + 7

Total 159 111 99 55 25 15 14 2 3 475 1584 775 47.7

Source: Household forms as defined in the text.

Table 9. Death rates by size of family, Colyton, November 1645-November 1646 Size of family % dying in + 95% confidence interval 1 21 + 13 2 17 + 6 3 12 + 4 4 20 + 4 5 25 + 6 6 22 + 6 7 22 + 7 8+ 23 + 6

All 20 Source: Family reconstitution forms, selected if the date of marriage known, or if the date of baptism of the first child was known , and it was before the beginning of the epidemic, and at least one parent was alive at the beginning of the epidemic.

In the case of the English parish of Colyton, the association was even weaker, and this was taken as evidence that person-to-person transmission of the disease through the human flea was not an important element.42 In the case of the Bräkne-Hoby, however, people in Group A, and people in Group A plus the minimum of Group B, showed a marked tendency for their death rate to be a function of their household size. This would seem to confirm either the human, but not the rat flea, or plague spread pneumonically, as a serious contender for the spread of the epidemic. In fact for ‘Group A’, it is possible to work out corresponding figures for two separate periods, from December to February and from March to July in 1711. The results are set out in Table 11 and show a weak correlation in the earlier period, and an even stronger correlation for the later period than was obtained from the results for the period as a whole. Thus the period for which the human flea, or the pneumonic form of the disease, seems a relevant form appears to run from March 1711 until the end of the epidemic. If we now stay with ‘Group A’, divided into its two sub-periods (December to February and March to July), we can investigate the second aspect of the pattern of incidence of an epidemic, that is how far burials were clustered in certain families and not in others. In order to achieve this, we need to have some idea of what the distribution of burials would have been like in the absence of any clustering by family. If we first assume that everyone’s chance of dying was the same, and was independent of whether someone else in the same family had died or not, we can calculate for each size of family the number of families there would be with 0, 1, 2, 3, 4 and 5 and above deaths. With the outcomes of death and survival amongst families of say, size 3, there are four possibilities: all three survive, two survive and one dies, one survives and two die, and all three die. If we then calculate what is known as the ‘binomial expansion’ of the death and survival rates amongst those living in a given household size, we obtain the number of families that are ‘expected’ to have 0,1,2 and 3 deaths.43 The result of applying the ‘binomial expansion’ assumes the individual’s chance of dying is the same, and that it is independent of whether someone else, in the household happened to die or survive. In other words the ‘binomial expansion’ gives us what we are seeking: namely the distribution of deaths by household we could expect in an epidemic if the deaths had occurred randomly throughout the population and quite independently of each other.

42 Schofield, ‘Anatomy of an epidemic’, pp. 105-6. 43 That is the cube the probability of surviving for the chance of there being no death; the probability of surviving squared times the probability of dying for the chance of there being one death; the probability of surviving times the probability of dying squared for the chance of their being two deaths; and the probability of dying cubed, for the chance of their being three deaths.

Table 10. ‘Group A’: Observed frequencies of deaths and percentage dying by family size, December 1710-February 1711 and March-July 1711. Household Deaths per household No. No. No. deaths % dying in + size households persons 95% confidence interval 0 1 2 3 4 5+ 6 7 8+ December-February 1 18 2 20 20 2 10.0 + 13 2 28 2 0 30 60 2 3.3 + 5 3 32 2 2 0 36 108 6 5.6 + 4 4 27 9 2 2 0 40 160 19 11.9 + 5 5 17 2 1 1 2 0 23 115 15 13.0 + 6 6 15 4 1 0 1 0 0 21 126 10 7.9 + 5 7 5 0 0 0 0 0 1 0 6 42 6 14.3 + 11 8+ 7 1 0 0 0 2 1 1 0 10 83 14 16.9 + 8 Total 149 22 6 3 3 0 2 1 0 186 714 74 10.4

March-July 1 25 3 28 28 3 10.7 + 11 2 27 4 3 34 68 10 14.7 + 8 3 13 12 13 4 42 126 50 39.7 + 9 4 7 4 12 5 2 30 120 51 42.5 + 9 5 5 2 6 5 2 2 22 110 47 42.7 + 9 6 2 2 3 0 5 2 1 15 90 44 48.9 + 10 7 0 0 1 1 1 3 0 0 6 42 24 57.1 + 15 8+ 1 0 0 1 1 1 1 0 2 7 58 34 58.6 + 13

Total 80 27 38 16 11 8 2 0 2 184 642 263 41.0

Source: Table 9.

Table 11. Group A: expected frequencies Household Deaths per household No. size households 0 1 2 3 4 5+ December-February 1 18 2 20 2 28.1 1.9 0 30 3 30.3 5.4 0.3 0 36 4 24.1 13 2.6 0.2 0 39.9 5 11.5 8.6 2.6 0.4 0 0 23.1 6 12.8 6.6 1.4 0.2 0 0 21 7 2 2.4 1.2 0.3 0.1 0 6 8+ 2.3 3.5 2.5 1.1 0.4 0.1 9.9

Total 129.1 43.4 10.6 2.2 0.5 0.1 185.9

March-July 1 25 3 28 2 24.7 8.5 0.7 33.9 3 9.2 18.2 12 2.6 42 4 3.3 9.7 10.7 5.3 1 30 5 1.4 5.1 7.5 5.6 2.1 0.3 22 6 0.3 1.5 3.7 4.7 3.4 1.5 15.1 7 0 0.1 0.6 1.5 1.8 2.1 6.1 8+ 0 0 0.3 0.9 1.5 4.1 6.8

Total 63.9 46.1 35.5 20.6 9.8 4.9 183.9

Source: Table 9

Table 11 shows, for each period of time, the results of the calculation of the ‘binomial expansion’ of the survival and death rates of each household size in Table 10. Here the figures in the ‘0’ to ‘5+’ columns in the Table give the ‘expected’ number of deaths one would expect given an independent random distribution of deaths between the households. The discrepancy between the ‘expected’ and observed number of deaths are brought out more clearly in Table 12, in which the observed frequencies are expressed as a ratio of the ‘expected’ frequencies. A ratio of 1.00 means that the same number of burials was observed as would have been expected under independent random conditions. A ratio of 2.00 indicates an observed frequency that is twice the expected frequency, a ratio of 0.50 an observed frequency half the expected frequency, and so on. Thus, if clustering of deaths occurs, so that some households are badly affected whilst others escape, the ratios in Table 13 would be expected to be greater than 1.00 towards either end of each row and less than 1.00 in the middle terms of the rows.

Table 12. Group A: Ratio of observed to expected frequencies Household size Deaths per household 0 1 2 3 4 5+

December-February 1 1.00 1.00 2 1.00 1.05 0.00 3 1.06 0.37 6.67 0.00 4 1.12 0.69 0.77 10.00 0.00 5 1.48 0.23 0.38 2.50 1.67 0.00 6 1.17 0.61 0.71 0.00 L 0.00 7 2.50 0.00 0.00 0.00 0.00 L 8+ 3.04 0.28 0.00 0.00 0.00 20.00 Total 1.15 0.51 0.56 1.36 5.56 0.00

March-July 1 1.00 1.00 2 1.09 0.47 4.29 3 1.41 0.66 1.08 1.54 4 2.12 0.41 1.12 0.94 2.00 5 3.57 0.39 0.80 0.89 0.95 6.67 6 6.67 1.33 0.81 0.00 1.47 2.00 7 0.00 0.00 1.67 0.67 0.56 2.14 8+ L 0.00 0.00 1.11 0.60 0.97 Total 1.25 0.58 1.07 0.78 1.12 1.63 Note: L means very large. Source:

A glance at the first panel of Table 12 shows that the grouping of deaths by household was very different from what one would expect on the basis of an independent random allocation. For every household size the number with extreme experiences (no one dying; everyone, or almost everyone, dying) was much greater than expected, and the number of households with some people dying and others surviving was correspondingly far fewer than expected. There was, therefore, a marked tendency for deaths to cluster in some households, and for other household to escape the epidemic altogether.44 Thus in Bräkne-Hoby, in 1711, an individual’s chance of dying in the epidemic was dependent on what happened to others in the same household, a pattern which grew stronger in the second period of March to July 1711. Furthermore, if the poll tax listing can be considered to give an approximate, although imperfect, indication of household size, the chance of dying did have something to do with the size of household in which the individual lived. If our hypothetical arguments, associating different patterns of mortality with different modes of transmission of diseases are correct, then this combination of a clustering of deaths with an apparent association between mortality and household size, suggests that in Bräkne-Hoby in 1711, the epidemic is likely to have been caused either by diseases spread by contagion, or by human parasites, or by plague in its pneumonic form. In view of the very high level of mortality in this latter phase, the last possibility is highly likely. In Bräkne-Hoby there was effectively no correlation with household size during the early days of the epidemic, followed by a very strong correlation from March to July. One possible explanation of this difference is that initially the plague reached Bräkne-Hoby and was spread by a vector, the rat flea, which caused no correlation with household size. In Bräkne-Hoby, however, the build-up in the plague mortality was accompanied by a very marked correlation between mortality and household size. This would appear to indicate that the disease was largely

44 There is one exception to this pattern and this lies in the second panel of the table referring to the epidemic in the March-July period. Here, despite the general sharpening of the picture presented for the previous period (December-February), is destroyed by the large households of size 7 and above. But this relates only to 13 households of this size and they may well have been subjected only to random variation.

spread only to a few households at a time, all the members of which died after infection by one person. Thus the spread of plague, that is, became pneumonic, and it would have reached the isolated hamlets and farmsteads, to which did not lie on the tracks of rats. For example, in the west of the parish surrounded by woodland and steep hills, lay Klingsmåla a hamlet with a few fields, lying approximately 10 kilometers from the main settlement of Bräkne-Hoby. On 6 April 1711 there was recorded in the burial register the death of a Svän Månsson, along wth two of his daughters and a son. The next day, the 7 April, the burial of a further son was recorded. There were burial entries from elsewhere in the parish from the 8 to the 12 April 1711, then a final son and daughter of Svän Månson were buried.45 So we have seven people from one family in quite a remote location, and a large proportion of this family was infected and buried within one week, the greater part of the burials occurring within a two- day period. Thereafter, despite the fact that the epidemic ran for a further 3½ months there were no burials occurring in Klingsmåla. This was a settlement which seems to have been a long way off the roads linking markets, and probably very distant from the tracks of rats. We can only surmise that someone, possibly Svän Månson himself, caught the plague by being in direct range of a cough from someone that already had it, and then introduced it into the farmstead at Klingsmåla. It should be noted that this was, indeed, likely to be overcrowded with a high concentration of sputum particles lying in the air, as for example would have been produced by several children asleep in the kitchen.46 In Bräkne-Hoby, therefore, it seems that we have a relatively short period of bubonic plague, spread by the rat flea. And then a period in which it was also spread, especially in the remote settlements in the north of the parish, in the classic manner, but this time it was a pneumonic plague, with a very high mortality.47 After all, the plague spread throughout the parish, infecting almost everybody, and killing three-quarters of the inhabitants. So great an infection, and so great a mortality during the first six months of 1711, could only be achieved by plague. In this case, though, it was through a combination of two radically different means of spreading the disease; initially through a classic bubonic means of spread through the rat flea; but latterly through the pneumonic means of spread ironically it was through an infection of the inhabitants by their own friends and neighbours.

Cambridge Group for the History of Population and Social Structure University of Cambridge

Footnote references Audoin-Rouzeau, F, ‘Le rat noir (Rattus rattus) at la peste dans l’occident antique etmedievale, Bulletin de la Societe Pathologie Exotique, (1999), 92; 5, Pt. 2, pp. 422-26. Audoin-Rouzeau, F, & J-D Vigne, ‘La colonisation de l’europe par le rat noir (Rattus), Révue de Paléologie, (1994), 1; pp. 125-145. Benedictow, O.J., ‘Morbidity in historical plague epidemics’, Population Studies, 41 (1987), pp. 401-431. Biraben, J-N., ‘Current medical and epidemiological views on plague’ in P. Slack (ed), The plague reconsidered, pp. 25-36. This is a translation of Les hommes et la peste en France et dans les pays européen et mediterranéens, vol 1, pp. 9-21. Bradley, L, ‘Some medical aspects of plague’, in P. Slack (ed.), The plague reconsidered, pp. 11-23. Bradley, L., ‘The most famous of all English plagues: a detailed analysis of the plague at Eyam, 1665-6’, in P. Slack (ed.), The plague reconsidered, pp. 63-94. Bradley, L., ‘The geographical spread of plague’, in P. Slack (ed.), The plague reconsidered, pp.127-32. Coale, A. J. and P. Demeny, Regional model life tables and stable populations, Princeton University Press, 1966. Hult, O.T., ‘Pesten i Sverige 1710’, Hygienisk Tidskrift, VIII, Stockholm, 1916, pp. 96-185. Laslett, P., ‘Introduction: comparing illegitimacy over time and between cultures’, in P. Laslett, K.sterveen and R. M. Smith (eds.), Bastardy and its comparative history, 1980, pp. 1-65. Moseng, O. G., ‘Nordens siste pestepidemi: en punkstudie av Allerum 1710-1711’, Hovedoppgave i historie, Universitetet i Oslo, 1990.

45 Bräkne-Hoby, Parish register, pp.154-5. 46 Wu, Epidemiology of pneumonic plague, pp.180-1. 47 Pitkänen, in his study of the plague epidemic which struck in south and west Finland in 1710-1711 considers pneumonic plague, but thinks that it is unlikely to have been the operative factor in Finland.

Morris, C., ‘Plague in Britain’, in P. Slack (ed.), The plague reconsidered, pp. 37-38. Persson, B.E.B, Pestens gảta; farsotens i det tidiga 1700-talets Skåne, 2001. Pitkänen, K., ‘Pesten i Finland 1710-1711- en tvivel underkastad historia’, Historisk Tidskrift fðr Finland, 1977, pp. 201-214. Plague Research Commission, ‘On the spread of epidemic plague through districts with scattered villages. Part 1, Journal of Hygiene, (1910), pp. 349-415, with 48 maps. Plague Research Commission, ‘Reports on plague investigations in India. XXII. The epidemiological observations made by the Commission in Bombay City, Journal of Hygiene, vol. 7, 1907, pp. 724-98. The age and sex- specific incidence rates are given in table XIV, on pp. 763-4. Schofield, R., ‘Microdemography and epidemic mortality: two case studies’, in J. Sundin and E. Söderlund (eds.), Time, space and man: essays on microdemography, Stockholm, 1979, pp. 53-67. Slack, P. (ed.), The Plague reconsidered: a new look at its origins and effects in sixteenth and seventeenth century England, Local Population Studies Supplement, 1977. Swellengrebel, N.H. and Hoesen, H.W., ‘Über vorkommen von rattenpest ohne menschenpest in klandestinen herden, Zeitschriften för Hygienen und infektuosen krankheiten, vol.79, p. 436. Wieselgren, P. Ny Smålands beskrifning, vol. 2, 1844-5. Wu, Lien Teh, A treatise on pneumonic plague, Geneva, 1926. Wrigley, E.A., ‘Marriage, fertility and population growth in eighteenth century England’, in R.B. Outhwaite (ed.), Marriage and society: studies in the social history of marriage, 1981, pp.137-185. Wrigley E.A. and R. Schofield, The population history of England, 1541-1871: a reconstruction, Cambridge University Press, 1993.