Transgenerational effects of early-life experiences on descendants’ height and life span. An explorative study using Texel Island (Netherlands) genealogies, 18th-21st centuries

ABSTRACT Early-life experiences can have lasting effects on health across multiple generations. The pathways of these transgenerational transmissions are difficult to explore, because of the complex interactions of social and biological factors involved. This study explores the potential role of one such pathway – inherited epigenetic modifications to gene expression – by controlling for shared environmental factors. It uses a database constructed from descendant genealogies of six lineages from Texel Island, The Netherlands. Heights and life spans of respectively 2761 and 3279 19-year old boys are related to the early-life experiences of themselves, parents and grandparents. Adversity in early-life is studied through trauma and food deprivation. Adversity has clear effects, especially on heights, but few of these effects were transmitted to children and grandchildren.


Introduction
Starting in the second half of the nineteenth century, the Dutch experienced a remarkable increase in stature with each generation of men being on average four centimeters taller than their fathers (see Quanjer, 2023). This development raises the question of the relative importance of genetic versus environmental dimensions of growth. We know that heritability of height is very strong, but also that it take a long time span and selection processes for DNA to change. In other words, genetic changes cannot account for the rapid increase in height. Does that mean that the only explanation is a steady improvement of socio-economic conditions resulting in higher 'net nutrition' in each generation (Floud et al., 2011)? There may be an alternative explanation in the form of 'epigenetic' or 'imprinting' effects which imply changes in the expression of genes, which can be transmitted to descendants. Such effects constitute pathways linking environmental to biological changes and resulting in long-term or even lasting effects on growth and health of subsequent generations.
Epigenetic effects have been proven in animal studies, but, understandably, experiments involving humans are very limited (Lundborg & Stenberg, 2010). However, historical conditions, especially famines such as the Dutch Hunger Winter of 1944/1945, have been used as 'natural experiments' (e.g., Heijmans et al., 2008Stein & Lumey, 2000). A recent review of the literature on the transmission of parental (early) life experiences to children indicates that the most important factors are the parents' food intake and their metabolic diseases and their psychological traumas and stress (Jawaid et al., 2021). To some extent, early-life dietary exposure and traumas are still visible in the growth and health of the grandchildren. So far, the research on epigenetic mechanisms has focused mainly on all-and cause-specific mortality as well as on premature births and birth weight, with only a few studies looking at adult height outcomes (e.g., Van den Berg & Pinger, 2016). However, uncovering the role and reach of such mechanisms may well offer a missing piece of the secular growth puzzle (Pembrey, 1996). Martorell & Zongrone speculate that the secular increase in height should be interpreted as the 'washing out' of epigenetic effects of the food crises of the first half of the nineteenth century (Martorell & Zongrone, 2012, p. 308). Indeed, if epigenetic effects can be demonstrated, this would be highly relevant for our understanding of the timing of the nineteenth-century increase in heights. It would mean that not only the conditions during the growth period of a specific generation mattered, but also the experiences of previous generations.
However, not only are studies of epigenetic mechanisms in humans still rare, the concept itself is also strongly debated. Some geneticists argue that epigenetics are a simplistic 'get out of genetics free card'. 1 Others point at the enormous methodological challenges of isolating epigenetic pathways from other mechanisms of transmission across generations. For instance, problems in childhood can also be transmitted through poverty and bad parenting (Bowers & Yehuda, 2016;Martorell & Zongrone, 2012). Such other pathways of transgenerational transmission may well prevail over the epigenetic pathway (Horsthemke, 2018, p. 3). Finally, there is considerable confusion as to the processes involved, which differ dependent on the timing of exposure by parents or grandparents.
This article explores how historical datasets can be used to offer insight in the potential role of epigenetics mechanisms. In principle, datasets spanning multiple generations which are observed in their (early-life) experiences of adverse conditions and health outcomes offer the material for a 'natural experiment' of epigenetic mechanisms. But how to set up such as experiment? The first aim of this article is to suggest an approach to distinguish between epigenetic and other transgenerational effects (from grandparents to grandchildren), while also taking account of intergenerational effects (parents to children). The second aim is to test this approach in an experiment with descendant genealogies. These genealogies pertain to families originating from the Island of Texel in the northwest of the Netherlands. I constructed a database with descendants from six founder couples which were traced from the late seventeenth century onward even when living away from the Island. The genealogies were then linked to the exceptionally well-preserved nineteenth-and twentieth-century records of conscripts (heights measured at age 19 or 20) as well as the militia (measured at age 25). The data allow me to run regressions of height and lifespan on parental as well as grandparental early-life experiences, effectively spanning four generations of descendants including those living on the mainland. The experiment may offer insights useful for future studies of transgenerational effects. Finally, the results will be evaluated from the perspective of the secular increase in height and longevity: what is the potential impact of previous generations?

Transmission of early-life experiences
How can early-life experiences be physically transmitted to offspring without involving changes in DNA? One import pathway is through the development of mother's reproductive organs, which already begins in utero. Undernourishment of a female fetus leads to low birth weights of her own children and grandchildren, possibly through the way uterine vasculature is formed affecting placentation (Scott & Duncan, 2002, p. 138).Thus, there is a strong correlation between mother's birthweight and the birthweight of her offspring (Martorell & Zongrone, 2012, p. 305). A recent Norwegian study suggests that food deprivation during pregnancy may also lead to 'culling' or loss of relatively unhealthy fetuses. The health of surviving fetuses may actually translate into longer life spans of the grandchildren (Sari et al. 2021).
Another pathway relates to how information in DNA is passed on, which might also imply the temporary suppression of information. This may involve changing the histones or proteins packing and ordering DNA (Jawaid et al., 2021). An important process is DNA methylation, the adding of methyl groups to DNA molecules thus changing their activity (but not the sequences themselves). Methylation of genes responsible for metabolism and dealing with stress can be caused by exposure to toxins (e.g., through smoking), but also by hunger or trauma (Burton & Metcalfe, 2014;Jawaid et al., 2021).
It may also be relevant for height. A 2014 study showed that among (by then ascertained) 87 height-associated genes no less than 72 were susceptible to methylation (Simeone & Alberti, 2014).
It makes sense that the growing body responses to important cues from the environment (which is known as 'developmental plasticity'), but how do the early-life experiences of the parents or grandparents fit in this? Several authors point at the 'environmental information' a fetus needs to 'calibrate' its growth trajectory. The mother's situation during gestation may not be a reliable indicator of a fetus' future prospects. The 'thrifty phenotype' hypothesis posits that the fetus adjusts its growth to conditions experienced by the mother -but this can lead to a 'mismatch' when conditions after birth are much better (or worse) resulting in health problems (Heard & Martienssen, 2014, p. 105). Kuzawa had described the dangers of the mismatch as follows: '(. . .) the fetus walks a metabolic tightrope, as it faces the challenge of setting its growth trajectory high enough to reap the fitness benefits of high productivity and growth rate, but not so high as to overshoot what available resources can support, and thereby increase the frequency or severity of negative energy balance' (Kuzawa, 2005, p. 9). Therefore, the calibration process has to start in advance and implies that the fetus, through the -epigenetically transmitted -messages it had received, adjusts its growth trajectory to the average ecological conditions experienced by its parents and grandparents. This (hypothetical) mechanism is labeled the intergenerational phenotypic inertia model (see, Kuzawa & Eisenberg, 2014).
These 'messages' for the future generations operate through reprogramming the germ cells, which takes place in early-life. Prepubertal 'slow growth period' (SGP) are seen as such a window of opportunity for reprogramming (Wu et al., 2015). In most researches, SGP is defined as 8-12 although in historical populations this period might have to be extended as pubertal growth spurts occurred later (Gao& Schneider, 2021;Quanjer & Kok, 2021). This interpretation of epigenetic reprogramming suggests counterintuitive outcomes: adverse experiences result in altered gene expressions that, eventually, improve the health of next generations. Several studies report results that support this idea. Van den Berg and Pinger (2016) studied health outcomes, including adult height, of grandchildren of persons who had experienced the German famine of 1916-1918 during their SGP. Although the grandchildren's height had the expected (positive) sign, the results were not statistically significant. However, they did find positive mental health scores of grandchildren. An important case study in this field is the research on Överkalix, an isolated municipality in northeast Sweden (e.g., Bygren et al., 2001;Kaati et al., 2007;Pembrey et al., 2006). From a relatively small number of probands (N = 146 men and 125 women) born in 1895, 1905 or 1920 all parents and grandparents were traced. Effects were found of the food availability (based on local harvest figures) during their SGP on grandchildren's cause-specific and all-cause mortality. Interestingly, the children were not affected, thus the effect skipped a generation (Bygren et al., 2001, p. 57). Transmission appears to be gender-specific: the mortality rate of men was linked to their paternal grandfather's food supply during his SGP, whereas for women only their paternal grandmother's food supply mattered (Kaati et al., 2007;Pembrey et al., 2006). Researchers were not only interested in famines; in fact abundant crops apparently leading to 'overfeeding' produced the strongest results, such as a higher mortality of diabetes of grandchildren (for an overview, see, Pembrey et al., 2014). Recently the results could be replicated on a much larger sample (N = 11,561 probands) drawn from the Uppsala Birth Cohort Multigeneration Study. The study confirmed that good food supply of the paternal grandfather during his SGP increased the mortality risk of the grandson (Vågerö et al., 2018).
Epigenetic effects of childhood trauma also appear to be gender-specific, but seem to work quite different from food shortages. Boys (but not girls) having experienced the death of a parent during their SGP tended to father children with lower birthweights and children with a higher likelihood of premature births. This research used no less than 800.000 Swedish probands born between 1976 and 2002 (Vågerö & Rajaleid, 2016). It is not clear why childhood trauma translates to lower birthweights of children. The authors do not speculate on the mechanisms, but simply conclude 'Our study fails to refute the hypothesis that a male-line epigenetic mechanism exists which may be triggered by trauma during boys' slow growth period'(ibid., p. 10).
Possibly, there are some evolutionary advantages in responding to a traumatic experience of death in the environment. Individuals growing up in a high-mortality environment may be set on a 'fast life strategy' of reproductive maturation, which however may come at a price to their own growth, life span and offspring's quality as energy normally devoted to growth is diverted to speedy reproduction (Hill & Kaplan, 1999;Kuzawa, 2005;Stulp & Barrett, 2016). If this also affects offspring's development it would imply that children and grandchildren of individuals who experienced parental death in early youth were shorter (because energy was diverted to reproduction) as well as unhealthier (resulting in a shorter life span).
Although epigenetic effects in humans have been ascertained in many studies (Jawaid et al., 2021), their transmission to next generations is heavily debated. Some authors state that reprogramming in the germline will undo such effects (Heard & Martienssen, 2014, p. 105). The Barbados Nutrition Study, covering 168 persons in two generations, has shown that although many genomic regions are methylated because of malnutritionleading to severe health effects in later life -only few of them were carried over to the next generation (Peter et al., 2016). Furthermore, the (expected) outcomes of such transmitted effects are far from clear. For instance, do adverse early-life experiences of grandparents lead to a longer life span of their grandchildren, because of the 'intergenerational phenotypic inertia' mechanism? Or to a shorter life span, when a 'fast life strategy' has somehow been imprinted? For this exploratory article, it is more relevant to discover whether we can actually find such effects than to explain how they work. In doing so, we face two more challenges. First, how to disentangle transgenerational and intergenerational epigenetic effects and, second, how to distinguish epigenetic effects from genetic effects as well as from socio-cultural pathways of transmission?
The first challenge implies we have to separate effects from direct exposures to cells impacting the second generation from effects on the third generation without exposure of the second generation (Sales et al., 2017). In other words, the imprinting should persist from the first to the third generation. Men form germ cells of their children before puberty, but women already form those cells while still a fetus. When the pregnant mother experiences adversity, this affects her fetus as well as the germ cells of her grandchildren. Thus a lasting effect has to be visible in her great-grandchildren. In the diagram in Figure 1 effects across four generations are displayed, but for practical reasons I will limit the transgenerational transmission of early life experiences to three generations. The Figure shows how experiences of the grandparents (GP) lead to epigenetic modifications, which may directly affect their grandsons, the 19-year old boys central to this study. However, intergenerational epigenetic effects can result from the early life experiences of the parents -meaning we should control for them if we aim to find transgenerational effects -the green lines in the diagram. The same is true for the early life experiences of the 19-year old boys themselves.
In the end, we should be able to compare 'gross' transgenerational effects from 'net' effects, that is, after controlling for the early-life experiences of recruits and their parents. The second challenge is to separate epigenetic from other lasting effects of bad earlylife conditions. In this study childhood adversity is defined as having experienced the death of a parent and/or periods with very high food prices. Such troubles may set in motion other chains of events, which can also impact the life courses of (grand) children. Impoverished or undernourished children can suffer life-long health effects (Quanjer, 2023;Quaranta, 2013) and end up in lower socio-economic positions than their parents with less to offer to their own children. Also, bad economic periods or becoming a (half) orphan can induce migration. Indeed, the economic crisis of the early nineteenth century caused many Island families to leave for the mainland (see next section). Thus, the geographical environment can also be part of the chain of transmission of early-life effects. The diagram shows those pathways, and the need to control for them in order to isolate epigenetic effects. Obviously, not all pathways can be traced, let alone controlled. For example, early childhood adversity can have psychological effects resulting in a different style of parenting (Van den Berg & Pinger, 2016, p. 105).
Finally, what about genes? As the diagram shows, height and longevity of the parents is influenced by both genes and epigenetic modifications which cannot be separated. I will run models with and without controls for parental stature and longevity. Unfortunately, height information is only available for men. In the next section, I will elaborate how the dataset for this experiment was created and how the variables were constructed.

Setting and data
This research is based on published genealogies listing all descendants of six seventeenth-century founder couples (Dijt & Dijt, 1970, 1973. My data collection starts with the third or fourth generation, that is persons born from 1785 onward, who were enlisted in 1813 (all adult unmarried men) and from 1817 onwards all those conscripted at age 19 (after 1862 age 20). The published genealogies end with the children of the sixth generation, most of them born in the last quarter of the nineteenth century. All male and female descendants were listed, often with dates of death, including those who had left the Island and had emigrated, for instance, to North America. Most departed families, however, remained in the province of North-Holland, especially in the nearby harbor town of Den Helder. This town offered plenty of employment, e.g., on the large naval dockyards. For the database I selected men who had reached at least the age of conscription. Using data on descendants implies I have relatively few observations in the early nineteenth century. Therefore, I added husbands of female descendants born before 1840, their brothers, and if possible, their fathers. In order to include the first half of the twentieth century, I also added the sons in the seventh generation, which could be found on an extensive genealogical website which is also based on the published genealogies. The website, however, contains relatively more information on people who had not left the Island and/or were part of the site's manager's family tree (http://www.robgomes.nl/ tng/) 2 . My data collection ends with boys born in 1920 because it is not possible to collect individual heights after 1940.
This procedure resulted in a database of 3968 boys who could have been measured during the medical examination which was part of the draft. To find their heights, I consulted the municipal archives of Texel and Den Helder for the cohorts born before 1850. Boys born later could be traced online as the examination registers from 1870 onwards have been digitized for the entire province of North-Holland (including Amsterdam) as well as for some large cities, such as Rotterdam. This implies that I have found more heights of offspring living elsewhere from birth cohorts after 1850. In total, heights for more than 1000 boys are lacking for several reasons. First, they had emigrated or their early death had been missed by the genealogists. Second, the officials failed to note the recruits´ heights in the records for Texel in entire periods (1824-1828, 1831-1832, 1837, 1847, 1859-1860). Third, boys could be absent from the examination, for instance, because they were at sea. In principle, they would be enlisted when they failed to show up. In other research, it has been shown that boys from elite groups managed to avoid the examination altogether (Quanjer & Kok, 2020). Finally, after about 1900, boys who were to be exempted because a brother had been enlisted were no longer measured. In total I have 2761 heights of conscripted descendants and in-laws in the database. For the cohorts 1800-1855 it is also possible to include the heights at age 25 when men still living on the Island were included in the ´civic guards´ militia. This source yields another 860 heights. The life span after age 19 is known for 3279 boys.
Texel Island was strategically located on the main roadstead of the Dutch East India Company, where ships had to be piloted and supplied (e.g., with fresh water). The Island economy prospered in the seventeenth and a large part of the eighteenth centuries. Furthermore, the Island was very suitable for farming, especially sheep grazing. The decline in shipping precipitated by the Fourth Anglo-Dutch War (1780-1784) was accelerated in the Napoleonic era (1794-1814). The early nineteenth century brought no recovery, as harbor activities shifted to the nearby town of Den Helder. Many family dependent on naval activities left the Island, which became culturally and socially isolated from the mainland. However, extensive land reclamation in the 1830s brought newcomers to the north of the Island and the Island's farmers remained in general prosperous due to sheep breeding and cheese exports ( Van der Vlis, 1975). In 1795, the municipality counted 4950 inhabitants and in 1840 4924.
The increasing isolation probably shielded the inhabitants from infections. The combination of geographical isolation with good drinking water and relatively abundant supplies of fresh food (dairy products, fish, birds, eggs, rabbits) seems to have produced a rather healthy population. In the mid-1850s an observer wrote: ´The altitude and fresh sea breeze make Texel a healthy and fertile place´ [an expert has said] . . . I do not believe . . . there is a healthier region, people grow old here and except for the smallpox there are few infectious diseases´ (Allan, 1856, p. 20, translation JK). Indeed, with infant mortality rates (IMR) of 11,9% in 1840-1860 declining to 4,4% in 1914-1923 the Island contrasted favorably with the national averages of respectively 18,9% and 8,0%. Around 1900, it even had the lowest IMR of the country (calculation based on data in Ekamper & Van Poppel, 2008). In relative terms, the child mortality rates of the Island were just as favorable (calculated on the basis of the data in Van den Boomen, 2021).
The Island also stood out with the stature of its 20-year old recruits. Based on aggregate statistics, Bolk (1909) calculated their average height in the period 1898-1907 as 1.73 meters, which made Texel men the tallest in the country (Kok, 2020). In Figure 2, I plot the median heights of all Island conscripts (not just the men in the database) and compare them with the national average as calculated by Drukker and Tassenaar (1997). We can see that the Island´s recruits (total N with heights = 6172) were always taller than the national average, but the advantage was especially impressive in the second part of the 19 th century. The figure also shows the adult heights culled from the civic guards registration (total N = 2306). There appears to be a slight decline in adult heights in the first half of the 19th century, but they still compare favorably to men in cities such as Rotterdam, which in the same period measured slightly below 1.70 meters (Oppers, 1963). For those Texel men for whom we have two measurements we witness an average growth after age 19 of seven centimeters.
The aim of my experiment is to find out whether transgenerational epigenetic effects on heights and longevity can be detected. Thus, do early-life experiences of grandparents influence height and longevity of grandsons, as suggested by the literature on epigenetics? This research is mainly inspired by the famous case study of Överkalix but there are notable differences with their research. First, the food availability on Texel Island did not depend on regional harvest yields, but on price levels on the national level since the Netherlands formed a relatively well-integrated market economy. 'Overfeeding' on bread in periods with relatively low prices does not seem very likely, thus I will focus on food shortages in periods with high prices of rye only. Different price series are available 3 and I calculated averages for 1740-1800 and 1800-1870. Years in which prices were higher by more than 15% were labeled as 'bad years'. The methods yields bad years that mostly match the famine years indicated by Curtis et al. (2017). In this way, I created variables indicating whether the recruits, his parents, or grandparents experienced at least one bad year during specific periods in their youth. For the variables related to childhood trauma, I ascertained whether a father or mother had died between birth or their 18 th birthday. I created separate age categories to discover whether the timing of an adverse experience mattered.  category when the father changed jobs between the birth of his sons. On the Island, swapping between naval and agricultural activities was not uncommon. The second difference with the Överkalix research design is that they used ascendant genealogies, whereas I work with a combination of descendant and ascendant (through the in-laws) genealogies. This implies that my data is clustered more strongly, with brothers, cousins and second cousins sharing sets of (great) grandparents. This clustering implies that I need to add random effects. For practical reasons I have chosen to do this at the level of the parental household. Models with clustering at higher levels (e.g., father's father, mother's father) did not yield different results. Third, the Överkalix probands probably all stem from sedentary families. This means that health selection effects involved in (longdistance) migration have been missed (see, Puschmann et al., 2016). The Texel dataset makes it possible to trace descendants after out-migration. Finally, my research of mortality effects is limited to the life span after age 19, whereas the Överkalix research included infant and child mortality, although the epigenetic effects appear to have been strongest in later life (Kaati et al., 2007, p. 786). But as I deal with survivors, possible effects leading to early deaths of grandchildren remain invisible. Figure 1 presented the different pathways along which grandparents could have influence on the physical growth and life span of individuals. Before testing the 'epigenetic' effects of early-life experiences (Tables 3 and 4), we need to take a closer look at the other pathways. The socio-economic situation in which one grew up is among the most important predictors of height (Quanjer & Kok, 2019). Probably it had less effect on laterlife mortality (Bengtsson et al., 2020). Table 1 shows that when the father was a farmer around the time the son was born, the son's height at age 19 was 11 millimeters taller than others. This is not surprising. But we also see an additional (but borderline significant) effect of mother's father being a farmer. This may mean that the mother herself was relatively healthy, having grown up on a farm, and passed her health on to her child. It can also mean that a farm was more viable when the mother also brought in a dowry in the form of land or money. We see the reverse of this in the model in which we contrast unskilled workers with the rest.

Health and height across three generations
How about the transmission of stature itself? As discussed, genes play a major role in determining one's height. Ideally, we would put the heights of the parents as well as of the grandparents in the model. However, for the (grand)mothers we have no data at all, whereas for the (grand)fathers our data suffers from many missing observations. To avoid ending up with a too small sample, I make liberal use of the category 'height unknown' as a control variable. Therefore, I create dummy variables indicating whether the (grand)father belongs to the top 10% of his birth cohort. Table 1 shows that the son was more than 3,5 centimeter taller when his father was in the top 10% of his cohort at age 19 or 20, and even 5 centimeters taller when his father was in the top 10% of the height distribution at age 25. The latter seems to imply that adult height expresses the 'genetic potential' for growth better than height in late adolescence. After age 19/20 stunted boys could experience catch-up growth -sometimes 20 or 30 centimeters -, depending on whether conditions had improved (also Beekink& Kok, 2017). Because of the many unknowns, the table should be read with caution. A regression using only those men (N = 207) with all anthropometric data (in absolute values) available for themselves, their fathers and both grandfathers, still shows a statistically significant effect of the height of the father but not of the paternal grandfather. On the other hand, the height of maternal grandfather does have an independent effect. This was to be expected, as father's height already captures the effect of the paternal grandfather.
Finally, we look at the role of the geographic environment. Out of the 2761 measured boys, 537 or 19,5% were not born on the Island. They were significantly shorter that the Island-born. But it seems that the effect of the environment was not carried over to the next generation: the birth place of the parents had no significant effects. The birth place of the mother's parents has significant although puzzling effects.A similar exercise was done in Table 2 in which the life span after age 19 is the dependent variable. Although the overall model has less explanatory power, the results generally point in the same directions as in the model on heights. Being born in a farming family increased the life span, as was being born on Texel Island. Being born in an unskilled worker's family also raised life expectancy which may have been caused by selection effects, with the weakest children culled in infancy. However, when mother's father had been an unskilled worker this lowered life expectancy. Father's, mother's and father's mother's longevity (as defined by being in the top 10% of their cohort) 4 had independent effects on the life span of the 19-year boys. When the paternal grandfather was born on the Island, this decreased the life span.
In Table 3, I focus on the effects on height of recruits of the early-life experiences of themselves, their parents and grandparents, with as many controls in place as possible. Being born in a particularly bad year negatively affected height. Childhood traumas also had a strong impact on height, especially the death of the mother between ages 3 and 12. It is likely that effects at earlier ages were even harsher resulting in high infant mortality. Here, we witness the effects on the survivors (see for similar results Quanjer et al., 2023). Father's death had no effect on late adolescents height, indicating that actually not trauma but maternal care was the major factor here (also Quanjer & Kok, 2019).
The table shows few intergenerational effects. Vågerö and Rajaleid (2016) found that men having experienced parental death between 8 and 12 has sons with lower birth weights. We find no such effect here on late adolescent height, although this may be due to smaller absolute numbers. Mother's food deprivation during her adolescence seems to have had a (borderline significant) negative effect on son's height. What about transgenerational effects? Some outcomes for the grandparents are interesting. When father's 1489 † p < 0.1; *p < 0.05; **p < 0.01;*** p < 0.001. Controlled for birthplace, SES, birth cohort, and adult height of the father (top 10%). Furthermore, in the models of the father and mother, the early life experiences of the recruit is controlled. In addition, the models for the grandparents also include controls for the early life experiences of father and mother. mother had experienced the death of her own mother in infancy, this was associated with taller stature in her grandson. And when mother's father had lived through a bad economic period between age 8 and 12, his grandson was shorter. This seems to contradict the positive health effects predicted in the literature, but epigenetic processes occurring in the paternal or maternal line differ strongly for reasons still largely unexplored, let alone understood (M. Pembrey et al., 2014). Table 4 repeats this procedure with the life span of the 19-year old boys. We see that fluctuations in food availability in early youth had almost no impact on the life span, apart from being born in a bad year which leads to a longer life span. Perhaps selection played a role here. The death of the father at an early age had a quite dramatic impact on adult life span of boys, but effects of mother's death were not significant at an older age. The literature indicates that children who lost their mother at a young age died more often than when they lost their father (e.g., Beekink et al., 2002), so again a selection effect probably plays a role. Still, a father's death at an early age set in motion a downward cycle, which has also been demonstrated for the social status of bereaved sons (Rosenbaum-Feldbrügge, 2019). However, the finding that we still see a negative effect of the paternal grandfather's death during father's childhood on the grandson's life span -while controlling for SES, birthplace, own trauma and father's longevity -does suggest a biological rather than social mechanism of transgenerational inheritance. Surprisingly, when father's mother had died in his infancy, his own son appears to have lived longer. Similar 1650 † p < 0.1; *p < 0.05; **p < 0.01;*** p < 0.001. Controlled for birthplace, SES, birth cohort, and longevity of the father and mother (top 10%). Furthermore, in the models of the father and mother, the early life experiences of the recruit is controlled. In addition, the models for the grandparents also include controls for the early life experiences of father and mother.
contradictory findings appear among the grandparents as well. For instance, the death of father's mother's father is associated with a longer life span of her grandson, but the death of her mother with a shorter life span. Tables 3 and 4 shows the 'net' effects of food deprivation and bereavement during the grandparents' youth, after controlling for parental health and parental early-life effects on children. It can be argued, however, that epigenetic effects of grandparents flow through the parents and that controlling for the parental health obscures this. To estimate 'gross' transgenerational effects, I ran models without parental height and longevity, without their early-life experiences, and without all these factors. Although some of the findings in Tables 1 and 2 changed in significance levels, there were no major changes in outcomes (tables available upon request).

Discussion and conclusion
This explorative study aimed to answer the question whether we can detect transgenerational epigenetic effects on height and longevity in past populations. More specifically, it explored the possibility of using descendant genealogies for this purpose.
The previous section has shown that early-life exposure to food availability and parental -especially maternal -death affected individual variation in height and adult life span. But few of those effects were transmitted intergenerational and even fewer transgenerational (that is, when controlling for the parental effects). Moreover, the effects are not very consistent with the hypotheses. The 'intergenerational phenotypic inertia' mechanism predicts that effects on food availability during the Slow Growth Period is translated to same-sex grandchildren in reverse. Thus, same-sex descendants of persons having experienced shortages should display better health. We only see such an effect of the maternal grandmother's food availability on life span, but this defies the gender specification. With respect to height the only significant outcome goes in the 'wrong' direction: grandsons of mother's fathers who had experienced a bad period during their SGP were relatively short. Building on the 'fast life strategy hypothesis', we can predict shorter stature and shorter life span in response to (grand)parental trauma. But again, few effects are statistically significant and they defy consistent conclusions. For instance, when father's mother had experienced the death of her father between age 3-7, the grandson was shorter (Table 3), but he also lived longer (Table 4).
These outcomes can add to the skepticism regarding the heritability of epigenetic effects in humans, or at least regarding the possibility to detect this amid the myriad of other pathways of inter-and transgenerational transmission (see, section 2). But my research design also requires critical evaluation. Why were so few early-life experiences of parents and grandparents carried over to next generations when compared to the Swedish studies? First, I have discussed only statistically significant effects on rather crude indicators (all-cause mortality and height). With a larger and more complete dataset, in particular with more information on the (great) grandparents more effects may come to light. This is a clear caveat of working with a mainly descendant genealogy. Second, the yearly food price indicators used here were obviously not very precise, whereas monthly fluctuations may be needed to trace the timing of effects on, for instance, fetal development. Third, and most important, the Texel Island population may have been quite different from the nineteenth-century Swedes in that they were much less vulnerable to harvest fluctuations. Diets could be enriched with animal protein through e.g., fish, eggs, dairy and meat. And, as on the mainland, grain could easily be imported. In the Netherlands, famines threatened in particular when trade was obstructed, which happened during the Napoleonic era and in 1855 (due to the Crimean War). Also, Dutch poor relief was still able to supplement incomes for the needy, although less so than in the seventeenth and eighteenth centuries (Curtis et al., 2017). Local poor relief may also have mitigated the effects of parental loss, especially by providing for a widow with children (Beekink et al., 2002). Probably due to prospering agriculture and isolation, Texel islanders were among the healthiest and tallest of the country. Their favorable environment may have undone, e.g., through 'reprogramming', the effects of grandparental food deprivation or trauma.
This experiment with genealogies aimed to test some intriguing ideas on how adverse early-life experiences could lead to biological changes, which are supposedly transmitted to future generations. In this test, I aimed to control for other, environmental pathways. I did find clear effects on late adolescent height and adult life span of own early-life experiences with food deprivation and with parental death, socio-economic and geographic environment and parental health. But additional, independent effects of (grand)parental early-life experiences appear to have occurred rarely, they were rather weak and, finally, difficult to understand.
The conclusion could be that epigenetic mechanisms likely played no important role in the Dutch secular increase in stature and longevity. But that verdict is premature and awaits a replication of this research with a more vulnerable population, such as the urban poor, and a larger dataset built on ascendant genealogies.