So, in a sense, the book is de-drama-tizing the classical French migration history, by rejecting the appealing notion of rural exodus. However, the way this is done is very exciting. The book offers many new perspectives on migration as such. Les sentiers invisibles leaves us with a whole new arsenal of testable hypotheses and new methods about the functioning of family networks and its relationship with migration. Just to give a few examples of insights that were new, at least to me. Rosental introduces the concept of family territory, generally a cluster of villages where family member have lived for several generations. Sedentarity and mobility are discussed in terms of this family territory. Socially significant mobility, or migration, can be viewed as leaving the family territory. Equally important is his concept of the migratory project. Individuals are not seen as responding simply and immediately to push and pull forces from the outside. Their migratory decisions are placed against the background of their previous life history, including their trajectory. The combination of their experiences, their anticipations and their perceived opportunities, and only finally some external stimulus, explain why a particular move is made at a particular time.

Rosental exhaustive survey of entire dynasties leads to many interesting fin-dings. The interaction between branches of families that live in different places, proves to be very important for the broa-dening of the occupational horizon of family members. The extention of occupational potential was often crucial for a long distance migration later in life. The same role was played by family in law. Marrying sisters often played key roles in family networks, by introducing their brothers to new occupations and new information networks. What is also impressive in the book is the equal treatment of sedentarity and mobility. Rosental shows how sedentarity acquires a totally different meaning within a relatively short time when the majority of the family had left the family territory. Finally, apart from all the inte-resting case studies and theory building, the book offers an excellent discussion of the literature, not only on migration but also on microhistory and network analysis.

Although the book does not refer to the life course perspective and to life course studies, I find this a splendid example of one is central starting points: the interdependence of lives across both time and space. This is one of the rare studies that really show how macro developments (for instance, the diversification of occupational structure) have repercussions on and acquire meaning at the level of the individuals, always me-diated by their families.

Although I am very impressed by this book, there are still some questions and doubts that I would like to share with you. These doubts are related to the selection of the reseach population. Also, I will discuss one of the central indicators of the book, the level of premarital mobility. Then I will go into an even more technical, but also very crucial issue, the use of the witnesses on marriage certificates. Finally, I would like to review the conclusions of the book.

To begin with the research population. The 97 dynasties are selected and built up from the research project of Dupâquier and Kessler called 3000 Families, in which all descendents of couples married between 1803 and 1832 were traced, by a nationwide checking of all indexes on certificates of marriage, death and birth. This enormous effort at record-linkage was possible, because volunteers only had to look for surnames beginning with the letters TRA. One obvious bias, properly discussed by the author, is the truncation of female lineages. Paul-André Rosental has concentrated his research on families from comparable villages in a number of rural areas, who actually had enough descendants to make a long term study of the family possible. This means that they had to have at least one male descendant who married after 1860. My first question has to do with the beginning of the period of observation. Initial families are made up of children marrying after 1803, who are linked to their parents, the so-called founding couples. The analyse is based on the occupations and residences of the members of the different branches of these dynasties, and their influence on residential choices of family members. The intensity of the influence depends on the strength of the intrafamily ties. However, it is not clear to what extent Rosental included in this analysis the brothers or sisters who happened to marry before 1803, and their descendants as well. The same goes for unmarried brothers and sisters. Are their wherabouts known other than from their appearance of the certificates of the core group? In other words, is it possible that migrational choices have been influenced by unobserved family members, other than the descendants on the female line?

My second question is about the representativeness of the final selection. Obviously, persons should not marry too late, in order not to run the risk of reduced fertility or even childlessness. Their families should not suffer too much from child mortality. And finally, they should be capable and willing to secure formal marriages and not resort to concubinage. To conclude, the analysis is based on relatively successful and large families, which makes it possible to give the element of interfamily networks such a prominent place in migration behaviour. The bias is not only demographic, but social as well. It seems to me that the poorer and more proletarian parts of the rural population are missing in these dynasties. Their members seem practically all peasants and craftsmen. I am aware that propertyless workers were not very numerous in the French countryside, but in this story they seem to be lacking entirely. A final problem has to do with proletarian concubinage. As is well known, proletarian migrants from the French countryside to Paris and other cities often lived in consensual unions. One of their major problems seems to have been the expenses and the administrative wheeling and dealing involved in an official marriage. Many documents from the place of birth had to be procured, which proved too great a difficulty for many. To conclude, because Rosental’s migrations analysis depends so heavily on official marriage, it is possible that many long distance moves have been missed and that entire families were therefore dropped from the selection.

My second point of discussion deals with the measurement of premarital mobility. This may seem a bit technical to you, but I would like to emphasize here that in this book methodology takes central place. The author claims to have offered a new perspective on the analysis of migration, and therefore I like to take a closer look at his innovations. Rosentals criticizes conventional approaches to the study of migration for what he calls their “ponctiformité”, their tendency to reduce mobility to just one movement from one place to another, in just one moment in time. Only the attributes of this particular move, these particular places and this particular moment are studies and related to one another. Instead, Rosental himself looks are individual trajectories, that he builds up from the civil registers, the certificates of marriage, death and birth. Mobility between birth and marriage play a part on many levels throughout the book. They are used to specify the research topics in the first chapters, that is firstly, the slow increase in long distance mobi-lity and, secondly the shift in direction from rural destinations to the Paris area. All this is based on premarital mobility. Of course, these figures are not very precise, as they are influenced by migration of families with children, by individual mobility of the still unmarried and even by the widening of the marriage market. Further on in the book, mobility or sedentarity between birth and marriage is related to indivi-dual position within sibling sets and, even more important, to the histories and relations between respective branches of dynasties.

In my opinion, by concentrating so heavily on the residence at marriage, Rosental has fallen into a reductionist trap of his own making. I would like to demonstrate this by making a comparison with the Netherlands in the second half of the xix^th century. In 1810, the Dutch were forced to use the French system of civil registration. However, this proved to be such an improvement, that they decided to keep the system after the defeat of Napoleon. This means that we have basically the same laws and regulations at to the registration of marriages, births and deaths. As far as I can see, the marriage certificates contain the same information which means that in principle a similar re-search on migration could be done. However, in the 1850's another system was adopted, alongside the civil registration, which is known as the population register. These registers are truly dynamic and record all change of adress, civil status and so on of all the inhabitants. I have compared a mobility analysis on the basis of marriage certificates to one on the basis of population registers. I have made use of material from the Historical Sample of the Netherlands, which is a project of life course reconstruction starting with a sample from the birth certificates. The results are in table 1.

Tab. 1
Mobility between residence at birth and residence at marriage. Marriage certificates and population registers compared
(N=108 women born in villages and small towns in a central Dutch province, sampled from birth certificates 1853-1882)

Marriage certificates	Population registers
Pre-marital mobility= 30%	Pre-marital mobility=65%
76 women living at marriage in the place of their birth	38 women truly sedentary38 had experienced migration(s), but returned
32 women lived in another place	13 had made just one move19 made more than one move
Post-marital mobility: 40% of the brides left their place of residence within two months after the wedding.

Many of these girls had worked as domestic servants, and not just in neighbouring communities. Quite a few of them had lived in one or several cities before returning home. Chain migration was not of particular importance in these moves. At least, these girls seldom lived with relatives. What are the implications from these findings? First, we know that many girls had experiences of their own outside their family territory well before marriage. They could bring their connections and information into their marriage and in this way contribute to decisions about future migrations. Secondly, the choice of a residence at marriage should not be given too much importance. We can see them as one in a sequence of temporary moves. As we have seen, many of the women had lived in another place before they married. Also, 40% of them left immediately or soon after the wedding. Thirdly, instead of asking ourselves why people were sedentary, we should ask why unmarried youngsters returned so often to their parents. In fact, in concentrating on the extended family, Rosental seems to neglect the mechanisms inside nuclear families, for instance the residential choices surrounding the care of the elderly. To conclude, I am obviously not reproaching Rosental for not having population registers and therefore mis-sing a number of moves. However, his concentration of the place of living at marriage in relation to the presence of kin leaves little room for connections and experiences other than those having to do with the family.

My third point is still more metho-dological. In order to assess the intensity of family ties, Rosental makes extensive use of the witnesses on marriage certificates. Apparently, little is known on the role of these witnesses in marriage ceremonies, because no literature on the subject is cited. Rosental divided these witnesses into four groups, those that are kin, those that may be kin because of the similarity of family names, those that are not related but shared by family members in successive weddings and finally those that are simply non-kin and not shared. He uses these witnesses in a real formula that makes it possible to distinguish between autocentric families on the one hand, exocentric families on the other. This distinction then plays a major parts in his elaboration of hypotheses on migration.

These are all very interesting and plausible hypotheses. However, I cannot help thinking that there is some circular reasoning involved here. In my view, migration and the recruitment of witnesses are not independent variables. To put it simply, migrants will be less likely to attract family members to their new residence to be a witness than sedentary persons, therefore they will appear as exocentric. Exocentrism, however, is seen here as the cause of migration, not as its effect. To demonstrate my point, I have analyzed a number of marriage certificates myself, in particular the witnesses on the certificates. Before I do this, I must emphasize that although Dutch marriage certificates were very similar to the French, this is not to say that marriage ceremonies were the same or had the same meaning to those involved. Still, the following exercise might be instructive. I have analyzed 300 mar-riages certificates from one ordinary Dutch village, in the period of 1830 to 1880. They are all certificates of first marriages of locally resident grooms. Dutch civil law required the presence of four witnesses, so we have information on almost 1200 witnesses. On closer inspection, however, they prove to be no more than 700 individuals. First, there is number of what we could call “professional witnesses”. Apparently, signing a certificate was one of the jobs of the local constable. Second, many witnesses who had come for one particular wedding, were asked to sign the certificate for the next one as well. This probably happened when several marriage were secured in the town hall at the same time. The relatives of one couple were helping others who did not bring enough witnesses of themselves. In other words, sharing of witnesses was very common, but not in a Rosentalian sense. Therefore, it is not surprising that only 40% of the witness were relatives of the bride or groom, the other one being non-related. Comparison of their surnames does not reveal any hidden kin-relationship. I now come to my point, and that is that recruiting relatives as witnesses was inversely related to migration.

Tab. 2
Recruitment of marriage witnesses by migration status
(N certificates=298, municipality of Akersloot, North Holland, 1830-1879, first marriages of grooms living in Akersloot)

	Witnesses
	Not related	Relatives
Grooms	Mean number	% resident in Akersloot	Mean number	% resident in Akersloot
Local born (N=152)	2,13	76	1,05	68
Born elsewhere (N=146)	2,76	78	0,60	34
Brides
Local born and resident (N=101)	2,26	77	0,94	71
Born elsewhere but resident (N=123)	2,50	77	0,64	32
Born elsewhere, not resident (N=68)	2,65	81	0,56	5

To summarize the table, locally born men recruited most family members, whereas immigrants tended to lean on non-related, local witness (men and women). To be sure, there is no long distance migration involved here. My exercise could of course point in the direction of cultural differences between the French and Dutch as to the value attached to civil marriage.

The Dutch had a separate church wedding afterwards and I do not know about this in France. However, Rosentals calculations and reasoning are based on the assumption that persons from families with strong internal ties would witness for one another not withstanding the distance. I think this assumption should be made more explicit and tested with a closer analysis of the phenomenon of witnessing.

Finally, some last comments. Basically, my remarks have come down to the same thing. In his admirable effort to shift attention to family relations, Rosental is putting too much emphasis on their importance as an independent variable. His selection of his research population and his choice of methods tend to overestimate the impact of family factors. First, he has concentrated his research on demographically successful families and secondly he concentrates on moves that are geared to family transitions, marriages, births and deaths. He links these movements to relation with wider kin. However, many other moves and motives are not observed. Thirdly, he suggests that maintenance of kin relations (by witnessing) is not altered by migration. Fourthly, macro developments leading to increases in migration are described rather roughly. In fact, the only macro development involved seems to be the rise of the state and the diversification of occupations, that incited individuals to move depending on the information and opportunies mediated by their families. The gradual increase of long distance migration to Paris was a cumulative effect of this development. But how about other factors, such as the role of literacy of that of conscription in ope-ning up new horizons. Finally, changes in the social meaning of family networks themselves are not discussed. Rising standards of living, changes in legislation, cultural changes, changes in welfare arrangements may all have worked to make integration in family networks less important for individual survival. In time, maintaining the family network may have become less urgent and so many family would become exocentric.