version 20/5/92, June 92

DRAFT. From Paul Stirling, University of Kent at Canterbury.

Two Turkish Villages and their Emigrants: 1950-86.
Comments and Advice
Paul Stirling

This database owes its existence to two ESRC grants, "Thirty-five Years of Transformation in Central Anatolia", G00 232121 (1985-7), and "Longitudinal Study: Turkish Villagers and their Emigrants", R000 23 1951 (1989-90). I have received help from Dr. Emine Onaran Incirlioglu, Mr. Mehmet Arikan, and many others in the field; and from Dr. Michael Fischer, Dr. Nick Ryan and other colleagues at Kent in the organisation of the data and the use of computers.


1.An Evolving Project
2. Fieldwork, Ethnography and Data Collection
3. Sampling and Coverage
4. Inserting and Organising: A Dilemma
5. Personal and Household Codes
6. Main Definitions and Comments: Individual life Histories
7. Main Definitions and Comments: Households

NOTE: Every user table and column is listed in the Data Dictionary which follows, with brief explanations where necessary.

1. An Evolving Project, 1949 - 1990

In the 1970s, I found myself with a lot of detailed information about two villages of Anatolian Turkey gathered in 1949/52 and 1971, by no means all of it published. [. Stirling 1965, Stirling 1974.] My note books seemed to me unique historical sources, worth making available to future scholars. I conceived the ambition to return to the villages to attempt another full study of the two villages and the emigrants from them. Turkey had undergone vast changes, - in effect, an industrial revolution - since 1949. An anthropological parable of the greatest changes in human history?

I had two difficulties. First, I could not, and still do not accept a single `theory' in which to set the project. [.Stirli 1951, Stirli 1965, Stirli 1974.] Second, I have never rid myself of the idea that an ethnographer has an altruistic duty to record and bequeath as much information as he can, whether it seems relevant to his own, necessarily ephemeral, current interest or not. `Everything that happens in the village concerns you' said Meyer Fortes to me in Oxford in 1948. And I also learned that asking direct questions, while absolutely essential, is unreliable, and less valuable than watching, and listening to what people say spontaneously. Such ideas impose impossible standards and thus a sense of continuous failure.

Inspired by the example of Laslett and the Cambridge Centre for the Study of History, Demography and Social Structure, and by the growing interest in detailed local demographic histories all over the world, I conceived the idea of providing others with an `archive', a body of data for future comparative work both in Turkey, and between different countries. [.Duben Istanbul.]. It seemed a modest and not too difficult aim. And it has - almost? - come to pass.

In practice, it has not turned out to be modest. Perhaps due to my own inexperience in this kind of work, perhaps to trying to do too much, it has become something of a nightmare.

In the field, I had the illusion that I was planning for putting the data into a computer, and that a computer could help me solve the immense difficulties of trying to define and thus quantify what I saw going on around me. I now know that it is essential to get the data orderly and quantifiable before going near computers; and that it is advisable to be skilled in computer programming before starting on databases.

Moreover, my conviction that an ethnographer should be ready for surprises, and record all he learns from listening and watching people is hard to combine with eliciting preplanned ostensibly hard specific information on a comprehensive range of topics.

I returned from the field in 1986 with a large amount of messy data, and tried to get it typed in to the university main frame, about which I then knew nothing. Two major errors; I should first have spent months learning about computers; second, I should have spent months tidying up the data and facing the complexities of making it rationally countable. Because I did not do either, this DB is less useful and user friendly than I hoped. I still have much detail in various note books which could, and I still intend will, enhance it.

2. Fieldwork and Ethnography

My wife and I went to do anthropological field work in Turkey in March 1949. From November 1949 to August 1950, with interruptions, I, and later both of us, lived in S, a village 30 kilometers east of Kayseri on the road to Tomarza. We returned to Oxford where I wrote a thesis, and then returned to Turkey. We spent July to December in E, a larger village in the county (ilce) of Bunyan, a further thirty kilometres east of S, and 5 kilometers off the main road from Kayseri to Malatya. In 1952, I spent my summer vacation in E.

I paid a brief visit to both villages in 1955; and then not again till another brief visit in 1970. On this occasion, the changes struck me forcibly. In 1971, I arranged to be in Turkey for two months, hoping to do serious field work. But Turkey was under military government, and I was finally given research permission only for one week, though I managed to extend it to two. I was, by law, allocated an official 'guide', Mr. Ali Cayirdag, who was extremely helpful. [. Stirli 1974.]

From 1974 on, I visited Turkey for various professional reasons every year except 1977, and paid brief visits to the villages when I could. From 1983 to 1986, I was Visiting Professor at the Middle East Technical University in Ankara. I held official research permission from July - October 1985, when it was suddenly rescinded without any explanation; and from January to September 1986. I was explicitly asked to combine research with my teaching duties in Ankara, which made life more difficult.

I returned to Turkey for six weeks, June-July 1989, and for six weeks, April-June, 1990, to check on and expand the 1986 data, mainly in the towns.

Mr. Mehmet Arikan, graduate student in Sociology at the METU, was my Research Officer from July 1985 to September 1986, and again in England from December 1986 to July 1987.

Dr. Emine Onaran Incirlioglu was my Research Officer from January 1986 to August 1986, and again in June 1989. We continue to co-operate. Her thesis at the University of Florida was based on her field work with me, and I was a member of her jury.

Our official guide in 1986, Mr. Vahap Tastan, then Research Assistant at the Faculty of Theology in Erciyas University, helped greatly with field work in the summer of 1986. Ms. Hulya Demirdirek also contributed to the research team in July and August 1986.

The data were collected direct from informants from the two villages; but often by asking people about other people. Much of the information on migrants is from village kin or neighbours, but I and my colleagues met many migrants in the village, and we also went to Turkish cities, and I briefly to Pforzheim in Germany.

In 1950 and 1951-2, my wife and I collected a simple census on every household in each village. I also collected genealogies in rough note books.

In 1971, I had only three days in each village, and with the help of Mr. Ali Cayirdag, I was able to ask the constantly changing room full of people who came to greet me about every person listed in my 1950 survey of S, and about every person in every third household in the 1951 survey of E. ( See below ).

In 1985, in Ankara, I devised and had printed two sets of cards, one for individuals, one for households. (See Appendix 4.) My plan was not to use these openly and systematically in interviews, but to carry a mnemonic and a rough note book, and to ask as much information as seemed compatible with good manners and the social situation of an encounter. We would then fill the cards in from the rough notes back in our office. I intended that we should consciously remain receptive to other information, and not limit informants to the questions on the card, nor attempt to make what they told us fit into the layout on the cards.

We began in March 1986 visiting S regularly by car from Talas, about 25 minutes drive. But it became clear that this method was clerically inefficient, and we were soon far behind schedule. Completing the writing up of one visit before the next was often difficult. I did not wish to abandon the informal friendliness of our village visits; nor could we have done. We then devised a simpler, improved schedule of questions (see Appendix 4) , the responses to which, with the help of mnemonics, we wrote directly on to blank A4 pads. This was vastly quicker, but postponed many problems to a later stage. Even so, we were running late, and because of the rigidities of grants and bureaucracies, we were not able to extend the field work.

We moved to E late.We were able to reside in the village, because of the extraordinary kindness of Vahap Tastan, my official guide, and his wife Metiye; but only for a month. We only had time to collect data on just over half the 1971 sample of households.

Besides these relatively systematic notes on households and individuals, we all kept personal notes of all events, on a wide range of topics. Many of these contain information relevant to the DB, which I plan in due course to insert.

3. Sampling and Coverage

The Table shows the total number of households and of individuals, by village and by household migration in the database.
Village and Migrant Households and Populations
Village ObservationVillageMigrant

In 1949-50, In S, we collected information on every household. I decided the range of topics ad hoc; `common sense'. We visited most households, or asked people as we encountered them. In a few cases, we asked close neighbours or kin. In 1951, we did the same for E; but with less time and twice as many households, we were less thorough, and used more hearsay evidence.

In 1971, I and my official guide, who willingly acted as scribe, had only three days in each village. In each, I sat in a guest room, and people flooded in to greet me. Amid the tea, ayran and general hubub, I simply went through the 1950/51 censuses asking about each individual listed, including information about migration and occupations. In S, I was able to cover the whole village. In E, for lack of time, I selected on the spot every third household from my 1951 list, which were arranged roughly in alphabetical order.

Plainly, this was not a reliable method. But with several people present, most of it was checked on the spot; and by the use of the 1950/51 and the 1986 data, we have been able further to check and augment it. Some of it remains well informed guesses.

In 1986, I set out both to be much more systematic, and to cover an even wider range of topics. In the event, I and the team had less time in the field than I planned; and because of the informality, the range of our interests, and the conditions of work, the survey went slowly. It was not easy to get to the towns to collect first hand information from migrants; and even in S, we ran late and left one or two village households incomplete. These we completed in 1989. All the same, I reckon I know about virtually all the emigrants from S. I may have missed one or two, and for some migrant households, the data is scanty.

As I have said, in E in July 1986, with only one month, we halved the 1971 sample. We did this by selecting every alternative household from the 1971 list. This was a mistake. On this list, where two or more households were derived from one household on the 1951 list, they were listed contiguously. So in such cases, this method of selection eliminated approximately half of them. For example, if a man who headed a household in 1951 had died, and three of his sons had established new households, then at least one of the three households, and perhaps two would inevitably be omitted from the 1986 sample list. But one important set of questions is precisely the generation of new households from the old ones in the 1951 list. We should have gone back to the sample list of 1951 household heads, selected every other name from that list, and then included all the patrilineal descendants of that household in the sample.

When, too late, I realised this mistake, I attempted to add to the 1971 list used for constructing the 1986 sample, all patrilineally collateral households of households already on the list. We were not able to this in all cases. The effect on the 'randomness' of the sample is probably very small; but it does introduce some bias in favour of prolific households. (Appendix 2)

My 1971 and 1986 sampling procedures then divide the 1951 E households into three kinds. On two thirds, we collected no information after 1952. One third formed the basis of the 1971 sample. Of this sample, slightly more than half, just over one sixth of the original 1951 total, formed the basis of the revised 1986 sample. The key table for distinguishing these three kinds of E households is called 'iobs'. ( See Data Dictionary below.)

My ad hoc method of sampling produces another minute bias. One household head (pcode 1185, hh_code EL5V7067), whose name came up in the sample in the village in 1971, had died, leaving daughters who had married to town.

Certain consequences follow from the fact that the 1986 E survey is only a a sample of a sample. First, in S, I listed all persons mentioned in the 1950 and 1971 surveys whose names did not occur in later surveys, and asked about them. Some were dead, some were migrants. But in E, I never made such a list. So some migrants from the E 1986 sample may have been missed. Second, where persons in the sample have married E villagers who do not appear in the sample, we have two complications. If a daughter of a man in the sample marries an E man who is not, she is recorded neither as a married woman in the sample, nor as married out to another place. If a son of a man in the sample marries a daughter of a man who is not, she comes back into the sample although her natal household is not in it. This may affect certain counts.

4. Inserting and Organising: A Dilemma

In late 1986, I did not know where or how to begin. On Dr.Fischer's advice, we began by getting computer operators to type in data direct from the cards which we had used in the field. The household cards (sample attached) had been filled in by clerical assistants in Ankara in 1985 from my old note books, and then used in the field until mid June 1986. A set of individual cards (sample attached), with no details, had also been created in Ankara, by clerical assistants simply writing in names.These had proved more difficult to use in the field. By 1986, only a few of them were reasonably complete, and many had not been used at all. As I explained above, we had completed the research in S, after mid June, by using a schedule of questions (copy attached) and A4 pads, abandoning any further use of cards. Thus the 1986 data for some S households and all E households was on these unedited A4 sheets, and could not be used by computer operators. This data was inserted by Mr. Mehmet Arikan, who knew the research and the people, and by me, into files matching the household files. Individual details could not be entered for want of a suitable individual files. This work was slow; neither of us were skilled typists, nor familiar with computers. The A4 pads were handwritten and messy; and some had not been checked over in the field.

In fact, with hindsight, I now see that we faced three dilemmas. The first was whether to go ahead at once, or to take time first to correct the cards and the A4 record sheets. I made the wrong choice without recognising that I had one to make.

Second, what form should the files takes? With such messy data, and with many problems of definition still ahead, the use of any of the ready made database formats then available was out of the question. I needed to remain firmly in charge, with the ability to change the arrangements as we gained experience. Dr. Fischer assumed that I would become fairly quickly and easily an accomplished computer programmer able to manipulate my own files as I wished. This I have failed to do, and thus I remain dependent on more expert colleagues.

We created the data files simply by using the word processor, and using a space as a separator between records. This meant that I - and Mehmet - had constantly to remember which record we were dealing with. It was easy to make not only occasional miscounts, but even systematic errors. We worked on the household files, which were more or less inserted by late 1987, for more than another year, correcting errors, rearranging layout, and adding data from my various field notes, and from field work in 1989. The records of this process have been archived at Kent. In all this time, we did not manage to create an easily usable individual life history format into which to insert the numerous relevant items of information from the interview schedules and field notes.

The third dilemma only sank in slowly. Am I dealing with three separate one off surveys, or with continuous historical processes? From the beginning, I had intended the second, but not realised the full implications, still less the problems of the representation of time in databases. [.Ryan 1990.] Life histories cover all the main events for each individual; birth, marriage , children and death, residences, education, occupation, employment, migration, assets. These occur over time, and differently for each person. Years, or at least the sequences of events, are essential. In most cases these dates were not entered, or even known. But it is possible to work out reasonable guesses for most of them, giving ranges where necessary. Another very large task.

All this became clearer after, in the autumn of 1989, I accepted an offer from Dr.Ryan to build a relational database for the whole body of information, using the latest Ingres DBMS, which UKC had just acquired.

Three comments. First, it took a lot of time to go through and discuss the whole data set again from the new angle, and to write the complex software necessary. Second, precisely because this was a solution to the absence of usable individual life history files, a mass of data still remained to be entered, from rather unsystematic sources; this still has not been finished. Corrections and changes to a complex data set are often difficult.

Third, the problem of individual life histories creates the problem of household histories. One major difficulty in studying households is that authors normally have to deal with surveys and censuses confined to single points in time, and thus not with the year by year, week by week fluctuations in household - or individual - affairs and fortunes. The problems of representing this constant flux on computers have not been solved. But I do have more data, and we can claim that this project has made substatial progress with the problems both of time and of messy complexity.

5. Personal and Household Codes

In 1985, I had had little experience of survey research, and none at all of working with computers. So I devised solutions to problems as they arose. One misguided aim, which caused a lot of inconvenience later, was to avoid clerical repetition and redundancy by ingenuity in combining information in symbols, instead of breaking it down into minimal units.

Personal Codes.

Every person or individual in the DB must have a personal code, or 'pcode'. In 1985, lists of persons were typed out from the 1950/51 and 1971 lists, and allocated 4 digit codes. At this point, children under fifteen were excluded. We used the following scheme:

S males 0001 - 0999
S females 5001 - 5999
E males 1001 - 2999
E females 6001 - 7999.
In 1986, each new person whom one of us came across was given a new code in numerical sequence by hand on a communal list available in our office, which all research workers shared. Clerical errors and duplication were all too easy. Later, in England, we generated pcodes for children by computer programming.
S boys 4001 - 4399
S girls 4400-4799
S unknown 4950 - 4999
S mixed 4670 - 4949
E boys 9001 - 9399
E girls 9400 - 9799
E unknown 9940 - 9960
E mixed 9900 - 9916
In the field in 1989 and 1990, we did not have full records with us. So to avoid any danger of inadvertently duplicating pcodes, we gave new adult persons who came to light in our checking, pcodes in the form N000, and new children, C000. Later we made a decision to give C001 and on codes only to children of unknown sex, and changed other C codes accordingly.

In early 1990, when Dr Nick Ryan set up the relational database on Ingres, we discovered the need for partial information on some persons not in the above samples. These persons were needed to define the life histories or relationships of the main sample as parents, spouses or close kin. Since the Pcode is the sole definition of a person in the database, we had to give these auxiliary persons Pcodes. For this purpose we introduced three new kinds of Pcodes.

(i) D for Dead Persons.

Some persons who were already dead in 1949 are entered as D001 and on.

(ii) B for Bride codes

Some women born in S or E, but who left to marry before 1950, are entered as B001 and on.

(iii) S for Stranger codes

Some husbands of those women of S or E who married to other villages are entered as S001 and on.

(iv) M for Migrants before 1950

Persons whom I know left S as household migrants (see below) before 1950 are entered as M001 and on.

Persons with D, B, or S pcodes represent no universe, and should in no circumstances be counted. Persons with M pcodes have been given as much detail as is known. I have visited some of them. I am sure there were very few, but also sure I did not find out about all of them. But they might with caution be counted for certain purposes. I do not have this information for E.

Household Codes

Households have two kinds of codes, which arose in the development of the research. Both remain necessary. In the field in 1986, I coded the three surveys, 1950/1, 1971 and 1986, separately. Every household in each has an 8 character code, `hh_code'; e.g. SK8VF656, EL7M7438. These are a case of my frustrated ingenuity; against the principle that items of information should be minimal and distinct.

SK or EL indicates the village.

5,7,or 8 indicates the survey, or yr_obs; 1950/1 or 1971 or 1986.

V or M indicates village household or migrant household.

In S, the next single capital letter ( the 5th character) indicates the patrilineal lineage of the household head within the village, according to my 1950 information. Although Eh as a similar lineage organisation, my information on it was partial, and was not included in the surveys of 1971 and 1986. So all E hhcodes have a redundant 7 at this point, to keep the pattern uniform. It would be possible to recover lineage information for most E households from my notes.

The last three digits are a unique number for each hshd. They also indicate the year of observation and village or migrant status.


001 - village 1950
200 - village 1971
400 - migrant 1971
500 - village 1986
700 - migrant 1986


001 - village 1951
300 - village 1971
400 - migrant 1971
500 - village 1986
700 - migrant 1986

Household Numbers

The construction of the database brought out a problem. The data covers the history of households through time, but the household codes do not show continuity between the three surveys. Moreover, we do in fact know quite a lot about events between the three surveys. We therefore decided to define households (see Definition of households) by their heads. Thus defined, a household does not only exist in a survey, but as an entity continuing through time which may appear in up to three surveys. Since the household is defined by its head, we used the Head's pcode as a code for the household over time, and called this code household number, hh_no. So hh_no = household head's pcode.

6. Main Definitions and Comments: Individual Life Histories

Words are notoriously slippery, and boundaries almost always fuzzy. The totals on which the statistics are based vary not only with definitions, but with interpretations of definitions. Many definitions are controversial; different researchers in the team apply rules differently, or forget them; and over a period of eight months, let alone thirty five years, persons change their ideas. So all totals are approximate, and sometimes inconsistent with other totals.

I discuss here the important, controversial or unorthodox definitions, concerned with individuals. I discuss those concerned with households in the next section. The Data Dictionary gives all other definitions.

Individual An individual is a person who appears in the database. This raises two points.

First, there are two sets of pcodes (see 5. above), and thus two kinds of individuals in the databases: main individuals, whose life histories are given, and who belong to households in the samples; and auxiliary individuals, who are in the database solely to provide information about kin relations between main individuals. For auxiliaries, information is minimal, and they represent no kind of universe. Their presence requires vigilance to exclude them from any counting by programming. Those whom I know left S before 1950 are anomalous, but belong with main rather than auxiliary individuals, because they are in the data base for their own intrinsic reasons. It might be useful later to add all dead persons from all my genealogical information. (See 5.Coding).

Second, at what age does an infant become an individual? Two related questions. What is worth calling a life history? What are users of the database likely to want to count with moderate confidence? Briefly, large numbers of the pregnancies which occurred failed to produce a viable young adult. ( see below, Birth, Years) It was normally difficult to find out when a child died, and indeed, how many infant deaths had occurred. Roughly, if we know a child lived for more than one year, or if a mother uses a personal name, then we allocate a pcode. But the criteria depend on the chance circumstances of interviewing, on the sources of information, and on the personality of the mother.

Lineage In 1950/51, both villages contained explicit and active patrilineal groups. [.Stirli 1965{chap 8,11}.] As usual in such patrilineal societies, people gave slightly different accounts of the system, and large lineages had named segments. But I was able to allocate every S household to a lineage, though not all households were equally interested in their lineage affiliation. Thus, it was possible to allocate every individual born in S to a lineage, through their recognised father. I hold that lineages were already declining in importance in 1950, and they had certainly declined much further by 1986. Enmity between lineages as such was not dead, and interlineage violence had occurred in both villages in recent years. But by and large lineage groups were barely noticeable, though they do retain some significance for some people. Certainly patriliny remains important for households, for economic co-operation, and for marriage.

My data on lineages in E was less complete, and I did not feel confident about allocating all households and individuals to lineages. I decided that to attempt to ask questions about lineage affiliation in 1986 would provoke time wasting discussion and confusion on a topic which seemed merely of a kind of antiquarian interest to the villagers. So I omitted lineage affiliation from the E data. In many ways I now regret this decision.

Years Years, that is, dates, are seldom fully reliable. People are not normally accurate about years. Questions about the year of a birth, a marriage, a house move, starting work, meet with three kinds of response: uncertainty or ignorance ("am I my mother that I should know when I was born?"); a discussion involving related events; or a definite date, age or distance in years. Sometimes, when informants give such specific answers, they know; sometimes they are saving face, time or bother. When, as often, we were asking about dates of events in other people's lives, the answers were even less reliable. Since people remember personal events by other events, a it is sometimes possible by long discussions to settle fairly accurately on dates. Where this was not possible, interviewers sometimes accepted doubtful estimates, sometimes guessed, and sometimes left a blank. And in many cases, questions about when things happened simply did not get asked.

In 1950/51, most people were illiterate. By 1986, virtually all the young had been to school, and some men and a very few women were highly sophisticated, running businesses or holding jobs which demanded writing and counting. So in 1986, we were able to get more accurate years, especially for recent events. But very often 1986 data from the old and middle aged conflicted with data from 1950/51 or 1971.

In many cases, we are able to use existing information, such as birth order, or knowledge about migration and marriages in families, to reconstruct years or ranges of possible years.

In the light of such a range of problems, we have wherever years are given for any purpose, adopted the following convention:
"1930" means either a reliable year as given by an informant, or a reliable estimate by a researcher.
"1930-35" means a range. Any year, including or between those given.
"1930+" means after 1929
"1930-" means before 1931
"1930?" means probably close to 1930
"?" implies doubt


In several tables information about time is recorded both in terms of age and in terms of year. These are logically equivalent, and recorded thus for convenience. Each is readily convertible to the other, but care may be necessary when extracting data about time.

Birth By and large, years of birth in the database are not very reliable. Four points. Most Turks do not celebrate birthdays, and the children did not on the whole talk about how old they are in the way the children I know in middle class Ankara do; nor indeed did adults. Second, I found that birth dates recorded on different occasions frequently differed, sometimes widely. Third, people usually give their age as the ordinal number of the current year, not as the number of years completed since birth. So a child which has completed one year is said to be two, not one. We did not systematically allow for this. Fourth, as I discuss below, births are not normally registered immediately.

Informants sometimes gave ages, rather than birth year. These were often approximate guesses, and in rounded numbers. Many adults, especially older ones, gave their birth years in the Ottoman calendar, that is, in 1300s, not 1900s. I have subtracted 16 from the last two digits in these cases; thus 1333 becomes 1917.

Registration Of course, all births must by law be registered within a brief period with the Registrar, or Population Officer of the Republic. But in practice, babies were usually registered late, from months to a year or two. Occasionally, a child simply took on the registration of a dead elder sibling of the same sex. In 1950/51, these practices were virtually universal, except for men whose jobs carried child allowance. Village headmen did not in practice register births, and to make the journey to the official Registrar was a trouble, and slightly risky, since not all adults were sure that their papers were entirely in order. Moreover, people waited to see whether the new child would survive. Some infant deaths were simply never registered. When they did finally register it, they reported it as newly born to avoid a fine. Many thought it an advantage for a son to go late to military service.

The unreliability of registered age is also illustrated by another practice. If a person's age is bureaucratically awkward, it can be changed. If the headman and two witnesses declare the registered birth date to be mistaken, a new birth date can be registered. This is not common, and may no longer be as easy as it was. But there were cases known to me in the villages where this had been done. We also knew of cases where people had no official existence at all, since they had never been registered. The military authorities are said to make sure they find out about all unregistered men. When people gave a precise year of birth, they were normally giving their registered age, as recorded on their official identity card. I deliberately set out to distinguish real birth year from registration, though in interviews this was often difficult, and in estimating impossible.

Such practices have of course declined steadily since 1951, and many children, especially in the cities are registered promptly. But I was surprised that late registration is still quite common, especially if registration involves trouble or problems, and does not involve a loss of a child allowance.


In almost all cases, the Year of Observation, yr_obs, is one of the three surveys. However, I also visited the villages briefly in 1955, 1970, and then every year except 1977 from 1974 to 1985. I also did 6 weeks field work in 1989, and again 1990. In a very few cases, I give one of these years as the year of observation.

Residence Residence is used arbitrarily, against common sense, linked to household membership. For women and most children, this raises few problems, but it does for migrant male workers, and for children studying away from home. A male household head is defined as resident wherever his household home is; a male non household head, wherever his household head's home is. Even though he lives and works in another place, in Turkey or abroad, for months or years, for this DB he remains resident in his recognised household. Equally, children attending schools out of reach of home are defined as resident in their parents', or household head's, household.

This definition corresponds to the way the villagers themselves normally define a man's responsibilities. But it does cause anomalies; and there is one exception to my own usage of 'residence'. First, the anomalies.

i: In a very few cases, unmarried young men who leave the village do not contribute to or acknowledge membership of the household which they have left. Such cases are difficult to discover, because they are normally a source of disgrace to both the emigrant and his household. Where we know of such cases, we define the man as resident in his urban home.

ii: In one or two cases in each village, a man openly supports two households. In these cases I have made an arbitrary decision, giving his residence as the household where he spends most of his time. ( see types of household below). This inevitably creates anomalies in the data base.

iii: In two or three cases, a pendular migrant has deserted his household, and ceased to send money or recognise responsibility. In these cases, I have defined any relevant marriage as ended by separation, and the place in which he currently lives as his residence.

The exception is the use of the word `residence' under education. When a child or young person attends a school or university away from his home, he has to live somewhere during term time. In this context I have regrettably used the word residence for this place. See below.

Place We have used as placenames in the database the names of other villages in the

Country We have used the names of nation states, including a redundant Turkey even where this is obvious. Thus to count the destinations of migrants, it is advisable to use both place and country.

Marriage Our informants define marriage simply by the Islamic religious ceremony, `imam nikahi', which confers formal propriety on cohabitation and on sexual activity. [.stirling 1965{p.209}.] In practice, all persons recognised as spouses are assumed to have gone through this ceremony, by proxy if not in person. Of course, both a formal wedding, and a secular state legal ceremony are accepted, but as evidence of the quasi-secret `imam nikah', rather than as conferring moral legitimacy in their own right. Neither is necessary.

The only borderline cases occur with separation. Formal legal divorce is rare, and in any case takes a long time. If a couple break up, the woman leaves her husband's home. She normally goes to her father, or a brother; or to some other kinsperson. If she goes to another man, religious divorce and remarriage are assumed to have taken place. But where she goes to her natal or some other home, it is not clear whether she may yet be reconciled and return, or whether the marriage is finished. Her husband's kin are likely to maintain that she is still his wife; but she may well remarry without more ado from her natal home. None of this has to be legal to be accepted and respectable. One result of this arrangement is that within the villages and their emigrants, there can be no unmarried cohabitation; an imam `nikahi' can easily be arranged, and is automatically assumed. Adultery is between persons living separately.

Of course, with growing education and more awareness of legal implications, the number of legal marriages and even of legal divorces is increasing all the time.

Divorce Marriages, whether legally registered with the State or not, may be terminated by separation. Normally, the woman leaves her husband's house. (See Marriage above). Such separation is ambiguous, since the wife may or may not return. Thus for our informants, there is in practice no state of formal separation.

But if a couple have been legally married, they may decide to get legally divorced in court; expensive and drawn out. Where we know of legal divorces, we have recorded them.

Literacy The formal criteria for calling a person literate are notoriously complex. We rely on what we knew, or on what they or others said. Some are self taught. Anyone with 3 or more years education is defined as literate, except for one or two cases, where we happen to know they were not.

"Migration for Education" * Some village children are sent to urban schools. They may stay with migrant kin, in commercial lodgings, or in hostels run by the State or by religious charities. S had relatively few of these, and the details are often incomplete in our notes.

Occupations* This very fundamental subject causes many problems, and the tables are not yet entirely satisfactory. The issues are discussed in an Appendix.

Migration* A migrant is a person who leaves one of the two villages, either in order to earn in another place, or as a member of a household of someone who earns, or seeks to earn, in another place; or a patrlineal descendent of such a person. This definition excludes women who leave S or E to marry a man not born in the research villages. It includes women who are not born members of one of the the two villages, but who marry husbands from the two villages, except those who are born aand remain in towns. In practice, we might also use the word migrant for people who move out of other villages, so that a woman who married out of the village to a man from another village who migrated would in general terms be a migrant, but is not counted as a migrant in this database.

I distinguish sharply between two kinds of migration, but I am not satisfied with the two words I am currently using to distinguish them. Pendular migrants are persons who retain the responsibilities of membership of a village household, and who send or take part of their earnings back to that household. Household migrants are heads or dependent members of households which have left the village and moved to a town. The movements of pendular migrants are recorded as part of individual life histories. The movements of household migrants are recorded partly under residence in individual life histories, and partly under household place. This weakens the close link for most men between pendular migration and household migration.

These definitions imply that women are not pendular migrants. So far this has been true of both villages. Up to 1986, none of the few trained professional women in E - teachers and a midwife - as far as I know, lived on their jobs unmarried and sent remittances to a village household.

There are 2 anomalous cases, in which a villager moved to another village on marriage. The number is insignificant. In theory, if a man becomes a farmer of his wife's land, then this is not migration in my DB. If, as one did, he moved ( on marriage to a close patrilineal kinswoman) nearer Kayseri, and then earned his living in Kayseri, then I call him and his household migrant.

Return In this context, return has several distinguishable meanings.

i Pendular migrants, both national and international, come home from time to time to rest, see their families, and conduct family business; or between jobs. These visits are often at harvest, and to a lesser extent during the winter. But men may come at any time for a whole variety of purposes, and it is impossible to keep or ask from informants any accurate record for measuring such movements. We do of course have a lot of general and anecdotal information on such comings and goings.

ii Men who migrated regularly for years may at some point stop doing so. This is not exactly return; and in any case, they may start up again. So again, we have anecdotes and interesting cases, but no measures.

iii Very rarely, households return to the village from Turkish towns, due to special circumstances.

iv Pendular migrants to Saudi Arabia may come home on leave, with permission, and even a commitment to return, sanctioned by money owed to them. But some come back terminally, either because they have failed to get permission to return, or because they do not wish to go there anymore. Equally, pendular migrants to Europe may also decide to stop going, or find their job terminated. Such men may then become pendular migrants within Turkey; a few may return to Saudi Arabia or Europe..

v In the literature on Turkish migration to Europe, `return' often implies that a household which has been living in Europe decides to move back to Turkey. In such cases, they may settle back in the village, or more likely, they may move to a Turkish town where they have used European earnings to acquire a house or a building site. Many attempt to start up a business of some kind with their savings.

vi Sometimes a man moves his household back to Turkey, but continues to earn in Europe as a pendular migrant.

While these six possibilities are clear in retrospect, we did not question all relevant persons systematically on these topics. So significant counting of kinds of returns is not satisfactory.

7. Main Definitions and Comments: Households

Household General definitions of households are not easy. But for this DB, a household is one or more persons with common rights in a domestic space - a home - , and sharing a basic income, basic assets, and basic consumption, especially cooked food. There are two kinds of marginal cases. First, a very small number of single persons with household affiliations: old persons claiming independence, and men who apparently do not contribute to the household from which they last separated. Secondly, married daughters living within the parental compound, but claiming to run an independent economy and cuisine. Sensible decisions in these rare cases are not a problem.

Head A married man who lives with his wife and any children whom either of them may have, apart from his father or elder brother, is always household head. The villagers normally define an old man in a farming household as head, however ill or senile. Very recently, it has become possible for old men to be left alone by their married children, and then to be taken in to a son or daughter's household, as a dependent. No case has yet occurred in our sample. So we continue to define all such senior men as household heads.

Where a wife of a household head is widowed, however, a problem may arise. If she has no sons, and does not remarry, she remains household head. If she has young sons, she also remains household head until they grow up. But at what stage does she relinquish the headship to a son? A married son will certainly be treated formally as head by other men, including officials. But a strong woman may continue to give orders, and describe herself as household head. In at least two cases, a woman plainly dominated her sons and their wives, and ran the household economy. The team disagreed about these cases. I decided that for statistical reasons it was simpler not to introduce our own estimate of a woman's power within the household. So all men of 18 years or above at the time of the survey in question, who are the oldest male in their household, are defined in the database as Head, regardless of their actual relation with their mother, where she is still alive. This in no way should be taken to mean that in practice they give all the orders. But it does conform to the official and public, and male, assumptions of Turkish society; and not all household heads give all the orders anyway.

Member The members of a given household are all current resident. Residents include pendular migrants; all men who contribute to its income - heads, heads' sons, and other kin - even when these are away working for months or even years. Borderline cases occur when a man has not been in touch for a long period, but these are few, and often easy to decide.

In Turkey, households sometimes contain non kin strangers who are explicitly there as servants or as agricultural labourers. Several cases had occurred in these villages in the recent past, but none occur in the sample. Had they done so, I would probably have counted them as household members, but given them a label making their exclusion or inclusion easy for statistical purposes.

Type Classifying households by type is notoriously difficult. In the end, I decided on a classification into only 5 types. It is not difficult for users to construct their own more complex classifications from the database.

i F Fragmentary: A household with no married couple.

ii S Simple: A married couple, with or without the children of their union .

iii S+ Simple Plus: A married couple, with or without the children of their union, and other persons.

iv M Multiple: Two or more married couples, linked by a close kin tie, with or without the children of any adult member.

v M+ Multiple PLus: Two or more married couples, linked by kinship, with other close kin, with or without the children of any adult spouses of the couples.

Property We have no one heading for property. Information occurs under land, machinery, animals, home, assets, business assets. With a very small number of exceptions, every household has some property other than clothes and chattels. Property takes many forms, and it is not possible to find out by simple questions who owns what, and how much it is worth. Perhaps by a carefully planned specific enquiry on this point on a small sample, reliable indicators of personal and household wealth might be established; difficult and time consuming. I contemplated using a general subjective estimate, say on a five point scale, but when we tried it out, I realised it was morally unpleasant as well as highly unreliable.

Land My teachers warned me in 1948 that land is both central and extremely sensitive; never a subject for simple truth. The first difficulty was the unit of measurement, which remains unresolved.The unit of all land estimates in the database is decares. The word \fI donum \fP means the amount of land ploughed by one pair of oxen in one day. In practice, villages had fairly standard ideas of how much this is. But estimates differed from village to village and from person to person. [.Stirli 1965 {p.52}.] All agreed that a village was between two and three times a 'government' donum, which is fixed at one decare, about one quarter of an acre. At the time, no educated or official person whom I met in Turkey seemed to be aware of the implication of the discrepancy between village donums and government donums. By 1986, everyone was clear about the difference, and usually put the relation at 3 government donum to 1 village donum. I suspect that my 1950 estimate was more accurate than the new rule of thumb.

The second difficulty arises from the village system of fallow. In each village, territory was, and still is, divided into two halves. Each household owned land on both sides. The two sides were cultivated and fallowed alternately, and the side which was not under cultivation was used as pasture for the villages herds and flocks.

I simply asked how much land a household owned, without attempting to go into details, beyond checking that the answer included both sides. The question was sensitive. Sometimes, at first, I asked publicly, and men did not want others to know how much land they wanted me to think they owned. Even in private, many men did not want to give a precise answer; and I am convinced that many, though they knew of course what they ploughed, genuinely did not have a ready answer in numbers of donums. Why should they? Moreover, the question is not clear. Is it how much you ploughed this year and last year? This varies with small scale sharecropping and renting; and some marginal land may be left uncultivated. Or does it mean how much land does your household own? This immediately raises a host of questions about rights to land, and within the household who owns what and what claims have others to it? I was taught - rightly I think - that to start measuring land was unwise; and in any case how could I? So I am well aware that the detailed 1950/1 answers about land are unreliable. And more so in E than in S. All the same, I was in the village for months, I talked about this issue in many conversations, and while specific details are unreliable, the general picture is not. {Stirling 1965 p.53}

In 1971 I asked nothing about land holdings.

By 1986, the situation was even more complex, because many sons in the village were separate from their farming fathers, and many household migrants had legal rights to land, which they were not exercising, might never exercise, and indeed might never even claim. Moreover, both villages had been visited by officials of the State Cadastral Survey, who had mapped and fixed all holdings, often by simply listing joint owners' complicated fractional rights to land which had not yet been divided, and was still farmed by one or two households who technically did not own it. Some land was registered to dead persons. Some informants in 1986 were sophisticated about the complexities. The data we have on land is sometimes inconsistent with older data; and sometimes there is doubt about what units are being used, and what question the informant thought he was answering, or avoiding. Indeed, in a few cases, we were unable to ask questions about land holding at all. I have used my general knowledge of the villages and households to estimate amounts of land. I plan in due course to revise the land data carefully, using notes, memory, and commonsense.

Thus the estimates of land holdings, whether given in numbers or in verbal estimates, are either reports of replies by persons who should know, or sensible guesses of the amount of land a household had available at the time of each survey.

I hold a cadastral survey for S, with maps. So far I have not found time to attempt to use this or to relate it to my data.

"land use" What do households do with their land? They may simply work the land they claim to own, using their family labour and resources. But they may also give it to tenants or sharecroppers, leave it fallow, or lend it. Thus in this context 'land use' does not refer to techniques or to types of crop, but to the arrangements under which it is or is not cultivated. See Appendix.

tenure Tenure is also used in an unusual way. It tells us what rights if any a household has, or claims, over land. See Appendix.

machinery Does the household own a tractor; or a half share in one? All tractor owners own equipment to go with it; the list can be long, and we did not record their details in more than a few cases. ( Thus Other is redundant.)

animals In 1950/51, in S, I asked every household about animals, which were an almost indispensable part of a household's economy. In E, this information was a little less reliable. In 1971, I asked no questions about animals, except incidentally. In 1986, we intended to be systematic, but were not able to be so in all households.

Home Home is the built domestic space in which the main part of a household lives.

type In this context, 'type' is the kind of building, and which forms the home, and its legal status. See Data Dictionary below. codes.

tenure What kinds of rights does the household have to its home? In some cases only, we have information about dates of acquisition or moving in.

Assets My ambition to record and include measurable information on assets proved unachievable. But we do have quite a lot of information. Village assets are to be found under home and agriculture. A small number of villagers own two village houses. Otherwise, the assets recorded are urban. These take three main forms, buildings, building land, and businesses. Some people have cash in banks, but apparently few, and not much. One man has bought agricultural land in eastern Turkey.

With rapid inflation, cash values of assets increase rapidly. With rapid urban growth, urban land values increase faster than inflation. So far, investment in urban real estate has invariably proved profitable.

Business Many households have succeeded in accumulating capital. Many informants insist on their preference for independence, for running their own lives, and not taking orders from an employer. Some succeed, doing everything from running back street shops on a shoe string to running profitable businesses. From my definition, I exclude subcontracting by building craftsmen; but include building contractors.

Incompleteness. My note books include a lot of casual data relevant to occupations, assets, and businesses which is not yet entered in the database. I expect to update the database at regular intervals so long as I am working on writing up, and remain intellectually competent.