In this thesis I have asked "Do anthropological models ever work?" In a rather simplistic and general sense historians are often reluctant to generalise as much and construct comparative models to the same extent, sociologists tend to operate at a more macro level regarding complex extended nation state societies and economists like to focus on the 'economy' often at an aggregated macro level using governmentally derived statistics. So "Do anthropological techniques - that is micro level analyses over an extended period combined with a comparative intentionality, when analysed, offer any other insights into 'social systems?'"
Stirling's work, based as it is over many years is for all the reasons outlined in the above, a good case study to use when asking such questions. He was perplexed by social change for much of his life and asked:
What is it that changes? Or is it more appropriate to make the question plural? What are the things that change? All things change all the time, including people and communities and societies. How much change then, and in what, do we need before we talk of 'social change'? (Stirling, 1974)
As a result of this study I can begin to offer some answers. I do not believe these 'answers, could have been arrived at easily any other way.
I have conducted a coarse grain study of one 'hypotheoretical' model generated from Stirling's 'wiring diagram. Starting from 'the basics' I have systematically deconstructed & reassembled many times the ideas behind these models. During each 'reassembly' I gained a clearer picture of what lay behind them. This is not a 'reductionist' process in the usual perjorative sense but rather an 'engagement' with the data that leads to 'refinement'. As a result, rather than design a complex program that simulated many aspects of Stirling's models, I have designed a very simple program which tests a very simple hypothesis from one part of his 'wiring' diagram, however the simplicity of this program does not mean that it does not represent complexity. It does not mean that we can reduce all social phenomena to simple variables, constants and relationships. Representing social phenomena as variables, constants and processes is just a formal way of analyising conceptual 'objects' and their relationships such as those in the 'wiring diagram' They can be physical, abstract conceptual - whatever!. What these variables and constants are and why they are related in the way we suspect is still open to debate. All attempts at intellectual understanding posits relationships between 'things'. Doing it this way has the advantages that we can play with the relationships dynamically in a computer rather than simply make insightful
'if this, then that... or....that....else the other'
statements based on our 'intuitions' that cannot be verified other than by political assertion. By analyising the predicate and prepositional logics of the language used in argumentation we can get away from the often confusing, agended and multidimensional associative properties of the 'Written Word'[27]. Although often, we might not want to do this because of the qualities of the domains in which we are working, for this kind of analysis it is a sensible thing to do. The ethnography provides the 'context' and the mathematics provides the logical analysis. Anthropological quests are ideally suited to these complementary methods & approachs. They are always in one form or another implicit if not explicit in an intellectual 'theory' of anything!
As can be seen from the data in Chapter Three, the correlation between Stirling's census data and the simulation data in this initial experiment is remarkably good. There can be no doubt that there is something to investigate further but what is it? What are they?
If we examine the implemented model(Appendix F - derived directly from Proto sim1 p35) carefully, there are only variables (held within objects in Java) for birth rate (SKPop.Java) brate; death rate (SKPop.Java) drate; the minimum amount of land per household before migration occurs (Household.Java) minland, a variable for the population of the village (Household.Java) pop; a variable for the migrant population (Household.Java) mig and a variable for the total number of households hh. For any single run of the simulation all other values are constants. They can be changed before a cycle is started but not during (at least not yet). The constants are (Household.Java) hhnormp which is the normal number of people per household; (Household.Java) vland which is the total amount of village land available and (Household.Java) hhMaxp which is the maximum number of people allowed in a house hold before the house hold splits. These variables are related to each other and modified by 'processes'. The population varies through each month of the simulation according to a standard equation relating brate & drate. These variables are changed each month by multiplying them by a random number (range +/- 1) generated from a normal distribution whose rate & variance can be set. The birthrate is graphed as output. This allows the 'stability' of the sim to be assessed as regards normal variability in survival rates (no infant mortality rate is required here as it has been aggregated to the birthrate variable and no male / female differentiation is needed as the problem is really one of food resources per household and is not gender specific). For most runs these were set at 5% birthrate and 3% deathrate. These figures come from the census data and are in line with Stirling's observations. The population is calculated[28] at the end of each cycle. This number is divided by hhnormp to give the number of households. When a household has more than the normal membership but the population as a whole has more than the required threshold of land then the household membership is just reported. If the membership of a household is greater than hhMaxp and the total number of people greater than minland then the excess become migrants and mi = mig + 1 and pop = pop- 1. All useful parameters are graphed as output and logged for analysis.
Experiment shows that there is only one really sensitive variable in this simulation and that is the minimum amount of land needed by each household to avoid migration. This is the causal focus of the simulation. The birthrate sets the total number of people; hhMaxp sets the total number of households but minland sets the migration profile. In the simulation runwhich best corresponded with the census data from 1950 to 1971, the initial settings were:
minland(MinLandHH on the front panel)= 84
hhMaxp(HHMax) = 5.9
vland (Land) = 12026 decares (3006 acres)
hhnormp (HHNorm) = 5.8
hh (HouseHolds) = 105
pop (Population) = 633
brate (BirthRate) = 5.3 %
drate (DeathRate) = 2%
From this we can say that the average minimum land available for a household of 6 people to avoid the need for migration47[29] was approx. 21 acres or 3.5 acres per person. This would constitute a 'subsistence' level amount and according to experts[30] is not an unreasonable figure to have arrived at. However, in order to get the simulation to fit the data from 1971 to 1986 this value had to be altered, no other variable helped. By altering birthrates; deathrates; maximum household size, we can take views on different possible outcomes for our 'village'. Only the minimum land available variable changed the profile of the results. The rest just prolonged or delayed the final effect. So what could this mean?. There is plenty in the ethnography to suggest that the seventies were as different to the sixties as the sixties had been to the forties for the villagers. A metalled road was built which cut the journey time from Sakaltutan to Kayseri from Two hours to twenty minutes. Tractors and new fertilisers were introduced which increased crop yields from 5:1 to 8: 1. These changes are reflected in the census data. By adjusting the MinLandHH variable (Household.Java p3) It was possible to see that the only way to account for the data was if something had happened that had the same effect as increasing the productivity of the land i.e decreasing the amount of land needed by each household. This could have been fertilizers - tractors or a general diversification of the economic base of the village i.e. remittances from pendular migration work. Effectively, the village carrying capacity increased slightly and more households could be supported. In a more sophisticated simulation events like this are held in a 'historic' modifier list which triggers a change of a variable according to a specific criteria rather than a regular cycle. Perhaps in a future study this could be taken up as an investigation into the 'finer' structures of the village system.
The average amount of land per household variable is a variable that relates the social to the physical. It is a 'hard' biological determinant i.e. I f people do not eat they die, it has little 'elasticity'. However another variable that has influence on the results was the N=6 (average) house hold membership variable. This variable is surely culturally defined. It has a high degree of elasticity and can vary depending on all sorts of things like how you define the concept 'household'. Once alerted to this 'constant' (Fig. 24) it is possible to look further. The ideal of the 3 generation full joint household (Stirling, 1963) which splits on the death of the father is an immediate place to look. An analysis of this structure is consistent with this number of 6 so we can expect that unless this 'ideal' changes '6' will remain important. Often attitudes change more slowly than the contexts in which they have been formed and this seems to be the case here. The migrant population average household size (Fig. 24) starts small but increases. One can extrapolate and suggest that by the late 1980's - if this ideal from the rural context has remained part of the village 'tradition' in the new urban context & the environment allows it, it will have stabilised at around N=6 again.
From the discussion so far I argue that it is possible to gain some predictive insights into social systems as long as one is sensitive to the domains in which one is working. These domains are typically areas in which some regularity or 'structural ordering' is present and can be monitored. The kind of approach outlined above does not for example help one to be more intuitive to others or better at interpreting the 'mood' of a crowd in a concert or a political arena. It does not really help us to understand the 'noise' of events that whistle around us all the time. There are many areas in which only the sensitive human mind can make judgements and be creative in understanding. As a prosthetic device however it can unlock the secrets of certain domains in ways which our own minds are just not set up to do. Our intelligence rarely lies in doing millions of accurate calculations in a split second. We work differently but both us and the machines we make are subject to the same physical laws and must conform to the same physical constraints of material existence and the logic of these structures. Just as indeed the villagers are constrained by their biological needs for food.The machine just helps us to deal with parts of that existence that are normally hidden from our gaze. What we can do is to look at the physical fundamentals and evaluate much in the social - environmental overlap. Once we have an understanding of the constraints we are in a position to make statements about more social & cultural phenomena which must ultimately be built upon these constraints. Thus we can gain perspective.
Was Stirling right? It is not fashionable to assert any form of 'positivism' in the present academic climate but what this shows is no simple positivism but rather that there is 'something' out there which we can get a grasp of. Stirling 'knew' this and although he was problematised by what 'it' was, over a life time managed to get a good understanding of it. In this sense, certainly this part of his model is 'correct' that is, his argument stands up, is consistent and extremely plausible. I find this quite refreshing. A simulation like this sets a framework for maintaining consistency in our modelling. It offers a coarse beginning from which can be built up a more 'grounded' and refined understanding. By requiring our arguments to be internally consistent according to the 'natural' logics that we have to live by, it stops us from spinning off in interesting but often irrelevant directions. It forces our ideas to 'hang together'. It is obvious now that the village displays systemic behaviour, we have 'evidence, however this behaviour is certainly not simple.
I had hoped to be able to get Paul to critique this work so that in the process we could both dig deeper into what this all 'means' but he died just two weeks before this initial phase was completed. I'm sure that he would be pleased in many ways - He has been shown to be 'right' in the sense of the word outlined above and this, especially in social science is cause for celebration. He also however wrong, at least in some details. He weighted migration heavily as a 'casual' agent and certainly it is of fundermental importance but really the casual agent is a decrease in the infant mortality rate. This is the basic starting point, after that the main determinant of the outcome is the amount of land resource available to the expanding population. Migration then becomes a consequence not a cause. The other 'important' categories double lined in the 'wiring diagram'; Knowledge beliefs and skills; Pendula & Permanent migration are in a sense methods and solutions to these problems not generators of change in themselves. As he noted, there was no liking, at least at not at first, of migrant labouring . The preference would have been to stay in the village. Nothing would have changed unless forced to by external events. Of course once the wheels were moving Stirling makes a point of saying that on the whole the villagers were better off and he was glad of it. In 1950 after a bad winter and bad harvest some people had no shoes on their feet. Certainly in the eighties pregnancies were saved that just ten years earlier would have most certainly have died. Reading his accounts one is left feeling that there can be no easy judgement of the good & the bad of the changes. The infant mortality rate probably dropped because of better medicines and conditions after the war. The uptake of medicine and governmental change was brought about in many ways by a tripartite American agreement with Turkey that released aid in terms of cash, heathcare and infrastructure. One could argue that America 'caused' Turkey's migrations but it is difficult to applaud or blame them.
What we can ask is was this a new thing? The villagers seem to have been going through booms and busts of population certainly since the sixteenth century. A population explosion seems to have had a devastating effect48 on the Ottoman Empire for example.
Another piece of evidence came to light during this study in one of Stirling's field notes concerning another village:
{com: An extraordinary note. Pity I have ignored it. I should have done a
vast amount more local history. This implies that over many years - 60?
20? - Travsun has exported houses which have multiplied in other
villages. This confirms that my impression of population stability in 1950
was something both new and shortlived. I suspect that before say 1935
households moved between villages to gain land advantage from
temporary advantages -good vacant land, or kin links to rights. Tragic
missed opportunity}
Perhaps in the past land resource problems were solved not by having fewer children, a risky strategy in difficult environments, but by migrations from village to village. This time however other opportunities & technologies in the post war boom period have meant that people moved to cities and abroad, which in turn flooded the country with remittances that combined with aid and changed the operating constraints of the 'system' to the degree that more children survived and the state grew. These questions now can be addressed on the provisional understanding that Stirling's anthropological work has 'power'. I will now indulge in some speculative conjecture.
The debates surrounding concepts of causality, the 'real' world and whether there is anything to 'model' out there anyway are everywhere and complex (Hann, 1994) (Boudon, 1996) (Kuhn, 1962) (Feyerabend, 1975) (Foucault, 1984) (Park 1996) (Stirling, 1974) (Habermas, 1983; 1989) (Horkheimer & Adorno, 1972) (Fischer 1994), These debates, situated at the overlapping junctions of the enlightenment, post-enlightenment 'modern and so called postmodern' periods are ensconced in debates about the philosophy of science and knowledge; development theory, notions of globalisation, migration - transnationalism; systems dynamics(Forrester 1968; 1969; 1973) (Roberts 1983) etc. It would seem to me that we live in a complex and confusing time. What to ask?
The question seems in one sense 'relatively' clear. If we are given a number of individual interpretations of any set of perceived social 'events' in the world, are there any ways other than by 'faith' (Stirling 'credulity' 1998) & argumentative ability alone to distinguish between them? Are all interpretations equally valid or are some more valid than others?
Boudon (1996) outlines a set of criteria for the 'good theory' as opposed to the 'bad' theory and shows that it can apply to the social as well as the natural sciences. (He cites the work of Durkheim & Torqueville as being good examples of this) In contrast, Hübner (1985) argues that there is no rational way of distinguishing between scientific theories and myths. Harvey (1990) cites Eagleton:
In experimenting with the ideas, 'models' and conclusions of the anthropologist Paul Stirling I have tried to ask "What did he do?" "How did he represent his ideas" "Were they plausible?" "did they account for or explain anything?" "did the ideas work?".
The provisional conclusion is that yes they did work and the proof is above.
He predicted the development and unfolding of a social system in 1953 and had the 'pleasure'? of watching it unfold. He saw some socio - physical fundamentals and understood their consequences but how do we fair without having ourselves such intimate and ongoing experiences? Many times in debate he would correct my speculations on his work because he had the background to do so. With his death there are many many pieces of information that have been lost and what is left is his archives. Trying to work with this alone as most people undertaking research have to do is very different than when the living source is there to 'tell it as it was'. The absence of source opens up a space for more personal speculation ungrounded by living evidence, as we recede from direct experiences the easiest thing to do is fill this space with our own interpretations. The hardest thing is to see clearly but there are points at which we can take anchorage.
If we take ourselves as being at one level of reality embedded in an objectively 'real' world it follows that we must share a degree of common experience. Although we may represent the 'objects' of these experiences differently in individual and cultural domains we are able to recognise and represent the 'objects existence' between differing frameworks with a degree of comparative ease. i.e. the symbolic 'value' of a cow to a Goan Hindu is very different to that of a Kentish beef farmer but they both recognise the existence of the cow. That is we can find correspondences between our differing human centered world views. If we could not then we would indeed lack referentiallity (Harvey, 1992) and be paradigmatically isolated (Kuhn, 1962), not just culturally as some define, it but also individually. This is obviously to my mind not the case! If we look carefully we can come to the conclusion that our representations are recognisably underpinned by a layer of cross cultural realities and we can posit this to be true whether we realise it or not[32]! The areas of these correspondences can be seen in the comparative linguistic analysis of the noun and verb classes of language representing as they largely do the 'objects' of the physical world and the 'actions' we make in the physical world. However when we move into 'cultural' spaces, especially 'philosophical' or 'artistic' spaces and analyse such concepts as 'meaning' and the multidimensional associations we have about objects and concepts at that level, then we are talking about a terrain not so easily accessible by others who are at a 'thinking' distance from us. These 'abstracted objects' and 'concepts about concepts' are not as 'shared' as other 'objects' in our common realities. Here ideas are less bound - more fluid and flexible - multidimensional - associational and contextual & therefore just as likely to disappear as appear.
Now having differentiated between different levels of conceptual relationships. It is possible to see at what level Stirling's main analysis is based. It is not at the cosmological, religious or ritualistic level but rather the more material, economical, structural level. It is at the level of interplay between the villagers and their constraining environment. He is able to predict34 because he has mentally 'simulated' results in the real - persistently patterned world, of a change in the binding constraints which underpin village existence. These identified constraints and processes are therefore more 'fundamental' to an understanding of what actually was happening. Simply put Stirling could see 'what mattered25' in the village in respect to his personal interest & analytical purpose.
The wiring diagram is a diagrammatic representation of ideas abstracted from his texts and experiences. It is in a sense a tertiary level analysis where:
1. Primary - field notes.
2. Secondary - analysis into articles books and papers
3. Tertiary - abstraction into causes, events and abstracted concepts.
With one more level of formalising it could have been already in a form to be simulated and in a loose sense 'tested'.
Stirling was concerned that his models were "accurate" but the real question is - accurate for what? Here is the paradox. He was only able to generate these 'simple' models after a long and intimate affair with the complexity from which they came. A quick immersion will not do. Any analytical objective requires simplification. If he had merely stuck to the Geertzian ethnographic ideal of "thick description" he would never start to cut through the endless detail.
Stirling wanted to leave his 'collected information' for others to use and decided to store it in an electronic database. He had to re - engage with his data in order to do it. The database had to be designed. This design process requires a reformulation and organisation of his data because the database itself is a classificatory device. It represents specific intentions and reflects these intentions in its structure as did Stirling's primary data albeit in a much broader sense. The Turkish Village Database is orientated towards questions of migration. It presents some views of the gestalt. Its existence betrays the reason for its existence. Stirling had broad intentions and therefore perceived a lot of data. He then had more specific research interests and had to simplify this data but, having done that, the point was not to try and deconstruct the simpler models into atomic variables, constants and relationships - this is obviously unnecessary and probably impossible. All that is needed is that the variables chosen are aggregated at a level appropriate for the problem in hand. If you can't make it appear simple, you don't understand it! The question is how many variables do you need in order to address the problem without losing what matters? This is to me the art of science. I don't think there are shortcuts to insight and this is why longitudinal studies of Stirling's caliber are so important.
If we examine the government or UN or CIA Fact file or World Resource Institute[33] econometric statistics for Turkey over this time period - GNP, GNP per capita etc., these indicators of macro level effects gathered at a distance quickly, tell you something about what might be happening at the village level, but not much. They are variables aggregated at too high a level of abstraction and do not have the resolution to see into the village.
Stirling's objectives were amongst others, to understand the causes and effects of migration. He has already in his diagram shown 9 separate boxes connecting with the boxes for pendular and permanent migration. It other words this is an 11 variable schema. I have shown that we can 'test' these hypothesis using mathematics for the analysis and the ethnography for the context. It is these multiple approaches that anthropology could benefit from I believe. We can 'test' the arguments - get results and then debate what it means having rooted & filtered the noise.
I have described, based on the earlier work of people like Fischer (1994), a method
for tackling the 'how' of it all but perhaps, unlike Dawkins and Hawkins who are
content to lay open the mechanisms of this world, assess likely outcomes and leave
it at that, we can also engage with the 'why' and the 'so what!' of it. What can
we do & what should we do with the knowledge of our analyses? These have always
been the real questions!