Director Christine Shaw
That different cereal grain plants have emerged naturally during, and perhaps before, the period of human farming activity is undisputed. The exact timing and location when different types arose is less firmly established. The manner in which knowledge of their existence and how seeds came to be translocated is a matter for anthropology and cultural study, aided perhaps by insights from the modern science of information transfer. Growing trials can never give insight into any of this.
Principally, experiments can only show how each genotype behaves in a range of circumstances set to illuminate particular physically based questions, including important matters of climate and location , (Climate and Weather, Soil). The results from experiments may give insight into the economic benefits and sustainability of each crop plant. This may inspire discussion on wider aspects of farming communities and their livelihoods, extending the anthropological and cultural debate but providing no proofs, as such.
It would be a mistake to presume that past farmers took their decisions on the results of anything resembling modern crop trials ! There is a revealing Open University programme in the Information Science module, wherein a North African government (?Tunisia/Algeria/Morocco) wished to learn how best to influence farmers to take advantage of the potential benefits of a newly completed irrigation scheme. The studies showed that this would best be done through the village elders. However, it should be recognised that this modern day situation represents a well-established "stable" circumstance. In earlier societies, other forces are likely to have ruled, such as kinship, inter-marriage, migration, conquest and/or abduction and even, perhaps, theft.
Before embarking on setting the protocol to test defined objectives, with respect to differences in the performance of various genotypes, it is crucial to absorb all the lessons outlined in the associated pages in this section, dealing with the impacts of controlled and uncontrolled factors. The experimental objective will, equally, set the selection of Output Variables to be measured.
No efficient experiment in agriculture could ever assess the response of factors one at a time. Thus test designs must be established, from the outset, with the aid of a statistician or a sound appreciation of statistical design AND data analysis (including an understanding of the underlying assumptions in statistical methodology). This applies both to plot and factor allocation and to data collection protocols, which should all be carefully documented before a single action is taken. It is often overlooked that experimenters should have their own individual instructions about procedures. There is a misguided belief that telling someone to do something "randomly" is both meaningful and achievable. There are numerous studies to show that even experienced workers are incapable of doing anything "at random" without a specific procedure being advanced in clear detail.
It is a necessary caution in all multifactorial work to know that the methods used presume linear responses and additivity, including any interactions between variables that may exist. There are many reasons for supposing this may not be true. A simple example is that of sowing rates. Continuous increases in sowing rate will not give proportionate and ever increasing yields. Light, food and water will become less and less available for each increment, up to the point of failure. There is therefore a need to understand the range of each variable for which linearity might be a reasonable supposition.
One reason why such analyses give the impression of meaningfulness is that, even where non-linearity is known or reasonably expected to exist, a disproportionate amount of data may be needed to prove it. Thus the methods tend to be self-fulfilling, insofar as the standard cross-checks do not indicate anything may be awry. It is only when attempts are made to predict outside the original data range (another simple concept too readily forgotten !) that things may go amiss.
Another important feature of agricultural experiments is the wide range of variables experienced, no less so than in the weather and its effects. To meet this, especially in view of the great uncertainties in projecting back to the putative conditions of past millennia, results for anything less than trial periods of up to 20 years merely add uncertainty upon uncertainty. Results from work covering only a few can at best be described as indicative, at worst valueless and potentially misleading. This time constraint inevitably restricts our ability to study the outcomes for a wide range of climatic zones and soil types.
An essential point from modern agricultural research practice, that cannot be sidestepped, relates to plot size. It is routine to have a 1 metre planted strip along each side of a trial plot but with no samples collected from it. This is to avoid "edge effects" i.e. distortion of any target outcomes because of differing influences from the surrounding community compared with the test community. Thus it is easily seen that the MINIMUM overall plot size must be 3 metres by 3 metres.
The Butser test plots are 30 metres by 30 metres, in line with the size of Iron Age fields.
Test plot size ultimately dictates the quantity of seed required and this combines with aspects of storage and viability from year to year.
It is a reality that most, if not all, crop cereals have a maximum vital period of seven years, after which most of the seed is dead. Studies of farming practice around the globe show that grain for planting is routinely retained for only three years, by which time germinability has fallen from about 95% to around 75%. It is therefore essential to incorporate this into any test protocol, checking and recording both the history and germinability of source seed.
Created 01 August 2001 - Updated 21 January 2002