THE SPIESS EXPERIMENTS
Since 1984, reports of the negative results obtained in an experimental investigation of the lunar-sidereal 'trigons' in crop yield by Spiess have obtained widespread publicity and a major biodynamic textbook discussed the Spiess conclusions as having refuted the 'Thun-effect' (Koepf et al., 1996). It seems, however, that the negative nature of these results (Spiess, 1994) has been exaggerated. The effect present in the Spiess data appears as of relatively small amplitude, though in the case of his carrot data it achieved a high level of statistical significance if the three years of his trial were pooled together. For one year of his radish trials (1982), two sets of radish were sown sequentially, one with added fertilizer and the other without, where the sidereal rhythm was more strongly present in the data-set without added fertilizer.
Spiess developed a distinctive two stage method of data
transformation, here applied to results of his 1982 radish sowing trial
and his three years of carrot trials, 1978-80. The discussion here is
thus confined to his radish and carrot yields (Spiess, 1990b). His winter
rye sowing trials (Spiess 1990a) have not been analysed because seeds
sown in winter-time could fail to germinate within a day of being sown.
It is crucial to the theory of these experiments, that germination occurs
on the day of sowing.
Radish trials, 1982
To compare the yields (Y) of two or more trials, normalized yield values Y1= 100Y/Ymean were used. For investigating whether lunar-monthly rhythms are present in crop yield data, it is, in general, necessary to allow for the seasonal or year trend, as this may be larger in magnitude than rhythms of a monthly nature. To do this, one can put second-order regression lines through the data, i.e. best-fit parabolae, as did Spiess, or one can apply a moving average. Such a moving average here includes two sowings before a given date and two after, taking the mean of these five values. The latter method was used here (Kollerstrom & Staudenmaier, 2001). Both techniques, a regression line, plus a five-point moving average, are illustrated in Figure 7, applied to the Y1 data of the 1982 radish experiment.
The trend line values (T) are then subtracted out from the Y1 data-points. That subtraction gives a new data-set, which is called Y2 (i.e., Y2=Y1-T). These Y2 data points are used to test for lunar-sidereal rhythms present in final crop yield. We used this methodology as applied by Spiess, but modelling the seasonal trend differently as described, to go from Y1 to Y2. As Figure 7 indicates, the way one allows for seasonal trend can affect final values considerably. The 1982 experiment involved one lot of sowings per Moon-constellation (i.e. 12 per sidereal month) over a 39-day period, so that 19 rows were sown in all. Two separate sets were conducted in parallel, one with added fertilizer and one without. The experiment had randomized, complete block designs with four replications' at each sowing (Koepf et al., 1996).
The next step involved making a separation within the Y2 data-set. The hypothesis tested was that one 'trigon', in this case root-day sowings, gives a better yield than do the others. The data was therefore separated into two groups, root-day sowings and the rest. Spiess separated the data into four groups, by the four trigons, performing a four-way analysis of variance. These two approaches are both valid, but the former may have simpler statistics. Separating the Y2 yield values into these two groups gave the results shown in Table 2. These Y2 mean values are percentages, because the Y1 groups have means of 100. Thus the no-fertilizer group had 7.6% excess for the root-day yields, while the other rows had 2.4% deficit, so that, overall, the root-day sowings achieved 10% more in weight yield than did the other sowings.
Mean yield-deviations (Y2) normalised to a mean of 100
for Spiess's two 1982 radish trials, root-trigon sowings vs. others.
Owing to the large scatter in the data (Figure 7) the effect was not statistically significant (t=1.7, p = 0.1). It is preferable that mean yields should not increase so sharply during the course of an experiment. Even with a 10% excess as predicted, this increase prevented the data from attaining significance. Trial II with fertilizer did not show a statistically relevant yield distribution.
Thun has claimed that the sidereal trigons show up best
if crops are not treated with fertilizer (Thun 1964), which Graf in
her PhD at Zurich claimed to have confirmed (Graf 1977). Though of smaller
magnitude than others have found earlier from investigations of the
'Thun effect', this yield excess may nonetheless be large enough to
be of interest to commercial growers, as the result of a well-designed
and carefully performed test of the 'Thun-effect'.
Carrot trials, 1978-80
This same technique of data analysis was applied to the three years of carrot sowing data, which Spiess performed in 1978-80 (Spiess, 1993, 1994). For these three years of trials the same experimental procedure was used, sowing approximately once per Moon-zodiac constellation as for the 1982 radish trial. The trials extended over a month, which meant that 14-15 rows were sown each year. However, the last six rows sown in 1979 had to be omitted, owing to a two-week discontinuity in the data.
The carrot trials achieved a more uniform growth than did Spiess's radish trials. An indicator of how well trials have been conducted is the standard deviation of the yields expressed as a percentage of the mean. This is shown Table 3, where the variance of the carrot-yields is compared with that of the radish-trial discussed above. For whatever reason, these are a mere one-third the value for the carrot trials, so that better results might be expected from analysing his carrot-data. Five-point 'moving average' trendlines were put through these groups as before, subtracting which gave the Y2 values.
Standard deviations in the Y1 and Y2 data groups of Spiess
trials here analysed
The Table shows how two-thirds of the variance is removed by subtracting out the seasonal trend line, passing from Y1 to Y2, also how the number of rows per group was reduced in passing from Y1 to Y2 values, by use of the moving average.
The transformed carrot data 1978-80 were separated as
before into two groups, root-day sowings and others. Over the three
years of these trials, the Y22 root-day values averaged 6.3 ±
4.1 (n=5) while other sowings averaged -1.6 ± 2.9 (n=22), an
8% excess, with a t-value of 5.4, as was highly significant (p = 0.001).
In the 1990s, discussions in print of the biodynamic calendar in Europe, America and New Zealand, have alluded to the experiments conducted by Spiess as having tested the Thun-hypothesis and failed to replicate it (e.g., N.Z. Biodynamic Association 1989; Llewellyn, 1993). Enjoying widespread publicity, and published by the Forschungsring of the German biodynamic movement, the Spiess results have worked to discredit biodynamic calendars.
It appears, however, that although the experiments were well designed, this was not matched by a corresponding care in the data analysis. There were two other radish trials, which Spiess performed in 1979 and 1980, over 30-day periods, where in the first case yields increased by a factor of seven from start to finish of the experiment, and in the second case they more than doubled. But, in experiments conducted over only one month, containing such large seasonal trends within the data, it is unrealistic to expect low-amplitude sidereal rhythms to be detectable. The Dottenfelderhof farm near Frankfurt, where the trials were performed, is in a quite highly industrialized area of Germany, as could here be relevant.
While the Spiess sowing trials and methodology had commendable features, use of parabolic curves to model the seasonal trend was inappropriate, and the alternative here used of moving averages to model the seasonal trend gave Y2 element-means of considerably smaller standard deviations. Results published to-date suggest that the 'Thun-effect' is a testable and verifiable hypothesis. The current analyses endorse Spiess's general conclusion that 'lunar factors' may have a practical significance for agriculture.