Page 94 - J. C. Turner "History and Science of Knots"
P. 94

The Peruvian Quipu                      83

        group, these were divided into three unequal parts which then were encoded
        in the first three groups. As the exact result of the calculation is not always
        an integer, the numbers had to be rounded in order to be encoded on the
        quipu . Therefore we get an error which we state in percent of the number
        on the quipu. For the first group this error is about 0.6% for j 0 3 (i.e.
        0.33796 < (ply/pry ) < 0.34204) and 1 .4% for j = 3, in the second group 0.7%
        for j : 1, and in the third group 0.9%, again for j # 1. That means that the
        actual numbers are close to those we must expect when the quipu really is a
        division table.
        Multiplication of integers by integers and fractions : AS55 and AS56
        These are two small quipus with seven and three pendants respectively which
       were found together. Though the pendants are not grouped by spacing or
        colour, an implicit grouping appears when examining the values. We name
       the pendants on the first quipu p, q1, q2i q3, rl, r2, r3 and those on the second
        quipu s1, s2, s3. With this notation, we find several attractive relations:

                     glg2q3  rlr2r3  = SlS283
                            glr283  = r1s2g3 = s1g2r3
                            818283 = g1r2s3
                          qq ' ri+l = sj • sj+l  for  j = 1, 2, 3

       where the addition in the indices is mod 3, e.g. 3+1=1. We even find a more
       general pattern, namely

                ql =   B3Cx       r1 =    B6C2x      s1  =   B4Cx
                q2 =     x        r2 =    BCx        82 =     Cx
                q3 =     y        r3 =    B2C2y      S3  =   B2Cy

       where B = 7/8 and C = 34/33. Hence the numbers are multiples of the values
       on the third and fourth cord on the larger quipu by fractions. Of course, we
       cannot be sure that this is indeed the pattern behind the quipu. But it fits very
       well, as the error we make is remarkably small, namely in all cases smaller than
       0.4 %, in ten out of the twelve cases even smaller than 0.2 %. For a further
       examination of these quipus including the first cord and the three subsidiaries
       on the larger quipu, see [6], pp. 149-151.
       Geometric interpretation: AS120, AS143 and AS149 The quipus
       AS143 and AS149 have the same underlying structure as AS120, which we
       considered when dealing with division into unequal parts. AS143 has four and
       AS149 five pendant groups. The ratios pig/pry are the numbers bl, b2, c, a and
       b, c1, c2, c3, a in the following table, thus they stand in the same order as on
   89   90   91   92   93   94   95   96   97   98   99