Most users of the Wheatstone ledgers visit the Horniman website to try to determine the manufacturing date or selling date of a specific Wheatstone concertina from its serial number. For many instruments this can involve extensive searching, as some of the early ledgers show an apparently random jumble of serial numbers listed by the date that instruments were sold, exchanged or hired. Even if repeating serial numbers caused by exchange or hire are ignored, the ledgers still show a jumble of numbers. This note contains the results of some basic investigation to show what is happening, using the first four sales ledgers, C104a, C1046, C1047, and C1048. I show that a major source of the jumbled sequence is the practise of selling (and we presume also making) multiple ranges of serial numbers at the same time, a feature which the ledgers show to have changed character and become much more pronounced after 1850. (This is a preliminary version, and will be replaced by a fuller report and further data files as work progresses.)

Examination of the ledger data, confirmed by other readers, shows that the earliest date written in the ledgers is "22 Oct 26" (i.e. 1826) for serial number 381 (Ledger C104a, page 20). Apart from being illogical (a date prior to the first Wheatstone patent in 1829), it can be shown that this is an error, since ledger C1046 contains an almost identical entry on page 51, but for 21 Oct 1846. Further examples of differences between C1046 and C104a can be found, and there is no reason to suspect that similar errors may not have equally occurred in later ledgers. The data from this investigation makes no attempt to correct any of these errors. Errors will also almost certainly have occured in transcribing the investigation data from the photographs of the ledgers, and reports of any of these errors will be gratefully received.

The ledgers investigated so far cover the period May 1834 (although this may also be an original ledger error) to March 1854. However, some parts of this range are not covered, and only the dates June 1839 to April 1848 and January 1851 to March 1854 are covered in depth. Ledger C104a covers the first 1500 serial numbers, and provides some of the missing coverage. According to ledger C104a, one instrument is sold in 1834, three in 1835, and fourteen in 1836. No instruments are listed for the first half of 1837, suggesting that those records may have already been lost when C104a was compiled. Sixteen instruments are listed for the second half of 1837, so the Wheatstone concertina output was beginning to increase considerably.

Searching for serial numbers between 1 and 1500 appears to be simple, as ledger C104a is ordered by serial number. However, investigation shows that dates given in C104a tend to be for a 'final sale' (by 9th September 1849, the latest date in C104a), and cross-checking with entries from ledger C1046 shows that some instruments were first sold years before the date entered in C104a. For example, serial 759 first appears in C1046 in July 1844, and makes two more appearances (December 1844 and July 1846) before its entry on 13th Oct 1846, the date given in C104a. Many instruments have earlier entries than their C104a dates, with their manufacture much nearer that of other instruments with similar value serial numbers.

The plot below shows the 'earliest dates' for serial numbers that appear in the ledgers used so far.

Earliest-seen dates for Wheatstone serial numbers,
in the Wheatstone factory ledgers, 1834–1854.
(Click to view a larger graph.)

In this graph, serial numbers are scatter-plotted from numbers near 0 at the top, down to serial numbers above 6000 at the bottom; each division is 1000 serial numbers. From left to right, each division is two years, running from 1834 to 1856. We see immediately a major difference roughly around 1850 (which is also, unfortunately, the break in the data caused by a missing ledger); the occurrence of serial numbers for 1834–1850 is very different from the occurrence of serial numbers after 1850. In the earlier period, we see basically a single main descending diagonal line, showing a single sequence in which serial numbers get larger as time gets later. But in the later period, we see multiple descending diagonal lines, showing more than one sequence of serial numbers being produced at the same time. (In both periods, there is a scatter of numbers around the major features.)

It has been suggested that the muddle of serial numbers in the ledgers has been partially caused by multiple production sources for instruments, with allocations of groups of serial numbers to particular sources causing years of difference between instruments of similar serials.1 Although this investigation cannot demonstrate multiple sources, it can confirm that multiple 'range-lines' of serial numbers were occuring in the sales recorded by the ledgers. If we look more closely, we can actually discern that the sub-pattern of multiple overlapping production ranges begins before 1850, and seems to be evident as early as 1845 (see below). The results also show that in the early years this effect is not as pronounced as later.

Earliest-seen dates for Wheatstone serial numbers,
in the Wheatstone factory ledgers, 1842–1849.
(Click to view a larger graph.)

In this graph, serial numbers are scatter-plotted from numbers near 400 at the top, down to serial numbers above 1400 at the bottom; each division is 200 serial numbers. From left to right, each division is a single year, running from 1842 to 1849. At this scale, we can discern an internal structure of overlapping ranges, usually up to three ranges being produced simultaneously over periods of up to a year each.

By the end of 1851, these multiple range-lines dominate the Wheatstone output and at least seven range-lines seem to be active. Within a few years the spread of range-lines becomes even wider—the first instrument in the 7000s appears at the end of C1048 (March 1854) even though few instruments with numbers in the 6000s have been sold. There is also a significant gap below 5640 with instruments around 5560/5570 only just starting to appear in sales. The lowest range-line still operating in March 1854 is producing instruments at approximately serial number 4390, so there is a gap of almost 2700 between highest and lowest range-lines in active sales.

Earliest-seen dates for Wheatstone serial numbers,
in the Wheatstone factory ledgers, 1851–1855.
(Click to view a larger graph.)

In this graph, serial numbers are scatter-plotted from numbers near 2500 at the top, down to serial numbers above 6000 at the bottom; each division is 500 serial numbers. From left to right, each division is a single year, running from 1851 to 1855. At this scale, we can see more overlapping ranges, up to seven or more being sold simultaneously, over periods of typically multiple years. The random mass of points in the range 2500 to 3000 indicate an area where we do not have earlier data (roughly 1848 to 1851), and most of these serial numbers will have earlier origins.

In summary, the plots above have shown that a major contribution to the muddling of serial numbers in the early ledgers is that multiple range-lines of serial numbers are in use at any one date, and that this is further complicated by instrument exchanges and hires. The multiple range-lines are a feature of sales before 1850, but only a few ranges are sold at a time, and for briefer periods, so that they blend into a fairly-simple sequence. After 1850, the multiple range-lines dominate the structure of sales, with many simultaneous ranges extended over longer periods. The true causes of these range-lines cannot be defined by this investigation, and it is hoped that future research will provide some answers.

The investigations made here also lead to a further observation. The C104a entry for serial 334 gives serial, date and buyer. When C1046 is checked for the same date, an entry shows the same buyer name, without a serial, but adds the number of buttons of the instrument. The presence of single source data like this indicates that C104a and C1046 have been compiled independently of one another, using another source or sources.

Demonstration of the presence of these range-lines opens up at least two further aspects of ledger research. Ledger entries for sales without serial numbers are no longer subject to a totally random possible serial number, and more probable numbers may be suggested for these entries. In the future, it may also be possible to correlate sales prices with these range-lines and demonstrate if they represent distinct types of instrument.

Contents

• Serial Number Muddle
  in Early Wheatstone Ledgers
• Notes
• Author

• Other articles
  by Wes Williams

Links to related documents