310 Dr. F. Galton .
An Examination into the two divisions was obtained for the reflection reading .
From this an approximate idea of the intensity of the reflected component may be obtained .
Half the total radiation gave a deflection of 120 divisions .
The intensity of the reflected component , with a thickness of 0 45 mm. , is therefore l/ 120th part of the total amount of incident radiation , on the assumption , which is only approximate , that the galvanometer deflections were symmetrical .
When the thickness was reduced to 0'3 mm. , no reflected component could be detected , though the receiver was made extremely sensitive .
4 ' An Examination into the Registered Speeds of American Trotting Horses , with Remarks on their value as Hereditary Data,5 ' By Francis Galton , D.C.L. , F.R.S. Received November 29 , \#151 ; Read December 16 , 1897 .
It is strange that the huge sums spent on the breeding of pedigree stock , whether of horses , cattle , or other animals , should not give rise to systematic publications of authentic records in a form suitable for scientific inquiry into the laws of heredity .
An almost solitary exception to the disregard , shown by breeders and owners , of exact measurements for publication in stud books , exists in the United States with respect to the measured speed of " trotters " and " pacers under defined conditions .
The performance of 1 mile by a trotter , harnessed to a two-wheeled vehicle , carrying a weight of not less than 150 lbs. inclusive of the driver , in 2 minutes 30 seconds qualifies him for entry in the Trotting Register , giving him , as it were , a pass-degree into a class of horses whose several utmost speeds or " records " are there published .
To avoid prolixity I will not speak particularly of pacers ( pace = amble ) , since what will be said of the trotters applies in general principle to them also .
The great importance attached to high speed , and the watchfulness of competitors , have resulted in evolving a method of timing trotters which is generally accepted as authoritative .
The length of the track is scrupulously measured , and numerous other conditions are attended to , that shall ensure the record being correct , with an attempted exactitude to the nearest quarter of a second .
A race against time , even if exact to the nearest quarter of 8 second , is by no means so close a measure of the speed of a horse relatively to his competitors , as the differential method of ordinary races .
The speed of 1 mile in 2 ' 30 " , or of 1760 yards m 150 seconds , is equivalent to about 12 yards in 1 second .
Now , t e length of a horse when extended at full trot is half as long again a8 his'height at the withers\#151 ; as I gather from the instantaneous photographs of Muybridge\#151 ; and consequently is hardly ever as much afi registered Speeds of American Trotting Horses .
311 3 yards .
Therefore at a 2 ' 30 " speed a liorse travels through his whole length in a quarter of a second .
In an ordinary English race i winner by half a length gains a notable victory , while a neck or even a head in advance is sufficient to establish his priority .
Therefore the record of the speed of a horse to the nearest quarter of a second is by no means an absurd refinement It is , of course , very difficult under the exciting circumstances of a race to measure time with such precision as that .
I tested the value of these entries as follows:\#151 ; If quarter seconds were noted with exactness the entries of 0 , and f would be approximately equal in number ; they would also be equal if they were set down at random without bias , but if there be a bias towards favourite numbers its effects would be apparent .
I extracted a few hundred entries , and found the relative frequency of the 0 , and f to be almost exactly as 1 , 3 , 2 , and 1 .
Consequently the \ is on the average three times as great a favourite as either the 0 or the f , and the ^ is twice as much a favourite as they are .
It is evident that the \ seconds are not strictly trustworthy , but it may well be urged that their entry is preferable to their total disregard .
I was informed that a trifling laxity was tolerated when a horse had just but only just failed to qualify , an allowance of ^ of a second in his favour being commonly made .
So that a speed of 2 ' 30\#163 ; " would usually be reckoned as 2 ' 30 " .
I shall return to this point further on .
The system of timing and of registering records began more than fifty years ago , and was developed and improved by degrees .
In 1892 a considerable change was made in the conditions by the introduction of bicycle wheels with pneumatic tyres , wBjpli produced a gain of speed , the amount of which is much discussed , but which a prevalent opinion rates at 5 seconds in the mile .
Thenceforward the records are comparable on nearly equal terms .
All trotting performances up to the 2 ' 30 " standard are registered in the large and closely printed volumes of 'Wallace 's Year Book , ' published under the authority of the American Trotting Association .
Vols. 8\#151 ; 12 refer to the years 1892-6 , and it is from_the entries in these that the following remarks are based .
1 he object of my inquiry was to test the suitability of these trotting ( and pacing ) records for investigations into the laws of heredity .
1 heir trustworthiness was of course one point to be ascertained , another was to obtain a just notion of the proper principle on which marks for speed should be awarded , as , for instance , in the following example:\#151 ; Suppose a particular ancestor , whom we will call A , of a certain horse has a record of 2 ' 30 " , and that another ancestor in the same degree , whom we will call B , has a recoi 'd of 2 ' 10 " , how are their joint influences to be estimated ?
Will it be the z 2 Dr. F. Galton .
An Examination into the same on the average as that of two horses each having the speed of 2 ' 20 " , or will it be something altogether different ?
In short , is the arithmetical the most appropriate mean or not ?
It would he a strong presumption in the affirmative , if the relative frequency of the various speeds should correspond approximately with those determined by the normal law of frequency , because if they do so they would fall into line with numerous anthropometric and other measures which have been often discussed , and which , wdien treated by methods in which the arithmetic mean was employed , have yielded results that accord with observed facts .
Whether the speeds do or do not occur with the normal frequency had therefore to be ascertained .
So my inquiry had two objects : first , did the run of the observations suggest a tolerably smooth curve ?
Secondly , was that curve a tolerable approach to the curve of normal frequency ?
The investigation was troublesome and tedious .
It wras necessary to pick out from a large collection the names of those stallions , geldings , and mares ( all three being equally efficient trotters ) , whose records had been made in the year under consideration , and who also had arrived at maturity , that is , who were not less than five years old , and therefore had had time to show their full powers .
Had younger horses been included , the frequency of the slower records would have been much increased .
Assisted by a friend , the appropriate entries were underlined in the printed volumes , then one of us read them out , and the other ticked them down in the appropriate column of a page ruled for the purpose .
linally the marks in each column were counted .
In this way 5/ 05 extracts were made from the entries for the years 1892-96 ; they were not subsequently verified , so some few omissions are probable .
Anyhow they form a fair and large sample , and are quite sufficient for the present purpose .
T The discussion of this material resulted in rather bulky tables , which it is needless to reproduce here , because their contents are given in an adequate and much simpler manner by the accompanying diagrams .
The successive columns in the table are represent i\#187 ; the diagrams by imaginary columns that stand on corresponding bases .
They run us follows :-The first column , counting t* theleit , contains the percentage value of all observations recorded as 2 , 291 " 091 " or 29| " ; that is of all under 30 down to 29 mclusi ( the'minutes being here omitted for brevity ) .
The second column referred to 28-0 " , 28i " , 28* " , and 28| " , and so on with the rest .
Co -seqnently the dot in the diagram which indicates the percents number of observations , according to the side scale , stands m the middle of its own imaginary column .
For example , t ia 2 ' 28 " set stands vertically above the point that lies i\lt ; registered Speeds of American Trotting Horses .
313 between 28 and 29 on the scale along the base .
The dots are con-aected by thin lines to show the trace or curve of the observations .
The smooth curves are those of normal frequency , calculated from ohe values of the mean ( M ) and of the probable error ( P.E. ) , which ire given in the diagrams .
Leaving aside for the moment the strange pinnacle that rises on the extreme left of every diagram , we see that the traces of the observations run very roughly , but not intolerably so .
In each diagram they seem to be disposed about a fundamentally smooth curve .
Considering the smallness of the interval , namely , only 1 second , that separates the observations assigned to each pair of successive columns , together with the experience derived from other kinds of statistical curves , it seems to me that the run of the obser- Dr. F. Galton .
An Examination into the vations is good enough to certify their general trustworthiness .
As regards the pinnacle it is a different matter , and is one which .when beginning work , as I did , on the 1892 entries only , was very perplexing .
However , by persevering with the other years it became increasingly plain that the pinnacle was a false maximum ; in 1896 it was certain that the true maximum lay well within the portion of the curve included in the diagram .
The explanation of the pinnacle then became obvious ; it was that the tolerance granted to those horses who failed by only a little to qualify themselves , was extended considerably beyond the quarter second for which I was prepared.* The cases of 2 ' 3(H ) " were few ; they do not appear in the diagram , but their addition would be quite insufficient to remove the difficulty .
If the pinnacle were distributed among two adjacent columns outside and to the left of the diagram it would smoothaway the incongruity , so I suspect that cases of " under 32 " and down to 21 30 " " are habitually rated at a trifle less than 21 30 " .
Consequently I had no hesitation in wholly disregarding the entries that helped to make the pinnacle , namely , the whole of those contained in the first column to the left in every one of the diagrams .
The course thereupon became clear and straightforward .
When fixing upon the mean for each year , I was somewhat biased by the entries in the adjacent years ; similarly as to the probable error .
How that the curves are drawn I see that somewhat better fits might have been made , but they are close enough to show the existence of a. fair amount of correspondence between the observed values and those calculated according to the law of normal frequency .
It is near enough to remove hesitation m working with the arithmetic mean .
* { Jan. 20.\#151 ; I hare since learnt that the conditions of timing are too rigorous ( o justify this inference ; also that the very numerous efforts simply to secure* standard record , and thenceforward to cease training , may be a chief cause o t \#174 ; pinnacle .
] registered Speeds of A merican Trotting Horses .
315 I now come to the fundamental purpose of this memoir , which is to point out the existence in the registers of the American Trotting Association , of a store of material most valuable to inquirers into the laws of heredity , which accumulates and increases in value year by year .
Unfortunately it lies buried to a hopeless depth , partly ^because the published part of the registers refers only to standard trotters .
It appears to be buried simply through the omission of having its importance insisted on .
The published volumes of the 'Trotting Register ' contain numerous elaborate tables , but lacks one that should include the names and pedigrees of those horses concerning whose antecedents enough is known to make their pedigrees serviceable to investigators .
It is hardly worth while to discuss hereditary influence on speed , in the case of any horses , unless the records of at least their sires and of their dams , and those of each of their four grandparents , as well as their own record , are all known .
Even in this case ( according , at least , to my own theory ) one quarter of the hereditary influences are unknown and have to be inferred .
It is practically impossible to make an adequate collection of the names of horses who fulfil the above conditions out of the entries in the ' Trotting Register , each search requiring many cross references and occupying a long time , while the number of futile searches before attaining a success is great .
On the other hand , the breeders and possessors of these notably bred horses must be familiar with the required facts , and would assuredly be delighted to have them known .
There need , therefore , be little difficulty in obtaining materials for the much desired table .
In the meantime I am sending circulars to the chief breeders in America in hopes of making a start .
^ The great need for genealogical data of an exact numerical kind , by those who prosecute inquiries into the laws of heredity , is the justification that I offer for submitting these remarks to the Royal Society .

