Ann. Mo. Bot. Gard., 15(3): 241-332. (1928)

The problem of species in the northern blue flags, Iris versicolor L. and Iris virginica L.
Edgar Anderson

II. TAXONOMY AND MORPHOLOGY
HISTORY OF THE SPECIES

<skip>

The first five year's measurements have been summarized in tables I-IV. The figures in italics are the class containing the median or mid value. The median has been used as an average rather than the mathematical mean, since it is less influenced by occasional extreme values. The extra-small size of the frost-bitten or insect-mutilated petal, for instance, is not really significant, yet a single such extreme observation might seriously influence the position of the mathematical mean. It would have no more effect on the position of the median than would any petal of less than average size.

One fact is immediately apparent from an inspection of the summary. No single measurement will serve as a criterion for separating the two species. It is thus apparent at the outset that no biometric method of distinguishing the two species can be a simple matter. While it is certainly true, as Hall and Clements ('23) and McLeod ('26) have suggested, that taxonomy needs the development of more exact methods, there are serious limitations to a wholesale inclusion of biometry in every-day taxonomic procedure.

These limitations may be easily demonstrated by a simple example. Figure 6 shows five petals of Iris virginica and five of Iris versicolor. For the purpose of taxonomic description they may be conveniently and accurately separated by describing those of Iris versicolor as ovate-lanceolate and those of Iris virginica as obovate-spatulate. To separate them by biometric methods is by no means so easy, as table V shows.

Fig. 6. A, outlines of five petals each of I. versicolor; B, outlines of five petals each of I. virginica,

TABLE V

  Petal length Petal width Petal taper Taper width
Iris virginica
No.
cm. cm. cm.  
1 4.2 1.4 1.5 1.1
2 4.7 1.7 0.8 0.5
3 4.8 1.7 0.9 0.6
4 4.6 1.3 1.1 0.9
5 4.6 1.6 1.2 0.8
Iris versicolor
No.
       
1 4.7 1.3 2.0 1.5
2 4.8 1.5 1.7 1.1
3 4.1 1.2 1.3 1.1
4 3.0 1.2 1.2 1.0
5 4.5 1.4 1.7 1.2

Neither the length nor the width will suffice. The taper (i. e. the length between the point of maximum width and the tip) is only somewhat better. The ratio between this latter measurement and the width is still better though it fails to separate the entire lot. None of the measurements is as good for purposes of distinction as the single terms "ovate-lanceolate" and "obovate-spatulate." Only by combining all three measurements into a complex ratio would it be possible to demonstrate, mathematically, the discontinuity between the two sets of petals. As Minot (quoted by Thompson, '17) has said, "The fact that men of genius have evolved wonderful methods of dealing with numerical relations should not blind us to another fact, namely, that the observational basis of mathematics is, psychologically speaking, very minute compared with the observational basis of even a single minor branch of biology. * * * While therefore here and there the mathematical methods may aid us, we need a kind and degree of accuracy of which mathematics is absolutely incapable."

The above example illustrates the two chief reasons why biometry must necessarily be limited in its application to taxonomy. In the first place, mathematics is swift and efficient only in recording differences in number; it becomes cumbersome in record­ing differences in form, however useful it may be for a deeper analysis of the forces which produced the form. Yet it is just such differences in form which are most commonly met with in taxonomic work. This point is shown in a survey of the points on which specific distinction is based in two representative families. (one from the Monocotyledons and one from the Dicotyledons) in the seventh edition of Gray's 'Manual'. The differences were classified as being based on shape, on absolute size and number, and on comparative size and number. In all doubtful cases the preference was given to differences in number.

  Differences
in shape
Differences in absolute size
or number
Differences in comparative size and number
Iridaceae 21 13 12
Boraginaceae 31 11 9

There is an even more fundamental reason why the customary methods of mathematics are not well adapted to taxonomic work. Such methods are comparatively simple when the variations in one characteristic are being traced; they become involved and cumbersome when the simultaneous changes in a large number of variables are studied. Yet this last is essentially the method of the taxonomist. When he distinguishes between a group of sugar maples and a group of silver maples, for instance, he is summarizing a large number of differences--differences in form, size, color, and color pattern. Mathematics can deal with such problems through the study of correlation, but it is slow and laborious work and though it may be useful in the analysis of some particular problem it is not adapted to general taxonomic use.

An attempt has therefore been made to develop a new method of presenting biometric data which would combine the good points of the methods of mathematics and comparative morphology. Like mathematics, it is accurate and objective. Like morphology, it leaves something to the trained eye. This has been accomplished by diagramming the data in a series of ideographs. Figures 7 and 8 show how the four measurements on the petal and sepal are combined into a single figure. Essen­tially the ideograph consists of a diagrammatic white petal, superimposed upon a diagrammatic black sepal. By constructing an ideograph of this sort for each plant measured, it is possible to show simultaneously the variation in the four variables considered, for an entire population of plants, or to compare the variation in one population with that in another as in figs. 10 to 13.

Fig. 7. Diagram showing typical flower of I. virginica and resulting ideograph.
Fig. 8. Diagram showing typical flower of I. versicolor and resulting ideograph.

Such ideographs would seem to be of general usefulness in taxonomic work. They are capable of demonstrating slight differences in proportion which are not revealed by figures alone. In the genus Iris, for instance, though Iris fulva, Iris foetidissima, and Iris prismatica each belong to a different section of the genus, the three species have flowers of nearly the same size. The dimensions of the petals and sepals are so nearly the same that the species hardly appear distinct when the figures are compared, as in table VI. When these are arranged as ideographs, however, as in fig. 9, the essential differences in proportion between the three species are clearly demonstrated and they are seen to be morphologically distinct.

TABLE VI
COMPARISON OF FLORAL DIMENSIONS OF THREE SPECIES OF IRIS

  Sepal length
 in cm.
Sepal width
 in cm.
Petal length
 in cm.
Petal width
 in cm.
Iris prismatica 4.4 2.0 4.0 1.1
4.9 1.8 4.3 1.2
4.7 2.1 4.3 1.4
4.8 1.8 4.6 1.1
4.5 1.8 4.2 1.1
Iris fulva 4.9 2.2 3.7 1.4
4.9 2.4 4.0 1.2
5.6 2.6 4.5 1.6
5.3 2.4 4.1 1.3
5.0 2.2 3.9 1.1
Iris foetidissima 4.3 1.8 3.9 0.9
3.9 1.6 3.6 0.8
4.0 1.8 3.6 0.9
4.4 2.0 3.9 0.8
4.5 2.1 4.0 0.9
Fig. 9. A, ideographs of five plants of I. fulva; B, ideographs of five plants of I. foetidissima; C, ideographs of five plants of I. prismatica.

Ideographs for twenty individuals each of sixteen representative colonies are grouped in figs. 10-12 (I. virginica) and fig. 13 (I. versicolor). They give another proof of the striking variation in size and proportion which has been found in every colony studied. In marked contrast to the variation between individuals is the general resemblance between colonies of the same species. While several different colonies have slight individual tendencies, and while in the case of those colonies which have been measured in successive years (fig. 10) the peculiarities persist from year to year, there is practically no differentiation between one region and another. The only generalization that can be made is that Iris versicolor becomes on the average a little smaller as one goes from north to south and that Iris virginica becomes a little larger. Thus, although the colony at Stanton, Kentucky (fig. 12), seems slightly unusual by reason of its comparatively large petals the other colonies studied in Kentucky and Tennessee (Bonnieville and Camden) do not show these peculiarities. There is as much variation in proportion and almost as much in size in the colonies from southern Michigan as there is within the whole group of colonies of Iris virginica from the Great Lakes to the Gulf of Mexico.

Fig. 10. Ideographs of twenty individuals each of four colonies of I. virginica. Fig. 11. Ideographs of twenty individuals each of four colonies of I. virginica.
Fig. 12. Ideographs of twenty individuals each of four colonies of I. virginica. Fig. 13. Ideographs of twenty individuals each of four colonies of I. versicolor.

Above all, when the ideographs are considered as a whole, the two species remain completely and absolutely distinct. In spite of a wide range of variation in separate characteristics, when the combination as a whole is studied it is found to be strikingly constant. Iris versicolor remains always and unmistakably Iris versicolor. and Iris virginica remains always and unmistakably Iris virginica. There is not the slightest tendency for one species to merge into the other.

TABLE VII
MEDIAN MEASUREMENTS OF IRIS FLOWERS FROM DIFFERENT LOCALITIES

IRIS VIRGINICA
Number of
individuals
Locality Year Median sepal
length in cm.
Median sepal
width in cm.
Median petal
length in cm.
Median petal
width in cm.
25 Kimborough, Ala. 1927 6.3 2.8 5.7 1.8
21 Jackson, Miss. 1927 5.9 2.6 4.9 1.8
50 Camden, Tenn. 1926 6.7 3.0 5.7 2.0
35 Bonnieville, Ky. 1926 6.7 3.0 5.7 1.8
21 Stanton, Ky. 1926 5.5 2.6 4.9 1.6
20 Eastover, S. C. 1928 6.7 2.8 5.7 1.9
15 Maysville, N. C. 1928 7.5 3.0 6.1 2.0
38 Anna, Ill. 1926 6.3 2.8 4.5 1.8
27 Vulcan, Ill. 1925 6.3 2.6 5.3 1.8
15 Vulcan, Ill. 1926 6.3 2.6 5.3 1.6
38 Farmington, Ark. 1925 6.3 3.0 5.3 2.2
43 Wicks, Mo. 1925 5.9 2.6 4.5 1.6
23 Valley Park, Mo. 1926 5.9 2.8 5.3 1.8
40 Portage des Sioux, Mo. 1926 5.9 2.8 4.9 1.6
30 Portage des Sioux, Mo. 1927 6.3 3.0 5.7 2.0
28 Portage des Sioux, Mo. 1928 6.8 3.2 5.8 2.0
27 Louisiana, Mo. 1924 6.3 2.8 5.3 1.6
26 Fort Madison, Ia. 1924 5.9 2.8 4.5 1.6
30 Sunbury, O. 1927 6.7 3.0 5.3 1.9
28 Mill Creek, O. 1925 5.1 2.1 4.1 1.4
23 Huron, O. 1925 5.5 2.4 4.1 1.4
63 Bay Bridge, O. 1925 5.1 2.4 4.1 1.4
33 Middle Bass Is., O. 1925 5.1 2.6 4.5 1.6
30 Lawrence, Mich. 1926 5.9 2.8 4.9 2.0
30 Schoolcraft, Mich. 1926 5.9 2.6 4.9 1.8
30 Centerville, Mich. 1926 6.3 2.8 5.3 1.8
35 Hartland, Mich. 1926 5.9 2.8 4.9 1.8
25 Armada, Mich. 1924 5.5 2.2 4.1 1.4
45 Yale, Mich. 1924 5.5 2.4 4.5 1.7
43 Otisville, Mich. 1924 5.5 2.4 4.5 1.6
19 Otisville, Mich. 1926 5.5 2.4 4.5 1.8
35 Frankenmuth, Mich. 1926 5.9 2.6 4.9 1.8
30 Frankenmuth, Mich. 1927 5.9 2.8 4.9 1.8
30 Linwood, Mich. 1926 5.9 2.9 4.9 2.0
25 Linwood, Mich. 1927 5.9 3.0 4.9 1.8
25 Muskegon, Mich. 1924 5.5 2.4 4.5 1.6
52 La Crosse, Wis. 1924 5.9 2.6 4.9 1.6
27 Pardeeville, Wis. 1927 5.5 2.8 4.9 1.8
33 Slinger, Wis. 1924 5.9 2.4 4.9 1.8

 

IRIS VERSICOLOR
Number of
individuals
Locality Year Median sepal
length in cm.
Median sepal
width in cm.
Median petal
length in cm.
Median petal
width in cm.
28 Hood, Md. 1927 5.9 2.8 3.7 1.2
22 Liverpool, Pa. 1927 5.9 3.0 4.1 1.4
35 Harmonsburg, Pa. 1925 5.5 2.8 3.3 1.2
29 Conewango, N. Y. 1924 5.9 3.2 4.1 1.8
28 Villenova, N.Y. 1927 5.9 3.0 4.1 1.6
35 Hubbardsville, N.Y. 1925 5.5 2.6 3.3 1.0
27 Pownal, Vt. 1925 5.1 2.2 3.3 1.0
38 Pownal Center, Vt. 1925 5.1 2.4 3.3 1.1
37 Clarendon,Vt. 1925 5.1 2.6 3.3 1.0
34 New Haven Jct., Vt. 1926 5.9 3.0 4.1 1.5
25 Middlebury, Vt. 1926 5.9 3.0 4.1 1.4
20 Duxbury, Mass. 1927 5.5 2.8 3.7 1.4
26 Alberton, Ont. 1927 5.5 3.0 4.1 1.6
32 Ottawa, Ont. 1926 5.9 3.0 4.1 1.4
26 Timagami, Ont. 1926 5.9 3.2 4.1 1.2
22 Truro, N. S. 1927 6.3 2.8 4.1 1.4

By employing another mathematical concept, that of the average, we can construct a different sort of ideograph and produce something like a composite picture of each colony. If for each colony we take the average length of sepal, average width of sepal, average length of petal, and average width of petal, and construct an ideograph we will obtain a figure which will present graphically the averages of all four measurements for the colony in question. It will be a purely hypothetical figure; it will not necessarily present the kind of proportions most commonly met with in the particular colony but it will serve for convenient comparison between separate colonies and between successive measurements of the same colony. As in the earlier presentation of averages, the median is used rather than the mathematical mean because it is less influenced by occasional extreme values. The median values and also the number of individuals measured are presented in table VII. The resulting ideographs are shown in fig. 14.

The examination of these composite ideographs strengthens the conclusions already arrived at. The differences between colonies are very slight. In spite of the fact that the colonies measured extend from the Gulf Coast to the Great Lakes in the case of Iris virginica, and from Maryland to the north woods in the case of Iris versicolor, there is practically no evidence of regional differentiation; that is, of the formation of morphologically distinct geographical subspecies within either Iris versicolor or Iris virginica.

Composite ideographs of 39 colonies of I. virginica. Composite ideographs of 16 colonies of I. versicolor.

Section 1