Hybopsis gracilis is highly variable in several morphological characteristics, including size and shape of head, body, and fins, and number of scales, vertebrae, and fin-rays. The variations are correlated in a way that indicates the existence of two subspecies. One of these, H. g. gracilis, attains large size, and has 1) a slender, streamlined body, 2) a depressed head that is acutely wedge-shaped in profile, 3) strongly falcate fins with the dorsal and pelvic fins originating anteriorly, and 4) many scales, vertebrae, and pectoral fin-rays. The second subspecies, for which H. g. gulonella is the oldest applicable name, is small, and has 1) a deep, chubby body, 2) head convex in dorsal contour (less depressed than in H. g. gracilis), 3) fins less falcate than in the latter subspecies, with the dorsal and pelvic fins originating more posteriorly, and 4) fewer scales, vertebrae, and pectoral fin-rays than H. g. gracilis. These differences are consistently expressed throughout the size-ranges of the subspecies, and in series collected at the same or nearby localities in several different years. Considerable variability was found in features other than those mentioned above, but individual variation among specimens from the same locality and adjacent localities is so great that none is diagnostic of subspecies. For example, orbital size and length of fins (but not their falcate shape) are variables that have little diagnostic value, although both features seem to vary in clinal fashion, with the higher values in the north.

Variation in H. gracilis, as shown in the graphic analysis (Figs. 1 and 2) and distribution map (Plate 21), presents two clines: a north-south cline and a large-river to small-river (mainly east-west) cline. The absence of H. gracilis from certain portions of river systems is a matter of concern. The species has not been found in the lower Arkansas River and the Rio Grande, nor in sandy tributary creeks in eastern Kansas and Missouri that appear to provide suitable habitat. It has already been noted that H. g. gulonella seems to be the ecological equivalent of Semotilus atromaculatus in streams in which S. atromaculatus is not found. S. atromaculatus occurs in creeks of eastern Kansas and Missouri, and may provide interspecific competition that prevents establishment of the flathead chub in these creeks. Regardless of cause, the gaps in distribution of H. gracilis tend to limit gene flow.

Many characters used in the separation of the two subspecies are known to be influenced by environmental conditions, especially temperature. Hubbs (1922, 1926, 1941), Schultz (1927), Vladykov (1934), TÅning (1952) and Weisel (1955), among others, have pointed out a correlation between temperature (or developmental rate of fish) and the number of vertebrae, scales, and fin-rays. Likewise, Martin (1949) and Hart (1952) have shown that the proportions of some body-parts vary in response to temperature during early development. In H. gracilis, the general nature of the clines found in a majority of characters (but not all characters) suggests a temperature influence. However, temperature-dependent variability that has so far been demonstrated experimentally in fishes is generally of lesser magnitude than the differences distinguishing H. g. gracilis and H. g. gulonella. To our knowledge, the most extreme differences that have been induced by modification of temperature are those reported for Salmo trutta by TÅning (1952:181-182), who states: "Shock treatment produced by especially great changes in temperature (c. 10-14° C), especially during the super-sensitive period [of somatic differentiation that fixes vertebral number] may produce ... a difference of 3-4 vertebrae ... in offspring of the same parents." The difference cited approximates that which distinguishes natural populations of H. g. gracilis and H. g. gulonella. Although we cannot assume that the sensitivity of the brown trout is the same as that of the flathead chub, the causative conditions in TÅning's study could scarcely be expected in nature; furthermore, it seems significant that extremely high (as well as extremely low) mean numbers of scales and vertebrae were found at southern localities, and that low mean numbers of scales and vertebrae were found as far north as Wyoming and Montana. We think it likely that temperature does influence the expression of characters in H. gracilis, directly in individual development, and indirectly as a selective mechanism in the evolutionary process. The extent to which each kind of influence exists can be proved only by experimental work with both subspecies, which we hope to undertake at a later date.

Other environmental factors that may have selective influence in this species are rate of current, volume of flow, and turbidity. Interaction of these environmental factors could result in genetic fixation of morphological characters through natural selection. The characters that distinguish H. g. gracilis from H. g. gulonella seem adaptive to life in large rivers and small streams. Evidence that these characters are under limited, direct environmental influence is found among populations in the Arkansas River System. Although populations in the Arkansas River have no continuity with populations of H. g. gracilis, upstream-downstream variations like those found in other river systems are apparent, but in lesser degree. The direction of variation in the Arkansas River is the reverse of that in the Platte and other tributaries of the Missouri River. For example, the populations farthest upstream (Florence, Pueblo) have slightly higher mean numbers of lateral line scales than do populations from Kansas, downstream.

A remarkable effect of extreme parasitism in H. gracilis has been described by Hubbs (1927). Very young chubs that harbored numerous tapeworms (Proteocephalus) had unusually large numbers of lateral-line scales, large eyes, short snouts, small fins, small mouths lacking barbels, and coalescent nares (internarial bridge weak or absent). Some of these abnormalities presumably resulted from retention of larval characteristics of the fish, correlated with the degree of infestation by tapeworms. No teratological adults were found, indicating that severe infections prevent survival to maturity. H. g. gracilis occurs in three separate river systems (Mackenzie, Saskatchewan, Missouri-Mississippi) from latitude 36° N to 66° N, and longitude 89° W to 123° W. H. g. gulonella exists as several seemingly-isolated populations in the upper parts of the Rio Grande, Pecos, South Canadian, Cimarron, Arkansas, Platte, and upper Missouri basins, from latitude 35° N to 48° N, and longitude 97° W to 100° W.

There is evidence of high mobility on the part of both subspecies, based on irregularity of their occurrence in certain localities. Many collections have been made in the Cimarron River in the vicinity of Kenton, Oklahoma, from 1925 to the present, but only one of these (in 1957) contained flathead chubs. Bait dealers who seine the South Canadian River in Dewey County, Oklahoma, have taken flathead chubs in abundance in some seasons, but not at all in others. Seasonal variation in abundance in the lower Vermillion River, South Dakota (Underhill, 1959:100) has been cited, and the number collected in the lower Kansas River near Lawrence has varied similarly. Many rivers occupied by H. g. gulonella (and by intergrades) are intermittent, and in some years their sand-filled channels become wholly dry for many miles. These factors probably promote mixing of the two subspecies, and may account, over long periods of time, for the wide dispersal of H. g. gulonella in the Missouri Basin. Flathead chubs are known from Pleistocene beds at Doby Springs, Oklahoma (the Doby Springs local fauna) (Smith, 1958:177). Drainage connections between the Arkansas, Kansas and Platte river systems existed in Kansan and Nebraskan times (Frye and Leonard, 1952:189-190). Populations that have subsequently become isolated in those rivers could be accounted for in this way. Flathead chubs could have entered the Rio Grande-Pecos system by stream-capture from the Arkansas System, in northeastern New Mexico or southern Colorado. H. g. gracilis undoubtedly entered the Saskatchewan and Mackenzie basins from the upper Missouri Basin, following glacial retreat (Walters, 1955:347).