Judith M. Bradow, USDA, ARS, SRRC, New Orleans, LA

Philip J. Bauer, USDA, ARS, Florence, SC

Allen K. Murray, Glycozyme, Inc., Irvine, CA

Richard M. Johnson, Texas Tech Univ., Lubbock, TX

ABSTRACT

The book, Studies of Quality in Cotton, was published by W. Lawrence Balls in 1928 as a summary of a ten-year research project carried out in Egypt, at the Rothamsted Experimental Station, and in Lancaster, England. Initially, the project mission was the translation of cotton spinning into terms that would enable cotton breeders and growers to provide cotton fiber that had those properties which increased processing success. Later, Balls added the objective of constructing a fiber property-based "Prediction Formula" by which the cotton grower could anticipate the results of a spinning test. Over the ensuing seventy years, both fiber physiologists and textile technologists have frequently cited this book and other reports published by Balls between 1912 and 1928. Such citations are usually used to provide historical background for the modern studies performed with instruments that are more versatile or powerful than the light microscopes, mechanical balances, and prototypic fiber sorters available to Balls in the second decade of the Twentieth Century. However, it should be noted that, despite very real technical limitations, Balls laid out the basic premises underlying cotton 'fiber quality' and identified or even defined the rudiments of the quantitative relationships between fiber properties and yarn processing results. This report contrasts the methodology and results described by Balls in 1928 with results currently being obtained through modern fiber quality measurement techniques. In this review, as in Balls' research, particular emphasis has been placed on those results that provide predictive insights into the relationships between fiber properties and yarn strength.

INTRODUCTION

In 1928, Fellow of the Royal Society, W. Lawrence Balls, Sc.D., published a textbook-style summary of the ten-year research project he carried out in cooperation with the Khedivial Agricultural Society of Egypt and the Egyptian government (Balls, 1928). The Egypt-based portion of the project was terminated in 1914 when the political situation prior to World War I prompted Balls to return to England. There he continued his studies of cotton fiber quality at the Rothamsted Experimental Station and in Lancashire under the auspices of the Fine Cotton Spinners' and Doublers' Association, Ltd.

The original project mission was the 'translation' of cotton-spinning technology and phraseology into concepts and definitions that might aid cotton breeders and growers to provide the textile industry with cotton fiber with those properties deemed most likely to improve spinning success and yarn quality. The starting point and basis for this 'textile' project was the research in economic botany and cotton physiology of Egyptian cotton that Balls began in 1904 (Balls, 1912). The knowledge of cotton genetics and physiology that he acquired in his earlier project prompted Balls to expand the spinning-properties project to include the construction of a fiber property-based "Prediction Formula" to be used by both cotton producers and processors. It was his intention that such a formula would allow cotton growers [and the purchasers of cotton from those growers] to anticipate the results of a spinning test on the basis of a few 'simple' fiber-quality determinations (Balls, 1921).

Fundamental Relationships between fiber and yarn properties

For more than seventy years, the cotton quality reports published by Balls between 1912 and 1928 have been cited by fiber physiologists and textile technologists. In most instances, those bibliographic citations have been used as historical background for more 'modern' studies performed with new instruments that are represented as being more versatile or more powerful than the basic light microscopes, mechanical balances, and prototypic fiber sorters that were available to Ball. However, Ball's basic premises concerning the relationships between fiber and yarn properties have frequently been overlooked or ignored by the authors of 'high-tech updates' of fiber and yarn property research.

Working with simple, but extremely labor-intensive, methods, Balls hypothesized that the overall relationship between fiber and yarn properties depends almost entirely on three fiber properties.

1. Fiber fineness by weight.
   Balls realized quite early that fiber fineness and weight are important factors in yarn strength because, as stated in his First Paradox, "the weaker the hair, the stronger the yarn" (Balls, 1915). From his own data, he inferred that weaker fibers also weighed less and that lightweight, thinner fibers spun into better yarn because a higher number of finer fibers could be packed into the yarn cross-section. Balls defined the 'Intrinsic Strength' of a yarn as the maximum possible strength obtainable from spinning a given cotton bulk sample. He described Intrinsic Strength as corresponding to "hair-strength per equivalent weight" (Balls, 1928). His definition of yarn Intrinsic Strength also depended upon the assumption that each individual fiber "takes an equal share of the breaking load." That even distribution of the breaking load, in turn, was governed by the degree to which the fibers in the yarn cross-section adhered without slippage within the yarn.
   
   The adhesive properties of the fibers in the yarn cross-section also depended upon the yarn twist-angle, a yarn property that interacts with the fiber frictional coefficient or slipperiness, which Balls considered to be the second most important fiber property.
2. Fiber slipperiness
   Ball described 'slipperiness' as the coefficient of friction of one fiber moving against neighboring fibers, and he lamented the absence of a satisfactory method for quantifying the frictional coefficient of fibers (Balls, 1928). He found that rubbing single fibers on glass plates increased the fiber-glass contact area unrealistically. Also, the transfer of fiber surface wax to the plate during the test altered the frictional coefficient of both the fiber and of the glass test plate. The technology available for determining fiber-to-fiber frictional coefficients proved equally disappointing. Although Balls recognized that confounding factors were being introduced in the mill, he eventually resorted to an empirical method of using untwisted roving for estimating fiber frictional coefficients. After compensating for the twisting introduced during roving production, the average slipping load per fiber was calculated by dividing the slipping load of the roving by the average number of hairs in the roving cross-section. That latter value was obtained from replicated determinations of fiber fineness by weight.
   
   Balls saw fiber slipperiness and fineness as significant factors in the distribution of fiber 'weak links' and convolutions during drafting (Balls, 1928). For him, "the most important single fact about cotton-spinning "... is this, -- that singles yarn is unstable." Although he was aware that yarn doubling reduced this inherent instability, he maintained that the frequency and distribution of weak links in singles yarn depended on the slipperiness of the fiber. According to Balls, the more slippery the individual cotton fibers within a sample were, the more 'draftable' that fiber sample was and the higher the strength of the resulting yarn should be. Unfortunately, he found that, after the final drafting step, high fiber slipperiness became a negative factor in yarn strength and spinning success. At that point in the spinning process, fibers that adhere to each other with a minimum of twist are needed if the intrinsic tensile strength of the cotton fibers is to be realized. Balls called this the Second Paradox of cotton spinning.
   
   In the fiber-yarn paradigm that Balls developed, fineness by weight and slipperiness are both double-weight factors to which he added a third major factor, the 'strength' of the cellulose wall of the fiber (Balls, 1928).
   
   
3. Strength of fiber cell wall.
   During his earlier physiological studies of Egyptian cotton, Balls noticed that heaviest lint was repeatedly placed in the 'strong' fiber category by hand-classers (Balls, 1912). He also found a close correlation between lint and seed weights and noted that gin-out was highest for the strongest fiber class. Since the weight of a fiber necessarily varies with the length, Balls used 'weight per unit length' to eliminate genotypic and natural variations in length (Balls, 1928).
   
   Balls also examined the differences between the strengths of primary and 'secondary' fiber cell walls and found that the breaking strength of the secondary wall at harvest, 70 days after floral anthesis [dpa], was 2.5 times the strength of the 'primary' wall at 35 dpa. With the limited analytical power available to him, Balls had to confine his determinations of primary-to-secondary cell wall ratios to fibers collected just before boll cracking and fibers from fully opened bolls, which were deemed mature at harvest. He did, however, recognize that fiber walls began to thicken with secondary wall deposits at approximately 21 dpa, depending on environmental conditions. More recent reports indicate that secondary wall deposition commences around 14 to 16 days post anthesis (Stewart, 1986), indicating that Balls' primary-to-secondary wall weight and strength ratios may have been even more underestimated. The weight of the secondary wall deposits in Upland fibers is already significant at 35 dpa (Wartelle, et al., 1995).
   
   The period of Balls' research predated the invention of the electronic microbalances available now for weighing single fibers, and his comparatively crude light microscopes made fiber cross-section measurement both imprecise and extremely tedious. He found that even his best estimates of fiber cross-sections were further confounded by the wide variation in the shapes of fiber cross-sections, a natural fiber-shape property that he related to variations in the amount of wall thickening (Balls, 1928). Balls avoided the confusing implication that cotton fibers have circular cross-sections with true diameters by introducing the term, 'ribbon width' to represent the major axis of the ellipsoid formed by the collapse of a mature fiber with a thickened wall.
   
   For Balls, fiber length was the most easily quantified fiber property, but he found that measurement of this "simplest" fiber-quality factor was biased toward the longest fibers and that his length determinations were seldom reproducible (Balls, 1928). He also found gross changes in length introduced by fiber breakage or stretching during testing and processing. Further, he determined that the natural variability in fiber length was considerably greater than what he had expected, based on his experiences with skilled hand-classers and the 'halo' or watch-glass methods for measuring the length of fibers while they are still attached to the seed. Balls was forced to conclude that fiber length, although the most easily measured fiber property, was definitely less important than the three factors discussed above, i.e., fiber fineness, slipperiness, and cell wall thickness.

In the Introduction to his book, Balls apologized for departing from the norm for textbooks on cotton spinning. However, he felt strongly that his topic, the behavior of the fiber itself, could not be covered adequately if the prevailing focus on mill machinery were used (Balls, 1928). Throughout his book, he stressed that "the cotton hair is taken as the unit in which all spinning problems must be formulated and solved, and although this seems rather obvious it is really rather novel. As farmers, chemists or spinners, we have looked upon cotton lint as cellulose to be purified, or as bales of raw material; only the botanists have thought about it as discrete hairs until quite recently. Now, the uniform appearance of a lump of cotton is deceitful, for not two hairs in a thousand are reasonably similar; even in such a generous interpretation of similarity as would allow the reader to accept a filbert nut instead of a Kentish cob, not two in fifty are alike."

The inherent and undeniable variability of cotton fiber populations forced Balls to concentrate on developing and, frequently, inventing methods for sorting the complex population of fibers into groups with similar characteristics. Only then could he hope to resolve the spinning problems arising from either bulk fiber properties or from the natural variations in fiber quality that are combined in the bulk averages. "Meanwhile, we are the victims of this variety; the cotton behaves on the average as if it were uniform, and the peculiarities of individual hairs are not notable until the final yarn is reached..."

In 1928, quantitation of the three measurable fiber characters, fineness, slipperiness, and cell wall thickness, as well as fiber length measurement, depended upon classer's judgement, light microscopy, mechanical balances, and, eventually, upon the 'sledge' sorter that Balls invented for sorting fibers by length. Working as he did before World War I, Balls could only dream that future developments in fiber quality instrumentation would make obsolete his predictions that there "... is no likelihood that any elaborate system of hair testing can ever supersede the grader's rapid handling..." Based on his own experience, he did have hope, however, that "...it is not improbable that the quick and precise ... measurement of fineness by weight might become routine..."

In his studies of classers' standards and the precision of hand stapling, Balls constructed length by weight distributions (Balls, 1921, 1928). In Figure 1 are shown the 'actual' length distributions of three staple standards from 1918 USDA Arizona Gossypium barbadense cottons. The classers assigned staple lengths of 1-1/2, 1-5/8, and 1-3/4 inches to Samples 1 through 3, respectively. Sample 2, the 1-5/8-inch standard differed very little from Sample 3, the 1-3/4 inch standard, a situation with obvious commercial impact since differences of an eighth of an inch were [and still are] used in setting cotton lint prices.

As the high degree of variability in fiber length became more and more evident, Balls realized he must find ways to quantify and analyze fiber-quality variation if he were to examine effectively the "real relationships" between the fiber grades assigned by classers and the fiber properties that determined processing quality. "What does the grader mean by 'strength'? What relation does his opinion bear to the performance of the cotton in the spinning mill?" (Balls, 1928). Balls had shown that "grader's strength" had no direct connection to the breaking strength of a fiber (Balls, 1915). Further, he found that the precision of the classers' cotton grading system broke down when new cotton genotypes with different 'feel' were introduced (Balls, 1928). Thus, when confronted with lint of familiar genotypes, a skilled classer could make "brilliantly correct" inferences from the fiber of any genotype with which he had prior experience. Indeed, one classer both correctly identified a genotype that was no longer grown and named the region and town that produced the lint sample. However, that same highly skilled classer misjudged the strength of new genotypes with which he had no prior experience and recommended that those new cottons be eliminated from the breeding trials.

Balls was clearly impressed by the skills of experienced hand-classers, but he came to the conclusion " that hand and eye grading at its very best is unreliable." He also found that this unreliability was compounded by the assumption that fiber quality is uniform throughout a bale and, correspondingly, quality is uniform within a classer's sample drawn from that bale. Balls used his 'sledge' fiber length sorter, which he described in detail in his monograph on measuring cotton fiber length (Balls, 1921), to check classer's estimates of eight official length 'standards' representing successive thirty-seconds of an inch (Balls, 1928). However, the sorted samples did not fall into the length sequence determined from the classer's hand grading. Further, the fiber length variations within the examined samples were so great that Balls concluded, "Grading to steps of thirty-seconds would therefore seems to be quite impossible, even if a very modest degree of probability is desired."

In the early decades of the Twentieth Century, genotypic differences in fiber length were reported as the length of the longer hairs. Consequently, the accepted 'world cotton' fiber 'length' range was 0.5 to 2.5 inches (Balls, 1921). However, Ball's fiber-length sorter produced reproducible bell-shaped distributions of fiber length by weight for every genotype that he examined. The obvious variations in fiber length in genotypes that had been assigned 'nominal' staple lengths by hand-classers were inferred to be length modulations caused by growing conditions (Balls, 1928).

However, meaningful studies of the interactions between genotype and environment required more powerful instrumentation that could handle small fiber samples from individual bolls, locules, and even individual seeds and seed coat zones (Bradow et al., 1997a; 1997b; 1997c). The variations in fiber length within two locules from the same boll of PD3, a G. hirsutum genotype, are shown in Figure 2. The fiber length distribution within a single locule follows the same normal distribution that was seen by Balls in the 1918 USDA length profiles [Figure 1.] The data plotted in Figure 2 were collected using the AFIS Length & Diameter module (Bradow et al., 1997a, 1997c].

As Balls noted, the length of a fiber is "a fairly obvious thing." Since staple length could be determined with the unaided eye, the textile industry assigned more importance to fiber length than would appear to be warranted by the relatively small contribution of fiber length to yarn properties and processing success. Recognizing that fiber fineness by weight and cell wall thickening were important components of the fiber property yarn property relationship, Balls developed methods for measuring ribbon-width, wall thickness, and cross-sectional area (Balls, 1928). After his return to England, Balls had begun concentrating his study on "The Thirteen Samples" of American Upland cottons, which were given to him by USDA. The Thirteen Samples from the 1918 crop included seven new genotypes and three varieties, Express, Sunflower, and Polk, which were commercial in 1918, plus three duplicates.

Since yarn strength received the most attention in Balls' investigations, he performed seven sets of spinning tests in order to eliminate mill variability. The last two spinnings of 90s full twist yarn were in perfect agreement. Only then did Balls begin his exhaustive study of the yarn and roving by ranking the samples according to yarn strength and applying simple inspection methods to detect correlations among the quantified fiber characteristics and yarn strengths of the individual samples in the set of thirteen.

In Figure 3 and subsequent discussions of The Thirteen Samples, the order of presentation for the ribbon width and other data was determined by the ascending yarn strength. Thus, Sample 20 was the weakest yarn and Sample 16 was the strongest. [Sample numbers run from 10 to 22, not 1 to 13.] Fiber ribbon width was quantified by embedding the fibers in hot celloidin [pyroxylin or cellulose nitrate] followed by paraffin, and cutting fiber cross-sections with a microtome. Each ribbon width measurement was the mean of at least 100 fibers, and both ribbon widths and wall cross-sections were measured as close as possible to the middle of the longest fiber dimension.

Balls noted that fiber ribbon-width was not closely correlated to yarn strength unless one considered only the ends of the range, i.e., the weakest yarn was made from Sample 20, which had one of the higher ribbon widths and the strongest yarn was made from Sample 16, which had one of the narrowest ribbon widths. However, the broadest ribbon width was found in Sample 10, which was one of the stronger yarns. These ribbon-width data did, however, roughly approximate the empirical relationship that related finer fibers to stronger yarn.

The ribbon widths in Figure 3 are means of 100 fibers drawn randomly from the roving of The Thirteen Samples (Balls, 1928). The distributions of fiber diameters or ribbon-widths in Figure 4 were obtained from 10,000-fiber samples obtained by roller-ginning the individual locules of PD3 grown in Florence, SC, in 1992 (Bradow, et al., 1997c). These data show that the variability in diameter within a single boll of one genotype can be as large as the variability among The Thirteen Upland Samples characterized by Balls. Average PD3 fiber diameter ranges from ca. five to seven micrometers, considerably lower than the mean ribbon widths reported by Balls for The Thirteen Samples.

Just as Balls was limited by the rudimentary instrumentation available to him, he was also handicapped by the primitive state of statistical analysis during the period between the two World Wars. Simple means and standard deviations were useful for making basic comparisons among results, but he recognized that he could not acquire enough replicated data to make simple correlation analyses reliable. Therefore, Balls resorted to a series of visual comparisons among graphs of fiber properties as Intrinsic Strength, staple length, cell-wall thickness, and yarn strength for The Thirteen Samples (Balls, 1928).

Figure 5 shows overlay plots of the yarn strength [heaviest line, solid diamond symbol] of The Thirteen Samples and the corresponding staple lengths, cell-wall thicknesses, and Intrinsic Strengths of the fibers. The yarn strengths form a fairly linear progression from low-strength Sample 20 to high-strength Sample 16, and the actual range of the mean yarn strengths is 31.0 to 36.0 g breaking load [lea strength]. Plots of wall thickness [and ribbon width, shown in Figure 5] decreased across the sample yarn strength series. The fiber Intrinsic Strength trend was generally upwards from the lowest yarn strength found for Sample 20, but Samples 17, 13, 19, and 10 appeared as outliers in the plot of Intrinsic Strength. Staple length was completely unrelated to either yarn or fiber strength in these plots (Balls, 1928).

Balls was particularly concerned about the confounding artifacts found in his microscopic measurements of ribbon width, wall-thickness, and cross-sectional area (Balls, 1928). "In the ordinary routine of the testing laboratory it is only practicable to make these measurements upon undisturbed hairs, viewing in profile under the microscope, and measuring the ribbon and wall in at least a hundred hairs, preferably at one place only, near the middle of each hair." He felt that his ribbon width measurements were reasonably accurate, but he found accurate measurement of fiber wall-thickness much more difficult. Mounting the fiber in water resulted in a doubling of the true wall thickness, due to refraction of the light passing from the water to the wall and out again. The wall itself was also doubly refracting, and precise definition of the cell lumen [central canal] was, therefore, impossible.

Balls measured wall thickness in conjunction with ribbon-width determinations by mounting the fiber in a liquid [aniline and xylol] with the same refractive index as the wall and overcoming the near invisibility of the fiber lumen by polarizing the incident monochromatic light and viewing the fiber through a nicol prism. Thus, the width of the cell lumen could be determined, and the wall thickness calculated by subtracting the lumen width from the ribbon width and halving the difference. The cross-sectional area was calculated, with less than 5% error, by assuming that the cross-section can be approximated by two semicircles connected by a rectangle according to the formula: (Balls, 1928). This estimate was considered sufficient for the ordinary comparison of similar genotypes, but significant variations in fiber cross-section and wall thickening were obscured by the assumptions of fiber shape and limitations in the number of fibers that could be examined from each sample.

In 1928, Balls reported genotypic differences in fiber cross-section ranging from 130 to 450 µm² and a range for fiber-to-fiber cross-sectional variation of 50 to 500 µm² within a single sample. The variability within sample he attributed to "fiber life history", i.e., growth environment (Balls, 1928). Quantitation of genotypic variability and realization of the dependence of fiber cross-sectional area upon both genotype and growth environment required the development of instruments that could process statistically significant numbers of fibers while rapidly and reproducibly quantifying fiber properties (Bradow, et al., 1996). Figure 6 shows the range and distribution of PD3 fiber cross-sections in Position 1 and Position 2 bolls from fruiting branches seven through eighteen. The fiber cross-sectional areas of this modern cotton genotype fall considerably below the lower end of Balls' cross-section range for the genotypes in his 1928 study. The variation in fiber cross-sectional areas found among these fruiting positions that produce over 80% of the crop is comparable to the range of areas Balls reported for The Thirteen Samples.

As early as 1825, the extractable 'cellulose' from plant cell walls had been characterized as a mixture of substances called cellulose and pectose, a clear, but too frequently overlooked, acknowledgement that the extractable material was not pure cellulose (Preston, 1974). The first use of cuprous ammonium hydroxide as a swelling agent to remove cellulose from cell walls has been credited to Fr--my in 1859. Subsequent staining and solubility tests localized the 'pectose' in the wall middle lamella, which is found between the walls of adjoining plant cells [but does not occur between individual cotton fibers]. Mangin in the 1890's demonstrated the presence of two series of plant cell wall compounds - acidic and neutral - that both differed in chemical composition from cellulose and were often placed in the nonspecific class of hemicellulosic compounds. All these chemical analyses were performed on entire walls since the methods needed for compositional studies of separated primary and secondary fiber walls were not available.

Balls did not have access to the chemical analyses that would later reveal the compositional differences between the primary and secondary cell walls of plants (Goodwin and Mercer, 1983). He did, however, note the "spongy structure" of the fiber wall; and he was aware of the discrepancy between cotton fiber density determined by measuring the fiber cross-sectional area [p = 1.00 g cm^-3] and the density of 'pure' cellulose [p = 1.55 g cm^-3] (Balls, 1928). Using cross-sectional area measurements of a number of genotypes grown in a variety of years and locations, Balls arrived at a range of 0.9 to 3.6 µg per cm length for the fiber weight per unit length, a parameter he found useful in comparisons among genotypes.

Balls used sequential harvests during fiber maturation to track the evolution of the fiber cell wall. He followed the progression from the thin primary wall, which is covered by an external layer of waxy cuticle, to the thickened wall of the mature fiber in which the central, lumenal space is almost completely filled by secondary wall deposition (Balls, 1915; 1928).

He also determined that secondary wall deposition was a discontinuous or cyclic process that produced diurnal growth rings. Discrete, concentric rings could be seen in cross-sections of fibers that had been swollen in cuprous ammonium hydroxide. [See cartoon at right where the relative thickness of the fiber outer wall is a recurring artifact of sample preparation and differential response to the swelling agent. "The refractive index of the primary wall and cuticle is decidedly higher than that of the secondary wall."] Balls noted that the walls of mature fibers had approximately 30 rings, corresponding the four to five weeks needed for fiber to mature. Since he could partially suppress ring formation by growing the cotton plants under continuous light, Balls inferred that each ring represented the secondary wall deposit of a single 24-hour period. Thus he described these 'diurnal rings' as analogous to the annual rings of tree trunks with two exceptions: "that they represent each the growth-deposit of one night, and that the inner-most ones are the last to be formed."

Chemical analytical methods and instrumentation have, of course, been vastly improved over the past 70 years, and the chemical composition of plant cell walls, including those of cotton fibers has been more fully defined (Table 1). Resolution of the complex, water-insoluble mixtures that form the surface waxes and external cutin layer was made possible by advances in chromatography. The primary wall "cellulose" matrix was identified as a matrix of cellulose chains, the various xylans, mannans, and galactans of the cell wall hemicellulose, and the cell-wall matrix pectins, i.e., polyuronic acids, arabinans, and galactans (Goodwin and Mercer, 1983).

The calcium cation, listed in Table 1 as a component of the primary wall matrix, is a cross-linking agent in the pectic matrix of the primary wall. Because cotton-fiber calcium

is found only in the protoplasm and the primary wall, the primary-to-secondary wall weight ratio can be determined using x-ray fluorescence spectroscopy of Ca-XRF (Wartelle, et al., 1995). Immature fibers and callus tissue have high primary-to-secondary wall weight ratios and, therefore, high calcium concentrations, compared to those found in mature fibers in which the deposition of secondary wall polymers, which do not contain calcium, has 'diluted' the calcium present in the primary wall and fiber cell lumen.

During the seventy years since publication of Balls' 1928 book, a great deal has been learned about the crystalline structure of cellulose, the major component of the fiber secondary wall. However, there is still much that is not understood about the synthesis of cellulose and the role of cell-wall proteins in that synthetic process.

Recently, new instrumental techniques have allowed carbohydrate chemists to follow deposition of the fiber wall at the molecular level (Murray and Brown, 1997). Qualitative and quantitative monitoring of glycoconjugate polymer precursors (mer~ in cartoon at right) has led to characterization of the precursors and determination of the polymerization rates during secondary wall deposition and fiber maturation. Temporal shifts in the fiber glycoconjugate profiles coincide with shifts in wall composition and refractive index that would result in the diurnal rings such as those first seen by Balls in 1915 (Balls, 1928).

In his discussion of the light-microscopic pictures of the diurnal rings in cotton fiber cross-sections, Balls described the visible construction units of the secondary wall as stacks of "dominoes" adhering together along the ends, fronts, and backs but free at the sides (Balls, 1928). Each unit was seen to be approximately 0.3 µm thick. Balls offered these comments on the probable structure of a fiber wall fibril. "I am not at all confident that these domino shapes are real unitary structures, though there would seem to be good reason to think that the living cell does build its wall with bricks of this kind. The unit may be much shorter than the [visible] domino, for slip spirals may be more frequent; the fibril between one pit and the next may be compound, as if the domino were made up from thin spills lying side by side; lastly, the individual growth-ring is in all probability a series of molecular layers, as if the domino had been sawn out of a piece of laminated ply-wood. However, it can be confidently asserted that such unitary structures exist, and that they cannot be larger than the limits indicated.

The next step in this research will be to get inside these limits trying to split up this structure, if it can be split further, into the real units or 'micellæ.' Then, investigating still further, we can attend to the arrangement of the molecules within the pseudo-crystal of each micelle, ascertain whether all the micellæ are alike, and in so doing come down to the atomic structure. This will probably be done fairly soon; possibly even before this book is published." That last prediction of Balls proved overly optimistic, but significant progress is being made toward determining the glycoconjugate profiles of the cell wall units and the kinetics of cell-wall deposition.

Balls was well aware of the extent to which wall thickening was important to spinning success when he stated that it is wall thickness that determines the physical fineness of the fiber. Cell wall thickness, of course, varied markedly in the individual fibers that he examined microscopically. "Within a group of only a hundred hairs from the same seed we can find the wall-thickness varying four- or five-fold." (Balls, 1928) However, his microscopic studies did not allow Balls to demonstrate a clear quantitative relationship between fiber shape and yarn density, and he considered yarn fineness by weight one of the two most important fiber-quality factors in yarn strength. "...[When] further research has demonstrated the existence of connections between fiber shape and the density of the yarn ... it will be necessary to take the microscopic measurements more seriously."

Seventy year later, the electron-optical particle-sizing capabilities of the AFIS Fineness and Maturity Module are providing direct measurements of fiber shape [circularity] and cell wall thickness in the form of theta. Theta = 1.0 for a perfect circle, and theta 0.52 for fully mature fibers of Upland genotypes (Bradow et al., 1996). Rapid, reproducible quantitation of theta has allowed correlation of fiber circularity and yarn breaking strength for fiber from four different genotypes [Deltapine 20, 50, Acala 90 and 5690] that were grown in Florence, SC, in 1991 and 1992 (Bradow et al., 1997b). In Figure 7, the genotypic variation in fiber circularity within the two crops and among three planting dates can be seen in the scatter-plot data points. The linear correlations between the fiber circularity and yarn break strength in the two crop years are represented by the solid regression lines (Bradow et al., 1997b). Yarn strength increases with increasing fiber circularity and, therefore, with increased cell wall thickening. The more mature the fiber, as indicated by a higher theta, the higher the yarn breaking strength is.

Similar quantitative relationships can be derived for yarn strength and cross-sectional area measured as Area by number, A(n), by the AFIS Fineness and Maturity Module (Figure 8). Again, the experimental design involved the four Deltapine genotypes, the 1991 and 1992 crop years, and three planting dates. Ball's First Paradox, finer fiber makes stronger yarn, can be seen in quantitative form in the A(n) data of Figure 8. As the cross-sectional area, A(n) increased, the and the yarn breaking strength [and the Fine Fiber Fraction] decreased.

The term 'micronaire' was unknown to Balls. However, he did adapt the porometer used by plant physiologists to measure air-flow resistance of leaf stomatal for use in estimating the variations in the air-ways between parallel fibers and, thereby, the fiber cross-sections. Air-flow resistance of a fiber plug is, of course, the basis of micronaire, an empirical combination of fiber maturity and fineness factors that has become a major factor in modern cotton production and processing. Using the micronaire analog, micronAFIS, which is generated by the AFIS Fineness and Maturity Module, the relationship between micronaire and fiber strength can be visualized in Figure 9. Although there is considerable variability in micronaire within and between crops, micronaire and yarn breaking strength were positively correlated (Bradow et al, 1997b). The positive correlations between yarn strength and micronaire in Figure 9 suggest that the maturity of the fiber, represented by the increases in circularity in Figure 7, outweighs fiber fineness, which is seen as decreasing A(n) in Figure 8, in the determination of yarn strength.

When Balls considered fiber maturity at all, it was as chronological maturity, i.e., the number of days elapsed between flower anthesis and harvest. During his developmental studies based on sequential fiber harvests, he followed secondary wall deposition (analogous to increasing theta) and detected some anomalies. For example, "Secondary wall formation can fail altogether in the extreme case, so that the ripe hair consists only of primary wall. Hairs such as these, together with those that have but little secondary thickening, are industrially important because the primary cellulose reacts differently to dyes; further, these hairs are so flexible that they roll up easily into the knots and tangles called 'neps' in the mill. Such a 'nep' ... does not exist in the living boll, but is made by handling, by ginning, and especially by the carding machine, from the hairs with unduly thin walls." (Balls, 1928).

Balls went on to verify his statement concerning the increase in nep counts due to carding by counting the neps found in 0.1 g of fiber at various processing stages from the ginned lint to the finishing draw-frame (Balls 1928). His results are shown in Table 2, below. In his study, the card raised the nep count from 500 to 1,000 per gram with 25% of the post-card neps appearing in the card waste. The combs did not alter nep content but did eliminate 80% of the neps as waste after the first comb and 30% of the remaining neps after the second comb. "The nep-making proclivities of the card are, of course, well known, though I do not think that its inefficiency as a nep-remover is fully realised.

Any similar rubbing action will make a nep..." The truth of the last sentence was clearly demonstrated to Balls when an early model of his Sledge sorter converted the entire sample into nep.

Clearly, Balls was aware of the variability inherent in fiber properties, particularly in those characteristics related to maturity. He further inferred that significant linkages existed between those variations and both the spinning and dye-uptake properties of the fiber, yarn, and finished fabric. However, quantitation of fiber variability was far beyond the analytical instruments available to him. Today we can only wonder how Balls would react to the types and ranges of the variations in fiber properties found in a 1996 Site-Specific Management study done in Florence, South Carolina (Bradow et al., 1998). Table 3 shows the mean, standard deviation, and minimum-maximum range of some important fiber properties, including fiber-breaking strength from that study. Immature Fiber Fraction is that percentage of fiber with theta 0.25.

No strong correlations were found between fiber breaking strength and the individual fiber properties listed in Table 2. The strongest correlations were negative, i.e., between fiber strength and micronAFIS and between fiber strength and area. As Balls noted, the strength of the individual fibers contributes to the yarn strength, but those fiber properties that determine yarn strength are not necessarily the same properties that govern fiber strength.

This review of the cotton fiber research performed by W. Lawrence Balls, Sc.D., F.R.S. was initially envisioned as a demonstration of the advances made over seventy years in cotton fiber-quality measurement. However, comparisons of the results that Balls obtained using the rudimentary instruments of the 1920's and research results published in the 1990's showed repeatedly that Balls was the first to report fiber-quality phenomena and relationships that modern scientists are just beginning to quantify, interpret, and apply in textile manufacturing. Even more disconcerting to a scientist currently engaged in fiber-quality research are the scope, accuracy, and prescience of Ball's interpretations and inferences. Thus, this review became less a retrospective describing advances in fiber-quality research and more a catalog of has been forgotten or overlooked since Balls published his book in 1928.

Working as he did at the interface of 'pure' and 'applied' research, Balls must have frequently defended the appropriateness of his research as a botanist who was studying textiles. Indeed, he explained, at length, why and how he did his fiber-quality research (Appendix, Balls, 1928). He declared that the "three prime essentials of research work, in which it may be free [pure] or restricted [applied], can be denoted as the Method, the Subject, and the Aim." One or more to these 'essentials' may receive particular emphasis in of the experimental design. However, research initially conceived as a 'philosophical' exploration of a Subject, e.g., cotton fiber quality, may be transformed into Method development project that ultimately supplies solutions to problems that would have been emphasized as the Aim of applied research.

Balls also saw his research as providing a means for communication between agronomy [the Hoe] and textile technology [the Spindle]. His expressed research Aim was described thusly in the Introduction to his 1928 book: "...when scientists brought the cotton industry within the ambit of their curiosity they soon found that the Hoe could not do its best for the Spindle unless the Spindle sent word and description of its needs to the Hoe. The scientists had to examine them both... [I]n 1915 it was clear that the farmer never could produce good quality to order unless the spinner became able to explain exactly what was wanted... that the grower might know what kind of yarn his product was capable of making before ever it came near a spinning mill."

Working as he did at interfaces between various scientific disciplines and between basic and applied science, Balls gave considerable thought to the concept of teamwork in scientific research. Some clues to the success of his research can be found among his "maxims of research organization." For example, "The specialist in applied science must of necessity become a 'Jack of all trades,' widening his range of superficial knowledge as his specialisation increases." Also, "Every new problem presented to the team needs a new method. Perhaps the greatest advantage of teamwork lies in the availability of mutual help for this purpose." Balls put this maxim into practice when he adapted the Barr-Anderson monochromatic-light method for measurements of the internal diameter of glass tubing for measuring the thickness of fiber walls. He also used the porometer, which was developed by Sir Francis Darwin for measuring stomatal apertures in plant leaves, to determine the "variations in the air-ways left between parallel hairs" and thereby the changes in fiber cross-section.

The real reasons that Balls' research has stood the test of time can be found in the Appendix chapter in which he defines The Researcher's Code (Balls, 1928). That code can be summarized by the following key words.

1. Replicated - Every determination must be repeated. Every sample must be replicated.
2. Representative - Quantitative results must be presented as means with standard errors. Qualitative results must show examples of the 'average' phenomenon, not unique individuals or infrequently seen artifacts.
3. Reproducible - The phenomenon must reappear when experimental conditions are reproduced.
4. Rigorous - More than one method and approach must be applied so that confounding factors such as sample preparation artifacts and instrument errors are not reported as past of the phenomenon being investigated.

"The Code is not a mere convention. It is as essential to the advance of natural knowledge as the Experimental Method itself. The latter can produce isolated pieces of knowledge, the former provides automatic mechanism for co-ordinating such fragments. Also it provides the stimulus which drives men to continue to do research, to write books on the results, to organise such implements of communal knowledge as journals, conferences, indexes, and museums, as well as their own laboratories." (Balls, 1928)

Applying the necessary rigor of The Researcher's Code to his own work, Balls managed to recognize that the inherent and the introduced variability of cotton fiber required examination of as many samples as labor and time permitted. When quantitation of a defined fiber property was needed, he sought methods and instrumentation in every scientific and industrial specialty. Working with simple and, frequently, prototypical instrumentation and relying heavily on inference, Balls laid down the scientific foundations of modern cotton fiber quality and the defined the rudimentary relationships between fiber properties and processing success. Seventy years later, W. Lawrence Balls is the father of fiber quality research, the Robert Goddard of EFS® and the Zefram Cochrane of SPECK TREK™ .

Balls, W. L. 1912. The Cotton Plant in Egypt. MacMillan and Co., Ltd., London.

Balls, W. L. 1915. The Development and Properties of Raw Cotton. A.C. Black, Ltd., London.

Balls, W.L. 1919. The existence of daily growth rings in the cell wall of cotton hairs. Proc. Roy. Soc. London B 95:72-89.

Balls, W.L. 1921. A Method for Measuring the Length of Cotton Hairs. MacMillan and Co., Ltd., London.

Balls, W.L. 1924. The determiners of cellulose structure as seen in the cell walls of cotton hairs. World Cotton Conf., Liverpool and Manchester, UK.

Balls, W.L. 1928. Studies of Quality in Cotton. MacMillan and Co., Ltd. London.

Bradow, J.M., O. Hinojosa, L.H. Wartelle, G. Davidonis, G.F. Sassenrath-Cole, and P.J. Bauer. Applications of AFIS fineness and maturity module and x-ray fluorescence spectroscopy in fiber maturity evaluation. Textile Res. J. 66:545-554.

Bradow, J.M., P.J. Bauer, O. Hinojosa, and G.F. Sassenrath-Cole. 1997a. Quantitation of cotton fibre-quality variations arising from boll and plant growth environments. Eur. J. Agron. 6:191-204.

Bradow, J.M., P.J. Bauer, G.F. Sassenrath-Cole, and R.M. Johnson. 1997b. Modulations of fiber properties by growth environment that persist as variations of fiber yarn and quality. Proc. Beltwide Cotton Conf., 1351-1359.

Bradow, J.M., L.H. Wartelle, P.J. Bauer, and G.F. Sassenrath-Cole. 1997c. Small-sample cotton fiber quality quantitation. J. Cotton Sci. 1:48-58.

Bradow, J.M., R.M. Johnson, P.J. Bauer, and E. J. Sadler. 1998. Modulation of economically important cotton fiber properties by field spatial variability. Proc. 4^th Intl. Conf. on Precision Agriculture, St. Paul, MN. 18 -22 July, 1998. (In press)

Goodwin, T.W. and E.I. Mercer. 1983. Introduction to Plant Biochemistry, 2^nd Edition. Pergamon Press, Oxford, UK. Stewart, J. McD. 1986. Boll Development. p. 261-300. In J.R. Mauney and J. McD. Stewart (ed.) Cotton Physiology. The Cotton Foundation, Memphis, TN.

Murray, A.K. and J. Brown. 1997. Glycoconjugate profiles of developing fibers from different fruiting branches of the same plant. Proc. Beltwide Cotton Conference, pp. 1496-1498.

Preston, R.D. 1974. The Physical Biology of Plant Cell Walls. 1974. Chapman and Hall, London.

Wartelle, L.H., J.M. Bradow, O. Hinojosa, A.B. Pepperman, G.F. Sassenrath-Cole, and P. Dastoor. 1995. Quantitative cotton fiber maturity measurements by x-ray fluorescence spectroscopy and Advanced Fiber Information System. J. Agric. Food Chem. 43:1219-1223.