Dietary Effective Fiber, Particle Length and Sorting



It is well known by dairy producers, veterinarians, and nutritionists that dairy cows require fiber and also that cows require a portion of their diet to have adequate physical length to promote optimal rumen function. Because the long particles consumed by cows are virtually entirely from the more fibrous parts of plants, it is common to confuse the requirement for long fiber with total fiber. The long fiber is often called “effective fiber,” but I will always call it physically effective fiber and reserve the term effective fiber for something broader, more holistic, and admittedly vaguer. Physically effective fiber is essential to provide rumen fill and prevent abomasal displacement. Physically effective fiber forms a thick rumen mat that slows passage of smaller fibrous feed particles and increases their digestion. Because most structural fibers are degraded at a slower pace than starches, sugars, and soluble fiber, slowing these fibers down by matting is an important way ruminants have evolved to digest these more recalcitrant sources of carbohydrate. Physically effective fiber enhances rumination and salivation, providing a better buffer source for the rumen. The latter is a very important concern for cows eating large quantities of feed, much of which is rapidly fermented to organic acids in the rumen. Finally, physically effective fiber is related to rumen fill and can limit feed intake. It is possible that limiting feed intake on any given day may help reduce variation in day-to-day feed intake without drastically reducing average feed intake, but the potential for reduction in average feed intake is clearly present in overly long diets. Although limiting feed intake usually is a definite negative, it is also one way to reduce the likelihood of acidosis, so this feed intake limitation can also be one of the “positive” effects of physically effective fiber. This talk focuses on how we make sure cows consume an adequate physically effective fiber. I discuss this in terms of the role that coarse fiber has in complementing the remainder of the dietary NDF. I will not discuss the importance of many carbohydrate characteristics such as NDF digestibility, starch availability and rate of degradation, the effect of starch versus sugar, or the role of non-NDF fiber. All of these characteristics are important in determining the final carbohydrate “balance” of the diet.

Please check this link first if you are interested in organic or specialty dairy production.

While ignoring particle length is just plain wrong, focusing on only long fiber in balancing dairy diets leads to erroneous thinking about the requirement for fiber. One good way to express the requirement for long fiber was to suggest that the diet has 21% of diet dry matter as NDF from forage. This was a much better recommendation than requiring 28% NDF and 75% of the NDF from forage, but it still has its problems. This 21% forage NDF guideline alone would assume that all forage had identical length. Because particle size of forages varies, guidelines exist to ensure that forage sieving results meet some minimum requirement (say 15% of as fed forage mass retained on the top Penn State screen). Separating these two requirements is problematic. If the primary role of forage NDF, versus non-forage NDF, is to provide physically effective fiber, then there must be a trade-off where less of a coarser forage (plus some non-forage, fine fiber) would equal more of a finer forage. Having separate, rigid guidelines for % forage and % above a screen does not allow for this trade-off to be done quantitatively. A less obvious problem with 21% forage NDF as a requirement is that it essentially ignores the remainder of the NDF in the diet. There is very good information that NDF from by-product feeds or finely ground forages that have no physical effect in the rumen are nonetheless valuable in balancing the rumen chemistry by providing energy without providing more starch. It is important to separate out the concepts of physically effective (that is, particles long enough to affect rumen consistency and rumination) and other useful chemical attributes that belong to both long and short fiber alike. Many short fiber particles result from the grain pericarp and legume leaves and are relatively highly digestible sources of fiber. Once we accept that there is chemical value to fiber that is separate from the value of long particles which are fiber rich, we can begin to imagine a more systematic way of thinking of meeting fiber requirements. The work in this area is nowhere near complete, but advances continue to be made. Numerous ideas can be implemented now but also with an eye toward improved systems that incorporate (and suggest) more information that should be collected.

Myth: Only the particles on the very top screens are physically effective.

Reality: Any feed or diet has a distribution of particles of various sizes. These distributions can be hard to describe in simple, understandable terms. Mean particle length and standard deviation are technically the appropriate terms to use. However, given the unusual shape of the particle size distribution, mean particle length values are actually difficult to interpret. Having seen many diets and their calculated particle sizes, I know that my own intuitive assessment of the diet is much longer than the correctly calculated mean particle length. Because of this, one easy way to describe a target for how a forage or diet should “look” was to set a minimum for the percentage on the top screen. It is not surprising that many people understood this (incorrectly) to mean that the material on the top screen was what provided all the physically effective fiber.

Let’s look at some field data on particle sizing. These data are revealing and tell us that modern forage choppers were designed by engineers who apparently all read the same textbooks because they appear to provide a very similar distribution of particle sizes. As the theoretical cut of these machines is adjusted to make coarser or finer silages, the entire distribution is pretty well described by the mean particle size but also by the % of the particles above a middle point. This is shown in Figure 1 where the relationship of mean particle size and the % of particles above the 9 mm (diagonal) screen of the Wisconsin separator is shown. A similar tight relationship exists for particles above the 5.6 mm (next smallest) screen. Finding data to measure the physical effectiveness of any particular fiber length is actually quite difficult given the strong fit shown in Figure 1 and the fact that a true precision chopping is really impossible. If we were always dealing with feeds chopped by the same broad class of forage harvester, all this might be somewhat irrelevant as all the measures would be interchangeable. However, the danger of depending on only very long fibers with no “middle screen” will be highlighted in the remainder of this talk. Providing very long fiber (particles several inches long or more) that has not gone through a conventional forage harvester will distort the relationships shown in Figure 1 and render using the % on the top screen less useful than using the % above a more “central” screen.

Figure 1. Relationship of mean particle size and % of particles above the third screen (9 mm diagonal opening) of the Wisconsin separator. A positive relationship is obviously expected, but the quality of the fit is very good, indicating that one screen in the middle of the distribution does a very good job of predicting the entire distribution. The fourth screen (5.6 mm) also showed a good relationship, but larger and smaller screens were less predictive. The regression for alfalfa and corn silage were nearly identical, and a single regression equation is shown where x is a decimal value (50% = 0.5) and y is given in mm. The TMR regression (lower line and equation) is also given and is different because the non-forage feeds in the TMR follow a different distribution than the alfalfa and corn silage, yet this regression still explains 78% of the variation in TMR mean particle size, and most of the large deviations are above the regression line.

In order to test the effect of “greater mean particle length” versus “more long particles,” we conducted the trial described in Figure 2. Basically, we harvested oatlage in such a way as to provide two diets with similar mean particle size but one diet with lots of long and short particles and one with more medium particles. Our results suggest these diets performed about the same. Long particles will raise the mean particle length and therefore the physical effectiveness of the diet. Medium particles that raise the mean particle length to the same level (which requires more mass of medium particles) were just as effective, or at least too close to measure a difference. It is important to note that Figure 2 reports the average particle size of the diet the cows were offered, which brings us to the importance of sorting.

Figure 2: Results from a trial where particle size of oatlage-based TMR was distorted by mixing finely chopped (left-hand points) and long silage (right-hand point) to achieve about 5.4 mm in TMR with a bimodal distribution (mixed long and fine) or by using a middle setting on the forage harvester (medium). These diets had similar mean particle size and similar physically effective fiber, suggesting no greatly enhanced effect of longer fibers other than what is incorporated into mean particle length. See Figure 6 for particle size actually consumed.


We have conducted a series of trials to measure TMR sorting by dairy cows. Most of these trials, especially the early ones, focused on getting observations on numerous cows while also exploring diet composition effects. Therefore, most of the data were collected in tie stalls with the feeding behavior of individual cows measured. The data on individual cow behavior are summarized in Figure 3. Sorting was measured by determining the physical distribution of the feed refused by each individually housed cow. Actual as fed intake of each screen (total offered – amount in refusals) was calculated directly. This number was divided by the predicted as fed intake for that same screen, where predicted intake for a screen is the as fed intake for that cow (total offered – total refused) multiplied by the as fed distribution of the total mixed ration (TMR) on that screen. So the predicted intake is the “value on paper” for the diet, and the numbers we show are the actual cow intakes of each screen presented as percent of the predicted. When expressing data this way, screens less than 100% are being sorted against, and screens more than 100% are being selectively consumed. If one screen is less than 100%, some other screen will have to be more than 100%. In general, screens with a small amount on them (like the coarsest screen) can deviate from 100% more than screens with a lot of material on them, and it is important to remember that to avoid over-interpreting the data. A sample calculation for a simple one-screen separation is shown in Table 1.

Figure 3: Sorting of the three top screens of the Wisconsin separator. Many cows eat less than 50% of the material on the very coarsest screen (black bar), while the sorting expressed on a percentage basis is much less for the material passing the 18 mm screen and retained on the 9 mm screen (open bar). This is certainly partly due to the small amounts of feed on the very top screen, so a small absolute amount of sorting makes a large percentage change. But these feeds also represent material that cows can push away easily with the one tool at their disposal: a 6-inch-wide nose.

Table 1: Calculating a sorting index. Sieving data of the TMR offered (as %) are shown in the second column. Absolute amounts offered in column 3 are based on the amount of feed offered (in this case DM per cow but can be as fed per pen) and the % in column 2. Diet eaten (the difference between total offered and total orts) is calculated from orts weight, which must be measured. “Expected”* consumption of long particles is determined by percentage of diet offered that is long times total intake. This is what a 100% non-sorting cow would eat. If the orts are 10% long (indicating some sorting against long particles), the actual consumption of long particles is determined to be 0.6, which is 75% of the expected 0.8.
Offered % Offered kg Eaten kg Orts Really Eaten Sorting Index
Diet 100% 20 16 4 kg 100% 16
Long 5% 1 0.8* (0.05×16) 0.4 kg 10% 0.6 (1-0.4) 0.6/0.8 = 75%
Short 95% 19 15.2* (0.95*16) 3.6 kg 90% 15.4 (19-3.6) 15.4/15.2 = 101%

Two important facts are apparent in Figure 3. One is that sorting differs cow to cow. Some cows will sort extensively. There are the cows that consume less than 20% of the longest dietary particles, and more than one cow in our studies refused this portion of the diet entirely. The other fact is that not all cows do this, and in a pen of cows, it would be easy to miss the fact that a few individual cows were sorting aggressively when sieving pooled weigh-backs. Looking for “sorting holes” early in the feeding cycle may in fact be more informative. A second point is that the medium screen particles are not sorted extensively by any cows. Taken together what does this mean? Adding “top Wisconsin screen fiber” to the TMR will certainly increase the mean particle size of the diet offered and of the diet consumed by the “average cow.” However, it may do absolutely little for a small group of cows and nothing for a few outlier individuals. Since a few percentage points increase in the incidence of displaced abomasum is not a minor issue, this could be an important underlying factor. If, however, the diet mean particle size is raised by increasing the medium length particles, this will help all the cows. The latter approach is therefore much more desirable.

We did an experiment to compare sorting of tie-stall and free-stall housed cattle. We were not able to get individual cow behavior in the free-stall group, but we did a switch-back trial that provided some meager replication and at least an attempt at a statistical analysis. These data are shown in Figure 4. The data suggest more sorting by cows housed as a group than by the average of the same cows housed individually. This was what we predicted would happen. Cows housed in tie stalls that sort early in the feeding cycle will be presented with a more and more fibrous diet as the day proceeds. In a loose housing feeding situation, “sorter” cows may move to a relatively unsorted part of the TMR left by “non-sorter” cows. An aggressive cow, who also happens to be an aggressive sorter (we have no reason to believe they are linked), may be particularly susceptible to behaving in this manner.

Figure 4: Sorting by the same cows housed individually in tie stalls or as a group in a single free stall pen. Cows appeared to sort more in the free stall, and the difference was larger than could be attributed to chance based on observed variation of the cows in the tie stalls.

Measuring Sorting in the Field and Feed Bunk Management

The particle size of the feed refused by a cow (or a group of cows) is not only dependent on the sorting that has occurred but also very much by the amount of weigh-back remaining. Figure 5 shows this effect graphically for a theoretical situation. To put it simply, if there is no weigh-back at all, then the single cow has not sorted, and in the group of cows, average sorting was zero. Remember that average sorting of zero could mean that sorting against long particles by one cow is compensated for by another cow. We do not see a great deal of sorting in favor of long particles in our studies with individually fed cows, but in a group-fed, limit-fed pen situation, cows may be coerced into this behavior. The second point is that if weigh-backs are extremely large relative to the amount consumed, the effect of sorting is not so obvious. At “infinite weigh-back,” the orts will be the same particle size distribution as the diet, no matter how much sorting occurs. Obviously, we never get near to “infinite weigh-back” in practice, but the shape of the curve on the low weigh-back side is very informative. So if sieving of weigh-back is done, the amount of weigh-back must also be determined to have any real quantitative information. Also remember that the Penn State top screen pools the longest and next longest particles shown in Figure 3, so it will be less sensitive. Pulling out long particles from the top Penn State screen and counting them may be more informative than simply the total weight on the top screen.

Figure 5. A theoretical example: The TMR is 5% long particles. and the animal is consuming 16 kg. As more feed is offered and more orts (weigh-back) are left, the weigh-back begins to look just like the diet eaten though sorting has not changed. In this example, the animal was consuming only 75% of these long particles. Note that at low levels of orts, the composition of the weigh-back is very sensitive to amount of weigh-back even though we fixed the sorting at 75%. Weight and particle distribution of both the weigh-back and the TMR offered must be known to determine sorting. How to do the calculation is shown in Table 1.

The amount of weigh-back to leave is not a simple economic question, especially when no alternative feed use (heifers) of the refused feed is possible. Weigh-back is expensive and creates a nutrient management problem because unconsumed feed is used with an efficiency of 0. Careful feed management, dry matter measurements, cow number monitoring, observation at appropriate intervals in the feeding cycle, and anticipation of weather effects are all helpful but not automatic in minimizing wasted feed. On a hot diet, limited feed access is one way to control acidosis, but where cows have limited access to bunk space and have to wait in line to eat, this is a troubling approach even in a very well-managed operation. Using limited feed to minimize sorting is something I would only consider after all my other better options (i.e., proper chopping and proper diet moisture) were implemented, and even in that case, I think the cure would usually be worse than the disease.

How Much Physically Effective Fiber?

I believe the best way to determine physically effective fiber is to measure the amount of NDF above the 9 mm screen or some similar “middle” screen. Unfortunately, most research, including our own, has not used this measure. This measurement should replace the concept of calculating a “physical effectiveness factor” based on mean particle length and then multiplying this factor times the diet NDF. The reason I do not like the latter approach is that if you add very fine fiber to a diet in place of fine starch, the “physical effectiveness factor” will not change. However, because the NDF of the diet increases, the apparent content of physically effective NDF will increase. It is quite clear that adding fine NDF to low-forage diets is a positive step, but I think it is fundamentally incorrect to call this physically effective fiber. Clearly this fine fiber, because of its chemical difference from starch, is “effective,” but not “physically effective.” I believe a final system will have to consider both physically effective NDF and total NDF (in addition to other carbohydrate characteristics.

My firm impression of the literature (with mid-lactation cows primarily and often short feeding periods) is that the requirement for particle length on “normal” diets is actually lower than most people in the field think. Methods for average particle size distribution vary among studies based on the sieving device used and whether DM or as fed is used. My interpretation of the literature is that positive milk fat yield responses stop somewhere below 5 mm and that responses in chewing stop at about 7 mm. That is, we increase the physically effective fiber content between 5 and 7 mm, but the physically effective requirement was met by 5 mm. Intake reduction is common above 5 mm. In Figure 1, 5 mm corresponds to only 38% of as fed material on the top three Wisconsin screens. This would probably be somewhat higher % on the Penn State screen (maybe ~ 45%), but we do not know of a large body of comparative data for the two techniques. It is certainly reasonable to feed longer mean particle size to cows at calving, so when using a one group TMR system, this increased requirement would have to be accommodated. Diets that have a great deal of their digestible energy from NDF, and possibly from soluble fiber, can get by with less particle size in mid-lactation cows. The advantages of lower particle size can occur in the silo as well as in the cow. Diet economics will often favor adding by-product fiber in place of forages, and exploring this economic option requires consideration of the overall characteristics of the diet carbohydrate. Physically effective fiber consumed is only one aspect of this balancing act.

Figure 6: Particle size of diet offered, particle size of diets consumed by individual cows in tie stalls, and average particle size of diets consumed. These data are the same as used in Figure 2. Note the range of particle size consumption within a given diet versus the spread between fairly fine diets (4.5 mm) and coarse diets (6.7 mm). The lowest point for the 6.7 mm diet is actually the same as the distribution offered in the 5.4 mm diet.

Author Information

Lou Armentano
University of Wisconsin


Armentano, L.E., and D. Taysom. 2005. Short communication: Prediction of mean particle size and proportion of very long fiber particles from simplified sieving results. J. Dairy Sci. 88:3982-3985.

Kononoff, P.J., and A.J. Heinrichs. 2003. The effect of reducing alfalfa haylage particle size on cows in early lactation. J. Dairy Sci. 86:1445-1457.

Leonardi, C., F. Giannico, and L.E. Armentano. 2005. Effect of water addition on selective consumption (sorting) of dry diets by dairy cattle. J. Dairy Sci. 88:1043-1049.

Leonardi, C., K.J. Shinners, and L.E. Armentano. 2005. Effect of different dietary geometric mean particle length and particle size distribution of oat silage on feeding behavior and productive performance of dairy cattle. J. Dairy Sci. 88:698-710.

Leonardi, C., and L.E. Armentano. 2003. Effect of quantity, quality and particle length of alfalfa hay on selective consumption by dairy cows. J Dairy Sci. 86:557-564.

Leonardi, C., and L.E. Armentano. 2007. Short Communication: Feed Selection by Dairy Cows Fed Individually in a Tie-Stall or as a Group in a Free-Stall Barn. J. Dairy Sci. 90:2386–2389