Hysteresis, the disobedience of expiration toward inspiration, has never been the direct subject of a CICM exam question. It could reasonably form a part of Section F3(ii), "understand compliance (static, dynamic, and specific) and its measurement", or it could fit into Section F3(iii), "relate compliance to the elastic properties of the respiratory system", or it could legitimately come from Section F3(iv), "describe the pressure volume relationships in the respiratory system"; but otherwise it does not form a part of the 2023 CICM Primary Syllabus. However, it warrants a chapter all to itself, as it is a notoriously misunderstood topic with poor textbook exposure, featuring confusing throw-away explanations that conceal more than they reveal. The dim light cast on hysteresis by official CICM texbooks faces the exam candidate with a terrible choice, where either a trainee squanders their time resources by combing through obscure literature to synthesise their own explanation, or they resort to rote-learning poor explanations to regurgitate keywords without understanding them. Neither scenario is desirable for the future of the speciality, and so it is necessary to make this deliberate effort.
There is no single deposit of knowledge to mine for these data. The best papers (Escolar & Escolar of 2004, Harris from 2005, Brusasco & Pellegrino of 1995) all leave something unsaid, making it necessary to search through other references.
Under normal conditions (i.e where it is not filled with saline), the lung does not behave as an ideal system, i.e. the energy invested in its distension is not returned upon deflation. The upshot of this is that inflation and deflation have different pressure-volume relationships, and the difference between them is called "hysteresis", a term etymologically related to "lag" or shortcoming" which describes the dependence of a system's state upon its history. If one were completely unprepared for the questions "define hysteresis", one could easily break down and blather something like "the inspiratory thing does not look like the expiratory thing", so it would probably be worth investing some time in memorising a more solid definition. Here's one from an excellent article by Escolar & Escolar (2004):
"The energy applied to the lung in inspiration is not recovered in expiration. The property of dissipating energy receives the name of hysteresis."
A pithier, more memorable definition is available from a much less reputable source:
"Lung volume at any given pressure during inhalation is less than the lung volume at any given pressure during exhalation"
Textbooks differ in their interpretations, but only in terms of wording:
"The pressure–volume curves for air-filled lungs during inflation and deflation are different. This
phenomenon is called hysteresis"Kam (2015)
"rather more than the expected pressure is required during inflation, and rather less than the expected recoil pressure is available during deflation".
(Nunn, 9th ed, 2020)
"whilst at identical pressures during inhalation and exhalation, the volumes are actually different, with the volume being more during exhalation."
(Middleton, 2021)
It is of no use to the reader to list more similar items, as no additional clarity for the concept will be produced thereby, but they are assured that all the other sampled textbooks follow this trend, and differ.
Hysteresis can be framed as the improvement of lung compliance during expiration. It makes logical sense to expect something like this in a dynamic PV loop because of the effects of resistance (i.e. the effect on dynamic compliance is intuitive), but it is seen even in static compliance measurements. Here, a diagram from Harris (2005) demonstrates the hysteresis in a static PV loop using the supersyringe method. The added labels demonstrate that, for the same change in pressure, the expiratory compliance is higher:

But the attentive reader (thank you, Adam Kallmeyer) may object: going through the diagram, one is immediately impressed by the observation that the volume change during expiration is much smaller than the volume change during inspiration, over the same range of pressures. Surely that means that lung compliance is lower during expiration, i.e. a smaller volume is achieved per unit pressure?

So, what actually happens? Well: the textbooks are generally silent on this issue, but where they do make any comment, they tend to favour the interpretation that lung compliance is higher with expiration (for example this is the opinion of Edwards & Annamuraju, 2023). The basis for this seems to be a difference in perspective: the textbooks look at the total change in volume and the total pressure, or an instantaneous measurement (at this static pressure, we have this static volume) rather than a range of pressures (as the shaded box above represents). This is in fact correct: the average compliance is better on expiration over the course of the whole TLC manoeuvre.
But then, why does lung compliance seem to be decreased in expiration at the top of the volume range? The answer lies in the differences in lung compliance at different volumes. Compliance is maximal at FRC and lower at any other volume, i.e. a more distended lung has less compliance than a less distended lung. The pressure change in the inspiratory limb of the diagram above is happening at a much lower, more compliant volume than in the expiratory limb. When the pressure changes during expiration, the change in lung volume is smaller for the same change in pressure because the lung was more distended to begin with. Following from this, if both inspiration and expiration were observed over the same (uppermost) volume range, the compliance of each curve would be similar:

These real-life measurements were collected from ARDS patients by Mojoli et al ( 2023), and using their published data one can construct their own curves to illustrate the difference between inspiratory and expiratory compliance. At any given pressure and volume, the measured lung volume in expiration was higher, giving a consistently higher lung compliance during the expiratory phase of the cycle. Only when the volume was near-maximal (i.e. at close to 100% of the total lung capacity) did the compliance values converge:

Following from the above, one might expect hysteresis to be smaller close to the FRC, and in fact that is what one observes, eg. in this cat recording from Bachofen & Hildebrandt (1971):
_0.jpg)
In summary:
The reasons for this are one of the most confusing topics in physiology, potentially because well-meaning people tend to pull out complex diagrams to explain them.
There are several reasons for this change in lung behaviour between inspiration and expiration. The following explanation is an expansion and reinterpretation of another (already excellent) explanation by Harris (2005). Unfortunately the original paper is now paywalled by the corporate bastards that have taken over Respiratory Care, but it was possible to reconstruct the bulk of the content below.
In short, hysteresis occurs because of:
For reasons mostly related to the author's lack of intellectual alacrity, the explanation of how these work will be painfully elaborated in what feels like a logical order first, and then applied to different points in the pressure-volume curve to explain how these factors play a role at each stage.
We take all of this with a grain of salt, as the Laplace relationship is really only applicable to the even inflation of elastic spheres, and alveoli are neither evenly inflated, nor elastic, nor spheres (for the interested reader, Prange (2003) dismantles the Laplacian paradigm in a vicious takedown that culminates with dismissing it to "a place, not in our minds and textbooks, but in the museum of wrong ideas"). But bear with this broken model for a moment, as it still has a warm place in the hearts of the examiners. In short, the Law of Laplace describes the relationship between the pressure exerted on the contents of an elastic sphere by its walls:
σ ∝ Pr/T
where
The pressure acting to collapse the alveolus can therefore be represented as the vector diagram, relating this inward force to the magnitude and direction of the surface tension, i.e. the pull of each wall segment towards every other wall segment:

And so, as the alveoli distend, as long as the surface tension stays mostly the same, the force acting on the lung decrease in their magnitude:
The upshot of this is that small nearly collapsed alveoli are more prone to completely collapse and are harder to inflate. This genuinely appears to happen, at least in the parts of the respiratory cycle. Escolar & Escolar (2004) report a whole string of unobtainable references from the distant past which describe measurements of alveolar curvature, where at around 40% TLC the alveolar walls were on average more curved during inflation than during deflation.
Collapsed alveoli have walls which are stuck together and which require added mechanical energy to open. In contrast, well-inflated alveoli are relatively elastic and require relatively little energy to inflate further. Because of this, the pressure-volume relationship of alveoli changes after they have been fully inflated. This accounts for the difference in expiratory and inspiratory compliance in the low-pressure low-volume range of values. The pressure required to blow open fully collapsed alveoli is greater than the pressure required to maintain the inflation of already opened alveoli, which accounts for the different volumes for the same pressure in the low pressure, low volume corner of the PV graph.
The hysteresis of the lungs would be even greater if it were not for the positive rectifying influence of the chest wall, which lends some structure and discipline to those namby yieldsome hippies. Most of the time, only a small proportion of the alveoli are actually collapsed, because in an intact chest the pleural cavities is a fixed minimum volume beyond which further collapse would be impossible. But what if you are working with just a disembodied lung? Observe, these graphics borrowed from a lost Radford paper (1964):.jpg)
It was even worse in this rabbit model from Hoppin et al (1986), where the pressure-volume relationships from the first and third respiratory cycles are depicted:

That's right, the hysteresis was markedly increased at the lowermost volume ranges by the need to forcibly reopen every last alveolus, which is what you can achieve if you allow the lung to collapse completely. Notably, the shape of the expiratory curve was essentially untouched by this.
Surfactant increases lung compliance, because water on its own has a surface tension so high that the alveoli would collapse en masse and lung compliance of an air-filled surfactantless lung would be extremely poor.
How poor? The reader intent on taking their barest minimum to the exam will see it as sufficient to know that it is poor. The learner possessed by hyperfocus will also be interested to know we have numbers for this. Francis & Saha (2022) modelled mathematically and Schürch et al (1976) established empirically that when only water coats the alveoli, the surface tension has a value of 70 mN/m, whereas in the presence of surfactant, this drops to 25-30 mN/m during inspiration, and as low as 1 mN/m towards the end of expiration:
.jpg)
This rat lung data from Schürch et al (1976) does not quite represent what happens to compliance if the surfactant is missing. A suitable diagram to illustrate this comes from a paper by Lachmann et al (1980). The authors lavaged all the surfactant out of the lungs of rabbits, and thereby created conditions resembling ARDS (see their stolen graphs below).
.jpg)
One will note that some hysteresis remains, which suggest that some surfactant remains. It is possible to say this because the elimination of all surface tension tends to obliterate hysteresis, and this is the way in which most textbooks manage to fit lung surfactant into the category of things which affect static compliance. Usually, a graph is trotted out which compares a standard compliance curve with a compliance curve measured in a lung filled with saline. The reference for this is usually a famous 1971 paper by T.E. Morgan, but in fact the original experimental data weres published in 1929 by Kurt von Neergaard. Unfortunately, the original article from Zeitschrift fur die gesamte experimentelle Medizin can no longer be obtained by any reasonable (cheap) means, but the graph is sufficiently famous that one can find a version of it everywhere. In some cases, for example the diagram below from Radford (1964), one can even find the original experimental cat's serial number. For some reason, everybody always picks the graph from Cat 27.
.jpg)
Cat 27's lungs were first inflated and deflated with air. Then, they were submerged in saline and inflated with saline. The effect was substantial. In the saline-filled lung, the effect of the surfactant on the surface tension of the alveoli was obliterated, and only the elasticity of the lung itself was measured. The drowned lung was much more compliant than the air-filled lung, and had no hysteresis whatsoever because there was no air-water interface for the surfactant to influence. We are also grateful to Dr Hopkinson, one of the readers, who has submitted this video of a PV look from a surfactant-lacking neonate, as it illustrates the same concept (and apparently as they generate surfactant the loop "opens up"):
But at this stage the attentive reader will begin to complain that the previous section built their expectations for a lower surface tension in the presence of alveolar distension. Surely the direct application of the Law of Laplace should mean that the thinner stretched alveolar walls near TLC exert a lower inward pressure, as compared to the smaller alveoli? Why, then, do we observe a increased surface tension in these upper ranges of volumes and pressures, like in Schürch et al (1976)? The answer lies in the different distribution of surfactant during different phases of the respiratory cycle.
Surfactant, discussed elsewhere, is responsible for maintaining the low surface tension of alveoli. But it is said that the effect of surfactant is variable, and dependent on the dispersion of surfactant molecules across the alveolar surface. Without turning this already overlong chapter into a deep dive through the kinetics of phase transition in Langumuir monolayers, the interested reader is instead directed to Smith & Stamenovic (1986) and Wüstneck et al (2005), whereas the less interested reader is left with the following series of oversimplifications:
Ergo, lung hysteresis is produced (at least in part) by an asymmetry between inspiratory and expiratory surfactant behaviour, where surface phospholipid is concentrated on expiration, but diluted on inspiration. The result is a reduced surface tension for the expiratory limb of the pressure volume relationship, as demonstrated by these measurements from Smith & Stamenovic (1986):

Obviously, the lung has solid material in it, and members of this solid material are connected to one another in various flexible and inflexible ways. The work by Bachofen et al (1987) had tried to separate the contributions from structural tissue components and alveolar surface tension by comparing aerated and saline-filled lungs. The result, depicted below in a slightly mangled copy of the original diagram, was that the contribution of surface tension to the overall "recoil pressure" did not increase very much over the course of inflation and deflation, whereas the contribution of tissue tension increased dramatically.

This makes logical sense, as we recall that surface tension is nerfed by Laplacian factors in distended alveoli, and moreover the re-concentration of surfactant makes it fall quite quickly with deflation, even over a fairly small range of volume changes. On the other hand the effect of stretching collagen and elastin fibres in the walls of the alveoli, when one approaches the anatomical limits of their capacity, contribute a significant recoil pressure. Elastin specifically is about 30% of lung parenchyma, and is probably the most important of the protein elements involved in recoil at lower volumes; whereas collagen, according to Hoppin et al (1986), behaves like an accordion at lower volumes but then suddenly becomes straight and tense at high volumes. Once the critical collagen fibre stretch occurs (at around 80% TLC according to that diagram from Bachofen), the tissue tension plays a greater role in the pressure-volume relationship, and the inspiratory pressure-volume curve begins to "beak". This is the overdistension that is thought to play a role in VILI.
This high tissue recoil pressure, however, seems to mostly encumber the inflating lung. Not much of this energy is stored: fibres slide past one another, microstructural crosslinks are temporarily disrupted, and some energy is dissipated as heat. Faffe & Zin (2009) go into some of these details to describe some of the specific elastic and inelastic elements. In deflation, these barriers are already overcome, and the pressure applied by these tissue forces on the alveolus is diminished. One might say this is the very essence of the concept of hysteresis, the quantity of unrecovered energy which results from something being imperfectly elastic. As the result, the transpulmonary pressure can decrease without much change in the volume: the alveoli are not especially inclined to collapse.
This concept may be hard to parse as a drop in compliance, but that is in fact how it needs to be represented, compliance being the change in volume per change in pressure. The reader accustomed to operating ventilators will revolt against this counterintuitive position, as the term "compliance" is usually framed in terms of the inspiratory compliance (considering that the mechanical ventilation of ARDS is where most ICU trainees usually encounter it). However that is exactly what has happened. The noncompliant overdistended lungs are equally reluctant to change their volume, whether the change is up or down; and the relationship is flatter in expiration because tissue tension is no longer adding to the surface tension.
These are not really any properties of the lung parenchyma itself but rather errors or soruces of confusion in the process of making static measurements.

The tenacious exam candidate who has made it this far in the page will surely by this stage have wondered at least once whether any of this material is likely to appear in the CICM exams. The answer is clearly "yes", as several syllabus items point to this possibility, but the exact examinable content remains to be seen. It it likely to be derived directly from the official textbooks, which makes it essential to reflect on how they differ in their presentation of this topic. For example, the Kerry Brandis Physiology Viva does not have hysteresis as a specific item, which potentially explains why the examiners have not used it yet (it being a known resource for SAQ ideas).
A brief review reveals the following bibliography:
In short, the candidates can breathe easy, knowing that, if the exam questions are based on the material in official textbooks, the answers they produce cannot possibly be expected to be either coherent or detailed.
In the mistaken belief that a Grand Unified Diagram of All the Things will somehow automatically improve the understanding of the reader, an attempt was made to create a representative pressure-volume loop for a series of hypothetical static compliance measurements, and to label each section with the specific factors that shape it.

But at this stage the attentive reader will point to this relationship and complain that it makes no sense. The pressure-volume curve clearly demonstrates a reduced rate of volume change, for a pressure change from 40 to 35 cmH2O (as an example), where the surface tension is supposed to be highest.

But surface tension is maximal at 100% TLC, and is a force pushing the alveoli to collapse. So, should the increase in surface tension not result in improved compliance? As the stretched alveoli will have plenty of surface tension, surely the increased inward pressure should be making it easier to deflate these alveoli, leading to an improvement in compliance on expiration instead? This question seems to bedevil a number of readers.
Let us consider the net sum of the forces acting on the lung.
We will for the sake of this thought experiment assume that the alveolar pressure is somehow separate from the alveolar surface tension. The resting alveolus will therefore experience these pressures:
Alveolar surface tension is usually expressed in mN/m, and is ~ 25 mN/m at FRC. Assuming the average alveolar radius at FRC is ~ 100 µm (Hyde et al, 2004) and applying the Law of Laplace,
we calculate the surface tension in Pascals first, and then convert to mmHg:
Under normal circumstances, this Laplacian pressure is acting in opposition to the transpulmonary pressure, and the equilibrium between the two establishes whether the alveolus remains open or collapses. For various completely sensible reasons this balance is mostly in favour of alveolar patency under normal states of health.

As long as the intrapleural pressure remains below atmospheric, the transpulmonary pressure and Laplace pressure are subtracted from each other. During spontaneous respiration the pleural pressure decreases to entrain gas into the alveoli, but when each static volume measurement is performed at a different transpulmonary pressure, the alveolar pressure has equilibrated with atmospheric (that's why there is no flow in the airways), which means whatever transpulmonary pressure is recorded on the PV curve basically represents the pleural pressure. Using the maximum values from Smith & Stamenovic (1986), which is easy because dynes/cm = mN/m, and assuming the alveolar radius has increased in size to the theoretical maximum (~160 µm) at TLC,
Thus, at 100% TLC the pressures acting on the alveolus look like this:

Of course the separation of TPP and Laplace pressure are totally artificial; the TPP one measures is in fact the net pressure required to produce this change in volume and already incorporates the value of T. In other words, with T = 6.75 mmHg, a TPP of 30 mmHg produces the volume "100% TLC". If the T was 0 mmHg, the TPP required to produce the same volume would be 25.8 mmHg. In other words, the alveolar surface tension has already produced a counterpressure that has opposed the transpulmonary pressure and reduced the lung compliance, and the TPP measurement reflects this effect.
Now, consider the same alveolus being sequentially deflated from 100% TLC, in steps (as befits a good honest supersyringe measurement of static compliance). At each step, the TPP, alveolar radius and surface tension will reduce (the latter dropping rapidly to almost zero at FRC). A table of serial measurements can be produced, using the same data from Smith & Stamenovic (1986) to compare different volumes and pressures at different points in the cycle:
| r (µm) | t (mN/m) | T (mmHg) | (% TLC) | TPP (mmHg) | Volume (ml) |
Compliance (ml/mmHg) |
| On expiration | ||||||
| 160 | 0.045 | 4.22 | 100% | 30 | 5000 | 166.67 |
| 150 | 0.028 | 2.80 | 90% | 22 | 4500 | 204.55 |
| 140 | 0.02 | 2.14 | 80% | 12 | 4000 | 333.33 |
| 130 | 0.01 | 1.15 | 70% | 8 | 3500 | 437.50 |
| 120 | 0.0002 | 0.03 | 60% | 3 | 3000 | 1,000.00 |
| 110 | 0.00015 | 0.02 | 50% | 2.5 | 2500 | 1,000.00 |
| 100 | 0.0001 | 0.02 | 40% | 2 | 2000 | 1,000.00 |
| On inspiration | ||||||
| 100 | 0.015 | 2.25 | 40% | 8 | 2000 | 250.00 |
| 110 | 0.0325 | 4.43 | 50% | 14 | 2500 | 178.57 |
| 120 | 0.037 | 4.63 | 60% | 16 | 3000 | 187.50 |
| 130 | 0.04 | 4.62 | 70% | 18 | 3500 | 194.44 |
| 140 | 0.042 | 4.50 | 80% | 22 | 4000 | 181.82 |
| 150 | 0.044 | 4.40 | 90% | 26 | 4500 | 173.08 |
| 160 | 0.045 | 4.22 | 100% | 30 | 5000 | 166.67 |
From this, it follows that the opposition of the surface tension to the transpulmonary pressure is a constant feature throughout this sequence, and at each stepwise deflation, the transpulmonary pressure requirement for a given volume is higher if the surface tension is higher. For example, where the transpulmonary pressure is 8 mmHg during inspiration, the volume maintained by this pressure is 2000ml (40% TLC, i.e basically the FRC), whereas during expiration it is 3500ml (70% TLC).
Harris, R. Scott. "Pressure-volume curves of the respiratory system." Respiratory care 50.1 (2005): 78-99.
Fahlman, Andreas, et al. "Inflation and deflation pressure-volume loops in anesthetized pinnipeds confirms compliant chest and lungs." Frontiers in Physiology 5 (2014): 433.
Mojoli, Francesco, et al. "Tidal lung hysteresis to interpret PEEP-induced changes in compliance in ARDS patients." Critical Care 27.1 (2023): 233.
Romero, P. V., J. Lopez Aguilar, and L. Blanch. "Pulmonary mechanics beyond peripheral airways." Applied Physiology in Respiratory Mechanics. Milano: Springer Milan, 1998. 199-210.
Brusasco, V., and R. Pellegrino. "Hysteresis of airways and lung parenchyma." Respiratory medicine 89.5 (1995): 317-322.
Faffe, Débora S., and Walter A. Zin. "Lung parenchymal mechanics in health and disease." Physiological reviews 89.3 (2009): 759-775.
Bachofen, H., et al. "Relations among alveolar surface tension, surface area, volume, and recoil pressure." Journal of Applied Physiology 62.5 (1987): 1878-1887.
Horie, T., and J. Hildebrandt. "Dependence of lung hysteresis area on tidal volume, duration of ventilation, and history." Journal of Applied Physiology 35.5 (1973): 596-600.
Prange, Henry D. "Laplace’s law and the alveolus: a misconception of anatomy and a misapplication of physics." Advances in physiology education 27.1 (2003): 34-40.
Bachofen, H., and J. Hildebrandt. "Area analysis of pressure-volume hysteresis in mammalian lungs." Journal of applied physiology 30.4 (1971): 493-497.
Bachofen, H., J. Hildebrandt, and M. Bachofen. "Pressure-volume curves of air-and liquid-filled excised lungs-surface tension in situ." Journal of Applied Physiology 29.4 (1970): 422-431.
Francis, Isabella, and Suvash C. Saha. "Surface tension effects on flow dynamics and alveolar mechanics in the acinar region of human lung." Heliyon 8.10 (2022).
Zasadzinski, Joseph A., et al. "Overcoming rapid inactivation of lung surfactant: analogies between competitive adsorption and colloid stability." Biochimica et Biophysica Acta (BBA)-Biomembranes 1798.4 (2010): 801-828.
Schürch, Samuel, Jon Goerke, and John A. Clements. "Direct determination of surface tension in the lung." Proceedings of the National Academy of Sciences 73.12 (1976): 4698-4702.
Schürch, Samuel, Jon Goerke, and John A. Clements. "Direct determination of volume-and time-dependence of alveolar surface tension in excised lungs." Proceedings of the National Academy of Sciences 75.7 (1978): 3417-3421.
Smith, JEFFREY C., and D. I. M. I. T. R. I. J. E. Stamenovic. "Surface forces in lungs. I. Alveolar surface tension-lung volume relationships." Journal of applied physiology 60.4 (1986): 1341-1350.
Wüstneck, R., et al. "Interfacial properties of pulmonary surfactant layers." Advances in colloid and interface science 117.1-3 (2005): 33-58.
Schürch, Samuel, et al. "The surface-associated surfactant reservoir in the alveolar lining." Neonatology 67.Suppl. 1 (1995): 61-76.
Knebel, D., et al. "Scanning force microscopy at the air-water interface of an air bubble coated with pulmonary surfactant." Biophysical journal 82.1 (2002): 474-480.
Morris, J., et al. "Dynamic behavior of lung surfactant." J. Biomech. Eng. 123.1 (2001): 106-113.
Schürch, Samuel, et al. "Surface properties of rat pulmonary surfactant studied with the captive bubble method: adsorption, hysteresis, stability." Biochimica et Biophysica Acta (BBA)-Biomembranes 1103.1 (1992): 127-136.
Bhana, R. H., and A. B. Magan. "Lung mechanics: A review of solid mechanical elasticity in lung parenchyma." Journal of Elasticity 153.1 (2023): 53-117.
Vindin, Howard J., Brian GG Oliver, and Anthony S. Weiss. "Elastin in healthy and diseased lung." Current opinion in biotechnology 74 (2022): 15-20.
Hoppin, Frederic G., et al. "Lung recoil: elastic and rheological properties." Handbook of Physiology. The Respiratory System. Mechanics of Breathing. Vol. 3. American Physiological Society Bethesda, MD, 1986. 195-216.
Setnikar, I., and G. Meschia. "Proprieta elastiche del polmone e di modelli meccanici." Arch. Fisiol 52 (1953): 288-302.
Rubini, Alessandro, and Emanuele Luigi Carniel. "A review of recent findings about stress-relaxation in the respiratory system tissues." Lung 192.6 (2014): 833-839.
Cheng, Winchi, et al. "Contribution of opening and closing of lung units to lung hysteresis." Respiration physiology 102.2-3 (1995): 205-215.
Hyde, Dallas M., et al. "Total number and mean size of alveoli in mammalian lung estimated using fractionator sampling and unbiased estimates of the Euler characteristic of alveolar openings." The Anatomical Record Part A: Discoveries in Molecular, Cellular, and Evolutionary Biology: An Official Publication of the American Association of Anatomists 277.1 (2004): 216-226.