Formation, circulation and function of cerebrospinal fluid

This chapter tries to address Section K1(iii) of the 2017 CICM Primary Syllabus, which asks the exam candidates to "describe the physiology of cerebrospinal fluid". Somehow, out of all the possible options, this relatively narrow subject matter has commanded the CICM First Part examiners' attention the most; in the sense that there are more questions about CSF physiology then virtually anything else.  Historical examples of CSF questions are listed here as a reference:

 In summary:

  • Cerebrospinal fluid (CSF) is an ultrafiltrate of plasma contained within the ventricles of the brain and the subarachnoid spaces of the cranium and spine.
  • Amount:  Approximately 400-600ml/day, or 25ml/hr;
    • At any given time, a normal adult has about 150ml 
    • About 25ml is in the ventricles, and a further 30 ml in the spinal canal
    • The rest is in the subarachnoid space
  • Site: three main sources:
    • Choroid plexus within the ventricles (mainly lateral ventricles)
    • Brain interstitial fluid (minor source)
    • Circumventricular organs (very minor source)
  • Mechanism: By highly regulated ultrafiltration and active secretion 
    • Ultrafiltrate of plasma is formed by the fenestrated choroidal capillaries
    • It collects in the choroid interstitial space
    • Ions are actively transported out of this ultrafiltarte, and into the CSF:
      • Sodium is actively secreted from the apical membrane (Na/K/ATPase)
      • Carbonic anhydrase also provides H+ to power Na/H+ exchange at the basal membrane
      • The sodium gradient created and maintained in this way is then used to co-transport HCO3- and CL-  into CSF
    • This creates an osmotic gradient which pulls water across the membrane through aquaporin channels
    • This process is disconnected from ultrafiltration, i.e. CSF production is constant and is not pressure-dependent (though at low CPP, <55 mmHg, CSF prodction decreases)
  • Composition:
    • CSF contains minimal protein (~0.2g/L)
    • CSF sodium potassium and calcium are slightly lower than plasma
    • CSF chloride CO2 and bicarbonate are higher than plasma
    • CSF glucose is about 2/3rds of the plasma value
  • Function:
    • Barrier function (the blood-CSF barrier)
    • Chemical stability and waste removal
    • Buoyancy and mechanical cushioning of the CNS
    • Hydraulic pressure buffering of ICP with arterial pulse and respiration
  • Circulation:
    • Out of the choroid plexus on the floor of the lateral ventricles
    • Through the foramen of Munro into the third ventricle
    • Through the Aqueduct of Sylvius, into the fourth ventricle
    • From the fourth ventricle, into the cisterna magna via the lateral foramina of Luschka and then up to the basilar cisterns around the pons, and then to the rest of the cortex
    • Or, from the fourth ventricle, via the medial foramen of Magendie, down to the spinal subarachnoid space
    • It is then reabsorbed from the subarachnoid space
    • The motor force driving this flow comes from:
      • Constant production at the choroid plexus
      • Arterial pulsation of the central nervous system structures
      • Venous pulsation of the CNS related to the respiratory cycle
      • Constant reabsorption by the arachnoid granulations
  • Reabsorption:
    • ​​​​​​​Reabsorbed at 25ml/hr (CSF secretion rate = reabsorption rate) into the dural venous blood
    • Main sites are arachnoid granulations
    • Driving force is hydrostatic gradient between CSF and venous blood
    • Thus, CSF reabsorption stops when CSF pressure is less than ~ 7 cm H2O

For a paying reader of Springer journals, Segal (2000) covers this topic nicely, but the dumpster-diving freegan can get exactly the same information from Telano & Baker (2020), Sakka et al (2011), and literally a thousand other papers. If you really want to pay for something, you could pay for Herndon & Brumback's The Cerebrospinal Fluid (1989), or Desenheimer's Cerebrospinal Fluid in Clinical Neurology (2015). The CICM First Part candidate is warned against embarking on any 442-page journey just for the purpose of exam preparation, but the casual reader could potentially derive some sort of joy from it, and potentially the Tumani chapter on Physiology and Constituents of CSF (p 25-34) could be recommended on its own as a solid summary. 

Cerebrospinal fluid

This (usually) crystal-clear liquid is typically introduced to a reader by means of some kind of short sentence that pithily summarises the reasons for why it is interesting and how it is distinct from any other random fluid collecting in the human body. For shortness and pithiness, nothing beats Telano & Baker (2020):

"Cerebrospinal fluid (CSF) is an ultrafiltrate of plasma contained within the ventricles of the brain and the subarachnoid spaces of the cranium and spine."

That, reader, is exactly what and where it is. Most authors also follow these introductory statements with some blurb to the effect of "vitally important", "cushioning", "nourishment and protection" etc. These are not pointless motherhood statements, and could benefit with some supportive explanations. The presence of cerebrospinal fluid and some sort of system of ventricles and meningeal CSF cavities appears to be the precondition for even a very basic central nervous system. It is present in all vertebrates, all the way down to the level of a hagfish (whose vertebrate passport looks pretty fake, in all honesty, as they do not have a vertebral column per se).  Myxine glutinosa do not appear to have a choroid plexus and their blood-brain barrier looks very abnormal (Murray et al, 1975), suggesting that they secrete CSF directly out of their brain, but they do still have a system of cerebrospinal fluid spaces. Invertebrates don't seem to need them, but then they also often seem to lack a blood-brain barrier, and in many of them the fluid bathing major neural structures is basically the same gross toilet water that sloshes around everywhere in whatever passes for a circulatory system in those animals. In short, CSF seems to be an essential part of being a neurologically complex organism.

Formation of cerebrospinal fluid

The "amount, site and mechanism" were important to the examiners in the answer to Question 15 from the second paper of 2018. Thus:

  • Amount:  Approximately 400-600ml/day, or 25ml/hr;
    • At any given time, a normal adult has about 150ml 
    • About 25ml is in the ventricles, and a further 30 ml in the spinal canal
    • The rest is in the subarachnoid space
  • Site: three main sources:
    • Choroid plexus within the ventricles (mainly lateral ventricles)
    • Brain interstitial fluid (minor source)
    • Circumventricular organs (very minor source)
  • Mechanism: By highly regulated ultrafiltration and active secretion 

"Ultrafiltration and active secretion" here is an oversimplification which might make a person think about some sort of crude glomerulus-like sieving mechanism. Sure, ultrafiltration does take place, and the fenestrated capillaries are often compared to those of the glomerulus, but "ultrafiltration" does not even begin to describe what the choroid plexus does, and it would be offended by the comparison. The formation of a filtered plasma is just the first part. The ultrafiltrate in the choroidal interstitial compartment is then processed by the choroidal epithelium and some of it is secreted into the ventricular cavity by a series of carefully controlled active processes. Additionally, there is contribution to CSF formation from the extracellular fluid of the brain, with which it mingles passively, and from the various other circumventricular organs where the blood-brain barrier is broken (which are small, and their contribution minimal). 

The choroid plexus

Thomas H. Milhorat (1976) offers perhaps the best description of this structure in a style that is largely missing from modern scientific literature:

"When visualized directly at surgery, or viewed through a ventriculoscope, the human choroid plexus  appears as a velvety vascular membrane which floats lazily in the CSF. Gross pulsations are not apparent, but the ventricular pulse, reflecting transmitted cardiac and respiratory influences, causes the plexus to bounce to and fro like an anchored boat at sea. When the choroid plexus is excised and fixed in Formalin, it loses its gossamer appearance and becomes rather polypoid and villiform."

No high-resolution images or videos are easily available, but if one wanted to have a look at what this gossamer appearance looks like under the hot lights of an endoscope, Tamburrini (2017) offer some images snapped during endoscopic third ventriculostomy:

the choroid plexus seen during endoscopic ventriculostomy, from Tambourrini (2017).gif

When people say "the choroid plexuses are the main interface between the blood and CSF",  they usually mean this in terms of raw surface area as well as fluid output. The plexuses on the blood side are highly vascular structures, furnished with a vast number of fenestrated capillaries, and their ventricular side they are covered in an epithelium bristling with microvilli. Here's some SEM photographs of vascular casts by Zagórska-Świeży et al (2008), to demonstrate the rich choroidal blood supply. As you can see, the capillaries form almost glomerulus-like basket shaped structures.


The whole thing is only about 2g of tissue, sitting on the floor of the lateral ventricle, measuring no more than 4mm in thickness in the normal adult (İmamoğlu et al, 2013). Fortunately, the carpet of microvilli massively increase its surface area, such that the adolescent rat has about 75m2 of it (Keep & Jones, 1990). This microvillous layer is separated from the capillaries by an interstitial space where ultrafiltrate collects.  There do not seem to be a lot of images of this structure out in the easily available Googleable spaces, but one nice picture could be sourced and altered from Damkier & Praetorius (2020):

choroid plexus interstitial space TEM microscopy image from Damkier & Praetorius (2020)b.jpg

Those capillaries lack tight junctions. That interstitial space under the basal lamina of the choroidal epithelial cells is full of fluid which has extruded through the fenestrated capillaries, and is probably similar to glomerular ultrafiltrate, but in all honesty nobody knows - this is a difficult space to access. To say with a straight face that "an ultrafiltrate forms here" would require the measurement of protein and pressure, and both would be challenging in vitro, let alone while it "bounce to and fro like an anchored boat at sea". If you did have those measurements, you could plug them into the standard Starling equation (or some kind of computer model). Fortunately, it is not essential to consider this compartment in any great detail, as all the magic happens at the choroidal cells themselves, and so the entire capillary-interstitial-cellular complex can be consider en masse, as the choroidal secretory surface.

The most important upshot of this information is that the rate of ultrafiltration (dependent on things like pressure of the blood and the protein oncotic pressure) is disconnected from the rate of CSF secretion, which is independent of pressure. Only with very low cerebral perfusion pressure (where the choroid plexus becomes relatively underperfused) does the rate of production become affected. Experiments by Weiss & Wertman (1978) determined the range of pressures where this is likely to happen, and concluded that a CPP of less than 55mmHg will decrease the rate of CSF production. Conversely, if the CPP remains over 70mmHg, CSF will continue to be secreted at the same rate, irrespective of how high the ICP is.

How is the fluid actively transformed from a dumb soupy ultrafiltrate to the pristine aqua vita in the ventricular cistern? There are a ton of different ionic mechanisms, and the real trick for the CICM trainee is not to get bogged down in detail. To get an appreciation for the size and viscosity of that bog, the trainee is invited to peek at the likes of Brown et al (2004) or Speake et al (2001). It is not essential to become a master of this content; broad brushstrokes would be enough. For example:

  • The apical microvillous brush border contains Na+/K+ ATPase pumps
  • They constantly pump sodium out of the choroidal cells and into the CSF
  • This creates a sodium gradient across the epithelium
  • The sodium gradient is then used to drive other transport processes:
    • Na+/H+ exchange
    • Na+/Cl- co-transport
  • Chloride transport across the epithelium can then be used to drive a Cl-/HCO3- exchanger, taking bicarbonate out of the blood and bringing it to the CSF through bicarbonate-permeable anion channels in the apical membrane
  • Choroidal carbonic anhydrase also catalyses the conversion of H2O and CO2 into bicarbonate inside the cell and at the apical brush border
  • The extra protons produced in this process are used to pump more sodium via the Na+/H+ exchanger
  • Chloride is also pumped out into the CSF by a Na+-K+-2Cl- cotransporter
  • The net effect is the movement of cations (sodium) and anions (chloride and bicarbonate) across the epithelial layer
  • This movement of cations creates an osmotic gradient across the membrane
  • The resulting movement of water along this gradient is entirely transcellular, via aquaporins in the basal and apical membrane

Or, in an even shorter form,

  • Sodium is actively secreted from the apical membrane by Na+/K+ ATPase
  • Carbonic anhydrase facilitates this by providing H+ for more sodium reabsorption
  • Bicarbonate and chloride are pulled across the membrane by sodium co-transport
  • All these cations and anions are actively secreted across the apical membrane, also by mechanisms which take advantage of the sodium gradient
  • This creates an osmotic gradient which pulls water across the membrane

Or, in the form of a horrific ion channel diagram:

Ionic processes in the secretion of CSF

So: with all these ions being actively moved into the CSF, and other ions and proteins being selected out, it sounds like the CSF would be rather biochemically different from plasma.

This brings us to:

Composition of cerebrospinal fluid

At one stage, it was believed that there is actually nothing particularly special about the chemical composition of the CSF, and that it was a pure crystalloid ultrafiltrate of the plasma. A series of experiments had proved this theory wrong, such as the 1964 exploration by Adelbert Ames III et al. Here, their cat data are arrayed illustratively:

CSF composition from Ames III et al, 1964

It would be logical for us to think that modern intensive care trainees should be expected to know human values, but interestingly, if you dig deep, you will find that most textbooks produce a table of comparison values which is mainly based on animal data from as far back as the 1930s (eg. Merritt et  al, 1937) Here, a representative set is borrowed from Brown et al (2004), which is mainly dog and some cat:

Composition of CSF vs Plasma
Constituent (mmol/L) Plasma CSF
Na  155 151
K 4.6 3.0
Mg 0.7 1.0
Ca 2.9 1.4
Cl 121 133
HCO3 26.2 25.8
Glucose 6.3 4.2
Amino acids 2.3 0.8
pH 7.4 7.4
Osmolality 300 305
Protein (g/L) 65 g/L 0.25 g/L

Again, wherever you look, slightly different values are found, as everybody drained the CSF of different dogs and used different techniques. Those differences do not matter. They depend on how the blood and the CSF were collected, how the animal was anaesthetised, and whether the experiment was performed in the 1930s using some sort of brass-valved steampunk apparatus to analyse the samples. From the perspective of the modern CICM trainee, the most important things to remember are that in CSF, 

  • Sodium is slightly lower
  • Chloride is higher
  • Total calcium is lower (because less albumin)
  • Ionised calcium is similar
  • Potassium is lower
  • Protein is much lower
  • Glucose is about 2/3rds of the plasma value
  • pH is slightly lower
  • PCO2 and bicarbonate concentration is slightly higher

To memorise this list of differences is probably a good enough strategy if some sort of naked tooth-and-claw exam survival is the goal. Each of these differences is clearly important and has some sort of reason behind it, as CSF ions are all secreted actively and therefore their concentration is some sort of deliberate choice by the choroid plexus. Even though the reasons for these differences in concentration are not always well established, let us try to muddle through them anyway:

  • The pH of CSF, though slightly lower, should remain similar to that of blood, as the central chemoreceptors rely on the pH of cerebral interstitial fluid, with which the CSF communicates. We are somewhat dependent on the accurate sampling of body fluid pH by these receptors, as they use that data to determine the respiratory pattern. As such, it would make no sense to weirdly alkalinise or acidify the CSF. Ergo, the pH of the CSF roughly corresponds to the pH of the bloodstream (Siesjö, 1972) and rapidly follows changes in blood pH of PCO(Andrews et al, 1994). This is an important point: the lack of protein in the CSF makes it a poor buffer, and therefore highly responsive to changes in arterial CO2.  The exact value of the pH may be slightly different in various textbooks (for example Brandis reports a normal CSF pH of 7.33) but everybody agrees that there is a close relationship between blood and CSF pH.  Even more importantly, phenomena which pathologically and unilaterally alter the pH of the CSF can directly influence the ventrolateral medullary respiratory centres, and change the respiratory rate. When Pappenheimer et al (1965) irrigated the brains of goats with a markedly acidified CSF (HCO3- 15 mmol/L) , their alveolar ventilation rate increased fourfold. The same is observed when meningitis or encephalitis alters the pH of the CSF: the patients are seen to hyperventilate. and their systemic pH may rise impressively (Paulson et al, 1964).
  • The PCO2 of CSF is slightly higher than that of blood (by 4-11 mmHg, according to Siesjö, 1972). As there is no haemoglobin or other protein buffers in the CSF, there is no Bohr/Haldane binding of CO2, and it resembles the CO2 value of other bloodless and deproteinated interstitial fluids. Moreover, it is closer to the PCO2 of venous blood (Merril et al, 1961), which makes logical sense, as cerebral interstitial fluid and venous blood are both destinations for cellular waste products.
  • The bicarbonate value may be slightly higher in CSF, as there are no proteins in the CSF to act as buffers, but the difference is usually minor. From calculations by Merril et al (1961), the bicarbonate value is just exactly what you would expect from the Henderson-Hasselbalch equation with CSF variables plugged in.
  • The chloride is higher in CSF for probably a similar reason, i.e the main anionic proteins are not present, and somebody's got to maintain electroneutrality, especially if you don't want your precious pH to change.  There's probably a more scientific way to explain this, but most papers that discuss CSF chloride mainly do so from the standpoint of using it as a diagnostic marker (eg. for meningitis) and don't really try to explain why it is raised at baseline.
  • Potassium levels in the CSF are lower than plasma, and independent of plasma levels. When Schein (1964) infused electrolytes intravenously into dogs,  with enough potassium they obviously died (at a serum potassium level of around 10mmol/L), but the potassium remained around 2 until blood flow ceased and ATP-powered pumps stopped working. This slightly adjusted image from the unclearly photocopied original clearly demonstrates this:influence of serum potassium on CSF potassium from Schein et al, 1964
    Why? Potassium is pumped out of the CSF at the apical surface of the choroid plexus cells, but it can reenter again through various channels, and so theoretically there should be no difference between plasma and CSF. The fact that there is such a major difference, and that it is defended in the face of life-ending hyperkalemia, suggests that it is important. Potassium is an essential element of excitable tissue activity, insofar that the tissue remains excitable only while the potassium remains low (maintaining a good transmembrane gradient). It makes sense that you would want to carefully control this variable, so that the excitability of your synapses could be regulated by other, more intentional and intelligent mechanisms. 
  • Sodium concentration is slightly lower in the CSF. (Johnson & Orlowitz, 1985). CSF is a non-ideal solution, so sodium activity is the same, even if the concentration is slightly different. As plasma sodium varies, CSF sodium tends to also fluctuate. For example, in the aforementioned animal experiments by  Schein (1964) serum sodium and CSF sodium both went north of 180 mmol/L when an infusion of hypertonic saline was administered. The main reason for this is thought to be paracellular leak (Ghaffari et al, 2020): the concentration of sodium ions is so high that it stands a good chance of crossing the choroid epithelium by brute force diffusion. Moreover, sodium must be pumped into the CSF by the action of Na+/K+ATPase in order for CSF secretion to occur, which means there is only so much regulation that could happen before the secretion of CSF is affected.
  • CSF magnesium and calcium are also independent of plasma levels. As both ions (but especially calcium) are potent intracellular regulatory actors and co-factors in metabolism, it would make no sense to let them swill around the brain in a completely unregulated manner. The total calcium concentration of the CSF is lower than plasma, as plasma contains albumin to bind calcium and the CSF does not, and so CSF calcium levels resemble ionised plasma calcium levels (1.0-1.2)mmol/L.
  • CSF glucose is lower than serum glucose and is usually 60-70% of the blood concentration. It is therefore somewhere in the  2-4  mmol/L range. It appears that it tracks blood glucose levels with a lag- if the BSL suddenly spikes, CSF glucose levels will follow over the next 2-4 hours. One might be tempted to make the assumption that the CSF glucose is slightly lower because some has been eaten by neurons, but this is almost certainly not the case - profound anaesthesia with isoflurane or propofol, which greatly decreases cerebral metabolism, was not shown to affect the CSF glucose in any way (Seisdedos et al, 2019).
  • CSF protein is obviously decreased because proteins would rely on active transcellular transport into the CSF, being too large to sneak across through tight junctions. The choroid epithelium is a very effective barrier for protein. As the result, CSF protein concentration is usually very low (usually no higher than 0.30g/L).

So, what does it taste like?

Salty. According to patients with post-operative dural tears (classically following a transsphenoidal resection of a pituitary tumour) the post-nasal drip of CSF has a slightly metallic salty taste, similar to tears. This is sufficiently commonplace that highly respected publically available patient information material from centres of excellence include this as one of the things about which you should "ask your healthcare provider". 

Like everything in the history of early medicine, where the most sensitive chemical apparatus available to the physician was often their tongue, CSF sampling by taste seems to have been carried out routinely. Schwab & Green, in A Case of Cerebrospinal Rhinorrhoea" (1905)casually drop the information that "many specimens of the fluid were examined, [and were] clear, odorless, and tasteless". 

Function of cerebrospinal fluid

As was already mentioned in a series of highly alarming remarks, CSF appears to be a mandatory part of having a complex central nervous system. So: let's answer the question, what exactly does it do? For exam purposes, you could summarise this as:

  • Barrier function (the blood-CSF barrier)
  • Chemical stability
  • Waste removal
  • Buoyancy
  • Mechanical cushion function
  • Hydraulic pressure buffering

More detail is available below, but is really not essential.  

Biochemical defence: the brain-CSF barrier

The CSF bathes the brain tissue and penetrates widely, communicating with the extracellular interstitial fluid of the brain through fenestrations in the pia mater. From this, it should follow that its composition should be tightly controlled, because it would make no sense to have a highly sophisticated blood-brain barrier control every exit and entrance from the bloodstream, and then to bathe the brain in some random liquid of uncontrolled composition. The control of the composition of CSF is therefore carefully regulated to be as benign as possible. As already mentioned, the filtration of plasma which takes place at the choroid plexus is a very deliberate process that excludes various troublemakers (large proteins, bacteria and virus particles, neurotransmitters, hormones, metabolic toxins, etc). In this sense, the organs which secrete CSF can be thought of as barrier organs, and the secretory surface can be thought of as a "blood-CSF barrier".

Chemical stability and metabolic waste clearance

The effect of CSF can be viewed as similar to the effect of dialysate fluid. Having a constantly flowing liquid of a controlled concentration circulate around your brain is beneficial because it constantly exposes regions of the brain to this controlled concentration, maintaining the concentration gradient for the removal of waste products and weird electrolytes. Consider what might happen if this was not going on. The brain has no lymphatics, and needs to rely on the flow of interstitial fluid through perivascular spaces for the clearance of metabolic byproducts (which ultimately washes into CSF, with which is its basically continuous, barring a few steps). If CSF was motionless and stagnant, these spaces would become regionally highly concentrated pools of cellular wastes and electrolytes, like the bottom of an aquarium when the pump has failed for a week.  Moving CSF however acts as a sink into which these wastes can easily diffuse, and the rate of its circulation is sufficiently fast to always present the perivascular spaces with a ready supply of fresh fluid.

The term used to describe this waste clearance system is "glymphatics", a term coined by Maiken Nedergaard to incorporate the contribution of the (g)lial cells to the process. Iliff et al (2013) were actually able to visualise this glymphatic system in action by the use of contrast-enhanced MRI, and confirmed that yes, it does in fact flush the perivascular spaces (and in fact that there is some retrograde reflux, were CSF from the subarachnoid space washed down along those perivascular channels, deep into the brain tissue). For the purposes of CICM exams, it would not be essential to become a master of this material; but in case more detail is called for, this "beginner's guide" by Jessen et al (2015) would be a good starting point.

Mechanical cushioning and buoyancy role

The words "cushion"or "buffer" are often seen in medical textbooks which try to explain the reasons for the physical presence of the CSF. Others refer to "a buoyant force to support the brain", which - being a squishy soft organ made of fat and filled with fluid blood - would, under the effects of gravity, naturally tend to deform into some kind of flattened pie. The net weight of the brain, while suspended in fluid, is only 50g (whereas it's mass is 1400g), these official sources state. As Abdalla (2021) rightly pointed out, it is impossible to weigh the living brain as it floats in CSF, and so these numbers must be pure speculation. However, mathematically, the relationship between brain density and CSF density can be incorporated into models which do conform to this supposed weight reduction, and which are capable of estimating the buoyancy of the brain well enough to accurately predict a change in its position intraoperatively (when some CSF is drained out). The most important buoyancy-related role is played by the ventricles, which act in the same way as the swim bladder of fishes. By sitting centrally inside the brain they can act as a low-density bubble of variable volume, buffering quick changes in brain density which occur with the pulse or respiration. 

Hydraulic pressure buffer

Cerebrospinal fluid plays a major role in the compensation for raised intracranial pressure. It can be displaced from the cranium into the spinal canal in order to compensate for abrupt changes in ICP, and it can be gradually reabsorbed via the arachnoid granulations to compensate for a sustained change. Interestingly, CICM examiners are divided on this matter. In Question 16 from the first paper of 2015 ("describe the physiology of cerebrospinal fluid") they remarked that "Better answers also discussed raised ICP and CSF’s role in the compensation for raised ICP". In contrast, in the college comments for Question 2 from the first paper of 2013 (also "describe the physiology of cerebrospinal fluid"), they complained that "Some candidates described the displacement of CSF when intracranial pressure rises as a function of CSF", and reported that "no marks were given for this". 

So does the CSF in fact have some sort of ICP-defending function? Could the college examiners be wrong? Reader, they could. Di Terlizzi & Platt (2006) list this as literally the first function of the CSF (though, to be fair, their veterinary article was mainly aimed at readers interested in the biology of companion animals). Under normal circumstances, the intracranial CSF volume and intracranial blood volume are traded against each other, and this is demonstrated in the pressure fluctuations observed during lumbar puncture. When the manometer or pressure transducer is attached, there is a 1-2 mmHg fluctuation in lumbar CSF pressure with arterial pulse (Herndon & Brumback, 1989, p. 36) which occurs because about 1.5ml of CSF moves in and out of the spinal CSF spaces with every arterial systole (Haughton & Mardal, 2014).  Similarly, the same "spatial displacement" occurs when intracranial volume changes because of some other incompressible presence (I'm looking at you, subdural haematoma). If there is time, quite a substantial amount of CSF can be displaced in this way- for example, in the context of a slowly growing tumour (Tameem & Krovvidi, 2013)

Circulation of cerebrospinal fluid

Under the heading "Circulation", most textbooks tend to include a crossectional image of the central nervous system with little arrows tracking the path of CSF flow, like rows of marching ants.  Here's a particularly formic diagram from The Anatomy Colouring Book:

CSF circulation from the anatomical colouring book

In case this schematic is uninformative for the non-visual learner, it can be easily expressed in words:

  • Out of the choroid plexus on the floor of the lateral ventricles
  • Through the foramen of Munro into the third ventricle
  • Through the Aqueduct of Sylvius, into the fourth ventricle
  • From the fourth ventricle, into the cisterna magna via the lateral foramina of Luschka and then up to the basilar cisterns around the pons, and then to the rest of the cortex
  • Or, from the fourth ventricle, via the medial foramen of Magendie, down to the spinal subarachnoid space
  • It is then reabsorbed from the subarachnoid space
  • The motor force driving this flow comes from:
    • Constant production at the choroid plexus
    • Arterial pulsation of the central nervous system structures
    • Venous pulsation of the CNS related to the respiratory cycle
    • Constant reabsorption by the arachnoid granulations

Or at least that is what the examiners want you to write in the exam papers, and what is often seen in physiology textbooks. This origin myth of the CSF, where 75% of it is created by the choroid plexus and then pushed around unidirectionally, is based on experiments which go back to Cushing (circa 1914), and which have all been debunked in the modern era.  Realistically, the situation is much more complex. For example, it is possible to secrete plenty of CSF even without a choroid plexus, and flow at every point is often demonstrably reversible under normal physiological conditions. An excellent overview of the different theories and their controversies awaits the patient reader in Mantovani et al (2018). To summarise,  rather than a unidirectional flow and a single dominant site of CSF secretion, there is "a dynamic equilibrium among parenchyma, interstitial space, [subarachnoid space], vascular compartment, and ventricles", where the direction and magnitude of regional flow seems to be determined mainly by regional hydrostatic and osmotic pressure gradients. This concept will take some years to percolate into official college textbooks, and the exam-going candidate should cultivate some doublethink about it.

Absorption of cerebrospinal fluid

The rate of CSF production being around 25ml/hr, and there being about 150mls in the normal adult CNS, leads to the conclusion that there must be a constant turnover of this fluid, of around four exchanges per day. This cycling is possible because of the continuous reabsorption of the CSF by arachnoid granulations and cervical lymphatics.

So what are these "arachnoid granulations" and "arachnoid villi"? In short, these are weird little herniations of arachnoid space material that protrude into the cavity of dural venous structures. They are normally very small; villi can be microscopic, and to get called a "granulation" you'd have to be 2mm in diameter. They increase in size with age and anything larger than 10mm gets called "giant" because it can become clinically significant by obstructing venous flow through the sinus. Instead of gross cadaver images or uninspiringly grey electron microscopy, some truly excellent pencil artwork from Shabo & Maxwell (1968) will be used here to illustrate these structures:

 arachnoid granulation from Shabo & Maxwell

The absorption of CSF from the loose mesh space inside the granulation occurs mainly by ultrafiltration. The pressure of the CSF is usually higher than the venous pressure, particularly in the upright human, and even when supine there will be periods during the respiratory cycle when the venous pressure drops. When the CSF pressure drops below 7 cm H2O, the hydrostatic gradient driving this bulk flow of fluid becomes negligible, and CSF reabsorption stops (Orešković & Klarica, 2014).

Apart from arachnoid villi and granulations, multiple other sites of CSF absorption are recognised, including through the walls of the ventricles, and through the cervical lymphatics. The exact dynamics of how this works or what regulates it remains to be firmly established, and there's some excellent literature to discuss this evolving area (eg. Miyajima & Arai, 2015). As CICM examiners are naturally drawn to well-described phenomena found in textbooks, it would probably be better not to mention this, sticking to simple explanations with arachnoid granulations and hydrostatic gradients.


Sakka, Laurent, Guillaume Coll, and Jean Chazal. "Anatomy and physiology of cerebrospinal fluid." European annals of otorhinolaryngology, head and neck diseases 128.6 (2011): 309-316.

Segal, Malcolm B. "The choroid plexuses and the barriers between the blood and the cerebrospinal fluid." Cellular and molecular neurobiology 20.2 (2000): 183-196.

Brown, P. D., et al. "Molecular mechanisms of cerebrospinal fluid production." Neuroscience 129.4 (2004): 955-968.

Zagórska-Świeży, Katarzyna, et al. "Arterial supply and venous drainage of the choroid plexus of the human lateral ventricle in the prenatal period as revealed by vascular corrosion casts and SEM." Folia morphologica 67.3 (2008): 209-213.

İmamoğlu, Hakan, et al. "Evaluation of the normal choroid plexus size in adults with Magnetic Resonance Imaging." Cumhuriyet Medical Journal 35.1 (2013): 51-54.

Keep, R. F., and H. C. Jones. "A morphometric study on the development of the lateral ventricle choroid plexus, choroid plexus capillaries and ventricular ependyma in the rat." Developmental Brain Research 56.1 (1990): 47-53.1

Damkier, Helle, and Jeppe Praetorius. "Structure of the Mammalian Chorod Plexus." Role of the Choroid Plexus in Health and Disease; Praetorius, J., Blazer-Yost, B., Damkier, H., Eds (2020): 1-33.

Ariens-Kappers, Cornelius Ubbo, and Eduard H. Strasburger. Anatomie comparee du systeme nerveux, particulierement de celui des Mammiferes et de l'Homme. F. Bohn, 1947.

Jones, H. C. "Comparative aspects of the cerebrospinal fluid systems in vertebrates." Science Progress (1933-) (1979): 171-190.

Murray, M., et al. "The blood-brain barrier and ventricular system of Myxine glutinosa." Brain research 99.1 (1975): 17-33.

Abbott, N. Joan, Nancy J. Lane, and M. Bundgaard. "The Blood‐Brain Interface in Invertebrates a." Annals of the New York Academy of Sciences 481.1 (1986): 20-42.

Milhorat, Thomas H. "Structure and function of the choroid plexus and other sites of cerebrospinal fluid formation." International review of cytology 47 (1976): 225-288.

Orešković, Darko, and Marijan Klarica. "Development of hydrocephalus and classical hypothesis of cerebrospinal fluid hydrodynamics: facts and illusions." Progress in neurobiology 94.3 (2011): 238-258.

Buishas, Joel, Ian G. Gould, and Andreas A. Linninger. "A computational model of cerebrospinal fluid production and reabsorption driven by Starling forces." Croatian medical journal 55.5 (2014): 481-497.

Speake, Tracey, et al. "Mechanisms of CSF secretion by the choroid plexus." Microscopy research and technique 52.1 (2001): 49-59.

Ames III, Adelbert, Michio Sakanoue, and Shinichiro Endo. "Na, K, Ca, Mg, and Cl concentrations in choroid plexus fluid and cisternal fluid compared with plasma ultrafiltrate.Journal of neurophysiology 27.4 (1964): 672-681.

Merritt, H. Houston, and Frank Fremont-Smith. "The cerebrospinal fluid." The cerebrospinal fluid. 1938. 333-333.

Siesjö, Bo K. "The regulation of cerebrospinal fluid pH." Kidney international 1.5 (1972): 360-374.

Andrews, Russell J., John R. Bringas, and Gina Alonzo. "Cerebrospinal fluid pH and PCO2 rapidly follow arterial blood pH and PCO2 with changes in ventilation." Neurosurgery 34.3 (1994): 466-470.

Pappenheimer, J. R., et al. "Role of cerebral fluids in control of respiration as studied in unanesthetized goats." American Journal of Physiology-Legacy Content 208.3 (1965): 436-450.

Paulson, Olat'B., et al. "Regional cerebral blood flow, cerebral metabolic rate of oxygen, and cerebrospinal fluid acid‐base variables in patients with acute meningitis and with acute encephalitis." Acta Medica Scandinavica 196.1‐6 (1974): 191-198.

Merril, Carl R., Harry W. Seipp, and Peter C. Luchsinger. "Total CO2, PCO2, and pH in human spinal fluid." Journal of applied physiology 16.3 (1961): 485-487.

Johnson, DOUGLAS C., L. Y. N. N. Orlowitz, and BERNARD M. Hitzig. "Differences between CSF and plasma Na+ and K+ activities and concentrations." American Journal of Physiology-Regulatory, Integrative and Comparative Physiology 248.5 (1985): R621-R626.

SCHAIN, RICHARD J. "Cerebrospinal fluid and serum cation levels." Archives of neurology 11.3 (1964): 330-333.

Ghaffari, Hamed, et al. "Regulation of CSF and brain tissue sodium levels by the blood-CSF and blood-brain barriers during migraine." Frontiers in computational neuroscience 14 (2020): 4.

Joborn, Charlotte, et al. "Cerebrospinal fluid calcium, parathyroid hormone, and monoamine and purine metabolites and the blood-brain barrier function in primary hyperparathyroidism." Psychoneuroendocrinology 16.4 (1991): 311-322.

Fishman, R. A. "Studies of the transport of sugars between blood and cerebrospinal fluid in normal states and in meningeal carcinomatosis.Transactions of the American Neurological Association 88 (1963): 114-118.

Seisdedos, Alejandro, et al. "Anesthetic effects of isoflurane and propofol on cerebrospinal fluid biochemical markers in healthy dogs." Veterinary clinical pathology 48.2 (2019): 270-275.

Di Terlizzi, Roberta, and Simon Platt. "The function, composition and analysis of cerebrospinal fluid in companion animals: Part I–Function and composition." The Veterinary Journal 172.3 (2006): 422-431.

Tumani, Hayrettin. "Physiology and constituents of CSF." Cerebrospinal Fluid in Clinical Neurology. Springer, Cham, 2015. 25-34.

Gilpin, Sherman F., and Thomas B. Earley. "Drainage of Cerebrospinal fluid as a Factor in the Treatment of Nervous Syphilis." Journal of the American Medical Association 66.4 (1916): 260-262.

Iliff, Jeffrey J., et al. "Brain-wide pathway for waste clearance captured by contrast-enhanced MRI." The Journal of clinical investigation 123.3 (2013): 1299-1309.

Jessen, Nadia Aalling, et al. "The glymphatic system: a beginner’s guide." Neurochemical research 40.12 (2015): 2583-2599.

Abdalla, Mohamed Abdelrahman. "The Buoyant Brain." ScienceOpen Preprints (2021).

Tameem, Alifia, and Hari Krovvidi. "Cerebral physiology." Continuing Education in Anaesthesia, Critical Care & Pain 13.4 (2013): 113-118.

Mantovani, Giorgio, et al. "Controversies and misconceptions related to cerebrospinal fluid circulation: a review of the literature from the historical Pioneers’ theories to current models." BioMed research international 2018 (2018).

Weiss, Martin H., and Nancy Wertman. "Modulation of CSF production by alterations in cerebral perfusion pressure." Archives of neurology 35.8 (1978): 527-529.

Upton, Margaret L., and Roy O. Weller. "The morphology of cerebrospinal fluid drainage pathways in human arachnoid granulations." Journal of neurosurgery 63.6 (1985): 867-875.

Shabo, Alan L., and David S. Maxwell. "The morphology of the arachnoid villi: a light and electron microscopic study in the monkey." Journal of Neurosurgery 29.5 (1968): 451-463.

Orešković, Darko, and Marijan Klarica. "Measurement of cerebrospinal fluid formation and absorption by ventriculo-cisternal perfusion: what is really measured?." Croatian medical journal 55.4 (2014): 317-327.

Miyajima, Masakazu, and Hajime Arai. "Evaluation of the production and absorption of cerebrospinal fluid." Neurologia medico-chirurgica (2015): ra-2015.