I've been working on a protocol standardization project where, among other things, we want protocols to be able to be run equivalently by both humans and robots.

Something that I've noticed in doing so is that there's a radically different amount of information typically assumed needed by a human vs. a robot when it comes to pipetting.

With a robot, it is generally recommended to understand the nature of the liquid thoroughly, specify the liquid class (which includes information on viscosity, volatility, polarity, etc), and do quite a bit of quality control to make sure than when you want to pipette 100 µL you actually get 100 µL. A typical example is shown on this page by Hamilton.

With a human experimenter, however, the protocol pretty much always just says "pipette 100 µL" and assumes that the human can figure out all of the potential issues that we have to be so careful for with robots.

Is this actually a safe assumption, or just something that we don't typically check during protocol development for humans? Is it really intuitively obvious for nearly all liquids, or should our protocols be providing more guidance for humans as well?

  • 2
    $\begingroup$ I don't wish to be overtly critical, but how is this a question about biology? I have already removed the incorrect tag "synthetic biology", but even as a methods question it seems to be only answerable in a subjective sense as human psychology, or in a legal sense in relation to the laws of a particular jurisdiction relating to some form of industrial safety. $\endgroup$
    – David
    Commented Mar 5, 2021 at 17:24
  • 11
    $\begingroup$ @David I'm really not sure how you're coming to that conclusion: this is a problem of biological protocol replicability. $\endgroup$
    – jakebeal
    Commented Mar 5, 2021 at 18:18
  • 6
    $\begingroup$ @David I've added the lab-techniques tag as well; if you disagree with this question, then what do you find different from other lab-techniques questions on this site? $\endgroup$
    – jakebeal
    Commented Mar 5, 2021 at 18:33
  • 3
    $\begingroup$ @David Interesting; my own assessment, which supported my choice to ask, is that this falls solidly on the good subjective side of the bar, but of course that is, itself, a subjective judgement. $\endgroup$
    – jakebeal
    Commented Mar 6, 2021 at 0:45
  • 10
    $\begingroup$ For what it's worth, I find this a particularly interesting question which already has interesting answers and a very refreshing change from most questions we get. One of the long term problems we've had here is the relative dearth of professional biologists asking professional-level questions so I am really surprised to see this already has two close votes! I would have thought this is exactly the sort of question we want to attract and support. $\endgroup$
    – terdon
    Commented Mar 6, 2021 at 14:26

4 Answers 4


In my experience it is very rare to see a protocol for humans that describes how to pipette a certain liquid, but I don't have as many years pipetting as others do.

In general, it is left to the experimenter to observe the behavior of the liquid and act accordingly. It is quite easy to determine when a liquid is leaking (e.g. ethanol) or when it is very viscous (e.g. glycerol), and determine that a change in technique is required. In this case, I would ask someone else in the lab how to deal with the situation. This is something that robots cannot do.

However, not everyone has expert help available, and although it is fairly obvious when a technique isn't working, often the solution is not intuitive. Additionally, the advice of others in the lab could be unhelpful or even inaccurate. For example, I was told to cut the tip off a pipette in order to pipette Tween-20 more easily, however in reading for this answer have discovered that this is not recommended at all! (See eppendorf FAQ linked below).

Therefore, I would say that, unless the liquid is to be pipetted in the same way as water, it is always useful to provide information for those who are not knowledgeable or do not have easy access to expert practitioners.

For some examples of the sorts of problems that experimenters come up against, and therefore which liquids especially could use instructions in protocols, see this set of FAQs from eppendorf on handling liquids.

  • 4
    $\begingroup$ Wow... I think it's particularly revealing how answering this question actually ended up causing you to discover that an "oral tradition" answer was incorrect! It seems that pipetting actually has a lot of knowledge required in order to do it correctly? $\endgroup$
    – jakebeal
    Commented Mar 5, 2021 at 12:26
  • 5
    $\begingroup$ As with so many things we do in the lab, we are often in a rather high local minima of 'what worked for postdoc Bob 15 years ago and has been handed down ever since' $\endgroup$ Commented Mar 5, 2021 at 12:40
  • 5
    $\begingroup$ Many don't know that pipettes are calibrated for water solutions. Organic solvents (like ethanol) have a different partial pressure and can't be measured with a water calibrated pipette. This is crucial in some experiments. $\endgroup$
    – Hachiloni
    Commented Mar 5, 2021 at 14:01
  • $\begingroup$ The most finicky pipetting situations I've encountered are with regard to emulsions. The worst example I've encountered is the 10x Chromium, which asks you to dispense 95uL on a P-200 multichannel over the course of 30s to avoid breaking emulsions. But plenty of other emulsion-based protocols (often single-cell-related) specify slow pipetting $\endgroup$ Commented Mar 9, 2021 at 18:18

I work on software that controls one model of robotic pipetter. There are a few reasons why we use complex liquid class definitions:

  1. One of the selling points of using a robot is high speed. You can handle a wide range of viscosities if you set a conservative (slow) speed for aspiration/dispense, but for solutions that don't require this, we'd like to be able to move more quickly. In many cases, the different liquid classes are about optimizing the time required while still maintaining the desired precision and accuracy.

  2. Robotic pipetters will often dispense multiple times after a single aspiration of a larger volume. To do this accurately, we compensate for a variety of effects, including the effect of the column pressure, which depends on the reagent density. In the end, this is also largely about improving speed.

  3. In some applications, precision is especially important and you can use settings for the robot that improve precision. Let's say you have something that is extremely sensitive to getting the right pH. You could pre-titrate a buffer to the correct pH and load that into the system, but if you're trying to produce an assay grid on a microplate, you'd have to make a bunch of buffers. A well-calibrated robotic system can mix from stock buffers to produce the assay directly, since it has the necessary precision for mixing very small volumes.

Not having much personal lab experience since school, I can't speak to how precisely/accurately pipetting is done in labs by human beings. It probably varies quite a bit from lab to lab and likely depends on the skill and tiredness level of the person doing the work. I did it enough times to know that even basic pipetting of aqueous solutions has a bit of a knack to it and was glad when I was done with the labs that required it. In any case, I thought it might be useful to have an idea of why the robotic control seems so much more complicated; it's mostly because robots are trying to do it faster and without the benefit of being able to observe the reagent's behavior while using it.

  • $\begingroup$ I appreciate you sharing your expertise about the robotics, but that's the side that I'm more familiar with as well. I'm afraid however, that as currently written this answer isn't shedding terribly much light for me on the question about whether humans need more information than they are typically given. $\endgroup$
    – jakebeal
    Commented Mar 6, 2021 at 21:42
  • $\begingroup$ @jakebeal, it doesn’t quite say it explicitly, but I interpreted this answer as saying “humans don’t pipette in the quite the same way, and therefore may not need as many details.” $\endgroup$ Commented Mar 6, 2021 at 22:20
  • $\begingroup$ @MattKrause Agreed; I was hoping for a little bit more description and basis for that conclusion. $\endgroup$
    – jakebeal
    Commented Mar 6, 2021 at 22:24
  • $\begingroup$ @jakebeal I'm not aware of nay research on the topic, but I would suspect that even if you give explicit instructions to human operators, they might try to apply them, but most likely they'll largely fall back on their learned muscle memory and habits from their training/practice. In a sense, I'd say the humans already have the liquid class information in their brains, to the extent that they've encountered those types before. Humans also have the major benefit of being able to look at the pipette tip and see if there's still fluid stuck inside and can adapt accordingly. $\endgroup$
    – Dan Bryant
    Commented Mar 7, 2021 at 1:24

Just to be clear, it has been many years since I have done real lab work, so I am certainly not an expert on the specific question. However, I have a fair amount of experience writing clear instructions to cover complex (and often bespoke) processes.

As Dan Bryant has explained, robotic systems absolutely need the additional detail in order to complete their tasks optimized for speed (and a control method without the built in feedback loops that come with a human).

My experience writing procedures for humans is that same level of detail will actually be a significant negative, at least on techniques that are considered standard knowledge. Providing what humans consider 'unnecessary detail' tends to result in a problem where your procedure is 'skimmed' for key information, which may result in lower adherence rather than improved results.

Another consideration is your level of expertise - do you know every situation where a task specific pipette is required? Because if you specify it once, you create an expectation that you will specify it every time, so any situation you miss may cause an error, and the probability of error is higher with operators having less experience - the primary audience for detailed procedures.

There are solutions, but they tend to depend on your documentation standards and the documentation tools in use. The simplest is often to move the robot only detail into an appendix (or something similar, such as a collapsing box). This means the person programming the robot will be able to drag the detail out when required, but the 'visual noise' for the humans will be reduced.

  • $\begingroup$ Re: specifying robot details, most vendors will ship liquid class information with the tool covering the majority of cases, so you generally only need to go through a careful calibration process on your own if you need particularly tight performance or you have an unusual fluid. The parameters and configuration process will also vary between vendors/tools, since, among other things, there are a variety of different technologies in use that all operate slightly differently. It's worth specifying the liquid class you used for replication, but would be hard to distill to tool-independent info. $\endgroup$
    – Dan Bryant
    Commented Mar 7, 2021 at 14:25
  • $\begingroup$ That is a very good point about including unnecessary detail. If this information is included it should be in a subtle/optional way so that it won't distract people who don't need it. $\endgroup$ Commented Mar 7, 2021 at 17:23

In my experience, I've never seen more than "pipette 100 μl" or at most "caution viscous liquid". Also I've only been taught how to pipette in a general manner and never about how to do it with special liquids except for the "cut the tip off" advise so before Noah's answer I had never heard of reverse pipetting or using a multivette to accurately pipette glycerol for example as described on the eppendorf website. However, when pipetting glycerol for example I could tell that I needed to pipette it slowly and I could check visually that I had (about) the right volume using the marks on the tip. So compared to a robot I can use my eyes to adjust and it's good enough in my particular case and with a single tube it is not optimal and it may not hold true with a multichannel pipette so having good and better pipetting habits would be better for reproducibility.

Now if I can see the "cut the tip" advice and the standalone "pipette 100 μl " instruction work to some extend, "commercial" grade protocol would probably have the instruction written as "using a multivette (or using the reverse pipetting technique) pipette 100 μl".

Another thing to consider is the definition of protocol which if we take wikipedia's definition is not just a succession of steps to perform :

In addition to detailed procedures, equipment, and instruments, protocols will also contain study objectives, reasoning for experimental design, reasoning for chosen sample sizes, safety precautions, and how results were calculated and reported, including statistical analysis and any rules for predefining and documenting excluded data to avoid bias.

What we usually consider as a protocol is the "Quick protocol" with only the steps to perform where you're more likely to find the "pipette 100 μl" type instruction. But if you consider the more detailed protocol you'll find more information. I don't have an example for liquid handling but if you consider the qiagen miniprep protocol depending if you take the quick version or the full one you may not be doing the same thing. Let's take step 2 as an example :

Add 250 μl Buffer P2 and mix thoroughly by inverting the tube 4–6 times until the solution becomes clear. Do not allow the lysis reaction to proceed for more than 5 min. If using LyseBlue reagent, the solution will turn blue.

Add 250 μl Buffer P2 and mix thoroughly by inverting the tube 4–6 times. Mix gently by inverting the tube. Do not vortex, because this will result in shearing of genomic DNA and contamination of plasmid. If continue inverting the tube until the solution becomes viscous and slightly clear. Do not allow the lysis reaction to proceed for more than 5 min.

If LyseBlue has been added to Buffer P1, the cell suspension will turn blue after addition of Buffer P2. Mixing should result in a homogeneously colored suspension. If the suspension contains localized colorless regions, or if brownish cell clumps are still visible, continue mixing the solution until a homogeneously colored suspension is achieved.

If we consider only the mix part here, "gently" is the key word. If you've been handed only the quick protocol and you're an inexperience user even just by inverting too strongly you may end up shearing the genomic DNA and thus contaminate your sample what can impact your downstream application.

So the bottom line is that "pipette 100 μl" may be enough for experience users and/or some applications on a quick protocol but you may also want to have the instruction "pipette 100 μl using a multivette (or using the reverse pipetting technique)" on a more detailed protocol for inexperience user or troubleshooting purposes.

  • $\begingroup$ Thank you for the information about the miniprep protocol! I'd always assumed that thoroughly meant that it didn't need to be/shouldn't be gentle. I normally shake the tube viciously for that step... $\endgroup$ Commented Mar 9, 2021 at 15:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .