Skip to main content

Metaphors, Good and Bad

Block That Metaphor!

Phillip Ball had an interesting piece recently over at Nature News, which touches on a subject that I’ve also thought about: when does metaphorical thinking help, and when does it hurt? (I’ve got a whole category on the blog on this topic, although I haven’t filled it up with as many posts as I’ve meant to).
As he mentions, there’s no some empirical evidence that metaphors can influence the way we think about a situation, and not in ways that we’re consciously aware of. I think that we’re particularly vulnerable to this effect in scientific research, because so many of our concerns are outside day-to-day experience. I don’t see any way around this: we can’t see a G-protein-coupled receptor in action, so we come up with a mental picture to help. We can’t visualize the complexities of a biochemical pathway in toto, so we reduce it to a useful simplification. Well, we hope it’s useful, anyway.
I have a number of these mental constructs – for example, when I’m picturing a protein surface or a binding pocket, I have a tactile image of something like firm gelatin with a hard surface underneath (ball bearings or pool balls, depending on the scale I’m picturing). Thinking about it, I know where that image comes from – it’s from standard molecular graphics representations of van der Waals surfaces around atoms. The charge distributions on the surface come across to me mentally as warm and cold areas, or perhaps sour and sweet. The first of those is probably because many graphics programs represent charges as red and blue; the taste metaphor seems to be my own brain’s contribution – characteristically vivid, but of uncertain utility.
In case you’re wondering, I do audio, too. Protein-protein surfaces seem, in my mind’s eye, to be mildly sticky, which is probably my impression of an overall hydrophobic effect. The charged surfaces, when they come apart, do so in my head with a tactile peeling effect and a faint sound of Velcro.
Now, does my forebrain’s special effects budget make me a better medicinal chemist? Who knows? If I’ve got the wrong impressions, and if I act on them too strongly, they might make me a worse one. The same with other metaphors, both the internal ones and the ones we produce for others. A bad metaphor can do more harm to the people you’re trying to teach than good.
This also goes for the metaphors that people bring with them when they think about what we do in drug discovery. I think, for example, that people who design and build complex human-produced systems are prone (naturally enough) to believing that biochemistry and drug design must be similar processes, and thus subject to the same engineering approaches. Those of us wrestling with these problems are stuck trying to explain that not only are living systems more complex, they’re complex in a different way as well – you’re looking at differences of both degree and kind. But if you’re used to circuit diagrams, programming flow charts, or chip design, then you’re naturally going to see those when you look at diagrams of biochemical pathways.
The best “harmful metaphor” example I can think of at the moment is the importation of agonist/antagonist nomenclature into the nuclear receptor field. I’d like to find whoever did that and whack them on the head with a board. That misled me when I first started working in the area, and I’ve seen it mislead countless others since then. “They’re pretty much like GPCRs” is the impression given to the unwary, but that’s a tall glass containing 5% refreshment and 95% toxic sludge. You have to spend a lot of energy getting it out of your head if you want to have a chance to understand what’s going on with those targets, inasmuch as anyone does.
But there may be a larger example: the whole reductionist approach of target-driven drug discovery. That’ll be the subject of another post. . .

28 comments on “Block That Metaphor!”

  1. Monte Davis says:

    We’ll be another generation or three getting past the “blueprint” and “program” metaphors for DNA. That DNA bears information was an extraordinarily valuable insight — but the rest of the baggage from those metaphors does at least as much harm as good.

  2. NoDrugsNoJobs says:

    You hit the nail on the head with your mention of nuclear receptors. The metaphor in my head for a nuclear receptor is of a “black box” where we know a ligand might go in but what comes out by way of activity depends on so many things…..

  3. Metaphors help people start to understand a concept, but they don’t help them finish.
    Is the blueprint metaphor for DNA that bad? A house is the sum of the plans used to build it and the changes inevitably made by the builders and contractors either by mistake or on purpose. A living thing is the sum of the plans contained in the DNA and the execution of those plans.

  4. RM says:

    The most vivid metaphor-busting experience I’ve had was when I saw some of the HHMI replication/transcription/translation videos. (Like this one.) I’d been conditioned by molecular graphics programs to think of molecules as being smooth and hard – like a bunch of billiard balls stuck together. I was completely neglecting the scale of the variation due to entropy, thermal fluctuations and quantum delocalization effects. Molecules aren’t smooth and hard like billiard balls, they’re more fuzzy (furry even).
    Matthew@3 – The problem with the blueprint metaphor is that a blueprint is typically a scaled drawing of the building. Applying that to humans brings back echos of the homunculus concept – that there’s a tiny scaled down human inside sperm/eggs which then grows/is copied to make a complete human. There is no “heart” technical spec in the DNA which exactly specifies each component and where they go.
    I think a much better metaphor (though still incomplete) is DNA as recipe. When you make a cake, you don’t have a scale drawing of structure of the crumb and specification on the crumblines/elasticity, etc. Instead you have a description of what ingredients you use and the timing of when you add them, but ultimately you just throw them all together in a bowl and mix them up. The final product is not the result of the technical specification of the recipe itself, but more a function of what ingredients you used and how they interact. Little variations in quantity and timing, as well as how they are treated can have a big effect on the end result.
    A recipe metaphor captures more of the flexible, bottom-up, emergent properties of the DNA to human transition, whereas the blueprint or program metaphor is a more top-down, rigid, dictatorial type view.
    The really bad metaphors are the ones which “kinda” work. If there isn’t any obvious holes, you may get lulled into complacency about the non-obvious holes.

  5. Hasufin says:

    The reality is, communication is HARD. The goal with a metaphor is that the person understands the concept for which the metaphor is used. This is a good goal, and metaphors are useful tools. But…
    Well, is the person familiar with the metaphor? I can tell someone that DNA is the source code for a living organism, and RNA is the compiled code… and it’s a great metaphor – for a computer scientist. Everyone else is left scratching their heads.
    Or, worse yet, what if my hypothetical listener THINKS they understand it; they smile and nod, ask questions and seem to respond like they Get It… and then they ask how to decrypt the code. Oops.
    Or, they get more caught up in the metaphor than the actual lesson, and I could end up spending an hour explaining compilers and virtual machines, and never get to how DNA works.
    But these are all pitfalls inherent in any learning tool. THere’s no direct way to download comprehension into somebody’s brain, so all we can do is try to check our assumptions and seek after the understanding we lack.

  6. Chemjobber says:

    The lock-and-key (or hand-and-glove) metaphor for enzyme active sites is a major stumbling block for students from freshmen in high school biology until ~junior year of college (biochemistry.)
    There are probably MDs who still labor under that metaphor (that, as has been said) will only take you so far in understanding any particular problem.

  7. PTM says:

    Another example of a bad metaphor is the “histon code.”
    Histon modifications are not a real code and thinking of them that way is counterproductive. The meaning of each modification is very context sensitive, and there is no reason to believe all those individual modifications combine into some unified output.
    In other words they don’t encode anything concrete, rather each modification is an independent fingerprint of some particular class of molecular events locally taking place and may in turn be utilized by subsequent molecular events for varying purposes.
    A much better metaphor would be histon modifications as localized signals. But even this one is not completely satisfactory as it ignores the possibility that the purpose of certain modifications may not be to signal something but rather to modify some local physical properties of the packed double helix.
    Another terrible metaphor is “junk DNA.” It may not encode proteins, but it certainly has many important roles: structural, in the way it affects nuclear organization, regulatory as it affects local expression levels, as a pool from which to draw future genes, potentially as a defense against viruses/transposable elements, and likely many others.

  8. sutemi says:

    I am very bad at visual learning and visual metaphors; in fact I have no visual memory/imagination. I can easily imagine tactile memories or hear songs in my head or taste last night’s dinner but cannot imagine any visual scenes, faces or colors. I compensate using other senses as much as possible but many fields, including chemistry, are very visually-oriented and use visual metaphors.

  9. luysii says:

    Sorry guys — we always think in terms of metaphor, relating what we don’t know to what we do by them. A donkey is a kind of horse etc. etc. There are good metaphors and bad metaphors, however.
    I’ve always liked comparing an Sn2 reaction to an umbrella turning inside out in the wind. Who knows, it may even be original.
    #7 — the histone code IS a code, but as you note, it is context dependent. Computer codes initially were never context dependent (see Chomsky’s classification). But not all codes are context independent. The histone code is one such.

  10. Matt D says:

    It’s impossible for our puny minds to comprehend nature without sculpting it into a simplified model. Metaphor is one valuable aspect of that model.
    In science, it is vital that we re-evaluate our assumptions, whether they are assumptions of the experiment or of the nomenclature. If we call crime a “virus”, ask how we might treat it differently if it were a “beast”. I think that may be the most important quality we learn in our education, and probably not just in the scientific fields.

  11. bgg says:

    One must remember that all language is metaphorical. Mathematics included.

  12. bgg says:

    One can extend that thought to all sensory perception as well. A nice metaphor from Wittgenstein’s Tractatus relates:
    My propositions are elucidatory in this way: he who understands me finally recognizes them as senseless, when he has climbed out through them, on them, over them. (He must so to speak throw away the ladder, after he has climbed up on it.)
    He must surmount these propositions; then he sees the world rightly.”

  13. Pete says:

    I like to think of molecular recognition as a tango. Lots of surface contact and the movements of the dancers perfectly synchronised in space and time. However, this is a dance in which no partner leads.

  14. Anonymous BMS Researcher says:

    When I was listening to discussions of SAR in a meeting today, I fear the metaphor that leaped into my head as it frequently does in such meetings was the Three Witches scene from Macbeth. “Adder’s fork and blind-worm’s sting, Lizard’s leg and owlet’s wing, For a charm of powerful trouble, Like a hell-broth boil and bubble.”

  15. PTM says:

    luysii: “the histone code IS a code, but as you note, it is context dependent…”
    Yes, code is a code, and you can see codes everywhere if you are determined enough, but that is not the point, what matters is whether the code metaphor is useful or not.
    In the case of histone modifications it is not, thinking of them as a code doesn’t illuminate anything, there is no one thing being encoded, and there is no clear code to be deciphered, and without those two applying the code metaphor does more harm then good.

  16. Serpico says:

    What’s wrong with the agonist/antagonist concept? I don’t like it from the very start. What’s really in my mind is that a receptor functions by reaching one kind of its conformations. An agonist could anchor the receptor near this kind of conformations; An antagonist could forbid the receptor from reaching this kind of conformations.
    I was wondering if Derek thinks the same way as me. If not, what is it?

  17. Allchemistry says:

    Another example of a metaphor going bad: “Survival of the fittest”.

  18. KB says:

    As a molecular modeller I’d say the glossy hard sphere atomic representation is useful but quite misleading.
    Atoms are mostly vacant space – to put things into a larger scale I imagine something like a sphere of gas but with a fuzzy undefined outer limit (i.e. not like a balloon). I like the gas sphere analogy because I can envisage size and density, e.g. Bromine is larger but also less dense than a small relatively hard fluorine atom, which helps me imagine how the gas spheres would bump together in a tactile sense.
    One way to get a feel for these imaginary gas spheres is to play with magnets – gently pushing together and moving around poles of opposite sign.

  19. @4 RM — I agree wholeheartedly on the recipe metaphor, and I’ve used it. But there have been times when the recipe metaphor just wouldn’t work, and blueprint did.
    @12 bgg — wow. that quote is perfect.
    @17 — “survival of the fittest” isn’t really a metaphor, it’s a description/truncation/exaggeration/catchphrase.
    Metaphors are a trick to get your brain to start thinking about something. Understanding a concept often involves finding the places where the metaphor stops working.
    Language is a pretty inexact tool for conveying concepts — except we don’t have anything better.

  20. Luysii says:

    #15 — interesting set of ideas. Consider this. We’ve all accepted the idea that the 3 nucleotide codon for amino acids is a type of code. However, if you look at the nucleotides individually you will see that the amino acid they code for depends on the adjacent nucleotides (e.g. is context dependent).
    Moreover, the 3 nucleotide codons code for more than just amino acids — there’s been a lot of work showing that codons synonymous for a given amino acid have other functions (rate of protein synthesis varies with how much of the appropriate tRNA is around, determination of mRNA splicing variants etc. etc. ) So “there is no one thing being encoded” but it’s still a code.
    The histone code (whose complexity is still being determined) tells us when chromatin is active or inactive, and what particular proteins bind to a given nucleosome affecting transcription, replication and DNA repair. That it’s used to affect more than one cellular process doesn’t mean it isn’t a code.

  21. RKN says:

    My entry for a bad metaphor (presumably for mass action):
    When protein X becomes phosphorylated, it “recruits” other factors required to activate process Y.
    Recruits? What picture in our head is that intended to draw?
    “Hey everyone, look, I’m phosphorylated, come bind with me!”

  22. Boghog says:

    Concerning nuclear receptors, I think the agonist/antagonist concept is at least a useful starting point for discussion. In narrowly defined contexts (i.e., one response element in one cell type), the analogy to GPCRs holds up fairly well. The reality of course becomes much more complicated in a whole organism where the same ligand can be an agonist in one tissue and an antagonist in a second (i.e., mixed agonist/antagonists).

  23. Scott says:

    What’s that you say, you draw Lewis structures? You should think in terms of hybrid orbitals. What? Oh, yes, you should be looking at that protein from an MO perspective. Well, aaaaactually, MOs are nonsense…you really need to think in terms of *states*….
    I hate to say it, but ‘it’s metaphors all the way down’. We describe everything in chemistry (which is, after all, the study of a bunch of overlapping wave functions)in terms of metaphors, since we can’t actually see any of it, it *isn’t* like springs, or spheres, or balls and sticks, or whizzing electrons, or any of that at some level. You can have simple models or complex models. They’re all models.
    I think you can separate them into a 2×2 grid:
    I: Easy; Explains Lots (good metaphor)
    II: Hard; Explains Lots (too true to be good)
    III: Easy; Explains Little (too good to be true)
    IV: Hard; Explains Little (bad metaphor)
    Obviously, you want lots of Type I metaphors. You don’t ever want Type IV metaphors (those are the bad ones). Early in our careers, we rely mostly on I and III, and as we become more sophisticated, we lean more on I and II. There’s often an inverse correlation between how “true” a metaphor is, and how easy it is to understand its uses.
    Good scientists move up and down this ladder, and grow comfortable with holding multiple metaphors in their head at once.

  24. Still Scared of Dinosaurs says:

    If metaphors are going to work as educational tools we need to know two things about them. The first, obviously, is what they are good at explaining. The real education starts, however, when you can also explain where the metaphor breaks down.
    If all the breakdown points are trivial you call it a “theory”. Kind of… it’s just a metaphor.

  25. PTM says:

    @Luysii #20: “We’ve all accepted the idea that the 3 nucleotide codon for amino acids is a type of code. However, if you look at the nucleotides individually you will see that the amino acid they code for depends on the adjacent nucleotides (e.g. is context dependent)…”
    The genetic code is not context sensitive, the unit of the code is a nucleotide triple, and each triple has the same meaning regardless of the context. The fact that meaning depends on the starting nucleotide doesn’t change this fact. The same holds for many codes which use fixed word lengths instead of delimiters to identify individual words (for example assembly languages).
    The meaning of the code is also unambiguous – once you know mRNA sequence and starting codon you can use the code to determine protein sequence which has very high probability of being correct (as always in biology there are exceptions but very rare). The code metaphor here works very well.
    Luysii: “The histone code (whose complexity is still being determined) tells us when chromatin is active or inactive, and what particular proteins bind to a given nucleosome affecting transcription, replication and DNA repair.”
    Many things tell us when chromatin is active: it’s conformation; presence of certain transcription factors, polymerases, splicing factors and newly synthesized RNA; nucleosome density; nucleosome exchange rate; and so on. Yet no one talks about conformation code, transcription factor code, polymerase code, nucleosome exchange code, etc. Why? Because code metaphor doesn’t offer any benefit here and would only muddle things.
    Same goes for which proteins bind a certain nucleosome, it depends on many other factors beyond it’s modifications – DNA sequence, modifications and conformation; presence/absence of transcription factors, other bound proteins and RNA; state of DNA damage response; phase of cell cycle; local ion and small molecule concentrations; and so on.
    For example a certain histone modification may be required for recruitment of a certain complex in some place, but the same complex may be present in other locations despite lack of this modification (for example recruited by specific transcription factors) and yet in other places the modification may be present but not the complex (say a binding surface may be obscured).
    Histone modifications are just one small aspect of the whole picture.
    Using metaphors only makes sense when those to whom they are addressed understand them and have mental pictures ready to be utilized. What everyone imagines when you invoke “code” is a simple rule for translating one kind of well defined information into another, also well defined. Complexity, context sensitivity, low reliability, lack of well defined concept being encoded, and the fact that influenced processes are also impacted by great many other factors all make “histone code” completely at odds with this simple picture and therefore invoking code metaphor will only mislead people.

  26. luysii says:

    PTM: Thanks for your response. I don’t think we’ll ever agree, be cause my definition of code is far broader than yours — ” What everyone imagines when you invoke the word code “is a simple rule for translating one kind of well defined information into another”. Well, this is not what I imagine code to mean.
    I was around when the amino acid code was being worked out, and it was far clear how things would turn out. The rules didn’t appear to be simple (assuming they existed at all). I think the histone code presently is in a similar state. Even if it turns out that we could only use the panoply of histone modifications to predict chromosome conformation, transcriptional activity/inactivity etc. etc. with just 80% accuracy it would still be valuable. I’d call it a code, and you would not.
    Would you still call the other uses of the 3 nucleotide code for amino acids a ‘code’. I would. For a bit more detail – see
    What would you call the following codon usage? [ Science vol. 315 pp. 466 – 467, 525 — 528 ’07 ] Infrequently used codons for a given amino acid (with multiple redundant codons) slow the rate of protein synthesis, because tRNAs for a given codon are present in the cell in direct proportion to codon usage (makes sense). So proteins with from genes containing infrequently used codons are made more slowly. Amazingly, this can result in different protein conformations in the product, despite identical amino acid composition and ordering.
    I must say that it’s a pleasure to have a disagreement with someone in a nonpejorative, non ad hominem fashion.

  27. Baltic says:

    A couple of years ago (2nd or 3rd year undergrad) I had this harmful (or misleading, to be more exact) mental construct for Knoevenagel condensation. I pictured it as the carbonyl being brutally rammed, oxygen first, into the activated methylene group, producing water molecule instantly from the sheer force of the impact. All the while I knew that this was not how the reaction proceeds, but I simply couldn’t help having this mental image.

  28. HelicalZz says:

    I went on a date with a girl named Simile once. I met her phor dinner : )

Comments are closed.