eduKateSG · How Science Works
Choose your reading route
See how chemical sensing, short-term adaptation and biased run durations can produce navigation without a map. Follow the mechanism, calculate a simple drift and test alternative explanations.
Follow the cue into behaviour · Chapters 1–3 · Start this route
Connect molecules with drift · Chapters 4–6 · Start this route
Check evidence and limits · Chapters 7–10 · Start this route
Full contents · Check your understanding · How Science Works master guide
Some bacteria move towards favourable chemical conditions by adjusting a partly random search. In the well-studied bacterium Escherichia coli, swimming runs are interrupted by tumbles that change direction. When an attractant becomes more abundant along the cell’s recent path, signalling can reduce the probability of tumbling, extending movement in that favourable direction.
The bacterium does not need to see a food source or calculate a route across a map. A local molecular response changes the statistics of its motion. Across many runs, that small bias can produce net movement up a chemical gradient, even though an individual path remains irregular.
This process is chemotaxis. We will use the classic E. coli run-and-tumble system as our main example, while keeping its limits visible. Other bacteria use different swimming patterns or different forms of movement, and some do not move actively at all. The article develops the behaviour-and-mechanism branch of Microbiology within How Science Works.
Contents
Follow the cue into behaviour · Chapters 1–3
1. A chemical landscape can carry useful information
2. Runs and tumbles create a search before any bias is added
Connect molecules with drift · Chapters 4–6
4. Receptors connect chemical binding to motor switching
5. Adaptation keeps a persistent signal from becoming the only signal
Check evidence and limits · Chapters 7–10
7. A convincing test compares more than the final cluster
8. Sensing has limits, and environments keep changing
9. Keep chemotaxis distinct from related biological processes
10. A complete explanation fits on one carefully labelled page
CHAPTER 1 OF 10 · Follow the cue into behaviour
1. A chemical landscape can carry useful information
A chemical gradient is a change in concentration across space. Near a source, a substance may be more concentrated than farther away. Diffusion, fluid flow, chemical reactions and consumption by organisms can continually reshape that pattern. The gradient is a physical feature of the environment; it is not an arrow painted for a bacterium to follow.
An attractant is a chemical that elicits movement towards higher concentration under the relevant conditions. It need not be something the organism can actually consume. Researchers can use non-metabolisable attractants to investigate sensing while reducing the complication of the cells eating the cue. “Towards food” is a useful everyday entry point, not the definition of every chemotactic response.
Likewise, a repellent can favour movement away from particular conditions. The effect depends on the organism, its receptors and its state. A chemical does not carry an intrinsic universal label saying “good for all bacteria”. Nor does increasing concentration always improve performance indefinitely; sensing has a working range.
For a concrete diagram, draw a pale-to-dark band representing increasing attractant concentration. Add several starting positions and irregular tracks. The useful prediction is a statistical difference between motion up and down the gradient, not a promise that every track points neatly towards the darkest patch.
The existing diffusion guide explains spreading more generally. Here the extra mechanism is active movement whose direction statistics change in response to chemical information. Chemical molecules spreading and bacterial cells navigating are related processes, but they are not the same motion.
CHAPTER 2 OF 10 · Follow the cue into behaviour
2. Runs and tumbles create a search before any bias is added
Swimming E. coli typically carries several flagella, long helical filaments driven by rotary motors. During a run, coordinated rotation can organise the filaments into a bundle that propels the cell. Changes in motor behaviour disrupt that arrangement and produce a tumble, after which swimming resumes in a changed direction.
A tumble is not normally a deliberate turn to face the source. It changes orientation with a distribution of possible outcomes. The new direction can still be unfavourable. Runs also do not form perfectly straight ruler lines: random rotational motion and the surrounding fluid affect orientation.
In a uniform environment, many runs and tumbles produce an exploratory trajectory without a persistent chemical preference for one direction. Add a gradient and the cell can change how long it tends to continue particular runs. A directional population pattern can emerge from this altered persistence even when the reorientation process remains partly random.
Berg and Brown’s three-dimensional tracking study examined the trajectories of individual E. coli cells and helped establish the run-and-tumble account of chemotaxis. Tracking matters because an endpoint picture of accumulated cells cannot reveal how their individual motion created that accumulation.
It is helpful to separate three observations: how fast a cell swims during a run, how frequently it changes direction, and its net displacement over a longer interval. A cell can swim quickly while ending close to where it began. Efficient movement towards a source depends on the pattern of movement, not simply the largest instantaneous speed.
CHAPTER 3 OF 10 · Follow the cue into behaviour
3. Comparing now with recently can reveal improvement
An E. coli cell is small, and comparing chemical conditions at different points along its body can be difficult in many ordinary gradients. Its classic chemotactic system instead uses temporal information: current receptor activity is interpreted in relation to recent activity. As the cell swims, changing conditions along its path become changes through time.
Suppose a cell swims into increasing attractant concentration. The signalling system can respond by making a run persist longer. If subsequent movement fails to bring continued improvement, the response changes. The mechanism does not require remembering a visual scene or representing a distant destination.
“Memory” in this context refers to a molecular state influenced by previous stimulation. That state affects the response to the current stimulus. A thermostat with a changing reference or a chemical feedback circuit can illustrate the idea better than a human recalling yesterday’s lunch, although neither analogy reproduces all the biology.
A powerful experiment separates time from space. Macnab and Koshland’s 1972 study of Salmonella typhimurium changed attractant concentration over time while keeping the environment spatially uniform. A motility response to that change supports temporal sensing because there is no simultaneous spatial gradient across the apparatus telling cells which direction to choose.
The conclusion has a boundary. A uniform temporal increase can alter swimming behaviour without creating a preferred spatial direction in an otherwise symmetric environment. A response to change is not sufficient for net directional migration; the relationship between the cell’s movement and the spatial chemical pattern supplies that direction in the usual gradient-following situation.
CHAPTER 4 OF 10 · Connect molecules with drift
4. Receptors connect chemical binding to motor switching
The cue is detected through molecular interactions. Chemoreceptors and associated signalling proteins regulate the activity of a kinase called CheA. In the standard E. coli pathway, attractant stimulation can lower CheA activity, reducing the level of phosphorylated CheY. CheY’s phosphorylation state influences its interaction with the flagellar motor’s switching machinery.
Higher CheY-P generally favours clockwise motor switching in this system, which is associated with increased tumbling. A favourable attractant change can therefore favour longer runs through a chain from receptor activity to kinase activity, response-regulator state and motor behaviour. The cell does not need a central command centre that reads a concentration number.
Structural and biochemical work on CheY links phosphorylation with changes at a signalling surface and interaction with the motor component FliM. This is a molecular connection between a chemical modification and a change in movement probability.
The pathway should not be treated as a string of perfectly deterministic switches. Individual binding events fluctuate, proteins occur in finite numbers, and different cells can have different operating states. A signalling change alters probabilities and biases rather than guaranteeing an identical motor response at an identical instant in every cell.
Also separate the control signal from the energy supply. Phosphorylation helps communicate state; the E. coli flagellar motor is powered by an ion gradient across the membrane. Sensing an attractant does not mean that the sensed molecule is directly poured into the motor as fuel. Distinguishing information flow from energy flow makes the mechanism easier to understand and prevents a misleading food-powered steering analogy.
CHAPTER 5 OF 10 · Connect molecules with drift
5. Adaptation keeps a persistent signal from becoming the only signal
If the system responded indefinitely to one increase, it could become poor at detecting the next change. Adaptation changes the signalling baseline during sustained stimulation, helping the cell remain responsive within an appropriate concentration range.
In the canonical E. coli receptor system, enzymes including CheR and CheB modify receptor methylation state. Their opposing activities contribute to feedback that adjusts receptor signalling. After an initial response to a maintained attractant change, activity can move back towards its earlier operating level even though the attractant concentration remains different.
Experiments on chemoreceptor modification directly examine reactions catalysed by CheR and CheB and how changes to receptor structure affect those reactions. The underlying lesson is specific: a biochemical state can retain information about stimulation history and alter subsequent sensitivity.
Adaptation is not identical to forgetting the world or becoming unable to detect the chemical. It changes the reference against which further changes are evaluated. Nor is perfect restoration guaranteed across every stimulus, concentration and bacterial species. Motor-level adaptation and other regulatory effects also contribute to the complete response; research on adaptation at the pathway output shows why receptor modification is not the only adjustable element.
For an original thought experiment, compare a staircase of concentration increases with one single increase held constant. A system that only reports absolute concentration predicts one kind of output; a system sensitive to recent change and capable of adaptation predicts another. Drawing both predicted traces before examining data makes the distinction testable rather than merely attaching the word “memory” to a moving cell.
CHAPTER 6 OF 10 · Connect molecules with drift
6. Worked model: a random direction can still produce drift
Original one-dimensional teaching model. Imagine a swimmer that chooses right or left with equal probability after each reorientation. Let right be up the attractant gradient. Give every rightward run a duration of 2 seconds and every leftward run a duration of 1 second, with swimming speed fixed at 20 micrometres per second.
A rightward run contributes +40 micrometres; a leftward run contributes −20 micrometres. Because the two directions are equally likely at the start of a run, the mean displacement per run is half of +40 plus half of −20, or +10 micrometres. The mean duration per run is 1.5 seconds.
Over many independent runs in this simplified model, the long-time drift speed is the mean displacement divided by mean duration: 10 ÷ 1.5, approximately 6.7 micrometres per second to the right. The average swimming speed during a run is still 20 micrometres per second. Net progress is slower because some travel cancels other travel.
If both directions instead last 1 second, the mean displacement is half of +20 plus half of −20, or zero. The direction choices have not changed between the two models. The bias entered through run duration.
The model deliberately ignores tumble time, three-dimensional angles, gradual turning during runs, concentration dependence, walls and fluid flow. Real runs have distributions of durations rather than these two fixed values. Consequently, 6.7 micrometres per second is not a prediction for a real E. coli population. It is a transparent demonstration that equal direction choices and unequal persistence can produce directional drift without foresight.
For a paper activity, toss a coin to choose direction and draw the corresponding 40- or 20-unit step. Repeat on a second sheet with equal step lengths. Individual short trials may disagree with the average prediction. Comparing many trials reveals why a statistical mechanism should be assessed statistically.
CHAPTER 7 OF 10 · Check evidence and limits
7. A convincing test compares more than the final cluster
Suppose more bacteria are observed near an attractant source after some time. Chemotaxis is one explanation, but cells might also grow faster there, become trapped, adhere differently or be transported by fluid flow. The count at the destination contains the combined consequences of movement, reproduction, loss and retention.
A stronger investigation links the chemical landscape to individual trajectories and an appropriate control. In a spatially uniform cue, there should be no cue-defined direction in an otherwise symmetric setting. With a maintained gradient, the analysis can compare runs according to whether the cell is moving up or down the gradient. A disrupted sensing pathway can test the role of the proposed molecular mechanism, provided ordinary motility and cell condition are checked too.
Three-dimensional tracking is valuable because a two-dimensional image can hide motion out of the focal plane. Near a surface, the surface itself can alter swimming. A 2021 three-dimensional chemotaxis study measured both individual behaviour and population drift, while demonstrating how surface effects can complicate common two-dimensional measurements.
The correct control depends on the alternative explanation. Checking for fluid drift addresses passive transport; measuring growth addresses reproduction; confirming that cells still swim addresses a general motor defect. One control cannot be expected to exclude every competing cause.
This is also why “the mutant did not accumulate” is an incomplete mechanistic conclusion. If the mutant cannot swim, the experiment has not isolated gradient sensing. If it swims normally but no longer changes its run statistics with the cue, the evidence is more directly related to the proposed sensory pathway. The distinction concerns the chain of causation, not merely the appearance of the endpoint.
CHAPTER 8 OF 10 · Check evidence and limits
8. Sensing has limits, and environments keep changing
Chemical molecules arrive at receptors through fluctuating encounters, and noise also arises inside the signalling system. A weak gradient can be difficult to distinguish from those fluctuations, while the cell has limited time to compare conditions before its orientation changes. The environment can also change while the cell samples it, so an increase detected during a run need not always be caused by movement towards a fixed source.
Imagine a cell in a chemical field that is rising everywhere because a nearby process releases attractant. A temporal increase alone cannot tell the cell whether its chosen direction was favourable. Conversely, a patch carried past by flowing water may change faster than the cell can exploit it. Successful navigation depends on the relationship among movement, sensing, adaptation and environmental timescales.
Research on information limits in E. coli chemotaxis connects measured signalling information with gradient-climbing performance. It provides a useful reminder that sensitivity is not infinite. A 2026 follow-up study found that internal signal-processing noise, rather than random molecule arrivals, limited chemosensing accuracy in the E. coli conditions tested. The existence of a physical noise limit does not establish that a cell operates at that limit. A biological mechanism can work effectively while its internal processing leaves some environmental information unused.
Different cells can also behave differently under nominally identical conditions. Differences in protein abundance, motor behaviour or physiological state can affect trajectories. An average drift is therefore a population summary, not the speed or path of every member.
For an evidence-reading exercise, imagine two groups with the same mean displacement but very different spreads of individual outcomes. One group moves moderately consistently; the other contains a small number of strong movers and many weak responders. The identical mean does not establish identical biology. Reporting trajectory distributions, numbers of independent experiments and conditions helps readers see what the average conceals.
CHAPTER 9 OF 10 · Check evidence and limits
9. Keep chemotaxis distinct from related biological processes
Chemotaxis describes movement biased by a chemical cue. Chemokinesis describes a change in movement activity, such as speed or turning, without necessarily establishing directional movement relative to a gradient. A cue can affect several aspects of motility, so a complete experiment should measure the feature that its claim names.
Quorum sensing is different again. It concerns responses to chemical signals associated with the surrounding microbial population and environment, often affecting gene regulation. The Microbiology guide introduces that subject. A chemotactic response need not involve a message sent by another bacterium, and detecting a chemical does not automatically make the process quorum sensing.
Chemotaxis also differs from evolutionary adaptation across generations. Receptor adaptation can happen within the behaviour of an individual cell. Evolutionary adaptation involves changes in populations through inherited variation and differential reproduction. The same word can point to different timescales and mechanisms, so the context must be stated.
The E. coli example is not the only navigation strategy. Some bacteria reverse their swimming direction; others combine reversals with different reorientation events, and surface-associated movement can use other machinery. Large cells can use spatial comparisons more directly. We should not turn one successful model organism into a rule that all living cells sense gradients in the same way.
The transferable idea is a relationship among sensing, internal state and movement. To use it for another organism, first identify its actual motor behaviour and signalling system. Then ask which observations demonstrate that cue information changes movement, rather than assuming the familiar run-and-tumble diagram applies unchanged.
CHAPTER 10 OF 10 · Check evidence and limits
10. A complete explanation fits on one carefully labelled page
Draw an irregular track through a left-to-right attractant gradient. Label a run that moves towards increasing concentration, a tumble that changes direction, and a run that moves away. Make favourable runs longer on average across the drawing, while retaining some unfavourable motion. Caption it “Conceptual statistical pattern, not a measured trajectory or a guaranteed route.”
Beside the track, draw a second chain: receptor response → signalling state → motor-switching probability → run statistics → population drift. Add a feedback arrow from receptor modification to receptor activity. This connects a microscopic interaction with a larger pattern without inserting intention into the explanation.
Does the bacterium know where the food is? The mechanism does not require such knowledge. Recent chemical conditions influence movement probabilities.
Must every tumble point towards the source? No. Direction changes can remain partly random while favourable runs last longer on average.
Does swimming faster prove better chemotaxis? No. Speed, reorientation and net drift must be distinguished.
Can a uniform increase change behaviour without producing directional migration? Yes. The temporal response can occur without a spatial direction to favour.
Does a cluster prove chemotaxis? No. Growth, trapping, surface effects and fluid motion are alternatives that suitable observations and controls must address.
For safe learning, use published videos, supplied trajectory data or the paper-step model. Do not collect or culture unknown microorganisms to reproduce the effect. The central scientific skill is already available without a biological experiment: distinguish the chemical cue, the internal response, the resulting motion and the evidence linking them.
Return to Microbiology for the wider microbial context, or continue through How Science Works.
Sources and example boundaries
The linked original studies support three-dimensional trajectories, temporal-gradient sensing, molecular signalling, adaptation and limits on navigation. The one-dimensional calculation, paper activity, hypothetical comparisons and diagram are original teaching models. Their numerical choices are illustrative and should not be treated as measured bacterial speeds, timings or experimental specifications.