Identifying Errors and Limitations

No scientific investigation is perfect.

Even when an experiment is carefully planned and conducted, measurements contain some uncertainty, equipment has limitations, conditions may vary, and experimental methods may introduce errors.

Recognizing these problems does not mean an experiment was unsuccessful. In fact, identifying errors and limitations is an important part of good scientific work.

Scientists ask:

• How trustworthy are these measurements?
• How much variation is present?
• Could the method have shifted all the results?
• Were important variables controlled?
• Is there enough evidence to support the conclusion?
• How could the investigation be improved?

Evaluating these questions helps determine the quality of the experimental evidence.

Error Does Not Always Mean a Mistake

In everyday language, an error often means that someone did something wrong.

In experimental science, error has a broader meaning.

Experimental error describes the difference or uncertainty between a measured value and the value we are trying to determine.

For example, suppose you measure the length of an object three times:

15.2 cm

15.3 cm

15.1 cm

The slightly different measurements do not necessarily mean that someone made a mistake.

Small differences can occur because of:

• instrument resolution
• difficulty judging the exact end of the object
• viewing angle
• small changes in technique

Experimental science therefore involves understanding and managing measurement uncertainty.

---

Random Errors

Random errors cause measurements to vary unpredictably from one measurement to another.

Sometimes the measured value may be slightly too high. Other times it may be slightly too low.

For example, imagine timing a toy car using a handheld stopwatch.

The true travel time might be close to 2.50 s.

Repeated measurements could be:

2.46 s

2.54 s

2.49 s

2.57 s

2.45 s

The values vary around a central value.

This variation could be caused by small differences in:

• starting the stopwatch
• stopping the stopwatch
• releasing the car
• positioning the car
• environmental conditions

These are examples of random variation.

How Random Error Affects Results

Random errors make measurements more spread out.

If random error is large, repeated measurements may be quite different.

For example:

Set A

5.1 s
5.0 s
5.2 s
5.1 s
5.0 s

Set B

4.2 s
6.0 s
5.7 s
4.5 s
5.3 s

Set B has much greater variation.

This makes it harder to determine the underlying value confidently.

Random error therefore mainly affects the precision and consistency of measurements.

---

Reducing Random Error

Random error usually cannot be completely eliminated, but its effect can often be reduced.

Useful approaches include:

• repeat measurements
• calculate an average
• increase sample size
• improve measurement technique
• use equipment with finer resolution where appropriate
• use automated measurement equipment
• control environmental conditions more carefully

For example, instead of measuring a pendulum's period for one oscillation, you could measure the time for 10 oscillations.

If 10 oscillations take:

18.4 s

then:

Period = 18.4 ÷ 10

Period = 1.84 s

This reduces the relative effect of human reaction time compared with trying to time a single oscillation.

---

Repeated Measurements

Repeated measurements are particularly useful for identifying random variation.

Suppose you measure a reaction time three times:

42.1 s

41.8 s

42.4 s

The values are reasonably close.

Average = (42.1 + 41.8 + 42.4) ÷ 3

Average = 42.1 s

The average provides a useful estimate based on several measurements rather than relying on a single trial.

However, simply repeating measurements does not solve every type of experimental problem.

This becomes important when we consider systematic errors.

---

Systematic Errors

A systematic error shifts measurements consistently in the same direction.

Instead of producing unpredictable scatter, a systematic error may cause results to be consistently:

• too high
• too low

For example, suppose a balance incorrectly reads:

+2.0 g

even when nothing is placed on it.

An object with a true mass of approximately:

50.0 g

might be measured as:

52.0 g

If another object has a true mass of:

75.0 g

the balance might show:

77.0 g

The measurements may be very consistent, but they are systematically too high.

Repeating Does Not Automatically Fix Systematic Error

This distinction is extremely important.

Suppose a faulty balance gives:

52.0 g

52.1 g

52.0 g

51.9 g

52.0 g

The measurements are very consistent.

Their average is approximately:

52.0 g

But if the true mass is 50.0 g, the result is still wrong.

Repeating the measurement did not remove the systematic error.

That is because the instrument introduced approximately the same bias every time.

To address this problem, you might need to:

• zero the balance
• calibrate the balance
• replace faulty equipment
• apply an appropriate correction when scientifically justified

---

Random vs Systematic Error

Understanding this difference is essential when evaluating experimental evidence.

---

Accuracy and Precision

Accuracy describes how close a measurement is to the true or accepted value.

Precision describes how closely repeated measurements agree with one another.

Imagine measuring an object with a true mass of:

50.0 g

Measurements A

50.0 g
50.1 g
49.9 g

These measurements are both reasonably accurate and precise.

Measurements B

55.0 g
55.1 g
54.9 g

These measurements are precise but inaccurate.

They are tightly grouped, but they are centred around the wrong value.

A systematic error could produce this pattern.

Measurements C

48.2 g
52.0 g
49.3 g
51.4 g

These measurements show relatively poor precision because they are spread out.

Random error could contribute to this pattern.

Instrument Resolution

Every measuring instrument has a limited resolution.

Resolution is the smallest change that an instrument can detect or display.

For example, one ruler may have markings every:

1 mm

while another measuring device may measure to:

0.01 mm

The second instrument has finer resolution.

Similarly, a digital balance might display mass to:

0.1 g

while a more sensitive balance might display:

0.001 g

Using equipment with suitable resolution can improve the quality of measurements.

However, greater resolution does not automatically guarantee greater accuracy. An instrument can display many decimal places and still be poorly calibrated.

---

Measurement Uncertainty

Every measurement has some uncertainty.

Suppose a ruler is marked every millimetre.

A reported measurement of:

12.4 cm

should not be interpreted as infinitely exact.

The true value lies within some reasonable range determined by:

• instrument resolution
• measurement technique
• experimental conditions

This is why scientists should avoid reporting unrealistic numbers of decimal places.

For example, a simple classroom ruler cannot reasonably justify:

12.438271 cm

The number of digits reported should reflect the quality of the measurement.

---

Parallax Error

A common measurement problem is parallax error.

Parallax occurs when a scale is viewed from the wrong angle.

Suppose you are reading the volume of liquid in a measuring cylinder.

Looking from above or below can make the liquid level appear to align with a different scale marking.

The measurement should normally be viewed at the appropriate eye level.

Parallax can also occur when reading:

• rulers
• analogue meters
• thermometers
• measuring cylinders
• other graduated scales

A better experimental technique can reduce this source of error.

---

Zero Error

A zero error occurs when an instrument does not read zero when the quantity being measured is actually zero.

For example, an empty balance might display:

0.7 g

If this is not corrected, all subsequent mass measurements could be too high by approximately 0.7 g.

Before collecting measurements, check whether appropriate equipment is:

• zeroed
• calibrated
• functioning correctly

A zero error can create a systematic error.

---

Human Reaction Time

Experiments involving handheld stopwatches often include uncertainty caused by human reaction time.

A person does not press the stopwatch button at exactly the same instant as an event occurs.

This can produce random variation.

Suppose a pendulum completes one oscillation in approximately 1.5 s.

A reaction-time uncertainty of even a fraction of a second could be significant compared with the total measurement.

Instead, you could measure:

20 oscillations

and divide the total time by 20.

Even better, where available, automated equipment such as:

• light gates
• motion sensors
• electronic timers

can reduce the influence of human reaction time.

---

Environmental Conditions

The surrounding environment can affect experimental results.

Possible factors include:

• temperature
• humidity
• air movement
• light intensity
• vibrations
• atmospheric pressure
• background noise

For example, if you investigate evaporation but conduct different trials at very different room temperatures, temperature may become an uncontrolled variable.

The experiment might then measure the effects of both:

your independent variable + changing environmental conditions

This makes the evidence more difficult to interpret.

---

Biological Variation

Investigations involving living organisms often contain considerable natural variation.

Two plants of the same species may grow differently even when they receive similar treatment.

Variation can arise from differences in:

• genetics
• initial size
• health
• age
• nutrient availability
• environmental history

This is why biological experiments often benefit from larger sample sizes.

For example, using one plant for each treatment gives weak evidence.

Using several plants in each treatment allows natural variation to be considered more effectively.

---

Errors and Limitations Are Not the Same Thing

These ideas are related, but they are not identical.

An error affects measurements or results.

A limitation is a feature of the experimental design that restricts what can confidently be concluded.

For example:

Error: Stopwatch reaction time introduces variation into travel-time measurements.

Limitation: Only three ramp heights were tested.

The first affects individual measurements.

The second limits how completely the relationship between ramp height and travel time can be understood.

---

Common Experimental Limitations

Limitations may include:

• small sample size
• too few repeated trials
• narrow range of independent-variable values
• short investigation duration
• limited instrument precision
• uncontrolled variables
• limited equipment
• inability to measure something directly
• natural variation
• method that does not closely represent real-world conditions

A good evaluation identifies specific limitations that actually matter to the investigation.

---

Small Sample Size

Suppose you investigate:

How does salt concentration affect seed germination?

If you use only one seed at each salt concentration, one unusual seed could strongly influence your conclusion.

A better design might use:

10, 20, or more seeds per condition, depending on practical constraints.

Increasing sample size can reduce the influence of unusual individual cases and provide stronger evidence about the overall pattern.

---

Limited Range

Suppose you investigate temperature but test only:

20 °C and 25 °C

You may discover a difference, but your evidence tells you very little about what happens outside this narrow range.

Testing several suitable temperatures might provide a clearer picture:

10 °C, 20 °C, 30 °C, 40 °C and 50 °C

provided these values are appropriate and safe for the investigation.

A wider, scientifically sensible range can reveal relationships that two values might miss.

---

Too Few Data Points

Imagine plotting a graph using only two measurements.

Two points can always be connected by a straight line.

That does not necessarily mean the actual relationship between the variables is linear.

More independent-variable values can reveal:

• curves
• thresholds
• maximum values
• minimum values
• plateaus
• unexpected changes

Therefore, the number and distribution of data points affect the strength of the evidence.

---

Uncontrolled Variables

Suppose you investigate:

How does fertilizer concentration affect plant growth?

You carefully change fertilizer concentration, but the plants receive different amounts of sunlight.

Now plant growth may be affected by both:

• fertilizer concentration
• light availability

Light has become an uncontrolled variable.

This weakens your ability to conclude that fertilizer alone caused the observed differences.

---

Identify the Effect of the Error

Simply naming an error is not enough.

A strong evaluation explains how the error could influence the results.

Weak:

"There was human error."

Better:

"Human reaction time affected the stopwatch measurements."

Stronger:

"Human reaction time could cause measured travel times to be slightly too high or too low, increasing the spread between repeated trials and reducing precision."

The stronger explanation connects:

source of error → effect on measurement → effect on evidence

---

Avoid Saying "Human Error"

The phrase human error is usually too vague to be useful.

Instead, identify the specific problem.

Instead of:

Human error

write:

Difficulty judging exactly when the car crossed the finish line caused variation in stopwatch stopping times.

Instead of:

Measurement error

write:

The measuring cylinder had 5 mL divisions, limiting the precision of the volume measurements.

Specific explanations lead to specific improvements.

---

Suggest Improvements That Match the Problem

A good improvement directly addresses an identified error or limitation.

Problem

Human reaction time when using a stopwatch.

Improvement

Use electronic light gates.

---

Problem

Only one plant was used per treatment.

Improvement

Increase the sample size by using several plants in each treatment.

---

Problem

A balance has a zero error.

Improvement

Zero or calibrate the balance before taking measurements.

---

Problem

Water temperature changes during each trial.

Improvement

Use a thermostatically controlled water bath.

---

Problem

Only two concentrations were tested.

Improvement

Test several concentrations across a suitable range.

The relationship should be:

Problem → Effect → Improvement

---

"Be More Careful" Is Usually Not Enough

Suppose the problem is:

The stopwatch measurements varied because of reaction time.

Saying:

"Next time I would be more careful."

does not really solve the problem.

A stronger improvement would be:

"Use light gates connected to an electronic timer to remove most of the reaction-time uncertainty."

Improvements should be:

• specific
• practical
• connected to the problem
• likely to improve the quality of the evidence

---

Evaluating Experimental Evidence

After identifying errors and limitations, ask:

How strong is the evidence?

Consider several factors.

Quantity of data

Were enough measurements collected?

Repetition

Were trials repeated?

Consistency

Were repeated results reasonably similar?

Range

Was an appropriate range of values tested?

Control of variables

Were important variables controlled?

Measurement quality

Was suitable equipment used?

Anomalies

Were unusual results investigated?

Sample size

Was the sample large enough for the type of investigation?

Method

Did the method actually test the research question?

Conclusion

Does the conclusion match the evidence?

Evidence should be evaluated as a whole rather than judged by one feature alone.

---

Reliable Evidence

Evidence becomes more convincing when repeated measurements produce reasonably consistent results.

For example:

Trial 1 = 12.4 s

Trial 2 = 12.5 s

Trial 3 = 12.3 s

These measurements show relatively little variation.

Compare:

Trial 1 = 8.2 s

Trial 2 = 17.9 s

Trial 3 = 11.5 s

The second set raises questions.

You might investigate:

• whether the method was consistent
• whether variables were controlled
• whether the equipment was appropriate
• whether more trials are needed

---

Validity

An investigation has good validity when its method appropriately tests the question it is intended to answer.

Suppose your question is:

How does water temperature affect dissolving time?

But as temperature increases, you also stir faster.

The experiment cannot clearly separate the effect of:

temperature

from the effect of:

stirring

The validity of the investigation is weakened.

Good control of variables helps improve validity.

---

Reliability

Reliability concerns whether the evidence is reasonably consistent and repeatable.

Reliability can often be improved by:

• repeating trials
• increasing sample size
• standardizing procedures
• using consistent equipment
• calculating averages where appropriate

If repeated measurements vary enormously, confidence in the result decreases.

---

Accuracy

Accuracy concerns how close measurements are to the true or accepted value.

Accuracy can be affected by:

• poor calibration
• zero error
• incorrect measuring technique
• systematic errors
• unsuitable equipment

An experiment can be highly repeatable but still inaccurate.

This is why repeated measurements alone do not guarantee high-quality evidence.

---

Precision

Precision concerns the closeness of repeated measurements and the level of detail available from the measuring method.

Precision may be improved by:

• using higher-resolution equipment
• reducing random variation
• standardizing the procedure
• using automated measurement systems where appropriate

Again:

Precise does not automatically mean accurate.

---

Anomalous Results

An anomalous result or outlier differs noticeably from the general pattern.

Suppose repeated times are:

4.1 s

4.0 s

4.2 s

8.9 s

The 8.9 s measurement deserves investigation.

Ask:

• Was the procedure followed correctly?
• Was the measurement recorded correctly?
• Did something unusual happen?
• Does repeating the trial produce a similar result?
• Could this represent genuine variation?

Do not automatically delete an anomaly.

It may indicate:

• measurement error
• uncontrolled variables
• equipment problems
• natural variation
• an unexpected scientific effect

---

Error Bars and Variation

In more advanced scientific graphs, error bars may be used to show the variation or uncertainty associated with measurements.

For example, if several measurements are taken at each temperature, the graph may show:

• the average value
• a measure of spread or uncertainty

This allows the reader to see not only the average result but also how much the measurements varied.

The exact type of error bar used should always be stated because different error bars can represent different quantities.

---

Example Evaluation

Consider an investigation:

How does ramp height affect the average speed of a toy car?

The student used:

• one toy car
• five ramp heights
• three trials per height
• a 1.0 m travel distance
• a handheld stopwatch

Random error

Human reaction time caused small differences between repeated time measurements.

Effect: Increased variation and reduced precision.

Improvement: Use electronic light gates.

Possible systematic error

The stopwatch itself could theoretically run consistently fast or slow if faulty.

Effect: All measured times could be shifted in the same direction.

Improvement: Check the timer against a known reference or use calibrated timing equipment.

Limitation

Only five ramp heights were investigated.

Effect: The experiment provides limited information about what happens between or beyond those heights.

Improvement: Test additional heights across a suitable range.

Strength

Three repeated measurements were collected at each height.

Effect: This allowed consistency to be checked and average values to be calculated.

This is a much stronger evaluation than simply writing:

"There might have been some human error."

---

Evaluating the Conclusion

Finally, ask whether the conclusion is justified by the evidence.

Suppose your data show that bean seed germination decreased between 0% and 2% salt concentration.

A reasonable conclusion might be:

"Increasing salt concentration from 0% to 2% reduced bean seed germination under the conditions tested."

An unreasonable conclusion would be:

"Salt kills all plants."

The second statement goes far beyond the evidence.

A strong conclusion should be:

• supported by the data
• specific
• scientifically reasonable
• limited to what was actually investigated

---

Common Mistakes

Saying "human error"

Identify the specific action or measurement problem instead.

Assuming every unusual result is a mistake

Investigate anomalies before deciding what they mean.

Believing repeated trials remove all errors

Repetition helps with random variation but does not automatically remove systematic error.

Assuming precise measurements must be accurate

Measurements can be closely grouped around the wrong value.

Listing errors without explaining their effects

Explain how each problem influences the measurements or conclusion.

Suggesting "be more careful"

Give a specific improvement.

Claiming an experiment has no limitations

Every investigation has limitations.

Confusing a limitation with a failure

A limitation defines what the evidence can reasonably tell you.

Reporting unrealistic precision

The number of digits reported should reflect the measuring equipment.

Assuming more data automatically mean better evidence

Poorly collected data do not become strong evidence simply because there are many measurements.

---

Key Terms

Experimental error — Variation or difference between a measured value and the value being determined.

Random error — Unpredictable variation causing repeated measurements to differ.

Systematic error — An error that consistently shifts measurements in a particular direction.

Uncertainty — The degree of doubt associated with a measurement.

Limitation — A feature of an investigation that restricts the strength or scope of its conclusions.

Accuracy — How close a measurement is to the true or accepted value.

Precision — How closely repeated measurements agree or the level of detail available in measurement.

Reliability — The consistency or repeatability of experimental evidence.

Validity — The extent to which an investigation appropriately tests the intended question.

Resolution — The smallest change that a measuring instrument can detect or display.

Calibration — Comparing or adjusting an instrument using a known standard.

Zero error — A systematic measurement error caused when an instrument does not correctly read zero.

Parallax error — A reading error caused by viewing a scale from an incorrect angle.

Reaction time — The delay between observing an event and physically responding to it.

Repeat measurement — A measurement taken again under the same conditions.

Sample size — The number of observations, organisms, objects, or measurements included in an investigation.

Range — The span of independent-variable values investigated.

Outlier — A data value that differs substantially from most other values.

Anomalous result — A result that does not fit the general pattern.

Error bar — A graphical representation of variation or uncertainty in data.

Bias — A consistent influence that shifts measurements or conclusions in a particular direction.

Experimental evidence — Measurements and observations collected during an investigation.

Evaluation — An assessment of the strengths, weaknesses, errors, limitations, and quality of an investigation.

---

Key Takeaways

• No experimental measurement is perfectly exact.
• Experimental error does not necessarily mean someone made a mistake.
• Random errors cause unpredictable variation between measurements.
• Random errors often reduce precision.
• Repeated trials and averages can reduce the influence of random variation.
• Systematic errors shift measurements consistently.
• Systematic errors often reduce accuracy.
• Repeating measurements does not automatically remove systematic error.
• Calibration can help identify or reduce some systematic errors.
• Accuracy and precision are different.
• Measurements can be precise but inaccurate.
• Instrument resolution limits the detail that can reasonably be measured.
• Parallax occurs when a scale is viewed from an inappropriate angle.
• Human reaction time can introduce measurement uncertainty.
• Environmental conditions can become uncontrolled variables.
• Biological investigations often require larger samples because organisms naturally vary.
• Errors and limitations are related but not identical.
• An error affects measurements or results.
• A limitation restricts what can confidently be concluded.
• Avoid vague statements such as "human error."
• Identify the specific source of error.
• Explain how the error affects the results.
• Suggest an improvement that directly addresses the identified problem.
• Do not automatically discard anomalous results.
• Evaluate evidence using its quantity, consistency, range, measurement quality, control of variables, sample size, and experimental design.
• Strong conclusions remain within the limits of the evidence.

A strong experimental evaluation follows a useful pattern:

Problem → Effect on evidence → Specific improvement

And a strong scientist does not ask:

"Was my experiment perfect?"

Instead, they ask:

"How strong is my evidence, what limits it, and how could I make it stronger?"