PSLE Science Reality Lab Vol No.450 | “LOD90 = 2 ng” — Is Everything Above 2 ng Always Detected?

PSLE-SCI-REALITY-0450

Wait, What? A detection limit can describe probability, not a magic cliff.

A detector specification says: LOD90 = 2 ng. A student draws a line at 2 ng and says, “Anything above this line will always be detected. Anything below it can never be detected.”

That neat picture is tempting. It is also too strong.

Some scientific detection limits are defined using repeated measurements and probabilities. In the NIST material for an ASTM detection method, LOD90 is tied to a high probability of true detection under specified conditions, together with requirements about non-detection of a realistic blank. It is not described as a universal wall separating “always detected” from “never detected”.

This is exactly the kind of evidence transfer PSLE Science trains: interpret information, evaluate observations and methods, consider uncertainty and communicate a conclusion that does not outrun the evidence.

Quick Answer

No. A statement such as “LOD90 = 2 ng” does not mean every amount above 2 ng must be detected and every amount below 2 ng must be missed. LOD90 is a performance statement built from repeated detection behaviour under a defined method and set of conditions. To judge the claim, ask what was tested, how often the target was detected, how blanks behaved, what confidence statement is attached, and whether the real-world use matches the tested conditions.

Reality Lab habit: A threshold printed as one number can still describe a probability.

Owned Learner Job

This guide owns one real-world job: how to read a probabilistic detector threshold without turning it into a guaranteed yes/no boundary.

It does not replace existing owners for generic “detected” wording, measurement limits, false positives, repeated trials, uncertainty, sampling or graph reading. Those skills are routed into this case. The communication object here is the detector-performance label itself: LOD90.

The Original Reality Lab Case: Detector P

Imagine a fictional laboratory detector used to look for a harmless marker substance called P. Its product sheet states:

Detection performance: LOD90 = 2 ng under Method M.

A school science club tests prepared samples. The values below are original teaching data, not a copied laboratory protocol.

Amount of P	Detections in 20 repeated trials
0 ng blank	1 positive signal
1 ng	11 detections
2 ng	19 detections
3 ng	20 detections

A learner points to 2 ng and says, “See? Ninety-five per cent were detected, so 2 ng means guaranteed detection.”

The data themselves show why that statement fails: one of the twenty 2 ng trials was not detected. Meanwhile one blank produced a positive signal. Real detection systems can have variation. The scientific question is about the pattern of performance across repeated trials, not only one dramatic success or failure.

Observed, Claimed and Inferred

Layer	Example
Observed	In repeated trials, the detector returned positives and negatives at several prepared amounts.
Performance claim	The stated LOD90 summarises detector performance under the defined method and conditions.
Unsupported inference	Every real sample above 2 ng will always be detected.
Another unsupported inference	Every sample below 2 ng is definitely absent if the detector reports no detection.

Why One Number Can Hide a Curve

Imagine increasing the amount of a target little by little. At very low amounts, detection may be uncommon. As the amount rises, detection may become more likely. At sufficiently high amounts, the detector may respond in nearly every trial under the tested conditions.

If you graphed amount on the horizontal axis and fraction of trials detected on the vertical axis, you would often think in terms of a rising performance curve rather than a perfect vertical wall.

The reported LOD90 compresses part of that performance information into a useful threshold statement. The compression is convenient. It is also why a careful reader asks how the threshold was defined.

The “90” Does Not Mean 90% Accuracy

This is an important representation check. The number 90 in LOD90 is not a general accuracy grade. It does not say that every concentration measurement is “90% correct”. It belongs to a particular detection-performance definition.

That distinction is similar to many Reality Lab problems: the same symbol “%” can describe very different things. A percentage might describe a fraction of trials, a fraction of samples, a confidence statement, a concentration, a relative difference or an efficiency. Always identify the denominator and the scientific job before interpreting it.

The Blank Matters Too

A detector is not useful simply because it says “detected” frequently. Imagine a detector that reports a positive result on every sample, including every blank. Its “detection rate” for target-containing samples would look impressive, but the device would not be distinguishing target from no-target conditions well.

NIST’s LOD90 description therefore includes performance of a realistic process blank as part of the definition. For a Primary learner, the transferable idea is straightforward: good detection evidence needs both success when the target is present and appropriate behaviour when it is absent.

Comparison Check: Same Threshold, Different Conditions

Suppose two advertisements both say “LOD90 = 2 ng”. Are the instruments automatically equivalent?

Not unless the performance statements are genuinely comparable. Check the target substance, preparation, test method, environment, detector settings, sample type and other defined conditions. A threshold measured with a clean prepared sample cannot automatically be carried to a messy real-world sample if the method behaves differently there.

Detector A	Detector B
LOD90 = 2 ng in clean prepared samples	LOD90 = 2 ng in a different sample matrix
20°C laboratory test	Outdoor field test across varying temperatures
Method A	Method B

The headline number matches. The evidence objects may not.

Method and Variable Check

When evaluating a detection-limit claim, ask which variables could change performance:

the amount of target presented to the detector;
the material or mixture containing the target;
instrument condition and settings;
environmental conditions allowed by the method;
sample preparation;
interfering substances;
the rule used to decide whether a response counts as a detection;
the number and independence of repeated trials.

You do not need to memorise every laboratory detail. The learner habit is to notice that a compact performance number came from a method. If the use changes beyond the method, the claim may need new evidence.

Worked Case 1: One Miss Above the Reported Threshold

A detector has a reported probabilistic detection threshold. In a classroom demonstration, a prepared amount slightly above that threshold gives one negative result.

A student says, “The specification must be false.”

That conclusion is too fast. If the specification describes probability across repeated trials, a single miss may still be compatible with the stated performance. The right response is to inspect the definition, repeat the test appropriately and check whether the demonstration matched the specified conditions.

Worked Case 2: A Detection Below the Threshold

A sample below the reported LOD90 produces a positive detection. Another student says, “Impossible. Below the limit means the detector cannot see it.”

Again, a probabilistic threshold is not a statement of impossibility. Some lower amounts may sometimes produce detectable responses. The important issue is whether performance is reliable enough, according to the stated definition, for the intended use.

Worked Case 3: The Blank That Triggered

A detector gives a positive response to one blank sample. Does that prove the target was present?

Not by itself. A false positive, contamination, handling issue or another cause may be possible. A strong evaluation separates the detector response from the claim about what physically caused it. The method’s blank performance and follow-up evidence matter.

Worked Case 4: The Viral Demonstration

A video shows a detector successfully identifying a target in three demonstrations. The caption says, “Proven to detect everything above 2 ng.”

Three successful demonstrations show that detection occurred in those three cases. They do not establish the full probability-of-detection curve, blank behaviour, confidence statement or range of operating conditions. A demonstration can show a phenomenon without testing the general performance claim.

Alternative Explanations for a Missed Detection

the target amount was close to a region where detection is not certain;
the sample differed from the conditions used to establish the specification;
sample preparation or delivery varied;
the instrument response varied;
another substance interfered;
the detector was outside its intended operating condition;
the target was not actually present at the assumed amount.

These are possible explanations to investigate, not excuses to save a favoured claim. Science compares them against evidence.

What Evidence Strengthens the Claim?

a clear definition of the stated detection limit;
repeated trials across relevant target amounts;
blank trials that measure false-positive behaviour;
test conditions that match the intended use;
enough independent observations to estimate performance rather than showcase one success;
reporting of uncertainty or confidence associated with the threshold;
evidence that the method remains suitable in the relevant sample type;
conclusions phrased as performance probabilities rather than universal guarantees.

What Would Weaken It?

only one dramatic detection is shown;
no blank or negative condition is reported;
the specification is quoted without its method;
field use differs greatly from test conditions;
the word “limit” is treated as a perfect physical boundary;
the threshold is turned into a general accuracy percentage;
the conclusion ignores misses, false alarms or variability.

How Far Can the Conclusion Travel?

A careful conclusion might say: “Under the stated method and conditions, the LOD90 describes a level associated with a specified high probability of true detection, together with blank-performance requirements.”

It should not become: “All samples above this amount are always detected,” “all samples below it contain nothing,” or “the detector is 90% accurate at measuring quantity.” Each of those sentences changes the scientific job.

Tempting Reasoning That Fails

“Limit means hard wall.” Check the definition; a reported threshold can be probabilistic.
“90 means 90% accurate.” The percentage belongs to detection performance, not a universal accuracy score.
“One successful test proves the threshold.” Probabilistic performance needs repeated observations.
“No detection proves absence.” A negative result must be interpreted using the detector’s performance and conditions.
“Above the threshold means no uncertainty.” A performance threshold does not erase measurement and method uncertainty.

PSLE-Style Transfer Case

A fictional sensor has a stated performance level at 5 units. During 20 repeated tests at 5 units, it detects the target in 19 tests. A student concludes, “The one missed test proves the target was absent in that sample.”

Explain why the conclusion is not supported.

Explained answer: The detector can sometimes miss a target even when it is present. A negative detector response is an observation about the instrument response, not direct proof that the target was absent. The sensor’s detection performance and test conditions must be considered.

Practice: Threshold or Guarantee?

A specification reports a probability-based detection limit. Can one number alone guarantee every future trial?
A sample below the reported level gives a positive response. Is that automatically impossible?
Why are blank trials useful when evaluating a detector?
Why should you compare the test conditions with the real-world use?

Answers: (1) No; probability-based performance describes a pattern across trials under stated conditions. (2) No; lower amounts can sometimes be detected. (3) Blanks help show how often the detector responds when the target should be absent. (4) Performance can change when sample or operating conditions change.

Delayed Independent Return

Later, whenever you see a scientific “limit”, ask three questions before interpreting it: What exactly is being limited? How was the limit defined? Is it a certainty, a probability, a reporting rule, a specification or something else?

The word limit is not enough. The definition carries the evidence.

Where to Route Next

Parent and Tutor Teaching Guide

Use a simple bag experiment rather than a technical detector. Put coloured counters in an opaque bag and define a fictional “detector” rule in which the learner draws one counter from a mixture. Change the fraction of target-colour counters and repeat many times. The chance of drawing the target changes gradually; there is no magical point where chance instantly becomes certainty.

Then return to the scientific label. Explain that real detector standards are more sophisticated, but the transferable lesson is the same: repeated performance can be summarised by a threshold without becoming a universal guarantee.

Keep the teaching focus on evidence reasoning, not on specialist laboratory mathematics. The child should leave able to say, “I need the definition and the test conditions before I turn this threshold into a claim.”

Authoritative Sources

The Quiet Return

A single number printed beside a detector can look like a boundary drawn with a ruler. Scientific performance is often less mechanical than that.

Ask what the number was built from. Ask how often detection occurred. Ask what happened with blanks. Ask whether the conditions match.

Then let the conclusion be as precise as the evidence.

Small Group Tutorials