91921 · Science investigations

• Five ways to investigate, one toolkit

  Fair testing changes one variable and measures the effect. Pattern seeking looks for a relationship across natural variation you cannot control. Exploring and observing records what happens in a system over time without changing it. Classifying and identifying sorts things into groups using stated criteria, or names them with a key. Modelling builds a simplified version of a system to explain or predict how it behaves. Most real investigations lean on one approach but borrow from others - a fair test still needs careful observation, and a pattern-seeking study still needs a consistent way of classifying what is counted.

• Choosing the approach the question allows

  The approach is decided by two things: what the question asks, and what you are actually able to control. "Does mulch depth change how fast soil dries out?" is a fair test, because you can set the mulch depth yourself. "Are tūī more common where kōwhai is flowering?" cannot be a fair test - you cannot make trees flower on command - so it becomes pattern seeking. Justifying a choice means saying why the approach fits, and why the obvious alternative does not: it is the difference between naming a method and defending one.

• Fair testing: change one thing at a time

  In a fair test you deliberately change the independent variable, measure the dependent variable, and keep everything else the same so the change you see can be blamed on the one thing you changed. If two things change at once - say, you use deeper mulch and move that pot into the sun - the result is uninterpretable, because either change could have caused it. A control set-up (the version with no treatment, such as bare soil) gives you a baseline to compare against, so you can tell an effect from what would have happened anyway.

• Pattern seeking: relationships you cannot control

  Some things cannot be controlled, or should not be: weather, seasons, whole ecosystems, people's health. Pattern seeking takes many measurements across the natural variation that already exists and asks whether two things vary together. It needs a big enough sample and a consistent method - counting tūī for exactly ten minutes at the same time of day at every site, for instance. Because nothing was controlled, a pattern-seeking study can show that two things are related, but it cannot by itself prove that one causes the other.

• Exploring and observing: disciplined noticing

  Exploring and observing is used when you do not yet know enough to ask a sharp question, or when the point is to record what a system actually does. It is not casual looking. Good observation is systematic: fixed sites or quadrats, set times, an agreed recording sheet, measurements and drawings or photographs with a scale, and the conditions (tide, weather, season) recorded alongside. It produces qualitative and quantitative records that often generate the question the next investigation tests.

• Classifying and identifying: criteria someone else could use

  Classifying sorts things into groups; identifying names a particular specimen, usually with a dichotomous key that offers a choice of two features at each step. Both stand or fall on the criteria. "Big leaves" is useless - one person's big is another's medium - while "leaf longer than 60 mm" can be applied by anyone and gives the same answer. Grouping stream invertebrates by their tolerance of pollution, for example, only means something if every sorter uses the same features to decide what is a mayfly larva and what is not.

• Modelling: a deliberate simplification

  A model stands in for a system that is too large, too slow, too fast or too dangerous to investigate directly. Models can be physical (a stream table of sand and flowing water), mathematical (an equation for how a population grows), computer-based (a weather forecast), or an analogy (water in pipes standing in for current in a circuit). Every model leaves something out on purpose, and that is its strength and its limit: it makes the important relationship visible, but its predictions are only as good as the parts of the real system it kept.

• Variables, range and interval

  The independent variable is what you change, the dependent variable is what you measure, and controlled variables are everything you hold constant. Two planning decisions follow: the range (how far apart your lowest and highest values are) and the interval (the size of the steps between them). Testing mulch at 0 cm and 6 cm only gives you two points and no shape; testing at 0, 2, 4 and 6 cm shows whether the effect keeps increasing or levels off. A range that is too narrow can hide a real effect entirely.

• Reliability and validity are different questions

  Reliability asks: if I did this again, would I get a similar result? You improve it with repeats, a consistent method, careful measurement, and by averaging. Validity asks: am I measuring what the question is really about? You improve it by choosing a sensible measure, controlling the other variables, and sampling the right thing. They are independent. Judging soil moisture by how dark the soil looks can be perfectly repeatable and still invalid, because colour also depends on the soil type. Repeating an invalid measurement just gives you the wrong answer consistently.

• Conclusions and evaluations that stay inside the data

  A conclusion states the trend or relationship you found, quotes the data that shows it, and answers the original question - no more. "Water loss fell from 41 g to 18 g as mulch depth increased from 0 cm to 6 cm, so deeper mulch reduced evaporation over the two days tested" is defensible. "Mulch saves water in every garden" is not, because it goes beyond what was tested. An evaluation then names a specific limitation, explains how it could have affected the result, and proposes an improvement that would actually address it - not the generic "be more accurate next time".

Internal · marked per part.

This is an internal achievement standard. Whetū does not offer a sit-down Exam paper for it. Learn and Practise stay available.

3 authored Learn units for this standard.

• Pattern seeking

  Pattern seeking looks for a relationship in variation the investigator does not set. You sample what already varies. You do not run a fair test of a factor you never changed. Traffic on a state highway, the tide, a season, a whole catchment — those are not dials on a bench. That is the giveaway that this investigative-approach is the one in play.

• Exploring and observing

  Exploring and observing is a systematic record of what is there, or what happens, without changing the system to isolate a cause. You do not set a factor. You do not yet have a sharp relationship question. A casual look is not this investigative-approach, and it is not “the scientific method”.

• Choosing and justifying an approach

  Science answers questions with more than one kind of investigation. An investigative-approach is one of five named tools — pattern-seeking, exploring-and-observing, modelling, classifying-and-identifying, or fair-testing — not a step in a single “scientific method”. None of them is the real science and the others a warm-up. Each one is for a different shape of question, and this page is about choosing the ones that fit.

32 Practise questions in “Science investigations”. Feedback here is formative and is not an official NCEA grade.

• Choosing an approach

  Match the investigation type to the question and to what you can control.

• Fair testing

  Change one variable, hold the rest steady, compare against a baseline.

• Pattern seeking

  Find a relationship in variation you are not able to control.

• Exploring and observing

  Record what a system does, systematically and without changing it.

• Modelling

  Use a simplified version of a system to explain or predict.

• Classifying and identifying

  Sort and name specimens using criteria anyone could apply.

• Variables

  Name what you change, what you measure and what you hold constant.

• Reliability and validity

  Judge whether data would repeat, and whether it measures the right thing.

• Writing conclusions

  State what the data shows - and stop where the data stops.

• Evaluating the method

  Name a limitation, its effect on these results, and a fix that would work.

Sample questions

1. “I can't make the kōwhai flower on cue, so I'm going to survey twelve reserves and see whether tūī counts line up with how much is in flower.” Which approach is this student taking?
2. Which approach fits this question best?
3. Match each type of investigation to the question it suits best.
4. Name the approach you would use, and justify it in a sentence - including why the obvious alternative would be weaker.
5. What makes an investigation a fair test?
6. What is the main problem with this method?
7. Which changes would make this a stronger fair test? Select all that apply.
8. Pattern seeking is the right approach when…

Practise 91921 in Whetū