ICAS Science common mistakes
ICAS Science · Understand the traps and how to avoid them.
Reads a value from a scale (ruler, measuring cylinder, thermometer, balance) by assuming every smallest division equals 1 unit, or by counting marks from the wrong end, and records the number without a unit.
Candidates learn 'read between the marks' but never check what one mark is worth, and the scale numbering (0, 20, 40 on a cylinder; 0, 10, 20 on a thermometer) is not what they expect.
What to do instead
Before reading, calculate the value of one small division: (gap between two labelled marks) divided by (number of divisions between them). Then count up from the nearest labelled mark below the level, read the bottom of the meniscus at eye level, and always write the unit next to the number.
Writes an inference where an observation is asked for, e.g. 'the ice melted because the room was warm' or 'the plant is unhealthy' instead of 'the ice became liquid water' or 'the leaves are yellow and drooping'.
Everyday description blends what is seen with why it happened, and an inference feels like a more complete, more impressive answer.
What to do instead
Use a two-column see / think split. An observation must be something another person could confirm from a photograph or by repeating the look: a colour, a count, a shape, a position, a measured value, a change over time. Words like 'because', 'so that' and 'wants' mark the inference column.
Reads values off a graph by matching across to the wrong axis, or ignores the key and treats the second line/bar as if it were the first, producing a number off by a large factor.
The eye follows the nearest gridline or the visually dominant line, and the key or axis labels are skimmed rather than read.
What to do instead
Name both axes and their units before touching the data, state which series is which from the key, and trace with a finger in two steps: horizontal to the line, then vertical to the axis (or reverse for a read-off). Check the answer is sensible for that axis's range.
In sorting and classification items, groups objects by a criterion of the candidate's own choosing (e.g. colour or size) rather than the feature named in the question, or mixes two criteria in one group.
The most visually obvious feature dominates attention, and 'group these' is read as an open invitation rather than a test of a specified property.
What to do instead
Restate the given criterion explicitly, then test each object against that one property at a time, writing yes/no beside it. If a group contains members split on any other feature, the grouping is wrong.
Extends a graph line well beyond the plotted points and reports the extrapolated value as if it were measured, e.g. predicting a plant's height at day 60 from data ending at day 20.
A straight line looks like a rule, and answering with a number feels better than saying the data cannot tell us.
What to do instead
Predictions should sit inside, or just at the edge of, the measured range. If the question forces an extension, state the assumption in words ('if the same trend continued') and mark such answers as less reliable than interpolated ones.
Writes a conclusion of the form 'this proves X' from a single trial, from two values that differ by less than the spread of the repeats, or while ignoring one result that clearly does not fit the pattern.
The word 'proves' is absorbed from general reading, and once a pattern is seen the odd value is quietly deleted rather than confronted.
What to do instead
Before concluding, scan the results for an outlier and for whether repeats agree; a difference only counts if it is bigger than the variation between repeats. Use 'suggests', 'supports' or 'is consistent with' rather than 'proves', and mention the repeat trials as the basis.
Identifies the independent and dependent variables wrongly, or accepts an experiment as a fair test when two things were changed at once (e.g. watering one plant more and also moving it to a sunnier spot).
The variables are described in prose and candidates match keywords rather than reasoning about what the investigator deliberately changed versus measured.
What to do instead
Fill in three boxes every time: what I deliberately change (one thing only), what I measure or observe, what I keep the same. If two boxes are needed for 'what I change', the test is unfair and the conclusion cannot isolate a cause.
Accepts a causal claim from a comparison that has no baseline, e.g. concluding a fertiliser works when the treated plant grew 5 cm and no untreated plant was grown.
The treatment group's result looks like evidence on its own; the idea that a number needs something to be compared against is not intuitive.
What to do instead
Always ask 'compared with what?' A treatment needs a control that differs in exactly the one variable under test. If no control group exists, the conclusion must be limited to describing what happened, not attributing it to the treatment.
Compares quantities in different units by their bare numbers (0.5 m judged smaller than 45 cm, or 3 kg judged lighter than 2500 g) because no conversion is done before choosing.
The number is more salient than the unit, and conversion is treated as an optional extra step rather than part of the comparison.
What to do instead
Convert every quantity to a common unit and write the conversion out (0.5 m = 50 cm; 3 kg = 3000 g) before comparing. As a check, ask whether the answer's size makes physical sense for the object involved.
Answers the question that was expected rather than the one asked, missing limiting words such as 'least likely', 'not', 'best explains' or 'first', and selecting the first option that is true in general.
Under time pressure the stem is skimmed for topic keywords, and a familiar-sounding option produces a recognition feeling that stops further checking.
What to do instead
Underline the qualifier and the exact task word in the stem before looking at the options. Then mark each option T/F specifically against that qualifier, and only then choose. If two options both look true, re-read the stem: the qualifier is what separates them.