Two Sources, One Thesis
Lerato is drafting a 1,500-word essay arguing that moving the library's closing time from 10 pm to midnight would reduce student stress. She has two sources. Source A: a Student Council poll of 240 students recruited through the Council's Instagram account; 68% agreed that later hours 'would help me feel less rushed'. Source B: a university-wide survey of 1,200 randomly selected students, comparing self-reported stress scores on two campuses - one closing at 10 pm, one at midnight - with a difference of 0.4 points on a 20-point scale (95% confidence interval: -1.1 to +1.9). Her tutor marked an earlier draft down for overclaiming and asked her to use at least one piece of counter-evidence.
1. Which source should carry more weight for Lerato's causal claim, and why? Be specific about the properties of each study rather than labelling one 'unreliable'.
Read the worked answer
Source B carries more inferential weight; Source A carries almost none for the causal claim - but it is not worthless.
- Source B uses random selection (1,200 students), so it is far less exposed to self-selection bias, and it reports an uncertainty interval. The interval (-1.1 to +1.9) straddles zero, so the honest reading is no detectable difference between the two campuses. That is counter-evidence to the causal claim, and it satisfies the tutor's requirement.
- Source A is self-selected (Instagram followers of the Student Council), so the 68% describes that group, not students generally. It also measures an expectation ('would help me feel less rushed'), not an outcome. People routinely mispredict how a change will affect them.
What A can still do: support a narrower claim about perceived pressure. A well-argued essay uses A to establish that students believe there is a problem, then uses B to show the belief has not yet been confirmed as a real effect.
2. Rewrite Lerato's thesis so that it is defensible given these two sources. Keep it to one or two sentences and make the scope explicit.
Read the worked answer
A defensible version separates the perception from the demonstrated effect:
'Students report feeling rushed under a 10 pm closing, but campus-level data show no measurable difference in stress between a 10 pm and a midnight campus, so later hours should be piloted and evaluated rather than assumed to reduce stress.'
Why this works:
- It reports the 68% as a perception and names the population it came from.
- It states B's finding at the right strength - 'no measurable difference', not 'no effect', because the interval still permits a small benefit or a small harm.
- It converts the essay's purpose from advocacy to a testable recommendation, which is what the tutor's 'overclaiming' comment was pointing at.
A common failure here is to write 'later hours do not reduce stress', which overclaims in the opposite direction from the original draft.
3. Lerato proposes to strengthen her case by finding 'one more survey showing students want longer hours'. Evaluate that plan and describe what evidence would actually help.
Read the worked answer
The plan adds volume, not weight. A second self-selected poll of the same population would repeat Source A's two weaknesses (self-selection, measuring expectation rather than outcome) and would not touch Source B's null result. Two biased polls do not become one unbiased study.
Evidence that would actually move the argument:
- A within-campus comparison - e.g. the same campus before and after a staggered extension, with stress measured the same way in both periods. This removes the between-campus differences that could explain Source B's result.
- Behavioural rather than attitudinal data - late-night library entry counts, submission timestamps, or usage of study spaces, which do not depend on students predicting their own feelings.
- Ideally an intervention with a comparison group, so that exam-period timing and seasonal effects are not mistaken for the effect of hours.
What this would not fix: novelty effects (a new midnight closing may feel better for a term) and the fact that a single institution's result may not transfer. She should say so, briefly, rather than pretend the question is settled.