The passage establishes a controlled experiment where changes are introduced sequentially to identify their effects on food waste. In weeks 1-2, moving fruit to the start increased fruit consumption but total waste remained unchanged, suggesting students may have wasted less of other foods to compensate. In weeks 3-4, when pasta portions were reduced by 15% while keeping fruit at the start, total waste dropped by 18%. Since the menu itself never changed and fruit consumption stayed high, the logical inference is that reducing pasta portions contributed to the waste reduction. Choice A incorrectly suggests causation in the wrong direction, Choice C introduces an unsupported claim about student preferences, and Choice D overgeneralizes from a single school's results. When multiple changes occur together, look for the change that coincides with the observed effect—here, waste only dropped after pasta portions were reduced.

The passage presents a natural experiment where tree planting effects must be separated from broader environmental changes. Street A (trees on both sides) saw a 2.5°C drop, Street B (trees on one side) saw a 1.0°C drop, but the entire city experienced a 0.7°C cooling. To isolate the tree effect, we subtract the citywide cooling: Street A likely gained an additional 1.8°C cooling from trees, while Street B gained 0.3°C. This suggests trees provided cooling beyond the general trend. Choice B reverses causation—two streets cannot cause citywide cooling, Choice C contradicts the data showing different drops on different streets, and Choice D overgeneralizes from limited evidence. In inference questions involving confounding factors, the correct answer acknowledges both influences rather than attributing everything to one cause.

The passage reveals a correlation study with an important confounding variable: timing of preparation. Group 2 used practice quizzes, studied fewer hours, and scored higher—but they also started earlier. Because the researcher didn't control for when students began studying, we cannot determine whether the higher scores resulted from the study method or the earlier start (or both). The most logical inference acknowledges this ambiguity: practice quizzes may be associated with higher scores, but timing could also matter. Choice B incorrectly claims causation from correlation, Choice C overgeneralizes about rereading without evidence, and Choice D makes an unsupported claim about study duration and timing. When research doesn't control for all variables, avoid conclusions that assume only one factor matters.

The passage establishes that two changes occurred simultaneously: a new sign and a weekend discount. The key evidence is that the less-visited gallery's share of total visits increased, not just its raw count—this means it grew faster than the museum overall. If only the discount mattered, we'd expect proportional growth across all galleries. The disproportionate increase suggests the sign specifically directed people to that gallery. Choice A incorrectly attributes everything to the discount, Choice C contradicts the data showing increased attendance, and Choice D wildly overgeneralizes from one museum's experience. When multiple factors change together, look for evidence that helps separate their effects—here, the proportional shift points to the sign's influence.

The passage describes an experiment attempting to isolate light color effects, but with a confounding temperature variable. Plants showed different growth patterns under different colored lights (blue: taller; red: more leaves), with other factors supposedly controlled. However, the red-light lamp emitted more heat, introducing temperature as an uncontrolled variable. The logical inference acknowledges both possibilities: light color may affect growth patterns, though temperature differences could also play a role. Choice B overgeneralizes to all species and conditions, Choice C invents a water difference not mentioned, and Choice D incorrectly claims heat cannot matter. When experiments have imperfect controls, avoid conclusions that ignore potential confounding factors.

The passage traces visitor patterns suggesting substitution between time slots. Key evidence: evening visits increased by 50%, many evening visitors reported previously coming on Saturdays, and Saturday visits fell by 15%. The 10% increase in total daily visits means some growth came from new users, but the Saturday decline and visitor comments suggest many evening visitors shifted from weekends. The logical inference is that some evening visitors likely shifted from Saturday visits rather than being entirely new users. Choice B makes an unsupported claim about reading habits, Choice C invents an explanation not in the text, and Choice D overgeneralizes with false precision. When changes in one area coincide with opposite changes elsewhere, consider whether people are substituting rather than disappearing.

The passage reveals age-based differences in response to text reminders, with a clear mechanism explaining why. No-shows dropped from 14% to 9% overall, but the improvement concentrated among younger patients while older patients showed little change. The key insight: many older patients had landlines rather than mobile phones, preventing them from receiving text reminders. This technological barrier explains the age disparity better than assumptions about attitudes. Choice B makes an unfounded judgment about older patients' health concerns, Choice C illogically connects reminders to scheduling, and Choice D overgeneralizes. When interventions show demographic differences in effectiveness, look for mechanical explanations before assuming behavioral differences.

The passage uses a comparison group to isolate the effect of improved lighting from broader trends. Both improved-lighting blocks (12% decrease) and unchanged blocks (5% decrease) saw theft reductions, suggesting a general trend possibly due to the awareness campaign. However, the larger decrease in improved-lighting blocks (12% vs 5%) suggests lighting provided additional crime reduction beyond the general trend. Choice B incorrectly attributes the difference to the campaign affecting blocks differently, Choice C dismisses lighting effects despite the comparative evidence, and Choice D overgeneralizes with false precision. When using comparison groups, the difference between groups, not the absolute change in one group, best indicates the intervention's effect.

The passage presents simultaneous interventions that could both reduce disputes. The mobile app increased pre-boarding ticket purchases (60% to 85%) while disputes dropped 40%, but more fare inspectors were also deployed. Both changes logically reduce disputes: the app by reducing on-board purchases, and inspectors by catching more violations. Since both occurred together, we cannot isolate their individual effects. The logical inference acknowledges this ambiguity: the drop may be due to the app, more inspectors, or both. Choice B incorrectly attributes everything to the app, Choice C suggests an illogical causal chain, and Choice D overgeneralizes with false precision. When multiple plausible causes exist for an effect, the strongest inference acknowledges the uncertainty.

The passage presents A/B testing data revealing a trade-off between two metrics. Subject line Y achieved more opens (24% vs 22%) but fewer clicks (2% vs 3%). Since content and timing were identical, the difference must stem from how the subject lines set expectations or attracted different reader segments. The logical inference acknowledges both patterns: Y may attract more opens while X leads to more clicks—neither is universally 'better.' Choice B incorrectly declares Y superior based on one metric, Choice C contradicts the data showing different click rates, and Choice D makes an unfounded claim about sampling. When data shows trade-offs between metrics, resist choosing a 'winner'—the correct inference often acknowledges that different approaches serve different goals.