Questions: Longitudinal Designs and Study of Temporal Change Patterns

Question 1

A longitudinal study follows the same cohort from age 20 to age 60 and finds that vocabulary scores decline significantly. A researcher concludes that aging causes vocabulary decline. What is the strongest objection to this conclusion?

• A. The study lacks a control group of people who did not age
• B. Historical effects cannot be separated from aging — this cohort's entire lifespan coincided with specific historical events that may account for the change
• C. The study should have used a cross-sectional design to compare different age groups at the same time
• D. Longitudinal studies can only describe change, never support any causal inference whatsoever

Answer: (B)

Even with temporal ordering established, a single-cohort longitudinal study cannot distinguish developmental change from historical effects. A cohort followed from 1960 to 2000 experienced specific historical events — educational policy shifts, cultural changes, health interventions — that affected all participants simultaneously. The vocabulary decline may reflect those historical conditions rather than aging per se. Option D overstates the problem: longitudinal designs do support causal inference better than cross-sectional designs, but temporal ordering is necessary, not sufficient.

Question 2

A researcher compares 30-year-olds and 60-year-olds in a single survey and finds the 60-year-olds score lower on a technology fluency test. Why can this NOT be interpreted as evidence that technology fluency declines with age?

• A. The sample sizes may be too small to detect a real difference
• B. The two groups differ in both age and generational experience — the 60-year-olds grew up before the digital era, so lower scores may reflect cohort, not aging
• C. Cross-sectional studies can only describe current states, not compare groups at all
• D. The test may not be reliable across different age groups

Answer: (B)

This is the core weakness of cross-sectional age comparisons: they confound age with cohort. The 30-year-olds and 60-year-olds differ not only in age but in when they grew up — the older group came of age before personal computers, smartphones, and the internet. Lower technology fluency may simply reflect their developmental history, not any decline in their individual abilities over time. A longitudinal study following the same people over decades would be needed to establish whether fluency actually changes with age within individuals.

Question 3

In a longitudinal study of cognitive aging, healthier participants are more likely to remain enrolled at later waves. This pattern of dropout can make cognitive decline appear smaller than it actually is.

• T. True
• F. False

Answer: True

This is selective attrition: participants who drop out tend to differ systematically from those who remain. If sicker, more cognitively impaired participants leave the study, the surviving sample at later waves is healthier on average — not because everyone improved, but because those who declined most severely are no longer in the data. The result is a biased estimate of the population's trajectory that understates actual decline. Researchers address this by analyzing dropout patterns and using missing-data methods, but selective attrition remains a serious threat to longitudinal validity.

Question 4

A longitudinal design that measures participants at multiple time points eliminates most confounds from the study, making causal conclusions straightforward.

• T. True
• F. False

Answer: False

Longitudinal designs establish temporal ordering — a necessary condition for causal inference — but they introduce their own confounds rather than eliminating all of them. Historical effects (events affecting all participants simultaneously), practice effects (score improvements from repeated testing rather than genuine change), and selective attrition all threaten validity in ways absent from single-session designs. Temporal precedence is necessary but not sufficient for causal inference; it must be combined with ruling out plausible alternative explanations.

Question 5

Explain why temporal precedence — knowing that Variable A measured at Time 1 preceded Variable B measured at Time 2 — is necessary but not sufficient to conclude that A caused B.

Model answer: Temporal precedence rules out one alternative: B did not cause A (since A came first). But it cannot rule out that a third variable caused both A and B in sequence, that historical events produced both, or that the relationship is coincidental. Causation requires temporal ordering, but also the absence of plausible confounds and ideally some mechanism linking A to B.

A classic example: children's shoe size measured at Time 1 predicts their vocabulary score at Time 2, but shoe size doesn't cause vocabulary — both are caused by age and development. The temporal ordering is real, but a common cause explains the association. Longitudinal designs control for some confounds (e.g., stable individual differences) but cannot by themselves eliminate third-variable explanations. Experimental manipulation remains the gold standard for causal claims precisely because it can randomly assign A, breaking its connection with potential confounders.