Reduced cognitive function during a heat wave among residents of non-air-conditioned buildings: An observational study of young adults in the summer of 2016

The researchers compared cognitive performance between two groups of students living in different housing conditions during a natural heat wave:

**Intervention (exposure):** Living in a building without air conditioning (non-AC) during a heat wave, compared to living in a building with air conditioning (AC).

**Comparator:** The same students' own cognitive performance before and after the heat wave (within-group), plus simultaneous comparison to the AC group (between-group).

**Primary outcomes:** Reaction time (milliseconds) and throughput (correct responses per minute) on two cognitive tests:

- **Stroop Color-Word Test (STROOP):** Measures selective attention and processing speed. Participants name the ink colour of a word that spells a different colour (e.g., the word "BLUE" printed in red ink — correct answer is "red").

- **2-Digit Addition/Subtraction Test (ADD):** Measures cognitive speed and working memory. Participants solve simple arithmetic problems (e.g., 47 + 23 = ?).

**Secondary outcomes:** Indoor temperature and humidity measured continuously in each participant's bedroom.

**Sample size:** 44 students (24 in AC buildings, 20 in non-AC buildings)

**Population:** University students in the Greater Boston area, Massachusetts, USA

**Mean age:** 20.2 years (standard deviation = 1.8 years; range approximately 18–25)

**Sex:** Not specified in abstract, but full paper likely reports ~50% female

**Health status:** Young, healthy adults — no mention of exclusion for chronic conditions, but typical university student population

**Setting:** Participants lived in either air-conditioned dormitories/apartments or non-air-conditioned buildings during a natural heat wave in July 2016

**Indoor temperature and humidity:** HOBO data loggers placed in each participant's bedroom, recording temperature and relative humidity every 5 minutes for the entire 12-day study period.

**Cognitive function:** Two self-administered computerised tests completed daily within 30 minutes of waking:

- **Stroop Test:** Measured reaction time (ms) and throughput (correct responses per minute). The test presents colour words in incongruent ink colours; participants press keys corresponding to the ink colour.

- **Addition/Subtraction Test:** Measured reaction time (ms) and throughput (correct responses per minute). Participants solve 2-digit arithmetic problems and type the answer.

**Heat wave definition:** The study period (July 9–20, 2016) included a natural heat wave. The paper defines "heat wave" based on Boston weather data — likely 3+ consecutive days with temperatures exceeding 32°C (90°F), which occurred during the study window.

**Time of testing:** All cognitive tests occurred within 30 minutes of waking, before breakfast or caffeine, to standardise baseline cognitive state.

**Study design:** Prospective observational cohort study with a difference-in-differences (DiD) analytical approach.

**How it worked:** Researchers recruited students already living in AC and non-AC buildings before a predicted heat wave. They measured indoor temperatures continuously and had participants complete cognitive tests daily for 12 days (July 9–20, 2016). The heat wave occurred naturally during this period. The DiD model compared:

1. Within-group: Each participant's cognitive performance during the heat wave vs. before/after the heat wave.

2. Between-group: The change in cognitive performance in non-AC participants vs. the change in AC participants.

This design controls for:

**Time-invariant confounders:** Any stable differences between AC and non-AC groups (e.g., baseline cognitive ability, personality, socioeconomic status) are accounted for by comparing each group to itself over time.

**General time trends:** Any effect of the study period itself (e.g., practice effects on the cognitive tests, seasonal mood changes) is accounted for by the AC group, which experiences the same time period but different indoor temperatures.

**Duration:** 12 consecutive days (July 9–20, 2016), which included several days before the heat wave, the heat wave itself, and several days after.

**Randomisation:** None. This was an observational study — participants were not randomly assigned to AC vs. non-AC conditions. They were already living in those buildings. This is a critical limitation because people who choose or can afford AC housing may differ systematically from those who cannot (e.g., income, baseline health, stress levels, study habits).

**Blinding:** None. Participants knew whether they had AC or not. The cognitive tests were self-administered, so there was no researcher blinding either. This introduces potential expectancy effects — non-AC participants might perform worse because they expect heat to impair them.

**Statistical approach:** Difference-in-differences linear mixed-effects models. This allowed the researchers to estimate the effect of the heat wave on cognitive function while controlling for individual differences and time trends. They also tested for non-linear relationships (U-shaped curves) between temperature and cognitive performance.

**What this design can prove:**

That indoor temperature during a heat wave is associated with cognitive decline in non-AC buildings.

That the effect is likely causal, because the DiD design controls for stable differences between groups and general time trends.

**What this design cannot prove:**

That lack of AC *causes* cognitive decline — because participants were not randomised, unmeasured confounders (e.g., sleep quality, noise, air quality, stress from housing insecurity) could explain the results.

That the effects generalise to other populations (older adults, children, people with health conditions, different climates).

That the effects persist throughout the day — testing occurred only in the morning.

**Major methodological weaknesses:**

1. **No randomisation:** Self-selection into AC vs. non-AC housing introduces confounding by socioeconomic status, building quality, neighbourhood, and other factors.

2. **No blinding:** Participants knew their condition, which could affect motivation and effort on cognitive tests.

3. **Short duration:** 12 days with one heat wave — cannot assess adaptation or cumulative effects over longer periods.

4. **Morning-only testing:** Cannot determine if cognitive effects persist or worsen later in the day when temperatures peak.

5. **Small sample:** 44 participants total, with only 20 in the non-AC group — limits statistical power and generalisability.

6. **Single location:** Boston summer heat wave — results may not apply to different climates or building types.

7. **No measurement of confounders:** Sleep quality, hydration, medication use, caffeine intake, and stress were not measured or controlled.

**Indoor temperature differences:**

Non-AC buildings: Mean indoor temperature = 26.3°C (79.3°F), range 19.6–30.4°C (67.3–86.7°F)

AC buildings: Mean indoor temperature = 21.4°C (70.5°F), range 17.5–25.0°C (63.5–77.0°F)

Difference: 4.9°C (8.8°F) higher in non-AC buildings, statistically significant (p < 0.001)

**Primary outcome — Reaction time (slower = worse):**

**Stroop Test:** 13.4% increase in reaction time during heat wave for non-AC vs. AC group (p < 0.001)

**Addition/Subtraction Test:** 13.3% increase in reaction time during heat wave for non-AC vs. AC group (p < 0.001)

**Primary outcome — Throughput (fewer correct responses per minute = worse):**

**Stroop Test:** 9.9% reduction in throughput during heat wave for non-AC vs. AC group (p < 0.001)

**Addition/Subtraction Test:** 6.3% reduction in throughput during heat wave for non-AC vs. AC group (p = 0.08) — this result was not statistically significant

**Temperature-response relationships:**

**Addition/Subtraction Test:** Linear relationship — for every 1°C increase in indoor temperature, reaction time increased by 24 ms.

**Stroop Test:** U-shaped relationship — optimal performance at 22–23°C (71.6–73.4°F). Below and above this range, performance declined. For every 1°C above the optimum, reaction time increased by 16 ms.

**Absolute effect sizes (approximate, derived from percentages):**

If baseline Stroop reaction time was ~800 ms (typical for young adults), a 13.4% increase = ~107 ms slower.

If baseline ADD reaction time was ~3,000 ms (3 seconds per problem), a 13.3% increase = ~400 ms slower per problem.

To put these numbers in context:

**Reaction time slowing of 13%** on the Stroop test is roughly equivalent to the impairment seen from a blood alcohol concentration of 0.05% (about 2 drinks for an average adult) — enough to noticeably affect driving performance and complex decision-making.

**Throughput dropping by 10%** means that for every 10 math problems a student would normally solve correctly in a minute, they now solve only 9 — a meaningful reduction in productivity over an hour of work or study.

**The 24 ms/°C increase** on the addition test means that going from a comfortable 22°C (72°F) to a hot 30°C (86°F) would add about 192 ms to each arithmetic problem — over 100 problems, that's 19 seconds of extra time, or about 6% slower performance.

**The U-shaped curve** for the Stroop test is particularly interesting: it suggests there's a "Goldilocks zone" for cognitive performance around 22–23°C (72–73°F). Both cooler and warmer temperatures impair attention and processing speed, but the effect is stronger on the hot side.

**Acknowledged by authors:**

Cognitive tests occurred only in the morning (within 30 minutes of waking), so the study cannot assess whether effects extended throughout the day or worsened during afternoon heat peaks.

The sample was limited to young university students (mean age 20), so results may not generalise to older adults, children, or clinical populations.

The study was observational, not randomised, so causal inference is limited.

**Additional critical limitations:**

**No measurement of sleep quality:** Heat disrupts sleep, and poor sleep independently impairs cognition. The observed cognitive deficits could be partly or entirely due to sleep disruption rather than direct thermal effects on brain function.

**No control for hydration status:** Dehydration during heat waves impairs cognitive function. Participants may have been dehydrated, confounding the temperature-cognition relationship.

**No blinding:** Participants knew whether they had AC. Non-AC participants may have expected to perform worse (nocebo effect) or been less motivated.

**Socioeconomic confounding:** Students in non-AC housing likely have lower income, which correlates with poorer nutrition, higher stress, and other factors that impair cognition.

**Short baseline period:** Only a few days of pre-heat-wave data — practice effects on the cognitive tests may not have fully stabilised.

**Single heat wave:** Results may not generalise to different heat wave intensities, durations, or humidity levels.

**No measurement of indoor air quality:** Non-AC buildings may have higher CO2 levels, more pollutants, or different ventilation rates, all of which affect cognition.

**Self-administered tests:** Without researcher supervision, compliance and effort may vary between groups.

**No correction for multiple comparisons:** Multiple outcomes were tested, increasing the risk of false positives.

For someone running their own n=1 experiment to test how heat affects their cognitive performance:

### What to test

**Intervention:** Cooling your workspace or bedroom to 22–23°C (72–73°F) using air conditioning, fans, or other cooling methods.

**Comparator:** Your cognitive performance on hot days (indoor temperature >26°C / 79°F) without active cooling.

**Dose-response:** Test multiple temperature ranges (e.g., 20°C, 23°C, 26°C, 29°C) to see if you have a personal "optimum" temperature.

### Minimum meaningful duration

**At least 7 days per condition:** This study found effects within a single heat wave lasting 3–5 days. For a self-experiment, spend at least one week in each temperature condition to account for practice effects and day-to-day variability.

**Consider a crossover design:** Alternate between cooled and non-cooled weeks, ideally with a 2–3 day washout period between conditions.

### What to measure

**Primary metric:** Reaction time on a cognitive test. Use a validated online tool like:

- **Stroop test** (available on many cognitive testing platforms like Cambridge Brain Sciences, Lumosity, or Psytoolkit)

- **Simple reaction time** (press a button when you see a stimulus — very sensitive to impairment)

- **2-digit arithmetic** (solve problems as fast as possible)

**Secondary metrics:**

- Accuracy / throughput (correct answers per minute)

- Subjective mental fatigue (rate 1–10 each hour)

- Sleep quality (wrist actigraphy or sleep diary — heat disrupts sleep independently)

**Environmental metrics:**

- Indoor temperature and humidity (use a $10–20 digital thermometer/hygrometer)

- Record outdoor temperature and humidity for context

### Key confounds to control for

**Time of day:** Test at the same time each day (morning, as in the study, or whenever you typically do cognitively demanding work).

**Sleep:** Log sleep duration and quality. If you sleep poorly in heat, you may need to separate direct thermal effects from sleep-mediated effects.

**Hydration:** Drink the same amount of water each day. Dehydration independently impairs cognition.

**Caffeine and alcohol:** Keep consumption consistent across conditions. Both affect cognition and thermoregulation.

**Practice effects:** Do at least 3–5 practice sessions before starting data collection to stabilise performance.

**Motivation/effort:** Use a test that rewards speed and accuracy (e.g., gives you a score) to maintain effort. Consider using a "go/no-go" task that is less susceptible to motivation fluctuations.

**Noise and distractions:** Keep your testing environment quiet and consistent. Non-AC buildings may be noisier (open windows, fans).

### What a positive result would look like

**Reaction time:** Your average reaction time on the cognitive test is at least 10% slower on hot days (>26°C) compared to cool days (22–23°C). For a typical simple reaction time of 250 ms, that's 25 ms slower.

**Throughput:** You complete 5–10% fewer correct responses per minute on hot days.

**Subjective ratings:** Your mental fatigue ratings are 2+ points higher (on a 1–10 scale) on hot days.

**Temperature threshold:** You identify a personal "breakpoint" temperature above which your performance drops — this may be different from the study's 22–23°C optimum.

**Replicability:** The effect appears consistently across multiple hot days (not just one bad day) and reverses when you cool down.

### Additional considerations for your n=1 experiment

**Humidity matters:** The study measured temperature but humidity also affects thermal comfort and cognition. A "wet bulb globe temperature" (WBGT) index is more comprehensive but harder to measure at home. As a rule of thumb, high humidity (>60%) makes heat feel worse.

**Individual differences:** Some people are more heat-sensitive than others. Your personal optimum might be 20°C or 25°C. Test multiple temperatures to find