Effects of Physical Characteristics of Meeting Spaces on Discussion Activity, Psychological States, and Autonomic Nervous System Responses

The intervention consists of assigning one of six workplace meeting-room conditions that differ in acoustic privacy and visual openness: Panel, Frosted Glass, Glass, Open, Frame, and Louver. Meeting groups continue their regularly scheduled meetings under the assigned room conditions. Room conditions are randomly assigned to each meeting group on a weekly basis using a crossover design. Aside from the physical meeting environment, no aspect of meeting operation, agenda, or participant composition is intentionally altered.

2026-07-06

2027-03-31

1. Minute-level RMSSD during each meeting.
2. Minute-level LF/HF ratio during each meeting.
3. Total speaking volume per meeting.
4. Turn-taking intensity (speaker turns per minute).
5. Speech inequality across participants (Gini coefficient of speaking volume).
6. Average sentiment score of meeting utterances.

RMSSD and the LF/HF ratio are computed from beat-to-beat interval data recorded by wearable heart-rate sensors throughout each meeting.

Speaking volume is defined as the total number of spoken characters produced during a meeting based on speaker-labeled Microsoft Teams automatic transcripts.

Turn-taking intensity is defined as the number of speaker changes per minute.

Speech inequality is measured using the Gini coefficient of speaking volume across participants within each meeting.

Average sentiment is calculated by applying a pre-specified Japanese sentiment analysis model to the speaker-labeled meeting transcripts.

1. Subjective evaluation of external sound.
2. Subjective evaluation of visual distraction.
3. Subjective concentration.
4. Psychological safety (concern about being observed by others).
5. Meeting satisfaction.
6. Perceived liveliness of discussion.

All secondary outcomes are collected immediately after each meeting using a standardized post-meeting questionnaire with five-point Likert scales.

Behavioral observations extracted from video recordings, including reactions to external stimuli and participant posture, are treated as exploratory outcomes and are not included in the confirmatory analyses.

This study is a crossover field experiment conducted among recurring meeting groups within a firm, in which six spatial conditions (Panel, Frosted Glass, Glass, Open, Frame, and Louver) are randomly assigned at the meeting-group-by-week level. Each group is assigned a pre-specified order of the six conditions before the study begins (the order differs across groups), and conditions are fixed at the group-week level. Groups enter the study on different dates under a phased rollout, so a group's first treatment week may start mid-week. After a group enters the study, conditions are updated weekly, typically on Mondays.

The cluster is defined as the meeting group. The number of groups actually participating may increase over time as additional groups become operationally ready; eight groups are the planned upper bound.

Panel (the condition with the highest acoustic and visual privacy) serves as the reference condition, and the difference from each of the other five conditions is the primary object of estimation. The six conditions are not a fully factorial manipulation of acoustic and visual privacy, so the study does not interpret any single pairwise comparison as isolating the effect of acoustic privacy or visual privacy alone. Condition-specific treatment effects constitute the primary analysis; an aggregate comparison after reclassifying conditions by acoustic and visual openness is treated as an exploratory secondary analysis.

The primary analysis is restricted to recurring meetings with at least 3 in-person participants, an in-person participation rate of at least 70%, and an attendance rate of at least 80%. Meetings operated as standing hybrid meetings (routinely including remote participants) are excluded. For groups that meet more often than once per week, the primary analysis uses one pre-specified representative meeting per treatment week, in order to align the unit of observation with the weekly treatment assignment. Report-only meetings and unusually short meetings are excluded under pre-specified, quantitative definitions (rather than post hoc judgment); full definitions are provided in the Pre-Analysis Plan.

Estimation uses a two-way fixed-effects model with individual and week fixed effects, with condition-dummy coefficients identifying each condition's difference from the reference. In addition, a time-series fixed-effects model interacting treatment with elapsed time since meeting start (5-minute bins during the meeting, 10-minute bins afterward) is used to examine the timing and persistence of effects. Full estimation equations, the control-variable list, and sensitivity analyses are detailed in the Pre-Analysis Plan.

Not available

The research team conducted computer-generated randomization before the intervention. Each meeting group was assigned a pre-specified randomized sequence of the six meeting-room conditions using a computer-generated random-number algorithm.

Meeting group (cluster), with treatment assigned at the meeting-group-by-week level.

Yes

8 meeting groups (groups with substantial membership changes will be treated as new groups.)

Approximately 150 eligible meeting sessions. The primary analysis sample is defined at the meeting-session level, where one session is one representative meeting for a given meeting group in a given treatment week. Physiological outcomes are observed at the participant-by-minute level within each eligible session. Separately, the total number of meetings actually held across the 8 target groups during the study period (the raw count of occurrences, including all weekly occurrences for daily groups) is approximately 390; this is distinct from the primary analysis observation unit, one representative meeting per treatment week (see Sections 3 and 5 of the Pre-Analysis Plan). The total number of sessions is a planning estimate based on each group's meeting frequency and phased entry into the study, and may vary due to personnel changes, staggered group start dates, and meeting feasibility. The number of observations per condition depends on randomized assignment and operational feasibility; perfectly balanced allocation is not required.

Approximately 25 eligible meeting sessions per room condition:

* Panel: approximately 25 sessions
* Frosted Glass: approximately 25 sessions
* Glass: approximately 25 sessions
* Open: approximately 25 sessions
* Frame: approximately 25 sessions
* Louver: approximately 25 sessions

The final allocation may vary slightly because of staggered group entry, personnel changes, and operational constraints.

Pre-Analysis Plan (July 2, 2026)