Operational voice / Research review

Who gets to speak next?

Push-to-talk lets one person speak at a time. People, team rules, and radio settings determine who speaks next, what happens when voices collide, and when an exchange is complete.

Memo Research · 30 September 2026 · 5 min read

Figure 01 Schematic: speaker A releases a shared channel; speakers B and C then begin within the same short interval. The overlap and timing are illustrative, not a measurement from a working channel.

A push-to-talk channel lets people speak in both directions, but not at the same time. The radio may control access to the channel, but it cannot decide how a conversation should work. Crews still need rules for starting, yielding, replying, and recovering when two people speak together.

The radio sets the limit. People make it work.

Half-duplex communication works in both directions, but not at the same time on the same path. Push-to-talk makes that limit part of everyday work: one person transmits while everyone else listens. Direct radio, repeaters, and time-slotted systems handle the path differently, so the shared point is simple: only one direction is open at a time.[2]

That limit does not decide who speaks next, how long they speak, or how a listener knows the exchange is complete. Research on everyday conversation shows that one person usually speaks at a time. Near the end of a turn, the speaker may choose who answers, or someone else may step in. Debates, ceremonies, and other settings organize those turns differently.[7]

Radio is another speech-exchange system, but one with a difficult failure mode. People transmitting at the same time may be unable to hear one another, so the immediate repair available in face-to-face overlap can disappear. Procedure has to replace information that the medium withholds.[7][4]

Aviation writes the turn-taking rules down

The FAA tells pilots to listen before transmitting because keying over another transmission can jam receivers and force a repeat. It advises a slight pause after pressing the microphone button so the first word is not clipped, a wait of a few seconds after release before calling again, and attention to a stuck microphone that can block a frequency for an extended period.[1]

The same guidance connects a reply to a named person. Pilots should answer an air-traffic-control message with the right aircraft call sign, and the FAA calls for extra care when call signs sound alike. In plain terms, the rules cover when to speak, how to start, when to wait, and how to reply. They live in training and team procedure, not just in the radio.[1]

Listen-before-transmit is configurable

The current DMR air-interface specification supports three high-level channel-access behaviors. A radio configured as polite to all refrains while the channel is busy with any activity. Polite to its own colour code yields to DMR traffic from the same system but may transmit over other activity. Impolite access transmits regardless of activity already present. The standard also notes that voice can be configured impolite while packet data is configured polite.[2]

DMR also gives a call internal timing. Conventional voice begins with link-control signaling before speech, vocoder frames are carried in six-burst 360 millisecond superframes, and a base station may hold a two-frequency channel busy during a configurable call hang time. A gap in audible speech is therefore not always an invitation for a new call, and the same fleet can behave differently under different access settings.[2]

The standard describes what radios are allowed to do, not how real fleets are usually set up. We still need field evidence to know which settings industrial sites use and whether those settings reduce stepped-on calls.[2]

The collision may be hidden

EUROCONTROL describes a failure in air-ground systems that use best-signal selection. When two signals arrive together, the system may pass the stronger and suppress the weaker so the controller hears one intelligible transmission rather than a mixture. That improves intelligibility for the selected signal but can conceal the fact that a second person transmitted at all. This behavior belongs to systems using that selection method; it should not be generalized to every repeater or receiver.[4]

A European study examined 535 reported communication occurrences and surveyed 344 pilots and controllers, 90 percent of them pilots. About 60 percent of pilots reported encountering a blocked transmission weekly, compared with 20 percent of controllers. Respondents most often associated blocked transmissions with congestion, controller workload, untimely transmission, pilot workload, and long messages. These are subjective frequency estimates from an uneven sample, not measured channel rates.[3]

Respondents also reported that mixing local languages on a shared frequency made it harder for non-native listeners to judge when a transmission had ended, contributing to untimely starts. The finding comes from aviation and should not be treated as a measured result for multilingual worksites. It does identify a testable mechanism: turn-taking depends partly on recognizing linguistic completion.[3]

Sometimes the turn structure is part of the safety system

For covered crane signaling in construction, OSHA requires electronic signaling devices to be tested on site before operations and to provide effective, clear, and reliable transmission. Signals generally travel over a dedicated channel, subject to stated exceptions, and the operator must receive them hands-free. A dedicated path reduces unrelated contention; it does not prove that interference or simultaneous signaling cannot occur.[5]

The operator, signal person, and lift director must agree on voice signals before work begins and revisit them when personnel or signals change or confusion arises. Every voice signal must contain, in order, the function and direction, the distance or speed, and the function with a stop command. The participants must be able to communicate effectively in the language used.[6]

These rules apply to the covered crane operation, not to radio work generally. They show what a tightly specified voice protocol looks like when a spoken turn directly controls machinery: the path is tested, the vocabulary is agreed, and each command terminates explicitly.[5][6]

Study exchanges, not isolated transmissions

A field study should examine whole exchanges, not isolated transmissions. Measure the gap between calls, attempts to speak while the channel is busy, clipped openings, replies, retries, and how people recover from possible overlap. Record the channel settings with the audio, including access policy, colour code, and hang time where relevant, because the same pause can mean different things under different settings.[2]

Sample quiet and routine periods as well as congestion. Record the languages in use and whether listeners share them. Preserve unknowns: a receiver recording can contain the selected transmission without revealing a suppressed one, so absence of audible overlap is not evidence that nobody else keyed.[3][4]

This method could test whether access settings reduce collisions, whether crews develop stable local rules for allocating turns, and whether talkers and listeners experience the same channel differently. The reviewed sources do not answer those questions for industrial work.[2][3]

Evidence boundary

Limits & open questions

The title is an operational shorthand. Half-duplex is a communication constraint; the protocol is the combined set of technical access rules and human turn-taking practices built around it.
Most procedural and incident evidence comes from civil aviation, where radio work is trained, standardized, and reported. It cannot be assumed to transfer unchanged to industrial channels.
The DMR specification describes configurable behavior, not how industrial fleets are configured or how often each access policy is used.
The aviation survey reports subjective frequency estimates from 344 respondents, 90 percent of them pilots. It does not measure collision rates from recorded traffic.
OSHA's requirements apply to covered crane signaling in construction and are not a general template for every voice workflow.
Nothing in this review measures how often simultaneous transmissions occur on working industrial channels or what they cost.

How this article was researched

We reviewed primary sources on September 30, 2026. Claims were checked against the current FAA manual, the 2023 DMR air-interface specification, OSHA's crane-signal rules, EUROCONTROL reports, and the original paper on conversational turn-taking. We keep radio behavior separate from human procedure and label lessons carried from one field to another as interpretations or open questions.

Article type: research review. The diagrams are explanatory schematics. No new Memo field trial, corpus measurement or model benchmark is reported here.