What being heard does — the research on listening as an intervention

Most writing about communication focuses on what speakers say. A smaller body of research focuses on what listeners receive, and on what changes for both parties when the receiving is done well. That research consistently identifies a counterintuitive pattern: being heard appears to produce benefit separate from — and sometimes in the absence of — any response from the listener.

What is “being heard”?

In communication and clinical research, “being heard” refers to the experience of having one's speech received, processed, and registered by another mind. It is distinguished from related experiences such as being advised (receiving guidance), being validated (receiving agreement), and being responded to (receiving a reply). All three of those can accompany being heard; none is required for being heard to occur.

The phenomenon has been studied empirically primarily through two channels: comparisons of voice and text communication (which isolate what the channel transmits), and studies of attentive listening as a discrete behavior (which isolate what the listener does).

What does the research show about voice channels?

The most cited research on voice communication comes from Nicholas Epley's lab at the University of Chicago.

Schroeder, Kardas, and Epley's 2017 study in Psychological Science, titled The Humanizing Voice, demonstrated that listeners rate the same substantive arguments as more thoughtful, more rational, and as coming from a more capable mind when delivered by voice rather than text. The effect held across experimental conditions and across both audio recordings and live phone calls. Voice consistently produced a richer perception of the speaker than text did, despite the words being identical.

Kumar and Epley's 2021 follow-up, published in the Journal of Experimental Psychology: General, asked participants to predict how connected they would feel after voice versus text communication with someone they liked. Predictions were similar across the two conditions, and many participants predicted voice would feel more awkward. The actual outcomes showed the opposite: phone calls produced significantly stronger feelings of connection, with no detectable awkwardness penalty.

The mechanism the authors propose is paralinguistic: voice transmits tone, pace, breath, pause, and prosody — all information that text does not encode. The listener receives a fuller signal and responds to a more complete picture of the speaker.

What does the research show about listening itself?

A second strand of research examines listening as a discrete behavior independent of channel. Hall and colleagues' 2023 work in Communication Research followed adults across daily life and found that a single meaningful conversation per day — defined as an exchange requiring attentive listening from both parties — produced measurable increases in same-day wellbeing.

The effect did not require advice, problem-solving, or insight. The intervention was the conversational attention itself.

This finding aligns with a broader pattern in communication-and-relationships research: the reception of one's words by an attentive other appears to be itself a causally active component, distinct from whatever the listener says or does in response.

What lives in the pauses

Speech includes paralinguistic information that text cannot represent: the breath before a difficult sentence, the quarter-step drop in pitch on a hard word, the hesitation that precedes saying something true, the change in pace when attention shifts. A listener registers this information, often without conscious awareness, and responds to it as part of the perceived meaning.

Text encodes propositional content. It does not encode the speaker's state at the moment of speaking. For conversations in which the meaning lives mostly between the words — most emotionally substantive conversations — the channel determines what the listener receives.

What this means for AI conversation tools

The research has direct implications for the design of AI conversation products:

1. Text-based AI cannot receive paralinguistic information.The user's tone, pace, breath, and hesitation are stripped before the model sees the message. Whatever the model produces in response is operating on a degraded signal.
2. Voice-based AI receives the full signal. The architectural capacity is present. How much any particular product uses the signal is a function of how it is built.
3. Neither substitutes for human relationships. A 2024 study by Maples and colleagues found AI conversational support has specific use cases (notably stress-management) but functions as a complement to — not a replacement for — human relationships and professional care.

Voice-based AI conversation tools — including CallByrd, a phone-based AI designed for unstructured conversation — receive a structurally richer signal than chat-based AI. They are not substitutes for a human listener, but they operate on a different and more complete representation of the speaker.

The bottom line

The research literature converges on a consistent pattern: voice transmits information that text strips out; listeners perceive a fuller picture of the speaker when they hear rather than read; and the act of being heard appears to be itself an active ingredient — separate from any response — that contributes to wellbeing and connection. For conversation as a goal in itself, voice is the structurally appropriate medium.