Heard

Heard is a free iPhone app that scores how well one person listened in one real conversation. Two people put the phone between them and talk about something that matters, for up to 30 minutes. One of them is the listener. When the conversation ends, the phone turns the audio into text, works out who said what, and shortens the names, all on the phone. The shortened text is then scored, and the listener gets a Listening Report: a score from 0 to 100 built only from exact counts, and under it the full report of what they did, with their own words quoted back as evidence, and one thing to try next time. Record again and the app shows what changed, measure by measure.

The facts: free to download, iPhone only, iOS 26.5 or later, rated 4+, published by SoHo Creative LLC. It is on the App Store under the full name Heard: Who Really Listened, and at cdlab.io/heard.

Bring something that matters. Record the conversation. Find out who really listened.

Not that we could find, and we looked. On August 30, 2026 we read the pricing pages and App Store listings of dozens of products across the three categories people compare us to, and we re-checked the closest ones on September 13, 2026.

Speech coaches (Yoodli, Orai, Speeko, and others) score the speaker: pace, filler words, clarity, usually in solo practice or with an AI role-play partner. None of them scores the listener, and none scores a real conversation between two people.

Couples apps (Paired, Lasting, Agape, Flamme, Coral) sell daily questions and quizzes. One, Connected, records a real five-minute talk with both partners' consent and returns an AI reflection. Its own page says it is a reflection, not therapy, and the recording is deleted after analysis. It does not count anything.

Meeting note takers (Otter, Fireflies, Fathom, Granola, tl;dv, Plaud) transcribe and summarize what was said. They do not measure how anyone listened.

The tools that do count talk time, questions, and reflections on real recordings exist, and you cannot buy them at retail: Lyssn sells to healthcare organizations, Gong sells to sales teams, and the Gottman Love Lab records couples behind a clinician. Heard puts that kind of count in your own hand for $3.99. The one consumer product we found that measured talk share and interruptions in real meetings, Poised, announced on its own homepage that it is shutting down October 8, 2026.

No. Otter records a meeting and gives you back what was said: a transcript, a summary, action items. Heard gives you back how you listened: your share of the talking, your turns and your longest turn, how many questions you asked and how many were open, how often you reflected what you heard before asking the next thing, where you gave advice nobody asked for, and where you turned the attention back to yourself. Otter is a memory. Heard is a mirror. You can use both.

The other difference is where the audio goes. Otter, like the other note takers, transcribes your recording on its servers (see Otter's own privacy page). Heard never sends audio anywhere. The transcript is made on your iPhone by Apple's on-device speech model, and only the shortened text leaves the phone to be scored.

Five things, and every one of them is checkable.

1. It scores the listener, not the speaker.

2. It scores a real conversation between two people, not a rehearsal with an AI.

3. The numbers that matter are counted by software on the transcript and stated as fact, and the report tells you which numbers are judgment.

4. Audio never leaves the phone, and there are no accounts, no feeds, no leaderboards.

5. You pay per report, not by subscription, and reports you already ran stay on your phone free.

The three closest cousins, and how each differs. Lyssn counts talk time, questions, reflections, and rates empathy on real clinical recordings, sold to healthcare organizations for training clinicians; its published research is part of why we trust that these behaviors can be counted by software at all. Connected records a real short conversation with a pass-the-phone consent ritual, then deliberately returns a reflection instead of a score. Gong measures talk-to-listen ratio on sales calls for sales teams. We are not aware of another consumer app that counts listening behaviors on a real two-person recording. If one appears, we will say so here.

Two answers, because there are two things on the report.

The Listening Score, 0 to 100, is arithmetic we designed, and we print it. It has two dials. Room, up to 60 points: sixty times the share of the words the other person got, divided by fifty, capped. Give the other person half the words or more and the dial is full. Curiosity, up to 40 points: forty times the percentage of your questions that were open, divided by seventy, capped; ask no questions and it is zero. Both dials use only exact counts, so the same conversation scores the same number every time. The two marks are not ours. An even share of the talking is 50 percent by definition. Seventy percent open questions is the competency benchmark in the Motivational Interviewing Treatment Integrity manual, version 3.1.1, the coding system researchers use to measure how clinicians listen.

The measures under the score come from that same research tradition. Share of talking, turns, open and closed questions, simple and complex reflections, the reflection-to-question ratio, and empathy and partnership rated 1 to 5 are all defined in the MITI manual from the University of New Mexico's Center on Alcohol, Substance use, and Addictions, and its published thresholds are the developing and strong markers on the report. Two measures come from elsewhere and are labeled: support versus shift responses follow sociologist Charles Derber's research on conversational attention in The Pursuit of Attention, and discomfort tolerance is our own coaching lens, which the report says on its face.

So the definitions are the field's, the score's arithmetic is ours and it is printed, and nobody decides your score. It is counted.

A 100 means you gave the other person at least half of the words and at least 70 percent of your questions were open. Most first takes do not hit that, and that is fine: the MITI 3.1.1 manual puts beginning proficiency at 50 percent open questions and competency at 70. One conversation is one conversation. The other person shaped how it went more than anything else did, which is why there is no leaderboard and no comparing you to strangers. The comparison that means something is your own Take 1 to Take 2.

Step by step.

1. On your phone, the audio becomes a transcript with timings, using Apple's on-device speech model. Two voices are separated. You confirm which one is you. Full names become first name and last initial, or Speaker A and Speaker B if you choose anonymous mode.

2. The shortened transcript is read five separate times by the language model, independently, with the same fixed instructions each time. Each read returns a ledger: every turn coded, every question marked open or closed, every reflection marked simple or complex, plus three 1 to 5 judgments. Software checks each read the moment it arrives; a read that fails the check is thrown out and requested again.

3. Software on your phone merges the five ledgers by majority vote, counts every question with its own fixed grammar rules, and computes every count and ratio itself. The 1 to 5 judgments become the median of the five reads.

4. The model is asked once more to write the coaching prose: context, a strength, a growth edge, an experiment to try. Then the software overwrites every number in that prose with its own counted values, so no figure on the report comes from the model's arithmetic.

5. The report and a count of record are saved on your phone. Score the same conversation again and the app replays the saved count. Same numbers, forever, and no credit spent.

Consistency, measured and published. In July 2026 we re-scored the same calibration transcript three separate times from scratch and computed the run-to-run variance for every measure, the same statistic clinical laboratories publish for blood tests (the analytical coefficient of variation). Ten measures had zero variance, because they are counted rather than judged: share of talking, turn share, turns, longest turn, total questions, open questions, closed questions, and percent open, plus empathy and partnership, which did not move across the three runs. Five sat in single digits: reflections 5.2 percent, complex reflections 5.7, percent complex 3.8, reflection-to-question ratio 5.1, discomfort tolerance 7.5. Two were double-digit and are labeled as judgment on every report: unsolicited advice 32.7 percent and support responses 26.9. The report tells you which kind of number you are reading, every time.

A listening report is a box score, like the one printed after a ballgame. The game is your conversation. The stats are the things a careful observer could count from the tape: how much of the talking you did, how many questions you asked and whether they were open (what happened?) or closed (did it go okay?), how often you said back what you heard before moving on, how often you gave advice nobody asked for, and how often you turned the story back to yourself. A few stats need judgment, like how much empathy came through and whether you stayed with the hard moment, and those are shown as bands, the way a scout writes above average instead of inventing a decimal. Every read of the tape is done five times and the reads are merged by majority vote. Then a short coaching section: one strength, one growth edge, one experiment to try next time. Then you practice one thing, record again, and compare.

Three parts, and the middle one is the one most companies leave out.

What is solid. The behaviors Heard counts are the behaviors the listening research counts. The MITI coding system has been used and refined in published clinical research for two decades (Moyers et al., 2016). Automated coding of those same behaviors from transcripts is published science, not a startup's idea: Atkins et al., 2014, Imel et al., 2019, Flemotomos et al., 2022. Open questions pull more information out of the other person: in a study of 1,220 medical interviews, the use of open-ended questions predicted how much information patients gave (Takemura et al., 2005). And listening can be trained: a two-study field trial found listening training changed employees' listening and their sense of competence in difficult conversations (Itzchakov, 2020); the Annual Review of Organizational Psychology review by Kluger and Itzchakov (2022) summarizes the field.

What we will not claim. That a score predicts anything downstream. Not your promotion, not your marriage, not your sales. Nobody has shown that, and in the clinical literature, where these measures were born, a meta-analysis of 19 studies and 2,614 clients found that how counselors were rated on these behaviors was not significantly related to how their clients turned out (Pace et al., 2017). We measure what you did. We do not promise what it will get you. We have not run a controlled trial of Heard itself, and we will say so until we have.

What we promise instead. We do not promise you will be a better communicator. We promise you will know exactly what you did, and exactly what changed, measured on the recording and not on your opinion of yourself. The mechanism the app is built around is the one the training research keeps finding: feedback on real practice. In a randomized trial of 140 clinicians learning these skills, a workshop alone faded, while feedback on recorded practice, coaching, or both held the gains at four months (Miller et al., 2004). A meta-analysis of training studies found the skills erode without feedback and that three to four feedback sessions across six months sustained them (Schwalbe et al., 2014). Those studies used human coaches; Heard gives you the measurement half of that loop, on your own phone, for $3.99.

We think we are. The record says otherwise. Ralph Nichols, who ran the listening research program at the University of Minnesota, reported that right after listening to someone talk the average person remembers about half of what was said, and about a quarter two months later (Nichols and Stevens, 1957). Accenture surveyed 3,600 professionals in 30 countries: 96 percent rated themselves good listeners, and 98 percent said they multitask through their workday (Accenture, 2015). Listening is the largest single share of the day we spend communicating, about 24 percent in a time study of 680 college students, ahead of speaking, writing, and reading (Janusik and Wolvin, 2009). Count the classes you ever took in it. Zenger and Folkman's analysis of 3,492 managers found the best listeners are not silent sponges: they ask questions that make the other person think (Zenger and Folkman, 2016). That is a trainable behavior, and it is one Heard counts.

Motivational interviewing is a conversation method for helping people change, developed by William Miller and Stephen Rollnick, now in its fourth edition (Miller and Rollnick, 2023). We do not teach it, and Heard is not a motivational interviewing fidelity tool. We borrow its measurement system, the MITI, because it is the most mature published system for counting listening behaviors on a transcript: what counts as an open question, a simple reflection, a complex reflection, and how empathy and partnership are rated 1 to 5 (MITI 4.2.1). Its reliability has been studied, including where it is weak. Telling simple from complex reflections is the least reliable measure in the system even for trained humans, which is why the report shows that split as a soft band rather than a number (Atkins et al., 2014; Flemotomos et al., 2022). We also use the manual's own honesty: its authors write that the thresholds are based on expert opinion and lack normative validity data, and the report repeats that instead of hiding it.

On sales: no study we know of links motivational interviewing training to sales results, and we will not pretend one exists. What is true is that the behaviors the MITI counts are the behaviors sales research keeps finding in top performers. Neil Rackham's study of 35,000 sales calls found the best closers asked more questions of a specific kind (Huthwaite International). Gong's 2016 analysis of its customers' calls put the winning ratio at 43 percent talking to 57 percent listening; its 2025 re-analysis of 326,000 calls found won deals at 57 percent seller talk time against 62 percent for lost deals, found that sellers who lost asked more questions than sellers who won (about 20 a call against 15 or 16), and found that lost deals featured longer seller monologues. That is vendor data, not peer-reviewed research (Gong Labs), and it says the same thing the listening research says: it is not how many questions you ask, it is what kind, and it is how long you hold the floor. Heard counts exactly those things: your share of the talking, your longest turn, and the percentage of your questions that were open.

It will tell you what you did on the call, and it will tell you again after you changed something. Whether that makes you better at selling depends on you and your buyers, and we do not promise it. What we can say is what it counts and why a salesperson would care: your share of the talking, your longest turn, your open-question rate, how often you reflected before you pitched, and how many suggestions you made that nobody asked for. Gong's own data on 326,000 calls says the sellers who lost asked more questions and held the floor longer than the sellers who won (Gong Labs); Heard counts both. Record a practice discovery call with a colleague playing the prospect, score yourself, pick one number, and run it again in two weeks.

We do have our own server and our own software. Here is exactly who does what.

Your iPhone does the recording, the transcription (Apple's on-device speech model), the speaker separation (on-device models from FluidAudio), the name shortening, the counting, the merging, and the report. Our server, which we run, checks that a purchase is real, keeps the ledger of credits, and forwards the scoring requests. It stores no audio, no transcripts, and no reports. Anthropic's model, Claude, does the reading: it takes the shortened transcript and our fixed scoring instructions and returns the coded ledger and the written coaching sections. The current model is Claude Opus, through Anthropic's commercial API.

Why rent the reading instead of building it: a frontier language model costs tens to hundreds of millions of dollars in compute alone to train (the Stanford AI Index estimated $78 million for GPT-4 and $191 million for Gemini Ultra, reported by Fortune). Nobody should build one to score a conversation. We use Claude the way a lab uses an instrument it did not build, and we keep everything that touches your identity and your audio on your phone and off everyone's server, including ours.

What Anthropic gets: the shortened text and our instructions. What it never gets: your audio, your full names, or anything that identifies you, because we do not have your name, email, or phone number either. What Anthropic does with it: under its published policy, inputs and outputs from its commercial API are not used to train its models by default (Anthropic Privacy Center), and they are deleted from its systems within 30 days by default (Anthropic retention policy). We do not claim zero retention, because that would not be true.

No. Here is what leaves your phone and what does not.

Never leaves the phone: the audio, your full names, the report, and anything that identifies you. There is no account, no email, no phone number, no ads, no third-party analytics, and no trackers.

Leaves the phone once: the transcript text with names shortened to first name and last initial (or Speaker A and Speaker B), sent to our server, which passes it to Anthropic's commercial API for scoring and does not store it.

If you are thinking of the news about Anthropic, two things happened in 2025. It changed the terms of its consumer chat app so that consumer conversations may be used for training unless the user opts out (Anthropic, August 2025), and it paid $1.5 billion to settle a lawsuit over pirated books used to train its models (NPR). Neither touches Heard. Heard does not use the consumer chat app; it uses the commercial API, where inputs and outputs are not used for training by default and are deleted within 30 days by default (Anthropic Privacy Center).

Apple's privacy label for Heard on the App Store lists two categories of data that may be collected and not linked to your identity: purchase history and user content. That is the purchase receipt and the shortened text, exactly as described here. The full policy is at cdlab.io/heard-privacy. For your report to be all over the internet, someone would have to publish it, and the only person who can is you, from the share sheet, by your own tap.

Speaker separation runs on the phone and listens to the voices, not the words. The on-device models (FluidAudio, built on the open-source pyannote models) build a voiceprint for each speaker and assign every stretch of speech to one of the two. Then the app shows you sample lines from each voice and asks you to confirm which one is you before anything is scored, so a mislabeled line is caught by the person who knows.

The honest limits: it works best with two people, one phone between them, screen up, in a quiet room, one voice at a time. Crosstalk, a loud cafe, or a third voice degrade it. When the scoring finds a line that likely belongs to the other speaker, it does not quietly guess: the line is flagged, excluded from your counts and quotes, and disclosed in the report's reliability note. We would rather flag it than score it wrong.

You see your report. Nobody else does unless you share it. Reports belong to the person who was scored, they live on that person's phone, and the only way one leaves is the iPhone share sheet, operated by that person. There is no manager view, no partner view, no feed, no leaderboard. If your partner wants a report, your partner records a conversation as the listener and gets their own.

No. Heard trains nothing. The scoring runs through Anthropic's commercial API, and under Anthropic's published policy commercial inputs and outputs are not used to train its models by default (Anthropic Privacy Center). We never see your text, so we could not train on it if we wanted to.

The recordings, transcripts, and reports on that phone are deleted with it, so export any report you want to keep as a PDF from the share sheet first. Unspent credits are held on our server against an anonymous ID that lives in your iPhone's secure keychain rather than in the app's files, so they are still there when you reinstall on the same iPhone. They do not move to a different iPhone.

A report is not an app. It is a piece of work. Every report you buy is read six separate times by a frontier language model, five times to count and once to write, over your whole transcript each time, up to 30 minutes of conversation, and then counted, merged, and checked in software. We pay for that computation by the token every single time, and it is the largest cost in the business. The second largest is Apple's commission, 30 percent of every sale as standard and 15 percent for developers in Apple's Small Business Program. The app itself, the twelve lessons, the conversation prompt deck, and reopening any report you already ran are free.

Compare it to what else exists. The clinical tools that count these behaviors are sold to organizations, not people, with no retail price (Lyssn). Speech-coaching apps run about $8 to $30 a month whether you used them that month or not (Yoodli, Orai). One Heard report is $3.99, it never expires, and you keep it.

One conversation of up to 30 minutes, scored: the Listening Score, the counted measures (share of the talking, turns, longest turn, questions asked, open versus closed, interruptions), the careful reads (reflections, simple versus complex, reflection-to-question ratio, unsolicited advice, support versus shift), the coaching bands (empathy, partnership, discomfort tolerance), your own words quoted back as evidence, a written strength, growth edge, and experiment, and a PDF you can keep and share. Plus the Take 1 to Take 2 comparison whenever you score a second conversation. The report stays on your phone and reopens free forever.

Because you should not pay for a listening report in a month you did not have a conversation worth scoring. Credits are consumable: buy one, five, or ten, spend them when you record, and they sit there until you do. No renewal, nothing to cancel.

Because every report costs us the same to produce, and we would rather sell the tenth report at an honest price than sell the first at an inflated one. One credit is $3.99, five are $19.99, ten are $39.99. The packs buy convenience, not a markdown. Apple's fixed price tiers are why the packs land a few cents over an exact multiple rather than on it.

They do not expire. A credit is spent only when you score a new conversation. Two things to know before you buy a ten pack: credits are tied to the iPhone that bought them, held against an anonymous ID in that phone's secure keychain, because there are no accounts; and Apple treats them as consumables, which by Apple's own rules cannot be restored to another device (Apple Support). Unspent credits do not move to a new iPhone. Reports you already scored reopen free on the phone that holds them, and they ride along in your iPhone backup like the app's other files.

Yes, through Apple, which handles every purchase and every refund. Go to reportaproblem.apple.com, sign in with the Apple Account you bought with, choose I'd like to, then Request a refund, choose the reason, choose the Heard credit pack, and submit. Apple says to allow 24 to 48 hours for an update, and you can check the status on the same site. Apple decides refunds, not us. When Apple grants one, our server removes the refunded credits from your balance; if you already spent them, you keep the reports.

If you paid and the credits did not appear, that is not a refund case. Reopen the app. Apple holds the transaction open until our server confirms the credit, and the app retries on every launch, so a paid pack cannot be lost.

The hard conversation at home. One of you says the thing that has been sitting there for a month. The other does nothing but listen, and gets scored. The prompt deck in the app has sixteen openers for exactly this, under Bring the issue.

The argument about who listens. Score yourself on one conversation, them on the next, same counts both directions, and read the numbers instead of relitigating the memory.

The discovery call. Before the real one, run a fifteen-minute practice call with a colleague playing the prospect. Score yourself. Look at your share of the talking and your open-question rate.

The one-on-one. A manager records a real one-on-one with a direct report who agrees, and the manager is the one scored.

Anyone whose job is to ask and listen. Coaches, recruiters, teachers, clergy, journalists, clinicians in training: record a practice conversation with a willing partner and see the counts.

Parent and teen. Start light. The deck has openers like What did a Saturday look like when you were ten. The parent listens.

The lab room. In a Communication Dynamics Lab workshop, pairs record one conversation and every participant leaves with a report and one thing to practice.

Not as designed, and we will not pretend otherwise. Heard records through the iPhone's microphone with the phone between two people in the same room. There is no way to import a Zoom recording or a phone call, and that is deliberate: the consent step and the on-device pipeline are built around a live recording the app made itself. If you set the phone next to a laptop speaker on a call, it will record what the microphone hears, but we did not build for that, the speaker separation was built for two live voices in a room, and we cannot promise the counts. If the conversation matters, have it in the room.

One measure, two weeks, real conversations. That is the method the app teaches in its last lesson, and it is the whole thing. Read the report, pick the one measure that stung the most or matters most to how you want to be in a room, and practice only that. Ten measures practiced at once is a performance. One measure practiced for two weeks is a change.

Not every day. Record once, practice for a stretch of ordinary conversations with your one thing in mind, then record a fresh conversation and score it. The Take 1 to Take 2 view shows the point and percent change on every measure. If it moved, keep going or pick your next measure. If it did not, change the drill, not the conclusion about yourself.

How to read a report in one minute. The score, then the growth edge, then the experiment. Skip everything else on the first pass. Come back to the counts when you want to see the evidence.

Same prompt or a different one: it does not matter. The report measures how you listened, not what you talked about, so use whatever conversation you were going to have anyway. When you tap Record take two after a report, the app links the two takes so the comparison is automatic.

How big a change is real. Judgment measures carry scatter, and we publish how much. From our July 2026 precision runs, a change smaller than about 14 percent in your reflection count, 11 percent in your percent complex, 14 percent in your reflection-to-question ratio, or 21 percent in discomfort tolerance is within the instrument's own scatter, and support responses and unsolicited advice need much larger moves before we would call the change real. The exact counts (share of talking, turns, questions, open versus closed) carry no scatter at all: if they moved, you moved.

It can be used badly, the way a bathroom scale can. Four things in the product push the other way. Both people have to agree to the recording before the app will score anything. The report goes to the person who was scored, and it leaves their phone only by their own tap. The counts are the same in both directions, so the fair use is to score yourself on one conversation and them on the next, and the buyer risks their own number too. And the report ends with one thing to practice rather than a verdict; the Take 1 to Take 2 view exists so the second conversation goes better than the first. The people who get the most out of Heard read their own number first.

Then you do not record, and the app will not score. No is an answer, and passing on a prompt is an answer worth respecting; the app says so on the prompt screen. Two things that often change a no: offer to be the one who is scored, so the recording is about your listening and not theirs, and start with one of the light prompts instead of the hard thing.

Stop the recording whenever you want; one tap. Nothing is scored until you choose to score it, and the recording stays on your phone either way. If you do score it, the report will show what each of you did with the floor, and the discomfort tolerance panel will show whether the listener stayed with the hard moment or moved away from it. Heard is a communication assessment for capable adults. It is not therapy and never a diagnosis, and it is not built to referee a fight in progress.

No. Heard is a communication assessment for capable adults. It is not therapy, not counseling, not a diagnosis, and not a mental health service, and every report says so in its honest note. It counts what a listener did in one conversation and offers one thing to practice. If you need a therapist, see a licensed one.

Yes, with everyone's agreement, out loud, before you press record, and within your employer's own policy on recording. Managers and salespeople are exactly who it was built for. Do not record a client, a patient, or a customer unless they have agreed and your profession's rules allow it. Reports stay with the person who was scored; there is no manager view and nothing is shared unless that person shares it.

Yes. The app is rated 4+, the prompt deck starts light, and the parent is usually the one who should be scored. Your child has to agree to the recording like anyone else, and the recording stays on your phone.

It will not say that, because it does not grade people. It shows what one listener did in one conversation, with the words quoted back, and the person on the other side of the table shaped how it went more than anything else did. A light conversation with no hard moment prints Untested on the discomfort panel rather than a low mark, because nothing hard came up. What the report does give you is one growth edge and one experiment. The people who improve fastest read that, pick one thing, and record again in two weeks.

One credit scores one conversation of up to 30 minutes, and the recorder stops itself at 30:00, with a countdown through the last five minutes so nobody is surprised. Thirty minutes is more conversation than most people will sit through in one take, and the reading cost rises with every minute, so the cap is what lets one report cost $3.99. If you have more to say, stop at thirty, score it, and start a fresh recording with a second credit.

Michael Cinquino. He enlisted in the U.S. Navy at 17 and became a Rescue Swimmer on active duty. He trained as an actor at Rutgers in the William Esper and Meisner tradition and holds an MFA in acting. He worked for a decade producing the CIO Strategy Exchange, a closed-door community of Fortune 70 chief information officers. He has taught communication at the University of New Hampshire's Paul College of Business for five semesters, and he is completing a Master's in clinical mental health at Northwestern University. He founded SoHo Creative Studio and the Communication Dynamics Lab, and he built Heard to put the lab's scored conversation on a phone. Rooms he has taught in include BAE Systems, Red Hat, Blockchain Capital, Soho House, Grand Central Tech, and UNH.

Do not believe us. Check us. Every research claim in this document links to its source. Every price and policy claim links to Apple's or Anthropic's own page. The report prints its own limits: which numbers are exact, which are careful reads, which are judgment, and where its benchmarks come from. We publish our run-to-run variance, which nobody else in this category does that we know of. And we tell you what we will not claim: no promise that a score predicts anything. The whole product rests on one sentence. We do not promise you will be a better communicator. We promise you will know exactly what you did, and exactly what changed, measured on the recording and not on your opinion of yourself.

How To Talk So People Listen. (And Listen So People Talk.) The Communication Dynamics Lab is a one-day workshop that combines the art of acting with the science of psychology. Participants learn the same techniques the world's greatest stage and screen actors use to captivate an audience, the psychology of why they work, and how to use them in the next meeting, the next one-on-one, and the next hard conversation at home. Then they practice with a partner until the skills are theirs. The lab has its own app that scores your listening, and you are reading its FAQ. Details and dates are at cdlab.io.

Email michael@michaelcinquino.com and a person answers. For purchases and refunds, Apple is the place: reportaproblem.apple.com. The privacy policy is at cdlab.io/heard-privacy.