Every paper, venue and book underneath this curriculum — including a section for what is popular and does not hold up.
Every entry below was checked against a primary source. Where a claim is popular but not verifiable, it is listed under Discredited or unsupported rather than quietly omitted — knowing what to ignore is half the value here.
Huang, Yeomans, Brooks, Minson & Gino (2017), "It Doesn't Hurt to Ask: Question-Asking Increases Liking," JPSP 113(3), 430–452 The single most load-bearing paper in this workspace. Contains the six-type question taxonomy and the finding that only follow-up questions raise liking — full-switch questions (27.6% of all questions people ask) actively destroy perceived responsiveness, and mirror questions do nothing at all. Also: people ask a mean of 6.72 questions in a 15-minute conversation and have no idea questions make them likeable. Use for: the depth module, and any time you're tempted to teach "ask questions" without qualification.
Brooks & Yeomans (2025), "Boomerasking: Answering Your Own Questions," JEP: General 154(3), 864–893 The definitive "adequate conversationalist" failure mode: asking a question as a launchpad for your own answer. Feels great to do (median 82/100 preference), reads as insincere, and underperforms plain self-disclosure on second-date acceptance (56% vs 71%). Use for: the sharpening pass — this is what someone already "decent" is most likely to be doing.
Collins, Minson, Kristal & Brooks (2024), "Conveying and Detecting Listening During Live Conversation," JEP: General 153(2), 473–494 Five studies, N=1,225: people cannot tell whether they are being listened to. Minds wander 24% of the time and partners can't detect it. This is the counterweight that stops the workspace from teaching performative "active listening" theatre. Use for: justifying why the verbal/reflective components of listening are trainable and the nonverbal performance layer largely isn't.
Kardas, Kumar & Epley (2022), "'Overly shallow?': Miscalibrated expectations create a barrier to deeper conversation," JPSP 122(3), 367–398 People overestimate how awkward a deep conversation will be by d=2.23 — roughly four times the awkwardness they actually experience. Use for: the depth module, and as evidence that the discomfort signal is a forecasting error, not information.
Yeomans, Minson, Collins, Chen & Gino (2020), "Conversational receptiveness," OBHDP 160, 131–148 The most directly trainable recipe in the literature, at word level: positive phrasing over negations, explicit acknowledgement ("I understand," "I see your point"), naming points of agreement, and hedging. Teachable, with β=0.57 on perceived receptiveness. Use for: disagreement, and for the professional-presence half of the mission.
Craske, Treanor, Conway, Zbozinek & Vervliet (2014), "Maximizing exposure therapy: an inhibitory learning approach," Behaviour Research and Therapy 58, 10–23 The paper that reframes exposure away from "stay until the fear drops" toward expectancy violation. Eight optimization strategies: expectancy violation, deepened extinction, occasional reinforced extinction, removal of safety signals, variability, retrieval cues, multiple contexts, affect labeling. Use for: the entire architecture of the anxiety module — and specifically for why felt-calm must never be a success metric here. See the learning record “the block is somatic”.
Centre for Clinical Interventions, "Stepping out of Social Anxiety" (Government of Western Australia, Dept. of Health) Ten free modules. Module 4 is behavioural-experiment stepladders; Module 5 is safety behaviours; Module 6 is attention. Government-published, free, and follows the Clark & Wells model. The best value-per-dollar resource in this entire document. Use for: the structured version of what the pickup industry sells for £7,000.
Sandstrom, Boothby & Cooney (2022), "Talking to strangers: A week-long intervention reduces psychological barriers to social connection," JESP 102, 104356 — OSF preregistration · original mission list Full protocol now extracted — see The 29 Missions. Five days Mon–Fri, one stranger minimum per day, delivered as a 29-mission scavenger hunt via the GooseChase app. 286 analysed, 1,336 conversations. Outcomes: expected rejections 0.89 → 0.23 (ratio 3.82, p<.001, still 2.81 a week later); conversational ability d=0.37 and still rising at follow-up; awkwardness d=0.33; enjoyment d=0.34; positive impression d=0.38. Controls didn't move. The mechanism data is the valuable part: day 1, only 40% thought the first person would talk to them — 92% did. 87% of all 1,336 conversations happened with the first person approached. Predicted rejections fell steadily; actual rejections never changed (p=.37) — near-zero throughout. The intervention calibrates belief to an already favourable reality rather than teaching skill. Authors explicitly conclude a shorter version would not work — repeated experience is the active ingredient. Caveats: cohort-level not individual assignment; 30% of starters excluded; students averaging 20, 209 of 286 women; several treatment-vs-control contrasts ns.
Epley & Schroeder (2014), "Mistakenly seeking solitude," JEP: General 143(5), 1980–1999 Chicago commuters told to talk to a stranger had the most positive commute (d=0.63), with no productivity cost — while a matched sample predicted the exact opposite (d=0.41 in the wrong direction). Use for: the expectancy-violation framing. His prediction about how the approach will go is measurably, reliably wrong.
Boothby, Cooney, Sandstrom & Clark (2018), "The Liking Gap in Conversations," Psychological Science 29(11), 1742–1756 People consistently underestimate how much a new acquaintance liked them. Persisted across an entire academic year among suitemates. Larger for shyer people. Use for: the post-approach debrief — his read on how it went is biased downward and he should not act on it.
Mastroianni, Gilbert, Cooney & Wilson (2021), "Do conversations end when people want them to?" PNAS 118(10) 932 conversations. They almost never end when both people want. Two-thirds of people wanted out at some point; the average gap between desired and actual ending was ~half the conversation's length. Use for: the graceful-exit drill. Leaving early is a gift, not a rejection — this is the paper that makes exits emotionally cheap.
Antonakis, J., Fenley, M. & Liechti, S. (2011), "Can Charisma Be Taught? Tests of Two Interventions," Academy of Management Learning & Education 10(3), 374–396 (free full PDF, UNIL repository) The only randomised trial of charisma training worth citing. Study 1: 34 Swiss middle managers, randomised, control group, 360° ratings from 343 raters, d≈.47. Study 2: 41 MBA/EMBA students, within-subjects, 135 blinded raters, d≈.75. Weighted mean d = .62, comparable to leadership training generally (d=.60). Teaches 12 Charismatic Leadership Tactics — 9 verbal (metaphors; stories/anecdotes; contrasts; rhetorical questions; three-part lists; expressions of moral conviction; reflecting the sentiments of the collective; setting high goals; communicating confidence goals can be achieved) and 3 nonverbal (animated voice tone; facial expressions; body gestures). Use with care. Two studies from one lab, Swiss managers and MBA students, dose of 30–90 hours. One independent replication of Study 2 (N=140) appears in press as of 2026 but its result could not be confirmed — claim nothing either way. There is no meta-analysis of charisma training; too few independent trials exist to pool. And roughly half the tactics are podium-only — three-part lists, rhetorical questions and contrasts read as performed in dialogue; setting high goals and conveying confidence are the structured-work register he already has. See Charisma Baseline for the sorted version.
Antonakis, Bastardoz, Jacquart & Shamir (2016), "Charisma: An Ill-Defined and Ill-Measured Gift," Annual Review of Organizational Psychology and Organizational Behavior. Antonakis's own critique of the field, arguing the construct remains poorly defined and much questionnaire-based charisma research "cannot usefully guide policy." Use for: calibrating how hard to lean on the above. Unusually candid self-scrutiny.
Tskhay, Zhu, Zou & Rule (2018), "Charisma in Everyday Life: Conceptualization and Validation of the General Charisma Inventory," JPSP 114(1), 131–152 Six items, two factors, distilled from ~408 candidates. Influence (α .77–.85): has a presence in a room · has the ability to influence people · knows how to lead a group. Affability (α .75–.82): makes people feel comfortable · smiles at people often · can get along with anyone. Use for: the before/after outcome measure — implemented as a scored instrument in Charisma Baseline. Works as an other-rating too, which is worth more than self-report given the liking gap.
Fiske, Cuddy & Glick (2007), "Universal dimensions of social cognition: warmth and competence," Trends in Cognitive Sciences 11(2), 77–83 Warmth ≈ liked/trusted; competence ≈ respected/capable. The primacy of warmth: it is judged first and weighted more heavily, because assessing intent precedes assessing capability. Use for: the strategic point that competence is what he's already optimised and the dimension strangers weight less.
Added after the diagnosis was corrected — see the learning record “the scaffold is purpose”. This is the cluster that was missing entirely from v1.
Sacks, H., Schegloff, E. A. & Jefferson, G. (1974), "A Simplest Systematics for the Organization of Turn-Taking for Conversation," Language 50(4), 696–735 The formal grammar of turn-taking. At each transition-relevance place: (1) current speaker may select next; (2) if not, anyone may self-select — first to speak gets the turn; (3) otherwise current speaker continues. Use for: the belief that contributing to a group is "interrupting." Self-selection is Rule 2, a specified right. Paywalled at JSTOR but widely mirrored as free PDF on university course pages.
Stivers, T. et al. (2009), "Universals and cultural variation in turn-taking in conversation," PNAS 106(26), 10587–10592 Ten languages: mean gap +208ms, modal gap 0–200ms, unimodal near zero everywhere (fastest Japanese +7ms, slowest Danish +469ms). Use for: the single most useful number in the workspace — the clear opening he's waiting for does not exist, so "wait for a gap" structurally guarantees never speaking.
Schegloff, E. A. (2000), "Overlapping Talk and the Organization of Turn-Taking for Conversation," Language in Society 29(1), 1–63 Distinguishes terminal overlap (starting just before a projectable turn-end) from competitive interruption. Most talk that feels like interrupting is two people correctly anticipating the same transition point. Use for: reducing the cost of starting slightly early.
Schegloff, E. A. & Sacks, H. (1973), "Opening up closings," Semiotica 8(4), 289–327 The pre-closing sequence: "well…", "okay…", "so…", or a topic-final summary, which offers the other party the choice to close or add one more thing. Use for: the exit formula — token + forward-looking reason + half-beat. The direct answer to both "someone's monopolising me" and "they turned away and I'm stranded."
Kendon, A. (1990), Conducting Interaction: Patterns of Behavior in Focused Encounters, Cambridge University Press. F-formations. A standing group jointly maintains a protected o-space (empty centre), each person claims a p-space, and the r-space is the surround. A newcomer is admitted when the group physically reconfigures — a rotation or half-step opening a gap. Use for: the observable signal for group entry. The reconfiguration is the invitation and arrives before anyone speaks.
Barua, H. B. et al. (2020), "Let Me Join You! Real-Time F-Formation Recognition by a Socially Aware Robot," IEEE RO-MAN Operationalizes Kendon into real-time detection plus correct approach angle. Use for: the argument that group-joining is a specifiable procedure, not an innate gift.
Malinowski, B. (1923), "The Problem of Meaning in Primitive Languages" Coined phatic communion — language whose function is social bonding, not information transfer. Use for: why judging an opener on how interesting it is applies the wrong criterion entirely.
Trinh, E., Thio, N. & Klein, N. (2026), "Conversations About Boring Topics Are More Interesting Than We Think," JPSP 130(6), 1167–1189 Nine preregistered experiments, ~1,800 participants. People consistently underestimate enjoyment of mundane topics; they over-weight topic choice and under-weight interactional dynamics, and so decline conversations they'd have liked. Use for: the measured version of his "that's too lame to say" veto. Note: headline finding verified; the specific sub-claim that it holds even when both parties pre-agree the topic is boring was not confirmed against full text. Different research team from the Brooks/Yeomans/Cooney cluster — popular coverage conflates them.
Goffman, E. — Behavior in Public Places (1963), Forms of Talk (1981). Civil inattention — the respectful non-engagement that is the default between strangers — is well attested. Caveat: the "open persons / open situations" terminology could not be confirmed in these two titles and may belong to Relations in Public (1971). Use the framing; don't cite the terminology until checked.
Cooney, Gilbert & Wilson (2017), "The Novelty Penalty," Psychological Science 28(3), 380–394 Speakers pick the novel story 66.7% of the time and it lands worse (d=1.10, a full reversal). Familiar stories let listeners fill in gaps. Use for: killing the "be interesting / have great stories" instinct.
Gable, Reis, Impett & Asher (2004), "What do you do when things go right?" JPSP 87(2), 228–245 The four-quadrant response to good news. Only active-constructive predicts relationship quality — and the polite, mild "oh, nice" (passive-constructive) patterns with the destructive responses, not the good one. Use for: the single easiest high-leverage upgrade for someone already competent.
Bitterly, Brooks & Schweitzer (2017), "Risky Business: When Humor Increases and Decreases Status," JPSP 112(3), 431–455 Successful humor raises status via perceived competence and confidence. Inappropriate humor raises confidence but tanks competence, and nets out worse than just being serious. The risk that matters is appropriateness, not funniness. Use for: levity, carefully.
Bavelas, Coates & Johnson (2000), "Listeners as co-narrators," JPSP 79(6), 941–952 Distinguishes generic listener responses (nodding, "mhm") from specific ones (wincing, exclaiming, timed to content). Distracted listeners drop the specific ones and the narrator measurably tells the story worse. Use for: the listening drill — this is what to train instead of eye-contact theatre.
Collins & Miller (1994), "Self-Disclosure and Liking: A Meta-Analytic Review," Psychological Bulletin 116(3), 457–475 94 effect sizes. Disclosure→liking is real but modest (d≈0.27 in well-controlled experiments) and negative in public/field settings with strangers. Use for: calibrating how hard to push depth with strangers specifically.
Aron, Melinat, Aron, Vallone & Bator (1997), "The Experimental Generation of Interpersonal Closeness," PSPB 23(4), 363–377 The "36 questions." d=0.88 on closeness vs small talk. Scope limit in the authors' own words: the goal was "a temporary feeling of closeness, not an actual ongoing relationship." The viral "fall in love" framing is journalism, not science. Use for: escalation structure — the ordering principle, not the promise.
ASHA, Accent Modification Practice Portal The professional body's position, and useful authority for the framing here. Accents are not a communication disorder. ASHA explicitly discourages "accent reduction" and "accent elimination" as "inaccurate and stigmatizing… because every speaker has an accent" — preferred framing is accent modification / intelligibility enhancement. A recognised elective service, not remediation of a disorder. Note: on outcomes ASHA says only that "evidence exists that supports its efficacy" — vague, no effect sizes. Use for: checking the language any prospective coach uses.
Schwarz, Cielo, Spritzer et al. (2023), "A speech therapy for transgender women: an updated systematic review and meta-analysis," Systematic Reviews 12(1), 128 The strongest evidence base anywhere for deliberately changing habitual pitch, because it's the population with the most motivation and clinical support. 16 studies. Speech therapy alone: 39 Hz in reading but only ~25 Hz in spontaneous speech. Use for: two hard facts — modest gains are the ceiling even with months of supervision, and trained pitch transfers worst to live conversation, the only setting that matters here. Caveat: the entire evidence base concerns pitch elevation; extrapolating to deliberate lowering is an inference.
Barsties v. Latoszek, Watts & Neumann (2020), "The effectiveness of voice therapy on voice-related handicap: A network meta-analysis," Clinical Otolaryngology Network meta-analysis restricted to RCTs. Resonance and projection work has RCT evidence that pitch-lowering advice does not. Named, manualised methods: semi-occluded vocal tract exercises (straw phonation), Vocal Function Exercises, resonant voice therapy. Caveat: evidence comes from treating dysphonia in clinical populations; transfer to a healthy voice is an extrapolation — but far better grounded than the commercial tier.
Roger Love and the commercial vocal-coaching tier — not recommended. Credibility rests entirely on celebrity association. Zero scientific studies, zero peer-reviewed publications, zero clinical credentials cited. Sells a trademarked "VQ™ Vocal Intelligence" assessment with no published validity or reliability evidence — inventing and trademarking a proprietary unvalidated metric is a marketing device, not a psychometric one. An experienced coach can genuinely help, and repeat celebrity clients are a market signal; but a market signal cannot distinguish skill from fame. An ASHA-certified SLP doing resonance work costs similar money against an actual clinical literature.
Vacharkulksemsuk et al. (2016), "Dominant, open nonverbal displays are attractive at zero-acquaintance," PNAS The one body-language finding with a clean real-world effect. Postural expansiveness raised odds of a speed-dating "yes" by 76% per SD (OR=1.76); open-posture photos got 27% more yeses across 2,983 swipe decisions. Affiliation cues predicted nothing. Use for: the only posture instruction worth giving — take up the space you have.
Binetti, Harrison, Coutrot, Johnston & Mareschal (2016), "Pupil dilation as an index of preferred mutual gaze duration," Royal Society Open Science 3(7) Mean preferred mutual gaze 3.3 seconds, consistent with earlier dyadic estimates (2.95s, 4.66s). Note: no SD is given in running text — don't invent one. Use for: calibrating gaze away from both the unbroken lock and the flicker.
Clark, H. H. & Fox Tree, J. E. (2002), "Using uh and um in spontaneous speaking," Cognition 84(1), 73–111 Fillers are words, not accidents: uh announces a minor delay, um a longer one. Um was followed by an actual pause 61% of the time vs 29% for uh, and comprised 56% of fillers at sentence boundaries vs 37–39% within sentences. Use for: the reason "eliminate all fillers" is the wrong target. See the learning record “fillers are not noise”.
Azrin, N. H. & Nunn, R. G. (1973), "Habit-reversal: A method of eliminating nervous habits and tics," Behaviour Research and Therapy 11(4), 619–628 Three components: awareness training → competing response → social support. The best-evidenced framework available for the fluency track. Caveat: validated for tics and Tourette's (strong later RCT support via CBIT); applying it to filler words is a defensible extrapolation, not something the authors tested.
Lev-Ari, S. & Keysar, B. (2010), "Why don't we believe non-native speakers? The influence of accent on credibility," JESP 46(6), 1093–1096 Statements read in a non-native accent are rated less true, with the penalty growing with accent strength — but telling listeners the reduced clarity came from a bad phone line shrank the penalty. The mechanism is processing fluency, not accent identity. Use for: the reason to train clarity rather than accent reduction. (Citation verified; substantive finding not re-pulled from primary text this session — both routes were bot-blocked.)
Klofstad, C. A., Anderson, R. C. & Peters, S. (2012), "Sounds like a winner: voice pitch influences perception of leadership capacity," Proc. R. Soc. B 279(1738), 2698–2704 Listeners preferred the lower-pitched version of the same recording, for male and female speakers alike. Critical caveat: this was a clean digital pitch shift, not a person straining downward. Use for: why deliberately forcing a lower voice is not the implication people draw from it.
Bögels, "Task Concentration Training," Common Language for Psychotherapy (free PDF) Author-written protocol for redirecting attention outward when self-focus takes over. The trainable core: after each situation, log % attention on self / on task / on environment (must total 100). Use for: the in-conversation attention drill. Evidence caveat: the foundational 1997 paper is two case studies; the 2006 RCT (N=65) showed TCT's advantage over applied relaxation was domain- and time-specific and vanished once cognitive therapy was added. Solid on process measures (self-consciousness, self-focused attention), weaker on clinical outcome than usually claimed.
Wells, "Attention Training Technique," Common Language for Psychotherapy (free PDF) The ~11-minute auditory attention exercise (5 min selective / 5 min switching / 1 min divided, across 5–7 simultaneous sounds). Use for: optional attention-control work. Evidence caveat: weak. The social-anxiety evidence base is one case in Wells, White & Carter (1997), an uncontrolled crossover (Vogel et al. 2016), and a controlled trial (McEvoy & Perini 2009) that found no incremental benefit over group CBT alone. No random-effects meta-analysis of ATT exists. Treat as optional, not core.
Verified live on 31 July 2026. Language exchange nights are ranked first deliberately: they are the only venue type where talking to total strangers is the explicit, advertised purpose — which drops the rejection cost of an approach to approximately zero and removes all ambiguity about whether you're welcome. Nothing else on this list matches that.
Barcelona Language Exchange — 57,113 members. The strongest single venue in this entire document. Schedule published on their own site and verified 4 Aug 2026, with no August-break notice (this was checked specifically — August is a dead month in Spain and the risk was real):
| Night | Venue | Time |
|---|---|---|
| Mon & Thu | Space Cowboy, c/ Carders 31 (Born) | 19:00 |
| Tue | Jardinet d'Aribau, c/ Aribau 133 (Eixample) | 19:00 |
| Wed | Soda Bus, c/ Aribau 150 (Eixample) | 19:00 |
| Fri | Estació de França bar, Av. Marquès de l'Argentera 6 (Born) | 20:00 |
| Sat | Trafalgar Pizza Club, c/ Trafalgar 19 (Born) | 20:00 |
| Sun | Occasional yoga & brunch | 1–2×/month |
€4 minimum consumption weekdays, €5 Friday. Six nights a week, near-free, zero commitment, effectively unlimited stranger reps. Friday occasionally changes venue — confirm that day. Note: Meetup's events tab renders client-side and reads as empty via automated fetch; the group is active. Use the site above, not the Meetup listing.
Toastmasters, Barcelona — 15 clubs. Barcelona Toastmasters (Wed 7:30pm), 22@ Barcelona (Thu 6:45pm), BCN Pride (Thu 7:00pm), SOM-HI (2nd/4th/5th Tue 7:30pm). Most central clubs run in English given the international membership, but confirm language on the club's own page — the finder's list view doesn't show it.
Barcelona Improv Group — the city's dominant
English-language improv company, running since 2011. Correction: there is no standing
"Improv Jam" — the /jam page 404s because it isn't an active offering. The recurring
participatory drop-in is the Sunday Open Workshop, 17:00–19:00, €10, capped at 18,
new theme and instructor weekly, explicitly beginner-friendly ("from acting hot shots to
total newbies"). Verified 4 Aug 2026.
⚠️ Could not confirm it runs during August — their Term 3 starts 14 September and
course registration showed as closed. Email classes@barcelonaimprovgroup.com or
WhatsApp +34 633 066 986 before turning up.
Their festival runs 3–5 Sept 2026 at Calabria 66 (Mainstage €18, Headliner €12, Late
Night €7, 3-day pass €55) — a cheap way to meet the scene. Beginner course when wanted:
Foundations of Improv 1, €150, 6 weeks, Thursdays 19:30–22:00.
Their ticketed mainstage shows are watching, not doing — wrong product for this purpose.
Free English walking tours — Runner Bean Tours, Walkative!, SANDEMANs, GuruWalk. Multiple daily departures, pay-what-you-want. The lowest-stakes venue of all: everyone present is a stranger, talking is expected, and bailing costs nothing.
Every Toastmasters meeting includes Table Topics: a surprise prompt answered impromptu for 1–2 minutes. That is initiation-latency training with a stopwatch on it — you cannot prepare, and the window is externally enforced. Meetings also assign an Ah-Counter, a member whose entire job is tallying filler words and reporting the count back. That is a free, human-delivered version of exactly the measurement the fluency module needs. Guests attend free, so visiting costs nothing but an evening.
Toastmasters International — club finder The best cost-per-exposure option that exists. ~$72 USD per six months (rose from $60 on 3 Aug 2026) plus a $25 one-time joining fee, with prorated entry. Guests attend free — visit three or four clubs before joining. Two segments matter for this mission: Table Topics (a surprise question, answered impromptu for 1–2 minutes — direct initiation-latency practice) and the Ah-Counter role, where a member literally counts every filler word spoken in the meeting. Online-only clubs are officially supported and run the identical program. Honest caveat: club quality varies a lot and feedback culture skews encouraging over corrective.
Mind — Side by Side Free, 24/7, actively moderated by Mind staff under published guidelines. The best-moderated peer community verified. Peer support, not method.
Gillian Sandstrom — University of Sussex The legitimate academic version of "talk to more strangers." Runs How to Talk to Strangers workshops; her book is Once Upon a Stranger. Same behavioural prescription as the pickup industry, with an actual evidence base and none of the ideology.
Yoodli — free tier (5 lifetime sessions), Pro $8/mo annual. The only speech tool where the measurement set was verified: filler words, pacing, talk-to-listen ratio, vocabulary diversity. Use for: the objective filler-word count in the fluency module. This is a measurement instrument, not a teacher.
THIS WAY UP — Social Anxiety Program AUD $59 self-guided, six modules over 90 days. St Vincent's Hospital / UNSW. Only worth it if the free CCI workbook proves insufficient. Internet-delivered CBT for social anxiety has genuinely strong RCT support (g≈0.55 vs controls, g≈0.79 vs waitlist; one meta-analysis found remote CBT outperformed face-to-face, g=0.34).
All publication facts verified 4 Aug 2026. Books are the lowest-leverage item in this workspace — the venue reps matter far more. This ranking mostly exists to decide which one to buy.
Honest caveat on 5, 7 and 8: search engines were blocked during verification, so those three verdicts are reasoned from publication facts, author credentials and the state of the literature — not from cited published critiques.
Listed explicitly because each one is widely repeated and will be encountered.