Apology and Repair Assessment for Everyday Work and Personal MistakesSorry is the easy word. Find out which part of the repair you leave out.

Methodology: Thirty-six original exercises across five formats, interleaved: nine forced-choice blocks in which four things a person might say or do after a mistake are marked as the one most like you and the one least like you; eight single-choice exercises in which four messages are on offer and one is sent; six keep-or-cut decisions on a single draft line; five select-every-line exercises; and eight first-person agreement statements on a five-point scale, two per part, one keyed forward and one in reverse, the reverse item at least four positions from its twin. Every option is something a decent person would plausibly say. The distractors are the named failure modes of apologising, drawn from the literature and recorded on every option in the seed: the conditional apology ('sorry if you felt'), the passive ('mistakes were made'), the excuse attached, the counter-complaint, minimisation, the explanation offered in place of the acknowledgement, payment offered in place of the repair, the repair that is cheapest to give, the grand gesture that takes the choice away, the vague promise, the over-promise, the system blamed in place of the change, the over-apology that makes the other person comfort you, the request to be forgiven, the repair delivered before the other person has spoken, and the repair left until it has blown over. No scenario involves abuse, violence, self-harm, an eating disorder, a clinical matter or a legal matter; every mistake is ordinary. Construct statement: this measures which of four components of an effective repair a person includes in what they say after an ordinary work or personal mistake, in thirty-six written situations in one sitting, namely owning it plainly (naming the specific harm and taking responsibility with nothing attached), offering the repair the other person needs rather than the one that is cheapest to give, saying specifically what changes, and timing and restraint (apologising now, briefly, and not asking the other person to manage your feelings or to say it is all right). It does not measure whether the person is good, kind or sincere, whether they are forgiven, whether the other person was right, how they behave in a real conversation with a real voice and a real face, or anything about their character; it is not a relationship or couples instrument, it diagnoses nothing, and no exercise asks about any personal circumstance. Declared response instruction, one for the whole instrument: behavioural tendency, what the respondent would actually say, send, keep, cut or do; the agreement statements ask what the respondent generally does. Scoring design C13, the reliability-weighted composite: each part is scored on its nine exercises as a chance-corrected quality, where zero is a respondent answering the way people typically answer these exercises, from a prior declared on every option (the share expected to mark each option most and each option least on a forced-choice block, the share expected to choose each message, the share expected to keep each draft line, the share expected to tick each line of a select-every exercise, and the share expected at each point of the agreement scale, which clusters above the midpoint and is never a midpoint null). Forced-choice blocks score half for each pole; select-every exercises score balanced accuracy, keyed lines kept and unkeyed lines left, so ticking everything and ticking nothing both land below the typical respondent; the agreement statements are read by their position on the item's own scale, honouring the reverse keying, against the expected position under the declared response prior, and are never marked right or wrong. McDonald's omega, not alpha, is estimated for each part from its answered count and an assumed average inter-item correlation of .27, printed unrounded and compared with .70 unrounded; the four omegas decide the weighting of the composite by one visible rule, printed on the report in words: weights proportional to omega only when the spread between the highest and lowest omega exceeds .15, and unit weights otherwise, because differential weighting below that spread buys little and hand-tuned weights are usually overfitting. Refusal rules: an unanswered exercise leaves the numerator, the denominator and the chance term together; an empty sitting scores exactly zero everywhere and is refused; fewer than twenty answered exercises refuses the headline, the sentence and every part, and the refusal is printed where each would have gone; a part with fewer than eight answered exercises or omega under .70 carries a three-way placement and no figure, no band and no percentile, with the reason printed; flatness is tested on the raw pressed values of the eight agreement statements, never on the reverse-keyed values, and a raw standard deviation under .45 sets the eight statements aside from every part with a refusal saying, in the reader's words, that one press worded two ways is still one press. No percentile appears anywhere because no norm group exists. Every declared prior and the assumed standard deviations of 30 points on the composite and 38 on a part, taken from a modelled mixed-ability reference sitting run through this instrument's own scorer, are authored assumptions stated in the open, and observed answer shares and observed spreads will replace them once live data exist. A careless-responding flag count is computed for the operator and never shown to the respondent as a judgement. The item conventions for the agreement statements follow the International Personality Item Pool (Goldberg and colleagues), which is in the public domain; the IPIP is cited and attributed, it neither endorses nor is affiliated with this instrument, and every item here is an original work. Nothing is reproduced from any published apology, forgiveness, conflict or relationship inventory, and this instrument is not affiliated with any commercial instrument, certification body or counselling framework. Sources drawn on across the construct: Lewicki, Polin and Lount (2016) on the structure of effective apologies, in which acknowledgement of responsibility and offer of repair carry the most weight and a request for forgiveness the least; Fehr, Gelfand and Nag (2010), the meta-analytic synthesis of the situational and dispositional correlates of forgiveness, including apology and timeliness; Schumann (2014) on defensive responses to victims and on what a self-affirmed transgressor does differently, Schumann and Dweck (2014) on who accepts responsibility, and Schumann (2018) on the barriers to apologising and the costs of a partial or conditional apology; Scher and Darley (1997) on which components of an apology speech act do the work; Kim, Ferrin, Cooper and Dirks (2004) on repairing competence-based against integrity-based trust violations; Tax, Brown and Chandrashekaran (1998) and Tax and Brown (1998) on service-recovery justice, distributive, procedural and interactional; Smith, Bolton and Wagner (1999) on customer satisfaction after failure and recovery; Gottman (1994) and Gottman and Silver (1999) on repair attempts and their reception; Tavris and Aronson (2007) on self-justification and the passive voice of 'mistakes were made'; Leunissen, De Cremer, Reinders Folmer and van Dijke (2013) on the apology mismatch between what victims want and what offenders offer, and Leunissen, De Cremer, van Dijke and Reinders Folmer (2014) on how offenders overestimate how aversive apologising will be; Bohns (2016) on how people misjudge the way a direct approach to another person is received, which extends to the imagined cost of a repair conversation; Ohbuchi, Kameda and Agarie (1989) on apology as aggression control; McCullough, Worthington and Rachal (1997) on interpersonal forgiving in close relationships; Schlenker and Darby (1981) on the use of apologies in social predicaments; Gollwitzer and Sheeran (2006) on implementation intentions, the meta-analytic effect of about d = .65 across ninety-four studies that the report's one if-then sentence rests on; and Baumgartner and Steenkamp (2001) on acquiescence and balanced keying.