For most of the twentieth century, organisational psychology assumed that groups were resilient — that a talented, well-managed team could absorb one difficult member without much cost. The evidence built since 2006 says otherwise, though not in the way the folklore suggests.
The phrase "bad apple" was given empirical weight by Will Felps, Terence Mitchell and Eliza Byington, whose 2006 review in Research in Organizational Behaviour proposed that a single group member who persistently withholds effort, expresses negative affect, or violates interpersonal norms can measurably degrade a team's functioning — not through one dramatic act, but through a slow transfer of mood, trust and standards to everyone else at the table. The mechanism is emotional contagion, independently demonstrated in a laboratory study by Sigal Barsade (2002): mood spreads between people largely without their noticing, and negative mood spreads more readily and faster than positive mood.
This article follows that finding to where the evidence gets thinner. Felps himself later ran a widely retold experiment — planting a trained actor in forty-four-person teams — that is now the version of the story almost everyone knows, complete with a specific claim that teams lost 30–40% of their performance. That figure is real in the sense that Felps has described it consistently across interviews and public talks. It is not, on the evidence available, traceable to a peer-reviewed journal article that states it — the finding lives in a 2008 public-radio interview and the secondary literature that repeats it, not in a citable dataset. This article treats it accordingly: as a well-corroborated account from the researcher who ran the study, not as an established statistic to quote as fact.
What is better evidenced is who the more extreme "bad apples" tend to be, and what they cost. Sub-clinical psychopathic traits sit in roughly 0.6% of the general population (Coid et al., 2009) but appear at three to four times that rate among senior managers (Babiak, Neumann & Hare, 2010). A 2025 meta-analysis of 166 studies and over 49,000 employees found that psychopathy reliably predicts lower task performance, less cooperative behaviour, and substantially more counterproductive conduct at work (Roth & Klehe, 2025) — and that the damage is driven far more by impulsive, reactive "secondary" psychopathy than by the calculating "primary" form popular culture usually pictures. A Harvard Business School analysis of over 50,000 workers found that avoiding one toxic hire saves an organisation more money than landing a top 1% performer (Housman & Minor, 2015). And a single, carefully documented case study of a UK charity under a newly appointed psychopathic CEO recorded an eightfold-plus increase in staff sickness absence and the departure of the entire original workforce within a decade (Boddy, 2017).
None of this means every difficult colleague is a psychopath, or that one bad mood sinks a team. It means the folk wisdom pointed at something real, the size of the effect is genuinely large, and the specific numbers in circulation deserve more scrutiny than they usually get.