Industrial and organizational psychology — work, selection, and motivation

Intuition Beginner

Most adults spend about a third of their waking life at work. Industrial and organizational psychology — called I/O psychology — applies what psychology knows about thinking, feeling, and behaving to the place where that huge chunk of life happens. The question at the heart of the field is simple: how do we make work better, both for the people doing it and for the organizations that depend on it?

That question splits into two halves. The industrial half is the older, more mechanical side: how do we match the right person to the right job, measure how well they do it, and predict who will succeed? It grew out of early twentieth-century "scientific management," which treated hiring and training as engineering problems. The organizational half arrived later, drawing on social psychology: how do leadership, motivation, group dynamics, and job design shape how people feel and perform?

Suppose a company must choose fifty new hires from a thousand applicants. If it picks at random, it gets an average workforce. If it uses a good selection procedure — one that actually predicts performance — the fifty it hires will be substantially better than average. The difference, measured in output, is the payoff of getting selection right. I/O psychology is, in large part, the science of that difference.

The same logic extends to every other lever an organization pulls: how goals are set, how performance is appraised, how jobs are structured, how leaders behave. Each lever can be studied with evidence rather than gut feeling, and the evidence is often surprising.

Visual Beginner

The field maps onto a small number of applied problems, each drawing on a more basic area of psychology.

A useful organizing contrast: the industrial column treats performance as the outcome to predict, while the organizational column treats the person's experience of work as the outcome to understand. In practice the two feed each other — appraisal data drives training decisions, and job design changes both satisfaction and output.

Worked example Beginner

A warehouse needs to hire 100 packers and has 1,000 applicants. Historically, about 40% of randomly hired packers reach a satisfactory level of output. If the warehouse uses a cognitive ability test known to predict packing performance with a validity of $r=0.50$, what fraction of the 100 hired workers should reach satisfactory performance?

This is the Taylor-Russell expectancy question, worked out by Hazel Taylor and John Russell in 1939. The answer comes from the overlap between two bell curves — the test-score distribution and the performance distribution — restricted to the people you actually hire.

When you hire only the top 10% of applicants (a selection ratio of $100/1000=0.10$), and the test validity is $0.50$, and the base rate of satisfactory performance is $0.40$, the Taylor-Russell tables give an expected success ratio of roughly $0.69$. Of the 100 hires, about 69 should be satisfactory, against the 40 you would get by chance.

The gain is striking. With no test you get 40 satisfactory workers; with a moderately valid test you get about 69. That is 29 extra productive workers, purely from choosing a better predictor. Multiply that across a year of wages and output, and you can see why Frank Schmidt and John Hunter titled their 1998 review "85 years of research findings" — the cumulative case for valid selection is enormous ^{[SchmidtHunter1998]}.

A related finding from their meta-analysis: general mental ability (GMA) is the single best single predictor of overall job performance, with a corrected validity of about $r=0.51$, and combining GMA with a test of integrity pushes that to about $0.65$.

Check your understanding Beginner

Formal definition Intermediate

Industrial and organizational psychology is the application of psychological theory, measurement, and research methods to people at work and to the organizations that employ them. Following the standard split, personnel psychology covers job analysis, selection, training, and performance appraisal; organizational psychology covers motivation, leadership, job satisfaction, and organizational development; human factors / ergonomics covers the design of tools, workspaces, and systems to fit human capacities.

A predictor $X$ is any measured attribute used to forecast an outcome; the criterion $Y$ is the outcome of interest, usually a measure of job performance. The criterion-related validity of a predictor is the correlation

$$r_{XY}=\frac{\mathrm{Cov}(X,Y)}{{\sigma }_X{\sigma }_Y}$$

between predictor scores and criterion scores, usually corrected for the artifacts discussed below. A predictor is valid to the degree that $r_{XY}$ is large in the relevant population.

The base rate $BR$ is the proportion of the applicant (or incumbent) pool that is successful on the criterion when no selection is applied. The selection ratio $SR=n_{\text{hired}}/n_{\text{applicants}}$ is the fraction hired. The success ratio is the proportion of those hired who turn out successful. The Taylor-Russell expectancy is the success ratio expressed as a function of $(BR,SR,r_{XY})$ under the assumption that predictor and criterion are jointly normal.

A job analysis is a systematic description of the tasks, duties, responsibilities, and worker requirements (knowledge, skills, abilities, other characteristics — KSAOs) of a job. It is the foundation on which selection, training, and appraisal are built: you cannot predict performance without first specifying what performance is.

Reliability is the consistency of measurement. Under classical test theory an observed score $X=T+E$ decomposes into a true score $T$ and random error $E$, with $\mathrm{Cov}(T,E)=0$. The reliability of $X$ is

$${\rho }_{XX}=\frac{\mathrm{Var}(T)}{\mathrm{Var}(X)}=\frac{\mathrm{Var}(T)}{\mathrm{Var}(T)+\mathrm{Var}(E)}.$$

Reliability bounds validity: an unreliable measure cannot be highly valid, because random error washes out the signal shared with the criterion.

Key model Intermediate

The validity model and the Taylor-Russell expectancy

The single most important formal object in personnel psychology is the joint distribution of predictor and criterion. Assume $(X,Y)$ is bivariate normal with correlation $r=r_{XY}$. The firm hires every applicant with $X\ge c$, a cut score chosen so that the selection ratio $SR=P(X\ge c)$ matches the desired number of hires. The success ratio is the conditional probability $P(Y\ge y_0\mid X\ge c)$, where $y_0$ is the satisfactory-performance threshold.

Hazel Taylor and John Russell tabulated this conditional probability in 1939 as a function of $(BR,SR,r)$, where $BR=P(Y\ge y_0)$ is the base rate. Three qualitative facts follow directly from the geometry of the bivariate normal:

1. If $r>0$, then $P(Y\ge y_0\mid X\ge c)>BR$ for every $SR<1$. Selection on a valid predictor always beats chance.
2. For fixed $r$ and $BR$, the success ratio is a decreasing function of $SR$: the more selective the firm, the better the average hired worker, because the predictor does more filtering.
3. For fixed $SR$ and $BR$, the success ratio is an increasing function of $r$: a more valid predictor yields more lift.

The Brogden (1946) utility formula gives the per-hire payoff in performance units. Let ${\mathrm{SD}}_y$ be the standard deviation of performance expressed in dollars (the cash value of a one-standard-deviation performance gap). The average productivity gain per selectee is

$$\Delta \bar{Y}=r_{XY}\cdot {\mathrm{SD}}_y\cdot \frac{\varphi (z_{SR})}{SR},$$

where $\varphi (\cdot )$ is the standard normal density and $z_{SR}$ the z-score cutting off the top $SR$ of the distribution ^{[Brogden1946]}. The factor $\varphi (z_{SR})/SR$ is the selection ratio multiplier (the average predictor score of those hired, in standard units). Multiplied by the validity and the value of a standard deviation of performance, it gives the dollar value of using the test.

Two consequences make this model the backbone of the field. First, it converts a correlation — a number psychologists treat as effect-size currency — directly into money, which is the currency organizations actually use. A validity of $0.50$ sounds modest until you multiply it by a ${\mathrm{SD}}_y$ of $15,000 and a selection ratio multiplier of about $1.76$ (at $SR=0.10$): roughly $13,200 of extra output per hire, every year. Second, it shows why validity generalization matters: the same formula works whenever $r$ is known and the performance distribution is comparable, so a meta-analytic estimate of $r$ carries across organizations.

Bridge. This validity model builds toward the Schmidt-Hunter meta-analysis in the Master tier, and it appears again in the psychometrics strand 29.13.01 where reliability and validity are defined from first principles. This is exactly the place where the industrial half of the field meets classical test theory: the central insight is that a correlation is not just a descriptive statistic but a lever that, combined with a selection ratio and a dollar-valued standard deviation, moves organizational outcomes. The model generalises from selection to training evaluation and to the validity of performance appraisals, because in each case a predictor-criterion correlation is the engine that converts a measurement into a decision. The bridge is the recognition that every applied decision in I/O psychology — whom to hire, whom to train, how to score performance — reduces to the same predictor-criterion geometry.

Exercises Intermediate

Lean formalization Intermediate

This unit has lean_status: none. The psychometric core of the field — classical test theory, the attenuation inequality, the Taylor-Russell conditional probabilities, and the Schmidt-Hunter meta-analytic corrections — is genuinely formal and would be a reasonable Lean target, but it belongs to the probability and statistics strands rather than to a psychology module. A natural formalization project would state classical test theory as a probability setup ($X=T+E$ with the orthogonality assumptions), prove the attenuation formula $\rho (X,Y)=\rho (T_X,T_Y)\sqrt{{\rho }_{XX}}\sqrt{{\rho }_{YY}}$, and derive the bound $|\rho (X,Y)|\le |\rho (T_X,T_Y)|$; this is carried forward informally in the Full proof set below and would slot into section 26 (probability) or 45 (statistical learning). The substantive applied content — leadership styles, job design, organizational development, ergonomics — is empirical rather than theorem-shaped and is not a candidate for formalization.

Advanced results Master

Validity generalization and the Schmidt-Hunter synthesis

For most of the twentieth century the standing assumption in personnel psychology was that a selection procedure's validity was situationally specific: a test valid for one organization's nurses might be invalid for another's, because each job and organizational context was believed to be unique. Schmidt and Hunter overturned this assumption through validity generalization, an application of psychometric meta-analysis ^{[SchmidtHunter1998]}.

The mechanism is statistical. Observed validity coefficients in primary studies are dragged down and scattered by artifacts: unreliability of the predictor and the criterion, range restriction (because applicants are pre-selected, the predictor range is compressed, which shrinks the observed correlation), sampling error, and small-sample noise. Correcting each primary correlation for these artifacts — using the formulas of classical test theory and range-restriction theory — typically collapses a cloud of seemingly inconsistent validities into a single underlying value with a narrow confidence interval. The scatter was not "situational specificity"; it was mostly measurement error.

The headline finding: across 85 years of research and hundreds of jobs, general mental ability predicts overall job performance with a corrected validity of about $0.51$, and the best combinations do better still — GMA plus an integrity test reaches about $0.65$, as does GMA plus a structured interview, and GMA plus a work-sample test reaches about $0.60$. The single best single predictor of training success is GMA, at about $0.56$. Work-sample tests are the strongest single predictor of performance ($\approx 0.54$) but are expensive to build. Graphology, by contrast, corrects to about $0.02$ — essentially zero.

Two meta-analytic refinements matter at this tier. First, range restriction is often indirect: the variable used to select incumbents is not the predictor under study but a correlated one, and Hunter, Schmidt, and Le (2006) gave the corrected formula for the indirect case, which generally produces larger validity corrections than the direct formula. Second, the credibility interval (the between-study standard deviation of true validities) tests whether situational specificity survives the corrections; for GMA it is small, vindicating validity generalization, whereas for some situational judgment tests it remains large, indicating genuine context dependence.

Goal-setting theory: mechanisms and moderators

Edwin Locke's 1968 paper proposed that task motivation is driven by two properties of goals — their specificity and their difficulty — and Gary Latham turned this into a thirty-year program of laboratory and field experiments, summarized in Locke and Latham (1990) ^[Locke1968]. The empirical core is that specific and difficult goals produce higher performance than "do your best" instructions or vague exhortations, with effect sizes in the moderate-to-large range across a wide variety of tasks.

The theory identifies four mechanisms. Goals direct attention toward task-relevant activities and away from distractions; they mobilize effort in proportion to their difficulty; they increase persistence over time; and they activate task-relevant knowledge and strategies, including the search for new ones. The moderators refine the picture: goal effects depend on goal commitment (the goal must be accepted), self-efficacy (belief in one's capability amplifies the effect), task complexity (goals help on simple tasks more than complex ones, where strategy matters more), and feedback (without knowledge of results, even a specific goal cannot steer behavior). A crucial boundary is that specific difficult goals improve performance on well-defined tasks but can narrow attention and reduce performance on complex, creative, or learning tasks, where an explicit performance goal can induce premature fixation.

The Job Characteristics Model

Hackman and Oldham (1976) proposed that five core job dimensions drive three critical psychological states, which in turn drive motivation, satisfaction, and performance ^{[HackmanOldham1976]}. The dimensions are skill variety, task identity, task significance, autonomy, and feedback; the psychological states are experienced meaningfulness of the work, experienced responsibility for outcomes, and knowledge of results. The Motivating Potential Score

$$\mathrm{MPS}=\frac{\mathrm{SV}+\mathrm{TI}+\mathrm{TS}}{3}\times \mathrm{AUT}\times \mathrm{FB}$$

summarizes a job's motivating potential. The model also predicts that the strongest responses to enriched jobs occur in employees high on growth-need strength — the desire for challenge and development — which is why the same redesign can energize one worker and overwhelm another. Empirical support is moderate: the mediations predicted by the model are usually in the expected direction but weaker than the theory claims, and job enrichment reliably raises satisfaction and internal motivation more than it raises objective performance.

Leadership: traits, behaviors, and transformation

Leadership research has cycled through three broad programs. Trait theory asks what qualities distinguish leaders from non-leaders; Stogdill's (1948) review initially seemed to kill the trait approach by finding no consistent traits, but later meta-analyses (Lord, DeVader, and Alliger, 1986) showed that intelligence, dominance, and adjust modestly but reliably predict leadership emergence and perceptions of leadership ^[Lord1986]. Behavioral theory, rooted in the Ohio State (consideration and initiating structure) and Michigan studies, asks what leaders do rather than what they are; both consideration (people-oriented) and initiating structure (task-oriented) predict group performance and subordinate satisfaction, with initiating structure stronger for performance and consideration stronger for satisfaction. Contingency theories — Fiedler's, House's path-goal — argue that the right behavior depends on the situation. The most influential recent synthesis is transformational leadership (Bass, 1985), which distinguishes transactional leadership (contingent rewards and monitoring) from transformational leadership (idealized influence, inspirational motivation, intellectual stimulation, individualized consideration). Transformational leadership predicts follower effort, commitment, and organizational citizenship above and beyond transactional rewards, with meta-analytic correlations typically in the $0.4$ to $0.7$ range, though the construct has been criticized for conceptual overlap with its outcomes.

Synthesis. The advanced results show that the foundational reason I/O psychology works as an applied science is the same across its three pillars: noisy, situation-specific measurements collapse, under psychometric correction, into stable and combinable effect sizes. This is exactly the lesson of validity generalization, and it generalises to goal-setting (where effect sizes survive lab-to-field transfer and accumulate across moderators) and to leadership (where meta-analysis rescues trait and behavioral findings that single studies had buried). Putting these together reveals a field built on three reliable engines — a predictor-criterion correlation, a goal mechanism, and a leadership style — each measurable, each combinable, each translatable into organizational payoff. The bridge is that all three reduce the same way: a measured predictor, corrected for artifacts, multiplied by a context-dependent multiplier, produces a defensible estimate of human behavior at work — and that reduction is why an applied field can claim scientific status without sacrificing practical reach.

Full proof set Master

Proposition (Spearman's attenuation formula)

Let $X$ and $Y$ be observed scores obeying classical test theory, with $X=T_X+E_X$ and $Y=T_Y+E_Y$, where $T_X,T_Y$ are true scores and $E_X,E_Y$ are error scores. Assume the errors are uncorrelated with the true scores and with each other:

$$\mathrm{Cov}(T_X,E_X)=\mathrm{Cov}(T_Y,E_Y)=\mathrm{Cov}(T_X,E_Y)=\mathrm{Cov}(T_Y,E_X)=\mathrm{Cov}(E_X,E_Y)=0.$$

Let the reliabilities be ${\rho }_{XX}=\mathrm{Var}(T_X)/\mathrm{Var}(X)$ and ${\rho }_{YY}=\mathrm{Var}(T_Y)/\mathrm{Var}(Y)$. Then the observed validity is

$$\rho (X,Y)=\rho (T_X,T_Y)\sqrt{{\rho }_{XX}}\sqrt{{\rho }_{YY}},$$

and in particular $|\rho (X,Y)|\le |\rho (T_X,T_Y)|$.

Proof. By definition of covariance and the decomposition $X=T_X+E_X$,

$$\mathrm{Cov}(X,Y)=\mathrm{Cov}(T_X+E_X,T_Y+E_Y).$$

Expanding bilinearly gives four terms:

$$\mathrm{Cov}(T_X,T_Y)+\mathrm{Cov}(T_X,E_Y)+\mathrm{Cov}(E_X,T_Y)+\mathrm{Cov}(E_X,E_Y).$$

By the orthogonality assumptions, the last three terms vanish, so $\mathrm{Cov}(X,Y)=\mathrm{Cov}(T_X,T_Y)$. Now the observed correlation is

$$\rho (X,Y)=\frac{\mathrm{Cov}(X,Y)}{\sqrt{\mathrm{Var}(X)}\sqrt{\mathrm{Var}(Y)}}=\frac{\mathrm{Cov}(T_X,T_Y)}{\sqrt{\mathrm{Var}(X)}\sqrt{\mathrm{Var}(Y)}}.$$

Insert the true-score correlation $\rho (T_X,T_Y)=\mathrm{Cov}(T_X,T_Y)/(\sqrt{\mathrm{Var}(T_X)}\sqrt{\mathrm{Var}(T_Y)})$ and solve for the covariance:

$$\mathrm{Cov}(T_X,T_Y)=\rho (T_X,T_Y)\sqrt{\mathrm{Var}(T_X)}\sqrt{\mathrm{Var}(T_Y)}.$$

Substituting back,

$$\rho (X,Y)=\rho (T_X,T_Y)\cdot \frac{\sqrt{\mathrm{Var}(T_X)}}{\sqrt{\mathrm{Var}(X)}}\cdot \frac{\sqrt{\mathrm{Var}(T_Y)}}{\sqrt{\mathrm{Var}(Y)}}=\rho (T_X,T_Y)\sqrt{{\rho }_{XX}}\sqrt{{\rho }_{YY}}.$$

Since reliabilities satisfy $0\le {\rho }_{XX},{\rho }_{YY}\le 1$, the product $\sqrt{{\rho }_{XX}}\sqrt{{\rho }_{YY}}\le 1$, so $|\rho (X,Y)|\le |\rho (T_X,T_Y)|$. Equality holds only when both the predictor and criterion are perfectly reliable. $■$

This is the formal statement of the slogan "reliability bounds validity." It is the engine behind every correction in the Schmidt-Hunter program: a primary study that reports $r=0.30$ against a predictor of reliability $0.80$ and a criterion of reliability $0.60$ has observed only $\sqrt{0.80}\sqrt{0.60}\approx 0.69$ of the true validity, so the corrected estimate is $0.30/0.69\approx 0.43$. Without this correction the field would chronically underestimate its predictors and chronically over-estimate situational specificity.

Connections Master

The single strongest link runs to the cognition and intelligence unit 29.05.01, because the best single predictor of job performance — general mental ability, GMA — is precisely the $g$-factor that unit develops. Schmidt and Hunter's $0.51$ is, formally, the criterion validity of Spearman's $g$ applied to work; the same construct measured by the same instruments reappears here as an applied lever rather than a theoretical one, and the $g$-vs-multiple-abilities debate in 29.05.01 maps directly onto whether narrow abilities add incremental validity on top of GMA for specific jobs.

The organizational half of the field depends on social psychology 29.07.01: leadership is a phenomenon of social influence, group dynamics shape team performance, and the Hawthorne studies — historically central to I/O — are best read as a demonstration that being observed and being part of a cohesive social group change behavior, an idea formalized in the social-psychology treatment of group cohesion and conformity. Organizational development interventions likewise borrow from the group-dynamics and attitudes-change traditions developed there.

Personality psychology 29.08.01 supplies the second most studied class of personnel predictors after GMA: the Five-Factor Model, especially conscientiousness (corrected validity $\approx 0.31$ for overall performance, higher for jobs with rule-following and dependability) and emotional stability. The Big Five also underwrite research on person-organization fit, on counterproductive work behavior, and on the integrity tests that, combined with GMA, reach the $0.65$ benchmark. Trait leadership findings draw on the same personality dimensions.

Finally the psychometrics unit 29.13.01 provides the measurement foundation: the definitions of reliability and validity used throughout this unit, the meta-analytic apparatus behind validity generalization, and the attenuation formula proved in the Full proof set. The formal link is so direct that the two units are best read as one measurement theory applied twice — once to ability and personality tests, once to the job-performance criterion.

Historical and philosophical context Master

The field has two origin points. The first is scientific management: Frederick Winslow Taylor's The Principles of Scientific Management (1911) argued that work should be studied empirically — timed, motion-analyzed, decomposed — and that workers should be scientifically selected, trained, and paid by the piece ^[Taylor1911]. Taylor's "time-and-motion" studies, and the husband-and-wife team Frank and Lillian Gilbreth's parallel work on motion economy, established job analysis as a serious activity and gave the field its engineering flavor. The price was a reductive view of workers as cogs: Taylor's system paid well but controlled tightly, and the historical verdict credits the methodological contribution while condemning the human cost.

The second origin is the Hawthorne studies, run at Western Electric's Hawthorne Works in Cicero, Illinois between 1924 and 1932 ^{[RoethlisbergerDickson1939]}. The studies began as straightforward illumination experiments — does brighter lighting raise output? — and produced a puzzle: output rose when lighting improved, but it also rose when lighting was reduced, and it rose in a control group that experienced no change at all. The investigators, led by Elton Mayo and elaborated by Fritz Roethlisberger and William Dickson, concluded that the workers were responding less to the physical conditions than to the social fact of being studied, of being consulted, and of being part of a cohesive group. The "Hawthorne effect" — that observation itself changes behavior — entered the language, and the studies pulled I/O psychology toward the organizational half it had lacked. (Modern re-analyses, notably by Rice and by Levitt and List, have complicated the original story; the effect is real but smaller and less clean than the textbook version suggests.)

The post-war period saw three converging developments. Personnel selection matured into a quantitative science through the work of Brogden (1946) on utility, Taylor and Russell (1939) on expectancy, and Cronbach and Gleser on decision theory. Motivation became a theoretical target with Vroom's (1964) expectancy theory ^[Vroom1964], Locke's (1968) goal-setting theory, and Hackman and Oldham's (1976) Job Characteristics Model — three frameworks that still dominate the textbook treatment. Leadership moved from trait theories (abandoned after Stogdill's 1948 review) through behavioral and contingency theories to the transformational-transactional synthesis of the 1980s.

The methodological high-water mark was the validity generalization program of the 1970s and 1980s, in which Frank Schmidt, John Hunter, and their collaborators applied psychometric meta-analysis to four decades of selection-validity studies and argued that most of the scatter in the literature was artifact rather than substance ^{[SchmidtHunter1998]}. This was not a quiet methodological note: it overturned the legal and professional consensus that each organization had to validate its own tests locally (the "situation specificity" doctrine that had underpinned federal validation guidelines), and it is why a modern personnel psychologist can quote a number like $0.51$ for GMA with a straight face.

Philosophically the field sits on a tension it has never fully resolved. It is applied — paid by organizations to improve organizational outcomes — and that creates a structural pressure to treat people instrumentally, as means to productivity. Its humanistic wing (the organizational-development tradition, drawing on Kurt Lewin and Carl Rogers) insists that healthy organizations serve the people in them, not only the reverse. The same tension runs through performance appraisal (developmental feedback versus administrative control), through selection (matching people to jobs versus sorting them into hierarchies), and through job design (enrichment for the worker's sake versus for the firm's). Honest I/O psychology holds both commitments at once and refuses to collapse either into the other.

Bibliography Master

@book{taylor1911scientific,
  author    = {Taylor, Frederick Winslow},
  title     = {The Principles of Scientific Management},
  publisher = {Harper \& Brothers},
  year      = {1911}
}

@book{roethlisberger1939management,
  author    = {Roethlisberger, Fritz J. and Dickson, William J.},
  title     = {Management and the Worker},
  publisher = {Harvard University Press},
  year      = {1939}
}

@article{taylor1939expectancy,
  author  = {Taylor, Hazel C. and Russell, John T.},
  title   = {The Relationship of Validity Coefficients to the Practical Effectiveness of Tests in Selection},
  journal = {Journal of Applied Psychology},
  volume  = {23},
  number  = {5},
  pages   = {565--578},
  year    = {1939}
}

@article{brogden1946interpretation,
  author  = {Brogden, Hubert E.},
  title   = {On the Interpretation of the Correlation Coefficient as a Measure of Predictive Efficiency},
  journal = {Journal of Educational Psychology},
  volume  = {37},
  number  = {2},
  pages   = {65--76},
  year    = {1946}
}

@book{vroom1964work,
  author    = {Vroom, Victor H.},
  title     = {Work and Motivation},
  publisher = {Wiley},
  year      = {1964}
}

@article{locke1968toward,
  author  = {Locke, Edwin A.},
  title   = {Toward a Theory of Task Motivation and Incentives},
  journal = {Organizational Behavior and Human Performance},
  volume  = {3},
  number  = {2},
  pages   = {157--189},
  year    = {1968}
}

@article{hackman1976motivation,
  author  = {Hackman, J. Richard and Oldham, Greg R.},
  title   = {Motivation through the Design of Work: Test of a Theory},
  journal = {Organizational Behavior and Human Performance},
  volume  = {16},
  number  = {2},
  pages   = {250--279},
  year    = {1976}
}

@book{locke1990theory,
  author    = {Locke, Edwin A. and Latham, Gary P.},
  title     = {A Theory of Goal Setting and Task Performance},
  publisher = {Prentice-Hall},
  year      = {1990}
}

@article{schmidt1998validity,
  author  = {Schmidt, Frank L. and Hunter, John E.},
  title   = {The Validity and Utility of Selection Methods in Personnel Psychology: Practical and Theoretical Implications of 85 Years of Research Findings},
  journal = {Psychological Bulletin},
  volume  = {124},
  number  = {2},
  pages   = {262--274},
  year    = {1998}
}

@article{lord1986meta,
  author  = {Lord, Robert G. and DeVader, Christy L. and Alliger, George M.},
  title   = {A Meta-Analysis of the Relation between Personality Traits and Leadership Perceptions},
  journal = {Journal of Applied Psychology},
  volume  = {71},
  number  = {3},
  pages   = {402--410},
  year    = {1986}
}

@article{hunter2006indirect,
  author  = {Hunter, John E. and Schmidt, Frank L. and Le, Huy},
  title   = {Implications of Direct and Indirect Range Restriction for Meta-Analysis Methods and Findings},
  journal = {Journal of Applied Psychology},
  volume  = {91},
  number  = {3},
  pages   = {594--612},
  year    = {2006}
}

@book{spector2019iopsych,
  author    = {Spector, Paul E.},
  title     = {Industrial and Organizational Psychology: Research and Practice},
  edition   = {7},
  publisher = {Wiley},
  year      = {2019}
}

@book{muchinsky2024paw,
  author    = {Muchinsky, Paul M. and Culbertson, Samuel S.},
  title     = {Psychology Applied to Work},
  publisher = {Cengage},
  year      = {2024}
}