traditional factor analysis to structural-causal 8. Cronbach and Meehl, note 7, p. 291 15. For SFA issues in longitudinal data anal- modeling in developmenlal research, m Experi- 9. t^anonical regression models can be written ysis, see I J. McArdle and F Anderson, Lalenl mental Psychology in the ruiure, V. Sarris and A. as a restricted structural equation model where variable growth models in research on aging, in Parducci,'Eds. (Krlbaum, Hillsdale, NJ, 1984); j.J. we eliminate (a) the error variance D^ ^ 0 on the The Handbook of the Psychology of Aging, J.E. Birren McArdle, Principals versus principles in factor latent variable C and (b) the internal validity as- and K,W. Schaie, Eds. (Plenum Press, New York, analysis, Multivariate Behavioral Researth, 25(1), sumptions of uncorrelated residuals among the V" 1990); J.J. McArdle and F. Hamagami, Modehng Sl-87 (1990). variables; see McArdle and Prescott, note 7. incomplete longitudinal and cross-seclional data 2 For path analysis diagrams, see S. VVrighi, 10. For the debate on job performance, see using latent growth structural models. Experimen- On "Path analysis in genetic epidemiology: A cri- M,J. Ree and J.A. Earles, Intelligence is the best tal Aging Research. 18, 145-167 (1992) tique," American journal of Human Genetics. 35, predictor of job performance, Ciirrenl Directions in 16. For further issues about SFA missing-data 757-762 (1982); J.]. McArdle and S.M. Boker, Psi/chakigical Science, 1, 8^^-89 (1992); RJ. Stern- models, see I'.D. Allison, Estimation of linear RAMpath (Eribaum, Hilhdalc, N), 1990). bcrg and R.K. Wagner, The ^.'-ocentnc view of models with incomplete data, m Sociological Meth- 3. I'or classical factor analysis, see J R. Nes- intelligence and job performance is wrong. Cur odology, 1987, C.C. Clogg, Ed. (Jossey-Bass, San seiroade and R. B. Cattell, Handbook of Multivaruitr rent Directions in Psycholo^iictt! Science, 2, 1 4 Francisco, 1987): J.J. McArdle, Structural factor Experimental Psychotogy (Rand MciMally, New (1993). analysis with incomplete data, Multivariate Behav- York, 1988); J.B, Carroll," Human Cognitive Abilities: 11. For the debate on personality factors, see ioral'R.esearch, .-i, 409 454 (1994). A Survey of Pactor-Analytic Studies (Cambridge B. Mershon and R. Gorsuch, Number of factors in 17. For critiques of TEA, see S.J. Gould, 77ji' Universih' Press, New York, 1993), the personality sphere: Does increase in factors Mi^mcasure of Man (W W. Norton, New York, 4. For statistical testing in factor analysis, see increase predictability in real-life criteria? lournal 1981], R.J. Sternberg, Human intelligences The D.N. Lawley and A.E. Maxwell, fuctor Anaiysif^ (!,•= of Personality und Social Psycholo^ii, 55, 67.> .h8(l modo] IS the message. Science, 230, 1111-1118 a Statistical Method (BulterwoTihs, London, 1971); (1988); L.R,' Goldberg, An" alternative "descrip- (1985). M, Browne and R. Cudeck, Alternative ways of tion of personalily": I he Big-Five factor structure, 18. For critiques o! SFA, see N Cliff, Some assessing model fit, m Testin)^ Structural Ltjuatioii journal of t'enaiialittf and Social Psycholo'^y, 59, cautions concerning the application of causal Modeh, K. Bollen and S. Long, Eds. (Sage, New- 1216-1229 (1991)). modeling methods, Multivariate Behavioral Re- bury Park, CA, 1993). 12. For the debate on behavioral genetic mod- search, 18, 115-126 (1983); P. Cohen, J. Cohen, J. 5. Eor power analysis in SFA, see A. Sattora els, see J.J. McArdle and Fi.H Goldsmith. Some Teresi, M. Marchi, and C.N. Vekv., Problems in and W.E. Saris, Power of the likelihood ratio test alternative structural equation models for niulti- the measurement of latent variables in struclural in covariance structure analysis, Psychometrika, variatt? biometric analyses, Bt'havior CJeiit'ticf^, 20, equations causal models. Applied Psychological 50, 83-90 (19«5); M. Sobel and G.W. Bohrnstedt, 569-608 (1990); M.C. Neale and i .R. Cardon, Mcasuwment, 14, 183-196(199(1) Use of null models in evaluating the fit of covari- ,\^dliOiiolo^y for Genetic Studies of Twins and families 19. R.B. Cattell, I'actof Analysis' An Introduc- ance structure models, in SociologKai Methodoiof^y (Kluwer Academic, Boston, 1992), tion and Manual for Ihe Psychologist and Social Sci- 1985, N.B. Tuma, Ed, 0ossey-Bass, San Francis- 13. For current issues in SFA group selection entist (Harper & Brothers, New York, 1952), p. co, 1985). effects, see J,L, Fiorn and J.J, McArdle, A practi- 315 6. For factor analysis of item-level data, see R. cal and theoretical guide to measurement invari- Mislevy, Recent developments in the factor anal- ance in aging research, E.\pt'>imental Aging Re Recommended Reading ysis of categorical variables, journal of Educational search, 18, 117-144 (1992); J.J. McArdle and' R.B. Statistics, II, 3-31 (1986); D.H. Parry and ].]. Cattell, Structural equation models ol factorial in- Boilen, K.A , and Long, J.S. (Eds.). (1993). McArdle, An applied comparison of methods for variance in parallel proportional profiles and ob- Te-'ting Structural Filiation Models (Sage, Newbury least squares factor analysis of dichotomous vari- lique confactor problems, Multit'ariate Bchai'ioral Park, CA). ables. Applied Psychological Meiifiiretnoit, J5(l), 3>- Research, 290), 63-113(1994). Loehlin, J.C. (1992). Latent Variable Models: An 46 (1991) 14. For classical issues of group selection et- Intnhluclion !i< laclor. Path, and Struclura! Analysi> 7. For construct validity and SFA, see L.J. fects, ,see W. Meredith, A method lor studying (Eribaum, Fiillsdale, NJ|. Cronbach and P. Meehl, Construct validity in diflerences between groups, I'syehoiiu'trika. 3<J{1), McDonald, R.P. (1985). Factor Anah/sis and Ri-- psychological tests, Psycholo<,^ical Hulletui, 52, 2S1- "15-29 (1965); J.R Nesselroade^ Temporal selec- lated Methods (i'.rW-iaum, Hillsdale, NJ).' 302 (1955); J.J. McArdle and C.A. Prescott, Age- tion and factor invariance in the .study of devel- Miilaik,S A. (1988). Confirmalor\'factor anal- based construct validation using structural equa- opment and change, in l.ife-Sfim Devehpnient and ysis. In Ihc Himdbook of Multivariate Experimental tion modeling. Experimental Agini; Research, iS(3), Behavior, Vol. 5, P.B Dalles and O.G. Brim, Fds. 'I'sychology, J.R. Kesselroade and R.B. Caltell, 87-115 (1992). (Academic Press, New York, 1983). Eds. (Plenimi Press, New York). rected mainly by culture-specific Towards a Psychology of Food and traditions. This article deals with Eating: From Motivation to Module to three psychologically significant Model to Marker, Morality, Meaning, Paul Rozin is a Professor of Psy- and Metaphor chology at the University of Pennsylvania. His research inter- Paul Rozin ests are the interaction of biolog- ical, psychological, and cultural influences in relation to food at- titudes and food choice; cultural psychology; and cultural evolu- Each human being experiences from being a source of nutrition tion. Address correspondence to a food trajectory that begins with and sensory pleasure to being a so- Paul Rozin, Department of Psy- one food, milk, and expands to an cial marker, an aesthetic experi- chology, University of Pennsyl- vania, 3815 Walnut St., Philadel- incredibly varied set of foods and ence, a source of meaning and phia, PA 19104-6196; e-mail: preparations, attitudes, and food- metaphor, and, often, a moral en- [email protected]. related rituals. Food progresses tity. These transformations are di- Published by Cambridge University' Press aspects of human relations with on animal abilities and plasticity of the human brain, body form, food. First, although food relations depend very much on two things: and mind. constitute a major part of human the type of food and the range of Throughout human history, ob- life, they have been virtually ig- food. Specialists, who concentrate taining adequate or pleasing food nored by psychologists. Second, on one kind of food (such as car- has been a major force in deter- biological, psychological, and cul- nivores, insectivores, and some mining both daily events and tural adaptations to food problems fruit and leaf eaters), face few events of great significance. The become the basis (model) for other problems in selecting or balancing linkage between the Eastern and biological and cultural aspects of their diet. Because their range of Western Hemispheres, initiated humans, by the processes of pre- foods is narrow, their food must, by Columbus, had a massive effect adaptation in evolution and acces- of necessity, contain all essential on world politics and daily life. sibility in development, so that elements (the only way for a lion The explorations that led to this food adaptations become general- to be vitamin Bl deficient is for it great "Columbian exchange" were ized to other systems. This pro- to eat vitamin-Bl-deficient zebra). motivated in large part by a desire gression constitutes one of the For such animals, the identifica- for spices. great transformations in human tion of food can be programmed Today, expenditures on food cultural and individual history, genetically. For carnivores, most constitute the largest single cate- and represents the linkage of a ba- psychologically interesting adap- gory for the world at large.^ Food sic biological motivational system tations occur in the search process and drink constitute approximate- to aesthetic, social, and moral as- (because they often eat prey that ly 21% of total expenditures in pects of life. Third, human and an- are widely scattered and hard to Western developed countries, but imal food relations offer special ad- detect) or in the capture of moving more than 50% in many of the vantages for the study of general prey. Substantial social coopera- much more populous Third World processes. tion, motor and tracking skills, countries, notably, India and and elaborate constructions (e.g., China.