American Journal of Community Psychology,Vol.35,Nos.3/4,June2005(C 2005)
DOI:10.1007/s10464-005-3397-z
Getting the Big Picture in Community Science:
Methods That Capture Context
Douglas A.Luke1
Community science has a rich tradition of using theories and research designs that are con-
sistent with its core value of contextualism.However,a survey of empirical articles published
in the American Journal of Community Psychology shows that community scientists utilize a
narrow range of statistical tools that are not well suited to assess contextual data.Multilevel
modeling,geographic information systems(GIS),social network analysis,and cluster analysis
are recommended as useful tools to address contextual questions in community science.An
argument for increased methodological consilience is presented,where community scientists
are encouraged to adopt statistical methodology that is capable of modeling a greater pro-
portion of the data than is typical with traditional methods.
KEY WORDS:community science;context;multilevel modeling;GIS;social network analysis;cluster
analysis.
INTRODUCTION
A generation ago,Allen Barton had this to say about social science research:
For the last thirty years,empirical social research has
been dominated by the sample survey.But as usually
practiced,using random sampling of individuals,the
survey is a sociological meatgrinder,tearing the indi-
vidual from his social context and guaranteeing that
nobody in the study interacts with anyone else in it.
It is a little like a biologist putting his experimen-
tal animals through a hamburger machine and look-
ing at every hundredth cell through a microscope;
anatomy and physiology get lost,structure and func-
tion disappear,and one is left with cell biology....
If our aim is to understand people’s behavior rather
than simply to record it,we want to know about pri-
mary groups,neighborhoods,organizations,social
circles,and communities;about interaction,com-
munication,role expectations,and social control.
(Barton,1968as reported in Freeman,2004)
Although this statement came a few years after the Swampscott Conference,it is almost as if Barton were talking to the group of community scientists
1To whom correspondence should be addressed at Saint Louis University School of Public Health,3545Lafayette Avenue,Saint Louis,MO63104;e-mail:dluke@https://www.doczj.com/doc/128189081.html,.
who were in Massachusetts inventing a new?eld.As Kelly notes,“The conference was an occasion to ac-claim that beyond conventional methods and,with a focus beyond the individual,there were valid activi-ties and meaningful roles for a new kind of psychol-ogist,the community psychologist(Kelly,2002,p.44, emphasis added).
Thus,community scientists have put context front and center as one of the core values of community psychology.Shinn and Rapkin(2000) advised that“...a central tenet of community psy-chology is that human behavior must be under-stood in context.”So,for example,community sci-entists study domestic violence using methods and theories that are consistent with the view that do-mestic violence is not just an individual behavior, but a complex process shaped by historical,social,?nancial,and legal contexts.The types of groups that community scientists work with(e.g.,domes-tic violence victims,young mothers,gays and les-bians,drug users,inner city residents,etc.)are of interest not because of any de?ning psychological characteristics,but because these groups have been affected in speci?c ways by the economic,social, cultural,and physical situations in which they are embedded.
185
0091-0562/05/0600-0185/0C 2005Springer Science+Business Media,Inc.
186Luke
However,although community scientists fre-quently employ theories,models,and frameworks that take context into account,they seem to be less likely to employ contextual methods in their work. Community psychology does have a long tradition of analytic innovation(Revenson et al.,2002),but the majority of empirical work in community science utilizes a remarkably narrow array of analytic ap-proaches.
In an earlier set of articles examining science and community psychology,Kelly’s(2003)?rst sug-gestion for vitalizing scienti?c community psychology was to“demythologize statistics.”He argued that not only do we need to move beyond traditional methods such as ANOVA,regression,and factor analysis;but that community scientists need to be in control of the quantitative methods,and not vice versa.This paper is one small response to his call.
The goal of this paper is to provide examples of a number of useful analytic methods that can be used to capture community context.It is my hope that if community scientists have a wider variety of analytic tools in their toolbox,they will be more likely to get away from Barton’s sample survey approach to do-ing social science.To support this argument,the next section presents a review of statistical practices dur-ing six years of articles published in the American Journal of Community Psychology.The bulk of the paper describes four statistical methods appropriate for assessing context,and provides examples drawn from community science.The paper concludes with a discussion of the need for the development of con-silient statistical methods in community science. STATISTICAL PRACTICE IN
COMMUNITY SCIENCE
To provide context for the following discussion, I conducted a review of the types of statistical meth-ods used in the American Journal of Community Psy-chology(AJCP)during two time periods:from1981 to1983,and20years later from2001to2003.By re-viewing six years of AJCP articles,we can get a good sense of the types of statistical practices used by com-munity scientists.Also,by examining two different time periods we can see how statistical practice has changed over a20year period.For the purposes of this paper,this review can help ascertain the extent to which contextual methods have been used in the past, or are being used currently by AJCP authors.Al-though AJCP is not the only place that community-oriented empirical work appears,it is one of the more important settings for work that aims to advance community science.
A total of304articles were reviewed;144in the early period(1981–1983)and160in the current period(2001–2003).Table I shows how the types of articles published in AJCP have changed during this time.In the early period,the majority(88%) of published articles were traditional empirical stud-ies where original data were presented.The current AJCP publishes a wider variety of articles.Almost half of all articles(48%)came from themed special issues.More room was also made for Presidential and award addresses.Finally,on a percentage basis four times as many articles with qualitative content are be-ing currently published compared to20years earlier (17%vs.4%).
Figure1shows how often particular types of statistical analyses were used in the215empirical papers published in the early1980s and from2001 to2003.(These percentages add up to more than 100%,because most papers used multiple types of statistical analyses.)An empirical article is one that presents original data that have been analyzed either quantitatively or qualitatively.We can see that there have been some changes in statistical practice over the past two decades.Well more than half(59.5%) of all empirical articles presented correlations in the early80s;currently a little over a third(34.8%)of the empirical articles present correlations.The use of ANOVA methods shows a similar drop,from 59.5to37.1%.Use of structural equation modeling (SEM)and associated methods such as path analysis increased dramatically during this same period(from 2.4to11.2%).This is not surprising given that sophis-ticated SEM software did not appear until the1990s.
However,despite these changes,this?gure supports Kelly’s contention that as a?eld we
Table I.AJCP Article Characteristics
Years Total articles Empirical Qualitative Special issue Address 1981–1983144126(88%)6(4%)10(7%)5(4%) 2001–200316089(56%)27(17%)76(48%)19(12%) Totals304215(71%)33(11%)86(28%)24(8%)
Getting the Big Picture in Community Science:Methods that Capture Context
187
Fig.1.Statistics usage in the American Journal of Community Psychology from 1981to 1983(N =126empirical articles)and
from 2001to 2003(N =89empirical articles).
predominately use traditional statistical approaches such as ANOVA,regression,psychometrics,corre-lations,and categorical statistics (e.g.,chi-square).More specialized techniques such as cluster analysis,social network analysis (SNA),multilevel modeling (HLM),meta-analysis,non-linear modeling,and ge-ographic information systems (GIS)are used rarely.Of the four techniques that are the topic of this paper (i.e.,cluster analysis,social network analysis,multi-level modeling,and geographic information systems)only network analysis was used at all in the early 3-year period.Furthermore,none of these contextual techniques appeared in more than four studies from 2001to 2003.Although one would expect that a very general modeling technique such as ANOVA would be used more often than,say,network analysis,the generality of the technique is not the entire story.For example,structural equation modeling,multi-level modeling,and nonlinear modeling are also very general statistical approaches,but they have not been used widely in the community sciences.One possible interpretation of the results from this methodologic review is that community science simply does not need to use nontraditional statistical tools to “...make a positive difference in commu-nity living and activities ...”(Sarason,2003,p.209).However,the main point I hope to make in the fol-lowing sections of this paper is not that we should use different tools because they are newer,more in-novative,or more popular.Rather,I believe that the quantitative methods that we historically rely on are not always the most appropriate tools if community scientists are seriously interested in understanding how the physical,social,organizational,cultural,eco-nomic,and political context shape human behavior and health.Although this paper will talk much about statistical practice,it is not a statistics paper.At its heart,I believe this argument is more philosophical and political than technical.Put simply,the decisions we make about the tools we use in community sci-ence say something about the values we hold as com-munity scientists.
188Luke
CONTEXTUALISM AND
COMMUNITY SCIENCE
I should venture to assert that the most pervasive
fallacy of philosophic thinking goes back to neglect
of context.
–John Dewey As part of its mission statement,the Society for Community Research and Action(SCRA)recognizes that“Human competencies and problems are best understood by viewing people within their social,cul-tural,economic,geographic,and historical contexts”(SCRA,2004).This core value of contextualism can be seen in many ways,including the close ties of com-munity psychology to ecological psychology,the use of explicitly contextual theories such as Behavior Set-ting Theory(Wicker,1992),and the recognition that effective interventions based on community science can and should be aimed at the extra-individual level (Tseng et al.,2002).That is,changing families,neigh-borhoods,schools,churches,and organizations can be a more effective way to enhance health than sim-ply intervening at the person level.
The long history of contextualism in commu-nity science puts us at the leading edge of a wave of change in how human health and behavior should be studied.One very prominent example is the2001 report issued by the National Institutes of Health (NIH)entitled Toward higher levels of analysis: Progress and promise in research on social and cul-tural dimensions of health(Of?ce of Behavioral and Social Sciences Research,2001).This report pre-sented a new agenda for NIH research focusing on two goals:(1)expanding health-related social sci-ences research at NIH and(2)integrating social sci-ence research into interdisciplinary contextual and multilevel studies of health.Particularly relevant for this discussion is the recommendation to:
...support the development of state-of-the-art so-
cial science methods.Challenges include measure-
ment at the group,network,neighborhood,and
community levels;the further development of meth-
ods for longitudinal research;multi-level research
designs that integrate diverse qualitative and quan-
titative approaches...;and the development of
improved methods for data collection and analysis
(p.3).
The empirical work published in AJCP shows the importance and prevalence of contextual ap-proaches.Notable examples in the past few years include interventions to enhance the quality of life of battered women through a community-based ad-vocacy approach(Bybee&Sullivan,2002);a de-scription of the relationship of ethnicity to family networks(Hirsch,Mickus,&Boerger,2002);iden-ti?cation of supportive organizational characteris-tics for rape victim advocates(Wasco,Campbell, &Clark,2002),and a study examining the effects of organizational downsizing on the health of em-ployees(Kivim¨aki,Vahtera,Elovainio,Pentti,& Virtanen,2003).These(and many other)examples utilize explicitly contextual frameworks.However, closer examination of these studies reveals that al-though they use a contextual framework,the ac-tual data collection and analyses are restricted to the person-level.
This is a general pattern.One hundred twenty of the empirical studies reviewed above incorporate some type of contextual framework or construct to understand individual-level behavior or character-istics.However,less than one in four(22.5%)of these studies collected data directly from the extra-individual levels or settings,or analyzed these data using appropriate methods for multiple levels.In other words,although community scientists value contextual thinking,we are much less likely to actu-ally employ contextual methods.
There are a number of possible explanations for this disconnect between our theories and our meth-ods.First,community psychology is still a relatively young?eld,and many practitioners have received their training from people and places still rooted in the person-centered traditions of clinical and general psychology.Many of the research design and ana-lytic approaches which are appropriate for contex-tual studies are more commonly used in other dis-ciplines such as organizational behavior,sociology, urban planning,etc.
Another reason for this disconnect that I would like to discuss at more length has to do with the Rule of the hammer—if the only tool you know how to use is a hammer,every situation looks like it needs to be hammered.The traditional statistical tools used most often by community scientists(i.e.,regression, ANOVA,factor analysis,etc.)have been developed to be used with a single level of analysis.Although it is common to use predictor variables in ANOVA and regression that give contextual information(e.g., group membership,state of residence,classroom), these models actually work by disaggregating group-level information to the individual level so that all predictors in a model are tied to the individual unit of analysis.This leads to at least two problems.First, all of the un-modeled contextual information ends
Getting the Big Picture in Community Science:Methods that Capture Context189
up pooled into the single individual error term of the model(Duncan,Jones,&Moon,1998).This is problematic because individuals belonging to the same context will presumably have correlated errors, which violates one of the basic assumptions of the general linear model.The second problem is that by ignoring context the model assumes that the re-gression coef?cients apply equally to all contexts,“...thus propagating the notion that processes work out in the same way in different contexts”(Duncan et al.,1998,p.98).
So,since we mainly know how to use statistics that are limited to individual-level analyses,it is not surprising that we end up not collecting or analyzing true contextual,multilevel data.
STATISTICAL APPROACHES CONSISTENT WITH CONTEXTUALISM
The good news is that there are a number of quantitative tools available that are consistent with the value of contextualism,and that are extremely useful for describing or modeling the in?uence of ecological,environmental,or group-level factors on individual-level behavior or health.In the next sec-tions of this paper I would like to highlight four such methods:(1)multilevel modeling;(2)geographic in-formation systems(GIS);(3)social network analysis; and(4)cluster analysis.
Multilevel Modeling
Multilevel modeling is a general regression-based statistical tool that can build statistical mod-els of data that are multilevel in nature.Multilevel models are more statistically appropriate for multi-level data than are traditional regression or ANOVA techniques(Snijders&Bosker,1999).For example, if the data are truly multilevel in nature,then the level-1units are clustered within the level-2units, and the level-1error terms are not likely to be in-dependent,thus violating one of the basic assump-tions of the general linear model.More importantly for us,with multilevel modeling we can build statis-tical models that match our conceptual models.That is,there will no longer be the disconnect between our thinking and our methods.
Multilevel modeling,also known as hierarchical linear modeling,mixed-effects modeling,or growth-curve modeling,has been underutilized in commu-nity science compared to developmental psychology, education,sociology,or political science.Hedeker, McMahon,Sason,and Salina(1994),in a study of a worksite smoking cessation program,gave an early demonstration of the utility of these methods;focus-ing on how the clustering of the person-level data can be appropriately adjusted for in a hierarchical model.Coulton,Korbin and Su(1999),used multi-level methods in an examination of how neighbor-hood factors in?uence child maltreatment.They con-vincingly demonstrated that the parameter estimates for the predictor variables were different than would be expected if only individual or neighborhood data had been used.Speci?cally,adverse neighborhood conditions weakened the effects of previously estab-lished individual risk factors,such as violence in the family of origin.Finally,Mankowski,Humphreys, and Moos(2001)develop a multilevel model that shows that person–environment?t of treatment be-lief systems is an important predictor of the extent of 12-step group involvement.
The following more detailed example may help demonstrate the utility of multilevel modeling for community science.The data and analyses come from work that we have done in the area of tobacco control policy(Luke&Krauss,2004).The main goal of this study was to identify the important in?uences on voting on tobacco-related legislation by members of Congress from1997to2000.The unit of analysis is an individual Congress member.
A typical single-level regression model that could be developed for these data might look like this:
VotePct i=β0+β1(Party)i+β2(Money)i+r i where we want to predict the percentage of time that a member of Congress voted in a pro-tobacco indus-try direction based on his or her political party,and the amount of money received from a tobacco indus-try political action committee(PAC).
However,members of Congress are not ran-domly selected,and it is reasonable to expect that the data will exhibit cluster effects.That is,Sena-tors or Representatives within the same state such as Massachusetts are likely to be more similar to each other on any number of important characteristics than Congress members from two different states. Figure2illustrates this—we can see that the percent-age of time that a Congress member votes for a bill in a pro-tobacco industry direction varies from state to state.
190
Luke
Fig.2.Average pro-tobacco vote percentage by Congress members–1997–2000.
The single-level regression model above cannot
handle the clustering of the data(and the concomi-tant non-independence of error terms).More impor-tant than this statistical problem is the fact that as community scientists we would like to build a more contextual model.In particular,we would like to ac-count for state-level characteristics that may in?u-ence voting behavior apart from the individual party and money received.The size of the tobacco econ-omy in a state,operationalized as size of tobacco har-vest,is a type of contextual variable that might in?u-ence Congressional voting.
The following set of equations show how a mul-tilevel statistical model is structured:
VotePct=β0j+β1j(Party)ij+β2j(Money)ij+r ij β0j=γ00+γ01(Acres)j+u0j
β1j=γ10+γ11(Acres)j+u1j
β2j=γ20+γ21(Acres)j+u2j
Although this statistical model looks quite a bit more complicated than the earlier single-level model, it is mainly doing two important things for us.First, the beta-coef?cients at level-1become dependent variables at level-2—indicating that the level-1inter-cepts and slopes may vary from state to state.Sec-ond,this type of model clearly shows which predic-tor variables are acting at level-1(i.e.,political party, PAC money)and which at level-2(amount of to-bacco acreage produced in a state).
Table II presents the results of?tting this multi-level model.We can see that at the Congress member level being a Republican(?γ10=.5066)and receiv-ing more PAC money(?γ20=.0049)are associated
with voting more often in the pro-tobacco indus-try direction.Speci?cally,for every$10,000received by a member of Congress from a tobacco PAC, there is approximately a5%greater chance to vote in the pro-tobacco industry direction on a tobacco-related bill.However,we also see that there is a sig-ni?cant state-level contextual effect—members from states with a greater tobacco economy are also more likely to vote pro-tobacco(?γ01=.0027).The rela-tively large and signi?cant random variability com-ponent for the intercept term(u0j)suggests,more-over,that there may be other important contextual effects that have not been included in the model.This relatively simple multilevel model shows us that it is important to take both individual and contextual effects into account when modeling tobacco policy behavior in Congress.For more details on how to
Getting the Big Picture in Community Science:Methods that Capture Context191 Table II.Hierarchical Model Estimates of the Effects of Political Party,Total Contributions,
and State-Level Tobacco Acreage on Percentage of Pro-Tobacco Votes
Fixed effects Coef?cient SE T-ratio p
For intercept(β0j)
Intercept(γ00).1828.02058.90.000
Acres(γ01).0027.0005 5.10.000
For PARTY slope(β1j)
Party(γ10).5066.021523.53.000
Acres(γ11)?.0016.0004 3.60.001
For MONEY slope(β2j)
Money(γ20).0049.00058.80.000
Acres(γ21)?.00002.0000 5.50.000
Random effects SD https://www.doczj.com/doc/128189081.html,p.χ2p
Intercept(u0j).0978.009684.1.000
Party Slope(u1j).0705.005054.8.023
Money Slope(u2j).0009.000029.2>.50
Level-1(e ij).1628.0265
Model Fit Deviance Parameters AIC BIC
?353.813?327.8?272.3
Notes:Political party is coded0for Democrat and1for Republican.Contributions are
recorded in thousands of dollars and Acres are recorded in thousands of acres.
?t and interpret multilevel models,see Hox(2002), Luke(2004),or Snijders and Bosker(1994). Geographic Information Systems
The second useful quantitative tool for explor-ing context is geographic information systems(GIS). GIS is a set of database,mapping,and statistical tools that allow visual and quantitative assessment of geo-graphic information.(Geographic in the broad sense, meaning any type of information that has a physi-cal location.)Although the production of maps is a common end result of GIS methods,GIS can also be used to construct contextual variables and statis-tically analyze spatial relationships(Haining,2003). GIS methods emerged as a computing technology over the last several decades and are used extensively in a wide variety of areas including geology,meteo-rology,urban planning,public safety,marketing,po-litical science,and public health(Mark,Chrisman, Frank,McHaf?e,&Pickles,2004).There has been a growing movement towards a‘community GIS’where GIS methods are used for community devel-opment,mapping of community assets,community health assessments,and eco-development(Carver, 2001).However,community scientists have been rel-atively slow to take advantage of GIS techniques.
GIS has been used most often by health and social scientists to examine patterns of criminal and risky health behaviors.For example,Wieczorek and Hanson(1997)used GIS methods to examine pat-terns of driving-while-intoxicated(DWI)and reveal the geographic context of this behavior.The authors use GIS maps including contour plots which show that DWIs are not distributed randomly around the metropolitan area.This type of analysis leads to poli-cies and interventions that can be aimed more pre-cisely,leading to lower costs and hopefully enhanced effectiveness.
A second example comes from the area of to-bacco control policy.Figure3is a map showing the distribution of tobacco billboards in the St Louis metropolitan area,shortly before tobacco billboards were eliminated as a provision of the1998Master Settlement Agreement(Luke,Esmundo,&Bloom, 2000).We used GIS to collect the billboard data and analyze the location patterns of tobacco advertising. In the?gure,the billboards are coded by the type of image found on the billboard,and the underly-ing map shows the population mix by census tract. By combining the billboard and census data,we can see that billboards with African American images on them tended to be concentrated in neighborhoods with higher proportions of African American resi-dents.We used this evidence to support an argu-ment that the tobacco industry was targeting African Americans for their products.
A?nal example of the utility of GIS is pro-vided by Goldstein et al.(2003)in a case study of
192
Luke Fig.3.Targeted placement of tobacco billboards with African American images in1998.
Symbols indicate type of image found on the tobacco billboard.Census block groups
are shaded according to percentage of African American residents.Taken from Luke,
Esmundo,&Bloom(2000).
the adoption of100%tobacco-free school policies in North Carolina.In this study the authors used key informant interviews to identify important fac-tors in?uencing the successful adoption of tobacco-free policies in14North Carolina school districts. Although key informants suggested that the local to-bacco economy had little direct in?uence on policy adoption,GIS analyses revealed that all school dis-tricts passing policies were located in counties with relatively little tobacco production(Fig.4).This map not only reveals an important pattern that can be the subject of further research,it allows dissemination of this?nding in a way that is clear,powerful,and easy to understand.
Network Analysis
Social network analysis is a broad set of meth-ods for the systematic study of social structure (Degenne&Fors′e,2004).Network tools are based on the analysis of relational data—information about the connections among a set of actors,be they
Getting the Big Picture in Community Science:Methods that Capture Context
193
Fig.4.Location of14North Carolina school districts adopting100%tobacco-free policies.Taken from Goldstein
et al.(2003).
persons,agencies,etc.This distinguishes network analysis from much of the rest of community sci-ence where attribute data are typically collected and analyzed.The relations can be almost any type of information about a connection between actors—friendship,recognition,money exchange,kinship,in-formation exchange,and respect.The relation can be very speci?cally de?ned,such as sharing needles among a network of drug users.
Network analysis has been broadly used in the physical,social,and health https://www.doczj.com/doc/128189081.html,work meth-ods have been used to model disease transmission (Rothenberg et al.,1998),job-seeking(Granovetter, 1973),the diffusion of innovations(Valente,1995), inter-organizational behavior(Mintz&Schwartz, 1981),social capital and community development (Lin,2000),and teen smoking(Alexander,Piazza, Mekos,&Valente,2001),to name just a few.
The use of network analysis in community sci-ence grew out of the work on social support.So-cial support was thought to have a positive effect on a wide variety of important individual character-istics and behavior,including coping,bereavement, suicide,general physical and mental health(Cohen,Underwood,&Gottlieb,2000).Early on,social sup-port tended to be measured by simply asking indi-viduals to rate how much support they received from others.A network analysis approach to social sup-port became attractive as community scientists dis-tinguished between perceived social support and the relational or structural aspects of social support.That is,researchers began to‘move beyond the individ-ual’(Felton&Shinn,1992)and see that social sup-port was not just a psychological characteristic but an outcome of being embedded in a social network of friends,family,and other support providers.
We started seeing studies that had participants identifying their own support networks.This allowed investigators to determine the size of the supportive networks,distinguish support by source,and even get some very simple measures of network structure.For example,in an innovative study of social support sat-isfaction,Stokes(1983)had each participant identify up to20people who were important in their lives and with whom they had monthly contact.After identify-ing the network member,each participant drew lines connecting each pair of people in their lists who were signi?cant in each other’s lives.This allowed Stokes
194
Luke
https://www.doczj.com/doc/128189081.html,parison of ego-centric friendship network to complete
friendship network.
to not only look at source of support and satisfaction with support,but to also calculate the network mea-sures of size and density.Density is the proportion of ties observed in a network relative to the total possi-ble number of ties.
This type of network is known as an ego-centric network,because it is determined solely by the per-spective of a single member.To see the limitations of ego-centric networks,consider Fig.5.On the top is the type of network that you could get if you asked Alina who her friends were,and then asked her if her friends were friends with each other.However, this ego-centric network may not be representative of any real,complete friendship network.If you were to observe the children in action,or talked to all of the children,you might?nd out that you need to add two more members to the network(see bottom of Fig.5).Becca and Bobby are not directly friends of Alina,but they are friends with Alina’s friends.The characteristics of complete networks may or may not be similar to the ego-centric network.In particular, only direct relationships are examined in ego-centric networks.For example,you could not examine the indirect friendship support that Alina might receive from Bobby and Becca using only an ego-centric net-work.
The limitations of ego-centric networks are an-other illustration of Barton’s‘meatgrinder survey ap-proach.’If social networks are important because they in?uence people,then we need to observe and analyze complete networks that are not de?ned from the perspective of one member.Furthermore,most of the powerful network analysis techniques are only usable with complete networks.
Community scientists have started to recognize the utility of moving beyond ego-centric networks to examine complete networks,especially in the study of organizational relations and behavior.In an early classic example,Tausig(1987)used network meth-ods as an assessment tool for a community mental health service system in a New York https://www.doczj.com/doc/128189081.html,ing the presence or absence of interactions between all 45mental health service agencies,Tausig was able to identify different types of“cracks”in the service system,including missing links and con?icted links.
More recently,Pennie Foster-Fishman, Deborah Salem,and their colleagues have used network methods in a series of their organizational studies(Foster-Fishman,Salem,Allen,&Fahrbach, 1999,2001;Salem,Foster-Fishman,&Goodkind, 2002).In a particularly good example of using network tools on a complete network,they looked at the exchange relations within and between two interorganizational alliances within a single county. By collecting data on the complete interorganiza-tional network,the authors were able to determine that organizational membership on both alliances was associated with a broader variety of exchange relations.
For a more detailed network example,consider Fig.6.These data come from an evaluation study of10comprehensive state tobacco control programs conducted from2002to2004(Krauss,Mueller,& Luke,2004).In conducting the evaluations we took a systems perspective(Best et al.,2003),and viewed a state tobacco control program not as one activity di-rected by one lead agency,but rather a set of interre-lated activities coordinated by an organized network of tobacco control agencies and organizations.The network analysis data were collected to help us un-derstand the structure of these complex state tobacco
Getting the Big Picture in Community Science:Methods that Capture Context
195
https://www.doczj.com/doc/128189081.html,parison of contact networks for two states with divergent?nancial and political climates.
control programs,to identify other state characteris-tics related to program structure,and to determine if changes in program funding and political support were associated with changes in program structure.
Fig.6presents the contact networks for Indi-ana and Oklahoma as they existed in2002.Each of the nodes represents one of the core tobacco control groups in the state.For example,in Indiana AHA is the American Heart Association,DOH is the Indi-ana Department of Health,and ITPC is Indiana To-bacco Prevention and Cessation,the lead agency for the state program.Two groups are connected by a line if they have frequent contact,de?ned as contact at least once a month via meetings,phone,or e-mail.
One of the things that is apparent from these networks is that the individual agencies vary in how “central”they are in the communication structure. For example,the Boys and Girls Club of Indiana has monthly contact with only three other groups in the state,whereas ITPC meets monthly with every mem-ber of the tobacco network.Thus,ITPC is more cen-tral in the network,and can be interpreted as be-ing a gatekeeper,or having more control over the ?ow of information.This centrality is measured using Freeman’s betweenness index(Wasserman&Faust, 1994);and is indicated in these graphs with color cod-ing.Nodes with the highest degree of betweenness centrality are black.
Freeman’s actor betweenness index is de?ned as follows—for any individual actor i,the actor’s cen-trality is the proportion of times that actor i is in-cluded in the geodesics of actors j and k,summing across all j,k pairs.A geodesic is the shortest net-work path between any two actors.Thus,between-ness centrality measures how often an individual ac-tor i is involved in the communication between other pairs of actors in the network.Betweenness central-ity can range from0to1.The individual actor be-tweenness scores can be combined into an aggregate network centralization https://www.doczj.com/doc/128189081.html,work centralization can range from0,when all members have the same betweenness scores,to1(or100%)when one mem-ber has maximum betweenness,and all other mem-bers have no betweenness.High-network central-ization scores indicate hierarchical communications structures,whereas low-centralization scores indicate ?at communication structures.
We have used this centrality information to as-sess the extent to which state?nancial and politi-cal climates in?uence the structure of state tobacco control networks.States with positive climates are those that have high levels of per capita spending on tobacco control,have high cigarette tax rates, have numerous tobacco control“champions”in the state,have actively supportive governors and legisla-tors,and have a low tobacco industry presence.In-diana and Oklahoma’s networks are presented here because they have remarkably different climates.In 2002Indiana had a very positive climate for tobacco control,while Oklahoma’s was very challenging.
The pattern we observed across the states that we evaluated was that states who had a strong posi-tive climate tended to have communication networks that were relatively hierarchical,with the lead agency always being the most central actor in the network. Indiana’s network is a good example of this.ITPC is the most central node,and the overall network centralization index of22.7%indicates a moderately
196
Luke
Fig.7.Relationship of climate score to network centrality for10 state tobacco control programs in2002.(Note:the lowess line is a smoothed local regression line(Cleveland,1993)). hierarchical structure.States that had challenging cli-mates,on the other hand,tended to show a differ-ent pattern.The communication structure was?at-ter,and the lead agency was not always the most central member.Oklahoma illustrates this—TUPS is the lead agency,but two other agencies,the Latino Agency and the American Cancer Society,also are highly central.Also,the centrality index for Okla-homa is only6.6%,indicating a?at communication structure.Fig.7shows the relationship between state climate and centrality scores for all10states;as cli-mate improves,the centrality scores get higher,indi-cating greater centralization of communication.We are interpreting this?nding as suggesting that state networks are adapting to their environments.When times are good,a lead agency takes a central role in guiding activities and distributing?nancial resources. When times are not as good,there is less money?ow-ing into the program,and other non-lead agencies step up and become more actively involved.
Network analysis is the only tool that can pro-vide this type of structural information.From a pol-icy perspective,this information is important because it can help us understand how to most effectively sustain important health and social programs even when the environment is challenging.Although net-work analysis requires a very different“non-survey”approach to data collection and analysis,it is a very useful component of a contextual toolbox for a com-munity scientist.
Cluster Analysis
The bulk of the methods most often used by community scientists work by telling us information about how variables are related to each other(e.g., correlations,regression,factor analysis,etc.).How-ever,there is also a class of analytic tools that works by telling us how cases are related to each other. These methods include cluster analysis,multidimen-sional scaling,network analysis,and various aspects of data mining.
A little over a decade ago,Rapkin and I sug-gested that cluster analysis should be of interest to community psychologists because it could reveal di-versity and heterogeneity within our data(Rapkin &Luke,1993).Since then,cluster analysis has been used sporadically by community scientists;for exam-ple,clustering was used in three empirical articles in AJCP during2002and2003(Fleishman et al.,2003; Salem et al.,2002;Zapert,Snow&Tebes,2002). Cluster analysis works by grouping cases(often peo-ple)based on their similarities and dissimilarities across any number of variables.Cluster analysis is an exploratory technique that can be used to reveal un-known heterogeneity.This stands in contrast to tra-ditional modeling where heterogeneity is handled by using covariates to control or adjust for these indi-vidual differences.Controlling heterogeneity in this way has two problems.First,controlling heterogene-ity implies that the individual differences are not of scienti?c interest.Data heterogeneity may,in fact,be the most interesting part of the study.Second,this approach can sometimes lead to a type of method-ological laziness,where certain covariates(e.g.,age, SES,race,gender)are always included in a model because past studies have always included these co-variates.Controlling heterogeneity is a way to decon-textualize the data.By removing important individ-ual differences,we are once again adopting Barton’s meatgrinder approach.
What may not be as clear is that cluster analysis can be a useful tool for uncovering and describ-ing contextual patterns.By grouping cases together on the basis of patterns of similarity,cluster anal-ysis can be used to uncover naturally occurring groupings,social structures,or social types.Whereas network analysis and GIS methods are often used when the contexts are already de?ned(e.g.,the classroom or the neighborhood),cluster analysis can be used to discover new examples of social contexts.
In a notable example of this,Kuhn and Culhane (1998)used cluster analysis on administrative data on public shelter use by single adults in New York and Philadelphia.They were able to identify three clear types of homeless persons on the basis of patterns
Getting the Big Picture in Community Science:Methods that Capture Context197
of shelter usage.Transitional homeless were persons who entered the shelter system once and stayed for only a short period.The episodically homeless are persons who are frequently in and out of the shel-ter system.Finally,the chronically homeless used the shelter system less frequently,but stayed for much longer periods of time.Not only did these?nd-ings counter stereotypes of homelessness(the tran-sitional group made up80%of shelter use,and the chronic group constituted only10%),but they have in?uenced subsequent social policy at the federal level(Trutko&Barnow,2003).The new typology of homelessness identi?ed by Kuhn and Culhane is a useful description of the context of homelessness that helps us understand the interplay between poverty, housing,and social services.
Another recent example from AJCP shows the advantages of cluster analysis.In a longitudinal study of adolescent substance use,Zapert,Snow,and Tebes(2002)used cluster analysis to identify sub-groups of teenagers on the basis of their patterns of drug use from6th to10th grade.They were able to identify six distinctive clusters of substance users.These six clusters are shown in Fig.8,which is based on the data presented in Table I in the original ar-ticle.The graphs show that the clusters capture the complexity of teen substance use,with distinctive patterns across time and across different types of substances.These patterns may represent different classes of drug usage patterns.As such they repre-sent contextual information that may have implica-tions for community practice and policy.In particu-lar,interventions can be focused more speci?cally on the right type of high-risk adolescent at the right time in their developmental trajectory given these types of data.
Incidentally,this type of graph uses the tech-nique of“small multiples”(Tufte,1983)to facilitate interpretation of the patterns across clusters,time, and substance.It is my experience that community and social scientists tend to use graphics merely to report statistical results,such as regression lines or line-plots of ANOVA interaction effects.This leads to graphs based on a small number of data points,or, in Tufte’s terminology,a low“data-ink ratio.”The power of graphics lies in their ability to reveal
data Fig.8.Cluster pro?les of teenager substance use patterns.Based on results presented in Zapert,Snow,and Tebes(2002).
198Luke
complexity.Fig.8is a moderately complex graphic with three dimensions(i.e.,cluster,grade,and sub-stance)and is based on72data points.More impor-tantly,it is much easier for the reader to see and un-derstand the exciting structure of the data using this type of graph,than from a data table that simply lists the numbers.
DISCUSSION
In this paper I have argued that the uncritical use of traditional analytic methods can be inconsis-tent with core values and principles of community science.Speci?cally,traditional methods of the kind that Kelly refers to ignore or distort the diversity and contextual embeddedness of our data.
Toward Methodological Consilience
in Community Science
Consilience has been de?ned as a“...measure of how much a theory explains,so that we can use it to tell when one theory explains more of the ev-idence than another theory”(Thagard,1988).Thus, consilience is an epistemological construct that can be used to judge the adequacy of scienti?c theories. We can extend this idea to examine the adequacy of scienti?c methods for explaining the data.That is, methodological consilience is a measure of how much an analytic approach can explain.
A statistician usually thinks of a statistical model as a means for describing the variability in a dataset:
Data=Modeled variability+Unmodeled variability
Viewed this way,a model is successful when the Modeled variability is large relative to the Unmod-eled variability.However,Tukey suggested a differ-ent way to think about modeling(Hoaglin,Mosteller, &Tukey,1983):
Data=Smooth+Rough
Here,the purpose of modeling is to explore interest-ing structures and patterns(i.e.,smooth)in the data. Also,modeling is viewed more iteratively—what is considered“rough”in one analysis may be the sub-ject of another analysis to reveal previously unexam-ined patterns hidden in the rough.
By adopting Tukey’s approach to modeling,we can see more clearly the relevance of methodologi-cal consilience for community science.Although he did not use these same terms,Roger Barker(1968) clearly understood that much of the“smooth”part of behavioral data that we want to understand as com-munity scientists is accounted for by extra-individual settings,processes,and structures.
By using a wider array of statistical techniques such as cluster analysis,multilevel modeling,GIS, network analysis,and others,analysts are more likely to be able to explore and understand the complex data patterns of interest to community scientists.If context is a core value of community science,then our methods will be more consilient to the extent that they allow analysis of contextual effects.
Another way to frame this is to consider a lesson taught to me by my advisor in graduate school,Julian Rappaport.I remember a conversation in his of?ce where he told me that a good community scientist needed three things:a phenomenon of interest,a the-ory,and a set of research methods.His purpose at the time,as far as I can recollect,was to gently remind a young graduate student who was increasingly in-terested in quantitative analysis that community psy-chology at its heart was always about something real. Community science,in other words,starts with the phenomenon of interest,or the social problem that the scientist wants to study and change.After many years of taking his lesson to heart,I would like to sug-gest a corollary—in order for our work to have mean-ing and effect real change,we need to make sure to use tools that are consistent with our values,and that can capture the big picture of community context. ACKNOWLEDGMENTS
I would like to thank Jenine Harris for her help with collecting and coding the survey data used here. Thanks also to Nancy Mueller,Beth Shinn and the reviewers for comments that were extremely helpful in revising this paper.
REFERENCES
Alexander,C.,Piazza,M.,Mekos,D.,&Valente,T.(2001).Peers, schools and adolescent cigarette smoking.Journal of Adoles-cent Health,29,22–30.
Barker,R.G.(1968).Ecological psychology:Concepts and methods for studying the environment of human behavior.
Stanford,CA:Stanford University Press.
Best,A.,Moor,G.,Holmes,B.,Clark,P.I.,Bruce,T.,Leischow, S.,Bucholz,K.,&Krajnak,J.(2003).Health promotion dissemination and systems thinking:Towards an integrative model.American Journal of Health Behavior,27,S206–S216.
Getting the Big Picture in Community Science:Methods that Capture Context199
Bybee,D.I.,&Sullivan,C.M.(2002).The process through which an advocacy intervention resulted in positive change for bat-tered women over time.American Journal of Community Psy-chology,30,103–132.
Carver,S.(2001):Participation and geographic information:A po-sition paper.ESF-NSF Workshop on Access to Geographic Information and Participatory Approaches using Geographic Information.Spoleto,December2001.
Cleveland,W.(1993).Visualizing data.Summit,NJ:Hobart Press. Cohen,S.,Underwood,L.,&Gottlieb,B.H.(2000).Social support measurement and intervention:A guide for health and social scientists.New York,Oxford University Press
Coulton,C.,Korbin,J.,&Su,M.(1999).Neighborhoods and child maltreatment:A multi-level study of resources and controls.
Child Abuse&Neglect,23,1019–1040.
Degenne,A.,&Fors′e,M.(2004).Introducing social networks.
London:Sage.
Duncan,C.,Jones,K.,&Moon,G.(1998).Context,composi-tion and heterogeneity:Using multilevel models in health re-search.Social Science and Medicine,46,97–117.
Felton,B.J.,&Shinn,M.(1992).Social integration and social sup-port:Moving social support beyond the individual level.Jour-nal of Community Psychology,20,103–115.
Fleishman,J.A.,Sherbourne,C.D.,Cleary,P.D.,Wu,A.W., Crystal,S.,&Hays,R.D.(2003).Patterns of coping among persons with HIV infection:Con?gurations,correlates,and change.American Journal of Community Psychology,32, 187–204.
Foster-Fishman,P.G.,Salem,D.A.,Allen,N.A.,&Fahrbach, K.(1999).Ecological factors impacting provider attitudes to-wards service delivery reform.American Journal of Commu-nity Psychology,27,785–816.
Foster-Fishman,P.G.,Salem, D. A.,Allen,N. A.,& Fahrbach,K.(2001).Facilitating interorganizational collab-oration:The contributions of interorganizational alliances.
American Journal of Community Psychology,29,875–905.
Freeman,L.C.(2004).The development of social network analy-sis:A study in the sociology of science.Vancouver,Canada: Empirical Press.
Goldstein,A.O.,Peterson,A.B.,Ribisl,K.M.,Steckler,A., Linnan,L.,McGloin,T.,&Patterson,C.(2003).Passage of100%tobacco-free school policies in14North Carolina school districts.Journal of School Health,73,293–299. Granovetter,M.S.(1973).The strength of weak ties.American Journal of Sociology,78,1360–1380.
Haining,R.(2003).Spatial data analysis:Theory and practice.
Cambridge:Cambridge University Press.
Hedeker,D.,McMahon,S.D.,Jason,L.A.,&Salina,D.(1994).
Analysis of clustered data in community psychology:With an example from a worksite smoking cessation project.American Journal of Community Psychology,22,595–615.
Hirsch,B.J.,Mickus,M.,&Boerger,R.(2002).Ties to in?uen-tial adults among black and white adolescents:Culture,social class,and family networks.American Journal of Community Psychology,30,289–304.
Hoaglin,D.C.,Mosteller,F.,&Tukey,J.W.(1983).Understand-ing robust and exploratory data analysis.New York:Wiley. Hox,J.(2002).Multilevel analysis:Techniques and applications.
Hillsdale,NJ:Erlbaum.
Kelly,J.G.(2002).The spirit of community psychology.American Journal of Community Psychology,30,43–64.
Kelly,J.G.(2003).Science and community psychology:Social norms for pluralistic inquiry.American Journal of Community Psychology,31,213–217.
Kivim¨aki,M.,Vahtera,J.,Elovainio,M.,Pentti,J.,&Virtanen, M.(2003).Human costs of organizational downsizing:Com-paring health trends between leavers and stayers.American Journal of Community Psychology,32,57–68.Krauss,M.J.,Mueller,N.H.,&Luke, D. A.(2004)In-terorganizational relationships within state tobacco control networks:A social network analysis.Preventing Chronic Disease,1,[serial online].Available from:URL:http:// https://www.doczj.com/doc/128189081.html,/pcd/issues/2004/oct/040041.htm.
Kuhn,R.,&Culhane,D.P.(1998).Applying cluster analysis to test a typology of homelessness by pattern of shelter utiliza-tion:Results from the analysis of administrative data.Ameri-can Journal of Community Psychology,26,207–232.
Lin,N.(2000).Social capital.Cambridge:Cambridge University Press.
Luke,D.A.(2004).Multilevel modeling.Thousand Oaks,CA: Sage.
Luke, D. A.,&Krauss,M.(2004).Where there’s smoke there’s money:Tobacco industry campaign contributions and U.S.Congressional voting.American Journal of Preventive Medicine,27,363–372.
Luke,D.A.,Esmundo,E.,&Bloom,Y.(2000).Smoke signs:Pat-terns of tobacco billboard advertising in a metropolitan re-gion.Tobacco Control,9,16–23.
Mankowski,E.,Humphreys,K.,&Moos,R.H.(2001).Individ-ual and contextual predictors of involvement in twelve step self-help groups after substance abuse treatment.American Journal of Community Psychology,29,537–563.
Mark, D.M.,Chrisman,N.,Frank, A.U.,McHaf?e,P.
H.,&Pickles,J.(2004).The GIS history project.Re-
trieved February2004,from https://www.doczj.com/doc/128189081.html,/ ncgia/gishist/bar harbor.html.
Mintz,B.,&Schwartz,M.(1981).Interlocking directorates and interest group formation.American Sociological Review,46, 851–869.
Of?ce of Behavioral and Social Sciences Research.(2001).
Toward higher levels of analysis:Progress and promise in research on social and cultural dimensions of health.
NIH Publication No.21-5020.Available online at:http:// https://www.doczj.com/doc/128189081.html,/news/publications.html.
Rapkin,B.,&Luke,D.A.(1993).Cluster analysis in commu-nity research:Epistemology and practice.American Journal of Community Psychology,21,247–277.
Revenson,T. A.,D’Augelli, A.R.,French,S. E.,Hughes,
D.,Livert,D.,Seidman,
E.,Shinn,M.,&Yoshikawa,H.
(Eds.).(2002).Ecological Research to Promote Social Change: Methodological Advances from Community Psychology.New York:Plenum.
Rothenberg,R.B.,Potterat,J.J.,Woodhouse,D.E.,Muth,S.Q., Darrow,W.W.,&Klovdahl,A.S.(1998).Social network dy-namics and HIV transmission.AIDS,12,1529–1536. Salem,D.A.,Foster-Fishman,P.G.,&Goodkind,J.R.(2002).
Adoption of innovation in collective advocacy organiza-tions.American Journal of Community Psychology,30,681–710.
Sarason,S. B.(2003).The obligations of the moral-scienti?c stance.American Journal of Community Psychology,31,209–211.
SCRA.(2004).The SCRA mission.Retrieved February2004,from https://www.doczj.com/doc/128189081.html,/divisions/div27/goals.html.
Shinn,M.,&Rapkin,B.D.(2000).Cross-level research without cross-ups in community psychology.In J.Rappaport&E.Sei-dman(Eds.),Handbook of community psychology(pp.669–693).New York:Kluwer Academic.
Snijders,T.A.B.,&Bosker,R.J.(1999).Multilevel analysis:An introduction to basic and advanced multilevel modeling.Lon-don:Sage.
Stokes,J.P.(1983).Predicting satisfaction with social support from social network structure.American Journal of Commu-nity Psychology,11,141–152.
Tausig,M.(1987).Detecting“cracks”in mental health service sys-tems:Application of network analytic techniques.American Journal of Community Psychology,15,337–351.
200Luke
Thagard,P.(1988).Computational Philosophies of Science.A Bradford Book.Cambridge:MIT Press.
Trutko,J.,&Barnow,B.(2003).Core performance indicators for homeless-serving programs administered by the U.S.Depart-ment of Health and Human Services:Final Report.Available online:https://www.doczj.com/doc/128189081.html,/hsp/homelessness/perf-ind03. Tseng,V.,Chesir-Teran, D.,Becker-Klein,R.,Chan,M.L., Duran,V.,Roberts,A.,&Bardoliwalla,N.(2002).Promotion of social change:A conceptual framework.American Journal of Community Psychology,30,401–428.
Tufte,E.R.(1983).The visual display of quantitative information.
Cheshire,CN:Graphics Press.
Valente,T.W.(1995).Network models of the diffusion of innova-tions.Cresskill:Hampton.
Wasco,S.M.,Campbell,R.,&Clark,M.(2002).A multiple case study of rape victim advocates’self-care routines:The in?u-
ence of organizational context.American Journal of Commu-nity Psychology,30,731–760.
Wasserman,S.,&Faust,K.(1994).Social network analysis: Methods and applications.Cambridge:Cambridge University Press.
Wicker,A.W.(1992).Making sense of environments.In W.B.
Walsh&K.H.Craik(Eds.),Person-environment psychol-ogy:Models and perspectives(pp.157–192).Hillsdale,NJ:Erl-baum.
Wieczorek,W.F.,&Hanson,C.E.(1997).New modeling methods: Geographic information systems and spatial analysis.Alcohol Health and Research World,21,331–339.
Zapert,K.,Snow, D.L.,&Tebes,J.K.(2003).Pat-terns of substance use in early through late adolescence.
American Journal of Community Psychology,30,835–852.
福建省高等教育自学考试应用心理学专业（独立本科段） 《心理学研究方法》课程考试大纲 第一部分课程性质与目标《心理学研究方法》是福建省高等教育自学考试应用心理学专业（独立本科段）的一门专业基础必修课程，目的在于帮助考生了解和掌握心理学研究的理论基础和主要方法，检验考生对心理学研究理论基础与主要方法，检验考生对心理学研究方法的基本知识和主要内容的掌握水平与应用能力。 心理学研究的对象是心理现象。它的研究主题十分广泛：即涉及人的心理也涉及动物的心理；即涉及个体的心理也涉及群体的心理；即涉及有意识的心理也涉及潜意识的心理；即涉及与生理过程密切相关的心理也涉及与社会文化密切相关的心理。心理学研究是一种以经验的方式对心理现象进行科学探究的活动。由于心理学的研究方法是以经验的或实证的资料为依据的，因而使心理学与哲学相区别，也与人文学科相区别。 设置本课程的具体目的要求是，学习和掌握心理学研究方法的基本理论和基本技能，将有助于学生们理解心理学的基本概念、基本原理和基本理论。理解心理学家在探索心理与行动时所做的一切，有助于考生将来为心理学的发展做出有益的贡献。 第二部分考核内容与考核目标 第一编心理学研究基础 第一章心理学与科学 一、学习目的与要求 通过本章学习，要求考生了解心理学的性质，了解心理学科学研究的方法、特征及基本步骤，理解心理学研究的伦理问题和伦理规范。 二、考核知识点与考核目标 1、识记： （1）心理学的含义； （2）心理学科学研究的特征：系统性、重复性、可证伪性和开放性； （3）知情同意。 2、领会： （1）一般人探索世界的常用方法； （2）心理学研究主要包含哪几个步骤； （3）科学研究的开放性主要表现在哪几方面； 3、应用： （1）根据科学研究的特征来分析某些心理学的研究； （2）心理学研究的伦理问题及以人为被试的研究的伦理规范来分析是否可以在心理学研究中使用欺骗的方法。
心理学研究方法 一、科学研究的特征：系统性、重复性、可证伪性和开放性。 二、以人为被试的研究的伦理规范：知情同意、保护被试、保障退出自由与保密 三、研究课题的来源：（1）对日常生活的观察。（2）实际的需要。（3）理论。（4）技术发展的推动。 四、选题的原则：（1）课题是可行的。（2）课题是清楚的。（3）课题是有价值的。（4）课题是符合道德的。 五、取样的类型：（一）随机取样： （1）简单随机取样：又称纯随机取样，对研究总体单位不进行任何分组排列，只按随机原则直接从总体中抽取一定的样本，以使总体的每一个样本都有被同等抽取的可能性。常用抽签法和随机表取样法。（2）分层随机取样：按照总体已有的某些特征，将总体分成几个不同的部分，然后在每一个层或子总体中进行简单随机取样。优点：（1）样本的代表性强。（2）取样更灵活。（3）参数估计更加准确。 （3）系统取样：是以某种系统规则来选择样本的方法。优点：比较简单，易于实施，在实际生活中应用较多。缺点：样本对总体代表性不够，名单中个体的排列方式可能使系统取样产生问题。 （4）聚类取样：将总体按照某种标准划分为若干子群体，每个子群体作为一个取样单位，用随机的方法从总体中抽取子群体，将抽中的子群中的所有单位合起来作为总体的样本。优点：调查单位比较集中，调查工作的组织和进行比较方便。缺点：如果调查单位在总体上分布
不均匀，准确性相对随机取样要差。 （5）多段取样：适用于总体范围大、对象层次多的研究，能有效的节省调查所需的人力及费用。缺点：由于每级取样时都会产生误差，所以误差较大。 （二）非随机取样： （1）方便取样：研究者选择在方便的时间和方便的地点将所遇到的人作为研究样本的方法 （2）立意取样：研究者根据自己的经验，如对总体构成要素和研究目标的认识，主观判断、选取可以代表总体的个体作为样本。 （3）定额取样：按特定的标准将总体中的个体分成若干类或层，然后在各层中取样。 （4）滚雪球取样：先从总体中合适的调查对象开始进行调查，再通过他们得到更多的调查对象，如同滚雪球一样。 六、心理学研究中的主要变量： 1、变量：指在性质上或数量上可以变化、可以取两个或两个以上值的事物的特征。 2、刺激变量、机体变量和反应变量： （1）刺激变量：简称S变量，是指已知的对有机体的反应发生影响的刺激条件，包括研究者可以操纵和控制的环境特征。有机体接受的刺激具有多种形式：自然性刺激和社会性刺激，具体性刺激和抽象性刺激，外部刺激和内部刺激。 （2）机体变量：也称被试变量、属性变量。指被试本身对行为反应
《心理学研究方法》年月期末测验指导
————————————————————————————————作者：————————————————————————————————日期：
0297《心理学研究方法》2013年6月期末考试指导 一、考试说明 本课程为闭卷考试。考试题由四种题型组成，满分100分，考试时间为90分钟。 题型一：单项选择题（每题1分，20题，共20分） 题型二：简答题（每题10分，4题，共40分） 题型三：分析设计题（每题10分，2题，共20分） 题型四：综合设计题（每题20分，1题，共20分） 二、重点复习内容 第一章绪论 （一）心理学研究 1、知识的来源：权威人士、个人经验、注意凝聚、科学方法。 2、科学研究的特点和目的 （1）科学研究的特点：以客观事实为依据；有系统的理论框架和目的；具有一定的控制机制；有严密的分析。 （2）科学研究的目的：描述对象的状况；解释对象的活动过程；预测对象将来的发展；控制对象发展的方向 （二）构造主义 1、代表人物：冯特 2、采用实验内省法 （三）心理学研究方法分类 1、心理学研究的方法体系：哲学方法论、一般科学方法论、具体研究方法。 2、心理学研究的具体方法 （1）收集数据的方法：观察法、调查法、实验法、测量法 （2）分析数据的方法：数学方法、模型方法、逻辑方法 第二章心理学研究方法发展的新特点 （一）心理学研究方法的特点 1、研究背景的现场化 2、研究方法的综合化 3、研究手段的现代化： 1）出现大量新的研究技术与手段 2）建立了现代化实验室 3）计算机网络在心理学研究中大量应用 （二）计算机在心理学研究中的主要应用 1、对实验过程进行控制，如呈现刺激、控制其他仪器、对被试的反应进行自动记录 2、处理、分析数据 3、模拟心理过程 4、进行心理测验
心理学研究方法包含哪些内容？ 选择研究的问题→选择实验的方法→数据的分析与解释→研究报告的撰写 学习心理学研究方法的目标 ?做研究的消费者 能够更好地辨别媒体报道研究的质量和缺失的重要信息； 能够解读他人的研究成果，作出更有见解的个人和专业决策。 ?成为研究的生产者 学会如何去做一项研究，如何为心理学做出更多的贡献； 科学方法和非科学（一般）方法的特征和区别 研究问题的选择： 1.研究问题的来源。 2.社会和文化背景对问题选择的影响。 3.心理学伦理道德对问题选择的影响。 科学方法：目标： 描述——解释——预测——应用 描述性方法: 1观察法 2调查研究 3行为的掩蔽测量 实验研究: ?独立组设计 ?重复测量设计 ?复合设计 应用研究 ?个案研究法和单被试设计 ?准实验设计和项目评估 观察法_百度百科 观察法是指研究者根据一定的研究目的、研究提纲或观察表，用自己的感官和辅助工具去直接观察被研究对象，从而获得资料的一种方法。 非干预观察：
?描述正常发生的行为 ?建立实验室研究结果的外部效度 ?无道德和伦理上的特殊要求 干预观察 ?参与性观察 ?结构性观察 ?现场观察 一．☆参与性观察 ★无伪装的参与性观察常用来了解群体文化和行为； ★当研究者知道个体会因为被记录而改变行为时，通常使用伪装的参与观察；★观察不能被直接进行科学观察的行为和情境。 优点:研究者开始研究时不必使用特定的假设，因此能在研究过程中不受假设约束，研究者无先入为主的偏见，结果可靠。2.效果较好，参与。。更深入梗真实的了解，并且能更及时地发现新的信息。3.观察者始终参与被观察者的活动，因此不仅能收集动态信息，而且能收集静态的信息。 ☆参与性观察的的缺点 ★参与观察者必须意识到观察记录的客观性； ★观察者的行为对被研究者可能产生影响； ★被观察的群体越小，观察者的影响越大。 费时，费力，收集的资料琐碎且不易系统化。 二．☆结构化观察 ★用于记录自然观察法难以观察的行为； ★常用于临床和发展心理学研究； ★需要严格遵循相同的观察程序，控制变量。 ☆结构化观察的缺点 ★如每次观察不遵循相同的程序，则不同观察者很难获得相同的结论； ★未控制因素及未知变量可能对观察的行为产生重要影响。 三．现场观察 科学记录的有效性？ 内容的准确性和精确性→依赖对观察者所进行的培训→观察记录被独立观察者和其他研究 方法实证的程度 1观察者偏差：观察者对研究结果的期望会导致在行为记录时产生系统误差。 2观察者信度：不同的独立观察者针对同一观察所作记录的一致性程度 3时间抽样：研究者系统地或随机选择进行观察的时间间隔 4事件抽样：当研究者对那些不经常发生的事件感兴趣时，他们就会对行为
心理学有哪些主要的研究方法？这些方法的特点是 什么？ 种类：观察法、心理测验法、实验 法、个案法 特点： 观察法：在自然条件下进行，不为被观察者所知，他们的行为和心理活动较少或没有收到 “环境的干扰”。因此，应用这种方法有可能了解到现实的真实状 况。 心理测验法：用一套预先经过标准化的问题（量表）来测量某种心理品质的方法。为了保证 心理测验的信 度和效度，一方面要对某种心理品质进行深入的研究。另一方面在编制心理测量表时要注意严谨性和科学 性。 实验法：在实验中，研究者可以积极干预被试者的活动，创造某种条件使某种心理现象得以 产生并重复出 现。这是实验法和观察法的不同之 处。 个案法：是一种比较古老的方法。它是由医疗实践中的问诊方法发展而来的。个案法要求对 某人进行深入 而详尽的观察和研究，以便发现影响某种行为和心理现象的原因。由于个案法限于使用少数案例，研究结 果可能只适合于个别情况。 现代心理学的主要派别：经验主义、联想主义、构造主义、机能主义、行为主义、格塔 式、精神分析学派 什么叫神经元？它的基本功能是什 么？ 神经元即神经细胞，是神经系统结构和机能的单位。它的基本作用是接受和传送信息。 神经元是具有细长突起的细胞，它由胞体、树突和轴突三部分组 成。 试说明神经冲动传导的基本方式，什么是电传导和 化学传导？ 神经冲动：当任何一种刺激（机械的、热的、化学的或电的）作用于神经时，神经元就会由 比较静息的状 态转化为比较活动的状态，这就是神 经冲动。 神经冲动的传导（单个神经元内的传导）实际上是一个神经内局部产生的电位 差引起。 神经冲动的传递（两个或多个神经元之间）神经冲动由一个神经元到另一个神经元的传递，要经 过突触。 突触是神经元与另一个神经元之间的连接部位。 电传导：是指神经冲动在同一细胞内的传导。神经冲动沿着神经的运动，跟电流在导线内的运动不同。电 流按光速运动，每秒 30 万公里，而人体内神经兴奋每小时运 行的速度只有 3.2~320 公里。 化学传导：神经冲动在突触间的传递，是借助于神经递质来完成的。当神经冲动到达轴突末 梢时，有些突 触小泡突然破裂，病通过突触前膜的张口处将存储的神经递质释放出来。当这种神经递质经过突触间 隙后， 就迅速作用于突触后膜，并激发突触后神经元的分子受体（另一种化学物质），从而打开或关掉膜 内某些离 子通道，改变了膜的通透性，并引起突触后神经元的电位变化，实现神经兴奋的传递。这种以化学物质为媒介的突触传递，是脑内神经元信号传递的主要方式。 知觉人们通过感官得到了外部世界的信息。这些信息经过头脑的加工（综合与解释） 产生了对事物整体的认识，就是知觉。 知觉的对象与背景的关系是怎样？ 人在知觉客观世界时，总是有选择的把少数事物当成知觉的对象，而把其他事物当成 知觉的背景，以便更清晰地感知一定的事物和对象。知觉过程是从背景中分出对象的 过程。知觉的对象从背景中分离，与注意的选择性有关。当注意从一个对象转向另一 个对象时，原来的知觉对象就成为背景，而原来的背景便成为知觉的对象。知觉的对 象与背景不仅相互转化，而且相互依赖。人们知觉某一对象，不仅取决于对象本身的 特点，而且受对象所处背景的影响。在不同背景下，人们对同一对象的知觉可能是不 同的。知觉中对象和背景的关系，不仅存在于空间的刺激组合中，而且存在于时间系 列中。可见，发生在前面的知觉直接影像到后来的知觉，产生了对后续知觉的准备状 态，这种现象叫知觉定势。什么叫知觉的恒常性？知觉恒常性是怎样产生的？
第一章练习题 一、单项选择题： ( D )1、弗洛伊德的精神分析理论建立在________的基础之上。 A.观察法 B.实验法 C. 调查法 D.个案研究法( A )2、科学心理学之父是。 A．冯特 B．华生 C．马斯洛 D．弗洛伊德 ( C )3、以下说法中错误的是。 A.冯特、铁钦纳是结构主义的代表 B.弗洛伊德是精神分析学派的代表 C.魏特海墨是机能心理学派的代表 D.华生是行为主义心理学派的代表 ( B )4、在控制的条件下，系统的操作某种变量，研究该变量对其他变量所产生的影响，这种心理研究的方法是________。 A.观察法 B.实验法 C. 调查法 D.个案研究法( C )5、认为心理学的目的是研究有机体适应环境过程中心理的功能的心理学流派是。 A.结构主义 B.精神分析 C.机能主义 D.行为主义 ( B )6、被称作心理学的第三势力的是学派。 A. 行为主义 B. 人本主义 C. 精神分析 D. 认知心理学 ( C )7、能揭示变量之间因果关系的研究方法是。 A. 观察法 B. 实验法 C. 相关法 D. 访谈法 ( C )8、创建第一个心理科学体系的心理学家是。 A. 弗洛伊德 B. 詹姆斯 C. 冯特 D. 铁钦纳 ( B )9、操纵和控制变量的心理学研究方法是。 A观察法 B实验法 C个案法 D调查法 ( B )10、由实验者操纵的，被假定为可影响行为的可能原因的变量 是。 A．自变量 B. 因变量 C. 中介变量 D. 无关变量 ( C )11、用实验内省的方法分析各种经验是指，研究者操纵刺激（视觉的、听觉的、皮肤觉的等），使之有系统地变化，让被试根据自己的主观判 断作出反应（如按键）或报告自己对于某种刺激的感受。这种“内省法” 属于现代心理学的哪个流派？ A、功能主义 B、行为主义 C、结构主义 D、精神分析 ( C)12、认为“男性和女性的基本择偶标准（如年龄）在世界各种不同文化中具有明显的普遍性，根源在于早期人类男性和女性在繁殖和抚养方 面所面临的不同的适应性问题。”这一观点属于当代心理学取向中的 哪一个分支？ A、生理心理学 B、行为遗传学 C、进化心理学 D、认知神 经科学 ( A )13、由研究者直接观察记录研究对象的行为，从而了解事物的特征或规
心理学研究方法笔记 Document serial number【KKGB-LBS98YT-BS8CB-BSUT-BST108】
第一编心理学研究基础 第一章心理学与科学 1.心理学是一门兼有自然科学和社会科学的性质、具有不同分析单位以及 既重视内省也重视客观观察的学科。 2.一般人探索世界的常用方法归纳起来有常识、传统和权威三种。 3.系统性、重复性、证伪性和开放性是科学研究的四种典型特征。 4.心理学研究主要包括5个基本步骤：（1）选择课题和提出假设；（2） 设计研究方案；（3）搜集资料；（4）整理和分析资料；（5）解释结果以及检验假设。 5.科学研究的开放性主要表现在多视角、公开性、可争辩性以及科学研究 无禁区四个方面。 6.以人为被试的研究的伦理规范主要包括知情同意、保护被试以及保障退 出自由与保密。 第二章选题与抽样 1.心理学研究课题可以有多种来源。主要的来源是：（1）对日常生活的观 察；（2）实际的需要；（3）理论；（4）研究文献的提示；（5）技术发展的推动等；也可以是这些来源的某种结合。 2.好的研究问题应该具备4个基本特征：（1）问题是切实可行的；（2） 问题是清楚的；（3）问题是有意义的；（4）问题是符合道德的。 3.文献是记录、保存、交流和传播知识的一切材料的总称。心理学文献就 其保存方式可分为纸质文献和互联网文献。纸质文献的常见类型有图书、期刊、会议和学术论文等。
4.查阅文献常用参考文献查找法、检索工具法。检索工具法分为手工检索 和计算机检索。 5.取样和选题的关系密切，要依据课题的性质和研究结果的概括程度来选 择被试。取样的方法很多，根据是否按随机原则进行操作，取样可分为两类：非随机取样和随机取样。 6.简单随机取样又称纯随机取样，即对研究总体单位不进行任何分组排 列，只按随机原则直接从总体中抽取一定的样本，以使总体的每一个样本都有被同等抽取的可能性。分层随机取样，也称比率取样、分类取样和分组取样，是按照总体已有的某些特征，将总体分成几个不同的部分（每个部分叫一个层或一个子总体），然后在每一个层或子总体中进行简单随机取样。 7.系统取样，又称等距取样或机械取样，是以某种系统规则来选择样本的 方法。聚类取样，又称整群取样，是将总体按照某种标准（如班级、地区）划分为若干子群体，每个子群体作为一个取样单位，用随机的方法从总体中抽取子群体，将抽中的子群体中的所有单位合起来作为总体的样本。 8.多段取样，也称多级取样、分段取样，先将总体按照某种标准（如学 校、地区）分为若干层（组），称为第一层，然后从这一层中随机抽取出几个层作为第二层，依次类推，直至从最后一层中用随机取样的方式中抽取出一定数量的样本作为研究所需样本的取样方法。 9.方便取样：也称偶遇取样，研究者选择在方便的时间和地点将所遇到的 人作为研究样本的方法。立意取样，也称主观取样、判断取样，研究者根据自己的经验，如对总体构成要素和研究目标的认识，主观判断选取可以代表总体的个体作为样本。
二、 7.ANOVA中最简单的设计是(单向方差分析)。 1.Q分类通常要求客体数目在(60-120 )。 14.Q分类技术以人为分析单位，其重点在于分析(人际关系) 2.自然观察有许多变式，其中最常见的是抽样调查和( 个案研究)。 3.格式塔学派的创始人是(韦特墨)。 4.心理统计中用以表示分数离散度的主要指标有标准差和( 方差)等。 5.小样本研究范式的缺点之一是(研究的外部效度不易达到要求)。 6.控制变量是实验中保持恒定的某些潜在的(自变量)。 7.一致定位量表是以(心理物理测量)为基础。 8.访谈信度通常以(重测信度)作为评估指标。 2.以数据为基础，把数据反映的实际关系组织为理论原则的心理学理论构建类型是(.归纳)理论。 3.费希纳等人提出了( 心理物理法)，揭示了客观刺激量和主观感觉量之间的函数关系。 5.建立“实验学校”或“基地”，对同样对象反复多次的进行心理学或教育实验研究，容易产生 多重实验的干扰，从而削弱研究的(外部效度) 6.用来表示“测验结果能够预测人们将来行为的程度”的是(预测效度) 7.观察者在评级中出现的过宽的倾向，我们称之为(正宽大)效应。 8.“原始报告分析”指的是( 口语报告法) 9.统计功效是指在统计检验中不犯或避免犯( .Ⅱ)类错误的概率。 10.多元分析的两种不同的策略是( .标准分析和层次分析) 2.心理学研究中，外源变量相应于（自变量），内源变量相应于(因变量)。 3.小N设计的主要类型之一是(多重基线设计)。 )。4.平均差误法又叫(调整法 )。5.方差也称为变异，一组分数的方差正好是其标准差的(平方6.多元分析的基本方法之一是( 判别函数分析)。 8.等比量表的特点是(既有相等单位，又有绝对零)。 9.行为主义学派主张（实验方法），研究可观察、可测量、可控制的外显行为。 10.简单随机抽样法有两种主要的方式，它们是（.抽彩法和随机表取）样法． 1.避免严重损害内部效度，在研究中应特别注意控制，最有效的途径是采用（随机化程序）1 2.能够最有效运用实验控制策略的是（实验室实验） 14.问卷设计中的最高层次是（问卷量表的构思与目的） 15.两种常用的标准分数是标准九分数和（.z分数）
第1章心理学研究方法概论 一、单项选择题 1．下列心理学研究方法中可以确定因果关系的是（）。 A．自然观察法 B．相关法 C．实验法 D．个案法 【答案】C 【解析】实验的根本目的在于确定变量间的因果关系，因果关系是心理学研究的重点。因为实验对无关的额外变量进行了严格的控制，排除了其它的可能性，我们就可以确定自变量和因变量的关系。而其它的非实验方法都不能对环境进行有效控制，变量之间的关系很可能受到其它很多因素的影响，从而只能判定它们有关，但不能判定是否是因果相关。 2．普莱尔在三年里每天三次对其子进行长期观察，并辅之简单实验，完成其心理发展报告，从研究设计上看，属于（）。 A．横断设计 B．纵向设计 C．聚合交叉设计 D．混合设计 【答案】B 【解析】纵向设计是在一个比较长的时间内，对个体进行追踪，对其心理发展进行系统
的研究，因此也叫追踪设计。普莱尔采用的正是纵向研究。 3．在同一时间内对国内二十三个省市在校青少年思维发展的研究，属于（）。 A．横断设计 B．纵向设计 C．聚合交叉设计 D．双生子设计 【答案】A 【解析】横断设计是在同一时间内对某一年龄（年级）或几个年龄（年级）的被试的心理发展水平进行测查并加以比较。 4．聚合交叉研究的基本设计思想是在（）图式中分段进行（）。 A．横向设计纵向设计 B．双生子设计横向设计 C．纵向设计横向设计 D．纵向设计双生子设计 【答案】C 【解析】聚合交叉设计是将纵向设计和横断设计综合在一起的一种研究方法，其基本设计思想是在纵向设计图式中分段进行横断设计，取长补短，保留纵向设计和横断设计的优点，克服其缺点，既可以在较短的时间内了解各年龄阶段儿童心理发展的总体情况，也可以从纵向发展的角度了解儿童随年龄增长而出现的各种心理变化以及社会历史因素对儿童心理发展的影响。
心理学研究方法复习题 一、重要概念 1、研究的效度：即有效性，它是指测量工具或手段能够准确测出所需测量的心理特质的程度。 2、内部一致性信度：主要反映的是测验内部题目之间的信度关系，考察测验的各个题目是否测量了 相同的内容或特质。内部一致性信度又分为分半信度和同质性信度。 3、外推效度：实验研究的结果能被概括到实验情景条件以外的程度。 4、半结构访谈：半结构化访谈指按照一个粗线条式的访谈提纲而进行的非正式的访谈。该方法对访谈 对象的条件、所要询问的问题等只有一个粗略的基本要求，访谈者可以根据访谈时的实际情况灵活地做出必要的调整，至于提问的方式和顺序、访谈对象回答的方式、访谈记录的方式和访谈的时间、地点等没有具体的要求，由访谈者根据根据情况灵活处理。 5、混淆变量：如果应该控制的变量没有控制好，那么，它就会造成因变量的变化，这时，研究者选定 的自变量与一些没有控制好的因素共同造成了因变量的变化，这种情况就称为自变量混 淆。 6、被试内设计：每个被试接受接受自变量的所有情况的处理。 7、客观性原则：是指研究者对待客观事实要采取实事求是的态度，既不能歪曲事实，也不能主观臆断。 8、统计回归效应：在第一次测试较差的学生可能在第二次测试时表现好些,而第一次表现好的学生则 可能相反,这种情形称为统计回归效应.。统计回归效应的真正原因就是偶然因素变化导致的随机误差，以及仅仅根据一次测试结果划分高分组和低分组。 9、主体引发变量：研究对象本身的特征在研究过程中所引起的变量。 11、研究的信度：测量结果的稳定性程度。换句话说，若能用同一测量工具反复测量某人的同一种心理特质，则其多次测量的结果间的一致性程度叫信度，有时也叫测量的可靠性。 12、分层随机抽样：它是先将总体各单位按一定标准分成各种类型（或层）；然后根据各类型单位数与总体单位数的比例，确定从各类型中抽取样本单位的数量；最后，按照随机原则从各类型中抽取样本。13、研究的生态效度：生态效度就是实验的外部效度，指实验结果能够推论到样本的总体和其他同类现象中去的程度，即试验结果的普遍代表性和适用性。 14、结构访谈：又称为标准化访谈，指按照统一的设计要求，按照有一定结构的问卷而进行的比较正式的访谈，结构访谈对选择访谈对象的标准和方法、访谈中提出的问题、提问的方式和顺序、访谈者回答的方式等都有统一的要求。 15、被试间设计：要求每个被试者（组）只接受一个自变量处理，对另一被试者（组）进行另一种处理。
科学研究的特征：目的性、继承性、创新性、系统性、控制性 *心理学研究的特殊性？ ①研究对象特殊性； A心理学研究的对象是有意识、有心理的有机体——人，具有思想、情感、意志及特殊的气质、性格与学习能力；既是生物实体，也是社会实体；心理具有发展性和个体差异性。 B涉及的各种变量多且复杂，难以精确地解释和预测；不能对研究对象做出精确控制与操纵。C研究者和研究对象即主体与客体都是有意识、有目的的人类活动。 ②研究过程特殊性； A是主客体相互作用过程，研究对象控制与操纵欠精确；变量多且设计复杂、精细；易受社会政治因素的影响。 ③研究方法特殊性； A需要符合伦理性原则。 B“黑箱方法”，难以准确描述脑内心理活动的具体过程与变化。 C难以进行严格的控制实验，较难确认变量间的因果关系。 心理研究的目的：描述、解释、预测、控制 心理学研究的方法论原则：客观性原则、理论与实践相结合原则、发展性原则、系统性原则、伦理性原则、教育性原则
心理学研究方法分类： ①按数据形式分： A量化研究：研究涉及到可量化的变量，并且研究结果中变量之间的关系以一种数量化的方式来呈现。 B质化研究：研究涉及的是难以进行量化的材料，研究的结论也常常是描述性的。 ②按研究目的分： A因果研究B相关研究 ③按对变量的干预情况分： A真实验研究B准实验研究C非实验研究 心理学研究设计方法： 变量数据获得方法：
实验法研究心理现象标志：费希纳1860年《心理物理学纲要》 科学心理学诞生标志：威廉冯特1879德国莱比锡大学建立世界第一个心理学实验室 理论研究：它主要通过理论论证、列举材料、总结分析等方式对心理现象或规律提出新的见解。如青少年早恋心理分析、教师心理健康及调适、试论心理健康教育运作模式等。 描述研究：它主要通过观察、访谈、案例分析（个案）等方式获得事实材料（证据）以表明或证明关于心理现象与规律的看法。如课堂教学中师生互动的观察研究、有关心理健康教育的访谈研究、注意力不集中儿童的个案研究等。 实证研究：它主要通过测评、实验等方式获得数据材料以表明或证明有关心理现象与心理规律的看法。如不同类型学校学生自信心状况研究、教师期望改变影响中学生学习动机的实验研究等。 心理学研究方法的现状：多样共存 新特点：研究思路的生态化、研究方式的多分支化、研究方法的综合化、研究手段的现代化、研究结果的数学化、研究过程的计算机化 心理学研究基本环节： 1、问题的提出与课题的确定 2、文献的查阅与综述 3、研究类型的选择 4、变量的确定和控制 5、被试的取样 6、研究的实施 7、数据的收集和整理 8、结论的获得及其适用性 理论：一种能解释某些现象的具有逻辑关系的肯定陈述，是由一定的科学概念、概念间 的关系及其论证所组成的知识体系。即对一系列抽象变量之间可能关系的描述。 归纳：由事例出发形成理论 演绎：由理论出发形成假设 研究就是检验假设，修改理论的过程
第九章人格心理学研究方法主要内容观察法实验法问卷法投射法个案研究一观察法观察法有目的有计划地通过观察被试的外部表现来研究其心理活动的一种方法长期观察和定期观察全面观察和重点观察参与性观察和非参与性观察现场观察和人为场所情境观察二实验法实验法按研究目的控制或创设条件以主动引起或改变被试的心理活动从而进行研究的一种方法实验室实验指在特定的心理实验室里借助各种仪器设备严格控制各种条件以研究心理的方法现场自然实验是在日常生活条件下对某些条件加以适当控制或改变来研究心理的方法这一方法的实质就是把实验研究和日常活动结合起来一方面仍对实验条件有所控制另一方面又适当放松控制三问卷法高尔顿于1884年提出构成我们行为的品格是一种明确的东西所以应该加以测量他尝试通过记录心跳和脉搏的变化来测量情绪通过观察社会情境中人们的活动来评估人的性情脾气等人格特征这一切标志着科学地评估性格的开始克瑞普林 E．Krapelin 则是人格测验的先驱他最早将自由联想测验用于临床近几年来经过心理学家的努力已发展了许多科学地评估人格的技术并发表了具备优良特性的人格测验如明尼苏达多相人格问卷卡特尔16种人格因素问卷罗夏墨迹测验等等为评价人格提供了科学工具三问卷法问卷法问卷法多以自我报告的形式出现所以又称自陈量表通常采用文字叙述的方法列出一些问答题或一个命题后面有几种答案让受试者选择也可由
受试者根据自己的评判在不同评价等级上表明某种程度根据赋予不同评价等级程度的分值将原始数据统计处理后即得到评判结果最早的自陈量表是伍德沃斯 R．S．Woodworth 在第一次世界大战期间设计的个人资料调查表用于考察士兵对军队生活的适应性并淘汰军队中的情绪障碍者该调查表的问题涉及变态的恐怖反应强迫观念和强迫行为睡眠障碍以及其它身心症状等后来被奉为人格量表的蓝本自陈量表通常采用的题目形式有以下几种 1．是非式例如你曾经害怕自己发疯吗是否 2．折中是非式例如你喜欢户外活动吗是否不一定 3．选择式例如A当遇到失败时我会意志消沉 B 在大庭广众之下我会感到紧张 4．文字量表式例如我所喜欢的人大多是 A．拘谨缄默的 B-介于AC之间 C．善于交际的 5．数字量表式例如我担心考试失败 5 4 3 2 1 5代表经常4代表多次3代表偶尔2代表极少1代表从不自陈量表假设被试充分地了解自己并且能对题目作真实的回答这显然是靠不住的因为一个人不可能对自己的各个方面作全面而正确的观察同时在个人评估自己行为时往往会出现各种反应定势首先被试往往对社会赞许性强的题目作肯定回答对具有社会否定特征的题目作否定回答例如我从来不撒谎有些被试就会回答是其次还会出现反应的肯定定势极端定势谨慎定势及猜测定势等倾向例如被试会对题目大部分采用是是是"或否否否"的回答而不考虑内容如何这种反应方式有可能反映了个体的一部分个性特征
心理学研究方法复习 一、描述科学方法的两个重要特征。 科学是对客观事物及其运动变化规律的真理性认识，表现为系统变化的知识体系。科学有其特定的研究方法，即科学方法 怀疑的态度、客观的原则以及科学的程序是所有科学方法不可或缺的基本特征 两个重要特征,1、客观性原则是指研究者对待客观事实要采取实事求是的态度，要从客观事实中寻找和发现规律，并且从客观事实到研究结论的推理要建立在逻辑规则上。 2、科学的程序: 提出问题和假设 ?设计研究方案 ?实施研究并收集资料 ?处理分析资料 ?检验假设并作出结论 二、科学方法的四个目标是什么？简要描述科学方法每个目标相要实现的内容。科学方法的四个目标：描述、预测、解释和应用 一描述 ·心理上家试图描述事件和变量之间的关系；很多时候研究者用一般规律研究法和定量分析法。 描述指的是研究者用来定义、区分、分类以及把事件以及他们之间的关系归类的过程。 常规法则研究法：心理学家试图建立适用于多样性总体的广泛概括和一般规律。 平均水平可能描述了组内个体的表现，也可能没有描述组内个体的表现。 运用常规法则研究法的研究者承认个体之间存在较大差异，然而，他们试图强调相似性而非差异性。 有些心理学家，特别是戈登·奥尔波特认为独一无二的个体不能用平均值来描述，应使用个 人特质研究法，该方法的主要研究形式是个案研究法。 研究者会根据他们研究的问题来确定究竟用哪一种方法，有时两种方法都用。 二预测 相关关系允许心理学家对行为和事件做出预测，但不允许心理学家推断引起这些关系的原因。对事件及事件间关系的描述是预测的基础。 相关：当一个变量的分数能用来预测另一个变量的分数时，或当相同的人、事或物的两个不同的测量指标一起改变时。也就是说，当一个变量的一列分数趋向于和另一个变量的一列分数相关联时，此时的两列分数成为“共变”。 有效的预测并不在于知道为什么两列变量之间存在某种关系，如通过观察动物来预测地震。三解释 ·心理学家理解一个现象的原因需要满足三个条件：共变、时序关系，以及排除其他貌似真 实的可能原因。 ·实验方法就是研究者操纵自变量以确定其对因变量的影响，研究者可借此建立时间序列关 系并清晰地确定共变趋势。 ·如果研究中没有混淆的因素，那么就可以排除其他貌似真实的可能原因。 ·研究者尝试对研究结果进行推广，以使其研究发现可以描述不同的总体、情境和条件。
[06059] 心理学研究方法自学考试大纲 自学用书：《心理学研究方法》，王重鸣主编，人民教育出版社2001年2月第2版 第一章心理学研究方法的特点与发展 （一）识记 1．科学研究的特点2．科学研究的目的3．定性研究方法（1）个案分析法（2）文献综合法 4．信息论的定义5．控制论的定义6．系统和系统方法7．早期心理学的主要学派和研究方法 （二）领会 1．心理学研究方法的主要内容2．心理学研究方法的分类3．心理学研究的主要范式 （1）大样本范式（2）小样本范式 4．心理学研究方法发展的新特点5．计算机在心理学研究中的基本用途 （三）应用 1．举例说明科学研究的一般过程2．举例说明大样本范式在心理学研究中的应用3．举例说明小样本范式在心理学研究中的应用4．举例说明系统方法在心理学研究中的实际应用 第二章心理学研究的理论基础与思路 （一）识记 1．传统心理学研究中主要采取的三种理论思路2．项目反应理论3．理论的两种功能 4．外源变量5．一般递归系统6．自抑式函数方程7．理论边界 （二）领会 1．有关因果关系分析的5种理论2．心理学研究中因果关系分析的基本原则 3．相互作用心理学的基本观点4．心理学理论构建的类型与方式 （1）归纳理论（2）演绎理论（3）机能理论（4）模型 5．心理学理论评价中采用的标准6．因果模型的条件7．构建因果模型的基本步骤 （三）应用 1．举例说明因果关系分析的意义。2．举例说明理论思路在心理学研究中的重要性。 3．举例说明如何构建因果模型。 第三章研究设计与评价 （一）识记 1.自然观察 2.个案研究 3.抽样调查 4.相关研究 5.实验和准实验 6.发展研究的三种类型 7.心理学研究中常见的变(1)自变量(2)因变量(3)中介变量(4)干涉变量 8.变量的操作定义9.元分析的特点 (二)领会 1.对个案研究和抽样调查的评价 2.进行相关分析应注意的事项 3.实验的主要特点 4.发展研究的三种类型及其特点 5.研究的方法和过程 6.作出研究结论过程中应注意的问题 7.元分析的步骤 （三）应用 1．举例说明纵向研究。2．举例说明横切面研究。3．举例说明趋势研究。 4．举例说明元分析结果的解释和评价。 （一）第四章研究的取样 （一）识记 1．总体2．样本的代表性3．置换取样和非置换取样4．样本空间 5．随机化6．样本均值的分布7．均值的标准差误 （二）领会 1．几种主要的取样方法（1）简单随机取样法（2）分层随机取样法（3）系统取样法（4）聚类取样法（5）
1、准实验设计：由库克和坎贝尔提出。伴随现场实验的发展而发展起来。 →特点：介于真实验设计和非实验设计之间。从研究中对额外变量的控制程度出发，准实验设计差于真实验，但比非实验设计控制严格，一般准实验设计不容易随机挑选、分配被试，可能产生额外变量与自变量的混淆。另外，其外部效度不一定好。 →类型：①单组设计：时间序列设计、相等时间样本设计 ②多组设计：不相等组前后测设计、不相等组前后测时间序 列设计 2、实验研究： 定义：是在控制无关变量、操纵自变量的条件下研究心理现象的变化，从而确定变量之间关系的研究。对变量间的因果关系感兴趣。 特点：系统操纵或改变一个变量，观察这种操纵或改变对另一变量所造成的影响，并在此基础上揭示变量间的因果关系。 3、决定论：认为自然界和人类社会普遍存在客观规律和因果联系的理论。（不同于宗教的先决论） 4、被试变量：又称被试特性，一般作为自变量处理。 一是被试固有的。（如性别、婚姻状况、年龄、文化程度等） 二是暂时的。（如遭遇灾害否、吸毒否） 三是被试行为分类。（如喜欢早上锻炼还是喜欢晚上锻炼） 5、检查点：自变量的不同取值。 6、额外变量：会对结果产生影响，但是又不是研究者能够控制的变量。 7、混淆变量：无关变量的一种，该变量与自变量之间存在系统性关联，并对因变量产生影响的变量。这个影响的结果叫虚假效应。（PS.无关变量：与研究目的无关的变量） 8、因果关系研究：一个变量为操作变量，另一个为随自变量变化的因变量，可以揭示因果关系。 9、相关研究：两个变量都没有任何操作，只是调查、测量、观察的结果。如两个变量均为被试变量。（eg.与父母共处的时间与儿童的抑郁水平） 注：相关研究对个体差异感兴趣，而因果研究关心某类人的普遍心理规律；相关研究只研究机体间的方差，因果研究关心处理间的方差。 10、交互作用：反映的是两个或多个因素的联合效应。或者说是因素间相互影响和制约的关系。（当一个因素如何起作用受另一个因素的影响时，我们就认为存在二重交互作用。同理，当一个因素如何起作用受另两个因素的影响时，我们就认为存在三重交互作用。） 11、简单效应：指一个因素的水平在另一个因素某个水平的变异。（如，噪音大小与竞争对学习效果的影响，只考虑在竞争组噪音大小的简单效应） 12、简单简单效应：指一个因素的水平在另外两个因素的水平结合上的效应。（如，只考虑噪音强度在高难度和无竞争结合上的简单简单效应） 13、被试间设计：在不同被试之间比较。研究变量为被试变量时只能进行这种设计。 14、被试内设计：每个被试接受全部的水平或者水平结合的处理，自己和自己比较。 条件1 条件2
第二部分心理学研究方法 第一章导论 第一节心理学研究的特殊性与原则 1．心理学研究的特殊性 （1）心理学研究具有科学研究的四个基本特征：继承性、创新性、系统性、控制性 （2）心理学是介于自然科学和社会科学之间的科学，兼具二者的特点，这使得心理学具有以下几方面的特殊性：?探究对象与研究者的特殊性：心理学的研究对象（被试）和研究这同属一类型，都是人。 ?研究过程的特殊性：一方面研究对象要根据研究这的要求或实验控制作出反应，另外一方面，研究对象的反 映又反过来影响研究者的行为。实验者效应，皮格马利翁效应，霍桑效应 ?研究方法的特殊性：首先，任何实验处理、控制、操纵，都不能妨碍研究对象的心理健康发展，必须遵循人 道主义精神，符合伦理性原则，这研究会使某些研究的客观性受到影响。其次从现代科学的角度来看，目前心理学研究的方法多属“黑箱方法”，只能通过比较输入和输出的信息来推测心理过程。第三，心理学的研究由于缺乏严格控制的实验，因而难确认变量之间的因果关系。 2．心理学研究的原则 （1）客观性原则：是指研究者对待客观事实要采取实事求是的态度，不能歪曲事实，也不能主观臆测。 （2）系统性原则：要求研究者不仅要讲研究对象放在有组织的系统中进行考察，而且要运用系统方法去考察。（3）理论与实践相结合的原则：心理学研究中，理论与实践是辩证统一的。 （4）教育性原则：要求研究者在进行研究时要符合被试的身心发展规律，具有教育意义，有利于被试正常发展。（5）伦理性原则：应特别注意在创设情境时切忌采取违背伦理性原则的方法，如欺骗隐瞒、威胁恫吓。社会心理学中著名的实验“模拟监狱实验”因为了这一原则受到了广泛的批评。 除上述原则外，心理学研究者还应遵循发展性原则、科学原则和有效性原则。 第二节心理学研究的方法体系 1．心理学研究的哲学方法——唯物主义辩证法 2．心理学研究的一般科学方法论——系统方法论 所谓系统方法论是指“三论”（系统论、控制论和信息论）和“新三论”（耗散结构论、协同论、突变论）的基本思想和方法。 3.心理学研究的具体研究方法 在收集数据是可以采用观察法、调查法、实验法和测量法等；在分析数据时可采用数学方法、模型方法和逻辑方法等。 第三节心理学研究方法发展的历史与新趋势 1.心理学研究方法发展的历史 大致分为三个阶段：19世纪70年代之前为准备时期，主要是依靠不充分的观察和思辨的方法，1879年诞生了科学心理学；1879年至第二次世界大战广泛采用了定量研究的方法，心里试验得到迅速的发展；第二次世界大战以后，心理学从自然科学和社会科学借用了一些新的研究方法，如功能模拟方法、电生理方法是借用自然科学的，文献法、传记法是借用社会科学的。 2.心理学研究方法发展的新趋势 （1）研究思路的生态化：心理学研究出现了生态化的趋势，即强调在现实生活中、自然环境下研究个体与自然、社会环境中的各种因素的相互作用，从而揭示心理发展和变化的规律。生态化趋势使心理学研究更为客观、更接近自然、更接近真实，提高了研究成果的外部效度和生态效度。 （2）研究的多学科化：心理学往往需要从多个学科的角度进行研究，心理学多学科研究有两个水平：一种是心理学内部分支学科的协作；另一种是心理学与其他学科如哲学、神经科学等的协作。 （3）研究范围跨文化特点：研究者愈来愈重视在不同文化背景中给行为表现和心理发展的类似性和差异性，即相对于不同的文化背景，哪些心理现象及其发展规律具有普遍性，哪些则是特定的。 （4）研究方法的综合化：①主张使用多种方法去研究和探讨心理现象及其规律，以对不同方法所得到的结果进行比较、补充和验证；②强调大量采用多变量设计，以揭示心理活动各个方面的联系；③强调采用综合方式，如兼有纵向设计和横向设计特点的聚合式交叉设计；④注重定性和定量研究方法结合起来。
第一编心理学研究基础 第一章心理学与科学 1.心理学是一门兼有自然科学和社会科学的性质、具有不同分析单 位以及既重视内省也重视客观观察的学科。 2.一般人探索世界的常用方法归纳起来有常识、传统和权威三种。 3.系统性、重复性、证伪性和开放性是科学研究的四种典型特征。 4.心理学研究主要包括5个基本步骤：（1）选择课题和提出假设；（2） 设计研究方案；（3）搜集资料；（4）整理和分析资料；（5）解释结果以及检验假设。 5.科学研究的开放性主要表现在多视角、公开性、可争辩性以及科 学研究无禁区四个方面。 6.以人为被试的研究的伦理规范主要包括知情同意、保护被试以及 保障退出自由与保密。 第二章选题与抽样 1.心理学研究课题可以有多种来源。主要的来源是：（1）对日常生 活的观察；（2）实际的需要；（3）理论；（4）研究文献的提示；（5）技术发展的推动等；也可以是这些来源的某种结合。 2.好的研究问题应该具备4个基本特征：（1）问题是切实可行的；（2） 问题是清楚的；（3）问题是有意义的；（4）问题是符合道德的。 3.文献是记录、保存、交流和传播知识的一切材料的总称。心理学 文献就其保存方式可分为纸质文献和互联网文献。纸质文献的常见类型有图书、期刊、会议和学术论文等。
4.查阅文献常用参考文献查找法、检索工具法。检索工具法分为手 工检索和计算机检索。 5.取样和选题的关系密切，要依据课题的性质和研究结果的概括程 度来选择被试。取样的方法很多，根据是否按随机原则进行操作，取样可分为两类：非随机取样和随机取样。 6.简单随机取样又称纯随机取样，即对研究总体单位不进行任何分 组排列，只按随机原则直接从总体中抽取一定的样本，以使总体的每一个样本都有被同等抽取的可能性。分层随机取样，也称比率取样、分类取样和分组取样，是按照总体已有的某些特征，将总体分成几个不同的部分（每个部分叫一个层或一个子总体），然后在每一个层或子总体中进行简单随机取样。 7.系统取样，又称等距取样或机械取样，是以某种系统规则来选择 样本的方法。聚类取样，又称整群取样，是将总体按照某种标准（如班级、地区）划分为若干子群体，每个子群体作为一个取样单位，用随机的方法从总体中抽取子群体，将抽中的子群体中的所有单位合起来作为总体的样本。 8.多段取样，也称多级取样、分段取样，先将总体按照某种标准（如 学校、地区）分为若干层（组），称为第一层，然后从这一层中随机抽取出几个层作为第二层，依次类推，直至从最后一层中用随机取样的方式中抽取出一定数量的样本作为研究所需样本的取样方法。 9.方便取样：也称偶遇取样，研究者选择在方便的时间和地点将所 遇到的人作为研究样本的方法。立意取样，也称主观取样、判断取样，研究者根据自己的经验，如对总体构成要素和研究目标的