Printer Friendly

Women and minorities in science and engineering: a life sequence analysis.

A Life Sequence Analysis

At the outset we should emphasize that this article is organized in a somewhat atypical fashion; that is, we have merged the existing literature on this topic with our own original results, rather than having composed separate literature review and findings sections. This unusual organization reflects the explicit purposes of our study: (1) to synthesize what others have learned about this important topic and then (2) to test those themes by analyzing two national data sets. Further, we illustrate rather than report all the data that served as the bases of our empiricism, in keeping with the study emphases: We do not want to overstate the importance of our own results, compared to the much more extensive work of others, in reaching our conclusions. In sum, the conclusions that we report represent an integration of the work of previous researchers and of our own original efforts.

Background

Underutilization of women and minorities in science and engineering is a problem of national priority; social equity and the quality of U.S. labor are involved. Although the numbers for minorities remain discouraging, women have nearly reached or even exceeded parity with men as entrants into numerous scientific fields; nevertheless, their representation in many specialties remains substantially below their distribution in the population (National Research Council, 1991, 1994; Seymour & Hewitt, 1994; Commission on Professionals, 1994). Not only does the lack of parity raise obvious social concerns, the quality of the science and engineering labor force is at issue, for even though the supply of scientists and engineers appears, at least for the present, to be adequate to national needs, larger labor supply pools translate into higher quality scientific workers, thereby increasing marginal labor productivity and, in turn, the productivity of the overall economy.(1) Put another way, all else being equal, we would expect that the more women and minorities enter these fields, which are known to be powerful contributors to national productivity, the larger will be the nation's stock of scientists and engineers, the greater will be the quality of that stock, and the greater will be the productivity of the nations's labor force.

Underrepresentation of women and minorities in science and engineering is not a new topic in science policy research. A recent review of the literature identified no less than 120 empirical and theoretical undertakings related directly to the issue (Leslie & Oaxaca, 1997). During 1995 and 1996 alone, many new articles on the subject were published or presented at national conferences (e.g., Barrett, 1996; Ethington, 1995; Paulson, Hagedorn, Petrides, & Wenzel, 1995; Ross, Volkwein, & Vogt, 1995; Sax, 1996; Yaeger, 1995; Zaruba & Dey, 1996). In addition to the ubiquitous commentary-type articles, this substantial literature ranges from studies that consider specific questions by examining small samples in single institutions to major national investigations of virtually all the conceivable, related issues.

Although our efforts were targeted on the postsecondary education years, we quickly learned that most postadolescent behaviors in regard to science and engineering can be understood clearly only by reference to earlier life experiences, probably most notably to very early socialization and to behaviors that become manifest in the adolescent years. What emerged from the literature synthesis was a robust early-years-to-employment explanation of why women, and to a lesser extent minorities, are underrepresented in particular areas of science, mathematics, and engineering.

This "life-sequence" approach identifies three major concepts that seem to capture most of what distinguishes majority females and minorities from majority males in regard to science and engineering study and employment. After briefly describing our methods, for each of the three concepts we begin our treatment with the early years of life, progress on through the school and college years, and finish long after college matriculation, at a time when some individuals are employed and still others are in graduate school. We conclude the article with some policy implications.

Testing the Three Concepts: Data and Methods

The majority of our effort in this study was directed toward understanding and synthesizing the results of the extensive research completed by others. For this we employed standard integrative review strategies. Once we had discerned the lessons of previous work, we were ready to test those explanations ourselves.

It is important for the reader to understand that the data presentation below is illustrative only. The tables necessary to present all or even most of the data that yield the findings far exceed the space available (see endnote 3). This is important to know, because the reader may otherwise spend considerable time searching futilely for the data that demonstrate the finding being reported.

Our data sets included the 1971 and 1980 Cooperative Institutional Research Program (CIRP) files, which contained substantial information from the precollege years, taken at the time of college entrance and extended through undergraduate and graduate education and employment, and the National Longitudinal Survey of Youth (NLSY), which annually (through 1993) interviewed persons between the ages of 14 and 22 in 1979.(2) The CIRP data, and to a considerable extent the NLSY data, were almost entirely categorical as opposed to continuous. This necessitated an exclusively limited dependent variable approach to the empirical analysis. We estimated binomial logit, multinomial logit, and ordered logit models by maximum likelihood methods.

Because the estimated parameters of these models are not always easily interpreted, we focus our discussion on the estimated marginal effects of the explanatory variables (evaluated at the sample means) on the outcome (education and employment) probabilities. Because the outcome probabilities must sum to one, the marginal effects must sum to zero. Since the explanatory variables are typically dummy (indicator) variables, the notion of a marginal effect requires some elaboration. We calculate the marginal effect of each variable as the incremental effect of a change in the variable from its sample mean, holding constant all other variables at their sample means. The interpretation is that the sample means represent the probabilities of selecting someone with the given characteristics. We then estimate the effects on the outcome probabilities of marginal changes in the probabilities of drawing individuals with the given characteristics.

Although a particular marginal effect may appear to be small, this does not imply that the estimated coefficients on the given variable are statistically insignificant in the underlying logit model; indeed, many are significant. Further, it must be remembered that these are marginal effects; that is, the contributions of other variables are held constant. For example, it will be noted that having a parent who has a science or engineering occupation adds to the likelihood that one will major in science or engineering, quite aside from other aspects of that occupation, such as income level, and quite aside from one's personal traits, such as high-school rank and high-school preparation in science and math.

Where number of cases permit (CIRP), we estimate separate models for white males, white females, black males, black females, Hispanic males, and Hispanic females. Most often our analysis (CIRP) is limited to white males versus white females or males versus females (NLSY). Full information from the CIRP analysis is contained in our report to the National Science Foundation (1996); full information from the NLSY is in McClure, 1997.(3)

Because all of the explanatory variables are dummy (indicator) variables, each set of dummy variables must have a left out reference group in order to avoid a perfect correlation among the explanatory variables. Where only a single dummy variable is included, the left out reference category is obvious. For the high-school rank variable (HSRANK4 and HSRANK3), those in the bottom 50 percentile served as the left out reference group. For the parental income variable, PARINC7 served as the left out reference group and represented parental income in the middle of the income range (where PARINC1 is the lowest income category and PARINC12 is the highest income category).

We utilize the CIRP data for the illustration of our analysis; the NLSY analysis is virtually identical. For purposes of analyzing factors that determine the probabilities of selecting a first choice from among competing college majors in the "freshman" year, we aggregate CIRP choices into five broad categories of college major: physical sciences/engineering, biological sciences, liberal arts, business, and all others. The underlying choice framework is represented by a multinomial logit model. (For analyses of persistence to the degree, graduate study and employment in physical sciences, mathematics and engineering [PSME], ordered logits are used.) The final multinomial model includes seven sets of explanatory variables, each set consisting of one or more dummy variables. These variable sets are high school rank, self-rating of math and natural science preparation for college, the presence of a college prep program in high school, parental occupation in the science and engineering area, parental income, expectations regarding sources of college financing, and marital plans. The sample size is 9,628. We estimate a pooled model and control for gender and ethnicity with dummy variables. We also estimate separate models for each of the six gender/ethnic categories.

First we discuss findings from the pooled samples. The estimated multinomial logit model correctly predicts 46.7% of the first choices of college major. The estimated marginal effects for the pooled sample are reported in Table 1. The magnitude of the effects may be evaluated by recalling that theoretically the marginal effects must lie between -1.0 and +1.0; but also note that the variables may not be statistically significant. As reflected by the negative signs (-), holding constant other factors, white females (WFEMALE) and Hispanic females (HFEMALE) are less likely (than white males) to select science and engineering or business as their first choice major and are more likely to select a major in liberal arts. For example, if the probability that a student is a Hispanic female increases by 10 percentage points at the mean, the probability of selecting a major in physical science or engineering falls by 1.5 percentage points. Black females (BFEMALE) also are less likely (than white males) to select their first choice from physical science/engineering or business; however, black females are more likely to select a major in the biological sciences and in the liberal arts. Black males (BMALE) are less likely (than white males) to select their first choice from science and engineering and are more likely to select business. Finally, Hispanic males (HMALE) are less likely to select biological science or business but are more likely to select physical sciences/engineering.
TABLE 1

Marginal Effects for the First Choice of College Major: Total Sample
(CIRP)

Variable    PhSci/Eng    BioSci   Lib Arts   Business     Other

WFEMALE      -0.1115    -0.0155     0.1628    -0.0552    0.0195
BFEMALE      -0.1144     0.0216     0.1089    -0.0298    0.0137
HFEMALE      -0.1475    -0.0298     0.1686    -0.0032    0.0119
BMALE        -0.0217    -0.0367     0.0130     0.0251    0.0203
HMALE         0.0028    -0.0156     0.0231    -0.0493    0.0390
HSRANK4       0.0526     0.0589     0.0074    -0.0391   -0.0799
HSRANK3      -0.0050     0.0236     0.0059    -0.0024   -0.0221
HSPRIVUS     -0.0222     0.0195     0.0392     0.0059   -0.0424
PREPSE        0.0789     0.0782    -0.1615     0.0138   -0.0094
CPREPSCH     -0.0064     0.0015     0.0693    -0.0433   -0.0210
PAROCSE       0.0506     0.0283    -0.0257    -0.0373   -0.0159
PARINC1      -0.0075    -0.0390    -0.0456     0.0312    0.0610
PARINC2       0.0086    -0.0144    -0.0504     0.0035    0.0526
PARINC3       0.0110    -0.0367    -0.0476     0.0094    0.0638
PARINC4       0.0270    -0.0188    -0.0354    -0.0022    0.0295
PARINC5       0.0079    -0.0088    -0.0226     0.0025    0.0210
PARINC6       0.0101    -0.0216    -0.0072    -0.0025    0.0212
PARINC8      -0.0112     0.0007     0.0472    -0.0277   -0.0090
PARINC9      -0.0102     0.0205     0.0162    -0.0227   -0.0038
PARINC10     -0.0114    -0.0050     0.0623    -0.0048   -0.0411
PARINC11     -0.0701     0.0483     0.0310    -0.0132    0.0041
PARINC12     -0.0493     0.0362     0.1194    -0.0731   -0.0333
WORKSAV      -0.0043     0.0010    -0.0193     0.0091    0.0135
PARENT       -0.0129     0.0015     0.0220    -0.0006   -0.0099
GRANTS        0.0163    -0.0008     0.0544    -0.0255   -0.0445
LOANS        -0.0117     0.0048     0.0415    -0.0218   -0.0127
FUTREMAR      0.0032    -0.0320     0.0072     0.0106    0.0112

Variable Definitions: WFEMALE = white female; BFEMALE = black
female; HFEMALE = Hispanic female; BMALE = black male;
HMALE =Hispanic male; HSRANK4, 3 = 75th and 50th percentiles of
high-school class; PREPSE = high-school preparation in science &
math better than most; CPREPSCH = high-school college prep program;
PAROCSE =parental occupation in science and engineering;
PARINC1-12 = parental income from lowest to highest; WORKSAV = work,
savings, or GI benefits a major source of college financing;
PARENT = parental resources a major source of college financing;
GRANTS = scholarship or grants a major source of college financing;
LOANS = loans a major source of college financing;
FUTREMAR = reasonable chance of marrying in college or a year after
college.


Turning to the separately estimated multinomial logit models, we find the following predictive accuracy rates for first choice of college major: white males - 37.5%, white females - 55.0%, black males - 40.4%, black females - 51.7%, Hispanic males - 51.9% and Hispanic females - 56.7%. We report the separately estimated marginal effects in Tables 2 through 7. Our illustrations of the data are of the major differences, among groups, in the marginal effects of the determinants of first choice of college major as they pertain to selection of majors in science and engineering.

Among the six gender and racial/ethnic groups, being in the top 25% of one's high-school class has the largest positive effects on the selection of science and engineering for white males (Table 2). The effect of better than average preparation in math and natural science is the highest among Hispanic males (Table 6). A 10 percentage point increase in the probability of reporting better than average preparation increase the probability of selecting a major in science and engineering by 2.2 percentage points for Hispanic males and 1.7 percentage points for white males.(4) Among males, having a parent in a science and engineering occupation has the largest effect on the probability of selecting physical science and engineering for Hispanics. For example, increasing the probability of having a parent in a science and engineering occupation by 10 percentage points increases the probability of selecting a major in physical science and engineering by 8.0 percentage points for Hispanic males; next is a 1.1 percentage points effect for white males.

In the case of the biological sciences, blacks exhibit the largest positive effects of having a parent in science and engineering. A 10 percentage point increase in the probability of this variable increases the probability of choosing a major in biological sciences by 1.5 percentage points for black females (Table 5) and 1.4 percentage points for black males (Table 4). For Hispanic males the effect is to lower the probability of selecting biological sciences by 7.7 percentage points; for Hispanic females the effect is a 1.0 percentage point gain (Table 7). Anticipation of reliance upon work, savings, or GI benefits (WORKSAV) has the largest effect on selection of a major in physical science and engineering for Hispanic males. This effect is negative and indicates that a 10 percentage point increase in the probability of relying upon work, savings, or GI benefits reduces the probability of selecting a major in physical science and engineering by 2.0 percentage points.
TABLE 2

Marginal Effects for the First Choice of College Major: White Males
(CIRP)

Variable   PhSci/Eng    BioSci    Lib Arts   Business    Other

HSRANK4      0.0669     0.0972      0.0166    -0.0642   -0.1166
HSRANK4      0.0669     0.0972      0.0166    -0.0642   -0.1166
HSRANK3     -0.0158     0.0741     -0.0169    -0.0014   -0.0401
HSPRIVUS    -0.0452     0.0089      0.0553     0.0414   -0.0603
PREPSE       0.1022     0.0747     -0.1579     0.0102   -0.0291
CPREPSCH    -0.0122     0.0159      0.0621    -0.0360   -0.0299
PAROCSE      0.1073     0.0097     -0.0207    -0.0213   -0.0750
PARINC1     -0.0197    -0.0283     -0.0068    -0.0141    0.0689
PARINC2      0.0385    -0.0084     -0.0643    -0.0322    0.0663
PARINC3      0.0195    -0.0816      0.0231    -0.0219    0.0608
PARINC4      0.0484    -0.0364     -0.0588    -0.0140    0.0606
PARINC5      0.0037    -0.0127     -0.0106    -0.0087    0.0283
PARINC6      0.0399    -0.0289     -0.0447     0.0070    0.0266
PARINC8     -0.0168     0.0114      0.0375    -0.0503    0.0182
PARINC9     -0.0531     0.0081      0.0333    -0.0311    0.0428
PARINC10    -0.0533     0.0157      0.0586    -0.0003   -0.0206
PARINC11    -0.2316     0.0845      0.0773     0.0153    0.0546
PARINC12    -0.0923     0.0800      0.0970    -0.0942    0.0095
WORKSAV      0.0067    -0.0215     -0.0180     0.0057    0.0272
PARENT      -0.0372     0.0191      0.0341    -0.0164    0.0004
GRANTS       0.0190     0.0188      0.0426    -0.0571   -0.0234
LOANS       -0.0357    -0.0104      0.0664    -0.0191   -0.0012
FUTREMAR    -0.0105    -0.0063     -0.0173     0.0396   -0.0055

NOTE: See Table 1 for variable definitions.
TABLE 3

Marginal Effects for the First Choice of College Major: White
Females (CIRP)

Variable   PhSci/Eng    BioSci   Lib Arts   Business     Other

HSRANK4      0.0515     0.0211     0.0069    -0.0262   -0.0533
HSRANK3      0.0110    -0.0064     0.0290    -0.0011   -0.0326
HSPRIVUS     0.0020     0.0332     0.0096    -0.0149   -0.0299
PREPSE       0.0620     0.0856    -0.1724     0.0153    0.0094
CPREPSCH    -0.0041    -0.0219     0.0557    -0.0372    0.0074
PAROCSE      0.0235     0.0351    -0.0548    -0.0268    0.0231
PARINC1     -0.0016     0.0053    -0.0968     0.0371    0.0560
PARINC2      0.0140     0.0185    -0.0855     0.0205    0.0324
PARINC3      0.0170    -0.0273    -0.0799     0.0288    0.0615
PARINC4      0.0044     0.0018    -0.0192     0.0060    0.0070
PARINC5     -0.0013     0.0047    -0.0249     0.0118    0.0097
PARINC6     -0.0128    -0.0115     0.0318    -0.0178    0.0103
PARINC8     -0.0011    -0.0084     0.0530    -0.0183   -0.0253
PARINC9      0.0158     0.0330     0.0189    -0.0169   -0.0508
PARINC10    -0.0014    -0.0297     0.0919     0.0006   -0.0615
PARINC11    -0.0000     0.0470     0.0134    -0.0422   -0.0182
PARINC12    -0.0145    -0.0130     0.1343    -0.0384   -0.0684
WORKSAV     -0.0077     0.0179    -0.0222     0.0095    0.0025
PARENT       0.0018    -0.0109     0.0223    -0.0027   -0.0105
GRANTS       0.0198    -0.0096     0.0727    -0.0246   -0.0583
LOANS       -0.0044     0.0089     0.0159    -0.0222    0.0018
FUTREMAR     0.0138    -0.0508     0.0364    -0.0117    0.0124

NOTE: See Table 1 for variable definitions.
TABLE 4

Marginal Effects for the First Choice of College Major: Black Males
(CIRP)

Variable   PhSci/Eng    BioSci   Lib Arts   Business     Other

HSRANK4      0.0779     0.0587    -0.0606     0.0003   -0.0764
HSRANK3      0.0205    -0.0321    -0.0642     0.0128    0.0630
HSPRIVUS    -0.1006     0.0061     0.1537    -0.0267   -0.0325
PREPSE       0.0714     0.0550    -0.1167     0.0119   -0.0217
CPREPSCH    -0.0198     0.0324     0.0414    -0.0399   -0.0142
PAROCSE      0.1559     0.1367     0.7729    -1.3174    0.2518
PARINC1     -0.0207    -0.0831     0.0182     0.0346    0.0510
PARINC2     -0.0163    -0.0720    -0.0423     0.0686    0.0620
PARINC3      0.0267    -0.0156    -0.1644     0.0895    0.0637
PARINC4      0.1110    -0.0239    -0.1064     0.0604   -0.0411
PARINC5      0.1506    -0.1468    -0.1347     0.0787    0.0522
PARINC6      0.0699    -0.0278    -0.0871     0.0832   -0.0382
PARINC8     -0.0822    -0.0017     0.0732     0.0452   -0.0345
PARINC9     -1.1888     0.0786     0.5727     0.1570    0.3805
PARINC10     0.6319     0.4789     1.4123    -0.9142   -1.6089
PARINC11     0.0454    -0.1212    -1.2662     1.7006   -0.3586
PARINC12    -1.0878     0.2810     1.3762    -1.1246    0.5552
WORKSAV     -0.0298     0.0005     0.0028    -0.0081    0.0346
PARENT      -0.0141     0.0406    -0.0125     0.0099   -0.0240
GRANTS       0.0109    -0.0239     0.0053    -0.0020    0.0098
LOANS       -0.0136     0.0058     0.0821     0.0079   -0.0822
FUTREMAR     0.0093    -0.0219    -0.0075     0.0566   -0.0365

NOTE: See Table 1 for variable definitions.
TABLE 5

Marginal Effects for the First Choice of College Major: Black
Females (CIRP)

Variable   PhSci/Eng    BioSci   Lib Arts   Business     Other

HSRANK4      0.0340     0.0141    -0.0010    -0.0199   -0.0273
HSRANK3      0.0059    -0.0478     0.0188    -0.0016    0.0247
HSPRIVUS    -0.0181    -0.0113     0.0840    -0.0463   -0.0083
PREPSE       0.0384     0.0624    -0.0963    -0.0212    0.0167
CPREPSCH     0.0034     0.0146     0.1137    -0.0499   -0.0818
PAROCSE     -0.0008     0.1542     0.3809    -0.5504    0.0161
PARINC1      0.0117    -0.0850    -0.0186     0.0240    0.0679
PARINC2      0.0016    -0.0514    -0.0128     0.0015    0.0610
PARINC3      0.0021    -0.0249     0.0131    -0.0069    0.0166
PARINC4     -0.0062    -0.0165    -0.0380     0.0010    0.0597
PARINC5      0.0081    -0.0313     0.0448    -0.0155   -0.0061
PARINC6     -0.0023    -0.0290    -0.0469    -0.0165   -0.0947
PARINC8     -0.0169     0.0319     0.0121     0.0332   -0.0602
PARINC9     -0.2265     0.1826    -0.0736     0.0163    0.1013
PARINC10    -0.2307    -0.0796     0.2068     0.0109    0.0926
PARINC11    -0.1007    -1.2704     3.4304    -0.3371   -1.7222
PARINC12    -0.1771     0.6029     2.0266    -0.4184   -2.0340
WORKSAV     -0.0013     0.0333    -0.0153     0.0071   -0.0238
PARENT       0.0157    -0.0348    -0.0023     0.0149    0.0065
GRANTS       0.0102    -0.0196     0.0523     0.0023   -0.0452
LOANS        0.0024     0.0210     0.0450    -0.0152   -0.0532
FUTREMAR     0.0005    -0.0705     0.0341    -0.0072    0.0431

NOTE: See Table 1 for variable definitions.
TABLE 6
Marginal Effects for the First Choice of College Major: Hispanic
Males (CIRP)

Variable   PhSci/Eng    BioSci   Lib Arts   Business     Other

HSRANK4      0.0298     0.0799     0.1363    -0.0278   -0.2182
HSRANK3     -0.0827     0.0518     0.0716    -0.0437    0.0029
HSPRIVUS    -0.0920     0.0057     0.0808     0.0154   -0.0099
PREPSE       0.1692     0.0494    -0.2723     0.0615   -0.0078
CPREPSCH    -0.0240    -0.0049     0.1212    -0.0091   -0.0832
PAROCSE      0.7993    -0.7711     2.2908    -0.4706   -1.8484
PARINC1     -0.0373     0.1529    -0.4767     0.1257    0.2355
PARINC2      0.0887     0.0878    -0.4160     0.0851    0.1543
PARINC3      0.0678     0.0813    -0.4833     0.0681    0.2662
PARINC4      0.1612     0.0486    -0.2587     0.0267    0.0222
PARINC5      0.1286     0.0862    -0.3481    -0.0059    0.1392
PARINC6      0.0898     0.1422    -0.1781     0.0488   -0.1027
PARINC8     -1.1509     0.2133     0.3853     0.1286    0.4237
PARINC9     -0.2571    -0.2186    -2.3000    -0.2000    2.9757
PARINC10     1.2539    -0.4247    -2.9292    -0.2203    2.3202
PARINC11          -          -          -          -         -
PARINC12    -0.2580    -0.2389    -2.4046    -0.1068    3.0083
WORKSAV     -0.1954     0.0344     0.0574     0.0440    0.0596
PARENT      -0.0906     0.0890    -0.0204     0.1005   -0.0786
GRANTS       0.0321     0.0138     0.1075    -0.0304   -0.1230
LOANS        0.0729    -0.0066    -0.0134    -0.0376   -0.0154
FUTREMAR     0.0324     0.0153    -0.1913    -0.0139    0.1575

NOTE: See Table 1 for variable definitions.
TABLE 7
Marginal Effects for the First Choice of College Major: Hispanic
Females (CIRP)

Variable   PhSci/Eng    BioSci   Lib Arts   Business     Other

HSRANK4      0.0008     0.0678    -0.0579     0.0534   -0.0641
HSRANK3     -0.0064     0.0234     0.0551    -0.0338   -0.0383
HSPRIVUS     0.0002     0.0190    -0.0250     0.0026    0.0032
PREPSE       0.0004     0.0486    -0.0078     0.0201   -0.0614
CPREPSCH     0.0006     0.0870     0.0843    -0.0693   -0.1026
PAROCSE     -0.0066     0.0952     0.4827    -1.1502    0.5789
PARINC1     -0.0013    -0.1383     0.3973    -0.1575   -0.1002
PARINC2     -0.0017    -0.1031     0.3803    -0.2359   -0.0396
PARINC3     -0.0015    -0.1576     0.2834    -0.2449    0.1206
PARINC4     -0.0010    -0.1288     0.5309    -0.2239   -0.1772
PARINC5     -0.0009    -0.1252     0.3477    -0.1976   -0.0240
PARINC6     -0.0071    -0.1439     0.2595    -0.1270    0.0184
PARINC8      0.0007    -0.8464     3.4810    -0.8563   -1.7791
PARINC9           -          -          -          -         -
PARINC10          -          -          -          -         -
PARINC11          -          -          -          -         -
PARINC12    -0.0055    -0.7933     3.8022    -1.0998   -1.9036
WORKSAV     -0.0001     0.0358    -0.0224     0.0168   -0.0302
PARENT      -0.0004     0.0121     0.1459    -0.0678   -0.0898
GRANTS      -0.0009    -0.0393     0.1436    -0.0089   -0.0946
LOANS        0.0000     0.0090     0.0853    -0.0961    0.0018
FUTREMAR    -0.0007    -0.0204    -0.0134    -0.0118    0.0463

NOTE: See Table 1 for variable definitions.


Group differences in estimated marginal effects result from differences in parameter estimates as well as differences in the mean values of the explanatory variables. One way to separate out these effects is to construct Duncan dissimilarity indexes. These indexes measure the distance between any two distributions across mutually exclusive categories. The index is bounded between 0 and 1. A value of zero indicates that the two groups are identically distributed across the categories, whereas a value of 1 indicates no overlap in the distribution, i.e. perfect segregation. The index can be interpreted as an estimate of what proportion of one group or the other would have to shift categories in order to render the two distributions identical. If we predict a distribution for a given group based on the estimated model for white males and then calculate the dissimilarity between white males and the predicted distribution for the given group, the value of the index is a measure of how much dissimilarity remains after controlling for different parameter estimates between white males and the given group. In other words, the index measures the contribution of group differences in the values of the explanatory variables to overall dissimilarity. Similarly, we can calculate the dissimilarity between a group's actual distribution and its distribution as predicted from the white male model. In this case the value of the index is a measure of the dissimilarity arising from group differences in the models generating the outcomes, i.e., dissimilarity stemming from group differences in the parameter values or structure.

The Duncan Dissimilarity Index (DDI) analysis (Table 8) indicates that the disparity between white males and others almost always has more to do with differences in the effects of explanatory variables (i.e., their coefficients) than with differences in the values of the variables themselves. (For example, it is not so much differences in the number or set of math courses taken, that explain differences between male and female persistence, as it is differences in the utility gained from the particular courses that are taken.) This is true for all females in the first choice of college major. For example the DDI between white males and white females is 20.4, which means that 20.4% of the white female freshmen would have to change their first choice of college major in order to have the same college major distribution as white males. If the effects of the explanatory variables on choice of college major for white females were the same as those of white males, only 3.4% of white females would have had to change majors in order to produce identical distributions. By comparison the DDI between the actual college major distribution of white females and their distribution predicted on the basis of the estimated [TABULAR DATA FOR TABLE 8 OMITTED] model for white males is 19.0. In other words the gender differences in the effects of the explanatory variables require that 19.0% of white females change their first choice of college major in order to produce identical distributions for college major. This is much larger than the change required solely on the basis of gender differences in the values of the explanatory variables.

This illustrates the nature of the analysis, the form of the findings, and the bases for the results reported below. The following section integrates the extant literature with our own findings; the specific data reflecting our own finding are not presented, as is consistent with the purpose and emphases of our study, not to mention space limitations.

Results: The Integration of Previous Evidence and Our Own Findings

In preface we should note that centering our discussion on the most powerful concepts explaining the underrepresentation of women and minorities in science and engineering (a conclusion based upon a synthesis of the literature and our findings) results in a somewhat artificial grouping of particular variables and in omission (in the discussion) of others, which often are noteworthy by themselves. In a few cases the variables omitted may be of greater explanatory power than included variables, which were selected for discussion only if they contributed to an understanding of the most powerful concepts. (Again, we commend to the attention of the reader the full NSF Report and to McClure, 1997.)

The three concepts, self-concept/self-efficacy, peer influence, and goal commitment, are all closely connected. We begin with self-concept and self-efficacy, two closely related abstractions that constitute an important body of research in psychology and social psychology. An individual's development of the first of these, self-concept, is fundamental to development of the latter, self-efficacy. The early development of self-concept bears importantly on the achievement of science outcomes for women and minorities later in life, that is, through science and engineering self-efficacy. In turn one's self-concept/self-efficacy is impacted importantly by peers, all of which affect one's relative commitment to science and engineering. In each of the three sections below, we begin with definitions and a discussion of the relevant literature and then progress to a discussion of our own results.

Self-Concept and Self-Efficacy

Self-concept is perception of self; self-efficacy is belief in one's ability to perform a given behavior (Lent, Lopez, & Bieschke, 1991). The evidentiary literature is quite compelling that self-concept and self-efficacy strongly affect the achievement of science and engineering outcomes for women and probably for minorities, too, although the evidence regarding minorities is much more limited.

Self-concept is both hierarchical and temporal in nature. General self-concept is at the top of the self-concept hierarchy; positive, subordinate self-concepts generally contribute to positive general self-concepts, although the former may be complementary rather than complimentary. General or overall self-concept is observable in very young children and is quite stable, but with age becomes situation-specific as the hierarchy is descended. (Subordinate) verbal and Mathematical self-concepts and self-concept hierarchies begin to develop by about ages 5-8 (Marsh & Shavelson, 1985; Marion & Coladarci, 1993); they are distinct and separate by late adolescence. Self-concept has been linked empirically to achievement (e.g., Shavelson, Hubner, & Stanton, 1976). These statements pertain to the overall population. What of differences for identifiable student groups?

Generally, in elementary school boys and girls do not vary significantly in math/science ability, confidence, or interest; however, many math/science gender differences are evident by the end of high school, with the junior high-school years probably being transition years for most youth, but particularly for girls (AAUW, 1991; Hyde, Fennema, & Lamon, 1990; Linn & Hyde, 1989; Meece, Parsons, Kaczala, Goff, & Futterman, 1982). It is instructive to note that these patterns are consistent with self-concept theory and with empirical evidence: the absence of self-concept differentiation in the early years and the distinctiveness and separateness of math and verbal self-concept in later years.

Generally, self-efficacy, rather than self-concept, is employed by social scientists to explain behavior related to science and engineering outcomes. This choice is supported by the definitional differences. Self-efficacy has a more specific contextual meaning than self-concept. As typically used, self-concept might be seen as the general conception of self one brings to personal decision making. It is the basis for forming one's sense of self-efficacy, which relates to specific personal decisions, such as career choices in science and engineering.

Self-efficacy theory, which usually is credited to Bandura (1977), begins with the notion that human behavior is devised through cognitive processes, with performance-based experiences being most important. Learning about ourselves involves evaluating "differential consequences"; it involves acquiring and evaluating personal information. The elements of the learning process are four in number: We learn through "performance accomplishment," which is very important and is based upon mastery; "vicarious experience," which involves making social comparisons and is less powerful; "verbal persuasion," which is widely utilized but is of limited utility; and "emotional arousal."

An appreciation of these four factors is important to understanding the issues addressed herein: As specific obstacles arise, one's self-efficacy will be instrumental in determining whether one decides to cope with adversity, to what degree, and how persistently. Bandura observes significantly that the "strength of conviction" in one's self-efficacy will affect one's willingness to see a task or goal to completion and even whether one will make an effort: "Weak expectations are easily extinguishable by disconfirming experiences, whereas individuals who possess strong expectations of mastery will persevere in their coping efforts despite disconfirming experiences" (Bandura, 1977). Strong self-efficacy can turn threatening obstacles into events perceived as safe. Self-efficacy has been shown to predict performance accurately in 85% of tasks confronted (Bandura, 1977).

With these ideas in mind, let us return to gender differences as they may pertain to science and engineering. On essentially all characteristics hypothesized to affect science and engineering-related educational and career choices, girls do not differ from boys in the early years of life, years before the development of the distinctively different subordinate self-concepts; however, numerous, important self-concept/self-efficacy differences related to science and math are evident thereafter.

As a point of departure in examining gender differences related to science and engineering, ability differences between boys and girls in math and science are almost universally agreed to be minimal over all age ranges (e.g., Friedman, 1989; Orenstein, 1994). Although some differences usually are noted in empirical work, analytical models almost uniformly find no significance for ability measures when other factors, such as number of science and mathematics courses taken, are controlled. Even so, certain exceptions are instructive. Mathematical abilities do decline earlier and more steeply for girls, and at the high end of the ability continuum, differences clearly are noted; for example, males outnumber females 4:1 in the number of individuals earning more than 600 on the SAT-Math test (Benbow & Stanley, 1983). Nevertheless, girls tend to earn somewhat higher grades in science and mathematics. Overall, the conclusion is that ability differences fail to account for important gender disparities in science and engineering and that we must look to other explanations.

Such differences are noted on a number of other measures. The number of mathematics courses taken almost invariably is a primary predictor of majoring and persistence in science and engineering (e.g., Astin & Astin, 1992). By the end of high school, boys substantially exceed girls by this measure. One's expectations of the probability of success in a given endeavor is known to be an important factor in formulating behavioral choices. Levels of expectations in science and math are lower for females than for males (e.g., Lent, Lopez, & Bieschke, 1991). There are several reasons for this.

Males rate the usefulness of science and math higher than do females, who even rate science and math as more useful for males than for themselves. Female ratings of science and mathematics usefulness decline for girls as they mature, a phenomenon that does not characterize the maturation of boys (Brush, 1980; Parsons & Adler, 1983; Sherman, 1980). No doubt female "other directedness" plays some part in this, that is the desire to please others. Linn and Hyde (1989) conclude that of all variables examined, only usefulness of math and science appears to affect persistence, and that this difference by gender has existed for many years.

Usefulness of science and math has been connected empirically to valuing science and math, males valuing these subjects more than females (Betz & Hackett, 1983). The importance of "usefulness" to science-and-mathematics-linked academic and career choices is illustrated by analogy to the gender differences in the valuing of athletics: Because traditionally girls have valued athletics less than boys, it is hardly surprising that historically more boys than girls have engaged in athletics, although, of course, this is changing.

It is similarly no surprise that boys voice a greater interest in and more positive attitudes toward science and mathematics than do girls, although, again, the differences emerge only after the elementary school years (Betz & Hackett, 1983). Before adolescence, interest levels are essentially the same. In one study interest even surpassed ability in predicting self-efficacy in engineering, and self-efficacy was found to intervene to mediate effects of such factors as stress, coping, and gender and ethnicity on achievement (Hackett, Betz, Casas, & Rocha-Singh, 1992).

The same is true for confidence in science and math: Relatively and on average, women are more likely to lack such confidence (Manis, Thomas, Sloat, & Davis, 1989). The disparity in favor of boys emerges by high school, and this relative confidence may not be supported by ability differences (Betz & Hackett, 1983). In turn, confidence in science and mathematics has been connected to test-taking skills in math/science; for example, boys are more willing or able to attempt time-saving shortcuts in solving problems.

A caveat should be inserted here. Gender-linked variables do not necessarily translate into gender-linked effects on the critical outcomes. For example, the only causal relationship Sherman and Fennema (1977) could find among math usefulness, performance, and course plans was a moderate one between math being perceived as a male domain and math achievement, and the relationship was only for girls in the high school-years.

This caveat notwithstanding, by the end of high school, these perceptual differences are associated with gender disparities in science and engineering-related decisions. Perceptions of math usefulness generally predict intentions to take math courses and math achievement, although the greater course-taking propensity of males is not clearly detectable until grade 12 and is not strong in predicting enrollment in mathematics courses even at the collegiate undergraduate level (Sherman & Fennema, 1977; Meece, et al., 1982).

That female self-efficacy is involved in these disparate changes, is of little doubt. Girls' self-esteem drops precipitously after adolescence, with the drop for Latinas being the greatest, whereas African-American girls do not decline in self-esteem (they do dislike schools and teachers) (AAUW, 1991; Sadker & Sadker, 1994). (There may be an important clue here to understanding gender differences if it can be shown that the socialization of African-American girls differs from that of other girls and that African-American female socialization is similar to that of majority males.)

The manifestations of gender difference take several forms. Older boys rate their ability as the "secret"' of their success in science and engineering, whereas similarly situated girls cite their personal diligence, skill, and effort (e.g., Jones & Wheatley, 1990; Manis et al., 1989). If girls experience problems in math, they perceive this as personal failure (AAUW, 1991, p.13). Especially boys, but girls as well, believe that male understanding of science and math is superior, even when it is not (Linn & Hyde, 1989; Orenstein, 1994). Boys overestimate their abilities; girls underestimate theirs.

The connection between self-concept or self-esteem and ability in science and mathematics is important. Students are more likely to enroll in optional math courses when they perceive themselves to possess high math ability or feel confident in math (Meece, et al., 1982; Sherman, 1980), although relatively few studies attempt to determine whether the links are causal, and when they do, the results are far from clear (Meece, et al., 1982).

Math comes to be viewed as a male achievement domain (Meece, et al., 1982; Ernest, 1976); the same is true for science (Mason, Kahle, & Gardner, 1991).(5) Boys consider math to be a masculine subject (Brush, 1980); however, girls are less likely to gender-type math than are boys, and although girls may perceive math-related careers as masculine, they do not necessarily view math as inappropriate for them (Meece, et al., 1982). They do believe that studying science is more important for boys and that it will be of greater utility for boys (Linn & Hyde, 1989). But again, these differences rarely are detected until the self-concept differentiation years, and they are not clearly evident until late adolescence. For example, gender-linked mathematics enrollment patterns do not materialize until grade 12 and are still weak among college undergraduates (Meece et al., 1982).

Clearly, self-efficacy in science and math, on average, is less evident for females than for males. The explanations would appear to involve important gender-linked perceptual differences that take many forms but that usually involve perceived male-female role disparities. The important policy question is, what changes between preadolescence and adolescence? Of course, differences in socialization experience are thought to explain many if not most of the gender differences, but what are the specific mechanisms leading to gender differentiation in science and engineering?

Identification of specific causal agents is not far advanced. Attitudes of parents and other family members, teachers, counselors, and peers are variously known or believed to play some part but most evidence is more impressionistic than hard (Haven, 1971; Jacklin, 1979; Nash, 1979). In regard to race, it appears that self-efficacy and self-confidence are primary factors in African-American male consideration of math/science careers whereas for females personal interests are paramount; nevertheless, African-American male/female interest levels are approximately equal (Post, Stewart, & Smith, 1991). African-American males give broad consideration to career possibilities, relative to females. Disturbingly, African Americans do not appear to reflect on their personal abilities in these considerations.

Self-Concept and Self-Efficacy: Our Testing

What do our results say about these issues? First, our conclusion from the literature that self-concept/self-efficacy is important to understanding underrepresentation of women and minorities is supported strongly. The associated variables in our data sets are second only to measures of commitment (which we connect to self-concept/self-efficacy below) as predictors of science and engineering study and employment.

The best measure of science, math, and engineering self-efficacy available in the CIRP files is at the point of college matriculation, when, as our findings show, the perception that one's preparation in math and natural science is "better than most"' is associated about equally with the greater likelihood of selecting physical science/engineering and biological sciences, rather than "other" fields, as the first choice of college major. The marginal effect, that is, the effect holding all other variables in the analysis constant, is a substantial 1.6 percentage point increase in the probability of choosing science/engineering majors for each 10 percentage point increase in the probability of perceiving oneself to have a relatively good science/math background.

The most closely related outcome measure available in the NLSY is "major field of study in last college attended." Here, there are four presumably good measures of self-concept (as opposed to self-efficacy), and all are related to majoring in the physical sciences, mathematics, and engineering; for example, a personal sense of pride is strongly associated with majoring in these fields and is strongly and negatively associated with majoring in nonscience fields. The reverse is true in regard to the relative feeling of personal failure. These findings are for all matriculants.

What about selected subgroups? First, the fact that white men are most likely to perceive their math and natural science preparation to be better than most is noteworthy. The comparative mean values (from CIRP) are white men, 43%; white women, 31%; black men, 26%; black women, 16%; Hispanic men 23%, Hispanic women, 17%. These are substantial differences. It is little wonder that the numbers of women and minorities entering science and engineering are relatively small.

How do the effects of self perception vary by race and gender? Results from CIRP, such as those reported in the tables, yield answers to this question. The positive marginal effect of the science and math preparation variable on the probability of choosing a major in physical sciences/engineering is larger for white males than for white females, African Americans, and Hispanic females. Only Hispanic males show a greater positive effect of relatively strong math and science preparation on the probability of selecting physical sciences or engineering as the first choice major. In short, one of the important reasons white men are more likely than women and minorities (save Hispanic men) to select math/science majors is that they perceive themselves as being relatively well prepared in math/science; i.e., their self-efficacy in science and math is stronger. In the NLSY files, mathematics knowledge scores are available, whereas an independent measure of science and math self-efficacy is not; therefore, one cannot know to what extent math knowledge, which is a very powerful predictor of majoring in PSME for males but is insignificant for women, acts through self-efficacy, independently, or in tandem. The other pertinent findings from NLSY are more extensive and are somewhat different, in that they largely reflect general feelings of self-concept rather than specific science/math self-efficacy. (Of course, unless both science and math self-efficacy and knowledge are contained in the data set, it may not be possible to separate the effects of the two.) Two NLSY measures of self-perception are strong predictors of majoring in PSME for both men and women, as are the feeling that one is in control of one's direction in life and that "luck" is not an important factor in achieving one's ends; however, the predictive power of each of these variables is modestly stronger for women majoring in PSME than for men: generally, positive self-perceptions and a feeling that one controls one's own destiny has a greater utility for women in PSME than for men. Put another way, the women whose self-concepts and sense of control are less strong are more likely to major in non-PSME fields. Yet, as the CIRP results show, self-efficacy in science and math is most clearly demonstrated by white men in science and engineering.

Turning to the biological sciences, first our stated expectations in regard to math and science preparation are met: For the CIRP sample, self-perception of above average math and science preparation has a larger impact on the probability of selecting a major in PSME than in selecting a major in biological sciences. Further, we can see that a reported "stronger than most" math and science preparation has a larger positive effect on the probability of selecting biological sciences for white women than for white men. The same is true for black females versus black males. Hispanic males and Hispanic females are about equal in this regard. For the NLSY sample, too, the self-concept and sense of control variables are far less important for the biological sciences than is the case for PSME, both for men and women, although generally they are a bit more important for women in the biological sciences than for men. Math knowledge is essentially inconsequential for the biological sciences outcomes. The gender and (for the most part) racial/ethnic differences that exist in these regards are almost completely limited to PSME.

Consistently, for the self-efficacy/self-concept variables, easily the largest differences between groups are found between Hispanic males and females. These results are instructive, because gender roles tend to be most sharply distinguished among Hispanics; that is, traits held to be more masculine tend to be quite clearly associated with male behavior, and those held to be more feminine tend to be quite clearly associated with female behavior.(6) In other words, if in fact socialization experiences do vary by gender and racial/ethnic group, if in fact these differences do manifest themselves in self-concept/self-efficacy variations by gender and racial/ethnic group, and if in fact those differences do lead to female and minority underrepresentation in science and engineering, then we would expect those differences to be most clearly shown among Hispanics, especially Hispanic males, whose roles are most closely linked to being family providers. This is precisely what occurs. Not only is this math and science perception variable most strongly associated with physical sciences and engineering matriculation for Hispanic males, the absence of such perception is most strongly associated with their matriculation in the liberal arts.

There is one other variable in the CIRP data set that should impact directly on self-concept, if not self-efficacy. That variable is one's grades earned in college, and the outcome measure is degree attainment. We would expect that earning good college grades would be more important for white women than for white men, because self-concept/self-efficacy is supposed to be more critical for the former group than the latter in the pursuit of science and engineering majors. Indeed, that is what the results show. For the physical sciences and engineering, the effect of having earned undergraduate grades of B or higher has almost twice the effect on achieving an undergraduate degree and almost a 50% greater effect on earning a masters degree for white women compared with white men. The patterns are of the same general order in the biological sciences; however, here, the effects of higher grades are more beneficial for white males in earning doctoral and professional degrees.

Parental influences. One other set of variables in our data sets would appear to be important to the formation of science and math self-efficacy and, in turn, to science and engineering outcomes. Those variables relate to how parental occupations and educational backgrounds impact whether individuals come realistically to "see themselves" as scientists or engineers.

Others have concluded that parental backgrounds are germane to science and engineering achievement of their offspring, although most of this research is of a relational rather than causal nature. Jackson, Gardner, and Sullivan (1993) report that women who enter male-dominated fields such as science and engineering tend to come from intact families, have mothers who work, and have parents who are well educated and consider success to be important. Worthley (1992), too, reports that science persistence is associated with having highly educated parents. The Jackson et al. work finds that women who become engineers are likely to have fathers who are engineers and that they are likely to marry engineers. The AAUW study (1991) reports that families (and schools) are the greatest effectors of science and engineering aspirations, and Astin (1993), from more recent CIRP data, finds a positive relationship between majoring in engineering and having a father who is an engineer. The engineering students in Astin's sample are clear that their parents want them to become engineers. Our view is that parental backgrounds work their effects on their children primarily through demonstrating the feasibility of science and engineering careers, that becoming a scientist or an engineer is a reasonable expectation, in a word, that it is efficacious.

We, too, find that having a parent engaged in physical science and engineering occupations increases the probability of majoring in the physical sciences and engineering. The NLSY files show a greater effect for males than females. The differences among the six CIRP groups in these regards are particularly instructive, being uniformly greater for males than females and by far the greatest for Hispanic males (followed by African-American males). The connection to self-efficacy, and in turn to science and engineering career commitment, is suggested by our earlier discussion regarding Hispanic culture. We hypothesize that for minority men, especially Hispanics, having a parent who is employed in physical sciences or engineering is very instrumental in creating a perception that such an occupation is a realistic goal, and that this perception reinforces science self-efficacy and supports the student becoming committed to that goal, i.e., to science and engineering study. Our reasoning is that minority males encounter many inhibitors to majoring in science and engineering and that parental role models serve as powerful counterforces. The reason that the marginal effects are less for white men is probably because they are less likely than minority men to encounter obstacles to majoring in science and engineering; therefore, parental role models are less vital. The quite different effects for women, minority and majority, may well reflect their more "eclectic" decision structure; that is, in deciding whether to major in science or engineering, women may be less prone to the influences of parental occupations because of the attractiveness of family-oriented goals. For the biological sciences the patterns are far more mixed for women, although the magnitude of the effects often are strong.

The CIRP data reveal that the marginal effects of fathers' occupations in science and engineering on the science and engineering outcomes for their children are even larger later, in shaping employment patterns, than they are in selection of first majors (a pattern not found in the NLSY data). Clearly, individuals working in science and engineering occupations in 1980 were substantially more likely to have had fathers whose occupations were in science and engineering than in some other field (CIRP data). Further, the effects for the entire sample of white men and white women were almost totally a function of the effects on the males: increasing by 10 percentage points the probability of having such fathers increased the probability of white males being employed in science and engineering occupations by almost 5 percentage points, but for white females the effects were essentially zero. In a pattern that became clear when we examined all the "family influence" variables, white men appeared more responsive to such variables than did white women. Again, we hypothesize that female goal conflict is a major cause for the smaller effects for women.

An important exception to the pattern for family influence variables was the effect of mother's education level upon the physical science/engineering higher degree attainment of their daughters (CIRP). Mothers' college degrees affected female degree attainment positively and male attainment negatively. Because white women, on average, have weaker math/science self-efficacies than men, it may be that maternal role models are more essential for women in science and engineering than for men. Further, for those in the physical sciences or engineering, the effects of parental degrees persist at least through the master's degree; in fact, they grow modestly stronger for both men and women, before declining at the doctoral and professional level. For the biological sciences, the patterns for women are very weak, providing still more evidence that women (as well as men) who pursue degrees and work in the biological sciences are substantially different from those in the physical sciences and engineering.

Peers

Interacting with physiological change, peer influence, we believe, is the "missing link" that explains much of the great transformation of girls, during and after adolescence, from being full and equal participants in science and math to much smaller representation. We largely credit Holland and Eisenhart's, Educated in Romance (1990), for this insight. Holland and Eisenhart explain how female self-concept, self-efficacy, classroom experiences, and external goal orientation come together to deny women their representative places in the science and engineering professions: It is the influence of peers that is critical to how these forces play out. Although their focus is on the college years, they observe that the origins of the phenomena are in the onset of adolescence, which marks not only a clear demarcation between boys and girls in regard to science- and math-related behaviors, it signals the beginning of distinctly different paths of gender-linked general personality development.

From its nationwide survey, the American Association of University Women (AAUW, 1991) concludes that adolescence is a critical time for female self-identity development. It is a time of both dramatic biological and psychological changes; it is a "critical moment" in the development of one's lifetime choices and decisions (AAUW, p. 2). The AAUW survey shows that prior to adolescence, girls are "confident, assertive and feel authoritative about themselves" (p. 4). In subsequent years their self-esteem declines dramatically. Notably, adolescent girls rank "being popular" as the most important personal concern, whereas boys list competence and independence. Girls report not being happy "with the way I am" (p. 5), growing more timid and tentative, and becoming more conflicted. Their physical appearance is of great importance to them (AAUW, 1991).

The similarity of these observations about the female self to female self-efficacy changes in science and mathematics is striking. Lack of self-confidence, low self-esteem, timidity and tentativeness, are all descriptions that fit many postadolescent girls in regard to science and mathematics, as is the perception on the part of girls that boys are more able in math and science than they are.

Logically, the dramatic decline in female self-esteem that occurs during adolescence (AAUW, 1991) would be expected to cause girls to become more other- rather than self-directed. Their primary concern with popularity suggests strongly that girls are likely to be particularly mindful of what their peers think of them: how they behave, their values and orientations, what they like and do not like. This other-directedness might be expected to result in openness to the suggestions of parents and teachers, provided that those suggestions do not run counter to the values of the peer culture, in which case the outcomes might be far less certain. Nevertheless, generally, one might assume that this other-directedness influences girls to develop superior academic habits, for example to complete homework assignments, study dutifully, and generally be more diligent in academic matters than are boys. Clearly, the AAUW data suggest that girls may lack the self-confidence to set their goals independently, that they may pursue life goals reflective not of what they want, but of what others want them to do. Again, the importance of commitment to goals accomplishments is clearly noted.

If adolescent girls are so susceptible to the influences of others, especially peers, what is it that others cause girls to do? Teachers, and to some degree parents, of course, want girls to be good students, often even to study science and math. Female peers, on the other hand, according to the AAUW findings, primarily value popularity. Because this and related values emerge dramatically during adolescence, popularity almost certainly is in part a reflection of personal appearance and by implication, involves appeal or attractiveness to boys (Smith, 1992).

Holland and Eisenhart (1990) demonstrate that females gain much of their self-esteem through their relationships with males. The effects are circular: Validation of girl's and later women's behaviors by males leads to greater acceptance by peers, all tending to reinforce female perceptions that their sense of self-worth is connected importantly to the perceptions of males. This means conforming to what both female and male peers conceive of as appropriate female behaviors and values. On average, this suggests a lessening, not a strengthening, of girls' commitment to science and math; why, to a considerable degree, girls begin to lose interest in science and mathematics beginning in the junior high years; and why girls come to view science and mathematics as the domain of boys.

In discussions of our findings our academic colleagues and our students often expressed initial skepticism about this conclusion. They questioned the implicit assumption that girls were more concerned about and affected by boys than vice versa. Certainly, adolescent as well as postadolescent boys are vitally interested in girls. What differs, we believe, is how this interest plays out in relative commitment to science and engineering occupations, how males come to view work as an obligation because they see themselves as family providers, whereas girls often come to view work and family as equally attractive alternatives.

Peers: Our Testing

At the time of college entrance (CIRP data), individuals were asked whether there was "some chance or a very good chance that he or she will marry while in college or within a year after college." At first glance this variable may appear to be a very indirect measure of peer influence; however, the variable directly reflects the precise effect peers are held to have on young women: the promotion of the importance of relations with men over academic values. Indeed, plans for marriage may be viewed as a very potent test of goal conflict/commitment (the third and final concept, which is discussed in the next section) for women versus men. Our reading of the literature is that it is indeed peers who are most fundamental to greater goal conflict for women than for men, specifically as that conflict reflects science and engineering versus family goals.

Generally, students who were relatively less interested in marriage and thus more committed to careers were more likely to plan to major in biological sciences than in other fields. Among white females, having plans for marriage in college had the largest positive effect on the probability of selecting liberal arts, with much smaller positive effects on the probabilities of selecting physical sciences and engineering. Among white males, the only positive marginal effect of having plans for marriage in college was on the probability of selecting business; the marriage variable had a negative effect on the probability of selecting each of the other majors, including science and engineering. Black women with expectations of marrying while in college were especially less likely to declare biological sciences as their first choice of major. Among Hispanic men marriage had a positive effect on the probability of selecting science and engineering, whereas for Hispanic women marriage plans exhibited the largest negative effect on the probability of selecting biological sciences and the largest positive effect on the probability of choosing "Other," which includes the "undecided major" category and might have indicated a relative lack of commitment to earning a degree. Our question here is how college peers might have impacted those plans.

In the 1980 CIRP follow-up survey, respondents were asked whether they had ever married. The results for the total sample were strongly in support of the peer influence (and goal commitment) hypothesis: Across the board, marriage impacted the probability of obtaining a college degree far more for women than for men.(7) Among whites the impact was ten times as large for women; among black women it was twice as large; and, consistent with observations made elsewhere herein, Hispanic women were far less likely to have graduated if married, whereas Hispanic men were the only group to show a negative effect from never having married. In other words, having married increased substantially the likelihood that Hispanic men had achieved a degree. These results may suggest strong support for the importance of socialization and indirectly the role of peers in that process, but also for the role of culture - if indeed these influences can be separated at all.

The results from the NLSY are supportive, although the form of the data is somewhat different. Asked at the time of college entrance what they would "like to be doing at age 35," 20% of females, compared to less than 2% of males, selected "married, family." This difference directly demonstrated goals differences by gender, but indirectly may have reflected how those goals came to differ, that is, to some extent at least, through differences in socialization and, specifically, to differences in the influences of peers.

What about the science and engineering students? If we read the literature correctly, peer influence should have a strong effect on the underrepresentation of women, but at what stage of life? Are the effects felt primarily before college, during college, after college, or are they spread evenly across all stages? The answer would appear to be primarily before college, largely because the behavioral changes that begin for women at adolescence are cumulative; deficiencies such as fewer math courses taken often preclude later progress in these fields. The evidence from our empiricism is in two parts. First, female attrition in PSME is roughly the same as for males (NLSY). Indeed, although female attrition is 73% for women in the PSME, overall, compared to 72% for males, attrition in the physical sciences and engineering is actually slightly greater for males than for females, being 75% and 72%, respectively. Only in the biological sciences are the attrition rates uneven, favoring males to a considerable degree.(8) Second, the marginal effect of the marriage variable upon degree attainment in physical sciences and engineering among white women is small, which is perhaps a rather obvious result in light of the fact that female and male attrition rates are about even.

The original CIRP survey in 1971 contained some information that was highly relevant to the peer influence hypothesis, although a direct connection to the importance of peer values is lacking. The primary variables in question reflected incoming students' self-ratings in regard to general popularity and to popularity with the opposite sex. Among five related self-concept variables, popularity with the opposite sex was clearly the most potent predictor of not achieving degrees in the physical sciences and engineering; general popularity also was a negative factor. The literature would suggest that peer influence likely was a primary culprit. The fact that these patterns did not hold, but rather were very weak for white women in the biological sciences, was consistent with observations made elsewhere herein regarding why these women were less at risk than women in physical sciences and engineering. Finally, the relatively small coefficients for these variables among women in science and engineering almost certainly reflected major self-selection effects; that is, those who place a high value on attractiveness to men are less likely to enter science and engineering in the first place, once again drawing attention to behavioral changes that occur prior to college entrance.

We may also examine labor force status of science and engineering majors. Overall, employment status essentially is unrelated to having been married; however having never been married reduces the probability of full-time employment in science and engineering and increases the probability for the other categories (part-time employment, unemployment, and out of labor force) for white males, while increasing the probability of the former and decreasing the probability of the latter for white females.

Clearly, the effects of peers in socialization take their toll primarily prior to college admission. Perhaps collegiate peers help to prevent women from "catching up" later on. Although the "marriage variable" is at best an indirect measure of possible peer influence, marriage and child-rearing could be viewed as the ultimate tests of the influence of peers in leading women toward what Holland and Eisenhart (1990) call being Educated in Romance and away from science and engineering.

Goal Commitment

Necessarily, much has already been said about goal commitment. Obviously, self-efficacy and peer influence contribute directly to personal goal commitment. From the literature and from our findings, we infer that science and math self-efficacy largely form in elementary school but are often retarded for women during adolescence by the interaction of physiological changes and peer influences, sometimes resulting in reduced or conflicted commitment to science and engineering study and work.

Commitment is known to be the most potent predictor of persistence in almost all human endeavors (e.g., Ethington, Smart, & Pascarella, 1988; Sarkar, 1993). For women, their interactions with parents, K-12 science and math teachers, and peers appear often to lead to formulation of goals that are externally directed; that is, their important life goals frequently are formed largely in response to the desires of others. Women are socialized to seek to please. Thus it is not surprising that most women tend to be interested in careers emphasizing human interaction (as are many men), even if within science and engineering. The evidence is noteworthy.

Both men and women tend to view math and science more as "masculine" fields of study. For men this perception probably reinforces their tendency to persist in these fields; for women the reverse probably is true. Women are attracted to fields seen as more nurturing, fields such as the social sciences and humanities (Hackett & Betz, 1981). Men exhibit nurturing behaviors too, but predominantly when that nurturing is in fields perceived to be masculine. For example, men are quite willing to be helpful and supportive of women in math and science classrooms, laboratories, and on math and science homework.

From her study with Hewitt, Seymour (1992) concludes that female self-worth in science and engineering is, in fact, extrinsically based, and that for many females, selection of a college major in science and engineering is externally rather than internally driven. Persistence, too, is connected directly to degree of goal commitment, specifically in engineering (Jackson et al., 1993). Thus, women, more so than men, require a "genuine interest" in their chosen careers (Dick & Rallis, 1991). In short for many women commitment to science and engineering goals may be quite tenuous.

Men are known to be more "single-minded" than women in regard to work and careers (Eccles, 1987; Manis et al., 1989), a characteristic suggesting greater science and engineering goal commitment. Women perceive an incompatibility between careers in science and family life, and they see raising a family as an attractive alternative (Ware, 1988). The effects may be direct. Marion and Coladarci (1993) demonstrate that those women who place a high future value on family are less likely to take science and engineering courses; Seymour (1992) finds that female science and engineering students switch majors due to lack of personal goal commitment; and Lewis (1991) reports that those women who do wind up studying mathematics are "extremely job oriented" (p. 722).

The literature demonstrates that male/female differences regarding career and commitment continue into employment. Employed female scientists and engineers are less involved with their work than are their male counterparts although the gap narrows for those holding master's and doctoral degrees; for both bachelor's and master's degree holders, time for personal lives is rated as more important by women, compared with men (DiTomaso & Farris, 1994). Tobias (1992) raises questions about faculty prejudices against women, in particular bias against women who have strong family commitments.

We would expect that any individual who is marginally committed to any goal, in this case a science and engineering degree, would be less likely to tolerate unsatisfactory or adverse conditions than one who is fully committed. In short, all else equal and on average, men should be more likely than women to persist in science and engineering. Of course, we already know that most female "attrition" in science and math occurs prior to college entrance.

Goal Commitment: Our Testings

We have already noted that one of the largest family influences on science and engineering outcomes was whether parents were engaged in science and engineering occupations and how the effects were quite different for women and men. We spoke of the more "eclectic" decision structures of women in these regards. Also, we have observed that female commitments to science and engineering goals may be reinforced by having a well-educated mother.

From the CIRP files, commitment was best expressed by one's stated purpose in attending college. For the NLSY sample, identifying the desire to be employed in science or engineering at the age of 35 was easily the strongest predictor of majoring in PSME. The relationships were much stronger for men and for PSME, but also were strong for women and for the biological sciences. A strong desire to gain a general education and to prepare for graduate or professional school were the measures. Among those who had been enrolled in degree programs in the physical sciences and engineering, white women were less likely than white men to have expressed these goals at the time of admission, and although the stated desire to attend graduate or professional school had positive effects on the probabilities of enrolling in physical sciences or engineering for both white males and white females, these effects were generally higher for white females. To us, these results emphasize the importance to white females of being committed to science and engineering, if they are to achieve these ends: At the time of admission, white women are less committed to science and engineering, generally, than are white men, but those white women who are committed, as attested to by their graduate or professional school goals, will achieve in science and engineering at relatively high levels.

Arguably, an even stronger indication of commitment is whether one plans science and engineering as a likely career. In fact, for the CIRP students, specifying science and engineering as one's likely career in 1971 was strongly and positively associated with being employed in a science or engineering occupation in 1980, and for the NLSY subjects, desiring to be "employed as a scientist or engineer at age 35" was the strongest predictor. Further, the effect was noticeably larger for males compared with females and for PSME as opposed to biological sciences. Our conclusion is that, on average, women who select science and engineering majors may be less committed to science and engineering careers, which may be to say that they are more conflicted about career and family.

Conclusion

Probably the greatest realization that has come to us from this work is that studying the phases of science and math education separately is problematic. One cannot understand why women, and to some degree minorities, are underrepresented in science and engineering unless one understands that the related behaviors are formed over at least half a lifetime, but especially in the years prior to college. Although collegiate interventions no doubt can increase female (and minority) participation rates, the critical damage is done much earlier. The essential proof of this statement is that attrition rates in postsecondary education are very similar for men and women; it is the difference in the numbers of men and women who enter science and engineering curricula, the much-discussed "pipeline effect," that largely explains gender differences in graduation rates, graduate study, and employment in science and engineering. These gender differences also are evident for minorities. Science and engineering matriculation patterns of minority men are far more similar to those of majority men than to minority women. The gender disparities among minorities are especially large in the PSME, but again it is differences in science and engineering matriculation rates, not persistence rates, that are at odds. The patterns for the biological sciences are more mixed by gender, not only in regard to minority matriculation but also in regard to attrition rates. It is important to note that most of the problems are in the physical sciences and engineering, not in the biological sciences. Students in the biological sciences may be at no greater risk than students in general, or at least causal patterns are not at all clearly evident, even for white females.

The core contributors to participation in science and engineering, particularly PSME, are self-concept/self-efficacy, the influence of peers, and commitment. General self-concept forms at an early age and differentiates into mathematics and verbal self-concepts before adolescence; prior to this time, socializing forces are relatively evenhanded in the treatment of boys and girls, although the seeds of later disparities clearly have been sown through differential socialization. Adolescence is the critical period, a time when differentiated self-concepts begin to transform into math and science and other self-efficacies. For girls numerous actions of family members and education personnel, but especially those of adolescent peers, promote values at odds with math and science pursuits. By the beginning of high school, boys possess superior science and math characteristics, compared with girls. As a result, by the end of high school, on average, females have studied less science and math than have males, and the former are behind on many factors that contribute to the pursuit of science and engineering careers. By this time, on average, female aspirations, expectations, and especially commitments to science and engineering careers are markedly less than for men. Even those women who do enter science and engineering curricula - a group that is already highly "pruned" - are less committed, on average; they are more likely to be conflicted about career versus family. Nevertheless, both men and women respond negatively to many features of science and engineering study, especially in the introductory courses, and most of both genders migrate out of science and engineering, the proportions remaining being about equal. What can be done?

The primary foci must be on early interventions aimed at improving female science and math self-perceptions and counteracting pertinent influences of peers, interventions that will lead to greater science and engineering self-efficacy and commitment. Science and math self-perception is a function of one's knowledge of math and science concepts and one's confidence in math and science. Special efforts to expose female and minority students to elective math and science courses in their pre-college years is important to enhancing both the skill acquisition and the confidence necessary to making science a feasible choice for a college major. Increasing math and science course requirements should not be precluded. To reap the benefits from skill acquisition, female and minority students must come to believe that they can use math and science tools effectively. We have some ideas about the kind of environments that might contribute to improved science and engineering self-efficacy and commitment, in particular for women.

Families clearly can be highly instrumental to the science and engineering related aspirations and commitment of their children. Special attention should be given to matters of early socialization. Proactive rather than reactive strategies probably will prove most useful. A specific example of the power of early socialization is seen in how the educational levels and science and engineering occupations of parents impact the science and engineering achievements of their children. Why not exploit this relationship (as some programs already are doing), not only for the children of such professionals but also for females and minorities who lack such parents? Why not expose many more female and minority youth to the work environments of female and minority scientists and engineers? And why not target such youth for recruitment to science and engineering careers?

Clearly, we must develop also more and better interventions for the adolescent years, especially in support systems. Strategies might include focus groups that would make girls and minorities aware of how their aspirations are formed, in particular the role of peers, and how powerful can be the detractors from science and engineering careers. Individual counseling sessions probably should follow. One obvious strategy is to make girls aware, early on, of the costs of adherence to peer norms, for example, the costs of avoiding math and science courses in high school and the way they come to gain their perceptions of their own science and math abilities, skills, and attitudes.

At the postsecondary level, it is not uncommon for some colleges and universities to structure student living arrangements according to various "themes." Consideration should be given to structuring housing arrangements so that female and minority science and engineering majors can live in proximity to one another, thus permitting the reinforcement of science and engineering goals and proactively working against detractions. Professional counseling and counseling by science and engineering faculty probably would be productive, too.

Space limitations have caused us to concentrate here on what appear to be the central factors contributing to underrepresentation of women and minorities in science and engineering; however, our primary omissions have been of additional aspects of our three central factors, rather than of some unidentified factors. That is, we believe that other factors, such as education personnel, teaching methods, and structural traits, exert their influences chiefly through our three central concepts. What follows are a few illustrations of these additional aspects.

Evidence exists that in the high-school years, though well-meaning, science and math teachers fail to challenge young women as much as they should. Treating women as "the weaker sex" is no favor to them, at least when science and engineering careers are considered. Science and math teachers should interact with their female students with no more delicacy than they do their male students. They should challenge them, engage them rigorously in discussions, and expect high performance of them. They should insist that females, as well as males, do their own lab work, answer difficult questions, and justify their responses; in short, they should engage their female students much more in the give-and-take of intellectual discourse and provide them much less in the way of deferential, unchallenging treatment.

If these and other such approaches were carried out in precollege years, the problems of the college years would be more manageable; however, there still would remain the need to address many structural issues in college-level science and engineering programs, especially in the introductory courses. In the end the issue for science and engineering education will be how to make necessary changes while maintaining high quality. Science and engineering course work is rigorous, on average requiting more time and effort than course work in most other fields. Science and engineering programs usually attract more highly able students than do most other fields and on average produce relatively competent graduates; yet, the criticisms of these programs by their students, white men as well as others, cannot be discounted. We wonder, for example, whether in introductory physical sciences courses there are not better alternatives to the almost exclusive reliance upon mathematically based problem solving, whether a greater mix of strategies might not yield superior results? We wonder why cooperative learning strategies could not be encouraged, and why science and engineering faculty might not be persuaded to encourage rather than discourage students to persist in science and engineering programs? Our sense is that if science and engineering faculty merely became aware of the implications of many of their present practices they would be able to design adaptive strategies that would increase, not decrease, their program quality, overall. Our observations in outstanding science and engineering programs in private liberal arts colleges tell us that the proposed changes are clearly possible, even likely, where faculty sense that course enrollments must be maintained in order to maintain program viability. Perhaps the incentives and disincentives are not yet adequate in larger and public institutions.

We note the promise of one particular technique for environmental enhancement of science and engineering units. We observe from our findings that coming to know a professor or administrator is especially important for women in terms of raising the probabilities of persisting to higher degrees in the physical sciences and engineering. This finding supports considerable personal testimony from science and engineering professionals, especially members of minority groups, who emphasize that it was a particular individual, usually a faculty member, who was instrumental in their science and engineering persistence and success.

Finally, in this work we have set aside value judgments on whether women and minorities ought to pursue educational and career goals identical to those of men, either in general or in science and engineering in particular. Rather, our intent has been to shed light on what factors can improve the prospects of underrepresented groups entering and succeeding in science and engineering fields, should they choose to do so, in other words, the enhancement of free and open choices.

This material is based upon work sponsored by the National Science Foundation under Grant Number SBR-9311351. The Government has certain rights in this material. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Notes

1 Predicting future supply of and demand for scientists and engineers is a most hazardous task, as is shown in a recent critique of past forecasts. Although there presently appears to be adequate supply, conditions can and usually do change both often and rather quickly. What is more useful to science policy is identification and quantification of the factors that drive supply and demand. Further, apparent balances in the supply of and demand for scientists and engineers may mask poor worker quality; what may well be more important to productivity is the quality of those so employed (Leslie & Oaxaca, 1993).

2 Full details of the samples are described in the annual CIRP reports published by the Higher Education Research Institute at UCLA, e.g., Higher Education Research Institute, 1995, and in NLSY publications of the Department of Labor and the Ohio State University, which collected the data.

3 Complete tables and results are available from the authors upon request.

4 Sample sizes for Hispanics often were relatively small and thus were relatively less reliable.

5 Many of the studies that compose the core of existing research and present thinking are now quite dated. It is almost certain to be true that many gender-linked differences have changed as the women's movement has grown.

6 The "macho male" construct is widely known and appears to be something more than stereotype, although the trait almost certainly is more characteristic of working-class than nonworking-class Hispanic males, who admittedly are likely to be present in greater numbers among college-goers.

7 Clearly, marriage is a manifestation of many, many forces, only one of which is peer influence. Further, some of those other forces may well be more potent than peer influences, both in the adolescent years and later, in explaining marriage and career relationships.

8 Because the CIRP surveys oversample women and minorities and define outcomes in different ways, the attrition rates from the CIRP data are not strictly comparable to those from the NLSY; nevertheless, CIRP female attrition rates are quite similar to those of men as well, being a few percentage points less for women than for men, in both PSME and the biological sciences. CIRP minority attrition rates follow the same general pattern; of those who begin in PSME the share who graduates in PSME is about the same as for the majority, with minority womens' graduation rates being a few percentage points lower than minority mens', when compared with female and male majority rates, respectively. In the biological sciences, minority graduation rates are a bit lower than they are for the majority. The largest disparities are between Hispanic women and men, and they clearly favor the latter.

References

American Association of University Women. (1991). Shortchanging girls, shortchanging America: A nationwide poll to assess self-esteem, educational experiences, interest in math and science, and career aspirations of girls and boys ages 9-15. Washington, DC: AAUW.

Astin, A. W. (1993). What matters in college?: Four critical years revisited. San Francisco: Jossey-Bass.

Astin, A. W., & Astin, H. S. (1992). Undergraduate science education: The impact of different college environments on the educational pipeline in the sciences. Final Report to the National Science Foundation (Grant Number SPA-8955365). Los Angeles: The Higher Education Research Institute, UCLA.

Bandura, A. (1977). Self-efficacy theory in human agency. American Psychologist, 37, 191-215.

Barrett, M. C. (1996). Students' perceptions of their undergraduate science and engineering classroom experiences. Paper presented at the annual meeting of the Association for the Study of Higher Education, Memphis, TN.

Benbow, C. P., & Stanley, J. C. (1983). Sex differences in mathematical reasoning ability: More facts. Science, 222, 1029-1031.

Betz, N. E., & Hackett, G. (1983). The relationship of mathematics self-efficacy expectations to the selection of science-based college majors. Journal of Vocational Behavior, 23, 329-345.

Brush, L. R. (1980). Encouraging girls in mathematics: The problem and the solution. Cambridge, MA: Abt Associates.

Commission on Professionals in Science and Technology. (1994). Professional women and minorities: A total human resource data compendium (11th ed.). Washington, DC: Commission on Professionals in Science and Technology.

Dick, T. P., & Rallis, S. F. (1991). Factors and influences on high-school students' career choices. Journal of Research in Mathematics Education, 22, 281-292.

DiTomaso, N., & Farris, G. F. (1994, April). Degrees and diversity at work. IEEE Spectrum, 38-41.

Eccles, J. S. (1987). Gender roles and achievement patterns: An expectancy value perspective. In J. M. Reinish, L. A. Rosenblum, & S. A. Sanders (Eds.), Masculinity/femininity: Basic perspectives (pp. 240-280). New York: Oxford University Press.

Ernest, J. (1976). Mathematics and sex. Santa Barbara: University of California Press.

Ethington, C. A. (1995). Navigating the pipeline: Women's interpretations of socialization influences in science, engineering, and mathematics. Paper presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL.

Ethington, C. A., Smart, J. C., & Pascarella, E. T. (1988). Influences on women's entry into male-dominated occupations. Higher Education, 17, 545-562.

Friedman, L. (1989). Mathematics and the gender gap: A meta-analysis of recent studies on sex differences in mathematical tasks. Review of Educational Research, 59, 185-213.

Hackett, G., & Betz, N. E. (1981). A self-efficacy approach to the career development of women. Journal of Vocational Behavior, 18, 326-339.

Hackett, G., Betz, N. E., Casas, J. M., & Rocha-Singh, I. A. (1992). Gender, ethnicity and social cognitive factors predicting the academic achievement of students in engineering. Journal of Counseling Psychology, 39, 527-538.

Haven, E. W. (1971). Factors associated with the selection of advanced academic mathematics courses by girls in high school. Dissertation Abstracts International, 32, 1747a. (University Microfilms No. 71-26027)

Holland, D.C., & Eisenhart, M. A. (1990). Educated in romance. Chicago: University of Chicago Press.

Hyde, J. S., Fennema, E., & Lamon, S. J. (1990). Gender differences in mathematics performance: A meta-analysis. Psychological Bulletin, 107, 139-155.

Jacklin, C. (1979). Epilogue. In M. Wittig & A. Petersen (Eds.), Sex-related differences in cognitive functioning: Developmental Issues. New York: Academic Press.

Jackson, L. A., Gardner, P. D., & Sullivan, L. A. (1993). Engineering persistence: Past, present, and future factors and gender differences. Higher Education, 26, 227-246.

Jones, M. G., & Wheatley, J. (1990). Gender differences in teacher-student interaction in science classrooms. Journal of Research in Science Teaching, 27, 861-874.

Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1991). Mathematics self-efficacy: Sources and relation to science-based career choice. Journal of Counseling Psychology, 38, 424-430.

Leslie, L. L., & Oaxaca R. L. (1997). Variables impacting supply of women and minorities in science and engineering. In J. C. Smart, (Ed.), Higher education: Handbook of theory and research. Vol. 12. New York: Agathon Press.

Leslie, L. L., & Oaxaca, R. L. (1993). Scientist and engineer supply and demand. In J. C. Smart (Ed.), Higher education: Handbook of theory and research. Vol. 9 (pp. 154-211. New York: Agathon Press.

Lewis, D. J. (1991). Mathematics and women: The undergraduate school and pipeline. Notices of the American Mathematical Society, 36, 721-722.

Linn, M. C., & Hyde, J. S. (1989). Gender, mathematics, and science. Educational Researcher, 18, 17-19, 22-27.

Manis, J. D., Thomas, N. G., Sloat, B. F., & Davis, C. S. (1989). An analysis of factors affecting choice of majors in science, mathematics, and engineering at the University of Michigan. Ann Arbor: University of Michigan, Center for the Education of Women.

Marion, S. F., & Coladarci, T. (1993). Gender differences in science course-taking patterns among college undergraduates. Paper presented at the annual meeting of the American Educational Research Association, Atlanta, GA. ED363512.

Marsh, H. W, & Shavelson, R. (1985). Self-concept: Its multifaceted, hierarchical structure. Educational Psychologist, 20, 107-123.

Mason, C. L., Kahle, J. B., & Gardner, A. L. (1991). Draw-a-scientist: Future implications. School Science and Mathematics, 91, 193-198.

McClure, G. T. (1997). Variables impacting the supply of majority female and male scientists and engineers. Unpublished doctoral dissertation, University of Arizona, Tucson, AZ.

Meece, J. L., Parsons, J. E., Kaczala, C. M., Goff, S. R., & Futterman, R. (1982). Sex differences in math achievement: Toward a model of academic choice. Psychological Bulletin, 91, 324-348.

National Research Council. (1991). Women in science and engineering: Increasing their numbers in the 1990s. Washington, De: National Academy Press.

National Research Council. (1994). Survey of earned doctorates. Washington, DC: National Academy Press.

Nash, S.C. (1979). Sex role as a mediator of intellectual functioning. In M. A. Wittig & A. C. Petersen (Eds.), Sex-related differences in cognitive functioning : Developmental issues. New York: Academic Press.

Orenstein, P. (1994). School girls. Bantam: New York.

Parsons, J. E., & Adler, T. F. (1983). Expectancies, values, and academic behaviors. In J. T. Spence (Ed.), Perspectives on achievement and achievement motivation. San Francisco: Freeman.

Paulson, K., Hagedorn, L. S., Petrides, L.A., & Wenzel S. A. (1995). Connecting what we know: The experience of women in engineering education. Paper presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL.

Post, P., Stewart, M. A., & Smith, P. L. (1991). Self-efficacy, interest, and consideration of math/science and non-math/science occupations among black freshmen. Journal of Vocational Behavior, 38, 179-186.

Ross, K. C., Volkwein, J. F., & Vogt, W. P. (1995). Women engineering students: Factors associated with their persistence, career commitment, satisfaction, and self-confidence. Paper presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL.

Sadker, D. M., & Sadker, M. (1994). Failing at fairness. New York: Scribner's & Sons.

Sarkar, G. (1993). SIAST retention study. Factors affecting retention of first-year students in a Canadian technical institute of applied science and technology. Saskatoon: Sasketchewan Institute of Applied Science and Technology. (ERIC Document Reproduction Service No. ED 356 359)

Sax, L. (1996). The impact of college on post-college commitment to science careers.' Gender differences in a nine year follow-up of college freshmen. Paper presented at the annual meeting of the Association for the Study of Higher Education, Memphis, TN.

Seymour, E. (1992, March/April). Undergraduate problems with teaching and advising in SME majors: Explaining gender differences in attrition rates. Journal of College Science Teaching, 22, 245-254.

Seymour, E., & Hewitt, N.M. (1994). Talking about leaving: Factors contributing to high attrition rates. Final Report to the Alfred P. Sloan Foundation. Boulder, CO: Authors.

Shavelson, R. J., Hubner, J. J., & Stanton, G. C. (1976). Self-concept: Validation of construct interpretations. Review of Educational Research, 46, 407-441.

Sherman, J. (1980). Mathematics, spatial visualization, and related factors: Changes in girls and boys, grades 8-11. Journal of Educational Psychology, 72, 476-482.

Sherman, J., & Fennema, E. (1977). The study of mathematics by high school girls and boys: Related variables. American Education Research Journal, 14, 159-168.

Smith, T. E. (1992). Gender differences in the scientific achievement of adolescents: Effects of age and parental separation. Social Forces, 71, 469-484.

Tobias, S. (1992). Revitalizing undergraduate science: Why some things work and most don't. Tucson, AZ: Research Corporation.

Ware, N. C. (1988). Sex differences in choice of college science majors. American Educational Research Journal, 25, 593-614.

Worthley, J. S. (1992). Is science persistence a matter of values? Psychology of Women Quarterly, 16, 57-68.

Yaeger, P.M. (1995). The "chilly climate for women" revisited: A review of the literature. Paper presented at the annual meeting of the Association for the Study of Higher Education, Orlando, FL.

Zaruba K., & Dey E. (1996). Women, science, and degree aspirations: Does type of institution matter? Paper presented at the annual meeting of the Association for the Study of Higher Education, Memphis, TN.

Larry L. Leslie is a professor at the Center for the Study of Higher Education, The University of Arizona; Gregory T. McClure was a doctoral student at the Center for the Study of Higher Education, The University of Arizona, when this article was written; he now is an adjunct professor in education at St. Martin's College, Lacey, WA. Ronald L. Oaxaca is a professor in the Department of Economics, University of Arizona.

207573530 03271M-J0098ITSO 0277 A Study of Student Involvement in Community Service

Introduction

I learn more through my volunteer work than I ever do in any of my classes at school. Talking to people from diverse backgrounds provides so much insight that people just can't imagine. I study all these different theories in political science and sociology, but until you get a chance to see how the social world influences people's everyday lives, it just doesn't have that much meaning.

I have been involved in volunteer work ever since I was in high school, and I'll probably continue to do stuff like Habitat [for Humanity] until I'm old and gray. I get a lot out of working to serve others, and it's a good feeling to know that I have helped someone even if it's in some small way. It helps me to cherish people more and understand what life is all about.

The preceding comments are from college students who discussed their involvement in community service and the meaning they derive from such activities. Both of these students give voice to a form of learning that may be termed "citizenship education" in that a concern for the social good lies at the heart of the educational experience (Delve, Mintz, & Stewart, 1990). These students are reflective of others described throughout this article who through participation in community service explore their own identities and what it means to contribute to something larger than their individual lives.

In recent years, the role of higher education as a source of citizenship preparation has come to the forefront. In this regard, higher education reflects a rising tide of concern for national service and the common good, as programs such as AmeriCorps, Learn and Serve America, Habitat for Humanity, and Big Brothers and Big Sisters have evoked our most prominent leaders as well as citizens across the country to commit themselves to the service of others. The influence this national movement has had on the academy is most apparent in the growth of organizations such as Campus Compact and Campus Outreach Opportunity League (COOL) whose memberships and influence increased dramatically in the early 1990s (Markus, Howard, & King, 1993). Professional organizations associated with the academic enterprise also have added fuel to the growing concern over social responsibility and citizenship. For example, in 1997 the call for proposals from the American Association for Higher Education Conference on Faculty Roles and Rewards specifically identified an interest in how community service and service learning contribute to a more engaged faculty. The 1996 Annual Meeting of the American Educational Research Association was organized around the theme of "Research for Education in a Democratic Society," and at the 1995 American College Personnel Association Annual Convention, one of the keynote speakers, Dr. Robert Coles, addressed the issue of moral education when he called for greater commitment to service learning and community service.

Although it is hard to argue with calls to foster social responsibility among our students, our future leaders, there also is a tremendous need for clarification. With this said, the following key questions offer a guide for addressing some of the confusion revolving around community service: (1) Are community service and service learning interchangeable concepts or are there important differences? (2) What is the role of community service in engaging students as democratic citizens in a culturally diverse society? (3) Are there variations in the structure of service activities which produce different experiences for students? The first question is examined as I explore the relevant literature on community service and service learning. The second and third questions are addressed primarily through discussions of the theoretical perspective, findings, and implications. Thus, the latter two questions form the heart of the theoretical and empirical analysis offered throughout this article. In weaving theoretical and empirical work together to address these questions, I follow the tradition of critical theory and support the argument that all research is theoretically rooted: Sometimes the perspective of the author is spelled out (as in this case), while at other times it must be interpreted based on the assumptions undergirding the work (Tierney & Rhoads, 1993). This is by no means a rejection of empiricism in favor of theory, but instead should be understood as an effort to bridge the gap separating the two.

Community Service and Service Learning

Over recent years there has been an incredible growth in attention paid to community service and service learning (Jacoby & Associates, 1996; Kendall, 1990; Kraft, 1996; Kraft & Swadener, 1994; Rhoads, 1997; Waterman, 1997: Zlotkowski, 1995). The increasing interest in service reflects to a large degree a concern that institutions of higher education be more responsive to society and that higher learning in general ought to have greater relevance to public life (Boyer 1987, 1994; Study Group 1984; Wingspread Group, 1993). A convincing argument could be made that for American colleges and universities a commitment to service "is a movement whose time has come" (Rhoads & Howard, 1998, p. 1).

The issue to be addressed in this brief review of the literature concerns distinguishing community service from service learning. The primary difference between these two concepts is the direct connection service learning has to the academic mission. Typically, service learning includes student participation in community service but with additional learning objectives often associated with a student's program of study. For example, a student majoring in social work may participate in service activities at a local homeless shelter in conjunction with a course of study on urban poverty. Specific activities designed to assist the student in processing his or her experience are included as part of the service learning project. The student, for example, may be expected to write a reflective paper describing the experience and/or there may be small-group interactions among students involved in similar kinds of experiences. The learning objective might be to help students interpret social and economic policies through a more advanced understanding of the lived experiences of homeless citizens. Seen in this light, service learning seeks to connect community service experiences with tangible learning outcomes. Assessing such outcomes becomes a central concern of research and evaluation (Boss, 1994; Giles & Eyler, 1994).

Although service learning often is specifically tied to classroom-related community service in which concrete learning objectives exist, some writers suggest that student involvement in community service may be tied to out-of-class learning objectives and thus constitute a form of service learning as well (Jacoby & Associates, 1996; Rhoads, 1997). From this perspective, student affairs professionals who involve students in community service activities may engage in the practice of service learning when there are clearly articulated strategies designed to bridge experiential and developmental learning. The confusion between "class-related" versus "out-of-class-related" service learning led Rhoads and Howard (1998) to adopt the term "academic service learning" to distinguish the formal curriculum (largely faculty initiated) from the informal curriculum (largely student affairs initiated). Howard (1998), for example, defined academic service learning as "a pedagogical model that intentionally integrates academic learning and relevant community service" (p. 22). For Howard there are four components of academic service learning. First, it is a pedagogical model and is therefore to be understood as a teaching methodology. Second, academic service learning is intentional; that is, there are specific goals and objectives tying the service experience to course work. Third, there is integration between experiential and academic learning. And finally, the service experience must be relevant to the course of study. As Howard explains, "Serving in a soup kitchen is relevant for a course on social issues, but probably not for a course on civil engineering" (p. 22).

From an educational standpoint, it makes sense to link community service activities with intentional learning objectives whenever possible. Obviously, when student participation in community service can be connected to specific learning activities involving reflection, group interaction, writing, and so on, the experience is likely to have a greater impact on student learning and move into the realm of service learning (Cooper, 1998; Eyler, Giles, & Schmiede, 1996).

In addition to varying degrees of connection community service may have to academic learning objectives, there are also differing opinions on which goals of higher education service ought to address. Whereas Howard stresses the role of service as a pedagogical model used to assist in course-related learning, others see service (community service and service learning) as a key strategy for fostering citizenship (Harkavy & Benson, 1998; Mendel-Reyes, 1998; Rhoads, 1998). This vision of community service and service learning is captured most pointedly in the philosophical work of John Dewey, in which education is fundamentally linked to the social good and what it means to exist in relation to others.

Theoretical Perspective: Dewey, Mead, and Gilligan

This article is grounded in the philosophical work of John Dewey and his contention that education has a vital role to play in a democratic society. In his classic work Democracy and Education, Dewey argued that a democratic society demands a type of relational living in which one's decisions and actions must be made with regard to their effect on others. "A democracy is more than a form of government; it is primarily a mode of associated living, of conjoint communicated experience. The extension in space of the number of individuals who participate in an interest so that each has to refer his own action to that of others, and to consider the action of others to give point and direction to his own" (1916, p. 93). Dewey's vision of democracy challenges all citizens to take part in a form of decision making that balances the interests of oneself with those of others. Democracy seen in this light demands that individuals understand the lives and experiences of other members of a society. How else can we weigh the effect of our actions if others remain distant and unknown?

Implied throughout Dewey's conception of democracy is an ethic-of-care philosophy akin to the work of feminist scholars such as Gilligan (1982) and Young (1990), in which caring for others forms a core component of identity (often discussed as the "relational self'). This is conveyed in Dewey's view of liberty: "Liberty is that secure release and fulfillment of personal potentialities which take place only in rich and manifold association with others" (1927, p. 150). Recent political theorists such as Battistoni (1985) also have recognized the importance of developing relational understandings of social life. For example, Battistoni supported Tocqueville's (1945) claim that American democracy is dependent upon "the reciprocal influence of men upon one another" (p. 117). For Battistoni, reciprocal influence is fostered through participatory forms of education, which he claimed are more likely to foster citizens who see themselves as active participants in the political process. Similarly, in discussing the relationship between citizenship and education, Barber argued that citizens must recognize their dependence upon one another and that "our identity is forged through a dialectical relationship with others" (1992, p. 4). Barber calls attention to the idea that citizenship is fundamentally tied to identity. Mead and Gilligan provide additional insight into the connection between citizenship and identity through their respective concepts of the "social self" and the "relational self."

Mead's (1934) idea of the social self derives in part from James (1890) and Cooley (1902), who both suggested that an individual's self-conception derives from the responses of others mirrored back to the individual. Mead argued that the self forms out of the interaction between the "I" and the "me." The "I" is the individual acting out some sort of behavior; the individual doing something such as talking, listening, interacting with others, expressing an idea. The "me" relates to the sense one has about the "I" who is acting out a behavior or set of behaviors. The sense we develop about the "I" derives from the interpretations we suspect that others have of us. We cannot develop an initial sense about ourselves without the help of others, who provide feedback and interact with the behaving "I." Through the imagined thoughts of others, we envision ourselves as a "me" as we become the object of our own thoughts. According to Mead, an individual cannot develop a sense of self without the interactive context of a social group or a community. Therefore, the other, either the particularized or generalized other, is essential to the development of the self.

Feminist theorists such as Gilligan also have developed a conception of the self strongly rooted in otherness. Gilligan (1979, 1982) was one of the first theorists to point out that women often make moral decisions based on a sense of connection with others. She argued that women's moral decision making reflected a fundamental identity difference based on gender. Whereas men tend to seek autonomy and make moral decisions founded on abstract principles such as justice, women, in general, seek connectedness and weigh moral decisions based on maintaining or building relationships.

As a result of early child-parent interactions and ongoing gender socialization (which arguably begins at birth), relationships become central to the social world of women (Chodorow, 1974, 1978). For men, the relational quality of social life is often displaced by a strong sense of individualism. The other is fundamentally a part of women's experience and kept at somewhat of a distance for men. The development of the self for females may be characterized by connectedness. Male development may be characterized by individuation. These general patterns (which obviously vary in degree from one individual to the next) have significant implications for how males and females relate to others and how they understand themselves in the context of the social world.

Based in part on early feminist work, various scholars have argued that regardless of gender differences, society is likely to benefit when its members develop a commitment to caring (Larrabee, 1993; Noddings, 1984, 1992, 1995; Oliner & Oliner, 1995). This is poignantly noted by Sampson (1989), who argued,

The feminist perspective should no longer be understood as developing a psychology of women but, I believe, is better seen as developing a psychology of humanity tomorrow. The real issue, therefore, does not involve gender differences per se, as much as it speaks to an emerging theory of the person that is appropriate to the newly emerging shape of a globally linked world system. (p. 920)

Of course, Sampson's point about the "globally linked world" reminds us of an earlier issue raised in this article concerning how cultural diversity might influence citizenship education (recall key question Number 2: What is the role of community service in engaging students as democratic citizens in a culturally diverse society?). Arguably, as a society grows increasingly diverse, communications are likely to become more challenging. Cultural differences, though they may be understood as a source of community for learning and sharing among citizens (Tierney, 1993), nonetheless pose a significant challenge to social interaction and an individual's ability to connect with the other, who, in the case of a heterogeneous society, is likely to be a diverse other.

Woven together, Dewey, Mead, and Gilligan, among others, provide insight into how citizenship education might encompass learning about the self, the other, and the larger society in which one exists. The "caring self" is the term I use to capture the synthesis of their work. The caring self is intended to convey the idea of a socially oriented sense of self founded on an ethic of care and a commitment to the social good. Furthermore, it is reasonable to assume that community service, with its focus on caring for others, would offer excellent settings to explore the development of the caring self. But is this the case, and if so, in what kinds of service contexts are the qualities associated with the caring self likely to be forged?

This brings me to the crux of my argument and what I intend to shed light on through a study of student involvement in community service. Arguably, unless individuals have a deep sense of caring for others, it is less likely that they will engage in interactions with diverse others in a meaningful way. Caring may be seen as the solution to the challenge presented by a postmodern society characterized by difference. In essence, I contend that fostering a deep commitment to caring is the postmodern developmental dilemma all of education faces, including higher education. If we are to promote democratic citizenship in these challenging times, then we must foster in our citizens a commitment to caring. Higher education has a major part to play in this process, and involving students in community service may be one vehicle for meeting this challenge. The question that needs to be asked then, is, How and in what kinds of community service settings is caring to be fostered? Before addressing this question through a discussion of the findings, I first clarify the methodology used in conducting the study.

Methodology

The primary goal of this article is to advance understanding of community service as a strategy for citizenship education. Through a qualitative study of college students involved in community service, I shed light on various facets of the service context that may be most beneficial to challenging students as caring citizens. The focus is not on student learning per se; instead, I target the kind of meaning students construct about their service encounters as a means to identify important aspects of community service associated with caring. I need to be clear here. This article does not attempt to assess developmental change by examining student involvement in community service. Although such a strategy is important and falls in line with the tradition of student outcomes research (Astin, 1979, 1993; Feldman & Newcomb, 1970; Pascarella & Terenzini, 1991), this article takes more of a phenomenological direction in which the essence of community service is the primary concern. Hence, the kind of experiences students describe are important in this study, not as learning outcomes, but as indications of the nature of the service context.

The data for this article were derived from research and participation in community service projects conducted in conjunction with three universities: Pennsylvania State University, the University of South Carolina, and Michigan State University. Community service projects ranged from week-long intensive experiences requiring travel to distant out-of-state communities to ongoing student service projects in the local communities or states in which these universities are situated. I participated as a volunteer in many of the service projects described throughout this article. My role ranged from a staff supervisor in a few cases to that of a graduate student volunteer with limited responsibility in other instances. In every case, my primary role was as a volunteer and not as a researcher; the data I collected was more of an outgrowth of the community service experience and was not the central objective. The comments here are not meant to shortchange the research strategy employed, but instead are intended to clarify for the reader the context of my interactions and involvement with the student volunteers. In fact, my role as a volunteer may actually add strength to the naturalistic strategies used in collecting data as I was able to engage in ongoing and meaningful dialogue with the research participants (Denzin, 1989).

Based on the methodological strategies associated with naturalistic inquiry, data were collected using a variety of techniques, including formal and informal interviews, surveys, participant observation, and document analysis (Lincoln & Guba, 1985). The principal documents used as a source of data were journals students were asked to keep as part of their community service experience. The use of multiple data collection techniques provides a degree of triangulation and offers the researcher an opportunity to confirm or reject tentative interpretations (Denzin, 1989).

The early phase of the study was conducted in conjunction with Pennsylvania State University and the data obtained was part of a formal evaluation of community service activities by students. This phase of the project involved surveys of students' experiences and was considered program evaluation and as such did not require human subject approval at Penn State. The second phase, which primarily involved interviews and observations, necessitated gaining human subject approval. Students were informed of the study and given the opportunity to participate or decline. It was during this phase of the study that student journals were used, but only with student approval.

During the six-year period (1991-1996) in which data were collected, 108 students participated in interviews, 66 students completed open-ended surveys, and more than 200 students were observed at various project sites in which participant observation was central. Approximately 90% of the students involved in the community service projects were undergraduates, and about 10% were graduate students. The vast majority (approximately 80%) of the undergraduates were traditional-age students in the range of 18 to 24 years old. Females represented approximately 60% of the sample, and in terms of race, the majority were Caucasian (roughly 85%), with African Americans constituting the largest minority group - about 8 to 10% of the overall group.

Interview transcripts (from both formal and informal interviews), open-ended surveys, field notes from participant observation, student journals, and documents collected in conjunction with various service projects form the entire data base for the study. Once collected, the data were read repeatedly in an effort to identify important and relevant themes. The process followed the kind of analytical strategy stressed in the work of cultural anthropologists and interpretivists (Geertz, 1973; Rosaldo, 1989). Specifically, themes were identified based on their contextual significance and relevance to the overall goal of the project: to better understand the context of community service and how such activities might challenge students' understandings of citizenship and the social good. In a procedure described by Lincoln and Guba (1985) as "member checks," themes and interpretations were shared with several students as part of a process to obtain feedback and incorporate student reactions into the final manuscripts.

Based on the data analysis, several themes were identified. Three of those themes - students' explorations of the self, understandings of others, and views of the social good - form the basis for this article. Other issues, such as "student motivation" for getting involved in community service and "attitudes toward community service," are examples of additional themes that emerged from the data analysis but are peripheral to this article and thus are not discussed in any substantive way.

Findings

In keeping with the theoretical concern of democratic citizenship and fostering more caring selves, the findings are organized around three general concerns suggested by students in discussing their participation in community service: self-exploration, understanding others, and the social good. These themes are highly interactive and, in general, students' exploration in all of these areas contributes to understanding what I describe as the caring self.

Self-Exploration

Participation in community service is an educational activity that lends itself to identity clarification. For example, a student who was part of an intensive week-long community service project in South Carolina talked about identity issues and her participation in the project: "I'm kind of in a search for my own identity, and this trip is part of that search. I just don't know quite who I am yet. I'm struggling to figure it all out. These kinds of experiences help. I'm most genuine in these kinds of settings." Another student added, "Getting involved in community service helps me to get back in touch with who I really am. It reminds me that I have more to live for than merely myself." A third student offered the following comments:

I've always done service work. During my freshman year at USC [University of South Carolina] I worked on the City Year project and the Serv-a-thon. I believe service is an important part of leadership. It's important to give back to the community. The last four weeks I've been totally into myself, like running for vice president of the student body. I signed up for this project because I wanted to get outside myself for awhile.

This student saw the service project as an opportunity to connect with others and in her words "get outside" of herself. For her, the service project offered a chance to become more other-focused and to contribute to her community.

A second student described her involvement in community service as part of a journey to better understand herself: "My work as a volunteer has really helped me to see that I have so much more to understand about myself in order to grow. I'm still on the journey and have a long way to go." And a third student discussed what he learned about himself: "I got involved in volunteerism because I wanted to learn more about myself. I've learned how to love a wide range of people despite differences between us. I've learned not to be judgmental." A fourth offered insight into the kind of soul searching students often go through as a result of service work:

Sometimes I feel like I'm only fooling myself and that I'm really only into service so that I can help myself. I list this stuff on my resume and I feel guilty because I know it will help me get a teaching job. Is that why I do this? I know it makes me feel better about what I do in my spare time, but who am I really serving?

This student recognized, like others, the positive returns of service, not only in terms of experience helpful for landing employment, but for the feelings reflected back to the self.

Self-exploration through community service often involved a kind of self-interrogation that helped students to think more seriously about their lives. Listen to the following student as she recalled her volunteer work with troubled youth:

I got involved in a lot of self-esteem work, primarily with teenagers. It helped me to think more seriously about my understanding of myself and how others think of me. I began to wonder about what kind of person I was and was going to be. I began to ask questions of myself: "Am I too judgmental? Am I open to others? Am I sensitive to how other people see the world?"

Once again, the role of community service in challenging one's sense of self is clear. Equally clear is how one's sense of self is tied to the social context and the views others hold of us.

Understanding Others

A significant learning experience associated with community service was the opportunity to better understand the lives students worked to serve. Students were able to put faces and names with the alarming statistics and endless policy debates about homelessness as well as rural and urban poverty. As one student explained,

Expressing what it has meant to me to actually have the chance to engage in conversations with people who used to be total strangers is next to impossible. It has been eye opening. My understanding of homeless people was based on what I'd see on the news, in magazines, or on TV shows. They were not real people and I could easily turn my back on them and the problem in general.

Similar comments were offered by Penn State and Michigan State students involved in community service projects working with homeless citizens in DC, Louisville, and New York City:

Every homeless person has a name, a story.

They just want to be recognized and treated as human beings. There are names behind the statistics.

Working with the people of the streets has transformed "those people" into real faces, real lives, and real friends. I can no longer confront the issue without seeing the faces of my new friends. This has an incredible effect on my impetus to help.

All the statistics about homeless people and the stories of people freezing to death in the winter never really sunk in until I made friends with Harry and Reggie. There are faces now.

Students who worked in rural areas with low-income families also derived benefit from personal interactions with those they worked to serve. One student commented on the general outcomes associated with having personal interactions in service settings: "The whole experience helps you to see that others are real people and have real problems and yet can come together to help one another. . . . When you work with the people on their houses or in their back yard it adds to the experience. You get a chance to know the people. You have a face or a personality to go with the work." A second student stated, "The fact that we were able to interact a great deal with the people in the community added so much to the overall experience. I've done volunteer work in the past where I never really got the chance to meet with the people who I was actually trying to help." A third student, who participated in a week-long service project in a low-income rural area, added, "This week has taught me so much about other people and the problems they face in life. You can read about growing up poor, but getting to share a conversation with someone who has overcome so much during their lifetime is quite a different matter. . . . It's made me much less judgmental of others and their place in life."

A common point made by students was the fact that community service work with people of diverse cultural backgrounds forced students to confront generalizations they had of the other. For example, students talked about various stereotypes they held about poor people and how such stereotypes were erased as a result of their service work. Several students noted how surprised they were to find so many intelligent and educated people without jobs or places to live. One student maintained that the only accurate stereotype relates to the amount of bad luck that most homeless people have experienced. A second added, "I learned that all people are innately afraid and that no one deserves to be without a voice and a safe place and that stereotypes can be more damaging than can be fathomed." A third student talked about how his preconceptions about homeless people had been shattered through his interactions with them. As he explained, "This experience gave my beliefs and convictions about the homeless a personal basis that I'll never forget."

Many of the preconceptions students had about the poor were rooted in their limited experience with cultural diversity. Although socioeconomic factors were the primary source of difference between students and community members, race was another factor. Interactions with a variety of low-income individuals and families often challenged students' conceptions of the diverse other. Because the vast majority of the student volunteers were Caucasians and many of the community members served by the students were African Americans, a number of racial issues emerged from time to time. A Penn State sophomore talked about the difference she felt between herself and the large number of homeless African Americans she encountered during her volunteer work in DC: "I definitely felt a major barrier between Blacks and Whites in this country. There were times working in the soup kitchens where I felt very uncomfortable." A college junior studying mathematics commented on a similar feeling: "It was an experience for me simply to be placed in the awkward environment of walking around in predominantly African American, poor neighborhoods. I want to remember that feeling of insecurity. It reminds me of the vast differences between races in our society."

Often, issues of race and class blended together and challenged students' prejudices in a multifaceted way. Listen to the following two students discuss their experiences:

There is something that I'm not proud of and I always considered myself open-minded and not prejudiced, but when I worked at Sharon's house [Sharon is an African American woman who needed repairs done to her home] it reminded me that some of my previous thinking about the poor had been based on stereotypes. I mean I've always kind of thought in the back of my mind that people become poor or destitute because they are not motivated or not as intelligent. But Sharon has a master's degree and is very articulate. I see now that there may be many causes or barriers that people face that can limit them. It was an eye opener and I see now that I was carrying this misconception about them being to blame for their plight.

Meeting homeless people and talking with them taught me that some of my stereotypes about the poor, about Blacks, have been rooted in my own life of White, middle-class privilege. I have never had to work that hard to get a college education, for example, yet I've bought into the idea that others who have less than me are somehow lazy because they are poor. Heck, they may have worked twice as hard as I have. I've never really had my views of the poor challenged until this experience working with homeless people.

The generalizations and stereotypes to which students referred were seen by several as the by-product of the media. As one student, a senior in geography, pointed out, "I learned that my perceptions of poverty, crime, and homelessness are influenced and perhaps shaped by misconstrued images that I see on television." Another student also talked about how television had played a major role in how she had come to envision African Americans. She pointed out that in her rural Pennsylvania community, "there wasn't a single African American family. I never even met an African American until I attended college."

The Social Good

As one might expect, given the context of caring for others, issues related to the social good often surfaced. Community service is ripe for such discussions and offers a context conducive to serious thought about the larger social body. One student offered an example of the kind of serious thought that may evolve from community service work:

There are a lot of people in this country who need help to make ends meet. You can choose to help them or you can turn your back on them. I want to help people, and I want those who choose not to help to know that there are consequences for walking away. There are children who will go hungry and people who will be living in the streets. I cannot live with that on my conscience.

For this student, the social good suggests a world in which no one starves or goes homeless. Giving up some of his own time and energy to help others "make it" is in line with his vision of social responsibility.

Other students offered similar remarks about the social good. For example, one student commented, "Intellectual exploration has been rewarding but also suffocating at times and so I find the desire to commit myself to experiential work. I found one way could be by working in a homeless shelter and understanding social issues from a political standpoint as well as from the perspective of those living and breathing poverty."

For another student, the common good included the role of education in assisting the poor. He saw service as important, but there were deeper issues underlying poverty. He explained,

Service activities are important, but we also have to help teach people how to fish. You just can't give people food or build houses for them without also helping them develop the skills to take care of their own lives and their own families. . . . Part of my goal is to help others to develop their own abilities so that they can lead productive lives.

This student alludes to the idea that simply providing "bandages," though important and necessary, may not heal deeper wounds. In this case, the student highlights how sometimes the problems rest with the poor and their limited skills.

Other students also concerned with the deeper roots of economic inequities chose to focus on social structure instead of individuals as part of their effort to make sense of the social good. For example, one student saw community service as a stepping stone to larger work for social change: "I need to be in community with people who are interested in radical social change. Together we can work and witness all kinds of changes, and perhaps come closer to finding some answers." Another student alluded to the structural aspects of poverty as she discussed her learning experiences:

Community service is something that I think everybody should get involved doing. You see a different side of our country when you see some of the struggles the poor face. You begin to understand the barriers to their economic situation and why it is so hard to get out of poverty. I talked to this one woman, and she explained to me how expensive day care is for her children and that in order for her to take a job she needs to make at least eight to ten dollars an hour. And no one will pay her that.

For the preceding student, community service experience helped her identify a structural problem that limits low-income workers - the lack of affordable day care.

Not everyone who participated in this study saw service as necessarily a positive force for improving society. Listen to the following student take issue with some of the general comments he heard about the positive aspects of service:

To be honest, and it's hard to say this around all these "do gooders," I'm not sure all this volunteer stuff really does a whole lot of good. I know, I'm one of those volunteers too. But I keep asking myself a bunch of questions: "Am I doing this to help the homeless or am I doing this to help myself? Who really benefits?" Maybe I'm being too skeptical, but I think most of the students here are like me but won't admit it. It makes them feel good to help feed someone, and that way they can go back to living their happy little lives without feeling too guilty.

Despite the biting cynicism of the preceding student's comments, he makes an important point that turns our attention back to the theoretical thrust of this article: The idea that one often develops positive feelings about oneself as a result of involvement in service reminds us that our sense of self indeed is tied to others. When warm feelings are shared with a student engaged in service, then logically, that student may see him or herself in a more favorable manner. The interactional context is one reason why community service is so critical to forging more caring selves. Through acts designed to serve others, students learn to feel better about themselves. At the same time, their relationship to others and to the larger social body is strengthened. Hopefully, reaching out becomes a way of life and the diversity that offers the potential to divide one from another becomes instead a source of sharing, This is the essence of the caring self.

Implications for Structuring Community Service

As noted earlier, this study was phenomenological in nature. The study did not seek to determine whether students become more caring citizens as a result of their service. Instead, by approaching the subject phenomenologically, I was able to identify aspects of the community service context that might contribute to students' considerations of the self, others, and the social good. The underlying assumption of course is that such considerations are likely to contribute positively to one's ongoing development as a caring citizen. Thus, in thinking about the implications of the findings I was able to identify three structural components of community service that appear to be critical to advancing citizenship as defined in this article. These key components are mutuality, reflection, and personalization.

There are two aspects of mutuality I stress: One aspect relates to a recognition that both parties - the so called "doers" and the "done to," in Radest's (1993) terms - benefit from the service encounter. Students involved in service receive incredible rewards for their work in the form of personal satisfaction. And, if their work is effective, community members also receive rewards in the form of a service provided. Thus, one might say that the experience is mutual.

The gifts that students receive through their community service offerings are not without complications. Students frequently expressed a degree of guilt for feeling good about themselves as a result of their service to others. A line from the great American poet, Delmore Schwartz, comes to mind here: "Nothing is given which is not taken." Taking or "receiving" the gifts offered by community members is something students engaged in service must learn to do. In fact, effective leadership training for students ought to prepare them to be recipients of the rewards of service. "In giving, one must learn to receive," noted one student who worked with homeless citizens in Washington, DC.

The second aspect of mutuality relates to the structure of the relationship between service providers and community members who may receive a specific service. Too often we are guilty of determining the needs of those to be served with little to no involvement on their part. For community service to be most effective for the development of caring citizens, then, the planning of such activities ought to include those to be served in an equal and empowering manner. After all, Dewey's conception of democracy entails each person taking others into consideration when making decisions affecting the public realm.

A second key to making community service most effective for citizenship development is the inclusion of reflection as part of the service work. By the term "reflection" I refer to activities designed to help students process their service experiences in a manner involving serious thought. Small-group discussions and writing assignments are common tools used to foster student reflection. As is noted earlier in this article, community service that incorporates reflection moves closer to what is typically considered service learning in that the reflective activity helps to link service to an educational outcome.

Several of the service projects observed through this study did not involve structured reflection and the students' experiences suffered. One example occurred in New York City, where a young woman became so intimidated by her interactions with a homeless man who screamed profanities at her that she vowed to never again work with the homeless. The project she worked on was led entirely by students and there was no opportunity for guided reflection. In interviewing this student, I was left to ponder how her reaction might have been different had she been able to interact with an experienced facilitator. Would she have been able to work through her feelings and perhaps take something positive from the traumatic encounter?

Other examples from this study reveal the power of reflection. Recently, I accompanied a group of 23 students from Michigan State University to Merida in the Yucatan where we worked at a Salvation Army shelter for children, a low-income health facility, and a women's resource center. As part of helping MSU students process their experiences, staff volunteers facilitated reflection groups each evening after students returned from their work sites. At the end of the week, we evaluated the project and consistently students described the reflection activities as one of the highlights of their cross-cultural experience (despite the "educational" overtones such activities carried!).

Perhaps the most significant aspect of community service that I found to contribute to caring is what may be called the personalization of service. For community service to be challenging to a student's sense of self, it seems most beneficial for service to involve opportunities for meaningful interaction with those individuals to be served. Time and time again students discussed how significant it was for them to have the opportunity to interact with individuals and families on a personal basis.

Conclusion

The challenge of education to foster caring citizens has taken on enormous proportion in contemporary society as the struggle between individualism and social responsibility has taken on new meaning (Bellah, Sullivan, Swidler, & Tipton, 1985; Coles, 1993; Palmer, 1993; Parks Daloz, Keen, Keen, & Daloz Parks, 1996; Wuthnow, 1991, 1995). Community service is one option educators can select to enhance the development of citizens concerned with the social good. Caring is central to the effectiveness of community service, and thus students are challenged to give serious thought to what it means to care as they struggle to evaluate their commitment to the lives of others. Because the relationship between individuals and their obligation to one another is a cornerstone of democracy, community service may be seen to contribute to citizenship in a democratic society.

The students in this study highlight how cultural diversity poses additional challenges to one's development as a caring citizen. They described how community service often is an interaction between diverse others. This is the essence of Radest's (1993) argument when he maintained that community service may be seen as an "encounter with strangers" in which the challenge of service is that we each learn from the other and we each give as well as receive. From this perspective, community service represents a dialogical encounter with diverse others and serves as a bridge to build communal ties. Thus, community service offers one vehicle for preparing students to communicate in a culturally diverse world.

Finally, because service encourages students to see themselves as intimately connected to the other, a learning context is created in which the caring self is more likely to emerge. Fostering a sense of self grounded in an ethic of care is one of the central challenges of education and becomes increasingly important as our society grows more diverse. By fostering an ethic of care, higher education encourages the sense of otherness needed for democracy to survive and, indeed, thrive in a complex and fragmented social world.

References

Astin, A. W. (1979). Four critical years: Effects of college on beliefs, attitudes and knowledge. San Francisco: Jossey-Bass.

Astin, A. W. (1993). What matters in college?: Four critical years revisited. San Francisco: Jossey-Bass.

Barber, B. (1992). An aristocracy of everyone: The politics of education and the future of America. New York: Oxford University Press.

Battistoni, R. (1985). Public schooling and the education of democratic citizens. Jackson, MS: University Press of Mississippi.

Bellah, R. N., Sullivan, W. M., Swidler, A., & Tipton, S. M. (1985). Habits of the heart: Individualism and commitment in American life. New York: Harper & Row.

Boss, J. A. (1994). The effect of community service work on the moral development of college ethics students. Journal of Moral Education, 23(2), 183-198.

Boyer, E. L. (1987). College: The undergraduate experience in America. New York: Harper & Row.

Boyer, E. L. (1994, March 9). Creating the new American college. Chronicle of Higher Education, p. A 48.

Chodorow, N. (1974). Family structure and feminine personality. In M. Rosaldo & L. Lamphere (Eds.), Women, culture and society (pp. 43-66). Stanford, CA: Stanford University Press.

Chodorow, N. (1978). The reproduction of mothering: Psychoanalysis and the sociology of gender. Berkeley: University of California Press.

Coles, R. (1993). The call of service: A witness to idealism. Boston: Houghton Mifflin.

Cooley, C. H. (1902). Human nature and the social order New York: Charles Scribner's Sons.

Cooper, D. (1998). Reading, writing, and reflection. In R. A. Rhoads & J. P. F. Howard (Eds.), Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73 (pp. 47-56). San Francisco: Jossey-Bass.

Delve, C. I., Mintz, S. D., & Stewart, G. M. (1990). Community service as values education. New Directions for Student Services, no. 50. San Francisco: Jossey-Bass.

Denzin, N. (1989). The research act (3rd ed.). New York: Prentice-Hall.

Dewey, J. (1916). Democracy and education. New York: Macmillan.

Dewey, J. (1927). The public and its problems. New York: Henry Holt and Company.

Eyler, J., Giles, D. E., & Schmiede, A. (1996). A practitioner's guide to reflection in service-learning: Student voices and reflections. Washington, DC: Corporation for National Service.

Feldman, K. A., & Newcomb, T. M. (1970). The impact of college on students. San Francisco: Jossey-Bass.

Geertz, C. (1973). The interpretation of cultures. New York: Basic Books.

Giles, D. E., Jr., & Eyler, J. (1994). The impact of a college community service laboratory on students' personal, social, and cognitive outcomes. Journal of Adolescence, 17, 327-339.

Gilligan, C. (1979). Woman's place in man's life cycle. Harvard Educational Review, 9(4), 431-446.

Gilligan, C. (1982). In a different voice: Psychological theory and women's development. Cambridge, MA: Harvard University Press.

Harkavy, I., & Benson, L. (1998). De-Platonizing and democratizing education as the bases of service learning. In R. A. Rhoads & J. P. F. Howard (Eds.), Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73 (pp. 11-20). San Francisco: Jossey-Bass.

Howard, J. P. F. (1998). Academic service learning: A counternormative pedagogy. In R. A. Rhoads & J. P. F. Howard (Eds.), Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73 (pp. 21-29). San Francisco: Jossey-Bass.

Jacoby, B., & Associates. (Eds.). (1996). Service-learning in higher education: Concepts and practices. San Francisco: Jossey-Bass.

James, W. (1890). Principles of psychology (Vol. 1). New York: Holt.

Kendall, J. C. (Ed.). (1990). Combining service and learning: A resource book for community and public service (Vol. 1). Raleigh, NC: National Society for Experiential Education.

Kraft, R. J. (1996). Service learning. Education and Urban Society, 28(2), 131-159.

Kraft, R. J., & Swadener M. (Eds.). (1994). Building community: Service learning in the academic disciplines. Denver: Colorado Campus Compact.

Larrabee, M. J. (Ed.). (1993). An ethic of care: Feminist and interdisciplinary perspectives. New York: Routledge.

Lincoln, Y. S., & Guba, E.G. (1985). Naturalistic inquiry. Beverly Hills: Sage.

Markus, G. B., Howard, J. P. F., & King, D.C. (1993). Integrating community service and classroom instruction enhances learning: Results from an experiment. Educational Evaluation and Policy Analysis, 15(4), 410-419.

Mead, G. H. (1934). Mind, self, & society (Charles W. Morris, Ed.). Chicago: University of Chicago Press.

Mendel-Reyes, M. (1998). A pedagogy for citizenship: Service learning and democratic education. In R. A. Rhoads & J. P. F. Howard (Eds.), Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73 (pp. 31-38). San Francisco: Jossey-Bass.

Noddings, N. (1984). Caring: A feminine approach to ethics and moral education. Berkeley: University of California Press.

Noddings, N. (1992). The challenge to care in schools: An alternative approach to education. New York: Teachers College Press.

Noddings, N. (1995). Teaching themes of care. Phi Delta Kappan, 76(9), 675-679.

Oliner, P.M., & Oliner, S. P. (1995). Toward a caring society: Ideas into action. Westport, CT: Praeger.

Palmer, P. (1993), To know as we are known: Education as a spiritual journey. San Francisco: Harper.

Parks Daloz, S., Keen, C. H., Keen, J. P., & Daloz Parks, L. A. (1996). Common fire: Lives of commitment in a complex world. Boston: Beacon Press.

Pascarella, E. T., & Terenzini, P. T. (1991). How college affects students: Findings and insights from twenty years of research. San Francisco: Jossey-Bass.

Radest, H. (1993). Community service: Encounter with strangers. Westport, CT: Praeger.

Rhoads, R. A. (1997). Community service and higher learning: Explorations of the caring self. Albany: SUNY Press.

Rhoads, R. A. (1998). Critical multiculturalism and service learning. In R. A. Rhoads & J. P. F. Howard (Eds.), Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73 (pp. 39-46). San Francisco: Jossey-Bass.

Rhoads, R. A., & Howard, J. P. F. (1998). Academic service learning: A pedagogy of action and reflection. New Directions for Teaching and Learning, no. 73. San Francisco: Jossey-Bass.

Rosaldo, R. (1989). Culture & truth: The remaking of social analysis. Boston: Beacon.

Sampson, E. E. (1989). The challenge of social change for psychology: Globalization and psychology's theory of the person. American Psychologist, 44(6), 914-921.

Study Group on the Conditions of Excellence in American Higher Education. (1984). Involvement in learning. Washington, DC: National Institute of Education.

Tierney, W. G. (1993). Building communities of difference: Higher education in the 21st century. Westport, CT: Bergin & Garvey.

Tierney, W. G., & Rhoads, R. A. (1993). Postmodernism and critical theory in higher education: Implications for research and practice. In J. C. Smart (Ed.), Higher education: Handbook of theory and research (pp. 308-343). New York: Agathon.

Tocqueville, A., de (1945). Democracy in America. New York: Alfred A. Knopf.

Waterman, A. S. (Ed.). (1997). Service learning: Applications from the research. Mahwah, NJ: Lawrence Erlbaum Associates.

Wingspread Group on Higher Education. (1993). An American imperative: Higher expectations for higher education. Racine, WI: Johnson Foundation.

Wuthnow, R. (1991). Acts of compassion: Caring for others and helping ourselves. Princeton, NJ: Princeton University Press.

Wuthnow, R. (1995). Learning to care: Elementary kindness in an age of indifference. New York: Oxford University Press.

Young, I. M. (1990). The ideal of community and the politics of difference. In Linda J. Nicholson (Ed.), Feminism/postmodernism (pp. 300-323). New York: Routledge.

Zlotkowski, E. (1995). Does service learning have a future? Michigan Journal of Community Service Learning, 2 (1), 123-133.

Robert A. Rhoads is assistant professor in the Department of Educational Administration, Michigan State University
COPYRIGHT 1998 Ohio State University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Leslie, Larry L.; McClure, Gregory T.; Oaxaca, Ronald L.
Publication:Journal of Higher Education
Date:May 1, 1998
Words:24598
Previous Article:Factors associated with student loan default among different racial and ethnic groups.
Next Article:In the service of citizenship: a study of student involvement in community service.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters