You are on page 1of 11

l~/chological Bulletin Copyright 1987 by the American PsychologicalAssociation, Inc.

1987, Vol. 102, No. 3, 346-356 0033-2909/87f$00.75

Helping Behavior in Rural and Urban Environments: A Meta-Analysis


Nancy Mehrkens Steblay
Concordia College

A recta-analytic review examined 65 tests of the hypothesis that "country people are more helpful
than city people?' Results demonstrated a significantly greater nonurban helping response, with an
effect size of.29. However,the effect was found to be more dearly a function of context factors than
of subject factors, a finding that led to refinement of the initial hypothesis. The data indicate that
helping is more likely to occur in a nonurban than an urban context. This effect was found to be
robust across variations in helping requests and experimenter and subject variables. Discussion fo-
cuses on the unexpected positive relationship found between population size and helping rate in
small to medium-size communities. Theoretical models of urban impact and weaknesses in our
understanding of differences in rural-urban social behavior are also considered.

Social psychologists have responded with a wealth of empiri- Finally, it must be noted that some additional tests of the rural-
cal data to a question posed by researchers Darley and Latan6 urban hypothesis have reported no differences in helping rate
(1968) almost two decades ago: Why will bystanders not help between people in urban and rural settings(e.g., Holahan, 1977;
other persons in distress? The research interest was initially Korte, Ypma, & Toppen, 1975; Rotton, 1977).
spurred by accounts Of actual incidents in which groups of by- A quick tally of available research results for and against the
standers failed to respond to obvious needs of emergency vic- hypothesis indicates 46 tests of the hypothesis in support of
tims. Although Darley and Latan6 went on to focus on other greater rural helpfulness, 9 in support of greater urban helpful-
components of this phenomenon, a minor point of their early hess, and 10 studies reporting no significant differences. Al-
laboratory study (Darley & Latant, 1968) noted a significant though the original expectation for a greater rural helping rate
negative relationship between size of community in which a seems to be supported by the bulk of the literature, those results
subject was raised and later levels of helping behavior. This that do not conform to this pattern lead one to question the
finding confirms what perhaps for many is simple folk wisdom: robustness of this social rule. More important, the exceptions
"Country people are more helpful than city people?' Stereotypi- to the rule may well provide an aid to our understanding of why
cal expectations are not uncommon regarding the unhelpful, in most eases a rural-urban difference does exist. Identification
aloof behavior of urbanites and the contrasting helpful, friendly of a conspicuously absent (or present) factor in those studies
demeanor of small-town residents (Krupat & Guild, 1980; that do not support the hypothesis is likely to provide insight
Schneider & Mockus, 1974). into the causal nature of the phenomenon. Thus it appears ap-
This rather simple hypothesis that "country people are more propriate to take time to sort out the studies, to identify relevant
helpful than city people" has itself generated a great deal of re- explanatory variables, and to arrive at a conclusion both as to
search interest within the broader realm of helping behavior the content issue (Who helps more after all, and why?) and to
studies in general. Evidence for greater helping rates in rural the state of the research on this matter.
and small-town vicinities comes from many empirical studies At the outset, let me note that a primary goal of the present
(e.g., Amato, 1981; House & Wolf, 1978; Merrens, 1973; Rush- investigation is the precise definition of the hypothesized rural-
ton, 1978). However, the opposite outcome has at times also urban (R-U) effect. A first step, then, is to operationally define
been documented; that is, some research has shown urban the rural-urban variable. As noted earlier, Darley and Latan6
dwellers to be the more helpful (e.g., Hansson & Slade, 1977; (1968) drew attention to the relationship between a subject's
Hansson, Slade, & Slade, 1978; Weiner, 1976). Drawing again hometown size and his or her later helping behavior. The R - U
from folk wisdom, it might be explained that it is rural persons variable in this case was conceptualized essentially as a subject
who are distrustful, reserved, and protective of their privacy. variable and specifically measured through a demographic fac-
tor, hometown size. This demographic factor may be more
broadly interpreted as an indication of personality; that is, the
This research was funded by the Bush Foundation of St. Paul, Minne- assumption is that an upbringing in a small town versus an ur-
sota, through the Bush Scholars Program of Concordia College, Moor- ban setting produces differential adult behaviors in the individ-
head, Minnesota. uals from these two environments. The R - U hypothesis from
The assistance of Anna Erickson, Mark Everson, Darla Haugen, and
this perspective implies a lack of prosocial response by urban
Marilyn Nelson in the data collection and analysis for this article is
gratefully acknowledged. Thanks are also extended to Arthur L. Bea- residents that can be traced back to a stable disposition within
man and Larry Hedges for their helpful statistical expertise. those persons. Indeed, Wirth (1938) has referred to such a dis-
Correspondence concerning this article should be addressed to Nancy position as the "urban personality."
Mehrkens Steblay, Department of Psychology, Concordia College, The understanding of the rural-urban variable and its impact
Moorhead, Minnesota 56560. on behavior may be approached from another perspective as
346
RURAL-URBAN HELPING 347

well. It is possible that helping behavior is affected in a more needed to reverse the meta-analytic conclusion. The present re-
transitory way; that is, the environmental context within which view applies these procedures to the rural-urban helping litera-
the request for help occurs may well have an impact on individ- ture in order to integrate findings objectively. No meta-analytic
ual response. In this light the operational definition of the R - review of the R - U hypothesis is available in the literature at
U variable would more appropriately involve an assessment of present, although less comprehensive literary reviews (Amato,
helping context, rural versus urban, and the phenomenon de- 1983; Korte, 1980) are available for comparison purposes.
scribed in the R - U hypothesis would be explained through con- The goals of the present review may be summarized as fol-
textual factors. Along these lines, the means to understand the lows: (1) to examine the R - U literature, defining (a) the overall
hypothesized lower helping rate of the urban resident have been probability of a rural-urban difference in helping behavior, (b)
offered through theoretical models that do not stress hometown the size of this effect, and (c) a confidence level that the research
size and inferred disposition as a causal variable. For example, reviewed accurately represents all (fugitive) documents; (2) to
Milgram (1970) suggested a picture of urban life based on the examine a number of factors that may influence and help to
concept of overload; his thesis argues that city residents out of explain the R - U phenomenon; (3) to tie this statistical sum-
necessity adopt an unresponsive behavior mode in that environ- mary to a theoretical network; and (4) to identify gaps in our
ment in order to protect themselves from the overwhelming lev- knowledge and to point to necessary areas for future research.
els of social and environmental stimulation in the city. From At this point, I might add that Amato (1983) has provided
another perspective, Fischer (1976) speculated that the hetero- an excellent introduction to this task. He surveyed 18 previous
geneity of the population within large cities creates a context tests of the rural-urban hypothesis and identified specific meth-
wherein residents feel insecure about such diversity and subse- odological problems apparent in the overall research picture.
quently avoid social contact with dissimilar others. Finally, Pili- In an attempt to remedy some ofthese issues (most specifically,
avin, Dovidio, Gaertner, and Clark (198 l) hypothesized that the the previous nonrandom choice of communities examined and
potential helper who encounters a request for aid quickly the lack of representative samples of helping behaviors and
weighs the costs and benefits o f a prosocial response by using a community characteristics), Amato designed and conducted six
kind of"bystander calculus?' A differential rural-urban proso- additional tests of the hypothesis. He reported an overall trend
cial response should be expected to the extent that costs and in his six tests that supported the notion of a lower rate ofproso-
benefits are perceived in different ways in these different envi- cial responsiveness on the part of urban dwellers, with the major
ronments. decline in helping rate beginning in communities of 20,000
Given these intuitively appealing theoretical works, a pri- people. Amato's inquiry into various samples of helping behav-
mary task of this project, then, is to evaluate rural and urban iors indicated a tendency for small-town (rural) dwellers to be
helping rates as they are affected by subject disposition, situa- more helpful across a range of helping situations that require
tional context, or both and to attempt to see this phenomenon informal, casual, spontaneous behavior and less helpful in help-
more clearly through a data-based theoretical framework. A re- ing situations that require highly formal and planned behavior
lated theoretical necessity is the definition of "rural" and "ur- (such as filling out a census form). Thus, Amato's ambitious
ban" using the criterion of population size and a subsequent work (including the Pierce & Amato, 1980, taxonomy of help-
examination of the relationship between these numbers and ing behaviors) has provided a helpful framework for further in-
helping rate. Again, as an aid to the theoretical understanding vestigation of the R - U hypothesis.
of the phenomenon, it is helpful to assess the predictive power
of population indicators alone and to fit the revealed relation-
Method
ship into a data-based model. Finally, a more precise under-
standing of the dependent measure of interestwhelping behav- Sample
iormmay also serve to direct attention to relevant explanatory
variables. The investigation of potential moderator variables An extensive search of the psychological literature was conducted to
such as type of behavior requested and victim characteristics obtain the most complete set of hypothesis tests possible for inclusion
may clarify the R - U phenomenon. in the analysis. Several steps were taken to locate potential studies, the
first being a computer search of two abstract services, PsycINFO and
The form of the present investigation is that of a statistical
SOCIALSCISEARCH~Next, a manual search of PsychologicalAbstracts,
review of all available literature. Thirty-five studies have been Sociological Abstracts, Social Science Citation Index. Monthly Catalog
located that address the issue of rural-urban differences in help- of U.S. Government Publications. and the library book catalog was un-
ing behavior; these include 65 tests of this R - U hypothesis. This dertaken. Subsequently, a complete search of relevant references cited
body of literature has examined conceptually similar variables by any of the previously located articles was performed. Finally, authors
and allows reliable application of the statistical procedures used of primary articles were contacted by mail and asked to send any rele-
in meta-analysis. Meta-anaiytic procedures allow reviewers to vant works, published or unpublished. The final sample included 32
provide more precise and confident conclusions than are possi- published reports, 2 convention papers, and I unpublished report. Pos-
ble with a "literary review" (Cooper, 1979), in which conclu- sible selection bias in published studies was a concern and, as will be
sions are drawn from one's overall impressions gathered from discussed later, was partially addressed through calculation of a fail-
safe N.
reading studies without applying statistical techniques. Among
An individual study was selected for inclusion in the review if the
other information, meta-analysis allows computation of the author(s) defined the study as a test of the difference in rural versus
overall probability that the pattern of results in a set of studies urban helping behavior and used a helping measure as the dependent
was due to chance, an overall effect size estimate, and a figure variable. At this initial point in the reviewthe range of helping behaviors
regarding the number of additional nonsupportive studies was quite diverse, including 17 separate behavioral categories ranging
348 NANCY MEHRKENS STEBLAY

from simple requests (e.g., give time of day) to quite demanding pleas marion results in differences between 0s that are equally detectable re-
(e.g., help an injured pedestrian). Similarly, the rural-urban compari- gardless of the relative sizes of the proportions. In its simplest form, the
son was operationally defined differentially by various authors as a sub- combined effect size of all studies is calculated as an arithmetic mean.
ject versus a context factoI; and the size of communities defined as rural For this review, the mean effect size thus represents the overall impact
or urban differed from study to study. In all but two eases, the R - U of the rural-urban variable on helping behavior.
hypothesis test was clearly a key concern of each of the studies; the re- A number of considerations are important, however, before the over-
maining two studies (Darley & Latant, 1968; Korte, 1970) addressed all effect size is computed. First, Hedges (1984) has shown the standard
the question only as a minor point but still in a manner relevant to this effect size estimator (g) of an individual study to be biased with respect
work. Although it was not a criterion for inclusion, all studies did define to population effect size, particularly in studies with small sample sizes.
the "victim" (or requester) as a stranger. The recommended correction (c,), however, is not suitable for effect size
The final sample consisted of 65 sets of data testing the R - U hypothe- estimators based on differences between proportions (h), as in this study.
sis, from 35 empirical reports representing 14,221 subjects. The review Thus the correction was not advised or calculated in this review
covered those studies available as of September 1985. (Hedges, personal communication, September 1985). It might be noted,
however, that the mean sample size in this selection of studies was 236;
thus the threat of this bias seems minimal.
Statistical Analysis A second step recommended by Hedges (1984)ma test of homogene-
ity of effect size--was also limited by the statistic (h) used in this review.
Z score. Following the work of Rosenthal (1984), I calculated an The direct calculation as recommended did not provide evidence of sig-
overall probability level associated with the observed pattern of results nificant differences in effect sizes across this group of studies. If existing,
by combining Z scores of individual tests of the hypothesis. Calculations such differences would violate assumptions about common sample
involved computing an exact (one-tailed) probability level for each test effect sizes and promote misleading conclusions about the population.
of the hypothesis, finding the corresponding Z score for each, summing However, because of the questionable validity of this specific test for use
these Z scores, and dividing by the square root of the total number of when dealing with differences between proportions, a distribution of
tests being combined, The probability level associated with this ob- effect sizes was also generated as an alternative indicator and in fact
tained overall standard normal deviate (Z~t~) then defined the likeli- proved to be useful. This step indicated groupings of studies with like
hood that the overall pattern of the data was generated by chance. effect sizes that could later be compared to subsets of data that had been
Given the convention of reporting imprecise p values (e.g. "p < .05" grouped for theoretical reasons.
o r " n o significant difference"), it was necessary to calculate a more pre- Finally, studies included in this review had different sample sizes. Be-
cise value for each test through reconstruction of the data or by return- cause the variance of an effect size estimator is a function of sample
ing to other statistics reported within the article. Recovery of sample size, individual effect size estimates will vary in accuracy from study to
sizes and proportion ofbelping responses per condition allowed calcula- study. To adjust for this, the effect size indicator was calculated with
tion of Z tests for the difference between proportions. For some studies, individual effect sizes weighted by the reciprocal of the effect size vari-
contact with the authors was necessary to obtain such information. For ance and the summation of these values divided by the sum of the
those tests (n = 8) in which retrieval of data was not possible, a conserva- weights. This procedure is considered analogous to that offered by
tive policy (Rosenthal, 1984) was followed: Reports of the absence of Hedges (1981) for mean differences and is appropriate for effect sizes
significant main effects that were not accompanied by test statistics were based on differences between arc sin transformed proportions (see Bea-
reconstructed as p = .50; reports of significant main effects with no man et al., 1983). In the remainder of this article, reference to "mean
statistics available were reconstructed as p = .05. effect size" denotes the weighted mean computed as above.
Following the technique of Beaman, Cole, Klentz, Preston, and In four cases, reported information was not adequate for calculation
Steblay (1983), I calculated three summary standard normal deviates of the effect size indicator h. In these cases a rough estimate of h was
(Z scores) with the sample. The overall Z score as just discussed (Z~t~) obtained through calculations of other estimators (d or 0). Due to the
provided an unweighted estimate of the overall probability level. In ad- imprecise nature of this calculation, care was taken in subsequent analy-
dition, a Z score was calculated that included weighted individual test ses to note the impact of these figures on overall results generated.
Zs (Z~vr~).A weighting by sample size of the study provided an estimate Fail-safe N. In conjunction with the computation of the recta-ana-
of population parameters that allowed greater emphasis on larger sam- lytic normal deviate, a fail-safe N was calculated following the proce-
ples (with accordingly more reliable estimates). dure of Rosenthal (1978). This method results in an estimate of the
A final weighted Z (ZMF) was computed using fractional weights number of additional tests showing a summed null relationship ( Z score
equivalent to the reciprocal of the number of tests contained within of zero) that would be needed in order to increase the probability figure
each study. This weighting technique essentially adjusted for the nonin- attained through the meta-analysis to a point beyond the desired level
dependence of hypothesis tests within single studies. of significance. Although the works reviewed did cover a wide spectrum
Study characteristics. Several methodological and theoretical charac- of publishing outlets, only three studies were unpublished. In addition,
teristics were recorded. Among the methodological were date of publi- a publication bias that favors confirming studies may exist. Thus the
cation, source (published, convention paper, unpublished manuscript), fail-safe calculation is likely to be somewhat inflated but nevertheless a
author, number of hypothesis tests per study, and sample size. best available estimate.
Theoretical characteristics included population location, population
size, type of helping behavior, and experimenter and subject age (adult/
child) and sex. Type of helping behavior was coded by using two for- Results
mats: (a) the Amato (1983) throe-dimension taxonomy (dimensions in-
clude formal-informal, serious-nonserious, active-passive) and (b) a Meta-Analysis 1: All Studies
simple list based on the 17 mutually exclusive categories encompassing
all of the behaviors from studies included. The 65 experimental tests o f the R - U hypothesis yielded a
Effect size. The effect size computed for each hypothesis test is the ZMA o f 13.15 (p < .0001), indicating a greater prosocial re-
coefficient recommended by Cohen (1977) for use when testing differ- sponse in the rural condition that was unlikely to have been
ences between proportions. This arc sin index (0 = 2 arc sinVP) corre- generated by chance alone. Weighting the individual tests by
sponds to effect size indices for other statistical tests in that the transfor- their sample sizes produced a ZMN o f 8.48, still highly signifi-
RURAL-URBAN HELPING 349

cant (p < .0001). A weighting based on the number of tests per "rural" and "urban" when these were available; thus there was
study produced ZMF = 6.26 (p < .0001). Thus the R - U differ- some overlap between categories in midrange cities. It was also
ence in helping behavior appears to be strongly evidenced by necessary, however, to use some basic decision rules. On the ba-
this group of studies. This is further indicated by calculation of sis, intuitively, o f a perceived trend and, more theoretically, on
the fall-safe N (Nvs). For ZMA, this figure is 4,089, which sug- the work of Amato (1983), I considered populations in the range
gests that this large number of unreported and unsupportive of 0-50,000 as rural and those in the 100,000+ range as urban.
studies would need to exist before the p figure would change to In some studies, researchers collapsed multiple population con-
a nonsignificant level. This well exceeds the tolerance level (335) ditions, using this rule o f thumb to produce a single R - U com-
computed as directed by Rosenthal (1984). Even the smallest Z parison. "Medium-size" towns, which were defined and com-
derived (ZMF = 6.26) still generated an N ~ o f 877. pared as such (usually 50,000-100,000), were not included in
The effect size indicated for the group o f 65 tests is 0.23. This the primary analyses but were addressed in a subsequent analy-
is a relatively small effect size; however, in practical terms it sis to be discussed later.
suggests that the mean helping rate moves from the 50th to the Subject involvement. The dependent measure of helping be-
59th percentile as urban versus rural environments are com- havior was conceptually defined in the majority of research re-
pared. Thus on the average the helping rate of 16.33% of this viewed as a subject's personal assistance to the requester's plea.
population is affected by this R - U variable (area of nonoverlap; In contrast, six of the hypothesis tests (Bar-Tal et al., 1981; Biz-
Cohen, 1977)) '2 man et al., 1978; Holahan, 1977) utilized paradigms in which
subjects were not asked to become personally involved in offer-
Meta-Analysis 2: Pure Tests of the Rural-Urban ing aid but rather were asked to verbally respond to a hypotheti-
Hypothesis cal scenario involving a person in need (e.g., judging the likeli-
hood that aid would be given by others or the appropriateness
As indicated previously, the inclusion of studies at the outset of the requested behavior). Intuitively, this type of test does not
of this review was quite unrestricted, the only criteria being that appear to be a clear test of the R - U hypothesis. Statistical analy-
a study be defined as a test of the rural-urban hypothesis and sis generating a ZMA of 0.0 and an h of 0.0 reinforced this as-
that a helping measure be assessed as the dependent variable. sumption that these six tests represent a separate behavioral
In reviewing the 35 studies, it became apparent that the defini- phenomenon.
tion of key variables differed among research teams. A first step Elimination o f those (l 3) tests that failed to define the helping
was then to identify the effects of differential variable definition or rural-urban variables in the normative (contextual) manner
on study outcome. should thus leave only "pure" tests o f the R - U phenomenon.
Context~backgroundcomparison. An immediate task was to Such a group is represented by the 52-test subset discussed in
operationalize the rural-urban variable, because the definition the previous section (ZMA = 13.85, h = .29), because the six
of this key component differed among authors. More specifi- "hypothetical scenario" tests discussed above are included in
cally, the primary mode of definition treated the context of the the 13 tests dismissed as noncontextual variables. Subsequent
study as the variable for comparison--that is, the helping test analyses reported are based on this subset o f 52 pure hypothesis
took place either in a rural area or in the city. A much smaller tests.
group of researchers )seated the comparison as a subject vari-
able, identifying a demographic variable, population of the city
of subject upbringing, as a critical factor (Bar-Tal, Raviv, & Meta-Analyses of Subsets of Data
Shavit, 1981; Bizman et al., 1978; Darley & Latanr, 1968; Gel-
It is valuable to go beyond the original primary analyses to
fand, Hartmann, Walker, & Page, 1973; Hansson et al., 1978;
seek information regarding relationships between various sub-
Holahan, 1977; Korte, 1970; Nadler, Romek, & Shapira-Fried-
sets of the data. This allows investigation into both methodolog-
man, 1979; Weiner, 1976). In both sets o f studies, of course,
these two factors may be somewhat confounded. Nevertheless
it appears important to review separately these two subsets of
The average effect size estimate of those four studies for which h
data. could not be calculated directly was .83, obviously a much higher figure
Thus analyzed, those 52 tests (from 26 studies) that defined than the .23 mentioned above. If these four studies are eliminated from
R - U as a context variable yielded a ZMA of 13.85, p < .0001, the above calculations of effect size, the end figures change very little,
an effect size of.29, a and an Nrs of 3,636. Quite different results however (N = 61, h = .21). In retrospect, it appears that the "rough"
were obtained from those 13 tests that defined R - U as a subject estimates obtained for these four studies were inflated.
variable. This group of data produced the following figures: 2 It might be noted that in the analysis, the weighted means differed
ZMA = 1.70, p = .04, h = .00, Nrs = 1.4 Given these figures, it only slightly from the arithmetic mean (in this case, the arithmetic
appears that rural-urban differences in helping behavior should mean was .26). Although the discrepancies between the two methods
be expected when the R - U variable is defined as a situational appear to be slight, the magnitude is important, especiallywhen discuss-
ing small effect sizes, and the more precise estimate is preferred.
variable, not necessarily when it is a subject factor.
3 Area of nonoveflap is 20.59% mean rural helping rate at 61st per-
A related issue in definition of the R - U variable was the size centile.
of communities represented. Population figures varied greatly 4 ZMFand ZMNwere also calculated for these and the following sub-
between studies, with "rural" populations ranging from 400 to groups, typically resulting in a reduced ZMN and an even somewhat
99,999 and "urban" populations from 60,000 to 7+ million. smaller ZMF. However, such figures are reported only if they differ in a
The meta-analytic calculations used researchers' definitions of noteworthy manner from the ZMA.
350 NANCY MEHRKENS STEBLAY

ical issues and theoretical underpinnings of the hypothesized 13 categories. For the purpose of identifying the relationship
R - U relationship. between type of behavior required and helping responses in ru-
Published versus unpublished data. Forty-eight R - U tests ral versus urban environments, the above subsets of studies were
were located in published format, whereas only four unpub- examined (see Table l).
lished tests were found. The comparison of statistics from these In all six Amato subsets, Z scores were significant (ps < .001),
two subsets is quite striking. For the published tests, ZM^ = which conveys the pervasiveness of the rural-urban difference
14.12 (p < .000 l) and h =.30; for the unpublished tests, ZMA = across quite different behavioral dimensions. Of particular in-
0.99 (p = . 16) and h = .04. The unpublished data are obviously terest are the strong effect sizes generated by the poles of
not supportive of the R - U difference, but this is not completely Amato's serious-nonserious dimension (hs = .56 and .42, re-
surprising given a probable bias toward publication of signifi- spectively), which would suggest that the R - U difference is not
cant effects. The fail-safe figures also suggest another important limited to trivial responses.
consideration: The unpublished set of tests presents a rather Experimenter and subject variables. The significant rural-
shaky conclusion (Nvs = -3), whereas the published studies urban difference (p = .003) was apparent whether the experi-
stand rather firm (Nvs = 3,939), suggesting reasonably strong menter (victim) was an adult or a child, male or female. How-
confidence in reliance on this published sample. ever, the difference was more pronounced for male victims com-
Date of publication. There was a significant relationship be- pared with females (hs = .55 vs..38, respectively) and for chil-
tween effect size and date of publication, with effect size increas- dren compared with adults (hs = .54 and .39). The mean
ing with more recent date of publication (r = .29, p = .02). An helping rate associated with female victims (70%) was higher
examination of mean R - U helping rates showed rural rates not than for male victims (42%), t(l l) = 2.54, p = .03, and child
to be associated with publication date but urban rates declining and adult victims received help at approximately equal helping
with time (r = -.15, p = . l 5). levels (67% and 62%, respectively), t(35) = 9 p = .72.
Author of publication. Two independent researchers, Amato All 52 samples included both male and female adult sub-
and Korte, have made major research contributions to this re- jects. Thus further subject variable analyses were not possible.
view sample. The work of each was analyzed as a subset. Relationship between population size and helping rate. A ba.
Amato's combined work, with 12 tests of the hypothesis, sic assumption regarding the rural-urban hypothesis is that as
yielded a lower effect size (h = .19) than did the overall review community size grows, the likelihood of prosocial behavior to-
sample. Perhaps this lessened effect size reflects Amato's in- ward a stranger should diminish; thus a negative correlation
tended diversity of sampled populations and behaviors. If should exist between tx~ulation and helping rate. However, a
Amato's studies are removed from the sample, overall effect size test of linear correlation for all population data points and their
rises to .38. The work of Korte (9 tests) produced the opposite corresponding percentage helping rates produced a correlation
pattern, a higher effect size (h = .42); however, if Korte's studies coefficient o f - . 1 0 (p = .17) and thus very limited support for
are removed from the review sample, the effect size of the re- the hypothesized relationship. A stronger association was found
maining studies does not change so dramatically (h = .25). between the population difference (rural versus urban) within
Geographicallocation of study. Only five countries are repre- each hypothesis test and the effect size achieved in that test (r =
sented in the data. Tests conducted in the United States (n = .33, p = .02). It appears that population level is related to the
24) and those in Canada (n = 9) produced similar results, with helping rate but not in the direct manner initially assumed.
Z scores of 8.30 and 6.49, ps < .000 l, and effect sizes of.36 and The lack of a strong negative linear relationship between
.32, respectively--these figures are a bit stronger than those of ulation size and helping rate is further indicated by a break-
the overall test sample. Israel (n = 2) also produced figures close down of the data into the subsets of rural and urban population
to the overall data, Z = 3.32, p < .0005, and h = .25; however, levels. Within the rural communities (0-50,000 residents) the
NFs = 6. Australia (n = 1 l) generated a Z of 6.32 and an h of correlation is .28 (p = .04), actually a significant relationship
914; Turkey (n = 3) produced the strongest figures, Z = 8.25, in the opposite direction from that predicted. In the urban com-
h = .57; and Holland's three studies produced a negligible munities (100,000+), there is virtually no linear relationship
effect, Z = -0.14, h = .0 l, Nvs = 0. between population and helping rate (r = .02, p = .45), which
Direct~indirect request. The type of help required of bystand- again does not support the expected pattern.
ers varied on two important dimensions, as discussed in this The overall pattern of the relationship between community
and the next section. One such dimension involved the mode size and helping rate is displayed in Figure 1. The seven levels
of request: Was the plea made directly to bystanders, or was of population size are based on Amato's (1983) five-level dis-
the request never explicitly verbalized, even though supposedly play, with further subdivisions added at the urban level (Levels
obvious? In both types of paradigms, the R - U difference in V and VI in the present figure)9
helping was readily apparent (Z scores = 13.21 and 5.30, ps < Community levels, population size, and numbers of commu-
.0001, direct vs. indirect, respectively), and produced similar nities represented at each level are shown in Table 2. It is note-
effect sizes (h = .28, n = 32, Nvs = 2,034, and h = .30, n = 19, worthy that the relatively lower levels of helping in communities
NFs = 178, respectively). of fewer than 5,000 residents are comparable to the level
Comparison of helping behaviors. The dependent variable of achieved in the very largest cities. In urban communities the
each study was coded in two ways: (a) according to Pierce and mean helping rate is still as high as 84% at population sizes of
Amato's (1980) six-category (three bipolar dimensions) taxon- 60,000-300,000; it then declines toward ranges closer to that of
omy of helping behaviors and (b) according to a simple list of the small towns. Also, if the data from medium-size communi-
mutually exclusive behaviors, which with these 52 tests yielded ties of 50,000-100,000 (not included in the analyses to this
RURAL-URBAN HELPING 351

Table 1
Comparisons of Helping Behaviors
Mean % helping rate
Amato (1983) code and
behavior code n ZMA h Nrs Urban Rural

Formal 4 6.59* .21 (2) 60 .76 .82


Participate in survey 4
Informal 2 2.99" .36 5 .53 .72
Correct overpayment 2
Serious 7 7.17* .56 126 .34 .59
Injured pedestrian 4 6.16* .56 52 .18 .45
Help lost child 3 3.90* .53 14 .56 .78
Nonserious 11 8.89* .42 (8) 311 .58 .74
Tell a color 1 4.34* .62 6 .55 .83
Change for coin 5 5.19" .39 45 .61 .75
Tell name 2 3.99* .38 10 .33 .51
Tell time 3 4.56* .37 20 .72 .84
Active 27 5.67* .14 (25) 294 .56 .62
Retrieve dropped item 5 0.69 .00 -4 .33 .36
Make phone call 7 3.70* .36 (6) 29 .70 .81
Return lost letter 9 3.08* .21 23 .46 .53
Give directions 6 3.60* - . t (5) 23 .79 .82
Passive 1 4.45* .76 6 .09 .40
Give donations 1

Note. Numbers in parentheses indicate number of studies included in calculations ofh if different from listed n. Not included are those with missing
data.
*p<.001.

point) are now examined, the mean helping rate of this group this notion would involve a comparison of number of bystand-
(68%) fits easily into the pattern established by this figure. ers and helping rates, across population levels. Although helping
Test of the bystander effect. A standard effect achieved in rates are available from the present sample, this group of studies
emergency helping research (e.g., Darley & Latan6, 1968) is was not constructed as tests of the bystander effect, and the lim-
that of a decreased likelihood of prosocial response as number ited reported information made it impossible to consider the
of bystanders increases. If urban areas are likely, just by virtue number of bystanders as a continuous variable. However, there
of their overall population size, to have more bystanders present was enough information available to allow definition of most of
in a given situation, then the lower helping rates of individuals the studies as either "bystanders absent" or "bystanders pres-
in urban areas may be attributable simply to the greater impact ent?' In cases rated as bystander present, these potential helpers
of the bystander effect in that location. Ideally, a direct test of had not intervened when the subject was tested, thus maximiz-
ing the likelihood of a bystander effect. In the cases judged by-
stander absent, each subject was led to believe that he or she
was the only available helper. Using this simple dichotomy, one
might predict that in those research paradigms in which by-
standers were present, an R - U difference in helping rate could
be attributed to differences in number of bystanders in the rural
versus urban context. However, a more critical comparison
would be between rural and urban helping rates within the
group of studies in which bystanders were not present. No R -

Table 2
Population Size and Numbers of Communities at Each Level
Level Population N

I <1,000 1
1I 1,000-4,999 4
III 5,000-19,999 23
IV 20,000-50,000 14
V 60,000-300,000 7
Figure L Mean percentage of subjects helping in communities VI 300,001-999,999 14
at seven levelsof population size. VII 1,000,000+ 29
352 NANCY MEHRKENS STEBLAY

U difference would be expected, because the bystander effect rural areas. This does not deny the power of the bystander effect
could not differentially affect this data. This was not the case. as a temporary inhibitor of prosocial behavior but rather sug-
Overall helping rates between the two subgroups were not sig- gests that the simple presence of more people is not directly a
nificantly different (Ms = .58 and .69, bystanders vs. no by- cause of the difference in helping rates between urban and rural
standers, respectively), t(44) = 1.30, p = .20, and the R - U communities.
difference was present in both groups: ZMA = 12.18, h = .30, The present data offer some additional information as to
and NFS = 1,939 for bystanders, and Z ~ = 7.50, h = .36, and those situational variables that are associated with strength of
Nr-s = 237 for no bystanders (ps < .001). the R - U effect. For example, female requesters were helped at
all population levels more than were male requesters; requests
Discussion from males produced the greater R - U difference in prosocial
response. Also, certain helping requests were likely to generate
Conclusions: Status of the Hypothesis a greater R - U difference in behavior, especially those that were
either very trivial or very serious. Additional factors did not play
A prominent feature of the meta-analytic results is the robust a role: The R - U effect was effectively generated through either
nature of the rural versus urban difference in helping behavior. a direct or indirect help request and for either adult or child
The highly significant difference between tested rural and urban requesters. Finally, geographical location produced only minor
groups was apparent from the first overall analysis through the variability in effect size (although this conclusion is based on a
many subset comparisons. Thus the difference in prosocial re- limited sample). It appears that the variables discussed thus
sponse is clearly not merely a chance finding but rather a reli- far--population size, victim characteristics, type of helping re-
able behavioral difference. Although the data were largely lim- quest, geographical location--may simply operate as modera-
ited to the published domain, the fail-safe N calculated lends tor variables in the R - U phenomenon. They appear to exert
support to the credibility and generality of this sample of stud- an impact but do not readily lend themselves as explanatory
ies. Finally, the effect size generated is not of great magnitude variables. Research efforts must now identify the relevant di-
but is well within a theoretically consistent range given that the mensions of urbanism that more directly underlie the R - U
rural-urban variable is only one of many that investigators rec- differences in helping response.
ognize as influential in the generation of a prosocial response. In a discussion of urban-nonurban differences in social be-
havior, Korte (1980) suggested that two complementary urban
Theoretical Implications models (Fischer, 1976; Milgram, 1970) may aid in our under-
standing o f the mechanisms of social contact between strangers
Critical to theoretical understanding of the R - U phenome- in the urban environment. The Milgram model stresses the
non is the operational definition of the effect. This recta-analysis characteristics of the city, particularly the high levels of stressful
clearly shows that the predominant R - U effect is achieved environmental input, that produce a generalized indifference to
through a comparison of context, not an examination of per- needs of strangers. According to this analysis, a city dweller's
sonality, at least as personality has been defined as hometown adaptive response to all types of outside stimulation is to ignore
size. The notion of context as a causal variable is also supported what is nonessential to personal goals. With respect to social
by results which reveal that behavior which subjects report as stimuli, it is those unknown persons (i.e., strangers) with no ob-
appropriate in a hypothetical helping situation (which generates vious claim on an individual's life who are most easily ignored.
no R - U difference) is not the same as the behavior produced It is important to note, relative to the data at hand, that the
9 when subjects are actually confronted with the real-life situa- nonresponsiveness to strangers that Milgram predicts is an
tion. This finding may be construed as a demonstration that effect limited to interactions between persons lacking social ties.
subjects' good intentions may be compromised by the situa- Thus the quality or closeness of personal relationships is not
tional factors involved as the real-life scenario is played out. The lessened in the city (as noted by Korte's, 1980, review of the
"urban personality" (Wirth, 1938) appears to be part of folklore relevant data) even though interactions between strangers may
rather than of reality; the hypothesis that "country people are be substantially restricted. The Milgram model concurs nicely
more helpful than city people" is simply not supported by the with the present data in its focus on a context rather than a
data. Rather, a more accurate statement is that "a stranger is subject variable as a causal factor in rural-urban behavior
more likely to be helped in the country than in the city." With differences.
this finding, the search for causal variables must narrow its fo- Fischer's (1976) explanation of urban impact stresses the di-
cus to examination of situational influences. versity of behavior and appearance that is likely to emerge in a
These data reveal the role of the most salient contextual fac- large city of heterogeneous population. Because an individual is
tor, population size, to be more complex than previously indi- less likely to feel secure in the midst of many dissimilar others,
cated. Although a relationship exists between the size of the the expectation is that social interactions will be inhibited. Thus
population difference in tested rural and urban samples and the the characterization of the needy stranger as a nonmember of
subsequent effect size of the R - U helping rate, there is not the the reference group or as unfamiliar, deviant, or potentially
anticipated direct inverse relationship between population and threatening produces nonresponsiveness of bystanders. Like the
helping rate. Rather, a positive relationship appears in nonur- Milgram (1970) model, the Fischer theory is appealing in its
ban samples, whereas no linear relationship is apparent in ur- recognition of the contextual origins of a social response. Per-
ban samples. In a related vein, the data do not suggest that a sonal subject factors may be outweighed by the presence of sa-
bystander effect is more likely to be a factor in cities than in lient situational cues.
RURAL-URBAN HELPING 353

Neither model can be critically tested with the present data, of the event as an emergency, (3) acceptance of responsibility
because the majority of studies did not report either stimulus for action, (4) decision regarding appropriate action, and (5)
load factors or subcultural patterns of the communities in- action. On the basis of the Fischer, Milgram, and Piliavin theo-
volved. One comparison to the present data is quite relevant, ries discussed above, it may be expected that this five-step sce-
however. Both models specifically address the basic R - U nario will be played out quite differently in the urban than in
difference in helping response and also offer the means to ex- the rural environment. First, the acquired nonresponse mode
plain the demonstrated relationship between population level oftbe city dweller as well as the excessive stimulation overall in
and helping rate. Milgram's (1970) model assumes that stimu- the city may make attention to an event less likely in that locale.
lus load is likely to increase with population size but not neces- Perhaps even if noticed, the event may next be construed as a
sarily in a linear fashion. Thus there should be an indirect con- common occurrence in the city rather than as an emergency
nection between population and helping rate, but other indica- demanding personal attention. The Fischer model might fur-
tors--density and noise level--are more directly connected to ther enlighten us as to the dynamics of this process by suggest-
helping rate. The Fischer (1976) model is also conducive to a fit ing that a city dweller might only feel comfortable taking re-
with these data, in that it also may declare population size to be sponsible action when certain about the appropriateness of his
only indirectly related to helping rate. This model assumes that or her response. This security may be difficult to come by if
the sheer size of an urban population will foster those subcul- social comparison reveals many dissimilar others besides the
tural differences that promote a nonhelping response. Thus ur- unknown requester. Evaluation apprehension is likely, as well
ban-nonurban differences in helping behavior can be expected, as uncertainty as to whether the self is an appropriate help giver,
but the heterogeneity of the urban environment is seen as the thus inhibiting helpful behavior at Steps 3 and 4. Finally, if the
relevant factor rather than population size. In fact, helping rates ratio of perceived risks to benefits is greater for anticipated pro-
may be quite high in isolated areas of urban subcultural domi- social intervention in the urban environment, the bystander's
nance, as indicated by the work of Korte and Ayvalioglu ( 1981). decision to act at Step 4 would predictably be in favor of eseape
Perhaps this model might even be stretched to account for the rather than intervention. The evaluation apprehension in the
data by assuming that a spirit of kinship or likeness and subse- presence of dissimilar others, mentioned above, may be just one
quent helpfulness may develop with increasing population size of those risks that become more salient in the urban context.
in rural to medium-size communities. Thus the key to under- As noted earlier, this uncertainty of the potential helper, or per-
standing the data from within these two frameworks would be ceived risk, may be diminished if the event is encountered
an assessment of the stimulus load and homogeneity oftbe com- within a subcultural neighborhood where strong norms for con-
munities tested. duct and supportive resources are perceived.
A third model may also be relevant here. Piliavin et al. (1981) In sum, the blending of these models appears appropriate as
suggested that in an emergency situation, the potential helper the helping process is elaborated in future research. Although
chooses a response after a quick appraisal of perceived costs the explanation of the R - U effect certainly is still speculative
and benefits. If the costs of providing help are greater than the regarding situational variables that directly foster it, these theo-
rewards of such behavior, the bystander is likely to pass respon- retical models, along with the present data, may guide us toward
sibility to others, to attempt escape from the unpleasant situa- future avenues of research. Consideration of factors characteris-
tion, or to distort perception of the actual need involved. As tic both of the physical and social environments is relevant. Sig-
with the previous two models, the size of the urban population nificant effects of environmental input factors (noise, density,
is not seen as a direct causal factor in the reduced helping rate traffic level) have been noted (Korte & Ayvalioglu, 1981; Korte
of urbanites. However, to the extent that perceived risk in- et al., 1975; Krupat & Coury, 1975; Milgram, 1970; Rushton,
creases with population and size, there may be an indirect con- 1978) along with related factors such as speed of activity (Gold-
nection between population and helping behavior. Other than man, Lewsandowski, & Carrill, 1982) and perceived vulnerabil-
in this limited sense, the Piliavin model does not well address ity to risk and crime (House & Wolf, 1978; Milgram, 1970).
the population-belping relationship demonstrated in this study. Similarly, the impact of victim characteristics and of subculture
However, it does agree nicely with the revealed greater response has been discussed in previous work (Fischer, 1976; Korte &
to female victims even in the city (as a salient low-risk group). Ayvalioglu, 1981 ). If we assume that social behavior within the
Finally, all three models may address the decreasing urban help- city is determined through a complex process of interacting fac-
ing rate across time. This effect may be a function of an increase tors, a logical future step would be to assess the relative impact
in the congestion of large cities (Milgram), the greater real or of each and their interactions on helping responses. Amato's
perceived risk (Piliavin), and/or the increased diversity, real or (1983) research effort provides an example of how such re-
perceived (Fischer), in our present-day urban areas. search might be approached. Amato attempted to assess the
In comparing these three models, it seems unlikely that one impact on helping response of various community-level vari-
alone might explain complex urban behavior. It is more reason- ables, including geographical isolation, tourism, residential sta-
able to assume that urban and nonurban prosociai response is bility, sex, age, and social class. These listed elements did not
affected by many contextual forces and cues. As illustration of contribute substantially to prediction; however, two additional
this complexity, the five-step model of helping behavior pro- variables measured--heterogeneity of population and environ-
posed by Latan6 and Darley (1970) may be considered. This mental load--demonstrated the relationship with helping be-
model suggests that potential helpers must advance through five havior predicted by the Fischer and Milgram models.
cognitive steps before a prosocial response can be expected. As a goal for future research, an effective explanatory vari-
These steps include (1) attention to an event, (2) interpretation able(s) should be able to account specifically for the R - U
354 NANCY MEHRKENS STEBLAY

difference in helping rate, as well as for the relationship between Canadian, and Australian locales. It is not clear to what extent
population size and prosocial rate, and, in addition, should lend the subject samples or helping behaviors investigated actually
insight into the reason for the increased effect size across time. represent the larger population of subjects and helping behav-
iors. The greater number of tests included in this review sample,
Comparison o f Meta-Analytic Conclusions With however, does add greater confidence to the conclusions.
Previous Literature This review does not address the need stated by Amato to
reduce reliance on city size as the sole indicator of urbanism.
The theoretical review by Korte (1980), discussed earlier, This is due to constraints of the original studies rather than to
concurs with the findings of this review. Korte noted limited a failure of the review method. The present review, however,
support for the idea that an individual's personal background does address a final methodological weakness of previous re-
(urban vs. nonurban) would predict prosocial response. Rather, search. It allows an assessment of the shape of the relationship
the data illuminate the urban-rural variable as a contextual fac- between urbanism and helping behavior, because it combines
tor and demonstrate strong evidence of its impact. 139+ communities of varying sizes. Indeed, the pattern of this
Perhaps the most conclusive empirical work prior to the pres- relationship is a critically important outcome of the review and
ent meta-analysis was reported by Amato (1983). Some com- points to the need to investigate alternative explanations for the
parisons between the present work and that of Amato may be R - U effect.
helpful to clarify theoretical and methodological issues. With In summary, the hypothesized difference in urban versus
his five field studies, Amato reported a clear trend for helping nonurban helping behavior is strongly evidenced by the avail-
rates to decrease as community size increased, with a major able research. This difference appears to be a function of situa-
decline in helping rate occurring at population sizes of about tional variables rather than a subject factor; the array of such
20,000. The present meta-analytic review, on the other hand, context variables involved in the effect is not completely de-
demonstrated a relationship between population size and help- fined but clearly includes both environmental and social char-
ing rate that is quite different. As noted earlier, the relationship acteristics. It is hoped that future research can identify such
established was actually positive in the nonurban sample, with variables and also extend findings to a broader realm of helping
a decline in helping rate only beginning at population sizes of behaviors in a wider contextual setting.
300,000. Thus the inclusion of more data in the meta-analytic
framework has produced conclusions at odds with Amato's References
smaller sample. This difference is particularly relevant theoreti- Amato, P. R. (1981). Urban-rural differences in helping: Behavior in
cally, because it necessitates a more complex explanation of R - Australia and the United States. Journal of Social Psychology, 114(2),
U effects. 289-290.
A point needing further clarification concerns the category Amato, E R. (1983). Helping behavior in urban and rural environ-
of formal helping behavior. Amato reported that cooperation in ments: Field studies based on a taxonomic organization of helping
completing the census was greater in communities of 4,999- episodes. Journal of Personality and Social Psychology, 45(3), 571-
586.
300,000 than in very small towns. Although for Amato this was
Bar-Tal, D., Raviv, A., & Shavit, N. (1981). Motives for helping behav-
an unusual trend, it actually concurs with the overall pattern of ior: Kibbutz and city children in kindergarten and school. Develol~
helping response indicated by this review. The slightly increased mental Psychology, 17(6), 766-772.
sample (n = 4) of the review also indicated that formal helping Beaman, A. L, Cole, C. M., Klentz, B., Preston, M., & Steblay, N. M.
behavior produced an effect similar to that of other behavioral (1983). Fifteen years of foot-in-the-door research: A recta-analysis.
categories. Personality and Social Psychology Bulletin, 9, 181-196.
There are similarities in the data as well. Amato recorded Bizman, A., Yinon, Y., Mivtzari, E., Shavit, R., Bar-Ilan, U., & Ramat-
high helping rates for a simple request--for a favorite colorm Gan, 1. (1978). Effects of the age structure of the kindergarten and
and for an unambiguous, easy request--giving directions. The altruistic behavior. Journal of School Psychology,, 16(2), 154-160.
present review also demonstrated high rates of helping for these Cohen, J. (1977). Statistical power analysis for the behavioral sciences
(rev. ed.). New York: Academic Press.
requests. Similarly, both reports indicated a much lower level
Cooper, H. M. (1979). Statistically combining independent studies: A
of helping for a request that demanded help that was difficult or recta-analysis of sex differences in conformity research. Journal of
costly to provide (hurt leg). Also, the five tests in this review Personality and Social Psychology, 37, 131-146.
that involved retrieving a dropped item were, as in Amato's Darley, J. M., & Latan~, B. (1968). Bystander intervention in emergen-
summary, unusually poor at producing a helping response or cies: Diffusion of responsibility. Journal of Personality and Social
an R - U difference. Psychologg, 8, 377-383.
Finally, the present review must concur with Amato's report Fischer, C. S. (1976). The urban experience. New York:Harcourt Brace
of methodological weaknesses in the overall research sample. Jovanovich.
Representative sampling of communities, subjects, and helping Gelfand, D. M., Hartmann, D. D., Walker, P., & Page, B. (1973). Who
behaviors was not common in the literature. This lack of sys- reports shoplifting? A field-experimental study. JournalofPersonality
and Social Psychology,, 25, 276-285.
tematic variations of key variables is perhaps most due to the
Goldman, M., Lewsandowski, H. E., & Carrill, R. E. (1982). Altruistic
nature of the research; field settings lend themselves well to con- behavior in rural and urban, residential and business areas. Basic and
venience samples rather than to perfect random selection. How- Applied Social Psycholog'g, 3(2), 155-160.
ever, it must then be recognized that the conclusions drawn Hansson, R. O., & Slade, K. M. (1977). Altruism toward a deviant in
must be limited in like manner. The R - U phenomenon has pri- city and small town. Journal of Applied Social Psychology, 7(3), 272-
marily been recorded with pedestrian traffic in various U.S., 279.
RURAL-URBAN HELPING 355

Hansson, R. O., Slade, K. M., & Slade, P. S. (1978). Urban-rural differ- Latan~, B., & Darley, J. M. (1970). The unresponsive bystander: Why
ences in responsiveness to an altruistic model. Journal of Social Psy- doesn't he help? New York: Appleton-Century-Crofts.
chology, 105(1), 99-105. Merrens, M. R. (1973). Nonemergency helping behavior in various
Hedges, L. V. (1981). Distribution theory for Glass' estimator of effect sized communities. Journal of Social Psychology,, 90(2), 327-338.
size and related estimators. Journal of Educational Statistics, 6, 107- Milgram, S. (1970). The experience of living in cities. Science, 167,
128. 1461-1468.
Hedges, L. V. (1984). Advances in statistical methods for meta-analysis.
Nadler, A., Romek, E., & Shapira-Friedman, A. (1979). Giving in the
In W. H. Yeaton & E M. Wortman (Eds.), Issues in data synthesis:
kibbutz: Pro-social behavior of city and kibbutz children as affected
New directionsfor program evaluation (Vol. 24, pp. 25--42). San Fran-
by social responsibility and social pressure. Journal of Cross-Cultural
cisco: Jossey-Bass.
Psychology, 10(1), 57-72.
Holahan, C. J. (1977). Effects of urban size and heterogeneity on judged
appropriateness of altruistic responses: Situational vs. subject vari- Pierce, P. L., & Amato, P. R. (1980). A taxonomy of helping: A multidi-
ables. Sociomet~ 40(4), 378-382. mensional scaling analysis. Social Psychology Quarterly, 43(4), 363-
House, J. S., & Wolf, S. (1978). Effects of urban residence on interper- 371.
sonal trust and helping behavior. Journal of Personality and Social Piliavin, J. A., Dovidio, J. E, Gaertner, S. L., & Clark, R. D., Ill. ( 1981).
Psychology, 36(9), 1029-1043. Emergency intervention. New York: Academic Press.
Korte, C. (1970). Effects of individual responsibility and group commu- Rosenthal, R. (1978). Combining results of independent studies. Psy-
nication on help giving in an emergency. Human Relations, 24, 149- chological Bulletin, 85, 185-193.
159. Rosenthal, R. (1984). Meta-analytic procedures for social research. Bev-
Korte, C. (1980). Urban-nonurban differences in social behavior and erly Hills, CA: Sage.
social psychological models of urban impact. Journal of Social Issues, Rotton, J. (1977). Sex, residential location, and altruism. Psychological
36, 29-51. Reports, 40(1), 102.
Korte, C., & Ayvalioglu, N. (1981). Helpfulness in Turkey: Cities, Rushton, J. (1978). Urban density and altruism: Helping strangers in a
towns, and urban villagers. Journal of Cross-Cultural Psychology,,
Canadian city, suburb, and small town. Psychological Reports, 43(3,
12(2), 123-131.
Pt. 1), 987-990.
Korte, C., Ypma, I., & Toppen, A. (1975). Helpfulness in Dutch society
as a function of urbanization and environmental input level. Journal Schneider, E W., & Mockus, Z. (1974). Failure to find a rural-urban
of Personality and Social Psycholog2z, 32(6), 996-1003. difference in incidence of altruistic behavior. Psychological Reports,
Krupat, E., & Coury, M. (1975). The lost letter technique and helping: 35(1, Pt. 1), 294.
An urban-nonurban comparison. Paper presented at the 83rd annual Weiner, E H. (1976). Altruism, ambiance, and action: The effects of
convention of the American Psychological Association, Chicago, IL. rural and urban rearing on helping behavior. Journal of Personality
Krupat, E., & Guild, W. (1980). Defining the city: The use of objective and Social Psychology, 34(1), 112-124.
and subjective measures for community description. Journal of So- Wirth, I. (1938). Urbanism as a way of life. American Journal of Sociol-
cial Issues, 36(3), 9-28. ogy, 44, 1-24.
(Appendix follows on next page)
356 NANCY MEHRKENS STEBLAY

Appendix

S t u d i e s I n c l u d e d i n t h e M e t a - A n a l y s i s ( N = 35)

Amato, P. R. (1978). City versus country: Friendliness and helping be- nication on help giving in an emergency. Human Relations. 24, 149-
havior in Queensland. Paper presented at the conference of the Socio- 159.
logical Association of Australia and New Zealand, Brisbane, Aus- Korte, C., & Ayvalioglu, N. (1981). Helpfulness in Turkey: Cities,
tralia. towns, and urban villagers. Journal of Cross-Cultural Psychology,
Amato, P. R. (1981). Urban-rural differences in helping: Behavior in 12(2), 123-131.
Australia and the United States. Journal of SocialPsychology, 114(2), Korte, C., & Kerr, N. (1975). Response to altruistic opportunities in
289-290. urban and nonurban settings. Journal of Social Psychology, 95(2),
Amato, P. R. (1983a). The effects of urbanization on interpersonal be- 183-184.
havior: Field studies in Papua, New Guinea. Journal of Cross-Cul- Korte, C., Ypma, I., & Toppen, A. (1975). Helpfulness in Dutch society
tural Psychology,, 14(3), 353-367. as a function of urbanization and environmental input level. Journal
Amato, P. R. (1983b). The helpfulness of urbanites and small town of Personality and Social Psychology, 32(6), 996-1003.
dwellers: A test between two broad theoretical positions. Australian Krupat, E., & Coury, M. (1975). The lost letter technique and helping:
Journal of Psychology, 35(2), 233-243. An urban-nonurban comparison. Paper presented at the 83rd annual
Amato, P. R~ (1983c). Helping behavior in urban and rural environ- convention of the American Psychological Association, Chicago, IL.
ments: Field studies based on a taxonomic organization of helping MeKenna, R. H. (1976). Good samaritanism in rural and urban set-
episodes. Journal of Personality and Social Psychology,, 45(3), 571- tings: A nonreactive comparison of helping behavior of clergy and
586. control subjects. Representative Research in Social Psychology, 7(1),
Bar-Tal, D., Raviv, A., & Shavit, N. (1981). Motives for helping behav- 58-65.
ior: Kibbutz and city children in kindergarten and school. Develop- Merrens, M. R. (1973). Nonemergency helping behavior in various
mental Psycholog~, 17(6), 766-772. sized communities. Journal of Social Psychology: 90(2), 327-338.
Bizman, A., Yinon, Y., Mivtzari, E., Shavit, R., Bar-llan, U., & Ramat- Nadler, A., Romek, E., & Shapira-Friedman, A. (1979). Giving in the
Gan, I. (1978). Effects of the age structure of the kindergarten and kibbutz: Pro-social behavior of city and kibbutz children as affected
altruistic behavior. Journal of School Psychology,, 16(2), 154- t 60. by social responsibilityand social pressure. Journal of Cross-Cultural
Darley, J. M., & Latant, B. (1968). Bystander intervention in emergen- Psychology, 10(1), 57-72.
cies: Diffusion of responsibility. Journal of Personality and Social Rotton, J. (1977). Sex, residential location, and altruism. Psychological
Psychology,, 8, 377-383. Reports, 40(1), 102.
Feinman, S. (1978). When does sex affect altruistic response? Psycho- Rushton, J. (1978). Urban density and altruism: Helping strangers in a
logical Reports, 43(3, Pt. 2), 1218. Canadian city, suburb, and small town. Psychological Reports, 43(3,
Forbes, G. B., & Gromoll, H. E (1971). The lost letter technique as a Pt. 1), 987-990.
measure of social variables: Some exploratory findings.Social Forces, Schneider, E W, & Mockus, Z. (1974). Failure to find a rural-urban
50, 113-115. difference in incidence of altruistic behavior. Psychological Reports,
Forbes, G. B., Tevault, R. K., & Gromoll, H. E (1978). Honesty as a 35(1, Pt. 1), 294.
function of geographic region and city size: Explorations of a thirty- Solomon, L. Z., Solomon, H., & Maiorca, J. (1982). The effects of by-
year old hypothesis. PsychologicalReports, 42, 647-652. stander's anonymity, situational ambiquity, and victim's status on
Franklin, B. J. (1974). Victim characteristics and helping behavior in a helping. Journal of Social Psychology, 117(2), 285-293.
rural southern setting. Journal of Social Psychology,, 93( 1), 93-100. Stem, T. D. (1974). A study of the relationships of helping behavior with
Gelfand, D. M., Hartmann, D. D., Walker, P., & Page, B. (1973). Who urbanism, sex, and socioeconomic status in two parts of the United
reports shoplifting? A field-experimentalstudy. JournalofPersonality States. Dissertation Abstracts International 34( 1 I-B), 5564.
and Social Psychology, 25, 276-285. Takooshian, H., Haber, S., & Lucido, D. J. (1977, October). Who
Goldman, M., Lewsandowski, H. E., & Carrill, R. E. (1982). Altruistic wouldn't help a lost child? You, maybe. Psychology Today, pp. 67-
behavior in rural and urban, residential and business areas. Basic and 68, 88.
Applied Social Psychology,, 3(2), 155-160. Weiner, E H. (1976). Altruism, ambiance, and action: The effects of
Hansson, R, O., & Slade, K. M. (1977). Altruism toward a deviant in rural and urban rearing on helping behavior. Journal of Personality
city and small town. Journal ofApplied Social Psychology,, 7(3), 272- and Social PsychologT,, 34( 1), 112-124.
279. Whitehead, G. I., & Metzger, S. C. (1981). Helping behavior in urban
Hansson, R. O., Slade, K. M., & Slade, P. S. (1978). Urban-rural differ- and nonurban settings. Journal of Social Psychology, 114(2), 295-
ences in responsiveness to an altruistic model. Journal of Social Psy- 296.
chology, 105(1), 99-105. Yinon, Y., Sharon, I., Azgad, Z., Barshir, I., Bar-llan, U., Ramat-Gan, I.
Holahan, C. J. (1977). Effects of urban size and heterogeneity on judged ( 1981). Helping behavior of urbanites, Moshavniks, and Kibbutniks.
appropriateness of altruistic responses: Situational vs. subject vari- Journal of Social Psycholog?g, 113( 1), t 43-144.
ables. Sociomet~ 40(4), 378-382.
House, J. S., & Wolf, S. (1978). Effects of urban residence on interper-
sonal trust and helping behavior. Journal of Personality and Social Received March 17, 1986
Psychology,, 36(9), 1029-1043. Revision received February 6, 1987
Korte, C. (1970). Effects of individual responsibility and group commu- Accepted February 19, 1987 9

You might also like