You are on page 1of 9

Review of General Psychology Copyright 2004 by the Educational Publishing Foundation

2004, Vol. 8, No. 1, 59 – 67 1089-2680/04/$12.00 DOI: 10.1037/1089-2680.8.1.59

Psychology’s Status as a Scientific Discipline: Its Empirical


Placement Within an Implicit Hierarchy of the Sciences
Dean Keith Simonton
University of California, Davis

Psychology’s standing within a hypothesized hierarchy of the sciences was assessed in


a 2-part analysis. First, an internally consistent composite measure was constructed
from 7 primary indicators of scientific status (theories-to-laws ratio, consultation rate,
obsolescence rate, graph prominence, early impact rate, peer evaluation consensus, and
citation concentration). Second, this composite measure was validated through 5
secondary indicators (lecture disfluency, citation immediacy, anticipation frequency,
age at receipt of Nobel Prize, and rated disciplinary hardness). Analyses showed that
the measures reflected a single dimension on which 5 disciplines could be reliably
ranked in the following order: physics, chemistry, biology, psychology, and sociology.
Significantly, psychology placed much closer to biology than to sociology, forming a
pair of life sciences clearly separated from the other sciences.

From time to time psychologists attempt to psychology displays the kind of disciplinary
take stock of how the discipline of psychology consensus and conceptual precision expected of
rates as a bona fide science. Sometimes this highly paradigmatic fields (Schachter, Christen-
self-evaluation takes the form of a qualitative feld, Ravina, & Bilous, 1991; Suls & Fletcher,
survey of its theoretical concepts and empirical 1983).
achievements. A case in point is Koch’s (1959 – These diverse self-appraisals might be sub-
1963) monumental Psychology: A Study of a sumed under a tradition dating back to Auguste
Science. On other occasions the self-assessment Comte (1842/1855), the French positivist phi-
assumes a more quantitative form. For instance, losopher who argued that the sciences could be
Rosenthal (1990) used the binomial effect size arrayed into a hierarchy. Comte’s own hierar-
display to address the question “How are we chy was based on the extent to which each
doing in soft psychology?” and Hedges (1987) discipline’s empirical findings were abstract,
applied meta-analytic techniques to respond to simple, and independent of the findings of other
the similar query “How hard is hard science, disciplines. Using these three criteria, he placed
how soft is soft science?” In a different vein are the principal sciences of his day in the following
various quantitative investigations that attempt order: astronomy, physics, chemistry, and phys-
to assess the extent to which psychology’s his- iology. Of course, psychology was not yet an
torical development follows a pattern similar to established science in the early 19th century.
that outlined in Kuhn’s (1970) classic account Even so, it is clear from Comte’s criteria that it
The Structure of Scientific Revolutions (Friman, would rank after physiology, especially insofar
Allen, Kerwin, & Larzelere, 1993; Robins, Gos- as early psychology emerged out of physiology.
ling, & Craik, 1999; see also Thagard, 1992). In Indeed, Wilhelm Wundt, the “father of psychol-
this latter group of inquiries can also be placed ogy,” originally distinguished the new disci-
those studies that attempt to determine whether pline from philosophical psychology by label-
ing it physiological psychology.
Although Comte’s disciplinary rankings were
more logical than empirical, Cole (1983) at-
I thank Laurence D. Smith for bringing to my attention tempted to provide a scientific test of the hy-
additional criteria for ranking scientific disciplines and for pothesized hierarchy. To do so, he developed a
providing me with some unpublished data. new set of criteria that reflected modern devel-
Correspondence concerning this article should be ad-
dressed to Dean Keith Simonton, Department of Psychol- opments in the sociology of science as well as
ogy, One Shields Avenue, University of California, Davis, Kuhn’s (1970) distinction between paradig-
CA 95616-8686. E-mail: dksimonton@ucdavis.edu matic and nonparadigmatic sciences. The crite-
59
60 SIMONTON

ria concerned the extent to which a discipline criterion based on empirical analyses of how
(a) has well-developed or highly “codified” the- scientific disciplines differ in their presentation
ories; (b) quantifies its ideas in a precise, math- of empirical findings (Cleveland, 1984; Smith,
ematical language; (c) exhibits high levels of Best, Stubbs, Archibald, & Roberson-Nay, in
consensus among its practitioners with respect press). Second, I investigated disciplines be-
to theory, methodology, and substantive ques- yond physics, chemistry, psychology, and soci-
tions; (d) displays high rates of obsolescence in ology insofar as it was permitted by the mea-
which recent work quickly replaces old work; sures. This greater inclusiveness allowed deter-
and (e) accumulates knowledge at a very rapid mination of the scope of the hypothesized
pace. Unable to provide quantitative measures hierarchy. In addition, by expanding the range
of all of these criteria, Cole confined his anal- of indicators and disciplines, it should be pos-
ysis to just two criteria, namely magnitude of sible to go beyond a mere ranking of the disci-
consensus and the speed with which a discipline plines. Specifically, it should be possible to
incorporates original results. On the basis of this obtain an interval score that reflects the distance
more limited analysis, Cole concluded that that separates each discipline along the under-
“there are no systematic differences between the lying dimension of implicit scientific status. In
sciences at the top and at the bottom of the other words, the analysis can generate a scale
hierarchy in either cognitive consensus or the rather than a mere ranking.
rate at which new ideas are incorporated”
(1983, p. 111).
Method
Nonetheless, the results of a more recent in-
vestigation conducted by Simonton (2002) cast The unit of analysis in this investigation was the
doubt on this conclusion. Simonton’s inquiry discipline rather than the individual scientist or
departed from Cole’s (1983) study in two sig- scholar. Depending on the disciplinary criteria exam-
nificant ways. First, Cole’s measures were com- ined, the number of disciplines could range be-
bined with additional indicators that represented tween 4 and 13. The measures of the attributes of
omitted criteria that had a reasonable associa- these several disciplines were divided into two cate-
tion with a discipline’s scientific status, namely gories: primary and secondary.
the theories-to-laws ratio (Roeckelein, 1997)
and rate of peer consultation (Suls & Fletcher, Primary Measures
1983). Second, Simonton determined the degree
of agreement among the several measures using The first set of indicators satisfies three specifica-
a correlational analysis, something that Cole tions. First, the measures must have a strong theoret-
failed to do. Simonton demonstrated that the ical or empirical connection with the supposed sci-
diverse criteria all exhibited a substantial entific status of a scientific discipline. After all, dis-
amount of agreement on the relative standing of ciplines can differ on a host of attributes that may
the four disciplines examined. Just as signifi- have nothing directly to do with a discipline’s place
in the hypothesized hierarchy (e.g., number of under-
cant, this summary indicator ranked the four graduate majors or expected income of doctorates).
disciplines in a manner completely consistent Hence, all of the indicators must have been explicitly
with the hypothesized hierarchy. Specifically, designed to assess the comparative scientific status of
physics was at the apex, followed by chemistry various disciplines. Second, the indicators must in-
and then psychology, with sociology at the clude assessments on the four disciplines in Simon-
bottom. ton’s (2002) study, namely physics, chemistry, psy-
In this article, I wish to replicate and extend chology, and sociology. In addition to the obvious
my earlier (Simonton, 2002) results in two di- fact that this quartet includes psychology, these two
rections. First, I incorporated additional mea- natural sciences and two social sciences have re-
sures that purport to gauge a science’s scientific ceived the most attention in prior research (e.g., Cole,
1983). The primary measures may apply to other
status, namely indicators of knowledge obsoles- disciplines as well, but these four could not have any
cence rate (McDowell, 1982) and graphic rep- missing values. Third, all indicators must be objec-
resentation (Cleveland, 1984; Smith, Best, tive rather than subjective. That is, they all entailed
Stubbs, Johnston, & Archibald, 2000). The straightforward counts or tabulations, thus avoiding
former is a standard listed but unmeasured by subjective judgments that might be biased by precon-
Cole (1983), whereas the latter represents a ceived notions regarding the relative merits of the
PSYCHOLOGY AS A SCIENCE 61

disciplines studied. The following seven indicators chemistry, biology, medicine, psychology, eco-
met these three standards. nomics, and sociology. Although Cleveland
(1984) did not aggregate the findings for the
1. Theories-to-laws ratio: Roeckelein (1997, Ta- disciplines, this aggregation was carried out by
ble 2, p. 137) assessed the number of theories Smith et al. (2000). This graph measure is also
and the number of laws mentioned in introduc- new to the current investigation (cf. Simonton,
tory textbooks in physics, chemistry, psychol- 2002).
ogy, anthropology, and sociology. These counts
were then used to compute the ratio of theories 5. Early impact rate: Cole’s (1983) Table 2 pro-
to laws; the higher the ratio, the more “soft” the vided the “proportion of scientists under 35
discipline. That is, exact sciences have many whose work received more than the mean num-
more laws in proportion to mere theories. In a ber of citations for their field” (p. 118). Those
sense, scientific status is a function of the ratio fields that incorporate most quickly the work of
of precise facts to vague conjectures. young scientists are assumed to rank higher in
the hierarchy, because such disciplines have a
2. Consultation rate: The next criterion was a stronger consensus about what can be regarded
consultation measure based on Festinger’s so- a significant contribution to the field. The dis-
cial comparison theory (Suls & Fletcher, 1983, ciplines covered were mathematics, physics,
Table 1, p. 578). According to this theory, when chemistry, geology, psychology, and sociology.
people are uncertain about their beliefs or per-
formance, they are more likely to engage in 6. Peer evaluation consensus: Cole’s (1983) Ta-
social comparison with similar others. The spe- ble 3 provided data indicating the “consensus
cific measure was the number of colleagues on evaluating scientists by field” (p. 120); 60
recognized in the acknowledgment section ad- scientists per field were rated by colleagues in
justed for the number of authors. In other the same discipline. The consensus was gauged
words, the measure is independent of the num- by the mean standard deviation of the ratings;
ber of collaborators. The higher this number, the lower the standard deviation, the higher the
the greater the apparent uncertainty about the consensus. The disciplines in this case were
quality of one’s work. This score was available physics, chemistry, biochemistry, psychology,
for physics, chemistry, psychology, and and sociology.1
sociology.
7. Citation concentration: The “concentration of
3. Obsolescence rate: On the basis of the relative citations to research articles” was presented in
frequency of citations to older publications, Cole’s (1983, p. 122) Table 5. The citations
McDowell (1982) determined the rate at which were to journals in mathematics, physics,
knowledge becomes obsolete for the disciplines chemistry, biochemistry, geology, psychology,
of physics, chemistry, biology, sociology, psy- and sociology. If the citations are all concen-
chology, history, and English. The specific trated in a single article, the disciplinary con-
measure used here was his calculation of the sensus must be very high, scientists concurring
expected publication cost of interrupting a ca- on what contributions deserve the status of “ci-
reer for just 1 year (McDowell, 1982, Table 2, tation classics.” In contrast, if the citations are
p. 757). For example, if their career is inter- more evenly distributed across articles, then the
rupted for a single year (e.g., by administrative consensus must be minimal. In the case of a
work or parental or health leave), the output of completely even distribution, the citations re-
physicists will be cut by about 17%, whereas ceived by articles would not differ from chance
the productivity of psychologists will be cut by expectations.
about 10% (because physicists will have much
Because the seven variables just described were
more “catching up on the literature” to do be-
measured on rather different scales, the raw scores
fore they can resuscitate their careers). This
were standardized to z scores (M ⫽ 0, SD ⫽ 1). In
measure was not used in Simonton’s (2002)
addition, those variables that were reverse indica-
investigation.
tors—namely theories-to-laws ratio, consultation
4. Graph prominence: Cleveland (1984) assessed
the extent to which graphs appear in articles 1
Simonton’s (2002) inquiry also measured peer evalua-
published in the professional journals, demon- tion consensus using the results in Cole’s (1983, p. 120)
strating that graphs are more extensively used study. However, that alternative measure was ignored here
in the “hard” disciplines (see also Smith et al., because (a) it was redundant with the first and (b) it ap-
in press). The specific disciplines were physics, peared to be less reliable according to a reliability analysis.
62 SIMONTON

rate, and peer evaluation consensus—were inverted 4. Age at receipt of Nobel Prize: Stephan and
by reversing the sign of the standardized scores. Levin (1993, Table 1, p. 395) provided the
median age at which scientists received Nobel
Prizes in the fields of chemistry, physics, and
Secondary Measures medicine (from 1901 to 1992). The information
provided at the official Nobel Prize Web site
The second set of measures all have one feature in (http://www.nobel.se) was used to obtain the
common: However many disciplines to which they same statistic for the recipients of the econom-
are applied, they have a missing value for at least one ics prize (from 1969 to 2001). The logic behind
of the disciplines included in Simonton’s (2002) including this indicator is the same as the early
study. As a consequence, none of these measures impact rate measure among the primary predic-
were used in the earlier investigation. In addition, tors. The more codified or paradigmatic a dis-
even though all are relevant to a discipline’s scientific cipline, the sooner it can recognize when a
status, not all of them are completely objective. Nev- scientist has made an exceptional contribution
ertheless, these secondary measures are useful for to the field.
validating the results obtained from the primary mea-
sures. There were five indicators in this group. 5. Rated disciplinary hardness: Smith et al.
(2000) had psychologists rate disciplines on the
1. Lecture disfluency: Schachter et al. (1991) de- degree to which they could be considered
termined the rate of filled pauses (“uh,” “er,” “hard” versus “soft.” The respondents used a
and “um”) during classroom lectures in under- 10-point Likert scale, with 10 indicating the
graduate courses in mathematics, chemistry, bi- highest degree of hardness. Seven disciplines
ology, psychology, economics, sociology, po- were so rated, namely physics, chemistry, biol-
litical science, philosophy, art history, and En- ogy, medicine, psychology, economics, and so-
glish. The higher the number of pause words ciology. Smith et al. (2000) showed that this
per minute, the greater the degree of speech subjective assessment correlated .97 with
disfluency, which presumably reflects the de- Cleveland’s (1984) measure of graph use. In
gree to which a discipline is less formal, struc- addition, the investigators showed that the as-
tured, and factual. This interpretation is bol- sessment correlated .94 with an independent
stered by the fact that the same set of lecturers measure of paradigm development in various
did not differ in disfluency when speaking on a disciplines (Ashar & Shapiro, 1990). This mea-
common subject. Hence, it is not a matter of the sure, although subjective, was included to de-
more inarticulate scientists being attracted to termine whether the objective assessments con-
the less rigorous disciplines. cur with more intuitive attitudes about the rel-
ative status of different scientific disciplines.
2. Citation immediacy: Cole (1983, Table 8, p.
126) calculated the extent to which the refer- As before, the preceding measures were all stan-
ences in published articles were confined to dardized to a mean of 0 and a standard deviation of 1.
recent work. In other words, this calculation Moreover, lecture disfluency was inverted by multi-
gauged whether citations emphasize contempo- plying by ⫺1. It should be noted that, unlike the case
rary research over classic studies. Scores on this of the primary indicators, there was sometimes little
immediacy factor were available for physics, overlap in the disciplines covered by these five indi-
chemistry, biochemistry, geology, and psychol- cators. In fact, some measures had as few as two
ogy (but not sociology, and hence its omission disciplines in common, a figure too small to compute
from the Simonton, 2002, inquiry). meaningful correlations. Even so, these measures
will prove useful in validating the primary indicators.
3. Anticipation frequency: Hagstrom (1974, Ta-
ble 1, p. 3) reported the results of a survey
of 1,718 scientists asked to report whether their Results
work had been anticipated by other scientists.
The percentage of scientists who had this expe- Generating the Composite Measure From
rience at least once during their career course
was gauged for mathematics, physics (combin- the Primary Indicators
ing theoretical and experimental), chemistry,
and biology (combining experimental and The four disciplines of physics, chemistry,
other). The greater the frequency of anticipa- psychology, and sociology were used to cali-
tion, the higher the consensus on what are brate an analytical baseline for devising a more
deemed the important and unimportant prob- comprehensive measure that would apply to a
lems in a discipline. wider range of disciplines. The first step was to
PSYCHOLOGY AS A SCIENCE 63

calculate the correlations among the seven mea- ences of anticipation, greater youthfulness of
sures for just these 4 sciences, the resulting Nobel Prize recipients, and higher rated disci-
correlations ranging between .63 and .998. plinary hardness.
These correlations were then subjected to a The corresponding Pearson product–moment
principal components analysis.2 Only one com- correlation coefficients ranged between .60 and
ponent had an eigenvalue exceeding unity, and .97, with a median of .88. Unfortunately, the
that lone component accounted for 86% of the number of cases underlying each correlation
total variance. Moreover, the loadings on the was often too small to permit statistical signif-
first component were uniformly high, ranging icance (however, p values were .0003 for the
from .86 to .99. The specific loadings were as subjective hardness rating and .0786 for the age
follows: theories-to-laws ratio, .99; consultation at receipt of Nobel Prize measure).3 Even so, it
rate, .99; graph prominence, .96; peer evalua- was possible to consolidate the five secondary
tion consensus, .93; early impact rate, .88; cita- indicators into a single composite in the same
tion concentration, .87; and obsolescence rate, manner as the primary composite. That is, a new
.86. As a consequence, the standardized scores score was created for each discipline by aver-
across all seven measures were averaged to aging across the indicators for which there were
produce a linear composite. The internal con- nonmissing scores. This secondary composite
sistency reliability (Cronbach’s alpha) for this correlated .87 (n ⫽ 11, p ⫽ .0004) with the
composite was .96. primary composite, thereby confirming statisti-
The next step was to extend this linear com- cally what is so apparent graphically in Fig-
posite to all disciplines that contained at least ure 1. The 12 primary and secondary indicators
one nonmissing value on the seven primary reflect a coherent latent variable on which dis-
indicators. This was accomplished by simply ciplines can be reliably differentiated.
averaging the standardized scores across all in-
dicators with nonmissing values for a given
discipline. This means that a discipline’s score
Ranking Five Disciplines Using the
on the linear composite may represent anywhere Composite Measure
between one and seven scores. Of course, the Because 5 disciplines had nonmissing values
expected measurement error will be greater for on at least three of the primary indicators, it was
those disciplines that have more missing values. possible to provide fairly reliable rankings for
The ratings based on a single component crite- this subset of the 13 studied. Figure 2 shows the
rion would be the least reliable. In any case, the outcome. Physics, chemistry, biology, psychol-
resulting composite measure was restandard- ogy, and sociology are arrayed according to
ized to a zero mean and unit standard deviation. standardized composite score and rank. The
composite scores for physics, chemistry, psy-
Validating the Composite Measure Using chology, and biology are based on all seven
Secondary Indicators
As a means of validating the resulting com- 2
Because the number of variables exceeds the number of
posite indicator, the scores on the secondary cases, it is impossible to extract more than three principal
measures were plotted as a function of scores on components (i.e., the number of nonzero positive eigenval-
ues cannot exceed N ⫺ 1). Nonetheless, the results still
the composite measure. The outcome is shown amply surpass what must be expected under a null (random)
in Figure 1, which also gives the lines of best model in which the eigenvalues for the first three compo-
least squares fit. If no association existed, these nents would all be equal. If there were two underlying
lines would be horizontal, whereas lines in- dimensions rather than one, the first two components would
have roughly equal eigenvalues. Contrary to either of these
clined by about 45° would represent perfect scenarios, the first eigenvalue is more than 8 times the size
positive relations. Clearly, a strong association of the second and 21 times the size of the third.
exists between the composite of the primary 3
One can readily argue that inferential statistics are not
indicators and each of the separate secondary appropriate anyway. The disciplines examined in this study
indicators. In particular, higher scores on the cannot be considered a random sample from a larger pop-
ulation of domains, nor is there any larger population of
composite are associated with lower lecture dis- disciplines to which the results are to be generalized. Ac-
fluency, higher concentration of citations on cordingly, the descriptive statistics have interpretative value
more recent literature, more frequent experi- apart from any desire to make population inferences.
64 SIMONTON

Figure 1. Scatterplot showing the relation between a discipline’s score on the composite measure and indicators of a
discipline’s lecture disfluency, citation immediacy, anticipation frequency, age at receipt of Nobel Prize, and rated
disciplinary hardness. The composite measure is defined by the discipline’s theories-to-laws ratio, consultation rate,
obsolescence rate, graph prominence, early impact rate, peer evaluation consensus, and citation concentration. Regression
lines of least squares best fit are also shown.

primary indicators and thus have an internal deviation separates the psychological from the
consistency reliability of .96. Because the stan- biological sciences, whereas nearly one stan-
dard error of measurement is equal to the square dard deviation separates psychology from soci-
root of 1 minus the reliability coefficient, the ology. Indeed, the biggest gap in the scores is
error for these four disciplines is only 0.2, or found between the latter two disciplines. Fur-
just one fifth of a standard deviation. The com- thermore, because a somewhat smaller gap also
posite score for biology, in contrast, was based separates biology from chemistry, it might be
on only three indicators, with a reliability of .89. more accurate to conclude that the disciplines
Even so, the standard error of measurement is fall into three clusters: the physical sciences of
still reasonably small, namely 0.3, or about a physics and chemistry at the top of the hierar-
third of a standard deviation. Consequently, the chy, the life sciences of biology and psychology
ordinal placement of these 5 disciplines is rea-
in the middle, and the social science of sociol-
sonably secure.
It is immediately apparent that the disciplines ogy at the bottom.4
are ordered in close conformity to the expected
hierarchy. In particular, the natural sciences
score higher than the social sciences. Psychol-
ogy itself falls almost exactly at the mean of the 4
Although the composite scores are based on fewer than
distribution, a placement that betrays its loca- three criteria, it is worthwhile to point out that the human-
tion at the junction of the natural and social ities are placed far below sociology. For example, scholar-
ship in English is farther removed from sociology than
sciences. Yet, according to the composite sociology is from psychology. Thus, it is apparent that the
scores, psychology is situated closer to biology criteria by which the sciences are here judged also distin-
than to sociology. Less than half of a standard guish the sciences from the nonsciences.
PSYCHOLOGY AS A SCIENCE 65

and predictable fashion, with psychology clus-


tering near biology at approximately the middle
of the hierarchy.
However, to offer a fair appraisal of the mer-
its of this composite assessment, it is necessary
to deal with three issues. First, although the
analysis treated all disciplines as if they were
homogeneous domains, it is manifest that each
field consists of subdisciplines that differ sub-
stantially in their scientific status. In psychol-
ogy, for example, there is a marked contrast
between those subdisciplines that lean toward
the natural sciences and those that lean toward
the human sciences (Coan, 1968; Kimble, 1984;
Simonton, 2000). Moreover, empirical evidence
demonstrates that these subdisciplines differ in
how they score on the criteria (e.g., Best, Smith,
& Stubbs, 2001; Cole, 1983; Smith et al., in
press). Indeed, these subdisciplinary contrasts
determine the odds that a psychologist will be
elected to the National Academy of Sciences
(Over, 1981). Even so, the placement seen in
Figure 2 can still be viewed as representing each
discipline’s “center of gravity.” Some subdisci-
plines may score higher and others lower, but
Figure 2. The disciplines of physics, chemistry, biology, the overall composite score indicates the central
psychology, and sociology placed in a Comtean hierarchy of tendency of the field. In fact, this solution is
the sciences. The horizontal axis indicates the rank, and the implicit in the construction of the measures,
vertical axis indicates the composite score on the seven
primary indicators (i.e., the discipline’s theories-to-laws
some of which are derived from scores averaged
ratio, consultation rate, obsolescence rate, graph promi- across numerous journals that represent the cen-
nence, early impact rate, peer evaluation consensus, and tral subdisciplines of each field (e.g., graph
citation concentration). prominence).
Second, disciplines are not static entities. In-
stead, disciplines change dynamically over
Discussion time. Sometimes these changes may be progres-
sive, as a discipline becomes increasingly more
Completely objective and previously pub- scientific. Comte (1842/1855) argued for the
lished measures were used to construct a com- essential nature of scientific progress, and some
posite measure of a discipline’s scientific status. empirical studies have attempted to document
This measure was defined by the field’s theo- its occurrence, including within psychology
ries-to-laws ratio, consultation rate, graph (e.g., Bruner & Allport, 1940; Roeckelein,
prominence, peer evaluation consensus, early 1996). Nevertheless, the hierarchical arrange-
impact rate, citation concentration, and knowl- ment shown in Figure 2 can still be considered
edge obsolescence rate. Moreover, this measure a fair “snapshot” of the comparative status of
correlated with alternative indicators of a disci- various disciplines at a particular point in his-
pline’s placement in the hypothesized hierar- torical time. Furthermore, the hierarchy of the
chy, namely lecture disfluency in undergraduate sciences may be transhistorically invariant ow-
courses, the degree to which citations are to ing to properties intrinsic to each discipline. For
recent literature, the frequency that one’s work example, those sciences that deal with highly
is anticipated by others, age at receipt of Nobel complex and concrete phenomena in the social
Prize, and subjectively rated disciplinary hard- sciences may always be at a disadvantage vis-
ness. Finally, the composite measure could à-vis the natural sciences, or psychology rela-
place five disciplines along a scale in a reliable tive to biology. Moreover, as Comte himself
66 SIMONTON

argued, the scientific standing of some disci- rather than the methods used to study those
plines is inevitably tied to foundational disci- phenomena.
plines, thus precluding any major changes in the Nonetheless, even this substantive limitation
hierarchy. might be overcome to the extent that psychol-
Third and last, according to Cole (1983), ogy becomes more fully integrated with the
most scientific disciplines can be partitioned biological sciences. The three most obvious
into two disparate parts. On the one hand is the routes by which this integration might be ac-
disciplinary core that consists of “fully evalu- complished are the cognitive neurosciences, be-
ated and universally accepted ideas which serve havioral genetics, and evolutionary psychology.
as the starting points for graduate education” (p. Although there already exist movements in this
111). On the other hand is the research frontier, direction (e.g., Barkow, Cosmides, & Tooby,
which includes “all research currently being 1992), there are also ample trends pushing the
conducted” (p. 111) at the leading edge of the field in the opposite direction, such as the pos-
discipline. It is conceivable that disciplines itive psychology movement, which has closer
might have very different statuses in the sup- ties to humanistic psychology than to the natu-
posed hierarchy depending on whether one ex- ral sciences (e.g., Snyder & Lopez, 2002).
amines the core or the frontier of each domain. Hence, it is possible that in the future the dis-
However, the present results imply that this cipline will simply continue to oscillate around
distinction would not seriously undermine the the position in Figure 2. In particular, psychol-
rankings exhibited in Figure 2. After all, the ogy may always find itself placed just below
primary and secondary measures included items biology, forming a pair of life sciences clearly
gauging both disciplinary aspects. Thus, with separated from the physical sciences, repre-
respect to the core, the theories-to-laws ratio sented by physics and chemistry, and the social
was based on introductory textbooks and lecture sciences, represented by sociology.
disfluency was calculated from undergraduate
courses, whereas the frontier was represented
by such measures as early impact rate, obsoles- References
cence rate, consultation rate, and citation con-
centration. Hence, those disciplines that have Ashar, H., & Shapiro, J. Z. (1990). Are retrenchment
decisions rational? The role of information in
more scientific cores tend to have more scien- times of budgetary stress. Journal of Higher Edu-
tific frontiers as well. cation, 61, 123–141.
Although the rankings depicted in Figure 2 Barkow, J. H., Cosmides, L., & Tooby, J. (Eds.).
still enjoy a reasonable amount of credibility, (1992). The adapted mind: Evolutionary psychol-
the composite measure on which they were ogy and the generation of culture. New York:
based cannot possibly be considered the last Oxford University Press.
word. The most obvious next step is to extend Best, L. A., Smith, L. D., & Stubbs, D. A. (2001).
the analysis to even more scientific disciplines. Graph use in psychology and other sciences. Be-
In addition, it would be highly desirable to havioral Processes, 54, 155–165.
Bruner, J. S., & Allport, G. W. (1940). Fifty years of
increase the number of criteria by which a dis- change in American psychology. Psychological
cipline’s status is assessed. Perhaps the most Bulletin, 37, 757–776.
critical of these additions would be an index of Cleveland, W. S. (1984). Graphs in scientific publi-
the degree to which a domain has been subject cations. American Statistician, 38, 261–269.
to mathematical analysis. Yet, even with this Coan, R. W. (1968). Dimensions of psychological
addition it would be highly unlikely that psy- theory. American Psychologist, 23, 715–722.
chology would dramatically alter its current po- Cole, S. (1983). The hierarchy of the sciences? Amer-
sition in the resulting hierarchy. In particular, it ican Journal of Sociology, 89, 111–139.
is most probable that psychology’s status re- Comte, A. (1855). The positive philosophy of Au-
guste Comte (H. Martineau, Trans.). New York:
sides between the natural science of biology and Blanchard. (Original work published 1842)
the social science of sociology, with a greater Friman, P. C., Allen, K. D., Kerwin, M. L. E., &
proximity to the former. To some undetermined Larzelere, R. (1993). Changes in modern psychol-
degree, psychology’s position in the hierarchy ogy: A citation analysis of the Kuhnian displace-
reflects the nature of the phenomena studied ment thesis. American Psychologist, 48, 658 – 664.
PSYCHOLOGY AS A SCIENCE 67

Hagstrom, W. O. (1974). Competition in science. knowledge. Journal of Personality and Social Psy-
American Sociological Review, 39, 1–18. chology, 60, 362–367.
Hedges, L. V. (1987). How hard is hard science, how Simonton, D. K. (2000). Methodological and theo-
soft is soft science? American Psychologist, 42, retical orientation and the long-term disciplinary
443– 455. impact of 54 eminent psychologists. Review of
Kimble, G. A. (1984). Psychology’s two cultures. General Psychology, 4, 1–13.
American Psychologist, 39, 833– 839. Simonton, D. K. (2002). Great psychologists and
Koch, S. (Ed.). (1959 –1963). Psychology: A study of their times: Scientific insights into psychology’s
a science (6 vols.). New York: McGraw-Hill. history. Washington, DC: American Psychological
Kuhn, T. S. (1970). The structure of scientific revo- Association.
lutions (2nd ed.). Chicago: University of Chicago Smith, L. D., Best, L. A., Stubbs, D. A., Archibald,
Press. A. B., & Roberson-Nay, R. (in press). Construct-
McDowell, J. M. (1982). Obsolescence of knowledge ing knowledge: The role of graphs and tables in
and career publication profiles: Some evidence of hard and soft psychology. American Psychologist.
differences among fields in costs of interrupted Smith, L. D., Best, L. A., Stubbs, D. A., Johnston, J.,
careers. American Economic Review, 72, 752–768. & Archibald, A. B. (2000). Scientific graphs and
Over, R. (1981). Affiliations of psychologists elected the hierarchy of the sciences. Social Studies of
to the National Academy of Sciences. American Science, 30, 73–94.
Psychologist, 36, 744 –752. Snyder, C. R., & Lopez, S. J. (Eds.). (2002). The
Robins, R. W., Gosling, S. D., & Craik, K. H. (1999). handbook of positive psychology. New York: Ox-
An empirical analysis of trends in psychology. ford University Press.
American Psychologist, 54, 117–128. Stephan, P. E., & Levin, S. G. (1993). Age and the
Roeckelein, J. E. (1996). Citation of laws and theo- Nobel Prize revisited. Scientometrics, 28, 387–
ries in textbooks across 112 years of psychology. 399.
Psychological Reports, 79, 979 –998. Suls, J., & Fletcher, B. (1983). Social comparison in
Roeckelein, J. E. (1997). Psychology among the sci- the social and physical sciences: An archival study.
ences: Comparisons of numbers of theories and Journal of Personality and Social Psychology, 44,
laws cited in textbooks. Psychological Reports, 80, 575–580.
131–141. Thagard, P. (1992). Conceptual revolutions. Prince-
Rosenthal, R. (1990). How are we doing in soft ton, NJ: Princeton University Press.
psychology? American Psychologist, 45, 775–777.
Schachter, S., Christenfeld, N., Ravina, B., & Bilous, Received November 15, 2002
F. (1991). Speech disfluency and the structure of Accepted January 10, 2003 䡲

You might also like