Chapter 5

Discussion and Implications

Overview

This chapter discusses results and implications derived during collection of the data. These outcomes are discussed using Bradford’s Law of Scattering, Garfield’s Law of Concentration, Goffman’s Epidemic Theory of Communication, the Bandwagon Appeal in publications, and Lotka’s Law of Author Productivity. Some nonbibliometric discussion is also incorporated into the implications, such as the evolution of autism literature and the appearance of “countries” that no longer exist. This chapter concludes by suggesting avenues of further research to pursue and how to continue autism causation research investigations.

General Discussion of Results

The null hypothesis offered in the first chapter is that no one cause will dominate the literature, that is, appear statistically more frequently than any other possible autism etiologies. Given the overall percentages from Table 4.2, at 35.71 percent, the neurological causation leads the group, with psychological causations next at 23.72 percent of the sources and genetic causations with 19.71 percent. With more than a third for neurological, and nearly a quarter and one fifth for psychological and genetic, respectively, these etiologies are each more prevalent than the one sixth (or 16.67%) hypothesized. The allergy/immunological, biological, and environmental causations each have less than half of this expected outcome at 6.74 percent, 6.68 percent, and 7.39 percent. The neurological causations number 551 of the 1,543 etiologies found, a plurality but not a majority. However, this 551 is more than twice the 257.17 instances expected for an equal distribution of causation. The 366 psychological and 305 genetic etiologies found are not dramatically greater than the 257.17 expected, but they are when contrasted with the 104, 103, and 114 causations found for allergy/immunologic, biological, and environmental causations respectively. More detailed discussion of this appears in the “Implications for Hypothesis” subsection.

Bradford’s Law of Scattering

Three Zones

Lancaster and Warner provide a compact explanation of Bradford’s Law in the midst of their explanation of how to determine the cost-effectiveness of databases. They state that a comprehensive literature search over a specific period will reveal that literature on a specific subject is scattered over a large number of journals. Step one is to arrange these journals by most productive to least, then divide this list into three zones, each with an equal number of journal articles. Although the number of articles in each zone will match, the number of journals yielding these articles will vary greatly. A few journals will have a large majority, while many journals will be represented only once or twice. The utility of these three zones was never pursued or even explained very well in Bradford’s work. However, Garfield’s “Law of Concentration” finds an application for the three zones in using Bradford’s Law specifically for scientific journals. All of the journals on a subject can be divided into multidisciplinary journals, more specialized journals, then “highly specialized” journals. The Bradford distribution for the autism causation data collected is shown in Figures 3 and 4, Appendix G.

The difference between the two figures will be explained fully in the next section, but concerns the differences that result when combining journal name changes rather than counting each journal name distinctly. For this data each of the three zones (indicated by the vertical lines in Figures 3 and 4) has approximately 404 of the 1,211 articles, letters, and reports that came from journals. Interestingly, the number in each zone approximates the constant discussed in the next section. (There is no reason this has to be true, but it is so only because of the current totals in the autism causation population. In ten more years the number of items in each zone could well be 500 or 600, but the constant may still be around 400.) In Figure 3, there are 15 highly specialized journals that account for the 404 articles in zone three. For Figure 4, this number is 13. The trivial many (zone one) in Figure 3 come from 300 journals: 298 journals for Figure 4.

The Formula

Bradford first published this “Law of Scattering” in 1934 but did not provide a mathematical statement of how and why it works or determine the minimal nucleus for which the law holds. He left this work to later scholars. In 1948, Vickery showed that this law “implies a J-shaped curve with an initial concave segment leading to a large linear segment and that such a curve gives a good fit to available data.” That Bradford’s Law is constant for a wide range of bibliographic situations supports its utility as a predictive tool. Especially for the linear middle portion of the chart, a constant can be derived using the formula:

R(n) = k x log(n)

Where R is the total number of articles and n is any cumulative number of articles found in the set, one can solve for the constant (k) by multiplying one side of the equation by the logarithm of n and dividing this by the product of R(n) that is on the other side of the equation.

Doing so at two points in each of the three zones in Figure 3 yields the following results (also see Table G-1, Appendix G). Where 223 is the cumulative number of articles (all of the journals publishing a single article), the k value is 515.76. Then at data point 377 the “constant” becomes 469.38. In the middle zone, where 526 is the cumulative number of journals k=445.22 and where 777 is the journal total, k=419.03. At the high end of the cumulative numbers, data point 892 yields a k value of 410.51 and at 1,108, the second to last data point, k=398.35.

For Figure 3, instead of remaining constant, the k value in the autism causation Bradford scatter drops as the number of journals increases. This progression is indicated by the slight curvature of the data line, even for the middle of the plot, which would be a straight line under ideal conditions. If the sample size were larger or otherwise less specific and limited, the expected constant might have been closer to a true constant. For example, if the researcher had used all autism literature (or possibly even all 2,086 of the DIALOG abstracts), the k values may have had a tighter grouping than the resulting numbers between 393.18 (the k value at 1,211 articles) and 515.76 (the k value at 223). (See Table G-1 for all k values.)

However, if we first combine the numbers of journals that have undergone name changes, the constant is only a little closer to a true constant (see Table G-2) and the shape of the curve becomes more like the elongated “S” expected of a Bradford scatter when the three regions include the “significant few,” the “middle class,” and the “trivial many.” While Figure 3 favors the J-shaped curve mentioned by Vickery, Figure 4 more closely matches the S-shape, with a more noticeable “Gross Droop” (or convex curvature) at upper right, the area that depicts the most abstracted journals. The “Bradford Restriction” (the concave curvature at the lower left of each graph) is very prominent since the field of autism causation is heavily represented by one-time author contributions. The “significant few” (see the upper right of both Figures 3 and 4) change most dramatically from Figure 3 to Figure 4 since the second most abstracted journal on the subject—the Journal of Autism and Childhood Schizophrenia—is the 1971 through 1978 name of the most cited, the Journal of Autism and Developmental Disorders (with 52 and 103 articles used, respectively). The other pair of journals that significantly impacts the change in the Bradford distribution is the combination of the Journal of the American Academy of Child and Adolescent Psychiatry (with 19 selections), and its predecessor, the Journal of the American Academy of Child Psychiatry (with 16 articles used). Three other name changes are reflected in Figure 4. Current Opinion in Pediatrics started out as Current Problems in Pediatrics in the 1980s. These two combine for 8 articles. The Japanese Journal of Child Psychiatry also added -and Adolescent after -Child (in 1981, approximately). However, this changes the combined total only from four to five. Similarly, the Psychiatric Clinics of North America became the Child and Adolescent Psychiatric Clinics of North America by 1995. The recent name change, however, adds only one to the combined total (nine instead of eight).

Solving for the constant by using two data points in each of the three zones of Figure 4 yields the following results (also see Table G-2, Appendix G). Where 221 is the cumulative number of articles (all of the journals publishing a single article), the k value is 517.52. At data point 375, the “constant” becomes 471.21. In the middle zone, where 517 is the cumulative number of journals k=446.86 and where 777 is the journal total,. k=419.03. At the high end of the cumulative numbers, data point 842 yields a k value of 413.31 and at 1,056, the second to last data point, k=400.99. Not only is the spread of what should be a constant somewhat wider in Figure 4 than it is in Figure 3 (starting at 517.52 versus 515.76), again instead of remaining constant the k value in the autism causation Bradford scatter drops as the number of journals increases. The only place the two graphs approach a constant is where the cumulative journal totals are in the 800s and the constant differs by less than two at each data point. As mentioned, though, this is a specialized application of Bradford’s Law since we are discussing only autism causation. A truer test would apply Bradford’s Law to all autism literature.

Although Wallace and Bonzi found a correlation between journal productivity (as measured by Bradford’s Law) and the frequency with which these journals were cited in later works, Wallace adds an important caveat to the Bradford scatter in stating “this approach requires acceptance of the unproven assumption that frequency of citation can be used as a measure of quality.” Donohue attempts to explain this dichotomy by stating that Bradford’s is a statistical—rather than social—law. That a certain journal carries the most articles on autism causation literature does not necessarily make it the voice of authority on the topic. It may be simply the most accessible, or in the case of the Journal of Autism and Developmental Disorders, it may be the only journal specific to the topic. Bradford’s Law of Scattering is intended to measure only quantity, not quality.

Goffman’s Theory

Goffman attempted to get to this next step and incorporate quality into his “epidemic theory of communication.” His theory is that ideas spread like diseases; they are transmitted mainly by direct communication and spread among a susceptible population. He adds that the journal population is the controlling factor; that is, the number of authors is directly related to the number of journals that publish on a subject. Also, the number of “quality” authors is proportional to the total number of authors in the field. However, the importance of Goffman’s theory lies in its “predictive power.” Just as statistics are predicted for epidemics, as a research topic emerges, Goffman’s mathematical analysis may be applied to predict controlling conditions, dimensions of growth, and the rate of slow down. The other value of Goffman’s work is that it points out the “infectious nature of ideas.”

It may be, however, that Goffman’s theory is simply a sophisticated restatement of the “bandwagon appeal” which Price applies to author productivity in scientific writings. He noticed that once a person has successfully published a paper, the likelihood of him or her publishing a second is .33; the likelihood of a third is .50; for a fourth paper the probability goes up to .60 (and so on). Simply put, it is more likely that prolific authors will continue to write and publish than that those who rarely do so will keep at it.

Bandwagon Appeal

Nicholas and Ritchie note that bibliographic history, especially of journals, can divulge much about shifts in the direction of the subject literature. Changes in a journal’s title, or more subtle changes in the terminology it uses, reflect changes in the subject covered. For that matter, they indicate that the subject itself is the most important bibliographic characteristic and the most troublesome. While subject-orientation of bibliometric studies is pre-eminent, that there are essentially two ways of thinking about subject literatures complicates matters. One literature is defined by custom, the other by use. “The boundaries of the first are dictated by conventional thought, educational syllabi and classification schemes. The boundaries of the second are much more fluid.” If the “subject” itself presents a bibliometric problem as a quantitative judgment, subject content is even more troublesome. Most bibliographic attributes are outward features of a work, but to discuss the subject content is to get at the document’s “inner self.” It is easier to discuss the container but much more interesting and useful to describe its contents.

Wallace, however, points out an even more fundamental problem with bibliometrics; it lacks a well-developed theoretical basis. He likens it to the “cumulative advantage theory” of Derek de Solla Price and explains Price’s theory in part by comparison to Saint Matthew’s “For to him who has will more be given, and he will have abundance; but from him who has not, even what he has will be taken away.”

While some evidence of the bandwagon appeal can be found in the growth of allergy, chromosomal, and brain research in recent years, this increase in percentages is more likely attributable to the improved technologies which had made such research impossible or prohibitively time consuming prior to the 1970s. Evidence of the bandwagon appeal in autism causation literature is further weakened by the fact that the growth in the other areas of research did not eliminate psychological etiology research, only slowed its growth (see Figure 2). Although technological advances in the other areas may have displaced psychological causations to a degree, even in the most recent decade of data used, psychological causations still account for 116 causations cited, well ahead of allergy-auto-immune (58), biological (52), and environmental (47) research.

Better evidence of the bandwagon appeal can be found not in the causes undergoing investigation, but in the emphasis on childhood and progression of opinion toward autism as a spectrum disorder. Rapin mentions many early deficit areas under research: sensory, cognitive, attention, social, behavioral, and language development. Hart adds that because autism shows hereditary as well as environmental factors, it is properly characterized as a syndrome. Even the DSM-IV describes autism as a “pervasive developmental disorder.”

As mentioned during the discussion of Bradford’s Law, at least five of the most-cited journals in autism etiology have undergone name changes, most to add “and adolescent” to the “child” already in their title. However, the 1978 name shift of the main autism journal from the Journal of Autism and Childhood Schizophrenia to the Journal of Autism and Developmental Disorders directly addresses what Nicholas and Ritchie mean by a name change indicating a shift in direction of the subject literature. In the late 1970s autism became separated from adult schizophrenia and linked with a failure to meet the developmental milestones expected during the toddler years. This shift in thought permeated all or most of the autism causations under investigation.

As for the causations themselves, there is not so much a concentrated adding to the etiology of the moment (characteristic of the bandwagon appeal) as the rapid growth of certain causations (like neurological and genetic) concomitant with the slower growth of other etiologies (as with biological and environmental). Although psychological causations have been declining as an annual percentage, the overall number of psychological etiology articles continues to have up and down years (even into the 1990s) rather than showing a true decline in numbers. The Figure 2 bar chart and Table C-1’s year by year summary of autism causations make these progressions obvious.

Epidemic Theory

Evidence of Goffman’s epidemic theory at work in autism causation research literature is more difficult to uncover than even the bandwagon appeal. The main evidence of it is that once a possible causation appears in the literature, it never goes away. This is especially surprising when considering that the causation studies are not limited to the six causal areas this researcher decided on. There are organic, cognitive, and gestational possibilities, as well as suspected associations with rubella, various viruses, and several other disabilities. All appear in different years, but none fade. Psychological etiologies appeared first, followed by neurological causations. Both continue to appear, along with all four of the other possibilities, including the latest addition to the argument, allergy/autoimmunological causations in 1965.

Even so, there are some hypotheses under investigation about the causation of autism that have been followed up by only a few later scholars. Many of these are narrowly focused studies within the broader psychological, environmental, and genetic fields already discussed. A study of the head circumference of children with autism compared with non-autistic peers is not as odd as it sounds initially. The theory that the brain, or at least certain regions of the brain, are physically larger that usual with autism started being studied post-mortem in the 1980s (mainly by Bauman and Kemper). However, some of the physical evidence for autism studies stretch the limits of scientific/medical research. Dermatoglyphics, the study of skin patterns, especially those of the hands and feet, has been useful in identification and genetic studies. Probably due to the genetic connection, dermatoglyphics has been applied to autism research. Six of the sources used specifically address this area, usually in connection with genetic or environmental causation research. Other genetic and environmental studies less grounded in science include one that investigates clumsiness as a genetic marker for Asperger Syndrome and another testing the prevalence of March birthdays among individuals with autism.

Research that is even more unusual comes from those investigating a psychological etiology for autism. Despite Coleman’s assertion in 1990 that “parental culpability. . .is no longer considered a viable hypothesis in an overwhelming majority of cases,” many countries and individuals continue to investigate psychological causations. As recently as 1994, Frances Tustin claimed autism resulted from a failure of the mother and child to bond. “Autism is a system of protective, but alienating autosensual aberrations . . . developed to deal with an infantile trauma.” As evident from the overall tally, Tustin is not alone. Such causes as a dysfunction of mother-infant metacommunication, nipple separation anxiety, and pre-Oedipal manifestations have been investigated. One study in 1958 had the mothers do Rorschach inkblot tests and concluded that “mothers of children with primary autism have less capacity for establishing social and emotional relationships than do mothers of children with secondary autism.” One study even implicated the father. The personality traits of the parents as a cause of autism has been investigated since the beginning of autism research, not as an epidemic but as an area that some researchers like Tustin keep coming back to. As early as the 1940s Asperger noted behavioral similarities between his clients and some of their parents.

Team authorship has played a role in the epidemic spread of causations. In the early years, much of the work was being done by individual investigators. In recent years, teams of researchers have broadened the scope of autism etiology possibilities. As mentioned in the “By Author” section of the previous chapter, even authors who show a causation preference while writing alone demonstrate a wider range of etiology investigations in their team summaries. This may be attributable to some group dynamic variant of the bandwagon appeal, but it may be a limited manifestation of the epidemic theory in action. The beliefs and background that each investigator brings to the project can have a mitigating or exacerbating influence on the other members of the research team in ways more subtle and profound than might only the prior publications of each. Of course, it may be simply that no individual or group research results to date have provided a genuine “front-runner” causation convincing enough for researchers in other areas to rally behind.

Lotka’s Law

Because of the limited topic, Lotka’s Law of Author Productivity does not hold up well in the field of autism causation research publications. Although the parabolic curve of Figure 5 (Appendix H) roughly matches the data points expected of an author productivity distribution (see Figure 6, also in Appendix H), solving for a constant results in even more disparate numbers than Bradford’s Law revealed for the journals. Koenig and Harrell’s article summarizes the arithmetic involved in Lotka’s Law so: “in a given body of literature, the number of authors contributing n papers is about 1/n² of the number of authors contributing just one paper . . .. Thus if two thousand authors have written just one paper, then roughly 1/10² x 2,000, or only about 20 persons would have written ten papers.” An equation to solve for the constant (k) in Lotka’s Law is

k = n² x p

where n is the frequency of occurrence for an author and p is instances of this number of authors being published.

Applying Lotka’s Law of Author Productivity to the data given in the “By Author” section of the previous chapter, Table 5.1 supplies the k values found for autism causation literature and what these values should be based on a total of 632 authors appearing once. Notice that the resulting k values are not constant. In fact, the data points do not show a clear progression or regression (as was true for the Bradford data), but rise and fall unexpectedly (see the third column of the following table). Modeling an expected constant based on the data having 632 authors who have been published only once clarifies how far from expectations the actual data are.

Table 5.1 Actual and Expected Author Productivity Outcomes

The decrease from 632 authors being published once to 87 authors being published twice is too rapid to be adequately explained by Lotka’s Law. The fourth column depicts the expected representation of authors having two, three, etc. articles published while the second column shows the actual values for single, double, triple, etc. publications by the same author. Either the 632 instances of an author being published only once is too high for a body of only 1,259 sources or the 38 times that Gillberg has been published is too high for the opposite end of the set. If the constant of 632 were to work out, then the author published most frequently (the asterisk in Table 5.1) should have only 25 or 26 articles published [632=25.14²(1)]. The other indication that the 632 k value is too high is that it would result in a total of 1,989 sources, 730 more than the actual number.

This is not to imply that the predictive value of Lotka’s Law is questionable, only that its utility in this specialized application is limited. This investigation addresses only autism causation literature, not the entire body of autism or even autism treatment literature. Perhaps with a broader sample to use, the formula would hold up better. This factor combined with the limits placed on counting authorship (as described in the “By Author” section of Chapter 4) mitigate the usefulness of Lotka’s Law as a predictive tool. However, its utility as an analytical tool—demonstrating expected outcomes versus actual—is enhanced by the uniqueness of this data set.

Does the Multidirectional Approach of Autism Causation Research Imply No Clear Consensus?

An apparent lack of consensus by time span, author, journal, or country prompted this research. The notion of consensus finding quickly gave way to simple counting in an attempt discover a percentage majority. However, the results of even this exercise show how diverse autism research has been and remains.

In addition to the causations and percentages discussed, the articles collected can be further subdivided into the number of causations that each produced. Of the 1,259 sources in the study, 991 are single-causation articles (78.71%). Two hundred forty-four (19.38%) of the sources present dual causations. There are also 22 articles (1.75%) that posit a triple factor etiology for autism. Two abstracts (0.16%) discuss four etiologies, but these are summary articles: one covering the “biochemistry” and “psychophysiology” of autistic syndromes and the other tracing autism back to possible genetic, neuronal, and virus developments concomitant with bleeding during the second trimester of gestation. The main point is that 268 of the 1,259 sources could not even limit themselves to a single etiology to support. It is almost the opposite of the epidemic theory or the bandwagon appeal.

Does the multidirectional approach to autism etiology research imply no clear consensus, or is this issue merely a symptom of a problem far worse? If the factional research being done is the result of something completely unrelated to autism (such as funding battles, pride, or politics) this would be unfortunate enough. However, the more likely reason for this lack of consensus is that autism involves such a range of developmental delays and ability levels that the researchers cannot focus on the root of this handicapping condition. High-functioning autism looks completely different from low-functioning autism. While neurologists and developmental pediatricians have made refinements in labeling a range of behaviors and abilities for each, the fundamental issue remains unresolved: whether autism is truly a spectrum disorder or many distinct kinds of retardation all classified together due to similar diagnostic behavioral traits. As a result, the even more fundamental issue of how autism occurs remains open to speculation and study.

Implications

Implications for Hypothesis

The null hypothesis of this research was that no one cause would claim a majority percentage of the literature. While this is supported, since the most cited causation is neurological (551 of the 1,543 causes collected) with 35.71 percent, the utility of deriving an expected average number of causations per etiology to discuss standard deviation boundaries for the causes found was not pursued. The causations arrived at are inherently contrived, if not arbitrary. A researcher with different thoughts and beliefs about autism could make a cogent case for combining the immunologic and allergy-induced causations with the environmental etiology. The same with biological and neurological causes. Conversely, difficulties during pregnancy (the pre-, peri-, and neonatal causation) could be tracked as its own category, instead of part of the environmental causation. Cognitive and organic could have been tracked individually also. The researcher could have separated social development from the psychological causation, or have aligned developmental delays with cognitive delays as examples of the neurological etiology. Actions a bit more drastic but justifiable would also result in a shift in the totals. For example, using only the past twenty or twenty-five years and dismissing earlier research as archaic would have greatly reduced the psychological causations.

Even so, to add more etiologies would make the original hypothesis even less likely. If an individual causation did not appear more than half of the time with only six causations, it is not very likely any would with even more causations tracked. Therefore, despite all of the variables and judgments involved, consistent constraints consistently applied should continue to yield consistent results. The causation set is not arbitrary through any fault in the data collection, but reflects the dispersion of autism causation literature. This is the most important finding of this bibliometric research.

Geographical Issues

The United States’ apparent dominance of autism causation literature (621 of the 1,259 sources and 746 of the 1,543 etiologies) is mitigated by two factors. The primary autism journal—the Journal of Autism and Developmental Disorders—is published in the U.S. but includes foreign authors and research findings. The other mitigating factor is that the data collection focused on English language abstracts. The latter limitation favors American and British journals, but the overall data show only limited indications of being pulled in the Anglo-American direction. The American percentages for neurological, biological, and psychological causations are lower than the overall percentages, while allergy/immunological, environmental, and genetic etiology percentages are slightly higher for the United States (reference Table 4.4). The neurological, psychological, genetic, environmental, allergy/immunologic, and biological rank order of the overall causations differs for the U.S. as well, with genetic etiologies second and psychological third. Also, although their order is the same, there is a larger gap between the allergy/immunological causations and the biological ones than is evident in the overall numbers. For causation articles published in England combined with those attributed to the “United Kingdom,” the percentages of each causation are one to two points higher for the neurological, psychological, and genetic etiologies, nearly the same for biological (6.84% versus the overall 6.68%) and about two points lower for environmental and allergy/immunologic causations. When the publications attributed to the United Kingdom are added to those attributed to England, the rank order of the top three is the same but the order of the bottom three causations is completely different. Interestingly, although the overall biological percentage is nearly the same as the British percentage, this etiology ranks last overall but fourth for England/United Kingdom. Allergy/immunologic etiologies place last in British research with environmental causations second to last. The 21 biological causatons attributed to the England and U.K. articles combined is considerably higher than the 16 environmental and 15 allergy/auto-immunological etiologies found. As represented in Table D-3, Appendix D, this is an accurate representation that the belief in a biological basis for autism is stronger in Europe than it is in the United States.

How the United Kingdom is distinguished from England in the databases brings to the fore another geographical issue worth noting: the geo-political lessons underlying this data collection. Some of the sources attribute the place of publication to the United Kingdom, others to England, the end result being 121 articles for the U.K. (and 155 causations) and 121 articles from England (152 causations). While articles listed as coming from England can be comfortably ascribed to just this country, the United Kingdom articles are not so clear. Especially with Ireland (there are no Northern Ireland items in this research, however) and Scotland appearing separately in the study (as well as Australia and Canada, which are part of the British Empire, though independent of the U.K.), how synonymously “United Kingdom” and “England” are being used from one database to the next is not clear.

The reason behind the division of certain other countries is more clearly a matter of when individual articles were published. Germany was partitioned into East and West for most of the 1943 through 1996 data span. The numbers found reflect this, with 42 of the 48 “Germany” sources coming from West Germany and only four (post 1990 ones) from the reunified Germany. Similarly, eleven articles are from the Union of Soviet Socialists Republics, with only one 1994 publication from Russia (and none from the other former Soviet Republics). Czechoslovakia has three sources included in the research; the recently created Czech Republic has one. The Yugoslavia that appears three times in this data set is the one that existed from 1943 to 1992, roughly the period of this research. (The present-day Federal Republic of Yugoslavia was created in 1992 from parts of the former country.)

Implications for Further Research

Short of some breakthrough in one causal area of research, clarification of what is and is not autism (then, possibly, what causes it) should become clear by further tracking of the growing body of research. Following up on the numbers and percentages as they were tracked herein will allow replication researchers to add later author, journal, country, and database statistics to these results and make the same comparisons.

Further research might do more with Goffman’s Epidemic Theory of Communication also. This research concluded only that since new causes surface and never disappear, the theory that specific causal categories will group together for a certain time span before giving way to autism etiology research in a different vein is partially true. Adding a comparison of the various names for autism at specific times and matching this to the causation statistics for these time spans may provide a correlation only touched upon in this research. In the previous chapter, the discussion of this aspect of autism focused on only how difficult the various labels make tracking autism etiologies. The more prominent result showing that every country has been investigating every cause in almost every decade reveals that instances of every etiology have always dominated the discussion. This point was the focus of this research. But investigating the divergence and uniqueness in autism’s labels could be a fruitful exercise if combined with grouping the causations popular when each term for autism was at its peak use. For example, one might expect to find that the psychological causation was the most investigated autism etiology when autism and childhood schizophrenia were considered synonymous.

In addition to clarifying what is and is not autism, another benefit of collecting the data for a few more years is that this would demonstrate the direction and influence of the most recent autism causation research. Along with advice to do this and create new Bradford and Lotka graphs, this researcher makes the following suggestions to improve the data collection phase.

Separate the country wherein the research was done (usually the native country of the authors) from the country that published the work. Most databases allow for this, or at least supply enough information for reasonable surmises. For example, after listing the authors, MEDLINE gives the university or institute and its location. After the journal title, the country of publication is parenthetically supplied. EMBASE similarly separates the place of scholarly research from the place of publication. BIOSIS Previews and PsycINFO supply the place the research was conducted but do not provide the country in which the named journal is published. The latter, however, does indicate in what language the article is written. Another advantage to this avenue of database information tracking is that it may clarify how having the majority of the research conducted in an individual country or by an individual university may be constricting an international consensus on what causes autism. However, conclusions could be difficult to draw since the print bibliographic sources are not as good about separating the place of the research from the place of publication. Only Excerpta Medica does; the rest make getting even this basic information challenging. However, as Goffman indicated, the information retrieval itself is a dynamic process, and source knowledge gained from data collection will allow the researcher to fill in most gaps. Perhaps the language in which a piece is written does not always provide a valid approximation of where the work occurred, but taken in conjunction with other abstract fields, this information could provide better accuracy.

Another suggestion for the data collection phase is that of designing the database to search multiple authors individually, since only 525 of the 1,259 pieces are by individual authors. Even before adding more years to the research, it would be interesting to contrast what such data might reveal through Lotka’s Law versus what was shown using the primary author method this research employed.

Similarly, a follow-on study should allow for rechecking the United States’ dominance of the publications by introducing true foreign medical and psychological abstract collections. Using not just translated sources but the National Library of Medicine’s sources as well (such as the African, Israeli, Korean, and Latin-American versions of Index Medicus) may supply a compelling contrast with the international data presented here.

Finally, pursuing other measures in addition to those of Bradford, Goffman, and Lotka might place more emphasis on the quality of the sources rather than their quantity. Another area of bibliometrics is citation analysis, which investigates the “practices and patterns of scholarly references.” The point is not merely to enumerate from bibliographic listings but that (as Nicholas and Ritchie elegantly put it) “[d]escriptive studies may seek to show how individual members of a literature possess common features and may be grouped according to these affinities.” As early as 1927, Gross and Gross conducted a study which looked at not only citation relationships and emerging patterns, but also each author’s possible motivation for citing a particular work. Other researchers have continued this, exploring the arithmetic of the “critical measure” of “coupling units.”

While research dynamics are important to keep in mind, the method for showing the “interconnectedness of citations” is the main thrust of citation analysis. Price claimed that each year there are 7 new papers published for every 100 previously published in that same subdiscipline. An average of about 15 references in each of the 7 papers will yield 105 references back to the original 100, for a citation average of a little over one each. In reality, though, some are never cited, a small group is heavily cited for a few years, and a small percent (the “citation classics”) continue to be cited year after year. Among the works cited heavily for a few years, there is a balance between the value of the information increasing with usage and the “half life” of the topic. “[H]alf life has been interpreted as the time during which half of the total use of individual items has been or is expected to be made.” However, these numbers do not account for re-citations; that is, newer works using the ideas first presented in works which are now obsolete.

Problems with measuring obsolescence stem from how this term—as well as the related term “half-life”—are defined. As mentioned above, the ideas from some obsolete works are reused in newer articles. Lancaster and Warner, however, describe the law of obsolescence from a database user’s perspective. They explain that, especially in science and technology, the probability of demand for materials declines as the materials age. However, they add that median use, or half-life is the number of years that are needed to satisfy half of the requests. That is, if fifty percent of the hits come from the last three years, this is the “half-life” for the subject being searched.

Two other problems in defining and using obsolescence are both related to quality. Frequency of citation is not necessarily a measure of quality. Related to this is that—while Bradford’s Law has limited utility as a collection development tool—most collection development is done on a qualitative basis, not the quantitative basis on which Bradford’s Law centers. Therefore, the main value of adding obsolescence or citation analysis to the discussion of autism causation literature is to introduce a look at quality that was not included in this thesis.

Summary of Discussion and Implications

In this final chapter, the outcomes from the previous chapter and their implications were discussed in the context of bibliometric information mapping using traditional information sciences methods. This application of bibliometric analyses involved a nonstandard use of standard information science tools such as Bradford’s Law of Scattering, Goffman’s Epidemic Theory, and Lotka’s Law of Author Productivity. Suggested methods of data collection and the use of citation analysis were also introduced as possible considerations for a follow-on study of autism causation literature.

The limits placed on the information mapped account for some of the nonconformance with expected outcomes. Even so, the results conformed to expected outcomes in more ways than they did not, and they show that autism etiology research has taken many roads and continues in many divergent directions. As the body of information concerning autism causation continues to grow, the application of bibliometric measures should grow more accurate and more predictive of the future of autism causation research, perhaps showing where all of these paths converge.

This page is maintained by
Jeff and Cathy Romanczuk, lukate@charter.net